From 38a8791aa7bfaa497e12e56b904a6b45986162d8 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Thu, 17 Aug 2017 16:37:49 -0400 Subject: drm/amdgpu: Fix huge page updates with CPU MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Correctly detect system memory mappings when using CPU and don't use huge pages for them. Avoid incorrectly translating a physical page table GPU address when splitting a huge page while mapping system memory. Signed-off-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 6b1343e5541d..ba475af99332 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1277,7 +1277,7 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, /* In the case of a mixed PT the PDE must point to it*/ if (p->adev->asic_type < CHIP_VEGA10 || nptes != AMDGPU_VM_PTE_COUNT(p->adev) || - p->func == amdgpu_vm_do_copy_ptes || + p->src || !(flags & AMDGPU_PTE_VALID)) { dst = amdgpu_bo_gpu_offset(entry->bo); @@ -1294,9 +1294,23 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, entry->addr = (dst | flags); if (use_cpu_update) { + /* In case a huge page is replaced with a system + * memory mapping, p->pages_addr != NULL and + * amdgpu_vm_cpu_set_ptes would try to translate dst + * through amdgpu_vm_map_gart. But dst is already a + * GPU address (of the page table). Disable + * amdgpu_vm_map_gart temporarily. + */ + dma_addr_t *tmp; + + tmp = p->pages_addr; + p->pages_addr = NULL; + pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo); pde = pd_addr + (entry - parent->entries) * 8; amdgpu_vm_cpu_set_ptes(p, pde, dst, 1, 0, flags); + + p->pages_addr = tmp; } else { if (parent->bo->shadow) { pd_addr = amdgpu_bo_gpu_offset(parent->bo->shadow); -- cgit From ddbb5313502f89fb41dbec209427f1a21f628441 Mon Sep 17 00:00:00 2001 From: Emily Deng Date: Mon, 21 Aug 2017 09:51:10 +0800 Subject: drm/amdgpu/virtual_dce: Virtual display doesn't support disable vblank immediately MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit For virtual display, it uses software timer to emulate the vsync interrupt, it doesn't have high precision, so doesn't support disable vblank immediately. BUG: SWDEV-129274 Signed-off-by: Emily Deng Reviewed-by: Michel Dänzer Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 4bdd851f56d0..538e5f27d120 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -221,8 +221,9 @@ int amdgpu_irq_init(struct amdgpu_device *adev) spin_lock_init(&adev->irq.lock); - /* Disable vblank irqs aggressively for power-saving */ - adev->ddev->vblank_disable_immediate = true; + if (!adev->enable_virtual_display) + /* Disable vblank irqs aggressively for power-saving */ + adev->ddev->vblank_disable_immediate = true; r = drm_vblank_init(adev->ddev, adev->mode_info.num_crtc); if (r) { -- cgit From 2d6fb10565ca13e7a0f6f4725f1c3da5a99a2bcc Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 18 Aug 2017 23:39:52 -0400 Subject: drm/amdgpu/gfx8: fix spelling typo in mqd allocation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 832e592fcd07..fc260c13b1da 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -4579,9 +4579,9 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_misc_reserved = 0x00000003; if (!(adev->flags & AMD_IS_APU)) { mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dyamic_cu_mask)); + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dyamic_cu_mask)); + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); } eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; @@ -4768,8 +4768,8 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); } else { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); - ((struct vi_mqd_allocation *)mqd)->dyamic_cu_mask = 0xFFFFFFFF; - ((struct vi_mqd_allocation *)mqd)->dyamic_rb_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); @@ -4792,8 +4792,8 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); - ((struct vi_mqd_allocation *)mqd)->dyamic_cu_mask = 0xFFFFFFFF; - ((struct vi_mqd_allocation *)mqd)->dyamic_rb_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); vi_srbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v8_0_mqd_init(ring); -- cgit From 83e74db6a81daff277732bdd00b438ede2107c68 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 21 Aug 2017 11:58:25 -0400 Subject: drm/amdgpu: add automatic per asic settings for gart_size MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We need a larger gart for asics that do not support GPUVM on all engines (e.g., MM) to make sure we have enough space for all gtt buffers in physical mode. Change the default size based on the asic type. Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 16 +++++++++++++++- 4 files changed, 20 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 12e71bbfd222..103635ab784c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -76,7 +76,7 @@ extern int amdgpu_modeset; extern int amdgpu_vram_limit; extern int amdgpu_vis_vram_limit; -extern unsigned amdgpu_gart_size; +extern int amdgpu_gart_size; extern int amdgpu_gtt_size; extern int amdgpu_moverate; extern int amdgpu_benchmarking; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 1a459ac63df4..f7ffb029f6d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1062,11 +1062,11 @@ static void amdgpu_check_arguments(struct amdgpu_device *adev) amdgpu_sched_jobs = roundup_pow_of_two(amdgpu_sched_jobs); } - if (amdgpu_gart_size < 32) { + if (amdgpu_gart_size != -1 && amdgpu_gart_size < 32) { /* gart size must be greater or equal to 32M */ dev_warn(adev->dev, "gart size (%d) too small\n", amdgpu_gart_size); - amdgpu_gart_size = 32; + amdgpu_gart_size = -1; } if (amdgpu_gtt_size != -1 && amdgpu_gtt_size < 32) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index e39ec981b11c..4ecf73787475 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -76,7 +76,7 @@ int amdgpu_vram_limit = 0; int amdgpu_vis_vram_limit = 0; -unsigned amdgpu_gart_size = 256; +int amdgpu_gart_size = -1; /* auto */ int amdgpu_gtt_size = -1; /* auto */ int amdgpu_moverate = -1; /* auto */ int amdgpu_benchmarking = 0; @@ -128,7 +128,7 @@ module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); MODULE_PARM_DESC(vis_vramlimit, "Restrict visible VRAM for testing, in megabytes"); module_param_named(vis_vramlimit, amdgpu_vis_vram_limit, int, 0444); -MODULE_PARM_DESC(gartsize, "Size of PCIE/IGP gart to setup in megabytes (32, 64, etc.)"); +MODULE_PARM_DESC(gartsize, "Size of gart to setup in megabytes (32, 64, etc., -1=auto)"); module_param_named(gartsize, amdgpu_gart_size, uint, 0600); MODULE_PARM_DESC(gttsize, "Size of the GTT domain in megabytes (-1 = auto)"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index 94c1e2e8e34c..b9b9f680fc08 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -65,7 +65,21 @@ */ void amdgpu_gart_set_defaults(struct amdgpu_device *adev) { - adev->mc.gart_size = (uint64_t)amdgpu_gart_size << 20; + u64 gart_size; + + if (amdgpu_gart_size == -1) { + /* make the GART larger for chips that + * dont' support VM for all rings + */ + if (adev->asic_type <= CHIP_STONEY) + gart_size = 1024; + else + gart_size = 256; + } else { + gart_size = amdgpu_gart_size; + } + + adev->mc.gart_size = gart_size << 20; } /** -- cgit From cf273a59ca3068caced2adaf2deeb44d2013c8a9 Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 18 Aug 2017 15:50:17 +0200 Subject: drm/amdgpu: fix and cleanup shadow handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Set the shadow flag on the shadow and not the parent, always bind shadow BOs during allocation instead of manually, use the reservation_object wrappers to grab the lock. This fixes a couple of issues with binding the shadow BOs as well as correctly evicting them when memory becomes tight. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 6 ---- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 46 +++++++++++++++--------------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 8 ------ 3 files changed, 23 insertions(+), 37 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index f7ffb029f6d5..e630d918fefc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2622,12 +2622,6 @@ static int amdgpu_recover_vram_from_shadow(struct amdgpu_device *adev, goto err; } - r = amdgpu_ttm_bind(&bo->shadow->tbo, &bo->shadow->tbo.mem); - if (r) { - DRM_ERROR("%p bind failed\n", bo->shadow); - goto err; - } - r = amdgpu_bo_restore_from_shadow(adev, ring, bo, NULL, fence, true); if (r) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index e7e899190bef..9e495da0bb03 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -91,7 +91,10 @@ static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, if (domain & AMDGPU_GEM_DOMAIN_GTT) { places[c].fpfn = 0; - places[c].lpfn = 0; + if (flags & AMDGPU_GEM_CREATE_SHADOW) + places[c].lpfn = adev->mc.gart_size >> PAGE_SHIFT; + else + places[c].lpfn = 0; places[c].flags = TTM_PL_FLAG_TT; if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) places[c].flags |= TTM_PL_FLAG_WC | @@ -446,17 +449,16 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, if (bo->shadow) return 0; - bo->flags |= AMDGPU_GEM_CREATE_SHADOW; - memset(&placements, 0, - (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); - - amdgpu_ttm_placement_init(adev, &placement, - placements, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC); + memset(&placements, 0, sizeof(placements)); + amdgpu_ttm_placement_init(adev, &placement, placements, + AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_SHADOW); r = amdgpu_bo_create_restricted(adev, size, byte_align, true, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC, + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_SHADOW, NULL, &placement, bo->tbo.resv, 0, @@ -484,30 +486,28 @@ int amdgpu_bo_create(struct amdgpu_device *adev, { struct ttm_placement placement = {0}; struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; int r; - memset(&placements, 0, - (AMDGPU_GEM_DOMAIN_MAX + 1) * sizeof(struct ttm_place)); + memset(&placements, 0, sizeof(placements)); + amdgpu_ttm_placement_init(adev, &placement, placements, + domain, parent_flags); - amdgpu_ttm_placement_init(adev, &placement, - placements, domain, flags); - - r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, - domain, flags, sg, &placement, - resv, init_value, bo_ptr); + r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, domain, + parent_flags, sg, &placement, resv, + init_value, bo_ptr); if (r) return r; - if (amdgpu_need_backup(adev) && (flags & AMDGPU_GEM_CREATE_SHADOW)) { - if (!resv) { - r = ww_mutex_lock(&(*bo_ptr)->tbo.resv->lock, NULL); - WARN_ON(r != 0); - } + if ((flags & AMDGPU_GEM_CREATE_SHADOW) && amdgpu_need_backup(adev)) { + if (!resv) + WARN_ON(reservation_object_lock((*bo_ptr)->tbo.resv, + NULL)); r = amdgpu_bo_create_shadow(adev, size, byte_align, (*bo_ptr)); if (!resv) - ww_mutex_unlock(&(*bo_ptr)->tbo.resv->lock); + reservation_object_unlock((*bo_ptr)->tbo.resv); if (r) amdgpu_bo_unref(bo_ptr); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index ba475af99332..96ec4e2b56e9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -165,14 +165,6 @@ static int amdgpu_vm_validate_level(struct amdgpu_vm_pt *parent, unsigned i; int r; - if (parent->bo->shadow) { - struct amdgpu_bo *shadow = parent->bo->shadow; - - r = amdgpu_ttm_bind(&shadow->tbo, &shadow->tbo.mem); - if (r) - return r; - } - if (use_cpu_for_update) { r = amdgpu_bo_kmap(parent->bo, NULL); if (r) -- cgit From a4da14cc6266a8b77e897750390cd4afe0e52344 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 22 Aug 2017 12:21:07 -0400 Subject: drm/amdgpu: refine default gart size MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Be more explicit and add comments explaining each case. Also s/gart/GART/ in the parameter string as per Felix' suggestion. Reviewed-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 38 +++++++++++++++++++++++++++----- 2 files changed, 33 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 4ecf73787475..0f16986ec5bc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -128,7 +128,7 @@ module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); MODULE_PARM_DESC(vis_vramlimit, "Restrict visible VRAM for testing, in megabytes"); module_param_named(vis_vramlimit, amdgpu_vis_vram_limit, int, 0444); -MODULE_PARM_DESC(gartsize, "Size of gart to setup in megabytes (32, 64, etc., -1=auto)"); +MODULE_PARM_DESC(gartsize, "Size of GART to setup in megabytes (32, 64, etc., -1=auto)"); module_param_named(gartsize, amdgpu_gart_size, uint, 0600); MODULE_PARM_DESC(gttsize, "Size of the GTT domain in megabytes (-1 = auto)"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index b9b9f680fc08..2027eb010a47 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -68,13 +68,39 @@ void amdgpu_gart_set_defaults(struct amdgpu_device *adev) u64 gart_size; if (amdgpu_gart_size == -1) { - /* make the GART larger for chips that - * dont' support VM for all rings - */ - if (adev->asic_type <= CHIP_STONEY) - gart_size = 1024; - else + switch (adev->asic_type) { +#ifdef CONFIG_DRM_AMDGPU_SI + case CHIP_HAINAN: /* no MM engines */ +#endif + case CHIP_TOPAZ: /* no MM engines */ + case CHIP_POLARIS11: /* all engines support GPUVM */ + case CHIP_POLARIS10: /* all engines support GPUVM */ + case CHIP_POLARIS12: /* all engines support GPUVM */ + case CHIP_VEGA10: /* all engines support GPUVM */ + default: gart_size = 256; + break; +#ifdef CONFIG_DRM_AMDGPU_SI + case CHIP_VERDE: /* UVD, VCE do not support GPUVM */ + case CHIP_TAHITI: /* UVD, VCE do not support GPUVM */ + case CHIP_PITCAIRN: /* UVD, VCE do not support GPUVM */ + case CHIP_OLAND: /* UVD, VCE do not support GPUVM */ +#endif +#ifdef CONFIG_DRM_AMDGPU_CIK + case CHIP_BONAIRE: /* UVD, VCE do not support GPUVM */ + case CHIP_HAWAII: /* UVD, VCE do not support GPUVM */ + case CHIP_KAVERI: /* UVD, VCE do not support GPUVM */ + case CHIP_KABINI: /* UVD, VCE do not support GPUVM */ + case CHIP_MULLINS: /* UVD, VCE do not support GPUVM */ +#endif + case CHIP_TONGA: /* UVD, VCE do not support GPUVM */ + case CHIP_FIJI: /* UVD, VCE do not support GPUVM */ + case CHIP_CARRIZO: /* UVD, VCE do not support GPUVM, DCE SG support */ + case CHIP_STONEY: /* UVD does not support GPUVM, DCE SG support */ + case CHIP_RAVEN: /* DCE SG support */ + gart_size = 1024; + break; + } } else { gart_size = amdgpu_gart_size; } -- cgit From c3db7b5a5591ede54fad5a4f5ea45f298e5d3470 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 22 Aug 2017 13:06:30 -0400 Subject: drm/amdgpu: move default gart size setting into gmc modules MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Move the asic specific code into the IP modules. Reviewed-by: Felix Kuehling Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c | 52 -------------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h | 1 - drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 19 +++++++++++- drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 22 +++++++++++++- drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 21 ++++++++++++- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 16 +++++++++- 6 files changed, 74 insertions(+), 57 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index 2027eb010a47..f4370081f6e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -56,58 +56,6 @@ * Common GART table functions. */ -/** - * amdgpu_gart_set_defaults - set the default gart_size - * - * @adev: amdgpu_device pointer - * - * Set the default gart_size based on parameters and available VRAM. - */ -void amdgpu_gart_set_defaults(struct amdgpu_device *adev) -{ - u64 gart_size; - - if (amdgpu_gart_size == -1) { - switch (adev->asic_type) { -#ifdef CONFIG_DRM_AMDGPU_SI - case CHIP_HAINAN: /* no MM engines */ -#endif - case CHIP_TOPAZ: /* no MM engines */ - case CHIP_POLARIS11: /* all engines support GPUVM */ - case CHIP_POLARIS10: /* all engines support GPUVM */ - case CHIP_POLARIS12: /* all engines support GPUVM */ - case CHIP_VEGA10: /* all engines support GPUVM */ - default: - gart_size = 256; - break; -#ifdef CONFIG_DRM_AMDGPU_SI - case CHIP_VERDE: /* UVD, VCE do not support GPUVM */ - case CHIP_TAHITI: /* UVD, VCE do not support GPUVM */ - case CHIP_PITCAIRN: /* UVD, VCE do not support GPUVM */ - case CHIP_OLAND: /* UVD, VCE do not support GPUVM */ -#endif -#ifdef CONFIG_DRM_AMDGPU_CIK - case CHIP_BONAIRE: /* UVD, VCE do not support GPUVM */ - case CHIP_HAWAII: /* UVD, VCE do not support GPUVM */ - case CHIP_KAVERI: /* UVD, VCE do not support GPUVM */ - case CHIP_KABINI: /* UVD, VCE do not support GPUVM */ - case CHIP_MULLINS: /* UVD, VCE do not support GPUVM */ -#endif - case CHIP_TONGA: /* UVD, VCE do not support GPUVM */ - case CHIP_FIJI: /* UVD, VCE do not support GPUVM */ - case CHIP_CARRIZO: /* UVD, VCE do not support GPUVM, DCE SG support */ - case CHIP_STONEY: /* UVD does not support GPUVM, DCE SG support */ - case CHIP_RAVEN: /* DCE SG support */ - gart_size = 1024; - break; - } - } else { - gart_size = amdgpu_gart_size; - } - - adev->mc.gart_size = gart_size << 20; -} - /** * amdgpu_gart_table_ram_alloc - allocate system ram for gart page table * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h index d4cce6936200..afbe803b1a13 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h @@ -56,7 +56,6 @@ struct amdgpu_gart { const struct amdgpu_gart_funcs *gart_funcs; }; -void amdgpu_gart_set_defaults(struct amdgpu_device *adev); int amdgpu_gart_table_ram_alloc(struct amdgpu_device *adev); void amdgpu_gart_table_ram_free(struct amdgpu_device *adev); int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 12b0c4cd7a5a..5be9c83dfcf7 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -332,7 +332,24 @@ static int gmc_v6_0_mc_init(struct amdgpu_device *adev) adev->mc.real_vram_size = RREG32(mmCONFIG_MEMSIZE) * 1024ULL * 1024ULL; adev->mc.visible_vram_size = adev->mc.aper_size; - amdgpu_gart_set_defaults(adev); + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_HAINAN: /* no MM engines */ + default: + adev->mc.gart_size = 256ULL << 20; + break; + case CHIP_VERDE: /* UVD, VCE do not support GPUVM */ + case CHIP_TAHITI: /* UVD, VCE do not support GPUVM */ + case CHIP_PITCAIRN: /* UVD, VCE do not support GPUVM */ + case CHIP_OLAND: /* UVD, VCE do not support GPUVM */ + adev->mc.gart_size = 1024ULL << 20; + break; + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } + gmc_v6_0_vram_gtt_location(adev, &adev->mc); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index e42c1ad3af5e..eace9e7182c8 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -386,7 +386,27 @@ static int gmc_v7_0_mc_init(struct amdgpu_device *adev) if (adev->mc.visible_vram_size > adev->mc.real_vram_size) adev->mc.visible_vram_size = adev->mc.real_vram_size; - amdgpu_gart_set_defaults(adev); + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_TOPAZ: /* no MM engines */ + default: + adev->mc.gart_size = 256ULL << 20; + break; +#ifdef CONFIG_DRM_AMDGPU_CIK + case CHIP_BONAIRE: /* UVD, VCE do not support GPUVM */ + case CHIP_HAWAII: /* UVD, VCE do not support GPUVM */ + case CHIP_KAVERI: /* UVD, VCE do not support GPUVM */ + case CHIP_KABINI: /* UVD, VCE do not support GPUVM */ + case CHIP_MULLINS: /* UVD, VCE do not support GPUVM */ + adev->mc.gart_size = 1024ULL << 20; + break; +#endif + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } + gmc_v7_0_vram_gtt_location(adev, &adev->mc); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 7ca2dae8237a..3b3326daf32b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -562,7 +562,26 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) if (adev->mc.visible_vram_size > adev->mc.real_vram_size) adev->mc.visible_vram_size = adev->mc.real_vram_size; - amdgpu_gart_set_defaults(adev); + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_POLARIS11: /* all engines support GPUVM */ + case CHIP_POLARIS10: /* all engines support GPUVM */ + case CHIP_POLARIS12: /* all engines support GPUVM */ + default: + adev->mc.gart_size = 256ULL << 20; + break; + case CHIP_TONGA: /* UVD, VCE do not support GPUVM */ + case CHIP_FIJI: /* UVD, VCE do not support GPUVM */ + case CHIP_CARRIZO: /* UVD, VCE do not support GPUVM, DCE SG support */ + case CHIP_STONEY: /* UVD does not support GPUVM, DCE SG support */ + adev->mc.gart_size = 1024ULL << 20; + break; + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } + gmc_v8_0_vram_gtt_location(adev, &adev->mc); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 2769c2b3b56e..d04d0b123212 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -499,7 +499,21 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) if (adev->mc.visible_vram_size > adev->mc.real_vram_size) adev->mc.visible_vram_size = adev->mc.real_vram_size; - amdgpu_gart_set_defaults(adev); + /* set the gart size */ + if (amdgpu_gart_size == -1) { + switch (adev->asic_type) { + case CHIP_VEGA10: /* all engines support GPUVM */ + default: + adev->mc.gart_size = 256ULL << 20; + break; + case CHIP_RAVEN: /* DCE SG support */ + adev->mc.gart_size = 1024ULL << 20; + break; + } + } else { + adev->mc.gart_size = (u64)amdgpu_gart_size << 20; + } + gmc_v9_0_vram_gtt_location(adev, &adev->mc); return 0; -- cgit From b249e18df151c9627af808321a8090c0b8d4d834 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 22 Aug 2017 16:39:30 -0400 Subject: drm/amdgpu: set sched_hw_submission higher for KIQ (v3) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit KIQ doesn't really use the GPU scheduler. The base drivers generally use the KIQ ring directly rather than submitting IBs. However, amdgpu_sched_hw_submission (which defaults to 2) limits the number of outstanding fences to 2. KFD uses the KIQ for TLB flushes and the 2 fence limit hurts performance when there are several KFD processes running. v2: move some expressions to one line change KIQ sched_hw_submission to at least 16 v3: bump to 256 Reviewed-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index 6c5646b48d1a..5ce65280b396 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -170,6 +170,16 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned irq_type) { int r; + int sched_hw_submission = amdgpu_sched_hw_submission; + + /* Set the hw submission limit higher for KIQ because + * it's used for a number of gfx/compute tasks by both + * KFD and KGD which may have outstanding fences and + * it doesn't really use the gpu scheduler anyway; + * KIQ tasks get submitted directly to the ring. + */ + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + sched_hw_submission = max(sched_hw_submission, 256); if (ring->adev == NULL) { if (adev->num_rings >= AMDGPU_MAX_RINGS) @@ -178,8 +188,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, ring->adev = adev; ring->idx = adev->num_rings++; adev->rings[ring->idx] = ring; - r = amdgpu_fence_driver_init_ring(ring, - amdgpu_sched_hw_submission); + r = amdgpu_fence_driver_init_ring(ring, sched_hw_submission); if (r) return r; } @@ -218,8 +227,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, return r; } - ring->ring_size = roundup_pow_of_two(max_dw * 4 * - amdgpu_sched_hw_submission); + ring->ring_size = roundup_pow_of_two(max_dw * 4 * sched_hw_submission); ring->buf_mask = (ring->ring_size / 4) - 1; ring->ptr_mask = ring->funcs->support_64bit_ptrs ? -- cgit From febb84a60304e7257d42b90cc191ef84279e5152 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 22 Aug 2017 12:50:46 +0200 Subject: drm/amdgpu: remove the GART copy hack MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This isn't used since we don't map evicted BOs to GART any more. Signed-off-by: Christian König Reviewed-by: Alex Deucher Reviewed-by: Roger He Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 22 +++++----------------- 1 file changed, 5 insertions(+), 17 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 96ec4e2b56e9..3bd430e180b5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1616,7 +1616,6 @@ error_free: * * @adev: amdgpu_device pointer * @exclusive: fence we need to sync to - * @gtt_flags: flags as they are used for GTT * @pages_addr: DMA addresses to use for mapping * @vm: requested vm * @mapping: mapped range and flags to use for the update @@ -1630,7 +1629,6 @@ error_free: */ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, struct dma_fence *exclusive, - uint64_t gtt_flags, dma_addr_t *pages_addr, struct amdgpu_vm *vm, struct amdgpu_bo_va_mapping *mapping, @@ -1685,11 +1683,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, } if (pages_addr) { - if (flags == gtt_flags) - src = adev->gart.table_addr + - (addr >> AMDGPU_GPU_PAGE_SHIFT) * 8; - else - max_entries = min(max_entries, 16ull * 1024ull); + max_entries = min(max_entries, 16ull * 1024ull); addr = 0; } else if (flags & AMDGPU_PTE_VALID) { addr += adev->vm_manager.vram_base_offset; @@ -1734,10 +1728,10 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_vm *vm = bo_va->base.vm; struct amdgpu_bo_va_mapping *mapping; dma_addr_t *pages_addr = NULL; - uint64_t gtt_flags, flags; struct ttm_mem_reg *mem; struct drm_mm_node *nodes; struct dma_fence *exclusive; + uint64_t flags; int r; if (clear || !bo_va->base.bo) { @@ -1757,15 +1751,10 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, exclusive = reservation_object_get_excl(bo->tbo.resv); } - if (bo) { + if (bo) flags = amdgpu_ttm_tt_pte_flags(adev, bo->tbo.ttm, mem); - gtt_flags = (amdgpu_ttm_is_bound(bo->tbo.ttm) && - adev == amdgpu_ttm_adev(bo->tbo.bdev)) ? - flags : 0; - } else { + else flags = 0x0; - gtt_flags = ~0x0; - } spin_lock(&vm->status_lock); if (!list_empty(&bo_va->base.vm_status)) @@ -1773,8 +1762,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, spin_unlock(&vm->status_lock); list_for_each_entry(mapping, &bo_va->invalids, list) { - r = amdgpu_vm_bo_split_mapping(adev, exclusive, - gtt_flags, pages_addr, vm, + r = amdgpu_vm_bo_split_mapping(adev, exclusive, pages_addr, vm, mapping, flags, nodes, &bo_va->last_pt_update); if (r) -- cgit From 9b0655e3adb6f741b79ce8758f77bf9e08e58243 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 22 Aug 2017 16:58:07 +0200 Subject: drm/amdgpu: fix amdgpu_ttm_bind MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use ttm_bo_mem_space instead of manually allocating GART space. This allows us to evict BOs when there isn't enought GART space any more. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c | 14 +++++-------- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 31 +++++++++++++++++++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h | 4 ---- 3 files changed, 30 insertions(+), 19 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c index 9e05e257729f..0d15eb7d31d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gtt_mgr.c @@ -108,10 +108,10 @@ bool amdgpu_gtt_mgr_is_allocated(struct ttm_mem_reg *mem) * * Allocate the address space for a node. */ -int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, - struct ttm_buffer_object *tbo, - const struct ttm_place *place, - struct ttm_mem_reg *mem) +static int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, + struct ttm_buffer_object *tbo, + const struct ttm_place *place, + struct ttm_mem_reg *mem) { struct amdgpu_device *adev = amdgpu_ttm_adev(man->bdev); struct amdgpu_gtt_mgr *mgr = man->priv; @@ -143,12 +143,8 @@ int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, fpfn, lpfn, mode); spin_unlock(&mgr->lock); - if (!r) { + if (!r) mem->start = node->start; - if (&tbo->mem == mem) - tbo->offset = (tbo->mem.start << PAGE_SHIFT) + - tbo->bdev->man[tbo->mem.mem_type].gpu_offset; - } return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 8b2c294f6f79..1efe1cba7e11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -824,20 +824,39 @@ bool amdgpu_ttm_is_bound(struct ttm_tt *ttm) int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem) { + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); struct ttm_tt *ttm = bo->ttm; + struct ttm_mem_reg tmp; + + struct ttm_placement placement; + struct ttm_place placements; int r; if (!ttm || amdgpu_ttm_is_bound(ttm)) return 0; - r = amdgpu_gtt_mgr_alloc(&bo->bdev->man[TTM_PL_TT], bo, - NULL, bo_mem); - if (r) { - DRM_ERROR("Failed to allocate GTT address space (%d)\n", r); + tmp = bo->mem; + tmp.mm_node = NULL; + placement.num_placement = 1; + placement.placement = &placements; + placement.num_busy_placement = 1; + placement.busy_placement = &placements; + placements.fpfn = 0; + placements.lpfn = adev->mc.gart_size >> PAGE_SHIFT; + placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; + + r = ttm_bo_mem_space(bo, &placement, &tmp, true, false); + if (unlikely(r)) return r; - } - return amdgpu_ttm_do_bind(ttm, bo_mem); + r = ttm_bo_move_ttm(bo, true, false, &tmp); + if (unlikely(r)) + ttm_bo_mem_put(bo, &tmp); + else + bo->offset = (bo->mem.start << PAGE_SHIFT) + + bo->bdev->man[bo->mem.mem_type].gpu_offset; + + return r; } int amdgpu_ttm_recover_gart(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index f22a4758719d..43093bffa2cf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -62,10 +62,6 @@ extern const struct ttm_mem_type_manager_func amdgpu_gtt_mgr_func; extern const struct ttm_mem_type_manager_func amdgpu_vram_mgr_func; bool amdgpu_gtt_mgr_is_allocated(struct ttm_mem_reg *mem); -int amdgpu_gtt_mgr_alloc(struct ttm_mem_type_manager *man, - struct ttm_buffer_object *tbo, - const struct ttm_place *place, - struct ttm_mem_reg *mem); uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man); uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man); -- cgit From 1cacc86a63abda5ff94a405765ca06b1c3a4031d Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 22 Aug 2017 21:04:47 +0200 Subject: drm/amdgpu: inline amdgpu_ttm_do_bind again MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The function is called only once and doesn't do anything special. Signed-off-by: Christian König Reviewed-by: Alex Deucher Reviewed-by: Roger He Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 45 +++++++++++++-------------------- 1 file changed, 18 insertions(+), 27 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 1efe1cba7e11..7ef6c28a34d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -761,35 +761,11 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) sg_free_table(ttm->sg); } -static int amdgpu_ttm_do_bind(struct ttm_tt *ttm, struct ttm_mem_reg *mem) -{ - struct amdgpu_ttm_tt *gtt = (void *)ttm; - uint64_t flags; - int r; - - spin_lock(>t->adev->gtt_list_lock); - flags = amdgpu_ttm_tt_pte_flags(gtt->adev, ttm, mem); - gtt->offset = (u64)mem->start << PAGE_SHIFT; - r = amdgpu_gart_bind(gtt->adev, gtt->offset, ttm->num_pages, - ttm->pages, gtt->ttm.dma_address, flags); - - if (r) { - DRM_ERROR("failed to bind %lu pages at 0x%08llX\n", - ttm->num_pages, gtt->offset); - goto error_gart_bind; - } - - list_add_tail(>t->list, >t->adev->gtt_list); -error_gart_bind: - spin_unlock(>t->adev->gtt_list_lock); - return r; - -} - static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem) { struct amdgpu_ttm_tt *gtt = (void*)ttm; + uint64_t flags; int r = 0; if (gtt->userptr) { @@ -809,9 +785,24 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, bo_mem->mem_type == AMDGPU_PL_OA) return -EINVAL; - if (amdgpu_gtt_mgr_is_allocated(bo_mem)) - r = amdgpu_ttm_do_bind(ttm, bo_mem); + if (!amdgpu_gtt_mgr_is_allocated(bo_mem)) + return 0; + + spin_lock(>t->adev->gtt_list_lock); + flags = amdgpu_ttm_tt_pte_flags(gtt->adev, ttm, bo_mem); + gtt->offset = (u64)bo_mem->start << PAGE_SHIFT; + r = amdgpu_gart_bind(gtt->adev, gtt->offset, ttm->num_pages, + ttm->pages, gtt->ttm.dma_address, flags); + + if (r) { + DRM_ERROR("failed to bind %lu pages at 0x%08llX\n", + ttm->num_pages, gtt->offset); + goto error_gart_bind; + } + list_add_tail(>t->list, >t->adev->gtt_list); +error_gart_bind: + spin_unlock(>t->adev->gtt_list_lock); return r; } -- cgit From a3ce364558faa12c4f25466dfc89eb3146b8063c Mon Sep 17 00:00:00 2001 From: Roger He Date: Thu, 24 Aug 2017 14:57:57 +0800 Subject: drm/amd/amdgpu: fix BANK_SELECT on Vega10 (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit BANK_SELECT should always be FRAGMENT_SIZE + 3 due to 8-entry (2^3) per cache line in L2 TLB for Vega10. v2: agd: fix warning Reviewed-by: Christian König Signed-off-by: Roger He Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 5 ++--- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 5 ++--- 2 files changed, 4 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index 4f2788b61a08..6c8040e616c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -124,7 +124,7 @@ static void gfxhub_v1_0_init_tlb_regs(struct amdgpu_device *adev) static void gfxhub_v1_0_init_cache_regs(struct amdgpu_device *adev) { - uint32_t tmp, field; + uint32_t tmp; /* Setup L2 cache */ tmp = RREG32_SOC15(GC, 0, mmVM_L2_CNTL); @@ -143,9 +143,8 @@ static void gfxhub_v1_0_init_cache_regs(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); WREG32_SOC15(GC, 0, mmVM_L2_CNTL2, tmp); - field = adev->vm_manager.fragment_size; tmp = mmVM_L2_CNTL3_DEFAULT; - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, field); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 6); WREG32_SOC15(GC, 0, mmVM_L2_CNTL3, tmp); diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 4395a4f12149..74cb647da30e 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -138,7 +138,7 @@ static void mmhub_v1_0_init_tlb_regs(struct amdgpu_device *adev) static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev) { - uint32_t tmp, field; + uint32_t tmp; /* Setup L2 cache */ tmp = RREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL); @@ -157,9 +157,8 @@ static void mmhub_v1_0_init_cache_regs(struct amdgpu_device *adev) tmp = REG_SET_FIELD(tmp, VM_L2_CNTL2, INVALIDATE_L2_CACHE, 1); WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL2, tmp); - field = adev->vm_manager.fragment_size; tmp = mmVM_L2_CNTL3_DEFAULT; - tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, field); + tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, BANK_SELECT, 9); tmp = REG_SET_FIELD(tmp, VM_L2_CNTL3, L2_CACHE_BIGK_FRAGMENT_SIZE, 6); WREG32_SOC15(MMHUB, 0, mmVM_L2_CNTL3, tmp); -- cgit From 06f10a537ec1d5fe68dc889a9a5d11afa49e6e0d Mon Sep 17 00:00:00 2001 From: Christophe JAILLET Date: Wed, 23 Aug 2017 07:52:36 +0200 Subject: drm/amdgpu: check memory allocation failure MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Check memory allocation failure and return -ENOMEM in such a case. 'num_post_dep_syncobjs' still has to be set to 0 before the test in order to have it initialized if 'amdgpu_cs_parser_fini()' is called to free resources. The calling graph would be, in such a case! failure in amdgpu_cs_process_syncobj_out_dep() ---> error code returned by amdgpu_cs_dependencies() --> amdgpu_cs_parser_fini() is called Reviewed-by: Christian König Signed-off-by: Christophe JAILLET Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 269b835571eb..d6ddd5562c16 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1079,6 +1079,9 @@ static int amdgpu_cs_process_syncobj_out_dep(struct amdgpu_cs_parser *p, GFP_KERNEL); p->num_post_dep_syncobjs = 0; + if (!p->post_dep_syncobjs) + return -ENOMEM; + for (i = 0; i < num_deps; ++i) { p->post_dep_syncobjs[i] = drm_syncobj_find(p->filp, deps[i].handle); if (!p->post_dep_syncobjs[i]) -- cgit From fd4b5f54e1afae4717ed716c8fdd72cb531f808b Mon Sep 17 00:00:00 2001 From: Colin Ian King Date: Wed, 23 Aug 2017 14:17:40 +0100 Subject: drm/amdgpu: remove duplicate return statement Remove a redundant identical return statement, it has no use. Detected by CoverityScan, CID#1454586 ("Structurally dead code") Reviewed-by: Felix Kuehling Signed-off-by: Colin Ian King Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index fb6e5dbd5a03..309f2419c6d8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -155,7 +155,6 @@ static const struct kfd2kgd_calls kfd2kgd = { struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void) { return (struct kfd2kgd_calls *)&kfd2kgd; - return (struct kfd2kgd_calls *)&kfd2kgd; } static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) -- cgit From 727030b0c6322b127b524dab1fbbe55a441a54b9 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Fri, 18 Aug 2017 16:46:47 +0800 Subject: drm/amdgpu: support polaris10/11/12 new cp firmwares MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Newer versions of the CP firmware require changes in how the driver initializes the hw block. Change the firmware name for new firmware to maintain compatibility with older kernels. Acked-by: Christian König Signed-off-by: Evan Quan Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 65 +++++++++++++++++++++++++++++------ 1 file changed, 55 insertions(+), 10 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index fc260c13b1da..775484639f79 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -918,8 +918,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) BUG(); } - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); - err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp_2.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); + err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.pfp_fw); @@ -929,8 +938,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); - err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me_2.bin", chip_name); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); + err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.me_fw); @@ -941,8 +959,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); - err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce_2.bin", chip_name); + err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); + err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); + err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.ce_fw); @@ -1012,8 +1039,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); - err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec_2.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); + err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); + } if (err) goto out; err = amdgpu_ucode_validate(adev->gfx.mec_fw); @@ -1025,8 +1061,17 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) if ((adev->asic_type != CHIP_STONEY) && (adev->asic_type != CHIP_TOPAZ)) { - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); - err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2_2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + if (err == -ENOENT) { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + } + } else { + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); + err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); + } if (!err) { err = amdgpu_ucode_validate(adev->gfx.mec2_fw); if (err) -- cgit From ca3670aa370e3cd01020decb308a0a5ddf193183 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Wed, 23 Aug 2017 15:33:40 -0400 Subject: drm/amd/amdgpu: Remove AMDGPU tracepoint and use new TTM tracepoint (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Switches the AMDGPU driver over to the TTM tracepoint and removes our old one. Now you can enable traces before loading the module and trace all mappings. Signed-off-by: Tom St Denis Reviewed-by: Christian König Signed-off-by: Alex Deucher (v2): Use struct device instead of pci in trace. --- drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 56 ------------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 21 ++---------- 2 files changed, 3 insertions(+), 74 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 1c88bd5e29ad..b1f97417241d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -14,62 +14,6 @@ #define AMDGPU_JOB_GET_TIMELINE_NAME(job) \ job->base.s_fence->finished.ops->get_timeline_name(&job->base.s_fence->finished) -TRACE_EVENT(amdgpu_ttm_tt_populate, - TP_PROTO(struct amdgpu_device *adev, uint64_t dma_address, uint64_t phys_address), - TP_ARGS(adev, dma_address, phys_address), - TP_STRUCT__entry( - __field(uint16_t, domain) - __field(uint8_t, bus) - __field(uint8_t, slot) - __field(uint8_t, func) - __field(uint64_t, dma) - __field(uint64_t, phys) - ), - TP_fast_assign( - __entry->domain = pci_domain_nr(adev->pdev->bus); - __entry->bus = adev->pdev->bus->number; - __entry->slot = PCI_SLOT(adev->pdev->devfn); - __entry->func = PCI_FUNC(adev->pdev->devfn); - __entry->dma = dma_address; - __entry->phys = phys_address; - ), - TP_printk("%04x:%02x:%02x.%x: 0x%llx => 0x%llx", - (unsigned)__entry->domain, - (unsigned)__entry->bus, - (unsigned)__entry->slot, - (unsigned)__entry->func, - (unsigned long long)__entry->dma, - (unsigned long long)__entry->phys) -); - -TRACE_EVENT(amdgpu_ttm_tt_unpopulate, - TP_PROTO(struct amdgpu_device *adev, uint64_t dma_address, uint64_t phys_address), - TP_ARGS(adev, dma_address, phys_address), - TP_STRUCT__entry( - __field(uint16_t, domain) - __field(uint8_t, bus) - __field(uint8_t, slot) - __field(uint8_t, func) - __field(uint64_t, dma) - __field(uint64_t, phys) - ), - TP_fast_assign( - __entry->domain = pci_domain_nr(adev->pdev->bus); - __entry->bus = adev->pdev->bus->number; - __entry->slot = PCI_SLOT(adev->pdev->devfn); - __entry->func = PCI_FUNC(adev->pdev->devfn); - __entry->dma = dma_address; - __entry->phys = phys_address; - ), - TP_printk("%04x:%02x:%02x.%x: 0x%llx => 0x%llx", - (unsigned)__entry->domain, - (unsigned)__entry->bus, - (unsigned)__entry->slot, - (unsigned)__entry->func, - (unsigned long long)__entry->dma, - (unsigned long long)__entry->phys) -); - TRACE_EVENT(amdgpu_mm_rreg, TP_PROTO(unsigned did, uint32_t reg, uint32_t value), TP_ARGS(did, reg, value), diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 7ef6c28a34d9..d1d94a14b089 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -34,6 +34,7 @@ #include #include #include +#include #include #include #include @@ -667,32 +668,16 @@ static void amdgpu_trace_dma_map(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned i; - if (unlikely(trace_amdgpu_ttm_tt_populate_enabled())) { - for (i = 0; i < ttm->num_pages; i++) { - trace_amdgpu_ttm_tt_populate( - adev, - gtt->ttm.dma_address[i], - page_to_phys(ttm->pages[i])); - } - } + ttm_trace_dma_map(adev->dev, >t->ttm); } static void amdgpu_trace_dma_unmap(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned i; - if (unlikely(trace_amdgpu_ttm_tt_unpopulate_enabled())) { - for (i = 0; i < ttm->num_pages; i++) { - trace_amdgpu_ttm_tt_unpopulate( - adev, - gtt->ttm.dma_address[i], - page_to_phys(ttm->pages[i])); - } - } + ttm_trace_dma_unmap(adev->dev, >t->ttm); } /* prepare the sg table with the user pages */ -- cgit From 7405e0dad4c75b33976ddd997513635d7a0204b1 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Fri, 18 Aug 2017 10:05:48 -0400 Subject: drm/amd/amdgpu: Use new TTM populate/map helper function MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 33 ++------------------------------- 1 file changed, 2 insertions(+), 31 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index d1d94a14b089..ce384ed9b9ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -926,7 +926,6 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned i; int r; bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); @@ -958,27 +957,7 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm) } #endif - r = ttm_pool_populate(ttm); - if (r) { - return r; - } - - for (i = 0; i < ttm->num_pages; i++) { - gtt->ttm.dma_address[i] = pci_map_page(adev->pdev, ttm->pages[i], - 0, PAGE_SIZE, - PCI_DMA_BIDIRECTIONAL); - if (pci_dma_mapping_error(adev->pdev, gtt->ttm.dma_address[i])) { - while (i--) { - pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i], - PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); - gtt->ttm.dma_address[i] = 0; - } - ttm_pool_unpopulate(ttm); - return -EFAULT; - } - } - - r = 0; + r = ttm_populate_and_map_pages(adev->dev, >t->ttm); trace_mappings: if (likely(!r)) amdgpu_trace_dma_map(ttm); @@ -989,7 +968,6 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) { struct amdgpu_device *adev; struct amdgpu_ttm_tt *gtt = (void *)ttm; - unsigned i; bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); if (gtt && gtt->userptr) { @@ -1012,14 +990,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) } #endif - for (i = 0; i < ttm->num_pages; i++) { - if (gtt->ttm.dma_address[i]) { - pci_unmap_page(adev->pdev, gtt->ttm.dma_address[i], - PAGE_SIZE, PCI_DMA_BIDIRECTIONAL); - } - } - - ttm_pool_unpopulate(ttm); + ttm_unmap_and_unpopulate_pages(adev->dev, >t->ttm); } int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, -- cgit From db95e2185523ee9d46a13ceee37bffe8442d2e1c Mon Sep 17 00:00:00 2001 From: Kent Russell Date: Tue, 22 Aug 2017 12:31:43 -0400 Subject: drm/amdgpu: Add debugfs file for VBIOS and version MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add 2 debugfs files, one that contains the VBIOS version, and one that contains the VBIOS itself. These won't change after initialization, so we can add the VBIOS version when we parse the atombios information. This ensures that we can find out the VBIOS version, even when the dmesg buffer fills up, and makes it easier to associate which VBIOS version is for which GPU on mGPU configurations. Set the size to 20 characters in case of some weird VBIOS version that exceeds the expected 17 character format (3-8-3\0). The VBIOS dump also allows for easy debugging v2: Move to debugfs, clarify commit message, add VBIOS dump file Signed-off-by: Kent Russell Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 62 ++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/atom.c | 5 ++- drivers/gpu/drm/amd/amdgpu/atom.h | 1 + 3 files changed, 67 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index e630d918fefc..da0a667a911b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -65,6 +65,8 @@ MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev); static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev); static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev); +static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev); +static int amdgpu_debugfs_vbios_version_init(struct amdgpu_device *adev); static const char *amdgpu_asic_name[] = { "TAHITI", @@ -2201,6 +2203,14 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) DRM_ERROR("registering firmware debugfs failed (%d).\n", r); + r = amdgpu_debugfs_vbios_dump_init(adev); + if (r) + DRM_ERROR("Creating vbios dump debugfs failed (%d).\n", r); + + r = amdgpu_debugfs_vbios_version_init(adev); + if (r) + DRM_ERROR("Creating vbios version debugfs failed (%d).\n", r); + if ((amdgpu_testing & 1)) { if (adev->accel_working) amdgpu_test_moves(adev); @@ -3754,6 +3764,50 @@ int amdgpu_debugfs_init(struct drm_minor *minor) { return 0; } + +static int amdgpu_debugfs_get_vbios_dump(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *) m->private; + struct drm_device *dev = node->minor->dev; + struct amdgpu_device *adev = dev->dev_private; + + seq_write(m, adev->bios, adev->bios_size); + return 0; +} + +static int amdgpu_debugfs_get_vbios_version(struct seq_file *m, void *data) +{ + struct drm_info_node *node = (struct drm_info_node *) m->private; + struct drm_device *dev = node->minor->dev; + struct amdgpu_device *adev = dev->dev_private; + struct atom_context *ctx = adev->mode_info.atom_context; + + seq_printf(m, "%s\n", ctx->vbios_version); + return 0; +} + +static const struct drm_info_list amdgpu_vbios_dump_list[] = { + {"amdgpu_vbios", + amdgpu_debugfs_get_vbios_dump, + 0, NULL}, +}; + +static const struct drm_info_list amdgpu_vbios_version_list[] = { + {"amdgpu_vbios_version", + amdgpu_debugfs_get_vbios_version, + 0, NULL}, +}; + +static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev) +{ + return amdgpu_debugfs_add_files(adev, + amdgpu_vbios_dump_list, 1); +} +static int amdgpu_debugfs_vbios_version_init(struct amdgpu_device *adev) +{ + return amdgpu_debugfs_add_files(adev, + amdgpu_vbios_version_list, 1); +} #else static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev) { @@ -3763,5 +3817,13 @@ static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev) { return 0; } +static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev) +{ + return 0; +} +static int amdgpu_debugfs_vbios_version_init(struct amdgpu_device *adev) +{ + return 0; +} static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev) { } #endif diff --git a/drivers/gpu/drm/amd/amdgpu/atom.c b/drivers/gpu/drm/amd/amdgpu/atom.c index d69aa2e179bb..69500a8b4e2d 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.c +++ b/drivers/gpu/drm/amd/amdgpu/atom.c @@ -1343,8 +1343,11 @@ struct atom_context *amdgpu_atom_parse(struct card_info *card, void *bios) idx = 0x80; str = CSTR(idx); - if (*str != '\0') + if (*str != '\0') { pr_info("ATOM BIOS: %s\n", str); + strlcpy(ctx->vbios_version, str, sizeof(ctx->vbios_version)); + } + return ctx; } diff --git a/drivers/gpu/drm/amd/amdgpu/atom.h b/drivers/gpu/drm/amd/amdgpu/atom.h index ddd8045accf3..a39170991afe 100644 --- a/drivers/gpu/drm/amd/amdgpu/atom.h +++ b/drivers/gpu/drm/amd/amdgpu/atom.h @@ -140,6 +140,7 @@ struct atom_context { int io_mode; uint32_t *scratch; int scratch_size_bytes; + char vbios_version[20]; }; extern int amdgpu_atom_debug; -- cgit From 5b41d94cc409610aa74b320b57a436df05a23f0f Mon Sep 17 00:00:00 2001 From: Kent Russell Date: Tue, 22 Aug 2017 12:31:43 -0400 Subject: drm/amdgpu: Move VBIOS version to sysfs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit sysfs is more stable, and doesn't require root to access Signed-off-by: Kent Russell Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 54 +++++++++++++----------------- 1 file changed, 23 insertions(+), 31 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index da0a667a911b..1c5c44acaad2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -66,7 +66,6 @@ static int amdgpu_debugfs_regs_init(struct amdgpu_device *adev); static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev); static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev); static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev); -static int amdgpu_debugfs_vbios_version_init(struct amdgpu_device *adev); static const char *amdgpu_asic_name[] = { "TAHITI", @@ -889,6 +888,20 @@ static uint32_t cail_ioreg_read(struct card_info *info, uint32_t reg) return r; } +static ssize_t amdgpu_atombios_get_vbios_version(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + struct atom_context *ctx = adev->mode_info.atom_context; + + return snprintf(buf, PAGE_SIZE, "%s\n", ctx->vbios_version); +} + +static DEVICE_ATTR(vbios_version, 0444, amdgpu_atombios_get_vbios_version, + NULL); + /** * amdgpu_atombios_fini - free the driver info and callbacks for atombios * @@ -908,6 +921,7 @@ static void amdgpu_atombios_fini(struct amdgpu_device *adev) adev->mode_info.atom_context = NULL; kfree(adev->mode_info.atom_card_info); adev->mode_info.atom_card_info = NULL; + device_remove_file(adev->dev, &dev_attr_vbios_version); } /** @@ -924,6 +938,7 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev) { struct card_info *atom_card_info = kzalloc(sizeof(struct card_info), GFP_KERNEL); + int ret; if (!atom_card_info) return -ENOMEM; @@ -960,6 +975,13 @@ static int amdgpu_atombios_init(struct amdgpu_device *adev) amdgpu_atombios_scratch_regs_init(adev); amdgpu_atombios_allocate_fb_scratch(adev); } + + ret = device_create_file(adev->dev, &dev_attr_vbios_version); + if (ret) { + DRM_ERROR("Failed to create device file for VBIOS version\n"); + return ret; + } + return 0; } @@ -2207,10 +2229,6 @@ int amdgpu_device_init(struct amdgpu_device *adev, if (r) DRM_ERROR("Creating vbios dump debugfs failed (%d).\n", r); - r = amdgpu_debugfs_vbios_version_init(adev); - if (r) - DRM_ERROR("Creating vbios version debugfs failed (%d).\n", r); - if ((amdgpu_testing & 1)) { if (adev->accel_working) amdgpu_test_moves(adev); @@ -3775,39 +3793,17 @@ static int amdgpu_debugfs_get_vbios_dump(struct seq_file *m, void *data) return 0; } -static int amdgpu_debugfs_get_vbios_version(struct seq_file *m, void *data) -{ - struct drm_info_node *node = (struct drm_info_node *) m->private; - struct drm_device *dev = node->minor->dev; - struct amdgpu_device *adev = dev->dev_private; - struct atom_context *ctx = adev->mode_info.atom_context; - - seq_printf(m, "%s\n", ctx->vbios_version); - return 0; -} - static const struct drm_info_list amdgpu_vbios_dump_list[] = { {"amdgpu_vbios", amdgpu_debugfs_get_vbios_dump, 0, NULL}, }; -static const struct drm_info_list amdgpu_vbios_version_list[] = { - {"amdgpu_vbios_version", - amdgpu_debugfs_get_vbios_version, - 0, NULL}, -}; - static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev) { return amdgpu_debugfs_add_files(adev, amdgpu_vbios_dump_list, 1); } -static int amdgpu_debugfs_vbios_version_init(struct amdgpu_device *adev) -{ - return amdgpu_debugfs_add_files(adev, - amdgpu_vbios_version_list, 1); -} #else static int amdgpu_debugfs_test_ib_ring_init(struct amdgpu_device *adev) { @@ -3821,9 +3817,5 @@ static int amdgpu_debugfs_vbios_dump_init(struct amdgpu_device *adev) { return 0; } -static int amdgpu_debugfs_vbios_version_init(struct amdgpu_device *adev) -{ - return 0; -} static void amdgpu_debugfs_regs_cleanup(struct amdgpu_device *adev) { } #endif -- cgit From 87f64a76b38acaa081b2bc46c3169884b9ccd6f2 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 23 Aug 2017 14:05:48 +0200 Subject: drm/amdgpu: fix amdgpu_vm_bo_map trace point MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit That somehow got lost. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 3bd430e180b5..db63ff5c80f2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2111,6 +2111,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, if (flags & AMDGPU_PTE_PRT) amdgpu_vm_prt_get(adev); + trace_amdgpu_vm_bo_map(bo_va, mapping); return 0; } @@ -2176,6 +2177,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, if (flags & AMDGPU_PTE_PRT) amdgpu_vm_prt_get(adev); + trace_amdgpu_vm_bo_map(bo_va, mapping); return 0; } -- cgit From 34d7be5dc28041b95254da517fd0f0f740544ece Mon Sep 17 00:00:00 2001 From: Christian König Date: Thu, 24 Aug 2017 12:32:55 +0200 Subject: drm/amdgpu: fix and cleanup VM ready check MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Stop checking the mapped BO itself, cause that one is certainly not a page table. Additional to that move the code into amdgpu_vm.c Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 33 ++------------------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 32 ++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 1 + 3 files changed, 35 insertions(+), 31 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 7171968f261e..9b1b6bdd4841 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -127,35 +127,6 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj, return 0; } -static int amdgpu_gem_vm_check(void *param, struct amdgpu_bo *bo) -{ - /* if anything is swapped out don't swap it in here, - just abort and wait for the next CS */ - if (!amdgpu_bo_gpu_accessible(bo)) - return -ERESTARTSYS; - - if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow)) - return -ERESTARTSYS; - - return 0; -} - -static bool amdgpu_gem_vm_ready(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct list_head *list) -{ - struct ttm_validate_buffer *entry; - - list_for_each_entry(entry, list, head) { - struct amdgpu_bo *bo = - container_of(entry->bo, struct amdgpu_bo, tbo); - if (amdgpu_gem_vm_check(NULL, bo)) - return false; - } - - return !amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_gem_vm_check, NULL); -} - void amdgpu_gem_object_close(struct drm_gem_object *obj, struct drm_file *file_priv) { @@ -189,7 +160,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, if (bo_va && --bo_va->ref_count == 0) { amdgpu_vm_bo_rmv(adev, bo_va); - if (amdgpu_gem_vm_ready(adev, vm, &list)) { + if (amdgpu_vm_ready(adev, vm)) { struct dma_fence *fence = NULL; r = amdgpu_vm_clear_freed(adev, vm, &fence); @@ -513,7 +484,7 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, { int r = -ERESTARTSYS; - if (!amdgpu_gem_vm_ready(adev, vm, list)) + if (!amdgpu_vm_ready(adev, vm)) goto error; r = amdgpu_vm_update_directories(adev, vm); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index db63ff5c80f2..67c37b22f8ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -231,6 +231,38 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, adev->mman.bdev.glob); } +/** + * amdgpu_vm_check - helper for amdgpu_vm_ready + */ +static int amdgpu_vm_check(void *param, struct amdgpu_bo *bo) +{ + /* if anything is swapped out don't swap it in here, + just abort and wait for the next CS */ + if (!amdgpu_bo_gpu_accessible(bo)) + return -ERESTARTSYS; + + if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow)) + return -ERESTARTSYS; + + return 0; +} + +/** + * amdgpu_vm_ready - check VM is ready for updates + * + * @adev: amdgpu device + * @vm: VM to check + * + * Check if all VM PDs/PTs are ready for updates + */ +bool amdgpu_vm_ready(struct amdgpu_device *adev, struct amdgpu_vm *vm) +{ + if (amdgpu_vm_check(NULL, vm->root.bo)) + return false; + + return !amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_vm_check, NULL); +} + /** * amdgpu_vm_alloc_levels - allocate the PD/PT levels * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index ba6691b58ee7..9347d28c3c1e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -225,6 +225,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); +bool amdgpu_vm_ready(struct amdgpu_device *adev, struct amdgpu_vm *vm); int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*callback)(void *p, struct amdgpu_bo *bo), void *param); -- cgit From 6ac7defb5ccf2c5e7b3fc9eb648535f3b3d17bc0 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 23 Aug 2017 20:11:25 +0200 Subject: drm/amdgpu: cleanup GWS, GDS and OA allocation MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Those are certainly not kernel allocations, instead set the NO_CPU_ACCESS flag. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 9b1b6bdd4841..ba012933e6aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -186,17 +186,17 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, { struct amdgpu_device *adev = dev->dev_private; union drm_amdgpu_gem_create *args = data; + uint64_t flags = args->in.domain_flags; uint64_t size = args->in.bo_size; struct drm_gem_object *gobj; uint32_t handle; - bool kernel = false; int r; /* reject invalid gem flags */ - if (args->in.domain_flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_NO_CPU_ACCESS | - AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_VRAM_CLEARED)) + if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_NO_CPU_ACCESS | + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_VRAM_CLEARED)) return -EINVAL; /* reject invalid gem domains */ @@ -211,7 +211,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, /* create a gem object to contain this object in */ if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) { - kernel = true; + flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS; if (args->in.domains == AMDGPU_GEM_DOMAIN_GDS) size = size << AMDGPU_GDS_SHIFT; else if (args->in.domains == AMDGPU_GEM_DOMAIN_GWS) @@ -225,8 +225,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, r = amdgpu_gem_object_create(adev, size, args->in.alignment, (u32)(0xffffffff & args->in.domains), - args->in.domain_flags, - kernel, &gobj); + flags, false, &gobj); if (r) return r; -- cgit From 925d5d798f465671c6b8011e80c636da46ef1a16 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 24 Aug 2017 16:46:29 -0400 Subject: drm/amdgpu/gfx8: apply dynamic cu mask to APUs as well Confirmed with the hw team. It's the same for all asics. Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 775484639f79..6666fcd8b08f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -4622,12 +4622,10 @@ static int gfx_v8_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se2 = 0xffffffff; mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; - if (!(adev->flags & AMD_IS_APU)) { - mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); - mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr - + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); - } + mqd->dynamic_cu_mask_addr_lo = lower_32_bits(ring->mqd_gpu_addr + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); + mqd->dynamic_cu_mask_addr_hi = upper_32_bits(ring->mqd_gpu_addr + + offsetof(struct vi_mqd_allocation, dynamic_cu_mask)); eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); -- cgit From ffe6d881e99413f3ebdfebf99570c2e840879e68 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 18 Aug 2017 23:52:37 -0400 Subject: drm/amdgpu/gfx9: adjust mqd allocation size To allocate additional space for the dynamic cu masks. Confirmed with the hw team that we only need 1 dword for the mask. The mask is the same for each SE so you only need 1 dword. Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 25 ++++++++++++++++++------- 1 file changed, 18 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 69182eeca264..7c06d1b99d99 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1346,7 +1346,7 @@ static int gfx_v9_0_sw_init(void *handle) return r; /* create MQD for all compute queues as wel as KIQ for SRIOV case */ - r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd)); + r = amdgpu_gfx_compute_mqd_sw_init(adev, sizeof(struct v9_mqd_allocation)); if (r) return r; @@ -2463,6 +2463,13 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) mqd->compute_static_thread_mgmt_se3 = 0xffffffff; mqd->compute_misc_reserved = 0x00000003; + mqd->dynamic_cu_mask_addr_lo = + lower_32_bits(ring->mqd_gpu_addr + + offsetof(struct v9_mqd_allocation, dynamic_cu_mask)); + mqd->dynamic_cu_mask_addr_hi = + upper_32_bits(ring->mqd_gpu_addr + + offsetof(struct v9_mqd_allocation, dynamic_cu_mask)); + eop_base_addr = ring->eop_gpu_addr >> 8; mqd->cp_hqd_eop_base_addr_lo = eop_base_addr; mqd->cp_hqd_eop_base_addr_hi = upper_32_bits(eop_base_addr); @@ -2695,7 +2702,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) if (adev->gfx.in_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; @@ -2707,7 +2714,9 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) soc15_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); } else { - memset((void *)mqd, 0, sizeof(*mqd)); + memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); + ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -2716,7 +2725,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); } return 0; @@ -2729,7 +2738,9 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) int mqd_idx = ring - &adev->gfx.compute_ring[0]; if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { - memset((void *)mqd, 0, sizeof(*mqd)); + memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); + ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; + ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, ring->me, ring->pipe, ring->queue, 0); gfx_v9_0_mqd_init(ring); @@ -2737,11 +2748,11 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) mutex_unlock(&adev->srbm_mutex); if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(*mqd)); + memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) - memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(*mqd)); + memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); /* reset ring buffer */ ring->wptr = 0; -- cgit From 08cab989f77582cb19df12d4a75a91b68b0017f6 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Tue, 29 Aug 2017 08:36:52 -0400 Subject: drm/amd/amdgpu: Add write() method to VRAM debugfs entry (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Allows writing data to vram via debugfs. Signed-off-by: Tom St Denis Reviewed-by: Christian König (v2): Call get_user before holding spinlock. Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 42 ++++++++++++++++++++++++++++++++- 1 file changed, 41 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index ce384ed9b9ef..6ea96e1fb273 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1671,10 +1671,50 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf, return result; } +static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result = 0; + int r; + + if (size & 0x3 || *pos & 0x3) + return -EINVAL; + + if (*pos >= adev->mc.mc_vram_size) + return -ENXIO; + + while (size) { + unsigned long flags; + uint32_t value; + + if (*pos >= adev->mc.mc_vram_size) + return result; + + r = get_user(value, (uint32_t *)buf); + if (r) + return r; + + spin_lock_irqsave(&adev->mmio_idx_lock, flags); + WREG32(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); + WREG32(mmMM_INDEX_HI, *pos >> 31); + WREG32(mmMM_DATA, value); + spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); + + result += 4; + buf += 4; + *pos += 4; + size -= 4; + } + + return result; +} + static const struct file_operations amdgpu_ttm_vram_fops = { .owner = THIS_MODULE, .read = amdgpu_ttm_vram_read, - .llseek = default_llseek + .write = amdgpu_ttm_vram_write, + .llseek = default_llseek, }; #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS -- cgit From 3d7d4d3a1b9f67c0caecf2b2aa8d7c347f074a33 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 23 Aug 2017 16:13:33 +0200 Subject: drm/amdgpu: rework moved handling in the VM v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead of using the vm_state use a separate flag to note that the BO was moved. v2: reorder patches to avoid temporary lockless access Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 13 ++++++++++--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 +++ 2 files changed, 13 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 67c37b22f8ef..1b36c62997b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1788,10 +1788,16 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, else flags = 0x0; - spin_lock(&vm->status_lock); - if (!list_empty(&bo_va->base.vm_status)) + if (!clear && bo_va->base.moved) { + bo_va->base.moved = false; list_splice_init(&bo_va->valids, &bo_va->invalids); - spin_unlock(&vm->status_lock); + + } else { + spin_lock(&vm->status_lock); + if (!list_empty(&bo_va->base.vm_status)) + list_splice_init(&bo_va->valids, &bo_va->invalids); + spin_unlock(&vm->status_lock); + } list_for_each_entry(mapping, &bo_va->invalids, list) { r = amdgpu_vm_bo_split_mapping(adev, exclusive, pages_addr, vm, @@ -2419,6 +2425,7 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, struct amdgpu_vm_bo_base *bo_base; list_for_each_entry(bo_base, &bo->va, bo_list) { + bo_base->moved = true; spin_lock(&bo_base->vm->status_lock); if (list_empty(&bo_base->vm_status)) list_add(&bo_base->vm_status, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 9347d28c3c1e..1b478e62a948 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -105,6 +105,9 @@ struct amdgpu_vm_bo_base { /* protected by spinlock */ struct list_head vm_status; + + /* protected by the BO being reserved */ + bool moved; }; struct amdgpu_vm_pt { -- cgit From cb7b6ec2f8b8759b6b5beb4d17ea6984867a3296 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 15 Aug 2017 17:08:12 +0200 Subject: drm/amdgpu: add bo_va cleared flag again v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We changed this to use an extra list a while back, but for the next series I need a separate flag again. v2: reorder to avoid unlocked list access Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 3 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 35 +++++++++++------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 --- 3 files changed, 16 insertions(+), 25 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index a288fa6d72c8..e613ba42f167 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -55,6 +55,9 @@ struct amdgpu_bo_va { /* mappings for this bo_va */ struct list_head invalids; struct list_head valids; + + /* If the mappings are cleared or filled */ + bool cleared; }; struct amdgpu_bo { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 1b36c62997b3..1334bbb82634 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1792,11 +1792,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, bo_va->base.moved = false; list_splice_init(&bo_va->valids, &bo_va->invalids); - } else { - spin_lock(&vm->status_lock); - if (!list_empty(&bo_va->base.vm_status)) - list_splice_init(&bo_va->valids, &bo_va->invalids); - spin_unlock(&vm->status_lock); + } else if (bo_va->cleared != clear) { + list_splice_init(&bo_va->valids, &bo_va->invalids); } list_for_each_entry(mapping, &bo_va->invalids, list) { @@ -1807,25 +1804,22 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, return r; } - if (trace_amdgpu_vm_bo_mapping_enabled()) { - list_for_each_entry(mapping, &bo_va->valids, list) - trace_amdgpu_vm_bo_mapping(mapping); - - list_for_each_entry(mapping, &bo_va->invalids, list) - trace_amdgpu_vm_bo_mapping(mapping); + if (vm->use_cpu_for_update) { + /* Flush HDP */ + mb(); + amdgpu_gart_flush_gpu_tlb(adev, 0); } spin_lock(&vm->status_lock); - list_splice_init(&bo_va->invalids, &bo_va->valids); list_del_init(&bo_va->base.vm_status); - if (clear) - list_add(&bo_va->base.vm_status, &vm->cleared); spin_unlock(&vm->status_lock); - if (vm->use_cpu_for_update) { - /* Flush HDP */ - mb(); - amdgpu_gart_flush_gpu_tlb(adev, 0); + list_splice_init(&bo_va->invalids, &bo_va->valids); + bo_va->cleared = clear; + + if (trace_amdgpu_vm_bo_mapping_enabled()) { + list_for_each_entry(mapping, &bo_va->valids, list) + trace_amdgpu_vm_bo_mapping(mapping); } return 0; @@ -2427,9 +2421,7 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, list_for_each_entry(bo_base, &bo->va, bo_list) { bo_base->moved = true; spin_lock(&bo_base->vm->status_lock); - if (list_empty(&bo_base->vm_status)) - list_add(&bo_base->vm_status, - &bo_base->vm->moved); + list_move(&bo_base->vm_status, &bo_base->vm->moved); spin_unlock(&bo_base->vm->status_lock); } } @@ -2516,7 +2508,6 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->reserved_vmid[i] = NULL; spin_lock_init(&vm->status_lock); INIT_LIST_HEAD(&vm->moved); - INIT_LIST_HEAD(&vm->cleared); INIT_LIST_HEAD(&vm->freed); /* create scheduler entity for page table updates */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 1b478e62a948..ff093d4b5e11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -129,9 +129,6 @@ struct amdgpu_vm { /* BOs moved, but not yet updated in the PT */ struct list_head moved; - /* BOs cleared in the PT because of a move */ - struct list_head cleared; - /* BO mappings freed, but not yet updated in the PT */ struct list_head freed; -- cgit From 00b5cc83c443dcd351cb2b21055656e007992b54 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 28 Aug 2017 14:46:40 +0200 Subject: drm/amdgpu: fix comment on amdgpu_bo_va MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Except for the reference count all other members are protected by the VM PD being reserved. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index e613ba42f167..42492e63b3a2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -49,9 +49,11 @@ struct amdgpu_bo_va { struct amdgpu_vm_bo_base base; /* protected by bo being reserved */ - struct dma_fence *last_pt_update; unsigned ref_count; + /* all other members protected by the VM PD being reserved */ + struct dma_fence *last_pt_update; + /* mappings for this bo_va */ struct list_head invalids; struct list_head valids; -- cgit From 3f3333f8a0e90ac26f84ed7b0aa344efce695c08 Mon Sep 17 00:00:00 2001 From: Christian König Date: Thu, 3 Aug 2017 14:02:13 +0200 Subject: drm/amdgpu: track evicted page tables v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead of validating all page tables when one was evicted, track which one needs a validation. v2: simplify amdgpu_vm_ready as well Signed-off-by: Christian König Reviewed-by: Alex Deucher (v1) Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 7 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 8 +- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 227 +++++++++++++---------------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 16 +- 5 files changed, 119 insertions(+), 141 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index d6ddd5562c16..8bf178a912f2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -636,9 +636,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved, p->bytes_moved_vis); - fpriv->vm.last_eviction_counter = - atomic64_read(&p->adev->num_evictions); - if (p->bo_list) { struct amdgpu_bo *gds = p->bo_list->gds_obj; struct amdgpu_bo *gws = p->bo_list->gws_obj; @@ -835,7 +832,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (!bo) continue; - amdgpu_vm_bo_invalidate(adev, bo); + amdgpu_vm_bo_invalidate(adev, bo, false); } } @@ -860,7 +857,7 @@ static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, } if (p->job->vm) { - p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.bo); + p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.base.bo); r = amdgpu_bo_vm_update_pte(p); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index ba012933e6aa..d02880640ee7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -160,7 +160,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, if (bo_va && --bo_va->ref_count == 0) { amdgpu_vm_bo_rmv(adev, bo_va); - if (amdgpu_vm_ready(adev, vm)) { + if (amdgpu_vm_ready(vm)) { struct dma_fence *fence = NULL; r = amdgpu_vm_clear_freed(adev, vm, &fence); @@ -481,10 +481,10 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, struct list_head *list, uint32_t operation) { - int r = -ERESTARTSYS; + int r; - if (!amdgpu_vm_ready(adev, vm)) - goto error; + if (!amdgpu_vm_ready(vm)) + return; r = amdgpu_vm_update_directories(adev, vm); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 9e495da0bb03..52d0109c0d9c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -929,7 +929,7 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, return; abo = container_of(bo, struct amdgpu_bo, tbo); - amdgpu_vm_bo_invalidate(adev, abo); + amdgpu_vm_bo_invalidate(adev, abo, evict); amdgpu_bo_kunmap(abo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 1334bbb82634..6ff3c1bf035e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -140,7 +140,7 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry) { - entry->robj = vm->root.bo; + entry->robj = vm->root.base.bo; entry->priority = 0; entry->tv.bo = &entry->robj->tbo; entry->tv.shared = true; @@ -148,61 +148,6 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, list_add(&entry->tv.head, validated); } -/** - * amdgpu_vm_validate_layer - validate a single page table level - * - * @parent: parent page table level - * @validate: callback to do the validation - * @param: parameter for the validation callback - * - * Validate the page table BOs on command submission if neccessary. - */ -static int amdgpu_vm_validate_level(struct amdgpu_vm_pt *parent, - int (*validate)(void *, struct amdgpu_bo *), - void *param, bool use_cpu_for_update, - struct ttm_bo_global *glob) -{ - unsigned i; - int r; - - if (use_cpu_for_update) { - r = amdgpu_bo_kmap(parent->bo, NULL); - if (r) - return r; - } - - if (!parent->entries) - return 0; - - for (i = 0; i <= parent->last_entry_used; ++i) { - struct amdgpu_vm_pt *entry = &parent->entries[i]; - - if (!entry->bo) - continue; - - r = validate(param, entry->bo); - if (r) - return r; - - spin_lock(&glob->lru_lock); - ttm_bo_move_to_lru_tail(&entry->bo->tbo); - if (entry->bo->shadow) - ttm_bo_move_to_lru_tail(&entry->bo->shadow->tbo); - spin_unlock(&glob->lru_lock); - - /* - * Recurse into the sub directory. This is harmless because we - * have only a maximum of 5 layers. - */ - r = amdgpu_vm_validate_level(entry, validate, param, - use_cpu_for_update, glob); - if (r) - return r; - } - - return r; -} - /** * amdgpu_vm_validate_pt_bos - validate the page table BOs * @@ -217,32 +162,43 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*validate)(void *p, struct amdgpu_bo *bo), void *param) { - uint64_t num_evictions; + struct ttm_bo_global *glob = adev->mman.bdev.glob; + int r; - /* We only need to validate the page tables - * if they aren't already valid. - */ - num_evictions = atomic64_read(&adev->num_evictions); - if (num_evictions == vm->last_eviction_counter) - return 0; + spin_lock(&vm->status_lock); + while (!list_empty(&vm->evicted)) { + struct amdgpu_vm_bo_base *bo_base; + struct amdgpu_bo *bo; - return amdgpu_vm_validate_level(&vm->root, validate, param, - vm->use_cpu_for_update, - adev->mman.bdev.glob); -} + bo_base = list_first_entry(&vm->evicted, + struct amdgpu_vm_bo_base, + vm_status); + spin_unlock(&vm->status_lock); -/** - * amdgpu_vm_check - helper for amdgpu_vm_ready - */ -static int amdgpu_vm_check(void *param, struct amdgpu_bo *bo) -{ - /* if anything is swapped out don't swap it in here, - just abort and wait for the next CS */ - if (!amdgpu_bo_gpu_accessible(bo)) - return -ERESTARTSYS; + bo = bo_base->bo; + BUG_ON(!bo); + if (bo->parent) { + r = validate(param, bo); + if (r) + return r; - if (bo->shadow && !amdgpu_bo_gpu_accessible(bo->shadow)) - return -ERESTARTSYS; + spin_lock(&glob->lru_lock); + ttm_bo_move_to_lru_tail(&bo->tbo); + if (bo->shadow) + ttm_bo_move_to_lru_tail(&bo->shadow->tbo); + spin_unlock(&glob->lru_lock); + } + + if (vm->use_cpu_for_update) { + r = amdgpu_bo_kmap(bo, NULL); + if (r) + return r; + } + + spin_lock(&vm->status_lock); + list_del_init(&bo_base->vm_status); + } + spin_unlock(&vm->status_lock); return 0; } @@ -250,17 +206,19 @@ static int amdgpu_vm_check(void *param, struct amdgpu_bo *bo) /** * amdgpu_vm_ready - check VM is ready for updates * - * @adev: amdgpu device * @vm: VM to check * * Check if all VM PDs/PTs are ready for updates */ -bool amdgpu_vm_ready(struct amdgpu_device *adev, struct amdgpu_vm *vm) +bool amdgpu_vm_ready(struct amdgpu_vm *vm) { - if (amdgpu_vm_check(NULL, vm->root.bo)) - return false; + bool ready; + + spin_lock(&vm->status_lock); + ready = list_empty(&vm->evicted); + spin_unlock(&vm->status_lock); - return !amdgpu_vm_validate_pt_bos(adev, vm, amdgpu_vm_check, NULL); + return ready; } /** @@ -326,11 +284,11 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, /* walk over the address space and allocate the page tables */ for (pt_idx = from; pt_idx <= to; ++pt_idx) { - struct reservation_object *resv = vm->root.bo->tbo.resv; + struct reservation_object *resv = vm->root.base.bo->tbo.resv; struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; struct amdgpu_bo *pt; - if (!entry->bo) { + if (!entry->base.bo) { r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, level), AMDGPU_GPU_PAGE_SIZE, true, @@ -351,9 +309,12 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, /* Keep a reference to the root directory to avoid * freeing them up in the wrong order. */ - pt->parent = amdgpu_bo_ref(vm->root.bo); + pt->parent = amdgpu_bo_ref(vm->root.base.bo); - entry->bo = pt; + entry->base.vm = vm; + entry->base.bo = pt; + list_add_tail(&entry->base.bo_list, &pt->va); + INIT_LIST_HEAD(&entry->base.vm_status); entry->addr = 0; } @@ -1020,7 +981,7 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm, int r; amdgpu_sync_create(&sync); - amdgpu_sync_resv(adev, &sync, vm->root.bo->tbo.resv, owner); + amdgpu_sync_resv(adev, &sync, vm->root.base.bo->tbo.resv, owner); r = amdgpu_sync_wait(&sync, true); amdgpu_sync_free(&sync); @@ -1059,10 +1020,10 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, memset(¶ms, 0, sizeof(params)); params.adev = adev; - shadow = parent->bo->shadow; + shadow = parent->base.bo->shadow; if (vm->use_cpu_for_update) { - pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo); + pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo); r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM); if (unlikely(r)) return r; @@ -1078,7 +1039,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, /* assume the worst case */ ndw += parent->last_entry_used * 6; - pd_addr = amdgpu_bo_gpu_offset(parent->bo); + pd_addr = amdgpu_bo_gpu_offset(parent->base.bo); if (shadow) { shadow_addr = amdgpu_bo_gpu_offset(shadow); @@ -1098,7 +1059,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, /* walk over the address space and update the directory */ for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { - struct amdgpu_bo *bo = parent->entries[pt_idx].bo; + struct amdgpu_bo *bo = parent->entries[pt_idx].base.bo; uint64_t pde, pt; if (bo == NULL) @@ -1141,7 +1102,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, } if (count) { - if (vm->root.bo->shadow) + if (vm->root.base.bo->shadow) params.func(¶ms, last_shadow, last_pt, count, incr, AMDGPU_PTE_VALID); @@ -1154,7 +1115,8 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, amdgpu_job_free(job); } else { amdgpu_ring_pad_ib(ring, params.ib); - amdgpu_sync_resv(adev, &job->sync, parent->bo->tbo.resv, + amdgpu_sync_resv(adev, &job->sync, + parent->base.bo->tbo.resv, AMDGPU_FENCE_OWNER_VM); if (shadow) amdgpu_sync_resv(adev, &job->sync, @@ -1167,7 +1129,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, if (r) goto error_free; - amdgpu_bo_fence(parent->bo, fence, true); + amdgpu_bo_fence(parent->base.bo, fence, true); dma_fence_put(vm->last_dir_update); vm->last_dir_update = dma_fence_get(fence); dma_fence_put(fence); @@ -1180,7 +1142,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - if (!entry->bo) + if (!entry->base.bo) continue; r = amdgpu_vm_update_level(adev, vm, entry, level + 1); @@ -1213,7 +1175,7 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent) for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - if (!entry->bo) + if (!entry->base.bo) continue; entry->addr = ~0ULL; @@ -1268,7 +1230,7 @@ void amdgpu_vm_get_entry(struct amdgpu_pte_update_params *p, uint64_t addr, *entry = &p->vm->root; while ((*entry)->entries) { idx = addr >> (p->adev->vm_manager.block_size * level--); - idx %= amdgpu_bo_size((*entry)->bo) / 8; + idx %= amdgpu_bo_size((*entry)->base.bo) / 8; *parent = *entry; *entry = &(*entry)->entries[idx]; } @@ -1304,7 +1266,7 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, p->src || !(flags & AMDGPU_PTE_VALID)) { - dst = amdgpu_bo_gpu_offset(entry->bo); + dst = amdgpu_bo_gpu_offset(entry->base.bo); dst = amdgpu_gart_get_vm_pde(p->adev, dst); flags = AMDGPU_PTE_VALID; } else { @@ -1330,18 +1292,18 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, tmp = p->pages_addr; p->pages_addr = NULL; - pd_addr = (unsigned long)amdgpu_bo_kptr(parent->bo); + pd_addr = (unsigned long)amdgpu_bo_kptr(parent->base.bo); pde = pd_addr + (entry - parent->entries) * 8; amdgpu_vm_cpu_set_ptes(p, pde, dst, 1, 0, flags); p->pages_addr = tmp; } else { - if (parent->bo->shadow) { - pd_addr = amdgpu_bo_gpu_offset(parent->bo->shadow); + if (parent->base.bo->shadow) { + pd_addr = amdgpu_bo_gpu_offset(parent->base.bo->shadow); pde = pd_addr + (entry - parent->entries) * 8; amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags); } - pd_addr = amdgpu_bo_gpu_offset(parent->bo); + pd_addr = amdgpu_bo_gpu_offset(parent->base.bo); pde = pd_addr + (entry - parent->entries) * 8; amdgpu_vm_do_set_ptes(p, pde, dst, 1, 0, flags); } @@ -1392,7 +1354,7 @@ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, if (entry->addr & AMDGPU_PDE_PTE) continue; - pt = entry->bo; + pt = entry->base.bo; if (use_cpu_update) { pe_start = (unsigned long)amdgpu_bo_kptr(pt); } else { @@ -1612,12 +1574,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - r = amdgpu_sync_resv(adev, &job->sync, vm->root.bo->tbo.resv, + r = amdgpu_sync_resv(adev, &job->sync, vm->root.base.bo->tbo.resv, owner); if (r) goto error_free; - r = reservation_object_reserve_shared(vm->root.bo->tbo.resv); + r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv); if (r) goto error_free; @@ -1632,7 +1594,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - amdgpu_bo_fence(vm->root.bo, f, true); + amdgpu_bo_fence(vm->root.base.bo, f, true); dma_fence_put(*fence); *fence = f; return 0; @@ -1927,7 +1889,7 @@ static void amdgpu_vm_free_mapping(struct amdgpu_device *adev, */ static void amdgpu_vm_prt_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { - struct reservation_object *resv = vm->root.bo->tbo.resv; + struct reservation_object *resv = vm->root.base.bo->tbo.resv; struct dma_fence *excl, **shared; unsigned i, shared_count; int r; @@ -2414,12 +2376,25 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, * Mark @bo as invalid. */ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, - struct amdgpu_bo *bo) + struct amdgpu_bo *bo, bool evicted) { struct amdgpu_vm_bo_base *bo_base; list_for_each_entry(bo_base, &bo->va, bo_list) { + struct amdgpu_vm *vm = bo_base->vm; + bo_base->moved = true; + if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) { + spin_lock(&bo_base->vm->status_lock); + list_move(&bo_base->vm_status, &vm->evicted); + spin_unlock(&bo_base->vm->status_lock); + continue; + } + + /* Don't add page tables to the moved state */ + if (bo->tbo.type == ttm_bo_type_kernel) + continue; + spin_lock(&bo_base->vm->status_lock); list_move(&bo_base->vm_status, &bo_base->vm->moved); spin_unlock(&bo_base->vm->status_lock); @@ -2507,6 +2482,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) vm->reserved_vmid[i] = NULL; spin_lock_init(&vm->status_lock); + INIT_LIST_HEAD(&vm->evicted); INIT_LIST_HEAD(&vm->moved); INIT_LIST_HEAD(&vm->freed); @@ -2551,30 +2527,31 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, r = amdgpu_bo_create(adev, amdgpu_vm_bo_size(adev, 0), align, true, AMDGPU_GEM_DOMAIN_VRAM, flags, - NULL, NULL, init_pde_value, &vm->root.bo); + NULL, NULL, init_pde_value, &vm->root.base.bo); if (r) goto error_free_sched_entity; - r = amdgpu_bo_reserve(vm->root.bo, false); - if (r) - goto error_free_root; - - vm->last_eviction_counter = atomic64_read(&adev->num_evictions); + vm->root.base.vm = vm; + list_add_tail(&vm->root.base.bo_list, &vm->root.base.bo->va); + INIT_LIST_HEAD(&vm->root.base.vm_status); if (vm->use_cpu_for_update) { - r = amdgpu_bo_kmap(vm->root.bo, NULL); + r = amdgpu_bo_reserve(vm->root.base.bo, false); if (r) goto error_free_root; - } - amdgpu_bo_unreserve(vm->root.bo); + r = amdgpu_bo_kmap(vm->root.base.bo, NULL); + if (r) + goto error_free_root; + amdgpu_bo_unreserve(vm->root.base.bo); + } return 0; error_free_root: - amdgpu_bo_unref(&vm->root.bo->shadow); - amdgpu_bo_unref(&vm->root.bo); - vm->root.bo = NULL; + amdgpu_bo_unref(&vm->root.base.bo->shadow); + amdgpu_bo_unref(&vm->root.base.bo); + vm->root.base.bo = NULL; error_free_sched_entity: amd_sched_entity_fini(&ring->sched, &vm->entity); @@ -2593,9 +2570,11 @@ static void amdgpu_vm_free_levels(struct amdgpu_vm_pt *level) { unsigned i; - if (level->bo) { - amdgpu_bo_unref(&level->bo->shadow); - amdgpu_bo_unref(&level->bo); + if (level->base.bo) { + list_del(&level->base.bo_list); + list_del(&level->base.vm_status); + amdgpu_bo_unref(&level->base.bo->shadow); + amdgpu_bo_unref(&level->base.bo); } if (level->entries) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index ff093d4b5e11..4e465e817fe8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -111,12 +111,12 @@ struct amdgpu_vm_bo_base { }; struct amdgpu_vm_pt { - struct amdgpu_bo *bo; - uint64_t addr; + struct amdgpu_vm_bo_base base; + uint64_t addr; /* array of page tables, one for each directory entry */ - struct amdgpu_vm_pt *entries; - unsigned last_entry_used; + struct amdgpu_vm_pt *entries; + unsigned last_entry_used; }; struct amdgpu_vm { @@ -126,6 +126,9 @@ struct amdgpu_vm { /* protecting invalidated */ spinlock_t status_lock; + /* BOs who needs a validation */ + struct list_head evicted; + /* BOs moved, but not yet updated in the PT */ struct list_head moved; @@ -135,7 +138,6 @@ struct amdgpu_vm { /* contains the page directory */ struct amdgpu_vm_pt root; struct dma_fence *last_dir_update; - uint64_t last_eviction_counter; /* protecting freed */ spinlock_t freed_lock; @@ -225,7 +227,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry); -bool amdgpu_vm_ready(struct amdgpu_device *adev, struct amdgpu_vm *vm); +bool amdgpu_vm_ready(struct amdgpu_vm *vm); int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*callback)(void *p, struct amdgpu_bo *bo), void *param); @@ -250,7 +252,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear); void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, - struct amdgpu_bo *bo); + struct amdgpu_bo *bo, bool evicted); struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, struct amdgpu_bo *bo); struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev, -- cgit From c5927537dd5706b17affa8aeea28c7b19c8fbb42 Mon Sep 17 00:00:00 2001 From: Himanshu Jha Date: Tue, 29 Aug 2017 18:51:27 +0530 Subject: drm/amd: Remove null check before kfree MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Kfree on NULL pointer is a no-op and therefore checking is redundant. Reviewed-by: Christian König Signed-off-by: Himanshu Jha Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 8d1cf2d3e663..f51b41f094ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -346,10 +346,8 @@ static void amdgpu_connector_free_edid(struct drm_connector *connector) { struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); - if (amdgpu_connector->edid) { - kfree(amdgpu_connector->edid); - amdgpu_connector->edid = NULL; - } + kfree(amdgpu_connector->edid); + amdgpu_connector->edid = NULL; } static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector) -- cgit From ea09729c930223edf492d0ca647c27e7eb0ccb12 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 9 Aug 2017 14:15:46 +0200 Subject: drm/amdgpu: rework page directory filling v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Keep track off relocated PDs/PTs instead of walking and checking all PDs. v2: fix root PD handling Signed-off-by: Christian König Reviewed-by: Alex Deucher (v1) Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 89 +++++++++++++++++++++++----------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 ++ 2 files changed, 63 insertions(+), 29 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 6ff3c1bf035e..faa08d5728da 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -196,7 +196,7 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, } spin_lock(&vm->status_lock); - list_del_init(&bo_base->vm_status); + list_move(&bo_base->vm_status, &vm->relocated); } spin_unlock(&vm->status_lock); @@ -314,8 +314,10 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, entry->base.vm = vm; entry->base.bo = pt; list_add_tail(&entry->base.bo_list, &pt->va); - INIT_LIST_HEAD(&entry->base.vm_status); - entry->addr = 0; + spin_lock(&vm->status_lock); + list_add(&entry->base.vm_status, &vm->relocated); + spin_unlock(&vm->status_lock); + entry->addr = ~0ULL; } if (level < adev->vm_manager.num_level) { @@ -1000,18 +1002,17 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm, */ static int amdgpu_vm_update_level(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_vm_pt *parent, - unsigned level) + struct amdgpu_vm_pt *parent) { struct amdgpu_bo *shadow; struct amdgpu_ring *ring = NULL; uint64_t pd_addr, shadow_addr = 0; - uint32_t incr = amdgpu_vm_bo_size(adev, level + 1); uint64_t last_pde = ~0, last_pt = ~0, last_shadow = ~0; unsigned count = 0, pt_idx, ndw = 0; struct amdgpu_job *job; struct amdgpu_pte_update_params params; struct dma_fence *fence = NULL; + uint32_t incr; int r; @@ -1059,12 +1060,17 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, /* walk over the address space and update the directory */ for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { - struct amdgpu_bo *bo = parent->entries[pt_idx].base.bo; + struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; + struct amdgpu_bo *bo = entry->base.bo; uint64_t pde, pt; if (bo == NULL) continue; + spin_lock(&vm->status_lock); + list_del_init(&entry->base.vm_status); + spin_unlock(&vm->status_lock); + pt = amdgpu_bo_gpu_offset(bo); pt = amdgpu_gart_get_vm_pde(adev, pt); /* Don't update huge pages here */ @@ -1075,6 +1081,7 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, parent->entries[pt_idx].addr = pt | AMDGPU_PTE_VALID; pde = pd_addr + pt_idx * 8; + incr = amdgpu_bo_size(bo); if (((last_pde + 8 * count) != pde) || ((last_pt + incr * count) != pt) || (count == AMDGPU_VM_MAX_UPDATE_SIZE)) { @@ -1135,20 +1142,6 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, dma_fence_put(fence); } } - /* - * Recurse into the subdirectories. This recursion is harmless because - * we only have a maximum of 5 layers. - */ - for (pt_idx = 0; pt_idx <= parent->last_entry_used; ++pt_idx) { - struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - - if (!entry->base.bo) - continue; - - r = amdgpu_vm_update_level(adev, vm, entry, level + 1); - if (r) - return r; - } return 0; @@ -1164,7 +1157,8 @@ error_free: * * Mark all PD level as invalid after an error. */ -static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent) +static void amdgpu_vm_invalidate_level(struct amdgpu_vm *vm, + struct amdgpu_vm_pt *parent) { unsigned pt_idx; @@ -1179,7 +1173,10 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm_pt *parent) continue; entry->addr = ~0ULL; - amdgpu_vm_invalidate_level(entry); + spin_lock(&vm->status_lock); + list_move(&entry->base.vm_status, &vm->relocated); + spin_unlock(&vm->status_lock); + amdgpu_vm_invalidate_level(vm, entry); } } @@ -1197,9 +1194,38 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev, { int r; - r = amdgpu_vm_update_level(adev, vm, &vm->root, 0); - if (r) - amdgpu_vm_invalidate_level(&vm->root); + spin_lock(&vm->status_lock); + while (!list_empty(&vm->relocated)) { + struct amdgpu_vm_bo_base *bo_base; + struct amdgpu_bo *bo; + + bo_base = list_first_entry(&vm->relocated, + struct amdgpu_vm_bo_base, + vm_status); + spin_unlock(&vm->status_lock); + + bo = bo_base->bo->parent; + if (bo) { + struct amdgpu_vm_bo_base *parent; + struct amdgpu_vm_pt *pt; + + parent = list_first_entry(&bo->va, + struct amdgpu_vm_bo_base, + bo_list); + pt = container_of(parent, struct amdgpu_vm_pt, base); + + r = amdgpu_vm_update_level(adev, vm, pt); + if (r) { + amdgpu_vm_invalidate_level(vm, &vm->root); + return r; + } + spin_lock(&vm->status_lock); + } else { + spin_lock(&vm->status_lock); + list_del_init(&bo_base->vm_status); + } + } + spin_unlock(&vm->status_lock); if (vm->use_cpu_for_update) { /* Flush HDP */ @@ -1601,7 +1627,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, error_free: amdgpu_job_free(job); - amdgpu_vm_invalidate_level(&vm->root); + amdgpu_vm_invalidate_level(vm, &vm->root); return r; } @@ -2391,9 +2417,13 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, continue; } - /* Don't add page tables to the moved state */ - if (bo->tbo.type == ttm_bo_type_kernel) + if (bo->tbo.type == ttm_bo_type_kernel) { + spin_lock(&bo_base->vm->status_lock); + if (list_empty(&bo_base->vm_status)) + list_add(&bo_base->vm_status, &vm->relocated); + spin_unlock(&bo_base->vm->status_lock); continue; + } spin_lock(&bo_base->vm->status_lock); list_move(&bo_base->vm_status, &bo_base->vm->moved); @@ -2483,6 +2513,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->reserved_vmid[i] = NULL; spin_lock_init(&vm->status_lock); INIT_LIST_HEAD(&vm->evicted); + INIT_LIST_HEAD(&vm->relocated); INIT_LIST_HEAD(&vm->moved); INIT_LIST_HEAD(&vm->freed); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 4e465e817fe8..c3753afe9853 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -129,6 +129,9 @@ struct amdgpu_vm { /* BOs who needs a validation */ struct list_head evicted; + /* PT BOs which relocated and their parent need an update */ + struct list_head relocated; + /* BOs moved, but not yet updated in the PT */ struct list_head moved; -- cgit From 570144c6522f5f332635d20dfa278cfcc764229c Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 30 Aug 2017 15:38:45 +0200 Subject: drm/amdgpu: cleanup the VM code a bit more MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The src isn't used any more after GART hack removal. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 20 +++++--------------- 1 file changed, 5 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index faa08d5728da..1582feba9289 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1476,7 +1476,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, * * @adev: amdgpu_device pointer * @exclusive: fence we need to sync to - * @src: address where to copy page table entries from * @pages_addr: DMA addresses to use for mapping * @vm: requested vm * @start: start of mapped range @@ -1490,7 +1489,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, */ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, struct dma_fence *exclusive, - uint64_t src, dma_addr_t *pages_addr, struct amdgpu_vm *vm, uint64_t start, uint64_t last, @@ -1508,7 +1506,6 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, memset(¶ms, 0, sizeof(params)); params.adev = adev; params.vm = vm; - params.src = src; /* sync to everything on unmapping */ if (!(flags & AMDGPU_PTE_VALID)) @@ -1548,13 +1545,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, /* one PDE write for each huge page */ ndw += ((nptes >> adev->vm_manager.block_size) + 1) * 6; - if (src) { - /* only copy commands needed */ - ndw += ncmds * 7; - - params.func = amdgpu_vm_do_copy_ptes; - - } else if (pages_addr) { + if (pages_addr) { /* copy commands needed */ ndw += ncmds * 7; @@ -1579,7 +1570,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, params.ib = &job->ibs[0]; - if (!src && pages_addr) { + if (pages_addr) { uint64_t *pte; unsigned i; @@ -1656,7 +1647,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, struct drm_mm_node *nodes, struct dma_fence **fence) { - uint64_t pfn, src = 0, start = mapping->start; + uint64_t pfn, start = mapping->start; int r; /* normally,bo_va->flags only contians READABLE and WIRTEABLE bit go here @@ -1711,8 +1702,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, addr += pfn << PAGE_SHIFT; last = min((uint64_t)mapping->last, start + max_entries - 1); - r = amdgpu_vm_bo_update_mapping(adev, exclusive, - src, pages_addr, vm, + r = amdgpu_vm_bo_update_mapping(adev, exclusive, pages_addr, vm, start, last, flags, addr, fence); if (r) @@ -1973,7 +1963,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, if (vm->pte_support_ats) init_pte_value = AMDGPU_PTE_SYSTEM; - r = amdgpu_vm_bo_update_mapping(adev, NULL, 0, NULL, vm, + r = amdgpu_vm_bo_update_mapping(adev, NULL, NULL, vm, mapping->start, mapping->last, init_pte_value, 0, &f); amdgpu_vm_free_mapping(adev, vm, mapping, f); -- cgit From 705e519e0ef1d1824c28ced3b1857a8608853dfb Mon Sep 17 00:00:00 2001 From: Christian König Date: Thu, 8 Jun 2017 11:15:16 +0200 Subject: drm/amdgpu: move hw generation check into amdgpu_doorbell_init v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This way we can safely call it on SI as well. v2: fix type in commit message Signed-off-by: Christian König Reviewed-by: Andy Shevchenko Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 1c5c44acaad2..77a32b79e8f2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -403,6 +403,15 @@ void amdgpu_pci_config_reset(struct amdgpu_device *adev) */ static int amdgpu_doorbell_init(struct amdgpu_device *adev) { + /* No doorbell on SI hardware generation */ + if (adev->asic_type < CHIP_BONAIRE) { + adev->doorbell.base = 0; + adev->doorbell.size = 0; + adev->doorbell.num_doorbells = 0; + adev->doorbell.ptr = NULL; + return 0; + } + /* doorbell bar mapping */ adev->doorbell.base = pci_resource_start(adev->pdev, 2); adev->doorbell.size = pci_resource_len(adev->pdev, 2); @@ -2075,9 +2084,8 @@ int amdgpu_device_init(struct amdgpu_device *adev, DRM_INFO("register mmio base: 0x%08X\n", (uint32_t)adev->rmmio_base); DRM_INFO("register mmio size: %u\n", (unsigned)adev->rmmio_size); - if (adev->asic_type >= CHIP_BONAIRE) - /* doorbell bar mapping */ - amdgpu_doorbell_init(adev); + /* doorbell bar mapping */ + amdgpu_doorbell_init(adev); /* io port mapping */ for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) { @@ -2304,8 +2312,7 @@ void amdgpu_device_fini(struct amdgpu_device *adev) adev->rio_mem = NULL; iounmap(adev->rmmio); adev->rmmio = NULL; - if (adev->asic_type >= CHIP_BONAIRE) - amdgpu_doorbell_fini(adev); + amdgpu_doorbell_fini(adev); amdgpu_debugfs_regs_cleanup(adev); } -- cgit From 0f2fc435d837213202bec3b8e26fbb67a4d6df24 Mon Sep 17 00:00:00 2001 From: Christian König Date: Thu, 31 Aug 2017 10:46:20 +0200 Subject: drm/amdgpu: fix new PD update code for Vega10 v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We need to refer to the parent instead of the root BO for multi level page tables on Vega10. Also don't set the PDE_PTE bit. v2: Don't set the PDE_PTE bit either. Signed-off-by: Christian König Reviewed-and-Tested-by: Roger He Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 1582feba9289..d3c48557555c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -309,7 +309,7 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, /* Keep a reference to the root directory to avoid * freeing them up in the wrong order. */ - pt->parent = amdgpu_bo_ref(vm->root.base.bo); + pt->parent = amdgpu_bo_ref(parent->base.bo); entry->base.vm = vm; entry->base.bo = pt; @@ -317,7 +317,7 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, spin_lock(&vm->status_lock); list_add(&entry->base.vm_status, &vm->relocated); spin_unlock(&vm->status_lock); - entry->addr = ~0ULL; + entry->addr = 0; } if (level < adev->vm_manager.num_level) { -- cgit From 4f5839c56ec38e2f8fcc59ca0e01defa8702987b Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 29 Aug 2017 16:07:31 +0200 Subject: drm/amdgpu: restrict userptr even more MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Don't allow them to be GEM imported into another process. Signed-off-by: Christian König Reviewed-by: Felix Kuehling Acked-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index d02880640ee7..e32a2b55b54f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -112,7 +112,13 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj, struct amdgpu_fpriv *fpriv = file_priv->driver_priv; struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_va *bo_va; + struct mm_struct *mm; int r; + + mm = amdgpu_ttm_tt_get_usermm(abo->tbo.ttm); + if (mm && mm != current->mm) + return -EPERM; + r = amdgpu_bo_reserve(abo, false); if (r) return r; -- cgit From 73fb16e7ebee12953de32a7a2552e0cf2bf74ebf Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 16 Aug 2017 11:13:48 +0200 Subject: drm/amdgpu: add support for per VM BOs v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Per VM BOs are handled like VM PDs and PTs. They are always valid and don't need to be specified in the BO lists. v2: validate PDs/PTs first Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 79 ++++++++++++++++++++++++---------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 ++- 3 files changed, 60 insertions(+), 26 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 8bf178a912f2..b57adc0723cb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -822,7 +822,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) } - r = amdgpu_vm_clear_moved(adev, vm, &p->job->sync); + r = amdgpu_vm_handle_moved(adev, vm, &p->job->sync); if (amdgpu_vm_debug && p->bo_list) { /* Invalidate all BOs to test for userspace bugs */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index d3c48557555c..26eb7dce5fe5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -189,14 +189,18 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, spin_unlock(&glob->lru_lock); } - if (vm->use_cpu_for_update) { + if (bo->tbo.type == ttm_bo_type_kernel && + vm->use_cpu_for_update) { r = amdgpu_bo_kmap(bo, NULL); if (r) return r; } spin_lock(&vm->status_lock); - list_move(&bo_base->vm_status, &vm->relocated); + if (bo->tbo.type != ttm_bo_type_kernel) + list_move(&bo_base->vm_status, &vm->moved); + else + list_move(&bo_base->vm_status, &vm->relocated); } spin_unlock(&vm->status_lock); @@ -1985,20 +1989,23 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, } /** - * amdgpu_vm_clear_moved - clear moved BOs in the PT + * amdgpu_vm_handle_moved - handle moved BOs in the PT * * @adev: amdgpu_device pointer * @vm: requested vm + * @sync: sync object to add fences to * - * Make sure all moved BOs are cleared in the PT. + * Make sure all BOs which are moved are updated in the PTs. * Returns 0 for success. * - * PTs have to be reserved and mutex must be locked! + * PTs have to be reserved! */ -int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_sync *sync) +int amdgpu_vm_handle_moved(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_sync *sync) { struct amdgpu_bo_va *bo_va = NULL; + bool clear; int r = 0; spin_lock(&vm->status_lock); @@ -2007,7 +2014,10 @@ int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amdgpu_bo_va, base.vm_status); spin_unlock(&vm->status_lock); - r = amdgpu_vm_bo_update(adev, bo_va, true); + /* Per VM BOs never need to bo cleared in the page tables */ + clear = bo_va->base.bo->tbo.resv != vm->root.base.bo->tbo.resv; + + r = amdgpu_vm_bo_update(adev, bo_va, clear); if (r) return r; @@ -2059,6 +2069,37 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev, return bo_va; } + +/** + * amdgpu_vm_bo_insert_mapping - insert a new mapping + * + * @adev: amdgpu_device pointer + * @bo_va: bo_va to store the address + * @mapping: the mapping to insert + * + * Insert a new mapping into all structures. + */ +static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, + struct amdgpu_bo_va *bo_va, + struct amdgpu_bo_va_mapping *mapping) +{ + struct amdgpu_vm *vm = bo_va->base.vm; + struct amdgpu_bo *bo = bo_va->base.bo; + + list_add(&mapping->list, &bo_va->invalids); + amdgpu_vm_it_insert(mapping, &vm->va); + + if (mapping->flags & AMDGPU_PTE_PRT) + amdgpu_vm_prt_get(adev); + + if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) { + spin_lock(&vm->status_lock); + list_move(&bo_va->base.vm_status, &vm->moved); + spin_unlock(&vm->status_lock); + } + trace_amdgpu_vm_bo_map(bo_va, mapping); +} + /** * amdgpu_vm_bo_map - map bo inside a vm * @@ -2110,18 +2151,12 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, if (!mapping) return -ENOMEM; - INIT_LIST_HEAD(&mapping->list); mapping->start = saddr; mapping->last = eaddr; mapping->offset = offset; mapping->flags = flags; - list_add(&mapping->list, &bo_va->invalids); - amdgpu_vm_it_insert(mapping, &vm->va); - - if (flags & AMDGPU_PTE_PRT) - amdgpu_vm_prt_get(adev); - trace_amdgpu_vm_bo_map(bo_va, mapping); + amdgpu_vm_bo_insert_map(adev, bo_va, mapping); return 0; } @@ -2148,7 +2183,6 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, { struct amdgpu_bo_va_mapping *mapping; struct amdgpu_bo *bo = bo_va->base.bo; - struct amdgpu_vm *vm = bo_va->base.vm; uint64_t eaddr; int r; @@ -2182,12 +2216,7 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, mapping->offset = offset; mapping->flags = flags; - list_add(&mapping->list, &bo_va->invalids); - amdgpu_vm_it_insert(mapping, &vm->va); - - if (flags & AMDGPU_PTE_PRT) - amdgpu_vm_prt_get(adev); - trace_amdgpu_vm_bo_map(bo_va, mapping); + amdgpu_vm_bo_insert_map(adev, bo_va, mapping); return 0; } @@ -2402,7 +2431,11 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, bo_base->moved = true; if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) { spin_lock(&bo_base->vm->status_lock); - list_move(&bo_base->vm_status, &vm->evicted); + if (bo->tbo.type == ttm_bo_type_kernel) + list_move(&bo_base->vm_status, &vm->evicted); + else + list_move_tail(&bo_base->vm_status, + &vm->evicted); spin_unlock(&bo_base->vm->status_lock); continue; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index c3753afe9853..90b7741d024b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -249,8 +249,9 @@ int amdgpu_vm_update_directories(struct amdgpu_device *adev, int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct dma_fence **fence); -int amdgpu_vm_clear_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm, - struct amdgpu_sync *sync); +int amdgpu_vm_handle_moved(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_sync *sync); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear); -- cgit From e1eb899b45781b9bb77e6d7772d6e67bb0bc1a18 Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 25 Aug 2017 09:14:43 +0200 Subject: drm/amdgpu: add IOCTL interface for per VM BOs v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Add the IOCTL interface so that applications can allocate per VM BOs. Still WIP since not all corner cases are tested yet, but this reduces average CS overhead for 10K BOs from 21ms down to 48us. v2: add some extra checks, remove the WIP tag v3: rename new flag to AMDGPU_GEM_CREATE_VM_ALWAYS_VALID Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 7 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 63 ++++++++++++++++++++++--------- drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 3 +- 4 files changed, 53 insertions(+), 22 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 103635ab784c..5809f55e0d9d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -455,9 +455,10 @@ struct amdgpu_sa_bo { */ void amdgpu_gem_force_release(struct amdgpu_device *adev); int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, - int alignment, u32 initial_domain, - u64 flags, bool kernel, - struct drm_gem_object **obj); + int alignment, u32 initial_domain, + u64 flags, bool kernel, + struct reservation_object *resv, + struct drm_gem_object **obj); int amdgpu_mode_dumb_create(struct drm_file *file_priv, struct drm_device *dev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index 9afa9c097e1f..b6cb276f0a70 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -149,7 +149,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_VRAM_CLEARED, - true, &gobj); + true, NULL, &gobj); if (ret) { pr_err("failed to allocate framebuffer (%d)\n", aligned_size); return -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index e32a2b55b54f..f1e61b3df640 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -44,11 +44,12 @@ void amdgpu_gem_object_free(struct drm_gem_object *gobj) } int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, - int alignment, u32 initial_domain, - u64 flags, bool kernel, - struct drm_gem_object **obj) + int alignment, u32 initial_domain, + u64 flags, bool kernel, + struct reservation_object *resv, + struct drm_gem_object **obj) { - struct amdgpu_bo *robj; + struct amdgpu_bo *bo; int r; *obj = NULL; @@ -59,7 +60,7 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, retry: r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain, - flags, NULL, NULL, 0, &robj); + flags, NULL, resv, 0, &bo); if (r) { if (r != -ERESTARTSYS) { if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) { @@ -71,7 +72,7 @@ retry: } return r; } - *obj = &robj->gem_base; + *obj = &bo->gem_base; return 0; } @@ -119,6 +120,10 @@ int amdgpu_gem_object_open(struct drm_gem_object *obj, if (mm && mm != current->mm) return -EPERM; + if (abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID && + abo->tbo.resv != vm->root.base.bo->tbo.resv) + return -EPERM; + r = amdgpu_bo_reserve(abo, false); if (r) return r; @@ -142,13 +147,14 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_list_entry vm_pd; - struct list_head list; + struct list_head list, duplicates; struct ttm_validate_buffer tv; struct ww_acquire_ctx ticket; struct amdgpu_bo_va *bo_va; int r; INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&duplicates); tv.bo = &bo->tbo; tv.shared = true; @@ -156,7 +162,7 @@ void amdgpu_gem_object_close(struct drm_gem_object *obj, amdgpu_vm_get_pd_bo(vm, &list, &vm_pd); - r = ttm_eu_reserve_buffers(&ticket, &list, false, NULL); + r = ttm_eu_reserve_buffers(&ticket, &list, false, &duplicates); if (r) { dev_err(adev->dev, "leaking bo va because " "we fail to reserve bo (%d)\n", r); @@ -191,9 +197,12 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_fpriv *fpriv = filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; union drm_amdgpu_gem_create *args = data; uint64_t flags = args->in.domain_flags; uint64_t size = args->in.bo_size; + struct reservation_object *resv = NULL; struct drm_gem_object *gobj; uint32_t handle; int r; @@ -202,7 +211,8 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, if (flags & ~(AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_NO_CPU_ACCESS | AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_VRAM_CLEARED)) + AMDGPU_GEM_CREATE_VRAM_CLEARED | + AMDGPU_GEM_CREATE_VM_ALWAYS_VALID)) return -EINVAL; /* reject invalid gem domains */ @@ -229,9 +239,25 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, } size = roundup(size, PAGE_SIZE); + if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { + r = amdgpu_bo_reserve(vm->root.base.bo, false); + if (r) + return r; + + resv = vm->root.base.bo->tbo.resv; + } + r = amdgpu_gem_object_create(adev, size, args->in.alignment, (u32)(0xffffffff & args->in.domains), - flags, false, &gobj); + flags, false, resv, &gobj); + if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { + if (!r) { + struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj); + + abo->parent = amdgpu_bo_ref(vm->root.base.bo); + } + amdgpu_bo_unreserve(vm->root.base.bo); + } if (r) return r; @@ -273,9 +299,8 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, } /* create a gem object to contain this object in */ - r = amdgpu_gem_object_create(adev, args->size, 0, - AMDGPU_GEM_DOMAIN_CPU, 0, - 0, &gobj); + r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU, + 0, 0, NULL, &gobj); if (r) return r; @@ -527,7 +552,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, struct amdgpu_bo_list_entry vm_pd; struct ttm_validate_buffer tv; struct ww_acquire_ctx ticket; - struct list_head list; + struct list_head list, duplicates; uint64_t va_flags; int r = 0; @@ -563,6 +588,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, } INIT_LIST_HEAD(&list); + INIT_LIST_HEAD(&duplicates); if ((args->operation != AMDGPU_VA_OP_CLEAR) && !(args->flags & AMDGPU_VM_PAGE_PRT)) { gobj = drm_gem_object_lookup(filp, args->handle); @@ -579,7 +605,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, amdgpu_vm_get_pd_bo(&fpriv->vm, &list, &vm_pd); - r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL); + r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates); if (r) goto error_unref; @@ -645,6 +671,7 @@ error_unref: int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { + struct amdgpu_device *adev = dev->dev_private; struct drm_amdgpu_gem_op *args = data; struct drm_gem_object *gobj; struct amdgpu_bo *robj; @@ -692,6 +719,9 @@ int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, if (robj->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) robj->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; + if (robj->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) + amdgpu_vm_bo_invalidate(adev, robj, true); + amdgpu_bo_unreserve(robj); break; default: @@ -721,8 +751,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv, r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_VRAM, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - ttm_bo_type_device, - &gobj); + false, NULL, &gobj); if (r) return -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 5b3f92891f89..7e0826469b5e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -136,7 +136,8 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, { struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj); - if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) || + bo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) return ERR_PTR(-EPERM); return drm_gem_prime_export(dev, gobj, flags); -- cgit From fd8bf087dffc0bce047c5aea2afcb8f821e48db1 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 29 Aug 2017 16:14:32 +0200 Subject: drm/amdgpu: bump version for support of local BOs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 0f16986ec5bc..792b11795a81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -69,9 +69,10 @@ * - 3.17.0 - Add AMDGPU_NUM_VRAM_CPU_PAGE_FAULTS. * - 3.18.0 - Export gpu always on cu bitmap * - 3.19.0 - Add support for UVD MJPEG decode + * - 3.20.0 - Add support for local BOs */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 19 +#define KMS_DRIVER_MINOR 20 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; -- cgit From 35c32f20a7e28b82b61d44a164b756d933bca4a4 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 08:41:54 -0400 Subject: drm/amd/amdgpu: Tidy up register list formatting. Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 96 +++++++++++++++++++++++------------ 1 file changed, 64 insertions(+), 32 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 7c06d1b99d99..8b31f8427d9a 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -66,38 +66,70 @@ MODULE_FIRMWARE("amdgpu/raven_rlc.bin"); static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = { - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14)}, - {SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), - SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15)} + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID0), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID0) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID1_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID1), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID1) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID2_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID2), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID2) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID3_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID3), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID3) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID4_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID4), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID4) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID5_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID5), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID5) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID6_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID6), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID6) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID7_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID7), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID7) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID8_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID8), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID8) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID9_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID9), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID9) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID10_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID10), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID10) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID11_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID11), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID11) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID12_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID12), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID12)}, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID13_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID13), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID13) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID14_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID14), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID14) }, + { SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_BASE), + SOC15_REG_OFFSET(GC, 0, mmGDS_VMID15_SIZE), + SOC15_REG_OFFSET(GC, 0, mmGDS_GWS_VMID15), + SOC15_REG_OFFSET(GC, 0, mmGDS_OA_VMID15) } }; static const u32 golden_settings_gc_9_0[] = -- cgit From 91629eff745e4c27d6501f1949e502868d9d5706 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 08:50:23 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_ngg_en() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 26 +++++++++----------------- 1 file changed, 9 insertions(+), 17 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 8b31f8427d9a..44885ffaadb7 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1152,30 +1152,22 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev) { struct amdgpu_ring *ring = &adev->gfx.gfx_ring[0]; int r; - u32 data; - u32 size; - u32 base; + u32 data, base; if (!amdgpu_ngg) return 0; /* Program buffer size */ - data = 0; - size = adev->gfx.ngg.buf[NGG_PRIM].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, size); - - size = adev->gfx.ngg.buf[NGG_POS].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, size); - + data = REG_SET_FIELD(0, WD_BUF_RESOURCE_1, INDEX_BUF_SIZE, + adev->gfx.ngg.buf[NGG_PRIM].size >> 8); + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_1, POS_BUF_SIZE, + adev->gfx.ngg.buf[NGG_POS].size >> 8); WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_1, data); - data = 0; - size = adev->gfx.ngg.buf[NGG_CNTL].size / 256; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, size); - - size = adev->gfx.ngg.buf[NGG_PARAM].size / 1024; - data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, size); - + data = REG_SET_FIELD(0, WD_BUF_RESOURCE_2, CNTL_SB_BUF_SIZE, + adev->gfx.ngg.buf[NGG_CNTL].size >> 8); + data = REG_SET_FIELD(data, WD_BUF_RESOURCE_2, PARAM_BUF_SIZE, + adev->gfx.ngg.buf[NGG_PARAM].size >> 10); WREG32_SOC15(GC, 0, mmWD_BUF_RESOURCE_2, data); /* Program buffer base address */ -- cgit From 0e5293d07b92dfc1ada2d4a6c9057c1ee8179399 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 08:59:00 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_save_restore_machine() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 44885ffaadb7..0d6a1b4075e9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1764,11 +1764,7 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) static void gfx_v9_0_enable_save_restore_machine(struct amdgpu_device *adev) { - u32 tmp = 0; - - tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL)); - tmp |= RLC_SRM_CNTL__SRM_ENABLE_MASK; - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL), tmp); + WREG32_FIELD15(GC, 0, RLC_SRM_CNTL, SRM_ENABLE, 1); } static void pwr_10_0_gfxip_control_over_cgpg(struct amdgpu_device *adev, -- cgit From e24c7f06db418c34c179233724656ec0efae6bc8 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:01:11 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_sck_slow_down_on_power_up() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 0d6a1b4075e9..5a301c865bf1 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1842,16 +1842,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_up(struct amdgpu_device *adev uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - if (default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PU_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + SMU_CLK_SLOWDOWN_ON_PU_ENABLE, + enable ? 1 : 0); + if (default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *adev, -- cgit From b926fe8efc54062f984b70e79bd71dddba176816 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:02:33 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_sck_slow_down_on_power_down() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 5a301c865bf1..1f95ca8e476b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1856,16 +1856,11 @@ static void gfx_v9_0_enable_sck_slow_down_on_power_down(struct amdgpu_device *ad uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data |= RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data &= ~RLC_PG_CNTL__SMU_CLK_SLOWDOWN_ON_PD_ENABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + SMU_CLK_SLOWDOWN_ON_PD_ENABLE, + enable ? 1 : 0); + if(default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev, -- cgit From 54cfe0fc54a16de265aac79fb8c7c7cd3131c964 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:12:47 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_cp_power_gating() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 1f95ca8e476b..61b3362d3c20 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1870,16 +1870,11 @@ static void gfx_v9_0_enable_cp_power_gating(struct amdgpu_device *adev, uint32_t default_data = 0; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - - if (enable == true) { - data &= ~RLC_PG_CNTL__CP_PG_DISABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } else { - data |= RLC_PG_CNTL__CP_PG_DISABLE_MASK; - if(default_data != data) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); - } + data = REG_SET_FIELD(data, RLC_PG_CNTL, + CP_PG_DISABLE, + enable ? 0 : 1); + if(default_data != data) + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev, -- cgit From f55ee212ee263de6f9a56530095bfafbecc5863d Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:14:41 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_gfx_cg_power_gating() Make it consistent in style with the other CG/PG enable functions... Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 61b3362d3c20..b0805b1e7b29 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1883,10 +1883,9 @@ static void gfx_v9_0_enable_gfx_cg_power_gating(struct amdgpu_device *adev, uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_POWER_GATING_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + GFX_POWER_GATING_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } -- cgit From 513f81332a498ce989559435a74491897dda8af2 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:27:22 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_gfx_pipeline_powergating() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index b0805b1e7b29..1a928f6dc801 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1896,10 +1896,9 @@ static void gfx_v9_0_enable_gfx_pipeline_powergating(struct amdgpu_device *adev, uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__GFX_PIPELINE_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + GFX_PIPELINE_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); -- cgit From 7915c8fd7ec70f65c66c268d8d19f563b5a0162d Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:31:01 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_gfx_static_mg_power_gating() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 1a928f6dc801..22346bbe9a4c 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1913,10 +1913,9 @@ static void gfx_v9_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__STATIC_PER_CU_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + STATIC_PER_CU_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } -- cgit From e567fa69f2045b5484a232a51ebbfe92f7c65386 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:31:23 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_enable_gfx_dynamic_mg_power_gating() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 22346bbe9a4c..16d1a429ec9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1926,10 +1926,9 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad uint32_t data, default_data; default_data = data = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL)); - if (enable == true) - data |= RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; - else - data &= ~RLC_PG_CNTL__DYN_PER_CU_PG_ENABLE_MASK; + data = REG_SET_FIELD(data, RLC_PG_CNTL, + DYN_PER_CU_PG_ENABLE, + enable ? 1 : 0); if(default_data != data) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_PG_CNTL), data); } -- cgit From b08796cee109b35996357286ca878611e9dae4a4 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:40:38 -0400 Subject: drm/amd/amdgpu: Tidy up gfx_v9_0_rlc_stop() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 16d1a429ec9f..e216aa68f51f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1968,13 +1968,8 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) void gfx_v9_0_rlc_stop(struct amdgpu_device *adev) { - u32 tmp = RREG32_SOC15(GC, 0, mmRLC_CNTL); - - tmp = REG_SET_FIELD(tmp, RLC_CNTL, RLC_ENABLE_F32, 0); - WREG32_SOC15(GC, 0, mmRLC_CNTL, tmp); - + WREG32_FIELD15(GC, 0, RLC_CNTL, RLC_ENABLE_F32, 0); gfx_v9_0_enable_gui_idle_interrupt(adev, false); - gfx_v9_0_wait_for_rlc_serdes(adev); } -- cgit From 78888cff5cbaf5ec712b20a216f500b7d4493e77 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:43:47 -0400 Subject: drm/amd/amdgpu: Fix indentation in gfx_v9_0_mqd_init() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index e216aa68f51f..2ba02739a228 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2489,10 +2489,10 @@ static int gfx_v9_0_mqd_init(struct amdgpu_ring *ring) DOORBELL_SOURCE, 0); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_HIT, 0); - } - else + } else { tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 0); + } mqd->cp_hqd_pq_doorbell_control = tmp; -- cgit From 2b9bdfa70faf8c00969f91d3c4548a0df6071b90 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 31 Aug 2017 09:48:11 -0400 Subject: drm/amd/amdgpu: Simplify gfx_v9_0_wait_for_idle() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 2ba02739a228..a21182debb3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2937,15 +2937,10 @@ static bool gfx_v9_0_is_idle(void *handle) static int gfx_v9_0_wait_for_idle(void *handle) { unsigned i; - u32 tmp; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { - /* read MC_STATUS */ - tmp = RREG32_SOC15(GC, 0, mmGRBM_STATUS) & - GRBM_STATUS__GUI_ACTIVE_MASK; - - if (!REG_GET_FIELD(tmp, GRBM_STATUS, GUI_ACTIVE)) + if (gfx_v9_0_is_idle(handle)) return 0; udelay(1); } -- cgit From 6849d47cabc36e8f2697043f8c81e7719876dfd3 Mon Sep 17 00:00:00 2001 From: Roger He Date: Wed, 30 Aug 2017 13:01:19 +0800 Subject: drm/amdgpu: handle all fragment sizes v4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This can improve performance for some cases. v2 (chk): handle all sizes, simplify the patch quite a bit v3 (chk): adjust dw estimation as well v4 (chk): use single loop, make end mask 64bit Signed-off-by: Roger He Signed-off-by: Christian König Tested-by: Roger He Reviewed-by: Felix Kuehling Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 55 ++++++++++++++++------------------ 1 file changed, 26 insertions(+), 29 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 26eb7dce5fe5..b83e0fa1f269 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1420,8 +1420,6 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, uint64_t start, uint64_t end, uint64_t dst, uint64_t flags) { - int r; - /** * The MC L1 TLB supports variable sized pages, based on a fragment * field in the PTE. When this field is set to a non-zero value, page @@ -1440,39 +1438,38 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, * Userspace can support this by aligning virtual base address and * allocation size to the fragment size. */ - unsigned pages_per_frag = params->adev->vm_manager.fragment_size; - uint64_t frag_flags = AMDGPU_PTE_FRAG(pages_per_frag); - uint64_t frag_align = 1 << pages_per_frag; - - uint64_t frag_start = ALIGN(start, frag_align); - uint64_t frag_end = end & ~(frag_align - 1); + unsigned max_frag = params->adev->vm_manager.fragment_size; + int r; /* system pages are non continuously */ - if (params->src || !(flags & AMDGPU_PTE_VALID) || - (frag_start >= frag_end)) + if (params->src || !(flags & AMDGPU_PTE_VALID)) return amdgpu_vm_update_ptes(params, start, end, dst, flags); - /* handle the 4K area at the beginning */ - if (start != frag_start) { - r = amdgpu_vm_update_ptes(params, start, frag_start, - dst, flags); + while (start != end) { + uint64_t frag_flags, frag_end; + unsigned frag; + + /* This intentionally wraps around if no bit is set */ + frag = min((unsigned)ffs(start) - 1, + (unsigned)fls64(end - start) - 1); + if (frag >= max_frag) { + frag_flags = AMDGPU_PTE_FRAG(max_frag); + frag_end = end & ~((1ULL << max_frag) - 1); + } else { + frag_flags = AMDGPU_PTE_FRAG(frag); + frag_end = start + (1 << frag); + } + + r = amdgpu_vm_update_ptes(params, start, frag_end, dst, + flags | frag_flags); if (r) return r; - dst += (frag_start - start) * AMDGPU_GPU_PAGE_SIZE; - } - - /* handle the area in the middle */ - r = amdgpu_vm_update_ptes(params, frag_start, frag_end, dst, - flags | frag_flags); - if (r) - return r; - /* handle the 4K area at the end */ - if (frag_end != end) { - dst += (frag_end - frag_start) * AMDGPU_GPU_PAGE_SIZE; - r = amdgpu_vm_update_ptes(params, frag_end, end, dst, flags); + dst += (frag_end - start) * AMDGPU_GPU_PAGE_SIZE; + start = frag_end; } - return r; + + return 0; } /** @@ -1562,8 +1559,8 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, /* set page commands needed */ ndw += ncmds * 10; - /* two extra commands for begin/end of fragment */ - ndw += 2 * 10; + /* extra commands for begin/end fragments */ + ndw += 2 * 10 * adev->vm_manager.fragment_size; params.func = amdgpu_vm_do_set_ptes; } -- cgit From f5830465967799de0334340d1888f7d2c0bc17f5 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Fri, 1 Sep 2017 15:07:41 +0800 Subject: drm/amdgpu: declare the new firmware files needed by polaris asics Signed-off-by: Evan Quan Reviewed-by: Flora Cui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 6666fcd8b08f..666a1545e949 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -125,24 +125,39 @@ MODULE_FIRMWARE("amdgpu/fiji_mec2.bin"); MODULE_FIRMWARE("amdgpu/fiji_rlc.bin"); MODULE_FIRMWARE("amdgpu/polaris11_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin"); MODULE_FIRMWARE("amdgpu/polaris10_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris10_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_rlc.bin"); MODULE_FIRMWARE("amdgpu/polaris12_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_pfp_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_me_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_mec_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris12_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_rlc.bin"); static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = -- cgit From 481c2e94897eb74abcfb4a3cdb87f5f89499b93f Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 1 Sep 2017 14:46:19 +0200 Subject: drm/amdgpu: fix moved list handling in the VM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Only move BOs to the moved/relocated list when they aren't already on a list. This prevents accidential removal from the evicted list. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index b83e0fa1f269..c9223a5184de 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1178,7 +1178,8 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_vm *vm, entry->addr = ~0ULL; spin_lock(&vm->status_lock); - list_move(&entry->base.vm_status, &vm->relocated); + if (list_empty(&entry->base.vm_status)) + list_add(&entry->base.vm_status, &vm->relocated); spin_unlock(&vm->status_lock); amdgpu_vm_invalidate_level(vm, entry); } @@ -2091,7 +2092,8 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) { spin_lock(&vm->status_lock); - list_move(&bo_va->base.vm_status, &vm->moved); + if (list_empty(&bo_va->base.vm_status)) + list_add(&bo_va->base.vm_status, &vm->moved); spin_unlock(&vm->status_lock); } trace_amdgpu_vm_bo_map(bo_va, mapping); @@ -2446,7 +2448,8 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, } spin_lock(&bo_base->vm->status_lock); - list_move(&bo_base->vm_status, &bo_base->vm->moved); + if (list_empty(&bo_base->vm_status)) + list_add(&bo_base->vm_status, &vm->moved); spin_unlock(&bo_base->vm->status_lock); } } -- cgit From 70a9c6b9010848e101ff67a886f2e75209b2f0d1 Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 1 Sep 2017 09:22:56 +0200 Subject: drm/amdgpu: fix placement flags in amdgpu_ttm_bind MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Otherwise we lose the NO_EVICT flag and can try to evict pinned BOs. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 6ea96e1fb273..841a5699bef0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -819,7 +819,7 @@ int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem) placement.busy_placement = &placements; placements.fpfn = 0; placements.lpfn = adev->mc.gart_size >> PAGE_SHIFT; - placements.flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_TT; + placements.flags = bo->mem.placement | TTM_PL_FLAG_TT; r = ttm_bo_mem_space(bo, &placement, &tmp, true, false); if (unlikely(r)) -- cgit From f6886c472654e12d9cab4c4462a1c38c732468b8 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Fri, 1 Sep 2017 09:13:04 -0400 Subject: drm/amd/amdgpu: Support full range of GFX ring names MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Right now there's only one but the rest of the code is being setup to support more so might as well fix this up too. Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index a21182debb3d..2ab049c45b1d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1330,7 +1330,10 @@ static int gfx_v9_0_sw_init(void *handle) for (i = 0; i < adev->gfx.num_gfx_rings; i++) { ring = &adev->gfx.gfx_ring[i]; ring->ring_obj = NULL; - sprintf(ring->name, "gfx"); + if (!i) + sprintf(ring->name, "gfx"); + else + sprintf(ring->name, "gfx_%d", i); ring->use_doorbell = true; ring->doorbell_index = AMDGPU_DOORBELL64_GFX_RING0 << 1; r = amdgpu_ring_init(adev, ring, 1024, -- cgit From ae6d1416fa1624a6a2e4f2d392b3c0bdcc570712 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Fri, 1 Sep 2017 09:27:31 -0400 Subject: drm/amd/amdgpu: Simplify gmc_v9_0_vm_fault_interrupt_state() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 51 ++++++++++++----------------------- 1 file changed, 17 insertions(+), 34 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index d04d0b123212..1cb7aa2af683 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -77,7 +77,7 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { struct amdgpu_vmhub *hub; - u32 tmp, reg, bits, i; + u32 tmp, reg, bits, i, j; bits = VM_CONTEXT1_CNTL__RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | VM_CONTEXT1_CNTL__DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT_MASK | @@ -89,43 +89,26 @@ static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, switch (state) { case AMDGPU_IRQ_STATE_DISABLE: - /* MM HUB */ - hub = &adev->vmhub[AMDGPU_MMHUB]; - for (i = 0; i< 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp &= ~bits; - WREG32(reg, tmp); - } - - /* GFX HUB */ - hub = &adev->vmhub[AMDGPU_GFXHUB]; - for (i = 0; i < 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp &= ~bits; - WREG32(reg, tmp); + for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) { + hub = &adev->vmhub[j]; + for (i = 0; i < 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp &= ~bits; + WREG32(reg, tmp); + } } break; case AMDGPU_IRQ_STATE_ENABLE: - /* MM HUB */ - hub = &adev->vmhub[AMDGPU_MMHUB]; - for (i = 0; i< 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp |= bits; - WREG32(reg, tmp); + for (j = 0; j < AMDGPU_MAX_VMHUBS; j++) { + hub = &adev->vmhub[j]; + for (i = 0; i < 16; i++) { + reg = hub->vm_context0_cntl + i; + tmp = RREG32(reg); + tmp |= bits; + WREG32(reg, tmp); + } } - - /* GFX HUB */ - hub = &adev->vmhub[AMDGPU_GFXHUB]; - for (i = 0; i < 16; i++) { - reg = hub->vm_context0_cntl + i; - tmp = RREG32(reg); - tmp |= bits; - WREG32(reg, tmp); - } - break; default: break; } -- cgit From 846347c9f527bc6a771dca568c866704c290e103 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Fri, 1 Sep 2017 09:52:21 -0400 Subject: drm/amd/amdgpu: Tidy up gmc_v9_0_gart_enable() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 1cb7aa2af683..d7cfee807287 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -719,14 +719,11 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) if (r) return r; - tmp = RREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL); - tmp |= HDP_MISC_CNTL__FLUSH_INVALIDATE_CACHE_MASK; - WREG32_SOC15(HDP, 0, mmHDP_MISC_CNTL, tmp); + WREG32_FIELD15(HDP, 0, HDP_MISC_CNTL, FLUSH_INVALIDATE_CACHE, 1); tmp = RREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL); WREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL, tmp); - if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) value = false; else @@ -734,7 +731,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) gfxhub_v1_0_set_fault_enable_default(adev, value); mmhub_v1_0_set_fault_enable_default(adev, value); - gmc_v9_0_gart_flush_gpu_tlb(adev, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", -- cgit From 4d9c333a464aef3341357035cb75617ebb052c65 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Fri, 1 Sep 2017 09:53:44 -0400 Subject: drm/amd/amdgpu: Tidy up gmc_v9_0_hw_init() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index d7cfee807287..0766dad1fafe 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -749,17 +749,11 @@ static int gmc_v9_0_hw_init(void *handle) gmc_v9_0_init_golden_registers(adev); if (adev->mode_info.num_crtc) { - u32 tmp; - /* Lockout access through VGA aperture*/ - tmp = RREG32_SOC15(DCE, 0, mmVGA_HDP_CONTROL); - tmp = REG_SET_FIELD(tmp, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1); - WREG32_SOC15(DCE, 0, mmVGA_HDP_CONTROL, tmp); + WREG32_FIELD15(DCE, 0, VGA_HDP_CONTROL, VGA_MEMORY_DISABLE, 1); /* disable VGA render */ - tmp = RREG32_SOC15(DCE, 0, mmVGA_RENDER_CONTROL); - tmp = REG_SET_FIELD(tmp, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); - WREG32_SOC15(DCE, 0, mmVGA_RENDER_CONTROL, tmp); + WREG32_FIELD15(DCE, 0, VGA_RENDER_CONTROL, VGA_VSTATUS_CNTL, 0); } r = gmc_v9_0_gart_enable(adev); -- cgit From f053cd478ecfe9cbc9988912fe63a6d723838b2e Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Fri, 1 Sep 2017 09:55:04 -0400 Subject: drm/amd/amdgpu: Cleanup gmc_v9_0_suspend() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Even though fini returns 0 always it could theoretically fail in the future. Might as well return it instead of 0. Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 0766dad1fafe..7ca9cbec3004 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -795,9 +795,7 @@ static int gmc_v9_0_suspend(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - gmc_v9_0_hw_fini(adev); - - return 0; + return gmc_v9_0_hw_fini(adev); } static int gmc_v9_0_resume(void *handle) -- cgit From 1430f73beb7fd4655d3fbd2e373615eca016f9c3 Mon Sep 17 00:00:00 2001 From: Thierry Reding Date: Fri, 1 Sep 2017 12:54:01 -0400 Subject: drm/amdgpu: Use correct path to trace include MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The header comment in include/trace/define_trace.h specifies that the TRACE_INCLUDE_PATH needs to be relative to the define_trace.h header rather than the trace file including it. Most instances get that wrong and work around it by adding the $(src) directory to the include path. While this works, it is preferable to refer to the correct path to the trace file in the first place and avoid any workaround. Reviewed-by: Christian König Signed-off-by: Thierry Reding Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/Makefile | 2 -- drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h | 2 +- 2 files changed, 1 insertion(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 658bac0cdc5e..25a95c95df14 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -133,5 +133,3 @@ include $(FULL_AMD_PATH)/powerplay/Makefile amdgpu-y += $(AMD_POWERPLAY_FILES) obj-$(CONFIG_DRM_AMDGPU)+= amdgpu.o - -CFLAGS_amdgpu_trace_points.o := -I$(src) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index b1f97417241d..213988f336ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -417,5 +417,5 @@ TRACE_EVENT(amdgpu_ttm_bo_move, /* This part must be outside protection */ #undef TRACE_INCLUDE_PATH -#define TRACE_INCLUDE_PATH . +#define TRACE_INCLUDE_PATH ../../drivers/gpu/drm/amd/amdgpu #include -- cgit From df1e63942063a0638a5813e5904988d834db2665 Mon Sep 17 00:00:00 2001 From: Rex Zhu Date: Fri, 1 Sep 2017 13:46:20 +0800 Subject: drm/amd/powerplay: delete eventmgr layer in poweprlay Reviewed-by: Alex Deucher Signed-off-by: Rex Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c | 10 +++++----- drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c | 2 +- 3 files changed, 8 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index 8c96a4caa715..3eba4137508b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -383,8 +383,8 @@ struct amdgpu_dpm_funcs { #define amdgpu_dpm_set_mclk_od(adev, value) \ ((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value)) -#define amdgpu_dpm_dispatch_task(adev, event_id, input, output) \ - (adev)->powerplay.pp_funcs->dispatch_tasks((adev)->powerplay.pp_handle, (event_id), (input), (output)) +#define amdgpu_dpm_dispatch_task(adev, task_id, input, output) \ + ((adev)->powerplay.pp_funcs->dispatch_tasks)((adev)->powerplay.pp_handle, (task_id), (input), (output)) #define amgdpu_dpm_check_state_equal(adev, cps, rps, equal) (adev)->pm.funcs->check_state_equal((adev), (cps),(rps),(equal)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 7df503aedb69..10c5d78081ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -119,7 +119,7 @@ static ssize_t amdgpu_set_dpm_state(struct device *dev, } if (adev->pp_enabled) { - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL); + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL); } else { mutex_lock(&adev->pm.mutex); adev->pm.dpm.user_state = state; @@ -330,7 +330,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev, if (state != POWER_STATE_TYPE_INTERNAL_BOOT && state != POWER_STATE_TYPE_DEFAULT) { amdgpu_dpm_dispatch_task(adev, - AMD_PP_EVENT_ENABLE_USER_STATE, &state, NULL); + AMD_PP_TASK_ENABLE_USER_STATE, &state, NULL); adev->pp_force_state_enabled = true; } } @@ -559,7 +559,7 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev, if (adev->pp_enabled) { amdgpu_dpm_set_sclk_od(adev, (uint32_t)value); - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL); + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); } else if (adev->pm.funcs->set_sclk_od) { adev->pm.funcs->set_sclk_od(adev, (uint32_t)value); adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; @@ -605,7 +605,7 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev, if (adev->pp_enabled) { amdgpu_dpm_set_mclk_od(adev, (uint32_t)value); - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_READJUST_POWER_STATE, NULL, NULL); + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); } else if (adev->pm.funcs->set_mclk_od) { adev->pm.funcs->set_mclk_od(adev, (uint32_t)value); adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; @@ -1496,7 +1496,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) } if (adev->pp_enabled) { - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_DISPLAY_CONFIG_CHANGE, NULL, NULL); + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL, NULL); } else { mutex_lock(&adev->pm.mutex); adev->pm.dpm.new_active_crtcs = 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c index b7e1c026c0c8..b0c4db8098ed 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -128,7 +128,7 @@ static int amdgpu_pp_late_init(void *handle) if (adev->pp_enabled && adev->pm.dpm_enabled) { amdgpu_pm_sysfs_init(adev); - amdgpu_dpm_dispatch_task(adev, AMD_PP_EVENT_COMPLETE_INIT, NULL, NULL); + amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_COMPLETE_INIT, NULL, NULL); } return ret; -- cgit From a2138eaf97b4e053b229fe07e1bb4ecbe07e6769 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 11 Aug 2017 17:49:48 +0800 Subject: drm/amdgpu: fix wait_any_fence first is incorrect if hit NULL/signaled fence Signed-off-by: Monk Liu Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index b57adc0723cb..233b6f2f8427 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1384,6 +1384,7 @@ static int amdgpu_cs_wait_any_fence(struct amdgpu_device *adev, array[i] = fence; } else { /* NULL, the fence has been already signaled */ r = 1; + first = i; goto out; } } -- cgit From a216ab09955d6b77f3af4f0aba9255c5ddf382f5 Mon Sep 17 00:00:00 2001 From: Christian König Date: Sat, 2 Sep 2017 13:21:31 +0200 Subject: drm/amdgpu: fix userptr put_page handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Move calling put_page into the unpopulate callback. Otherwise we mess up the pages reference count when it is unbound multiple times. Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 6 ++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 14 +++++++++++++- 3 files changed, 16 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 5809f55e0d9d..cc6de0b46326 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1793,6 +1793,7 @@ void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes, void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain); bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo); int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); +void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages); int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, uint32_t flags); bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 233b6f2f8427..e58db0c69c6a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -474,10 +474,8 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, /* Check if we have user pages and nobody bound the BO already */ if (lobj->user_pages && bo->tbo.ttm->state != tt_bound) { - size_t size = sizeof(struct page *); - - size *= bo->tbo.ttm->num_pages; - memcpy(bo->tbo.ttm->pages, lobj->user_pages, size); + amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, + lobj->user_pages); binding_userptr = true; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 841a5699bef0..28e121984332 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -664,6 +664,18 @@ release_pages: return r; } +void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages) +{ + unsigned i; + + for (i = 0; i < ttm->num_pages; ++i) { + if (ttm->pages[i]) + put_page(ttm->pages[i]); + + ttm->pages[i] = pages ? pages[i] : NULL; + } +} + static void amdgpu_trace_dma_map(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); @@ -738,7 +750,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) set_page_dirty(page); mark_page_accessed(page); - put_page(page); } amdgpu_trace_dma_unmap(ttm); @@ -971,6 +982,7 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); if (gtt && gtt->userptr) { + amdgpu_ttm_tt_set_user_pages(ttm, NULL); kfree(ttm->sg); ttm->page_flags &= ~TTM_PAGE_FLAG_SG; return; -- cgit From aa4ec7ce7ec52c7230cfa73b06d79288b45fe1c9 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 5 Sep 2017 15:10:50 +0200 Subject: drm/amdgpu: revert "fix deadlock of reservation between cs and gpu reset v2" MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This reverts commit 10e709cb296c98424c03408d23e3addeddcd4088. The patch doesn't work at all: 1. The CS can still be blocked because of amdgpu_ctx_add_fence(). 2. The order of submission isn't correct any more. 3. We could end up using freed up memory because we now drop the ctx reference to early. This needs to be fixed cleanly by doing the context handling after the BO handling, but this is a larger task just avoid the obvious crashes for now. Signed-off-by: Christian König Reviewed-by: Monk Liu monk.liu@amd.com Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index e58db0c69c6a..3fe816f6beca 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1148,7 +1148,6 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence); job->uf_sequence = cs->out.handle; amdgpu_job_free_resources(job); - amdgpu_cs_parser_fini(p, 0, true); trace_amdgpu_cs_ioctl(job); amd_sched_entity_push_job(&job->base); @@ -1206,10 +1205,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) goto out; r = amdgpu_cs_submit(&parser, cs); - if (r) - goto out; - return 0; out: amdgpu_cs_parser_fini(&parser, r, reserved_buffers); return r; -- cgit From 862095237c392887819e98018d6dc187d8c78726 Mon Sep 17 00:00:00 2001 From: Bas Nieuwenhuizen Date: Thu, 7 Sep 2017 13:23:21 +0200 Subject: drm/amdgpu: Account for shadow PTs in mapping update IB size. MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When amdgpu_vm_frag_ptes calls amdgpu_vm_update_ptes and the pt has a shadow PT we mirror all the write to the shadow PT too, which results in twice the commands. Signed-off-by: Bas Nieuwenhuizen Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index c9223a5184de..545531db66db 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1536,10 +1536,12 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, nptes = last - start + 1; /* - * reserve space for one command every (1 << BLOCK_SIZE) + * reserve space for two commands every (1 << BLOCK_SIZE) * entries or 2k dwords (whatever is smaller) + * + * The second command is for the shadow pagetables. */ - ncmds = (nptes >> min(adev->vm_manager.block_size, 11u)) + 1; + ncmds = ((nptes >> min(adev->vm_manager.block_size, 11u)) + 1) * 2; /* padding, etc. */ ndw = 64; -- cgit From b72cf4fca2bb786e20864b5e8755105aa9626fb4 Mon Sep 17 00:00:00 2001 From: Christian König Date: Sun, 3 Sep 2017 15:22:06 +0200 Subject: drm/amdgpu: move taking mmap_sem into get_user_pages v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This didn't helped as intended, just simplify the code. v2: unlock mmap_sem in the error path as well Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 12 +----------- drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c | 4 ---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 8 +++++++- 3 files changed, 8 insertions(+), 16 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 3fe816f6beca..283a216ee758 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -500,18 +500,14 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, struct amdgpu_fpriv *fpriv = p->filp->driver_priv; struct amdgpu_bo_list_entry *e; struct list_head duplicates; - bool need_mmap_lock = false; unsigned i, tries = 10; int r; INIT_LIST_HEAD(&p->validated); p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle); - if (p->bo_list) { - need_mmap_lock = p->bo_list->first_userptr != - p->bo_list->num_entries; + if (p->bo_list) amdgpu_bo_list_get_list(p->bo_list, &p->validated); - } INIT_LIST_HEAD(&duplicates); amdgpu_vm_get_pd_bo(&fpriv->vm, &p->validated, &p->vm_pd); @@ -519,9 +515,6 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (p->uf_entry.robj) list_add(&p->uf_entry.tv.head, &p->validated); - if (need_mmap_lock) - down_read(¤t->mm->mmap_sem); - while (1) { struct list_head need_pages; unsigned i; @@ -674,9 +667,6 @@ error_validate: error_free_pages: - if (need_mmap_lock) - up_read(¤t->mm->mmap_sem); - if (p->bo_list) { for (i = p->bo_list->first_userptr; i < p->bo_list->num_entries; ++i) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index f1e61b3df640..b0d45c8e6bb3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -318,8 +318,6 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, } if (args->flags & AMDGPU_GEM_USERPTR_VALIDATE) { - down_read(¤t->mm->mmap_sem); - r = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, bo->tbo.ttm->pages); if (r) @@ -334,8 +332,6 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, amdgpu_bo_unreserve(bo); if (r) goto free_pages; - - up_read(¤t->mm->mmap_sem); } r = drm_gem_handle_create(filp, gobj, &handle); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 28e121984332..ea0378c8b049 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -622,6 +622,8 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) flags |= FOLL_WRITE; + down_read(¤t->mm->mmap_sem); + if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) { /* check that we only use anonymous memory to prevent problems with writeback */ @@ -629,8 +631,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) struct vm_area_struct *vma; vma = find_vma(gtt->usermm, gtt->userptr); - if (!vma || vma->vm_file || vma->vm_end < end) + if (!vma || vma->vm_file || vma->vm_end < end) { + up_read(¤t->mm->mmap_sem); return -EPERM; + } } do { @@ -657,10 +661,12 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) } while (pinned < ttm->num_pages); + up_read(¤t->mm->mmap_sem); return 0; release_pages: release_pages(pages, pinned, 0); + up_read(¤t->mm->mmap_sem); return r; } -- cgit From ca666a3c298f838346ccea46ff542c605e68deb5 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 5 Sep 2017 14:30:05 +0200 Subject: drm/amdgpu: stop using BO status for user pages MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead use a counter to figure out if we need to set new pages or not. Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 11 +++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 14 ++++++++++++++ 3 files changed, 22 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index cc6de0b46326..f3e561136597 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1802,6 +1802,7 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, unsigned long end); bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, int *last_invalidated); +bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm); bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, struct ttm_mem_reg *mem); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 283a216ee758..4d3f8fbfa59d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -473,7 +473,8 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, return -EPERM; /* Check if we have user pages and nobody bound the BO already */ - if (lobj->user_pages && bo->tbo.ttm->state != tt_bound) { + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) && + lobj->user_pages) { amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, lobj->user_pages); binding_userptr = true; @@ -534,23 +535,25 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, INIT_LIST_HEAD(&need_pages); for (i = p->bo_list->first_userptr; i < p->bo_list->num_entries; ++i) { + struct amdgpu_bo *bo; e = &p->bo_list->array[i]; + bo = e->robj; - if (amdgpu_ttm_tt_userptr_invalidated(e->robj->tbo.ttm, + if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm, &e->user_invalidated) && e->user_pages) { /* We acquired a page array, but somebody * invalidated it. Free it and try again */ release_pages(e->user_pages, - e->robj->tbo.ttm->num_pages, + bo->tbo.ttm->num_pages, false); kvfree(e->user_pages); e->user_pages = NULL; } - if (e->robj->tbo.ttm->state != tt_bound && + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) && !e->user_pages) { list_del(&e->tv.head); list_add(&e->tv.head, &need_pages); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index ea0378c8b049..e67785191032 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -609,6 +609,7 @@ struct amdgpu_ttm_tt { spinlock_t guptasklock; struct list_head guptasks; atomic_t mmu_invalidations; + uint32_t last_set_pages; struct list_head list; }; @@ -672,8 +673,10 @@ release_pages: void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages) { + struct amdgpu_ttm_tt *gtt = (void *)ttm; unsigned i; + gtt->last_set_pages = atomic_read(>t->mmu_invalidations); for (i = 0; i < ttm->num_pages; ++i) { if (ttm->pages[i]) put_page(ttm->pages[i]); @@ -1025,6 +1028,7 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, spin_lock_init(>t->guptasklock); INIT_LIST_HEAD(>t->guptasks); atomic_set(>t->mmu_invalidations, 0); + gtt->last_set_pages = 0; return 0; } @@ -1077,6 +1081,16 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, return prev_invalidated != *last_invalidated; } +bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm) +{ + struct amdgpu_ttm_tt *gtt = (void *)ttm; + + if (gtt == NULL || !gtt->userptr) + return false; + + return atomic_read(>t->mmu_invalidations) != gtt->last_set_pages; +} + bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; -- cgit From 1b0c0f9dc5ca6c0c8be21eeac92c7aa77bbf1d33 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 5 Sep 2017 14:36:44 +0200 Subject: drm/amdgpu: move userptr BOs to CPU domain during CS v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead of moving them in the MMU notifier move them during CS. v2: still mark pages as accessed/dirty Signed-off-by: Christian König Reviewed-by: Felix Kuehling (v1) Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 6 ++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c | 5 +---- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 27 +++++++++++++++++++-------- 4 files changed, 27 insertions(+), 12 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index f3e561136597..134a049f87bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1794,6 +1794,7 @@ void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain); bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo); int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages); +void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm); int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, uint32_t flags); bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 4d3f8fbfa59d..4877df83b801 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -475,6 +475,12 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, /* Check if we have user pages and nobody bound the BO already */ if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) && lobj->user_pages) { + amdgpu_ttm_placement_from_domain(bo, + AMDGPU_GEM_DOMAIN_CPU); + r = ttm_bo_validate(&bo->tbo, &bo->placement, true, + false); + if (r) + return r; amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, lobj->user_pages); binding_userptr = true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 6558a3ed57a7..df85a1314799 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -137,10 +137,7 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, if (r <= 0) DRM_ERROR("(%ld) failed to wait for user bo\n", r); - amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (r) - DRM_ERROR("(%ld) failed to validate user bo\n", r); + amdgpu_ttm_tt_mark_user_pages(bo->tbo.ttm); amdgpu_bo_unreserve(bo); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index e67785191032..fe887f361be8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -685,6 +685,24 @@ void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages) } } +void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) +{ + struct amdgpu_ttm_tt *gtt = (void *)ttm; + unsigned i; + + for (i = 0; i < ttm->num_pages; ++i) { + struct page *page = ttm->pages[i]; + + if (!page) + continue; + + if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) + set_page_dirty(page); + + mark_page_accessed(page); + } +} + static void amdgpu_trace_dma_map(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); @@ -740,7 +758,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - struct sg_page_iter sg_iter; int write = !(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); enum dma_data_direction direction = write ? @@ -753,13 +770,7 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) /* free the sg table and pages again */ dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); - for_each_sg_page(ttm->sg->sgl, &sg_iter, ttm->sg->nents, 0) { - struct page *page = sg_page_iter_page(&sg_iter); - if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) - set_page_dirty(page); - - mark_page_accessed(page); - } + amdgpu_ttm_tt_mark_user_pages(ttm); amdgpu_trace_dma_unmap(ttm); -- cgit From 60de1c1740f390fe48141b54d04cc53a6073d347 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 5 Sep 2017 14:50:24 +0200 Subject: drm/amdgpu: use a rw_semaphore for MMU notifiers MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Allow at least some parallel processing. Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index df85a1314799..6d216abd0e1d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -50,7 +50,7 @@ struct amdgpu_mn { struct hlist_node node; /* objects protected by lock */ - struct mutex lock; + struct rw_semaphore lock; struct rb_root objects; }; @@ -74,7 +74,7 @@ static void amdgpu_mn_destroy(struct work_struct *work) struct amdgpu_bo *bo, *next_bo; mutex_lock(&adev->mn_lock); - mutex_lock(&rmn->lock); + down_write(&rmn->lock); hash_del(&rmn->node); rbtree_postorder_for_each_entry_safe(node, next_node, &rmn->objects, it.rb) { @@ -84,7 +84,7 @@ static void amdgpu_mn_destroy(struct work_struct *work) } kfree(node); } - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); mutex_unlock(&adev->mn_lock); mmu_notifier_unregister_no_release(&rmn->mn, rmn->mm); kfree(rmn); @@ -160,7 +160,7 @@ static void amdgpu_mn_invalidate_page(struct mmu_notifier *mn, struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; - mutex_lock(&rmn->lock); + down_read(&rmn->lock); it = interval_tree_iter_first(&rmn->objects, address, address); if (it) { @@ -170,7 +170,7 @@ static void amdgpu_mn_invalidate_page(struct mmu_notifier *mn, amdgpu_mn_invalidate_node(node, address, address); } - mutex_unlock(&rmn->lock); + up_read(&rmn->lock); } /** @@ -195,7 +195,7 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, /* notification is exclusive, but interval is inclusive */ end -= 1; - mutex_lock(&rmn->lock); + down_read(&rmn->lock); it = interval_tree_iter_first(&rmn->objects, start, end); while (it) { @@ -207,7 +207,7 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, amdgpu_mn_invalidate_node(node, start, end); } - mutex_unlock(&rmn->lock); + up_read(&rmn->lock); } static const struct mmu_notifier_ops amdgpu_mn_ops = { @@ -248,7 +248,7 @@ static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) rmn->adev = adev; rmn->mm = mm; rmn->mn.ops = &amdgpu_mn_ops; - mutex_init(&rmn->lock); + init_rwsem(&rmn->lock); rmn->objects = RB_ROOT; r = __mmu_notifier_register(&rmn->mn, mm); @@ -295,7 +295,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) INIT_LIST_HEAD(&bos); - mutex_lock(&rmn->lock); + down_write(&rmn->lock); while ((it = interval_tree_iter_first(&rmn->objects, addr, end))) { kfree(node); @@ -309,7 +309,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) if (!node) { node = kmalloc(sizeof(struct amdgpu_mn_node), GFP_KERNEL); if (!node) { - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); return -ENOMEM; } } @@ -324,7 +324,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) interval_tree_insert(&node->it, &rmn->objects); - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); return 0; } @@ -350,7 +350,7 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) return; } - mutex_lock(&rmn->lock); + down_write(&rmn->lock); /* save the next list entry for later */ head = bo->mn_list.next; @@ -365,6 +365,6 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) kfree(node); } - mutex_unlock(&rmn->lock); + up_write(&rmn->lock); mutex_unlock(&adev->mn_lock); } -- cgit From 3fe89771cb0a65d3b686bcafb5b7e3ebae0ea604 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 12 Sep 2017 14:25:14 -0400 Subject: drm/amdgpu: stop reserving the BO in the MMU callback v3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Instead take the callback lock during the final parts of CS. This should solve the last remaining locking order problems with BO reservations. v2: rebase, make dummy functions static inline v3: add one more missing inline and comments Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 11 +++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 30 ++++++++++++++++++++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c | 30 +++++++++++++++++++++--------- 3 files changed, 56 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 134a049f87bd..740683474a1c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -178,6 +178,7 @@ struct amdgpu_cs_parser; struct amdgpu_job; struct amdgpu_irq_src; struct amdgpu_fpriv; +struct amdgpu_mn; enum amdgpu_cp_irq { AMDGPU_CP_IRQ_GFX_EOP = 0, @@ -1057,6 +1058,7 @@ struct amdgpu_cs_parser { /* buffer objects */ struct ww_acquire_ctx ticket; struct amdgpu_bo_list *bo_list; + struct amdgpu_mn *mn; struct amdgpu_bo_list_entry vm_pd; struct list_head validated; struct dma_fence *fence; @@ -1201,9 +1203,18 @@ void amdgpu_test_moves(struct amdgpu_device *adev); * MMU Notifier */ #if defined(CONFIG_MMU_NOTIFIER) +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev); int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); void amdgpu_mn_unregister(struct amdgpu_bo *bo); +void amdgpu_mn_lock(struct amdgpu_mn *mn); +void amdgpu_mn_unlock(struct amdgpu_mn *mn); #else +static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {} +static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {} +static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +{ + return NULL; +} static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) { return -ENODEV; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 4877df83b801..c2310d4eebc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -513,8 +513,11 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, INIT_LIST_HEAD(&p->validated); p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle); - if (p->bo_list) + if (p->bo_list) { amdgpu_bo_list_get_list(p->bo_list, &p->validated); + if (p->bo_list->first_userptr != p->bo_list->num_entries) + p->mn = amdgpu_mn_get(p->adev); + } INIT_LIST_HEAD(&duplicates); amdgpu_vm_get_pd_bo(&fpriv->vm, &p->validated, &p->vm_pd); @@ -722,11 +725,7 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, { unsigned i; - if (!error) - ttm_eu_fence_buffer_objects(&parser->ticket, - &parser->validated, - parser->fence); - else if (backoff) + if (error && backoff) ttm_eu_backoff_reservation(&parser->ticket, &parser->validated); @@ -1127,14 +1126,29 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, struct amdgpu_ring *ring = p->job->ring; struct amd_sched_entity *entity = &p->ctx->rings[ring->idx].entity; struct amdgpu_job *job; + unsigned i; int r; + amdgpu_mn_lock(p->mn); + if (p->bo_list) { + for (i = p->bo_list->first_userptr; + i < p->bo_list->num_entries; ++i) { + struct amdgpu_bo *bo = p->bo_list->array[i].robj; + + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) { + amdgpu_mn_unlock(p->mn); + return -ERESTARTSYS; + } + } + } + job = p->job; p->job = NULL; r = amd_sched_job_init(&job->base, &ring->sched, entity, p->filp); if (r) { amdgpu_job_free(job); + amdgpu_mn_unlock(p->mn); return r; } @@ -1150,6 +1164,10 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, trace_amdgpu_cs_ioctl(job); amd_sched_entity_push_job(&job->base); + + ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence); + amdgpu_mn_unlock(p->mn); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 6d216abd0e1d..99edb40b5f99 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -106,6 +106,25 @@ static void amdgpu_mn_release(struct mmu_notifier *mn, schedule_work(&rmn->work); } + +/** + * amdgpu_mn_lock - take the write side lock for this mn + */ +void amdgpu_mn_lock(struct amdgpu_mn *mn) +{ + if (mn) + down_write(&mn->lock); +} + +/** + * amdgpu_mn_unlock - drop the write side lock for this mn + */ +void amdgpu_mn_unlock(struct amdgpu_mn *mn) +{ + if (mn) + up_write(&mn->lock); +} + /** * amdgpu_mn_invalidate_node - unmap all BOs of a node * @@ -126,20 +145,12 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, if (!amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm, start, end)) continue; - r = amdgpu_bo_reserve(bo, true); - if (r) { - DRM_ERROR("(%ld) failed to reserve user bo\n", r); - continue; - } - r = reservation_object_wait_timeout_rcu(bo->tbo.resv, true, false, MAX_SCHEDULE_TIMEOUT); if (r <= 0) DRM_ERROR("(%ld) failed to wait for user bo\n", r); amdgpu_ttm_tt_mark_user_pages(bo->tbo.ttm); - - amdgpu_bo_unreserve(bo); } } @@ -223,7 +234,7 @@ static const struct mmu_notifier_ops amdgpu_mn_ops = { * * Creates a notifier context for current->mm. */ -static struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) { struct mm_struct *mm = current->mm; struct amdgpu_mn *rmn; @@ -368,3 +379,4 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) up_write(&rmn->lock); mutex_unlock(&adev->mn_lock); } + -- cgit From 9cca0b8e5df0ac438c65eec5044bfa089d16fbbe Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 6 Sep 2017 16:15:28 +0200 Subject: drm/amdgpu: move amdgpu_cs_sysvm_access_required into find_mapping MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When we need to find the mapping we need sysvm access anyway. Signed-off-by: Christian König Acked-by: Leo Liu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 8 ++--- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 64 ++++++++++++--------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 16 ++++----- drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 11 +++--- 4 files changed, 36 insertions(+), 63 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 740683474a1c..5ee6cea8caaf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -179,6 +179,7 @@ struct amdgpu_job; struct amdgpu_irq_src; struct amdgpu_fpriv; struct amdgpu_mn; +struct amdgpu_bo_va_mapping; enum amdgpu_cp_irq { AMDGPU_CP_IRQ_GFX_EOP = 0, @@ -1900,10 +1901,9 @@ static inline int amdgpu_acpi_init(struct amdgpu_device *adev) { return 0; } static inline void amdgpu_acpi_fini(struct amdgpu_device *adev) { } #endif -struct amdgpu_bo_va_mapping * -amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, - uint64_t addr, struct amdgpu_bo **bo); -int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser); +int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, + uint64_t addr, struct amdgpu_bo **bo, + struct amdgpu_bo_va_mapping **mapping); #include "amdgpu_object.h" #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index c2310d4eebc8..c30110a3024a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -921,11 +921,11 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, uint64_t offset; uint8_t *kptr; - m = amdgpu_cs_find_mapping(parser, chunk_ib->va_start, - &aobj); - if (!aobj) { + r = amdgpu_cs_find_mapping(parser, chunk_ib->va_start, + &aobj, &m); + if (r) { DRM_ERROR("IB va_start is invalid\n"); - return -EINVAL; + return r; } if ((chunk_ib->va_start + chunk_ib->ib_bytes) > @@ -1475,15 +1475,16 @@ err_free_fences: * virtual memory address. Returns allocation structure when found, NULL * otherwise. */ -struct amdgpu_bo_va_mapping * -amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, - uint64_t addr, struct amdgpu_bo **bo) +int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, + uint64_t addr, struct amdgpu_bo **bo, + struct amdgpu_bo_va_mapping **map) { struct amdgpu_bo_va_mapping *mapping; unsigned i; + int r; if (!parser->bo_list) - return NULL; + return 0; addr /= AMDGPU_GPU_PAGE_SIZE; @@ -1500,7 +1501,8 @@ amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, continue; *bo = lobj->bo_va->base.bo; - return mapping; + *map = mapping; + goto found; } list_for_each_entry(mapping, &lobj->bo_va->invalids, list) { @@ -1509,44 +1511,22 @@ amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, continue; *bo = lobj->bo_va->base.bo; - return mapping; + *map = mapping; + goto found; } } - return NULL; -} + return -EINVAL; -/** - * amdgpu_cs_sysvm_access_required - make BOs accessible by the system VM - * - * @parser: command submission parser context - * - * Helper for UVD/VCE VM emulation, make sure BOs are accessible by the system VM. - */ -int amdgpu_cs_sysvm_access_required(struct amdgpu_cs_parser *parser) -{ - unsigned i; - int r; +found: + r = amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem); + if (unlikely(r)) + return r; - if (!parser->bo_list) + if ((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) return 0; - for (i = 0; i < parser->bo_list->num_entries; i++) { - struct amdgpu_bo *bo = parser->bo_list->array[i].robj; - - r = amdgpu_ttm_bind(&bo->tbo, &bo->tbo.mem); - if (unlikely(r)) - return r; - - if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) - continue; - - bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; - amdgpu_ttm_placement_from_domain(bo, bo->allowed_domains); - r = ttm_bo_validate(&bo->tbo, &bo->placement, false, false); - if (unlikely(r)) - return r; - } - - return 0; + (*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains); + return ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, false, false); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index e19928dae8e3..331e34ac61fd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -410,10 +410,10 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx) uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx); int r = 0; - mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo); - if (mapping == NULL) { + r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping); + if (r) { DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr); - return -EINVAL; + return r; } if (!ctx->parser->adev->uvd.address_64_bit) { @@ -737,10 +737,10 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx) uint64_t addr = amdgpu_uvd_get_addr_from_ctx(ctx); int r; - mapping = amdgpu_cs_find_mapping(ctx->parser, addr, &bo); - if (mapping == NULL) { + r = amdgpu_cs_find_mapping(ctx->parser, addr, &bo, &mapping); + if (r) { DRM_ERROR("Can't find BO for addr 0x%08Lx\n", addr); - return -EINVAL; + return r; } start = amdgpu_bo_gpu_offset(bo); @@ -917,10 +917,6 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx) return -EINVAL; } - r = amdgpu_cs_sysvm_access_required(parser); - if (r) - return r; - ctx.parser = parser; ctx.buf_sizes = buf_sizes; ctx.ib_idx = ib_idx; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index c855366521ab..b46280c1279f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -559,6 +559,7 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, struct amdgpu_bo_va_mapping *mapping; struct amdgpu_bo *bo; uint64_t addr; + int r; if (index == 0xffffffff) index = 0; @@ -567,11 +568,11 @@ static int amdgpu_vce_cs_reloc(struct amdgpu_cs_parser *p, uint32_t ib_idx, ((uint64_t)amdgpu_get_ib_value(p, ib_idx, hi)) << 32; addr += ((uint64_t)size) * ((uint64_t)index); - mapping = amdgpu_cs_find_mapping(p, addr, &bo); - if (mapping == NULL) { + r = amdgpu_cs_find_mapping(p, addr, &bo, &mapping); + if (r) { DRM_ERROR("Can't find BO for addr 0x%010Lx %d %d %d %d\n", addr, lo, hi, size, index); - return -EINVAL; + return r; } if ((addr + (uint64_t)size) > @@ -652,10 +653,6 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) p->job->vm = NULL; ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); - r = amdgpu_cs_sysvm_access_required(p); - if (r) - return r; - while (idx < ib->length_dw) { uint32_t len = amdgpu_get_ib_value(p, ib_idx, idx); uint32_t cmd = amdgpu_get_ib_value(p, ib_idx, idx + 1); -- cgit From aebc5e6f50f770ec9392c3ca804f18b30797dfa7 Mon Sep 17 00:00:00 2001 From: Christian König Date: Wed, 6 Sep 2017 16:55:16 +0200 Subject: drm/amdgpu: rework amdgpu_cs_find_mapping MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use the VM instead of the BO list to find the BO for a virtual address. This fixes UVD/VCE in physical mode with VM local BOs. Signed-off-by: Christian König Acked-by: Leo Liu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 42 +++++++----------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 4 files changed, 30 insertions(+), 32 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index c30110a3024a..5f19227b35e9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1479,46 +1479,24 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, uint64_t addr, struct amdgpu_bo **bo, struct amdgpu_bo_va_mapping **map) { + struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_va_mapping *mapping; - unsigned i; int r; - if (!parser->bo_list) - return 0; - addr /= AMDGPU_GPU_PAGE_SIZE; - for (i = 0; i < parser->bo_list->num_entries; i++) { - struct amdgpu_bo_list_entry *lobj; - - lobj = &parser->bo_list->array[i]; - if (!lobj->bo_va) - continue; - - list_for_each_entry(mapping, &lobj->bo_va->valids, list) { - if (mapping->start > addr || - addr > mapping->last) - continue; - - *bo = lobj->bo_va->base.bo; - *map = mapping; - goto found; - } - - list_for_each_entry(mapping, &lobj->bo_va->invalids, list) { - if (mapping->start > addr || - addr > mapping->last) - continue; + mapping = amdgpu_vm_bo_lookup_mapping(vm, addr); + if (!mapping || !mapping->bo_va || !mapping->bo_va->base.bo) + return -EINVAL; - *bo = lobj->bo_va->base.bo; - *map = mapping; - goto found; - } - } + *bo = mapping->bo_va->base.bo; + *map = mapping; - return -EINVAL; + /* Double check that the BO is reserved by this CS */ + if (READ_ONCE((*bo)->tbo.resv->lock.ctx) != &parser->ticket) + return -EINVAL; -found: r = amdgpu_ttm_bind(&(*bo)->tbo, &(*bo)->tbo.mem); if (unlikely(r)) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 42492e63b3a2..a4891bea2ca8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -35,6 +35,7 @@ /* bo virtual addresses in a vm */ struct amdgpu_bo_va_mapping { + struct amdgpu_bo_va *bo_va; struct list_head list; struct rb_node rb; uint64_t start; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 545531db66db..758bbb9e77f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2086,6 +2086,7 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, struct amdgpu_vm *vm = bo_va->base.vm; struct amdgpu_bo *bo = bo_va->base.bo; + mapping->bo_va = bo_va; list_add(&mapping->list, &bo_va->invalids); amdgpu_vm_it_insert(mapping, &vm->va); @@ -2263,6 +2264,7 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev, list_del(&mapping->list); amdgpu_vm_it_remove(mapping, &vm->va); + mapping->bo_va = NULL; trace_amdgpu_vm_bo_unmap(bo_va, mapping); if (valid) @@ -2348,6 +2350,7 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, if (tmp->last > eaddr) tmp->last = eaddr; + tmp->bo_va = NULL; list_add(&tmp->list, &vm->freed); trace_amdgpu_vm_bo_unmap(NULL, tmp); } @@ -2373,6 +2376,19 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, return 0; } +/** + * amdgpu_vm_bo_lookup_mapping - find mapping by address + * + * @vm: the requested VM + * + * Find a mapping by it's address. + */ +struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, + uint64_t addr) +{ + return amdgpu_vm_it_iter_first(&vm->va, addr, addr); +} + /** * amdgpu_vm_bo_rmv - remove a bo to a specific vm * @@ -2398,6 +2414,7 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, list_for_each_entry_safe(mapping, next, &bo_va->valids, list) { list_del(&mapping->list); amdgpu_vm_it_remove(mapping, &vm->va); + mapping->bo_va = NULL; trace_amdgpu_vm_bo_unmap(bo_va, mapping); list_add(&mapping->list, &vm->freed); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 90b7741d024b..c1accd15efc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -276,6 +276,8 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev, int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, struct amdgpu_vm *vm, uint64_t saddr, uint64_t size); +struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, + uint64_t addr); void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va); void amdgpu_vm_set_fragment_size(struct amdgpu_device *adev, -- cgit From 1ed3d2567c800eca053ef86fdd3fc27b72d0192e Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 5 Sep 2017 17:30:46 +0200 Subject: drm/amdgpu: keep the MMU lock until the update ends v4 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is quite controversial because it adds another lock which is held during page table updates, but I don't see much other option. v2: allow multiple updates to be in flight at the same time v3: simplify the patch, take the read side only once v4: correctly fix rebase conflict Signed-off-by: Christian König Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 +-- drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c | 59 +++++++++++++++++++++++++++++++--- 2 files changed, 57 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 5ee6cea8caaf..0758d9176ea1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1204,11 +1204,11 @@ void amdgpu_test_moves(struct amdgpu_device *adev); * MMU Notifier */ #if defined(CONFIG_MMU_NOTIFIER) +void amdgpu_mn_lock(struct amdgpu_mn *mn); +void amdgpu_mn_unlock(struct amdgpu_mn *mn); struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev); int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); void amdgpu_mn_unregister(struct amdgpu_bo *bo); -void amdgpu_mn_lock(struct amdgpu_mn *mn); -void amdgpu_mn_unlock(struct amdgpu_mn *mn); #else static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {} static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 99edb40b5f99..521a51b37f5d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -52,6 +52,8 @@ struct amdgpu_mn { /* objects protected by lock */ struct rw_semaphore lock; struct rb_root objects; + struct mutex read_lock; + atomic_t recursion; }; struct amdgpu_mn_node { @@ -125,6 +127,34 @@ void amdgpu_mn_unlock(struct amdgpu_mn *mn) up_write(&mn->lock); } +/** + * amdgpu_mn_read_lock - take the rmn read lock + * + * @rmn: our notifier + * + * Take the rmn read side lock. + */ +static void amdgpu_mn_read_lock(struct amdgpu_mn *rmn) +{ + mutex_lock(&rmn->read_lock); + if (atomic_inc_return(&rmn->recursion) == 1) + down_read_non_owner(&rmn->lock); + mutex_unlock(&rmn->read_lock); +} + +/** + * amdgpu_mn_read_unlock - drop the rmn read lock + * + * @rmn: our notifier + * + * Drop the rmn read side lock. + */ +static void amdgpu_mn_read_unlock(struct amdgpu_mn *rmn) +{ + if (atomic_dec_return(&rmn->recursion) == 0) + up_read_non_owner(&rmn->lock); +} + /** * amdgpu_mn_invalidate_node - unmap all BOs of a node * @@ -171,7 +201,7 @@ static void amdgpu_mn_invalidate_page(struct mmu_notifier *mn, struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; - down_read(&rmn->lock); + amdgpu_mn_read_lock(rmn); it = interval_tree_iter_first(&rmn->objects, address, address); if (it) { @@ -181,7 +211,7 @@ static void amdgpu_mn_invalidate_page(struct mmu_notifier *mn, amdgpu_mn_invalidate_node(node, address, address); } - up_read(&rmn->lock); + amdgpu_mn_read_unlock(rmn); } /** @@ -206,7 +236,7 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, /* notification is exclusive, but interval is inclusive */ end -= 1; - down_read(&rmn->lock); + amdgpu_mn_read_lock(rmn); it = interval_tree_iter_first(&rmn->objects, start, end); while (it) { @@ -217,14 +247,33 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, amdgpu_mn_invalidate_node(node, start, end); } +} + +/** + * amdgpu_mn_invalidate_range_end - callback to notify about mm change + * + * @mn: our notifier + * @mn: the mm this callback is about + * @start: start of updated range + * @end: end of updated range + * + * Release the lock again to allow new command submissions. + */ +static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, + unsigned long end) +{ + struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); - up_read(&rmn->lock); + amdgpu_mn_read_unlock(rmn); } static const struct mmu_notifier_ops amdgpu_mn_ops = { .release = amdgpu_mn_release, .invalidate_page = amdgpu_mn_invalidate_page, .invalidate_range_start = amdgpu_mn_invalidate_range_start, + .invalidate_range_end = amdgpu_mn_invalidate_range_end, }; /** @@ -261,6 +310,8 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) rmn->mn.ops = &amdgpu_mn_ops; init_rwsem(&rmn->lock); rmn->objects = RB_ROOT; + mutex_init(&rmn->read_lock); + atomic_set(&rmn->recursion, 0); r = __mmu_notifier_register(&rmn->mn, mm); if (r) -- cgit From 711becf0e6c5bf7cabd8f1ca528e9bbe085fda9a Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 8 Sep 2017 17:19:19 +0200 Subject: drm/amdgpu: move amdgpu_ttm_tt_* declarations into amdgpu_ttm.h MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just some cleanup. Signed-off-by: Christian König Reviewed-by: Michel Dänzer Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 15 --------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h | 16 ++++++++++++++++ 2 files changed, 16 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 0758d9176ea1..814df1a212c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1804,21 +1804,6 @@ void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes, u64 num_vis_bytes); void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain); bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo); -int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); -void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages); -void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm); -int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, - uint32_t flags); -bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); -struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm); -bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, - unsigned long end); -bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, - int *last_invalidated); -bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm); -bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); -uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, - struct ttm_mem_reg *mem); void amdgpu_vram_location(struct amdgpu_device *adev, struct amdgpu_mc *mc, u64 base); void amdgpu_gart_location(struct amdgpu_device *adev, struct amdgpu_mc *mc); void amdgpu_ttm_set_active_vram_size(struct amdgpu_device *adev, u64 size); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 43093bffa2cf..64709e041d5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -82,4 +82,20 @@ bool amdgpu_ttm_is_bound(struct ttm_tt *ttm); int amdgpu_ttm_bind(struct ttm_buffer_object *bo, struct ttm_mem_reg *bo_mem); int amdgpu_ttm_recover_gart(struct amdgpu_device *adev); +int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages); +void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages); +void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm); +int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, + uint32_t flags); +bool amdgpu_ttm_tt_has_userptr(struct ttm_tt *ttm); +struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm); +bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, + unsigned long end); +bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, + int *last_invalidated); +bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm); +bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); +uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, + struct ttm_mem_reg *mem); + #endif -- cgit From 9a18999640fa6aed0578c59c328dca9ca01a2d9a Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 12 Sep 2017 14:29:07 -0400 Subject: drm/amdgpu: move MMU notifier related defines to amdgpu_mn.h MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just some cleanup. Signed-off-by: Christian König Reviewed-by: Michel Dänzer Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 25 +--------------- drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h | 52 ++++++++++++++++++++++++++++++++++ 2 files changed, 53 insertions(+), 24 deletions(-) create mode 100644 drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 814df1a212c4..802fdc11944b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -65,6 +65,7 @@ #include "amdgpu_uvd.h" #include "amdgpu_vce.h" #include "amdgpu_vcn.h" +#include "amdgpu_mn.h" #include "gpu_scheduler.h" #include "amdgpu_virt.h" @@ -178,7 +179,6 @@ struct amdgpu_cs_parser; struct amdgpu_job; struct amdgpu_irq_src; struct amdgpu_fpriv; -struct amdgpu_mn; struct amdgpu_bo_va_mapping; enum amdgpu_cp_irq { @@ -1200,29 +1200,6 @@ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number); */ void amdgpu_test_moves(struct amdgpu_device *adev); -/* - * MMU Notifier - */ -#if defined(CONFIG_MMU_NOTIFIER) -void amdgpu_mn_lock(struct amdgpu_mn *mn); -void amdgpu_mn_unlock(struct amdgpu_mn *mn); -struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev); -int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); -void amdgpu_mn_unregister(struct amdgpu_bo *bo); -#else -static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {} -static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {} -static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) -{ - return NULL; -} -static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) -{ - return -ENODEV; -} -static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {} -#endif - /* * Debugfs */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h new file mode 100644 index 000000000000..d0095a3793b8 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h @@ -0,0 +1,52 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: Christian König + */ +#ifndef __AMDGPU_MN_H__ +#define __AMDGPU_MN_H__ + +/* + * MMU Notifier + */ +struct amdgpu_mn; + +#if defined(CONFIG_MMU_NOTIFIER) +void amdgpu_mn_lock(struct amdgpu_mn *mn); +void amdgpu_mn_unlock(struct amdgpu_mn *mn); +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev); +int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); +void amdgpu_mn_unregister(struct amdgpu_bo *bo); +#else +static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {} +static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {} +static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +{ + return NULL; +} +static inline int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) +{ + return -ENODEV; +} +static inline void amdgpu_mn_unregister(struct amdgpu_bo *bo) {} +#endif + +#endif -- cgit From f1ac0fc2f2355495628a45e90cbd88e3d2c40ef9 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Mon, 4 Sep 2017 17:42:28 +0800 Subject: drm/amdgpu: fixed raven psp cmd prepare and submit MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - fw_size in psp_v10_0_prep_cmd_buf is wrongly set as 0 - fixed the wrong calculation of psp_write_ptr_reg in psp_v10_0_cmd_submit Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Acked-by: Christian König Reviewed-by: Huang Rui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/psp_v10_0.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index f7cf994b1da2..b77feef829e7 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -136,15 +136,13 @@ int psp_v10_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, struct psp_gfx_cm { int ret; uint64_t fw_mem_mc_addr = ucode->mc_addr; - struct common_firmware_header *header; memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); - header = (struct common_firmware_header *)ucode->fw; cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); - cmd->cmd.cmd_load_ip_fw.fw_size = le32_to_cpu(header->ucode_size_bytes); + cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; ret = psp_v10_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); if (ret) @@ -245,15 +243,20 @@ int psp_v10_0_cmd_submit(struct psp_context *psp, struct psp_gfx_rb_frame * write_frame = psp->km_ring.ring_mem; struct psp_ring *ring = &psp->km_ring; struct amdgpu_device *adev = psp->adev; + uint32_t ring_size_dw = ring->ring_size / 4; + uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4; /* KM (GPCOM) prepare write pointer */ psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); /* Update KM RB frame pointer to new frame */ - if ((psp_write_ptr_reg % ring->ring_size) == 0) + if ((psp_write_ptr_reg % ring_size_dw) == 0) write_frame = ring->ring_mem; else - write_frame = ring->ring_mem + (psp_write_ptr_reg / (sizeof(struct psp_gfx_rb_frame) / 4)); + write_frame = ring->ring_mem + (psp_write_ptr_reg / rb_frame_size_dw); + + /* Initialize KM RB frame */ + memset(write_frame, 0, sizeof(struct psp_gfx_rb_frame)); /* Update KM RB frame */ write_frame->cmd_buf_addr_hi = upper_32_bits(cmd_buf_mc_addr); @@ -263,8 +266,7 @@ int psp_v10_0_cmd_submit(struct psp_context *psp, write_frame->fence_value = index; /* Update the write Pointer in DWORDs */ - psp_write_ptr_reg += sizeof(struct psp_gfx_rb_frame) / 4; - psp_write_ptr_reg = (psp_write_ptr_reg >= ring->ring_size) ? 0 : psp_write_ptr_reg; + psp_write_ptr_reg = (psp_write_ptr_reg + rb_frame_size_dw) % ring_size_dw; WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); return 0; -- cgit From 4ef72453311a697b3fb90da9c86c83012911ccf9 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Fri, 8 Sep 2017 13:04:52 +0800 Subject: drm/amdgpu: added api for stopping psp ring (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - v2: reuse the ring stop api in ring destory Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Acked-by: Christian König Reviewed-by: Huang Rui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h | 2 ++ drivers/gpu/drm/amd/amdgpu/psp_v10_0.c | 15 ++++++++++++++- drivers/gpu/drm/amd/amdgpu/psp_v10_0.h | 2 ++ drivers/gpu/drm/amd/amdgpu/psp_v3_1.c | 15 ++++++++++++++- drivers/gpu/drm/amd/amdgpu/psp_v3_1.h | 2 ++ 6 files changed, 36 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 8c2204c7b384..abd20819ba6b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -57,6 +57,7 @@ static int psp_sw_init(void *handle) psp->prep_cmd_buf = psp_v3_1_prep_cmd_buf; psp->ring_init = psp_v3_1_ring_init; psp->ring_create = psp_v3_1_ring_create; + psp->ring_stop = psp_v3_1_ring_stop; psp->ring_destroy = psp_v3_1_ring_destroy; psp->cmd_submit = psp_v3_1_cmd_submit; psp->compare_sram_data = psp_v3_1_compare_sram_data; @@ -69,6 +70,7 @@ static int psp_sw_init(void *handle) psp->prep_cmd_buf = psp_v10_0_prep_cmd_buf; psp->ring_init = psp_v10_0_ring_init; psp->ring_create = psp_v10_0_ring_create; + psp->ring_stop = psp_v10_0_ring_stop; psp->ring_destroy = psp_v10_0_ring_destroy; psp->cmd_submit = psp_v10_0_cmd_submit; psp->compare_sram_data = psp_v10_0_compare_sram_data; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 538fa9dbfb21..e79795b59797 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -66,6 +66,8 @@ struct psp_context struct psp_gfx_cmd_resp *cmd); int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type); + int (*ring_stop)(struct psp_context *psp, + enum psp_ring_type ring_type); int (*ring_destroy)(struct psp_context *psp, enum psp_ring_type ring_type); int (*cmd_submit)(struct psp_context *psp, struct amdgpu_firmware_info *ucode, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index b77feef829e7..6ec5c9f8074d 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -207,7 +207,7 @@ int psp_v10_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) return ret; } -int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +int psp_v10_0_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { int ret = 0; struct psp_ring *ring; @@ -227,6 +227,19 @@ int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), 0x80000000, 0x80000000, false); + return ret; +} + +int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + ret = psp_v10_0_ring_stop(psp, ring_type); + if (ret) + DRM_ERROR("Fail to stop psp ring\n"); + amdgpu_bo_free_kernel(&adev->firmware.rbuf, &ring->ring_mem_mc_addr, (void **)&ring->ring_mem); diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h index e76cde2f01f9..3af3ad1320ff 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h @@ -34,6 +34,8 @@ extern int psp_v10_0_ring_init(struct psp_context *psp, enum psp_ring_type ring_type); extern int psp_v10_0_ring_create(struct psp_context *psp, enum psp_ring_type ring_type); +extern int psp_v10_0_ring_stop(struct psp_context *psp, + enum psp_ring_type ring_type); extern int psp_v10_0_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type); extern int psp_v10_0_cmd_submit(struct psp_context *psp, diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index 2a535a4b8d5b..bcbe30dfff39 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -319,7 +319,7 @@ int psp_v3_1_ring_create(struct psp_context *psp, enum psp_ring_type ring_type) return ret; } -int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +int psp_v3_1_ring_stop(struct psp_context *psp, enum psp_ring_type ring_type) { int ret = 0; struct psp_ring *ring; @@ -339,6 +339,19 @@ int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), 0x80000000, 0x80000000, false); + return ret; +} + +int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + ret = psp_v3_1_ring_stop(psp, ring_type); + if (ret) + DRM_ERROR("Fail to stop psp ring\n"); + amdgpu_bo_free_kernel(&adev->firmware.rbuf, &ring->ring_mem_mc_addr, (void **)&ring->ring_mem); diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h index 9dcd0b25c4c6..5af2231b7099 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h @@ -41,6 +41,8 @@ extern int psp_v3_1_ring_init(struct psp_context *psp, enum psp_ring_type ring_type); extern int psp_v3_1_ring_create(struct psp_context *psp, enum psp_ring_type ring_type); +extern int psp_v3_1_ring_stop(struct psp_context *psp, + enum psp_ring_type ring_type); extern int psp_v3_1_ring_destroy(struct psp_context *psp, enum psp_ring_type ring_type); extern int psp_v3_1_cmd_submit(struct psp_context *psp, -- cgit From bcd6eab837fc9db67292c1d071ce2d96bb9689be Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Fri, 8 Sep 2017 13:09:50 +0800 Subject: drm/amdgpu: stop psp ring on suspend MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Otherwise, the ring will fail to create on next resume. Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Acked-by: Christian König Reviewed-by: Huang Rui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 10 ++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h | 1 + 2 files changed, 11 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index abd20819ba6b..0d530524ab21 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -455,6 +455,16 @@ static int psp_hw_fini(void *handle) static int psp_suspend(void *handle) { + int ret; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct psp_context *psp = &adev->psp; + + ret = psp_ring_stop(psp, PSP_RING_TYPE__KM); + if (ret) { + DRM_ERROR("PSP ring stop failed\n"); + return ret; + } + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index e79795b59797..1b7d12d88720 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -125,6 +125,7 @@ struct amdgpu_psp_funcs { #define psp_prep_cmd_buf(ucode, type) (psp)->prep_cmd_buf((ucode), (type)) #define psp_ring_init(psp, type) (psp)->ring_init((psp), (type)) #define psp_ring_create(psp, type) (psp)->ring_create((psp), (type)) +#define psp_ring_stop(psp, type) (psp)->ring_stop((psp), (type)) #define psp_ring_destroy(psp, type) ((psp)->ring_destroy((psp), (type))) #define psp_cmd_submit(psp, ucode, cmd_mc, fence_mc, index) \ (psp)->cmd_submit((psp), (ucode), (cmd_mc), (fence_mc), (index)) -- cgit From 50811c71fa1b7c2f8b422ad532d75a6860f15a8b Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Mon, 4 Sep 2017 17:48:27 +0800 Subject: drm/amdgpu: enable raven to load firmwares by psp at default (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - v2: share code with CHIP_VEGA10 case Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Acked-by: Christian König Reviewed-by: Huang Rui Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 -- drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 6 +----- 2 files changed, 1 insertion(+), 7 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 0d530524ab21..84bd6ed7a641 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -64,9 +64,7 @@ static int psp_sw_init(void *handle) psp->smu_reload_quirk = psp_v3_1_smu_reload_quirk; break; case CHIP_RAVEN: -#if 0 psp->init_microcode = psp_v10_0_init_microcode; -#endif psp->prep_cmd_buf = psp_v10_0_prep_cmd_buf; psp->ring_init = psp_v10_0_ring_init; psp->ring_create = psp_v10_0_ring_create; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 36c763310df5..f306374ff654 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -270,12 +270,8 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) else return AMDGPU_FW_LOAD_SMU; case CHIP_VEGA10: - if (!load_type) - return AMDGPU_FW_LOAD_DIRECT; - else - return AMDGPU_FW_LOAD_PSP; case CHIP_RAVEN: - if (load_type != 2) + if (!load_type) return AMDGPU_FW_LOAD_DIRECT; else return AMDGPU_FW_LOAD_PSP; -- cgit From d5884513a31df072879c89c80306d544467ee770 Mon Sep 17 00:00:00 2001 From: Christian König Date: Fri, 8 Sep 2017 14:09:41 +0200 Subject: drm/amdgpu: fix VM sync with always valid BOs v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit All users of a VM must always wait for updates with always valid BOs to be completed. v2: remove debugging leftovers, rename struct member Signed-off-by: Christian König Reviewed-by: Roger He Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 10 ++++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 15 ++++++++++----- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 +- 3 files changed, 17 insertions(+), 10 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 5f19227b35e9..ff61073b7181 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -761,10 +761,6 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (r) return r; - r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_dir_update); - if (r) - return r; - r = amdgpu_vm_clear_freed(adev, vm, NULL); if (r) return r; @@ -819,6 +815,12 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) } r = amdgpu_vm_handle_moved(adev, vm, &p->job->sync); + if (r) + return r; + + r = amdgpu_sync_fence(adev, &p->job->sync, vm->last_update); + if (r) + return r; if (amdgpu_vm_debug && p->bo_list) { /* Invalidate all BOs to test for userspace bugs */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 758bbb9e77f3..64baa3138965 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1141,9 +1141,8 @@ static int amdgpu_vm_update_level(struct amdgpu_device *adev, goto error_free; amdgpu_bo_fence(parent->base.bo, fence, true); - dma_fence_put(vm->last_dir_update); - vm->last_dir_update = dma_fence_get(fence); - dma_fence_put(fence); + dma_fence_put(vm->last_update); + vm->last_update = fence; } } @@ -1804,6 +1803,12 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, trace_amdgpu_vm_bo_mapping(mapping); } + if (bo_va->base.bo && + bo_va->base.bo->tbo.resv == vm->root.base.bo->tbo.resv) { + dma_fence_put(vm->last_update); + vm->last_update = dma_fence_get(bo_va->last_pt_update); + } + return 0; } @@ -2587,7 +2592,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->use_cpu_for_update ? "CPU" : "SDMA"); WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)), "CPU update of VM recommended only for large BAR system\n"); - vm->last_dir_update = NULL; + vm->last_update = NULL; flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_VRAM_CLEARED; @@ -2693,7 +2698,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) } amdgpu_vm_free_levels(&vm->root); - dma_fence_put(vm->last_dir_update); + dma_fence_put(vm->last_update); for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) amdgpu_vm_free_reserved_vmid(adev, vm, i); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index c1accd15efc8..cb6a6222fc3f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -140,7 +140,7 @@ struct amdgpu_vm { /* contains the page directory */ struct amdgpu_vm_pt root; - struct dma_fence *last_dir_update; + struct dma_fence *last_update; /* protecting freed */ spinlock_t freed_lock; -- cgit From 29c3035fe385b4214fc0515b9cd0ff53d23b4e82 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sat, 2 Sep 2017 02:16:35 -0400 Subject: drm/amdgpu/gfx9: properly set the hdp flush reg for Raven Was only being assigned for vega10. Reviewed-by: Junwei Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 2ab049c45b1d..72dbf890c65b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3504,7 +3504,9 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) u32 ref_and_mask, reg_mem_engine; struct nbio_hdp_flush_reg *nbio_hf_reg; - if (ring->adev->asic_type == CHIP_VEGA10) + if (ring->adev->flags & AMD_IS_APU) + nbio_hf_reg = &nbio_v7_0_hdp_flush_reg; + else nbio_hf_reg = &nbio_v6_1_hdp_flush_reg; if (ring->funcs->type == AMDGPU_RING_TYPE_COMPUTE) { -- cgit From 4e55eb3879fea6d8c7d414cebaa5bff1da58b4a1 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 11 Sep 2017 16:54:59 +0200 Subject: drm/amdgpu: fix amdgpu_vm_handle_moved as well v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit There is no guarantee that the last BO_VA actually needed an update. Additional to that all command submissions must wait for moved BOs to be cleared, not just the first one. v2: Don't overwrite any newer fence. Signed-off-by: Christian König Reviewed-by: Chunming Zhou Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 24 ++++++++++-------------- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 3 +-- 3 files changed, 12 insertions(+), 17 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index ff61073b7181..9f1202a4182f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -814,7 +814,7 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) } - r = amdgpu_vm_handle_moved(adev, vm, &p->job->sync); + r = amdgpu_vm_handle_moved(adev, vm); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 64baa3138965..2df254cc802e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1743,7 +1743,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, dma_addr_t *pages_addr = NULL; struct ttm_mem_reg *mem; struct drm_mm_node *nodes; - struct dma_fence *exclusive; + struct dma_fence *exclusive, **last_update; uint64_t flags; int r; @@ -1769,6 +1769,11 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, else flags = 0x0; + if (clear || (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv)) + last_update = &vm->last_update; + else + last_update = &bo_va->last_pt_update; + if (!clear && bo_va->base.moved) { bo_va->base.moved = false; list_splice_init(&bo_va->valids, &bo_va->invalids); @@ -1780,7 +1785,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, list_for_each_entry(mapping, &bo_va->invalids, list) { r = amdgpu_vm_bo_split_mapping(adev, exclusive, pages_addr, vm, mapping, flags, nodes, - &bo_va->last_pt_update); + last_update); if (r) return r; } @@ -1803,12 +1808,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, trace_amdgpu_vm_bo_mapping(mapping); } - if (bo_va->base.bo && - bo_va->base.bo->tbo.resv == vm->root.base.bo->tbo.resv) { - dma_fence_put(vm->last_update); - vm->last_update = dma_fence_get(bo_va->last_pt_update); - } - return 0; } @@ -2006,15 +2005,15 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, * PTs have to be reserved! */ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_sync *sync) + struct amdgpu_vm *vm) { - struct amdgpu_bo_va *bo_va = NULL; bool clear; int r = 0; spin_lock(&vm->status_lock); while (!list_empty(&vm->moved)) { + struct amdgpu_bo_va *bo_va; + bo_va = list_first_entry(&vm->moved, struct amdgpu_bo_va, base.vm_status); spin_unlock(&vm->status_lock); @@ -2030,9 +2029,6 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, } spin_unlock(&vm->status_lock); - if (bo_va) - r = amdgpu_sync_fence(adev, sync, bo_va->last_pt_update); - return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index cb6a6222fc3f..48c58ae4bb3a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -250,8 +250,7 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct dma_fence **fence); int amdgpu_vm_handle_moved(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_sync *sync); + struct amdgpu_vm *vm); int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, bool clear); -- cgit From a072c5f896beba806b4b867d478e1b90f94ba29b Mon Sep 17 00:00:00 2001 From: Michel Dänzer Date: Mon, 11 Sep 2017 17:04:41 +0900 Subject: amdgpu: Only destroy fbdev framebuffer if it was initialized MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Fixes crash when trying to unload the amdgpu module before the fbdev framebuffer was initialized, which can happen since the DRM fbdev helper code supports deferred setup. Acked-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Michel Dänzer Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index b6cb276f0a70..562930b17a6d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -303,10 +303,10 @@ static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfb if (rfb->obj) { amdgpufb_destroy_pinned_object(rfb->obj); rfb->obj = NULL; + drm_framebuffer_unregister_private(&rfb->base); + drm_framebuffer_cleanup(&rfb->base); } drm_fb_helper_fini(&rfbdev->helper); - drm_framebuffer_unregister_private(&rfb->base); - drm_framebuffer_cleanup(&rfb->base); return 0; } -- cgit From 9f0ed7aab60e3563bfe247bc2ad82db3a88c2d57 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 11 Sep 2017 15:51:30 +0200 Subject: drm/amdgpu: fix cgs alignment handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This always allocated on PAGE_SIZE alignment. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index fd435a96481c..892cd8b3483c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -121,7 +121,7 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, placement.busy_placement = &place; placement.num_busy_placement = 1; - ret = amdgpu_bo_create_restricted(adev, size, PAGE_SIZE, + ret = amdgpu_bo_create_restricted(adev, size, align, true, domain, flags, NULL, &placement, NULL, 0, &obj); -- cgit From 88531913a841a6354adfb40c78c86599639e5f32 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 11 Sep 2017 17:10:26 +0200 Subject: drm/amd: remove min/max addr handling from cgs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Nobody is actually using this and it causes a bunch of unused and buggy code. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 48 +++------------------------------ 1 file changed, 3 insertions(+), 45 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index 892cd8b3483c..df3bf22039d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -45,7 +45,6 @@ struct amdgpu_cgs_device { static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, enum cgs_gpu_mem_type type, uint64_t size, uint64_t align, - uint64_t min_offset, uint64_t max_offset, cgs_handle_t *handle) { CGS_FUNC_ADEV; @@ -53,13 +52,6 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, int ret = 0; uint32_t domain = 0; struct amdgpu_bo *obj; - struct ttm_placement placement; - struct ttm_place place; - - if (min_offset > max_offset) { - BUG_ON(1); - return -EINVAL; - } /* fail if the alignment is not a power of 2 */ if (((align != 1) && (align & (align - 1))) @@ -73,41 +65,19 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; domain = AMDGPU_GEM_DOMAIN_VRAM; - if (max_offset > adev->mc.real_vram_size) - return -EINVAL; - place.fpfn = min_offset >> PAGE_SHIFT; - place.lpfn = max_offset >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM; break; case CGS_GPU_MEM_TYPE__INVISIBLE_CONTIG_FB: case CGS_GPU_MEM_TYPE__INVISIBLE_FB: flags = AMDGPU_GEM_CREATE_NO_CPU_ACCESS | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; domain = AMDGPU_GEM_DOMAIN_VRAM; - if (adev->mc.visible_vram_size < adev->mc.real_vram_size) { - place.fpfn = - max(min_offset, adev->mc.visible_vram_size) >> PAGE_SHIFT; - place.lpfn = - min(max_offset, adev->mc.real_vram_size) >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_UNCACHED | - TTM_PL_FLAG_VRAM; - } - break; case CGS_GPU_MEM_TYPE__GART_CACHEABLE: domain = AMDGPU_GEM_DOMAIN_GTT; - place.fpfn = min_offset >> PAGE_SHIFT; - place.lpfn = max_offset >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_CACHED | TTM_PL_FLAG_TT; break; case CGS_GPU_MEM_TYPE__GART_WRITECOMBINE: flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; domain = AMDGPU_GEM_DOMAIN_GTT; - place.fpfn = min_offset >> PAGE_SHIFT; - place.lpfn = max_offset >> PAGE_SHIFT; - place.flags = TTM_PL_FLAG_WC | TTM_PL_FLAG_TT | - TTM_PL_FLAG_UNCACHED; break; default: return -EINVAL; @@ -116,15 +86,8 @@ static int amdgpu_cgs_alloc_gpu_mem(struct cgs_device *cgs_device, *handle = 0; - placement.placement = &place; - placement.num_placement = 1; - placement.busy_placement = &place; - placement.num_busy_placement = 1; - - ret = amdgpu_bo_create_restricted(adev, size, align, - true, domain, flags, - NULL, &placement, NULL, - 0, &obj); + ret = amdgpu_bo_create(adev, size, align, true, domain, flags, + NULL, NULL, 0, &obj); if (ret) { DRM_ERROR("(%d) bo create failed\n", ret); return ret; @@ -155,19 +118,14 @@ static int amdgpu_cgs_gmap_gpu_mem(struct cgs_device *cgs_device, cgs_handle_t h uint64_t *mcaddr) { int r; - u64 min_offset, max_offset; struct amdgpu_bo *obj = (struct amdgpu_bo *)handle; WARN_ON_ONCE(obj->placement.num_placement > 1); - min_offset = obj->placements[0].fpfn << PAGE_SHIFT; - max_offset = obj->placements[0].lpfn << PAGE_SHIFT; - r = amdgpu_bo_reserve(obj, true); if (unlikely(r != 0)) return r; - r = amdgpu_bo_pin_restricted(obj, obj->preferred_domains, - min_offset, max_offset, mcaddr); + r = amdgpu_bo_pin(obj, obj->preferred_domains, mcaddr); amdgpu_bo_unreserve(obj); return r; } -- cgit From c09312a6532a9a976ec4e72eb3b7fa10e87a8b07 Mon Sep 17 00:00:00 2001 From: Christian König Date: Tue, 12 Sep 2017 10:56:17 +0200 Subject: drm/amdgpu: fix and cleanup amdgpu_bo_create v2 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We adjusted the BO flags for USWC handling, but those never took effect because the placement was passed in instead of generated inside this function. v2: better commit message Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 83 +++++++++--------------------- drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 8 --- 2 files changed, 23 insertions(+), 68 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 52d0109c0d9c..726a662f43f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -64,11 +64,12 @@ bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) return false; } -static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, - struct ttm_placement *placement, - struct ttm_place *places, - u32 domain, u64 flags) +void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) { + struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); + struct ttm_placement *placement = &abo->placement; + struct ttm_place *places = abo->placements; + u64 flags = abo->flags; u32 c = 0; if (domain & AMDGPU_GEM_DOMAIN_VRAM) { @@ -151,27 +152,6 @@ static void amdgpu_ttm_placement_init(struct amdgpu_device *adev, placement->busy_placement = places; } -void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); - - amdgpu_ttm_placement_init(adev, &abo->placement, abo->placements, - domain, abo->flags); -} - -static void amdgpu_fill_placement_to_bo(struct amdgpu_bo *bo, - struct ttm_placement *placement) -{ - BUG_ON(placement->num_placement > (AMDGPU_GEM_DOMAIN_MAX + 1)); - - memcpy(bo->placements, placement->placement, - placement->num_placement * sizeof(struct ttm_place)); - bo->placement.num_placement = placement->num_placement; - bo->placement.num_busy_placement = placement->num_busy_placement; - bo->placement.placement = bo->placements; - bo->placement.busy_placement = bo->placements; -} - /** * amdgpu_bo_create_reserved - create reserved BO for kernel use * @@ -303,14 +283,13 @@ void amdgpu_bo_free_kernel(struct amdgpu_bo **bo, u64 *gpu_addr, *cpu_addr = NULL; } -int amdgpu_bo_create_restricted(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, - struct ttm_placement *placement, - struct reservation_object *resv, - uint64_t init_value, - struct amdgpu_bo **bo_ptr) +static int amdgpu_bo_do_create(struct amdgpu_device *adev, + unsigned long size, int byte_align, + bool kernel, u32 domain, u64 flags, + struct sg_table *sg, + struct reservation_object *resv, + uint64_t init_value, + struct amdgpu_bo **bo_ptr) { struct amdgpu_bo *bo; enum ttm_bo_type type; @@ -384,10 +363,11 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev, bo->flags &= ~AMDGPU_GEM_CREATE_CPU_GTT_USWC; #endif - amdgpu_fill_placement_to_bo(bo, placement); - /* Kernel allocation are uninterruptible */ + bo->tbo.bdev = &adev->mman.bdev; + amdgpu_ttm_placement_from_domain(bo, domain); initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); + /* Kernel allocation are uninterruptible */ r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, &bo->placement, page_align, !kernel, NULL, acc_size, sg, resv, &amdgpu_ttm_bo_destroy); @@ -442,27 +422,17 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, unsigned long size, int byte_align, struct amdgpu_bo *bo) { - struct ttm_placement placement = {0}; - struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; int r; if (bo->shadow) return 0; - memset(&placements, 0, sizeof(placements)); - amdgpu_ttm_placement_init(adev, &placement, placements, - AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_SHADOW); - - r = amdgpu_bo_create_restricted(adev, size, byte_align, true, - AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_SHADOW, - NULL, &placement, - bo->tbo.resv, - 0, - &bo->shadow); + r = amdgpu_bo_do_create(adev, size, byte_align, true, + AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_SHADOW, + NULL, bo->tbo.resv, 0, + &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -484,18 +454,11 @@ int amdgpu_bo_create(struct amdgpu_device *adev, uint64_t init_value, struct amdgpu_bo **bo_ptr) { - struct ttm_placement placement = {0}; - struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; int r; - memset(&placements, 0, sizeof(placements)); - amdgpu_ttm_placement_init(adev, &placement, placements, - domain, parent_flags); - - r = amdgpu_bo_create_restricted(adev, size, byte_align, kernel, domain, - parent_flags, sg, &placement, resv, - init_value, bo_ptr); + r = amdgpu_bo_do_create(adev, size, byte_align, kernel, domain, + parent_flags, sg, resv, init_value, bo_ptr); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index a4891bea2ca8..39b6bf6fb051 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -195,14 +195,6 @@ int amdgpu_bo_create(struct amdgpu_device *adev, struct reservation_object *resv, uint64_t init_value, struct amdgpu_bo **bo_ptr); -int amdgpu_bo_create_restricted(struct amdgpu_device *adev, - unsigned long size, int byte_align, - bool kernel, u32 domain, u64 flags, - struct sg_table *sg, - struct ttm_placement *placement, - struct reservation_object *resv, - uint64_t init_value, - struct amdgpu_bo **bo_ptr); int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, u32 domain, struct amdgpu_bo **bo_ptr, -- cgit From c30572814df55bc648ca449f69849b988abff54a Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Wed, 13 Sep 2017 12:35:15 -0400 Subject: drm/amd/amdgpu: Change vram debugfs to NO_KIQ for VM environments MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Reviewed-by: Christian König Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index fe887f361be8..b2b11e176667 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1696,9 +1696,9 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf, return result; spin_lock_irqsave(&adev->mmio_idx_lock, flags); - WREG32(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); - WREG32(mmMM_INDEX_HI, *pos >> 31); - value = RREG32(mmMM_DATA); + WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); + WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31); + value = RREG32_NO_KIQ(mmMM_DATA); spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); r = put_user(value, (uint32_t *)buf); @@ -1739,9 +1739,9 @@ static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf, return r; spin_lock_irqsave(&adev->mmio_idx_lock, flags); - WREG32(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); - WREG32(mmMM_INDEX_HI, *pos >> 31); - WREG32(mmMM_DATA, value); + WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)*pos) | 0x80000000); + WREG32_NO_KIQ(mmMM_INDEX_HI, *pos >> 31); + WREG32_NO_KIQ(mmMM_DATA, value); spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); result += 4; -- cgit From 376b6a1f4fbd4938042ec28bd9498c62a4bdcdf0 Mon Sep 17 00:00:00 2001 From: Allen Pais Date: Wed, 13 Sep 2017 13:31:09 +0530 Subject: drivers:gpu:Use ARRAY_SIZE() for the size calculation of the array. Signed-off-by: Allen Pais Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 74cb647da30e..7ff70762cfc8 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -273,7 +273,7 @@ static const struct pctl_data pctl0_data[] = { {0x135, 0x12a810}, {0x149, 0x7a82c} }; -#define PCTL0_DATA_LEN (sizeof(pctl0_data)/sizeof(pctl0_data[0])) +#define PCTL0_DATA_LEN (ARRAY_SIZE(pctl0_data)) #define PCTL0_RENG_EXEC_END_PTR 0x151 #define PCTL0_STCTRL_REG_SAVE_RANGE0_BASE 0xa640 @@ -309,7 +309,7 @@ static const struct pctl_data pctl1_data[] = { {0x1f0, 0x5000a7f6}, {0x1f1, 0x5000a7e4} }; -#define PCTL1_DATA_LEN (sizeof(pctl1_data)/sizeof(pctl1_data[0])) +#define PCTL1_DATA_LEN (ARRAY_SIZE(pctl1_data)) #define PCTL1_RENG_EXEC_END_PTR 0x1f1 #define PCTL1_STCTRL_REG_SAVE_RANGE0_BASE 0xa000 -- cgit From cf661ec1ff146183a6f8da1f1d41c1e3c4afe8bd Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Fri, 15 Sep 2017 17:36:19 -0400 Subject: drm/amdgpu/psp: declare raven psp firmware So it gets picked up properly by the kernel. Reviewed-by: Junwei Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/psp_v10_0.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index 6ec5c9f8074d..77cab1ff0254 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -35,6 +35,8 @@ #include "raven1/GC/gc_9_1_offset.h" #include "raven1/SDMA0/sdma0_4_1_offset.h" +MODULE_FIRMWARE("amdgpu/raven_asd.bin"); + static int psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) { -- cgit From 97bae49c44454549e7e22ef02020decafa099f04 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Thu, 14 Sep 2017 08:57:26 -0400 Subject: drm/amd/amdgpu: Support VM environments in amdgpu_ttm_access_memory() Signed-off-by: Tom St Denis Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index b2b11e176667..8ee16dfdb8af 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1201,14 +1201,14 @@ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo, } spin_lock_irqsave(&adev->mmio_idx_lock, flags); - WREG32(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000); - WREG32(mmMM_INDEX_HI, aligned_pos >> 31); + WREG32_NO_KIQ(mmMM_INDEX, ((uint32_t)aligned_pos) | 0x80000000); + WREG32_NO_KIQ(mmMM_INDEX_HI, aligned_pos >> 31); if (!write || mask != 0xffffffff) - value = RREG32(mmMM_DATA); + value = RREG32_NO_KIQ(mmMM_DATA); if (write) { value &= ~mask; value |= (*(uint32_t *)buf << shift) & mask; - WREG32(mmMM_DATA, value); + WREG32_NO_KIQ(mmMM_DATA, value); } spin_unlock_irqrestore(&adev->mmio_idx_lock, flags); if (!write) { -- cgit From cfa289fd4986c504b0396cce167802dfcf4943d0 Mon Sep 17 00:00:00 2001 From: Rex Zhu Date: Wed, 6 Sep 2017 15:27:59 +0800 Subject: drm/amdgpu: rename amdgpu_dpm_funcs to amd_pm_funcs renamed amdgpu_dpm_funcs and moved to amd_shared.h so can shared with powerplay. Reviewed-by: Alex Deucher Signed-off-by: Rex Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c | 4 +- drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h | 47 +-------------- drivers/gpu/drm/amd/amdgpu/ci_dpm.c | 100 +++++++++++++++++++++----------- drivers/gpu/drm/amd/amdgpu/kv_dpm.c | 53 +++++++++++------ drivers/gpu/drm/amd/amdgpu/si_dpm.c | 69 ++++++++++++++-------- 5 files changed, 150 insertions(+), 123 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c index 1cb52fd19060..e997ebbe43ea 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c @@ -960,8 +960,10 @@ u8 amdgpu_encode_pci_lane_width(u32 lanes) } struct amd_vce_state* -amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx) +amdgpu_get_vce_clock_state(void *handle, u32 idx) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (idx < adev->pm.dpm.num_of_vce_states) return &adev->pm.dpm.vce_states[idx]; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index 3eba4137508b..2f2bdb032d30 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -241,49 +241,6 @@ enum amdgpu_pcie_gen { AMDGPU_PCIE_GEN_INVALID = 0xffff }; -struct amdgpu_dpm_funcs { - int (*get_temperature)(struct amdgpu_device *adev); - int (*pre_set_power_state)(struct amdgpu_device *adev); - int (*set_power_state)(struct amdgpu_device *adev); - void (*post_set_power_state)(struct amdgpu_device *adev); - void (*display_configuration_changed)(struct amdgpu_device *adev); - u32 (*get_sclk)(struct amdgpu_device *adev, bool low); - u32 (*get_mclk)(struct amdgpu_device *adev, bool low); - void (*print_power_state)(struct amdgpu_device *adev, struct amdgpu_ps *ps); - void (*debugfs_print_current_performance_level)(struct amdgpu_device *adev, struct seq_file *m); - int (*force_performance_level)(struct amdgpu_device *adev, enum amd_dpm_forced_level level); - bool (*vblank_too_short)(struct amdgpu_device *adev); - void (*powergate_uvd)(struct amdgpu_device *adev, bool gate); - void (*powergate_vce)(struct amdgpu_device *adev, bool gate); - void (*enable_bapm)(struct amdgpu_device *adev, bool enable); - void (*set_fan_control_mode)(struct amdgpu_device *adev, u32 mode); - u32 (*get_fan_control_mode)(struct amdgpu_device *adev); - int (*set_fan_speed_percent)(struct amdgpu_device *adev, u32 speed); - int (*get_fan_speed_percent)(struct amdgpu_device *adev, u32 *speed); - int (*force_clock_level)(struct amdgpu_device *adev, enum pp_clock_type type, uint32_t mask); - int (*print_clock_levels)(struct amdgpu_device *adev, enum pp_clock_type type, char *buf); - int (*get_sclk_od)(struct amdgpu_device *adev); - int (*set_sclk_od)(struct amdgpu_device *adev, uint32_t value); - int (*get_mclk_od)(struct amdgpu_device *adev); - int (*set_mclk_od)(struct amdgpu_device *adev, uint32_t value); - int (*check_state_equal)(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, - bool *equal); - int (*read_sensor)(struct amdgpu_device *adev, int idx, void *value, - int *size); - - struct amd_vce_state* (*get_vce_clock_state)(struct amdgpu_device *adev, unsigned idx); - int (*reset_power_profile_state)(struct amdgpu_device *adev, - struct amd_pp_profile *request); - int (*get_power_profile_state)(struct amdgpu_device *adev, - struct amd_pp_profile *query); - int (*set_power_profile_state)(struct amdgpu_device *adev, - struct amd_pp_profile *request); - int (*switch_power_profile)(struct amdgpu_device *adev, - enum amd_pp_profile_type type); -}; - #define amdgpu_dpm_pre_set_power_state(adev) (adev)->pm.funcs->pre_set_power_state((adev)) #define amdgpu_dpm_set_power_state(adev) (adev)->pm.funcs->set_power_state((adev)) #define amdgpu_dpm_post_set_power_state(adev) (adev)->pm.funcs->post_set_power_state((adev)) @@ -485,7 +442,7 @@ struct amdgpu_pm { struct amdgpu_dpm dpm; const struct firmware *fw; /* SMC firmware */ uint32_t fw_version; - const struct amdgpu_dpm_funcs *funcs; + const struct amd_pm_funcs *funcs; uint32_t pcie_gen_mask; uint32_t pcie_mlw_mask; struct amd_pp_display_configuration pm_display_cfg;/* set by DAL */ @@ -551,6 +508,6 @@ u16 amdgpu_get_pcie_lane_support(struct amdgpu_device *adev, u8 amdgpu_encode_pci_lane_width(u32 lanes); struct amd_vce_state* -amdgpu_get_vce_clock_state(struct amdgpu_device *adev, unsigned idx); +amdgpu_get_vce_clock_state(void *handle, u32 idx); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index cb508a211b2f..bdf792822ff5 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -883,8 +883,9 @@ static int ci_power_control_set_level(struct amdgpu_device *adev) return ret; } -static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) +static void ci_dpm_powergate_uvd(void *handle, bool gate) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); pi->uvd_power_gated = gate; @@ -901,8 +902,9 @@ static void ci_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) } } -static bool ci_dpm_vblank_too_short(struct amdgpu_device *adev) +static bool ci_dpm_vblank_too_short(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 vblank_time = amdgpu_dpm_get_vblank_time(adev); u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 300; @@ -1210,11 +1212,12 @@ static int ci_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev) } } -static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev, +static int ci_dpm_get_fan_speed_percent(void *handle, u32 *speed) { u32 duty, duty100; u64 tmp64; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev->pm.no_fan) return -ENOENT; @@ -1237,12 +1240,13 @@ static int ci_dpm_get_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev, +static int ci_dpm_set_fan_speed_percent(void *handle, u32 speed) { u32 tmp; u32 duty, duty100; u64 tmp64; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (adev->pm.no_fan) @@ -1271,8 +1275,10 @@ static int ci_dpm_set_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) +static void ci_dpm_set_fan_control_mode(void *handle, u32 mode) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + switch (mode) { case AMD_FAN_CTRL_NONE: if (adev->pm.dpm.fan.ucode_fan_control) @@ -1292,8 +1298,9 @@ static void ci_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) } } -static u32 ci_dpm_get_fan_control_mode(struct amdgpu_device *adev) +static u32 ci_dpm_get_fan_control_mode(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (pi->fan_is_controlled_by_smc) @@ -4378,9 +4385,10 @@ static u32 ci_get_lowest_enabled_level(struct amdgpu_device *adev, } -static int ci_dpm_force_performance_level(struct amdgpu_device *adev, +static int ci_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); u32 tmp, levels, i; int ret; @@ -5291,8 +5299,9 @@ static void ci_update_requested_ps(struct amdgpu_device *adev, adev->pm.dpm.requested_ps = &pi->requested_rps; } -static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev) +static int ci_dpm_pre_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; struct amdgpu_ps *new_ps = &requested_ps; @@ -5304,8 +5313,9 @@ static int ci_dpm_pre_set_power_state(struct amdgpu_device *adev) return 0; } -static void ci_dpm_post_set_power_state(struct amdgpu_device *adev) +static void ci_dpm_post_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; @@ -5479,8 +5489,9 @@ static void ci_dpm_disable(struct amdgpu_device *adev) ci_update_current_ps(adev, boot_ps); } -static int ci_dpm_set_power_state(struct amdgpu_device *adev) +static int ci_dpm_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; struct amdgpu_ps *old_ps = &pi->current_rps; @@ -5551,8 +5562,10 @@ static void ci_dpm_reset_asic(struct amdgpu_device *adev) } #endif -static void ci_dpm_display_configuration_changed(struct amdgpu_device *adev) +static void ci_dpm_display_configuration_changed(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + ci_program_display_gap(adev); } @@ -6105,9 +6118,10 @@ static int ci_dpm_init(struct amdgpu_device *adev) } static void -ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, +ci_dpm_debugfs_print_current_performance_level(void *handle, struct seq_file *m) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amdgpu_ps *rps = &pi->current_rps; u32 sclk = ci_get_average_sclk_freq(adev); @@ -6131,12 +6145,13 @@ ci_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, seq_printf(m, "GPU load: %u %%\n", activity_percent); } -static void ci_dpm_print_power_state(struct amdgpu_device *adev, - struct amdgpu_ps *rps) +static void ci_dpm_print_power_state(void *handle, void *current_ps) { + struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps; struct ci_ps *ps = ci_get_ps(rps); struct ci_pl *pl; int i; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_dpm_print_class_info(rps->class, rps->class2); amdgpu_dpm_print_cap_info(rps->caps); @@ -6158,20 +6173,23 @@ static inline bool ci_are_power_levels_equal(const struct ci_pl *ci_cpl1, (ci_cpl1->pcie_lane == ci_cpl2->pcie_lane)); } -static int ci_check_state_equal(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, +static int ci_check_state_equal(void *handle, + void *current_ps, + void *request_ps, bool *equal) { struct ci_ps *ci_cps; struct ci_ps *ci_rps; int i; + struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) return -EINVAL; - ci_cps = ci_get_ps(cps); - ci_rps = ci_get_ps(rps); + ci_cps = ci_get_ps((struct amdgpu_ps *)cps); + ci_rps = ci_get_ps((struct amdgpu_ps *)rps); if (ci_cps == NULL) { *equal = false; @@ -6199,8 +6217,9 @@ static int ci_check_state_equal(struct amdgpu_device *adev, return 0; } -static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low) +static u32 ci_dpm_get_sclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps); @@ -6210,8 +6229,9 @@ static u32 ci_dpm_get_sclk(struct amdgpu_device *adev, bool low) return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk; } -static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low) +static u32 ci_dpm_get_mclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *requested_state = ci_get_ps(&pi->requested_rps); @@ -6222,10 +6242,11 @@ static u32 ci_dpm_get_mclk(struct amdgpu_device *adev, bool low) } /* get temperature in millidegrees */ -static int ci_dpm_get_temp(struct amdgpu_device *adev) +static int ci_dpm_get_temp(void *handle) { u32 temp; int actual_temp = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; temp = (RREG32_SMC(ixCG_MULT_THERMAL_STATUS) & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >> CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT; @@ -6551,9 +6572,10 @@ static int ci_dpm_set_powergating_state(void *handle, return 0; } -static int ci_dpm_print_clock_levels(struct amdgpu_device *adev, +static int ci_dpm_print_clock_levels(void *handle, enum pp_clock_type type, char *buf) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_single_dpm_table *sclk_table = &pi->dpm_table.sclk_table; struct ci_single_dpm_table *mclk_table = &pi->dpm_table.mclk_table; @@ -6618,9 +6640,10 @@ static int ci_dpm_print_clock_levels(struct amdgpu_device *adev, return size; } -static int ci_dpm_force_clock_level(struct amdgpu_device *adev, +static int ci_dpm_force_clock_level(void *handle, enum pp_clock_type type, uint32_t mask) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (adev->pm.dpm.forced_level & (AMD_DPM_FORCED_LEVEL_AUTO | @@ -6664,8 +6687,9 @@ static int ci_dpm_force_clock_level(struct amdgpu_device *adev, return 0; } -static int ci_dpm_get_sclk_od(struct amdgpu_device *adev) +static int ci_dpm_get_sclk_od(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_single_dpm_table *sclk_table = &(pi->dpm_table.sclk_table); struct ci_single_dpm_table *golden_sclk_table = @@ -6680,8 +6704,9 @@ static int ci_dpm_get_sclk_od(struct amdgpu_device *adev) return value; } -static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value) +static int ci_dpm_set_sclk_od(void *handle, uint32_t value) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps); struct ci_single_dpm_table *golden_sclk_table = @@ -6698,8 +6723,9 @@ static int ci_dpm_set_sclk_od(struct amdgpu_device *adev, uint32_t value) return 0; } -static int ci_dpm_get_mclk_od(struct amdgpu_device *adev) +static int ci_dpm_get_mclk_od(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_single_dpm_table *mclk_table = &(pi->dpm_table.mclk_table); struct ci_single_dpm_table *golden_mclk_table = @@ -6714,8 +6740,9 @@ static int ci_dpm_get_mclk_od(struct amdgpu_device *adev) return value; } -static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value) +static int ci_dpm_set_mclk_od(void *handle, uint32_t value) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct ci_ps *ps = ci_get_ps(adev->pm.dpm.requested_ps); struct ci_single_dpm_table *golden_mclk_table = @@ -6732,9 +6759,10 @@ static int ci_dpm_set_mclk_od(struct amdgpu_device *adev, uint32_t value) return 0; } -static int ci_dpm_get_power_profile_state(struct amdgpu_device *adev, +static int ci_dpm_get_power_profile_state(void *handle, struct amd_pp_profile *query) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (!pi || !query) @@ -6851,9 +6879,10 @@ static int ci_set_power_profile_state(struct amdgpu_device *adev, return result; } -static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev, +static int ci_dpm_set_power_profile_state(void *handle, struct amd_pp_profile *request) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); int ret = -1; @@ -6906,9 +6935,10 @@ static int ci_dpm_set_power_profile_state(struct amdgpu_device *adev, return 0; } -static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev, +static int ci_dpm_reset_power_profile_state(void *handle, struct amd_pp_profile *request) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); if (!pi || !request) @@ -6927,9 +6957,10 @@ static int ci_dpm_reset_power_profile_state(struct amdgpu_device *adev, return -EINVAL; } -static int ci_dpm_switch_power_profile(struct amdgpu_device *adev, +static int ci_dpm_switch_power_profile(void *handle, enum amd_pp_profile_type type) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct ci_power_info *pi = ci_get_pi(adev); struct amd_pp_profile request = {0}; @@ -6944,11 +6975,12 @@ static int ci_dpm_switch_power_profile(struct amdgpu_device *adev, return 0; } -static int ci_dpm_read_sensor(struct amdgpu_device *adev, int idx, +static int ci_dpm_read_sensor(void *handle, int idx, void *value, int *size) { u32 activity_percent = 50; int ret; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* size must be at least 4 bytes for all sensors */ if (*size < 4) @@ -7003,7 +7035,7 @@ const struct amd_ip_funcs ci_dpm_ip_funcs = { .set_powergating_state = ci_dpm_set_powergating_state, }; -static const struct amdgpu_dpm_funcs ci_dpm_funcs = { +static const struct amd_pm_funcs ci_dpm_funcs = { .get_temperature = &ci_dpm_get_temp, .pre_set_power_state = &ci_dpm_pre_set_power_state, .set_power_state = &ci_dpm_set_power_state, diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index 3bbf2ccfca89..f68d7abe1ed9 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -64,7 +64,7 @@ static int kv_set_thermal_temperature_range(struct amdgpu_device *adev, int min_temp, int max_temp); static int kv_init_fps_limits(struct amdgpu_device *adev); -static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate); +static void kv_dpm_powergate_uvd(void *handle, bool gate); static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate); static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate); static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate); @@ -1245,8 +1245,9 @@ static void kv_update_requested_ps(struct amdgpu_device *adev, adev->pm.dpm.requested_ps = &pi->requested_rps; } -static void kv_dpm_enable_bapm(struct amdgpu_device *adev, bool enable) +static void kv_dpm_enable_bapm(void *handle, bool enable) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); int ret; @@ -1672,8 +1673,9 @@ static int kv_update_acp_dpm(struct amdgpu_device *adev, bool gate) return kv_enable_acp_dpm(adev, !gate); } -static void kv_dpm_powergate_uvd(struct amdgpu_device *adev, bool gate) +static void kv_dpm_powergate_uvd(void *handle, bool gate) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); int ret; @@ -1868,10 +1870,11 @@ static int kv_enable_nb_dpm(struct amdgpu_device *adev, return ret; } -static int kv_dpm_force_performance_level(struct amdgpu_device *adev, +static int kv_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { int ret; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (level == AMD_DPM_FORCED_LEVEL_HIGH) { ret = kv_force_dpm_highest(adev); @@ -1892,8 +1895,9 @@ static int kv_dpm_force_performance_level(struct amdgpu_device *adev, return 0; } -static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev) +static int kv_dpm_pre_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; struct amdgpu_ps *new_ps = &requested_ps; @@ -1907,8 +1911,9 @@ static int kv_dpm_pre_set_power_state(struct amdgpu_device *adev) return 0; } -static int kv_dpm_set_power_state(struct amdgpu_device *adev) +static int kv_dpm_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; struct amdgpu_ps *old_ps = &pi->current_rps; @@ -1981,8 +1986,9 @@ static int kv_dpm_set_power_state(struct amdgpu_device *adev) return 0; } -static void kv_dpm_post_set_power_state(struct amdgpu_device *adev) +static void kv_dpm_post_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct amdgpu_ps *new_ps = &pi->requested_rps; @@ -2848,9 +2854,10 @@ static int kv_dpm_init(struct amdgpu_device *adev) } static void -kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, +kv_dpm_debugfs_print_current_performance_level(void *handle, struct seq_file *m) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); u32 current_index = (RREG32_SMC(ixTARGET_AND_CURRENT_PROFILE_INDEX) & @@ -2875,11 +2882,12 @@ kv_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, } static void -kv_dpm_print_power_state(struct amdgpu_device *adev, - struct amdgpu_ps *rps) +kv_dpm_print_power_state(void *handle, void *request_ps) { int i; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; struct kv_ps *ps = kv_get_ps(rps); + struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_dpm_print_class_info(rps->class, rps->class2); amdgpu_dpm_print_cap_info(rps->caps); @@ -2905,13 +2913,14 @@ static void kv_dpm_fini(struct amdgpu_device *adev) amdgpu_free_extended_power_table(adev); } -static void kv_dpm_display_configuration_changed(struct amdgpu_device *adev) +static void kv_dpm_display_configuration_changed(void *handle) { } -static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low) +static u32 kv_dpm_get_sclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); struct kv_ps *requested_state = kv_get_ps(&pi->requested_rps); @@ -2921,18 +2930,20 @@ static u32 kv_dpm_get_sclk(struct amdgpu_device *adev, bool low) return requested_state->levels[requested_state->num_levels - 1].sclk; } -static u32 kv_dpm_get_mclk(struct amdgpu_device *adev, bool low) +static u32 kv_dpm_get_mclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); return pi->sys_info.bootup_uma_clk; } /* get temperature in millidegrees */ -static int kv_dpm_get_temp(struct amdgpu_device *adev) +static int kv_dpm_get_temp(void *handle) { u32 temp; int actual_temp = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; temp = RREG32_SMC(0xC0300E0C); @@ -3222,14 +3233,17 @@ static inline bool kv_are_power_levels_equal(const struct kv_pl *kv_cpl1, (kv_cpl1->force_nbp_state == kv_cpl2->force_nbp_state)); } -static int kv_check_state_equal(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, +static int kv_check_state_equal(void *handle, + void *current_ps, + void *request_ps, bool *equal) { struct kv_ps *kv_cps; struct kv_ps *kv_rps; int i; + struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) return -EINVAL; @@ -3262,9 +3276,10 @@ static int kv_check_state_equal(struct amdgpu_device *adev, return 0; } -static int kv_dpm_read_sensor(struct amdgpu_device *adev, int idx, +static int kv_dpm_read_sensor(void *handle, int idx, void *value, int *size) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); uint32_t sclk; u32 pl_index = @@ -3312,7 +3327,7 @@ const struct amd_ip_funcs kv_dpm_ip_funcs = { .set_powergating_state = kv_dpm_set_powergating_state, }; -static const struct amdgpu_dpm_funcs kv_dpm_funcs = { +static const struct amd_pm_funcs kv_dpm_funcs = { .get_temperature = &kv_dpm_get_temp, .pre_set_power_state = &kv_dpm_pre_set_power_state, .set_power_state = &kv_dpm_set_power_state, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index d63873f3f574..05b3dbf585d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -3060,9 +3060,9 @@ static int si_get_vce_clock_voltage(struct amdgpu_device *adev, return ret; } -static bool si_dpm_vblank_too_short(struct amdgpu_device *adev) +static bool si_dpm_vblank_too_short(void *handle) { - + struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 vblank_time = amdgpu_dpm_get_vblank_time(adev); /* we never hit the non-gddr5 limit so disable it */ u32 switch_limit = adev->mc.vram_type == AMDGPU_VRAM_TYPE_GDDR5 ? 450 : 0; @@ -3871,9 +3871,10 @@ static int si_restrict_performance_levels_before_switch(struct amdgpu_device *ad 0 : -EINVAL; } -static int si_dpm_force_performance_level(struct amdgpu_device *adev, +static int si_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ps *rps = adev->pm.dpm.current_ps; struct si_ps *ps = si_get_ps(rps); u32 levels = ps->performance_level_count; @@ -6575,11 +6576,12 @@ static int si_fan_ctrl_stop_smc_fan_control(struct amdgpu_device *adev) } } -static int si_dpm_get_fan_speed_percent(struct amdgpu_device *adev, +static int si_dpm_get_fan_speed_percent(void *handle, u32 *speed) { u32 duty, duty100; u64 tmp64; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev->pm.no_fan) return -ENOENT; @@ -6600,9 +6602,10 @@ static int si_dpm_get_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static int si_dpm_set_fan_speed_percent(struct amdgpu_device *adev, +static int si_dpm_set_fan_speed_percent(void *handle, u32 speed) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct si_power_info *si_pi = si_get_pi(adev); u32 tmp; u32 duty, duty100; @@ -6633,8 +6636,10 @@ static int si_dpm_set_fan_speed_percent(struct amdgpu_device *adev, return 0; } -static void si_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) +static void si_dpm_set_fan_control_mode(void *handle, u32 mode) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (mode) { /* stop auto-manage */ if (adev->pm.dpm.fan.ucode_fan_control) @@ -6649,8 +6654,9 @@ static void si_dpm_set_fan_control_mode(struct amdgpu_device *adev, u32 mode) } } -static u32 si_dpm_get_fan_control_mode(struct amdgpu_device *adev) +static u32 si_dpm_get_fan_control_mode(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct si_power_info *si_pi = si_get_pi(adev); u32 tmp; @@ -6946,8 +6952,9 @@ static void si_dpm_disable(struct amdgpu_device *adev) ni_update_current_ps(adev, boot_ps); } -static int si_dpm_pre_set_power_state(struct amdgpu_device *adev) +static int si_dpm_pre_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps requested_ps = *adev->pm.dpm.requested_ps; struct amdgpu_ps *new_ps = &requested_ps; @@ -6984,8 +6991,9 @@ static int si_power_control_set_level(struct amdgpu_device *adev) return 0; } -static int si_dpm_set_power_state(struct amdgpu_device *adev) +static int si_dpm_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *new_ps = &eg_pi->requested_rps; struct amdgpu_ps *old_ps = &eg_pi->current_rps; @@ -7086,8 +7094,9 @@ static int si_dpm_set_power_state(struct amdgpu_device *adev) return 0; } -static void si_dpm_post_set_power_state(struct amdgpu_device *adev) +static void si_dpm_post_set_power_state(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *new_ps = &eg_pi->requested_rps; @@ -7103,8 +7112,10 @@ void si_dpm_reset_asic(struct amdgpu_device *adev) } #endif -static void si_dpm_display_configuration_changed(struct amdgpu_device *adev) +static void si_dpm_display_configuration_changed(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + si_program_display_gap(adev); } @@ -7486,9 +7497,10 @@ static void si_dpm_fini(struct amdgpu_device *adev) amdgpu_free_extended_power_table(adev); } -static void si_dpm_debugfs_print_current_performance_level(struct amdgpu_device *adev, +static void si_dpm_debugfs_print_current_performance_level(void *handle, struct seq_file *m) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *rps = &eg_pi->current_rps; struct si_ps *ps = si_get_ps(rps); @@ -7860,10 +7872,11 @@ static int si_dpm_set_powergating_state(void *handle, } /* get temperature in millidegrees */ -static int si_dpm_get_temp(struct amdgpu_device *adev) +static int si_dpm_get_temp(void *handle) { u32 temp; int actual_temp = 0; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; temp = (RREG32(CG_MULT_THERMAL_STATUS) & CTF_TEMP_MASK) >> CTF_TEMP_SHIFT; @@ -7878,8 +7891,9 @@ static int si_dpm_get_temp(struct amdgpu_device *adev) return actual_temp; } -static u32 si_dpm_get_sclk(struct amdgpu_device *adev, bool low) +static u32 si_dpm_get_sclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct si_ps *requested_state = si_get_ps(&eg_pi->requested_rps); @@ -7889,8 +7903,9 @@ static u32 si_dpm_get_sclk(struct amdgpu_device *adev, bool low) return requested_state->performance_levels[requested_state->performance_level_count - 1].sclk; } -static u32 si_dpm_get_mclk(struct amdgpu_device *adev, bool low) +static u32 si_dpm_get_mclk(void *handle, bool low) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct si_ps *requested_state = si_get_ps(&eg_pi->requested_rps); @@ -7900,9 +7915,11 @@ static u32 si_dpm_get_mclk(struct amdgpu_device *adev, bool low) return requested_state->performance_levels[requested_state->performance_level_count - 1].mclk; } -static void si_dpm_print_power_state(struct amdgpu_device *adev, - struct amdgpu_ps *rps) +static void si_dpm_print_power_state(void *handle, + void *current_ps) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_ps *rps = (struct amdgpu_ps *)current_ps; struct si_ps *ps = si_get_ps(rps); struct rv7xx_pl *pl; int i; @@ -7942,20 +7959,23 @@ static inline bool si_are_power_levels_equal(const struct rv7xx_pl *si_cpl1, (si_cpl1->vddci == si_cpl2->vddci)); } -static int si_check_state_equal(struct amdgpu_device *adev, - struct amdgpu_ps *cps, - struct amdgpu_ps *rps, +static int si_check_state_equal(void *handle, + void *current_ps, + void *request_ps, bool *equal) { struct si_ps *si_cps; struct si_ps *si_rps; int i; + struct amdgpu_ps *cps = (struct amdgpu_ps *)current_ps; + struct amdgpu_ps *rps = (struct amdgpu_ps *)request_ps; + struct amdgpu_device *adev = (struct amdgpu_device *)handle; if (adev == NULL || cps == NULL || rps == NULL || equal == NULL) return -EINVAL; - si_cps = si_get_ps(cps); - si_rps = si_get_ps(rps); + si_cps = si_get_ps((struct amdgpu_ps *)cps); + si_rps = si_get_ps((struct amdgpu_ps *)rps); if (si_cps == NULL) { printk("si_cps is NULL\n"); @@ -7983,9 +8003,10 @@ static int si_check_state_equal(struct amdgpu_device *adev, return 0; } -static int si_dpm_read_sensor(struct amdgpu_device *adev, int idx, +static int si_dpm_read_sensor(void *handle, int idx, void *value, int *size) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct evergreen_power_info *eg_pi = evergreen_get_pi(adev); struct amdgpu_ps *rps = &eg_pi->current_rps; struct si_ps *ps = si_get_ps(rps); @@ -8041,7 +8062,7 @@ const struct amd_ip_funcs si_dpm_ip_funcs = { .set_powergating_state = si_dpm_set_powergating_state, }; -static const struct amdgpu_dpm_funcs si_dpm_funcs = { +static const struct amd_pm_funcs si_dpm_funcs = { .get_temperature = &si_dpm_get_temp, .pre_set_power_state = &si_dpm_pre_set_power_state, .set_power_state = &si_dpm_set_power_state, -- cgit From cd4d74648b8021f0d0c7cc31a92fb4ea436a0019 Mon Sep 17 00:00:00 2001 From: Rex Zhu Date: Wed, 6 Sep 2017 18:43:52 +0800 Subject: drm/amdgpu: unify the interface of amd_pm_funcs put amd_pm_funcs table in struct powerplay for all asics. Signed-off-by: Rex Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5 +- drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h | 120 ++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c | 236 +++++++++++--------------- drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c | 3 + drivers/gpu/drm/amd/amdgpu/ci_dpm.c | 10 +- drivers/gpu/drm/amd/amdgpu/cik_dpm.h | 3 +- drivers/gpu/drm/amd/amdgpu/kv_dpm.c | 10 +- drivers/gpu/drm/amd/amdgpu/si_dpm.c | 10 +- drivers/gpu/drm/amd/amdgpu/si_dpm.h | 1 + 9 files changed, 161 insertions(+), 237 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 77a32b79e8f2..b0109ebe0a1b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3498,10 +3498,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf, valuesize = sizeof(values); if (adev->powerplay.pp_funcs && adev->powerplay.pp_funcs->read_sensor) - r = adev->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, idx, &values[0], &valuesize); - else if (adev->pm.funcs && adev->pm.funcs->read_sensor) - r = adev->pm.funcs->read_sensor(adev, idx, &values[0], - &valuesize); + r = amdgpu_dpm_read_sensor(adev, idx, &values[0], &valuesize); else return -EINVAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index 2f2bdb032d30..f79f9ea58b17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -241,134 +241,119 @@ enum amdgpu_pcie_gen { AMDGPU_PCIE_GEN_INVALID = 0xffff }; -#define amdgpu_dpm_pre_set_power_state(adev) (adev)->pm.funcs->pre_set_power_state((adev)) -#define amdgpu_dpm_set_power_state(adev) (adev)->pm.funcs->set_power_state((adev)) -#define amdgpu_dpm_post_set_power_state(adev) (adev)->pm.funcs->post_set_power_state((adev)) -#define amdgpu_dpm_display_configuration_changed(adev) (adev)->pm.funcs->display_configuration_changed((adev)) -#define amdgpu_dpm_print_power_state(adev, ps) (adev)->pm.funcs->print_power_state((adev), (ps)) -#define amdgpu_dpm_vblank_too_short(adev) (adev)->pm.funcs->vblank_too_short((adev)) -#define amdgpu_dpm_enable_bapm(adev, e) (adev)->pm.funcs->enable_bapm((adev), (e)) +#define amdgpu_dpm_pre_set_power_state(adev) \ + ((adev)->powerplay.pp_funcs->pre_set_power_state((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_set_power_state(adev) \ + ((adev)->powerplay.pp_funcs->set_power_state((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_post_set_power_state(adev) \ + ((adev)->powerplay.pp_funcs->post_set_power_state((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_display_configuration_changed(adev) \ + ((adev)->powerplay.pp_funcs->display_configuration_changed((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_print_power_state(adev, ps) \ + ((adev)->powerplay.pp_funcs->print_power_state((adev)->powerplay.pp_handle, (ps))) + +#define amdgpu_dpm_vblank_too_short(adev) \ + ((adev)->powerplay.pp_funcs->vblank_too_short((adev)->powerplay.pp_handle)) + +#define amdgpu_dpm_enable_bapm(adev, e) \ + ((adev)->powerplay.pp_funcs->enable_bapm((adev)->powerplay.pp_handle, (e))) #define amdgpu_dpm_read_sensor(adev, idx, value, size) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->read_sensor(adev->powerplay.pp_handle, (idx), (value), (size)) : \ - (adev)->pm.funcs->read_sensor((adev), (idx), (value), (size))) + ((adev)->powerplay.pp_funcs->read_sensor((adev)->powerplay.pp_handle, (idx), (value), (size))) #define amdgpu_dpm_get_temperature(adev) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle) : \ - (adev)->pm.funcs->get_temperature((adev))) + ((adev)->powerplay.pp_funcs->get_temperature((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_fan_control_mode(adev, m) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m)) : \ - (adev)->pm.funcs->set_fan_control_mode((adev), (m))) + ((adev)->powerplay.pp_funcs->set_fan_control_mode((adev)->powerplay.pp_handle, (m))) #define amdgpu_dpm_get_fan_control_mode(adev) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle) : \ - (adev)->pm.funcs->get_fan_control_mode((adev))) + ((adev)->powerplay.pp_funcs->get_fan_control_mode((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_fan_speed_percent(adev, s) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \ - (adev)->pm.funcs->set_fan_speed_percent((adev), (s))) + ((adev)->powerplay.pp_funcs->set_fan_speed_percent((adev)->powerplay.pp_handle, (s))) #define amdgpu_dpm_get_fan_speed_percent(adev, s) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s)) : \ - (adev)->pm.funcs->get_fan_speed_percent((adev), (s))) + ((adev)->powerplay.pp_funcs->get_fan_speed_percent((adev)->powerplay.pp_handle, (s))) #define amdgpu_dpm_get_fan_speed_rpm(adev, s) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_fan_speed_rpm((adev)->powerplay.pp_handle, (s)) : \ - -EINVAL) + ((adev)->powerplay.pp_funcs->get_fan_speed_rpm)((adev)->powerplay.pp_handle, (s)) #define amdgpu_dpm_get_sclk(adev, l) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l)) : \ - (adev)->pm.funcs->get_sclk((adev), (l))) + ((adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l))) #define amdgpu_dpm_get_mclk(adev, l) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l)) : \ - (adev)->pm.funcs->get_mclk((adev), (l))) - + ((adev)->powerplay.pp_funcs->get_mclk((adev)->powerplay.pp_handle, (l))) #define amdgpu_dpm_force_performance_level(adev, l) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l)) : \ - (adev)->pm.funcs->force_performance_level((adev), (l))) + ((adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l))) #define amdgpu_dpm_powergate_uvd(adev, g) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g)) : \ - (adev)->pm.funcs->powergate_uvd((adev), (g))) + ((adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g))) #define amdgpu_dpm_powergate_vce(adev, g) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g)) : \ - (adev)->pm.funcs->powergate_vce((adev), (g))) + ((adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g))) #define amdgpu_dpm_get_current_power_state(adev) \ - (adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle) + ((adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle)) #define amdgpu_dpm_get_pp_num_states(adev, data) \ - (adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data) + ((adev)->powerplay.pp_funcs->get_pp_num_states((adev)->powerplay.pp_handle, data)) #define amdgpu_dpm_get_pp_table(adev, table) \ - (adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table) + ((adev)->powerplay.pp_funcs->get_pp_table((adev)->powerplay.pp_handle, table)) #define amdgpu_dpm_set_pp_table(adev, buf, size) \ - (adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size) + ((adev)->powerplay.pp_funcs->set_pp_table((adev)->powerplay.pp_handle, buf, size)) #define amdgpu_dpm_print_clock_levels(adev, type, buf) \ - (adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf) + ((adev)->powerplay.pp_funcs->print_clock_levels((adev)->powerplay.pp_handle, type, buf)) #define amdgpu_dpm_force_clock_level(adev, type, level) \ - (adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level) + ((adev)->powerplay.pp_funcs->force_clock_level((adev)->powerplay.pp_handle, type, level)) #define amdgpu_dpm_get_sclk_od(adev) \ - (adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle) + ((adev)->powerplay.pp_funcs->get_sclk_od((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_sclk_od(adev, value) \ - (adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value) + ((adev)->powerplay.pp_funcs->set_sclk_od((adev)->powerplay.pp_handle, value)) #define amdgpu_dpm_get_mclk_od(adev) \ - ((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle)) + ((adev)->powerplay.pp_funcs->get_mclk_od((adev)->powerplay.pp_handle)) #define amdgpu_dpm_set_mclk_od(adev, value) \ - ((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value)) + ((adev)->powerplay.pp_funcs->set_mclk_od((adev)->powerplay.pp_handle, value)) #define amdgpu_dpm_dispatch_task(adev, task_id, input, output) \ - ((adev)->powerplay.pp_funcs->dispatch_tasks)((adev)->powerplay.pp_handle, (task_id), (input), (output)) + ((adev)->powerplay.pp_funcs->dispatch_tasks)((adev)->powerplay.pp_handle, (task_id), (input), (output)) -#define amgdpu_dpm_check_state_equal(adev, cps, rps, equal) (adev)->pm.funcs->check_state_equal((adev), (cps),(rps),(equal)) +#define amdgpu_dpm_check_state_equal(adev, cps, rps, equal) \ + ((adev)->powerplay.pp_funcs->check_state_equal((adev)->powerplay.pp_handle, (cps), (rps), (equal))) #define amdgpu_dpm_get_vce_clock_state(adev, i) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i)) : \ - (adev)->pm.funcs->get_vce_clock_state((adev), (i))) + ((adev)->powerplay.pp_funcs->get_vce_clock_state((adev)->powerplay.pp_handle, (i))) -#define amdgpu_dpm_get_performance_level(adev) \ - ((adev)->pp_enabled ? \ - (adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle) : \ - (adev)->pm.dpm.forced_level) +#define amdgpu_dpm_get_performance_level(adev) \ + ((adev)->powerplay.pp_funcs->get_performance_level((adev)->powerplay.pp_handle)) #define amdgpu_dpm_reset_power_profile_state(adev, request) \ - ((adev)->powerplay.pp_funcs->reset_power_profile_state(\ + ((adev)->powerplay.pp_funcs->reset_power_profile_state(\ (adev)->powerplay.pp_handle, request)) #define amdgpu_dpm_get_power_profile_state(adev, query) \ - ((adev)->powerplay.pp_funcs->get_power_profile_state(\ + ((adev)->powerplay.pp_funcs->get_power_profile_state(\ (adev)->powerplay.pp_handle, query)) #define amdgpu_dpm_set_power_profile_state(adev, request) \ - ((adev)->powerplay.pp_funcs->set_power_profile_state(\ + ((adev)->powerplay.pp_funcs->set_power_profile_state(\ (adev)->powerplay.pp_handle, request)) #define amdgpu_dpm_switch_power_profile(adev, type) \ - ((adev)->powerplay.pp_funcs->switch_power_profile(\ + ((adev)->powerplay.pp_funcs->switch_power_profile(\ (adev)->powerplay.pp_handle, type)) struct amdgpu_dpm { @@ -442,7 +427,6 @@ struct amdgpu_pm { struct amdgpu_dpm dpm; const struct firmware *fw; /* SMC firmware */ uint32_t fw_version; - const struct amd_pm_funcs *funcs; uint32_t pcie_gen_mask; uint32_t pcie_mlw_mask; struct amd_pp_display_configuration pm_display_cfg;/* set by DAL */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 10c5d78081ed..f6ce52956e6d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -74,7 +74,7 @@ void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev) adev->pm.dpm.ac_power = true; else adev->pm.dpm.ac_power = false; - if (adev->pm.funcs->enable_bapm) + if (adev->powerplay.pp_funcs->enable_bapm) amdgpu_dpm_enable_bapm(adev, adev->pm.dpm.ac_power); mutex_unlock(&adev->pm.mutex); } @@ -88,9 +88,9 @@ static ssize_t amdgpu_get_dpm_state(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; enum amd_pm_state_type pm; - if (adev->pp_enabled) { + if (adev->powerplay.pp_funcs->get_current_power_state) pm = amdgpu_dpm_get_current_power_state(adev); - } else + else pm = adev->pm.dpm.user_state; return snprintf(buf, PAGE_SIZE, "%s\n", @@ -140,13 +140,17 @@ static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - enum amd_dpm_forced_level level; + enum amd_dpm_forced_level level = 0xff; if ((adev->flags & AMD_IS_PX) && (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return snprintf(buf, PAGE_SIZE, "off\n"); - level = amdgpu_dpm_get_performance_level(adev); + if (adev->powerplay.pp_funcs->get_performance_level) + level = amdgpu_dpm_get_performance_level(adev); + else + level = adev->pm.dpm.forced_level; + return snprintf(buf, PAGE_SIZE, "%s\n", (level == AMD_DPM_FORCED_LEVEL_AUTO) ? "auto" : (level == AMD_DPM_FORCED_LEVEL_LOW) ? "low" : @@ -167,7 +171,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; enum amd_dpm_forced_level level; - enum amd_dpm_forced_level current_level; + enum amd_dpm_forced_level current_level = 0xff; int ret = 0; /* Can't force performance level when the card is off */ @@ -175,7 +179,8 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev, (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return -EINVAL; - current_level = amdgpu_dpm_get_performance_level(adev); + if (adev->powerplay.pp_funcs->get_performance_level) + current_level = amdgpu_dpm_get_performance_level(adev); if (strncmp("low", buf, strlen("low")) == 0) { level = AMD_DPM_FORCED_LEVEL_LOW; @@ -203,9 +208,7 @@ static ssize_t amdgpu_set_dpm_forced_performance_level(struct device *dev, if (current_level == level) return count; - if (adev->pp_enabled) - amdgpu_dpm_force_performance_level(adev, level); - else { + if (adev->powerplay.pp_funcs->force_performance_level) { mutex_lock(&adev->pm.mutex); if (adev->pm.dpm.thermal_active) { count = -EINVAL; @@ -233,7 +236,7 @@ static ssize_t amdgpu_get_pp_num_states(struct device *dev, struct pp_states_info data; int i, buf_len; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_pp_num_states) amdgpu_dpm_get_pp_num_states(adev, &data); buf_len = snprintf(buf, PAGE_SIZE, "states: %d\n", data.nums); @@ -257,8 +260,8 @@ static ssize_t amdgpu_get_pp_cur_state(struct device *dev, enum amd_pm_state_type pm = 0; int i = 0; - if (adev->pp_enabled) { - + if (adev->powerplay.pp_funcs->get_current_power_state + && adev->powerplay.pp_funcs->get_pp_num_states) { pm = amdgpu_dpm_get_current_power_state(adev); amdgpu_dpm_get_pp_num_states(adev, &data); @@ -280,25 +283,10 @@ static ssize_t amdgpu_get_pp_force_state(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - struct pp_states_info data; - enum amd_pm_state_type pm = 0; - int i; - - if (adev->pp_force_state_enabled && adev->pp_enabled) { - pm = amdgpu_dpm_get_current_power_state(adev); - amdgpu_dpm_get_pp_num_states(adev, &data); - - for (i = 0; i < data.nums; i++) { - if (pm == data.states[i]) - break; - } - if (i == data.nums) - i = -EINVAL; - - return snprintf(buf, PAGE_SIZE, "%d\n", i); - - } else + if (adev->pp_force_state_enabled) + return amdgpu_get_pp_cur_state(dev, attr, buf); + else return snprintf(buf, PAGE_SIZE, "\n"); } @@ -347,7 +335,7 @@ static ssize_t amdgpu_get_pp_table(struct device *dev, char *table = NULL; int size; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_pp_table) size = amdgpu_dpm_get_pp_table(adev, &table); else return 0; @@ -368,7 +356,7 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->set_pp_table) amdgpu_dpm_set_pp_table(adev, buf, count); return count; @@ -380,14 +368,11 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - ssize_t size = 0; - if (adev->pp_enabled) - size = amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf); - else if (adev->pm.funcs->print_clock_levels) - size = adev->pm.funcs->print_clock_levels(adev, PP_SCLK, buf); - - return size; + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_SCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); } static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, @@ -416,10 +401,9 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, mask |= 1 << level; } - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask); - else if (adev->pm.funcs->force_clock_level) - adev->pm.funcs->force_clock_level(adev, PP_SCLK, mask); + fail: return count; } @@ -430,14 +414,11 @@ static ssize_t amdgpu_get_pp_dpm_mclk(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - ssize_t size = 0; - - if (adev->pp_enabled) - size = amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf); - else if (adev->pm.funcs->print_clock_levels) - size = adev->pm.funcs->print_clock_levels(adev, PP_MCLK, buf); - return size; + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_MCLK, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); } static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, @@ -465,11 +446,9 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, } mask |= 1 << level; } - - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask); - else if (adev->pm.funcs->force_clock_level) - adev->pm.funcs->force_clock_level(adev, PP_MCLK, mask); + fail: return count; } @@ -480,14 +459,11 @@ static ssize_t amdgpu_get_pp_dpm_pcie(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - ssize_t size = 0; - - if (adev->pp_enabled) - size = amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf); - else if (adev->pm.funcs->print_clock_levels) - size = adev->pm.funcs->print_clock_levels(adev, PP_PCIE, buf); - return size; + if (adev->powerplay.pp_funcs->print_clock_levels) + return amdgpu_dpm_print_clock_levels(adev, PP_PCIE, buf); + else + return snprintf(buf, PAGE_SIZE, "\n"); } static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, @@ -515,11 +491,9 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, } mask |= 1 << level; } - - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask); - else if (adev->pm.funcs->force_clock_level) - adev->pm.funcs->force_clock_level(adev, PP_PCIE, mask); + fail: return count; } @@ -532,10 +506,8 @@ static ssize_t amdgpu_get_pp_sclk_od(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; uint32_t value = 0; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_sclk_od) value = amdgpu_dpm_get_sclk_od(adev); - else if (adev->pm.funcs->get_sclk_od) - value = adev->pm.funcs->get_sclk_od(adev); return snprintf(buf, PAGE_SIZE, "%d\n", value); } @@ -556,12 +528,12 @@ static ssize_t amdgpu_set_pp_sclk_od(struct device *dev, count = -EINVAL; goto fail; } + if (adev->powerplay.pp_funcs->set_sclk_od) + amdgpu_dpm_set_sclk_od(adev, (uint32_t)value); if (adev->pp_enabled) { - amdgpu_dpm_set_sclk_od(adev, (uint32_t)value); amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); - } else if (adev->pm.funcs->set_sclk_od) { - adev->pm.funcs->set_sclk_od(adev, (uint32_t)value); + } else { adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; amdgpu_pm_compute_clocks(adev); } @@ -578,10 +550,8 @@ static ssize_t amdgpu_get_pp_mclk_od(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; uint32_t value = 0; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_mclk_od) value = amdgpu_dpm_get_mclk_od(adev); - else if (adev->pm.funcs->get_mclk_od) - value = adev->pm.funcs->get_mclk_od(adev); return snprintf(buf, PAGE_SIZE, "%d\n", value); } @@ -602,12 +572,12 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev, count = -EINVAL; goto fail; } + if (adev->powerplay.pp_funcs->set_mclk_od) + amdgpu_dpm_set_mclk_od(adev, (uint32_t)value); if (adev->pp_enabled) { - amdgpu_dpm_set_mclk_od(adev, (uint32_t)value); amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_READJUST_POWER_STATE, NULL, NULL); - } else if (adev->pm.funcs->set_mclk_od) { - adev->pm.funcs->set_mclk_od(adev, (uint32_t)value); + } else { adev->pm.dpm.current_ps = adev->pm.dpm.boot_ps; amdgpu_pm_compute_clocks(adev); } @@ -621,14 +591,11 @@ static ssize_t amdgpu_get_pp_power_profile(struct device *dev, { struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; - int ret = 0; + int ret = 0xff; - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->get_power_profile_state) ret = amdgpu_dpm_get_power_profile_state( adev, query); - else if (adev->pm.funcs->get_power_profile_state) - ret = adev->pm.funcs->get_power_profile_state( - adev, query); if (ret) return ret; @@ -675,15 +642,12 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev, char *sub_str, buf_cpy[128], *tmp_str; const char delimiter[3] = {' ', '\n', '\0'}; long int value; - int ret = 0; + int ret = 0xff; if (strncmp("reset", buf, strlen("reset")) == 0) { - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->reset_power_profile_state) ret = amdgpu_dpm_reset_power_profile_state( adev, request); - else if (adev->pm.funcs->reset_power_profile_state) - ret = adev->pm.funcs->reset_power_profile_state( - adev, request); if (ret) { count = -EINVAL; goto fail; @@ -692,12 +656,10 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev, } if (strncmp("set", buf, strlen("set")) == 0) { - if (adev->pp_enabled) + if (adev->powerplay.pp_funcs->set_power_profile_state) ret = amdgpu_dpm_set_power_profile_state( adev, request); - else if (adev->pm.funcs->set_power_profile_state) - ret = adev->pm.funcs->set_power_profile_state( - adev, request); + if (ret) { count = -EINVAL; goto fail; @@ -745,13 +707,8 @@ static ssize_t amdgpu_set_pp_power_profile(struct device *dev, loop++; } - - if (adev->pp_enabled) - ret = amdgpu_dpm_set_power_profile_state( - adev, request); - else if (adev->pm.funcs->set_power_profile_state) - ret = adev->pm.funcs->set_power_profile_state( - adev, request); + if (adev->powerplay.pp_funcs->set_power_profile_state) + ret = amdgpu_dpm_set_power_profile_state(adev, request); if (ret) count = -EINVAL; @@ -831,7 +788,7 @@ static ssize_t amdgpu_hwmon_show_temp(struct device *dev, (ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return -EINVAL; - if (!adev->pp_enabled && !adev->pm.funcs->get_temperature) + if (!adev->powerplay.pp_funcs->get_temperature) temp = 0; else temp = amdgpu_dpm_get_temperature(adev); @@ -862,7 +819,7 @@ static ssize_t amdgpu_hwmon_get_pwm1_enable(struct device *dev, struct amdgpu_device *adev = dev_get_drvdata(dev); u32 pwm_mode = 0; - if (!adev->pp_enabled && !adev->pm.funcs->get_fan_control_mode) + if (!adev->powerplay.pp_funcs->get_fan_control_mode) return -EINVAL; pwm_mode = amdgpu_dpm_get_fan_control_mode(adev); @@ -879,7 +836,7 @@ static ssize_t amdgpu_hwmon_set_pwm1_enable(struct device *dev, int err; int value; - if (!adev->pp_enabled && !adev->pm.funcs->set_fan_control_mode) + if (!adev->powerplay.pp_funcs->set_fan_control_mode) return -EINVAL; err = kstrtoint(buf, 10, &value); @@ -919,9 +876,11 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev, value = (value * 100) / 255; - err = amdgpu_dpm_set_fan_speed_percent(adev, value); - if (err) - return err; + if (adev->powerplay.pp_funcs->set_fan_speed_percent) { + err = amdgpu_dpm_set_fan_speed_percent(adev, value); + if (err) + return err; + } return count; } @@ -932,11 +891,13 @@ static ssize_t amdgpu_hwmon_get_pwm1(struct device *dev, { struct amdgpu_device *adev = dev_get_drvdata(dev); int err; - u32 speed; + u32 speed = 0; - err = amdgpu_dpm_get_fan_speed_percent(adev, &speed); - if (err) - return err; + if (adev->powerplay.pp_funcs->get_fan_speed_percent) { + err = amdgpu_dpm_get_fan_speed_percent(adev, &speed); + if (err) + return err; + } speed = (speed * 255) / 100; @@ -949,11 +910,13 @@ static ssize_t amdgpu_hwmon_get_fan1_input(struct device *dev, { struct amdgpu_device *adev = dev_get_drvdata(dev); int err; - u32 speed; + u32 speed = 0; - err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed); - if (err) - return err; + if (adev->powerplay.pp_funcs->get_fan_speed_rpm) { + err = amdgpu_dpm_get_fan_speed_rpm(adev, &speed); + if (err) + return err; + } return sprintf(buf, "%i\n", speed); } @@ -1008,21 +971,21 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, return 0; /* mask fan attributes if we have no bindings for this asic to expose */ - if ((!adev->pm.funcs->get_fan_speed_percent && + if ((!adev->powerplay.pp_funcs->get_fan_speed_percent && attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't query fan */ - (!adev->pm.funcs->get_fan_control_mode && + (!adev->powerplay.pp_funcs->get_fan_control_mode && attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't query state */ effective_mode &= ~S_IRUGO; - if ((!adev->pm.funcs->set_fan_speed_percent && + if ((!adev->powerplay.pp_funcs->set_fan_speed_percent && attr == &sensor_dev_attr_pwm1.dev_attr.attr) || /* can't manage fan */ - (!adev->pm.funcs->set_fan_control_mode && + (!adev->powerplay.pp_funcs->set_fan_control_mode && attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't manage state */ effective_mode &= ~S_IWUSR; /* hide max/min values if we can't both query and manage the fan */ - if ((!adev->pm.funcs->set_fan_speed_percent && - !adev->pm.funcs->get_fan_speed_percent) && + if ((!adev->powerplay.pp_funcs->set_fan_speed_percent && + !adev->powerplay.pp_funcs->get_fan_speed_percent) && (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) return 0; @@ -1055,7 +1018,7 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) if (!adev->pm.dpm_enabled) return; - if (adev->pm.funcs->get_temperature) { + if (adev->powerplay.pp_funcs->get_temperature) { int temp = amdgpu_dpm_get_temperature(adev); if (temp < adev->pm.dpm.thermal.min_temp) @@ -1087,7 +1050,7 @@ static struct amdgpu_ps *amdgpu_dpm_pick_power_state(struct amdgpu_device *adev, true : false; /* check if the vblank period is too short to adjust the mclk */ - if (single_display && adev->pm.funcs->vblank_too_short) { + if (single_display && adev->powerplay.pp_funcs->vblank_too_short) { if (amdgpu_dpm_vblank_too_short(adev)) single_display = false; } @@ -1216,7 +1179,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) struct amdgpu_ps *ps; enum amd_pm_state_type dpm_state; int ret; - bool equal; + bool equal = false; /* if dpm init failed */ if (!adev->pm.dpm_enabled) @@ -1236,7 +1199,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) else return; - if (amdgpu_dpm == 1) { + if (amdgpu_dpm == 1 && adev->powerplay.pp_funcs->print_power_state) { printk("switching from power state:\n"); amdgpu_dpm_print_power_state(adev, adev->pm.dpm.current_ps); printk("switching to power state:\n"); @@ -1245,15 +1208,17 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) /* update whether vce is active */ ps->vce_active = adev->pm.dpm.vce_active; - - amdgpu_dpm_display_configuration_changed(adev); + if (adev->powerplay.pp_funcs->display_configuration_changed) + amdgpu_dpm_display_configuration_changed(adev); ret = amdgpu_dpm_pre_set_power_state(adev); if (ret) return; - if ((0 != amgdpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal))) - equal = false; + if (adev->powerplay.pp_funcs->check_state_equal) { + if (0 != amdgpu_dpm_check_state_equal(adev, adev->pm.dpm.current_ps, adev->pm.dpm.requested_ps, &equal)) + equal = false; + } if (equal) return; @@ -1264,7 +1229,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) adev->pm.dpm.current_active_crtcs = adev->pm.dpm.new_active_crtcs; adev->pm.dpm.current_active_crtc_count = adev->pm.dpm.new_active_crtc_count; - if (adev->pm.funcs->force_performance_level) { + if (adev->powerplay.pp_funcs->force_performance_level) { if (adev->pm.dpm.thermal_active) { enum amd_dpm_forced_level level = adev->pm.dpm.forced_level; /* force low perf level for thermal */ @@ -1280,7 +1245,7 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled || adev->pm.funcs->powergate_uvd) { + if (adev->powerplay.pp_funcs->powergate_uvd) { /* enable/disable UVD */ mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_uvd(adev, !enable); @@ -1302,7 +1267,7 @@ void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) { - if (adev->pp_enabled || adev->pm.funcs->powergate_vce) { + if (adev->powerplay.pp_funcs->powergate_vce) { /* enable/disable VCE */ mutex_lock(&adev->pm.mutex); amdgpu_dpm_powergate_vce(adev, !enable); @@ -1337,8 +1302,7 @@ void amdgpu_pm_print_power_states(struct amdgpu_device *adev) { int i; - if (adev->pp_enabled) - /* TO DO */ + if (adev->powerplay.pp_funcs->print_power_state == NULL) return; for (i = 0; i < adev->pm.dpm.num_ps; i++) @@ -1353,10 +1317,8 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) if (adev->pm.sysfs_initialized) return 0; - if (!adev->pp_enabled) { - if (adev->pm.funcs->get_temperature == NULL) - return 0; - } + if (adev->powerplay.pp_funcs->get_temperature == NULL) + return 0; adev->pm.int_hwmon_dev = hwmon_device_register_with_groups(adev->dev, DRIVER_NAME, adev, @@ -1634,8 +1596,8 @@ static int amdgpu_debugfs_pm_info(struct seq_file *m, void *data) return amdgpu_debugfs_pm_info_pp(m, adev); } else { mutex_lock(&adev->pm.mutex); - if (adev->pm.funcs->debugfs_print_current_performance_level) - adev->pm.funcs->debugfs_print_current_performance_level(adev, m); + if (adev->powerplay.pp_funcs->debugfs_print_current_performance_level) + adev->powerplay.pp_funcs->debugfs_print_current_performance_level(adev, m); else seq_printf(m, "Debugfs support not implemented for this asic\n"); mutex_unlock(&adev->pm.mutex); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c index b0c4db8098ed..5cd5b8ee9744 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -87,17 +87,20 @@ static int amdgpu_pp_early_init(void *handle) case CHIP_OLAND: case CHIP_HAINAN: amd_pp->ip_funcs = &si_dpm_ip_funcs; + amd_pp->pp_funcs = &si_dpm_funcs; break; #endif #ifdef CONFIG_DRM_AMDGPU_CIK case CHIP_BONAIRE: case CHIP_HAWAII: amd_pp->ip_funcs = &ci_dpm_ip_funcs; + amd_pp->pp_funcs = &ci_dpm_funcs; break; case CHIP_KABINI: case CHIP_MULLINS: case CHIP_KAVERI: amd_pp->ip_funcs = &kv_dpm_ip_funcs; + amd_pp->pp_funcs = &kv_dpm_funcs; break; #endif default: diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index bdf792822ff5..68ce1bdaf2fc 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -307,7 +307,6 @@ static int ci_set_power_limit(struct amdgpu_device *adev, u32 n); static int ci_set_overdrive_target_tdp(struct amdgpu_device *adev, u32 target_tdp); static int ci_update_uvd_dpm(struct amdgpu_device *adev, bool gate); -static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev); static void ci_dpm_set_irq_funcs(struct amdgpu_device *adev); static PPSMC_Result amdgpu_ci_send_msg_to_smc_with_parameter(struct amdgpu_device *adev, @@ -6282,7 +6281,6 @@ static int ci_dpm_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - ci_dpm_set_dpm_funcs(adev); ci_dpm_set_irq_funcs(adev); return 0; @@ -7035,7 +7033,7 @@ const struct amd_ip_funcs ci_dpm_ip_funcs = { .set_powergating_state = ci_dpm_set_powergating_state, }; -static const struct amd_pm_funcs ci_dpm_funcs = { +const struct amd_pm_funcs ci_dpm_funcs = { .get_temperature = &ci_dpm_get_temp, .pre_set_power_state = &ci_dpm_pre_set_power_state, .set_power_state = &ci_dpm_set_power_state, @@ -7067,12 +7065,6 @@ static const struct amd_pm_funcs ci_dpm_funcs = { .read_sensor = ci_dpm_read_sensor, }; -static void ci_dpm_set_dpm_funcs(struct amdgpu_device *adev) -{ - if (adev->pm.funcs == NULL) - adev->pm.funcs = &ci_dpm_funcs; -} - static const struct amdgpu_irq_src_funcs ci_dpm_irq_funcs = { .set = ci_dpm_set_interrupt_state, .process = ci_dpm_process_interrupt, diff --git a/drivers/gpu/drm/amd/amdgpu/cik_dpm.h b/drivers/gpu/drm/amd/amdgpu/cik_dpm.h index b1c8e7b446ea..c7b4349f6319 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/cik_dpm.h @@ -26,5 +26,6 @@ extern const struct amd_ip_funcs ci_dpm_ip_funcs; extern const struct amd_ip_funcs kv_dpm_ip_funcs; - +extern const struct amd_pm_funcs ci_dpm_funcs; +extern const struct amd_pm_funcs kv_dpm_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index f68d7abe1ed9..b57399a462c2 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -42,7 +42,6 @@ #define KV_MINIMUM_ENGINE_CLOCK 800 #define SMC_RAM_END 0x40000 -static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev); static void kv_dpm_set_irq_funcs(struct amdgpu_device *adev); static int kv_enable_nb_dpm(struct amdgpu_device *adev, bool enable); @@ -2961,7 +2960,6 @@ static int kv_dpm_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - kv_dpm_set_dpm_funcs(adev); kv_dpm_set_irq_funcs(adev); return 0; @@ -3327,7 +3325,7 @@ const struct amd_ip_funcs kv_dpm_ip_funcs = { .set_powergating_state = kv_dpm_set_powergating_state, }; -static const struct amd_pm_funcs kv_dpm_funcs = { +const struct amd_pm_funcs kv_dpm_funcs = { .get_temperature = &kv_dpm_get_temp, .pre_set_power_state = &kv_dpm_pre_set_power_state, .set_power_state = &kv_dpm_set_power_state, @@ -3345,12 +3343,6 @@ static const struct amd_pm_funcs kv_dpm_funcs = { .read_sensor = &kv_dpm_read_sensor, }; -static void kv_dpm_set_dpm_funcs(struct amdgpu_device *adev) -{ - if (adev->pm.funcs == NULL) - adev->pm.funcs = &kv_dpm_funcs; -} - static const struct amdgpu_irq_src_funcs kv_dpm_irq_funcs = { .set = kv_dpm_set_interrupt_state, .process = kv_dpm_process_interrupt, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index 05b3dbf585d5..9b8db6046271 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -1847,7 +1847,6 @@ static int si_calculate_sclk_params(struct amdgpu_device *adev, static void si_thermal_start_smc_fan_control(struct amdgpu_device *adev); static void si_fan_ctrl_set_default_mode(struct amdgpu_device *adev); -static void si_dpm_set_dpm_funcs(struct amdgpu_device *adev); static void si_dpm_set_irq_funcs(struct amdgpu_device *adev); static struct si_power_info *si_get_pi(struct amdgpu_device *adev) @@ -7944,7 +7943,6 @@ static int si_dpm_early_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; - si_dpm_set_dpm_funcs(adev); si_dpm_set_irq_funcs(adev); return 0; } @@ -8062,7 +8060,7 @@ const struct amd_ip_funcs si_dpm_ip_funcs = { .set_powergating_state = si_dpm_set_powergating_state, }; -static const struct amd_pm_funcs si_dpm_funcs = { +const struct amd_pm_funcs si_dpm_funcs = { .get_temperature = &si_dpm_get_temp, .pre_set_power_state = &si_dpm_pre_set_power_state, .set_power_state = &si_dpm_set_power_state, @@ -8083,12 +8081,6 @@ static const struct amd_pm_funcs si_dpm_funcs = { .read_sensor = &si_dpm_read_sensor, }; -static void si_dpm_set_dpm_funcs(struct amdgpu_device *adev) -{ - if (adev->pm.funcs == NULL) - adev->pm.funcs = &si_dpm_funcs; -} - static const struct amdgpu_irq_src_funcs si_dpm_irq_funcs = { .set = si_dpm_set_interrupt_state, .process = si_dpm_process_interrupt, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.h b/drivers/gpu/drm/amd/amdgpu/si_dpm.h index 51ce21c5f4fb..9fe343de3477 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.h @@ -246,6 +246,7 @@ enum si_display_gap }; extern const struct amd_ip_funcs si_dpm_ip_funcs; +extern const struct amd_pm_funcs si_dpm_funcs; struct ni_leakage_coeffients { -- cgit From 6df9855fe200d4e7e5cdd85575fb28cce808b2cc Mon Sep 17 00:00:00 2001 From: Rex Zhu Date: Fri, 8 Sep 2017 14:05:51 +0800 Subject: drm/amdgpu: add support for request SI/CI firmware in CGS Reviewed-by: Alex Deucher Signed-off-by: Rex Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 79 +++++++++++++++++++++++++++++++++ 1 file changed, 79 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index df3bf22039d5..e52192051507 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -633,6 +633,85 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, if (!adev->pm.fw) { switch (adev->asic_type) { + case CHIP_TAHITI: + strcpy(fw_name, "radeon/tahiti_smc.bin"); + break; + case CHIP_PITCAIRN: + if ((adev->pdev->revision == 0x81) && + ((adev->pdev->device == 0x6810) || + (adev->pdev->device == 0x6811))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/pitcairn_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/pitcairn_smc.bin"); + } + break; + case CHIP_VERDE: + if (((adev->pdev->device == 0x6820) && + ((adev->pdev->revision == 0x81) || + (adev->pdev->revision == 0x83))) || + ((adev->pdev->device == 0x6821) && + ((adev->pdev->revision == 0x83) || + (adev->pdev->revision == 0x87))) || + ((adev->pdev->revision == 0x87) && + ((adev->pdev->device == 0x6823) || + (adev->pdev->device == 0x682b)))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/verde_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/verde_smc.bin"); + } + break; + case CHIP_OLAND: + if (((adev->pdev->revision == 0x81) && + ((adev->pdev->device == 0x6600) || + (adev->pdev->device == 0x6604) || + (adev->pdev->device == 0x6605) || + (adev->pdev->device == 0x6610))) || + ((adev->pdev->revision == 0x83) && + (adev->pdev->device == 0x6610))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/oland_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/oland_smc.bin"); + } + break; + case CHIP_HAINAN: + if (((adev->pdev->revision == 0x81) && + (adev->pdev->device == 0x6660)) || + ((adev->pdev->revision == 0x83) && + ((adev->pdev->device == 0x6660) || + (adev->pdev->device == 0x6663) || + (adev->pdev->device == 0x6665) || + (adev->pdev->device == 0x6667)))) { + info->is_kicker = true; + strcpy(fw_name, "radeon/hainan_k_smc.bin"); + } else if ((adev->pdev->revision == 0xc3) && + (adev->pdev->device == 0x6665)) { + info->is_kicker = true; + strcpy(fw_name, "radeon/banks_k_2_smc.bin"); + } else { + strcpy(fw_name, "radeon/hainan_smc.bin"); + } + break; + case CHIP_BONAIRE: + if ((adev->pdev->revision == 0x80) || + (adev->pdev->revision == 0x81) || + (adev->pdev->device == 0x665f)) { + info->is_kicker = true; + strcpy(fw_name, "radeon/bonaire_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/bonaire_smc.bin"); + } + break; + case CHIP_HAWAII: + if (adev->pdev->revision == 0x80) { + info->is_kicker = true; + strcpy(fw_name, "radeon/hawaii_k_smc.bin"); + } else { + strcpy(fw_name, "radeon/hawaii_smc.bin"); + } + break; case CHIP_TOPAZ: if (((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x81)) || ((adev->pdev->device == 0x6900) && (adev->pdev->revision == 0x83)) || -- cgit From 780cffc599b640f1ea1ab051496ad1fed4532150 Mon Sep 17 00:00:00 2001 From: Rex Zhu Date: Tue, 12 Sep 2017 13:37:40 +0800 Subject: drm/amdgpu: add powerplay support for CI asics currently, for CI asics, use dpm by default, amdgpu.dpm=-1. when set amdgpu.dpm=1, enable powplay. when set amdgpu.dpm=0, disable both dpm and powerplay. when powerplay is stable on CI asics, ci_dpm will be removed. Reviewed-by: Alex Deucher Signed-off-by: Rex Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c index 5cd5b8ee9744..2d2f0960b025 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_powerplay.c @@ -93,8 +93,16 @@ static int amdgpu_pp_early_init(void *handle) #ifdef CONFIG_DRM_AMDGPU_CIK case CHIP_BONAIRE: case CHIP_HAWAII: - amd_pp->ip_funcs = &ci_dpm_ip_funcs; - amd_pp->pp_funcs = &ci_dpm_funcs; + if (amdgpu_dpm == -1) { + amd_pp->ip_funcs = &ci_dpm_ip_funcs; + amd_pp->pp_funcs = &ci_dpm_funcs; + } else { + adev->pp_enabled = true; + if (amdgpu_create_pp_handle(adev)) + return -EINVAL; + amd_pp->ip_funcs = &pp_ip_funcs; + amd_pp->pp_funcs = &pp_dpm_funcs; + } break; case CHIP_KABINI: case CHIP_MULLINS: -- cgit From ca290da8f6345c8e8e180256fbe092c751fa9654 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Fri, 25 Aug 2017 20:15:04 -0400 Subject: drm/amdgpu: Fix error handling in amdgpu_vm_init Make sure vm->root.bo is not left reserved if amdgpu_bo_kmap fails. Signed-off-by: Felix Kuehling Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 2df254cc802e..2196bca7331c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2615,9 +2615,9 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, goto error_free_root; r = amdgpu_bo_kmap(vm->root.base.bo, NULL); + amdgpu_bo_unreserve(vm->root.base.bo); if (r) goto error_free_root; - amdgpu_bo_unreserve(vm->root.base.bo); } return 0; -- cgit From 02208441cc3a5110191996bb129db39ff10e7395 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Fri, 25 Aug 2017 20:40:26 -0400 Subject: drm/amdgpu: Add PASID management Allows assigning a PASID to a VM for identifying VMs involved in page faults. The global PASID manager is also exported in the KFD interface so that AMDGPU and KFD can share the PASID space. PASIDs of different sizes can be requested. On APUs, the PASID size is deterined by the capabilities of the IOMMU. So KFD must be able to allocate PASIDs in a smaller range. Signed-off-by: Felix Kuehling Acked-by: Alex Deucher Reviewed-by: Oded Gabbay Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 2 + drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 75 ++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 14 ++++- 5 files changed, 91 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index b9dbbf9cb8b0..dc7e25cce741 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -169,6 +169,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_vmem_size = get_vmem_size, .get_gpu_clock_counter = get_gpu_clock_counter, .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, + .alloc_pasid = amdgpu_vm_alloc_pasid, + .free_pasid = amdgpu_vm_free_pasid, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_pipeline = kgd_init_pipeline, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 309f2419c6d8..c678c69936a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -128,6 +128,8 @@ static const struct kfd2kgd_calls kfd2kgd = { .get_vmem_size = get_vmem_size, .get_gpu_clock_counter = get_gpu_clock_counter, .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, + .alloc_pasid = amdgpu_vm_alloc_pasid, + .free_pasid = amdgpu_vm_free_pasid, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, .init_pipeline = kgd_init_pipeline, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index e16229000a98..79d9ab43d42c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -825,7 +825,7 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) } r = amdgpu_vm_init(adev, &fpriv->vm, - AMDGPU_VM_CONTEXT_GFX); + AMDGPU_VM_CONTEXT_GFX, 0); if (r) { kfree(fpriv); goto out_suspend; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 2196bca7331c..9b795915cab1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -27,11 +27,58 @@ */ #include #include +#include #include #include #include "amdgpu.h" #include "amdgpu_trace.h" +/* + * PASID manager + * + * PASIDs are global address space identifiers that can be shared + * between the GPU, an IOMMU and the driver. VMs on different devices + * may use the same PASID if they share the same address + * space. Therefore PASIDs are allocated using a global IDA. VMs are + * looked up from the PASID per amdgpu_device. + */ +static DEFINE_IDA(amdgpu_vm_pasid_ida); + +/** + * amdgpu_vm_alloc_pasid - Allocate a PASID + * @bits: Maximum width of the PASID in bits, must be at least 1 + * + * Allocates a PASID of the given width while keeping smaller PASIDs + * available if possible. + * + * Returns a positive integer on success. Returns %-EINVAL if bits==0. + * Returns %-ENOSPC if no PASID was available. Returns %-ENOMEM on + * memory allocation failure. + */ +int amdgpu_vm_alloc_pasid(unsigned int bits) +{ + int pasid = -EINVAL; + + for (bits = min(bits, 31U); bits > 0; bits--) { + pasid = ida_simple_get(&amdgpu_vm_pasid_ida, + 1U << (bits - 1), 1U << bits, + GFP_KERNEL); + if (pasid != -ENOSPC) + break; + } + + return pasid; +} + +/** + * amdgpu_vm_free_pasid - Free a PASID + * @pasid: PASID to free + */ +void amdgpu_vm_free_pasid(unsigned int pasid) +{ + ida_simple_remove(&amdgpu_vm_pasid_ida, pasid); +} + /* * GPUVM * GPUVM is similar to the legacy gart on older asics, however @@ -2539,7 +2586,7 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint64_t vm_size, uint32_ * Init @vm fields. */ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context) + int vm_context, unsigned int pasid) { const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE, AMDGPU_VM_PTE_COUNT(adev) * 8); @@ -2620,6 +2667,19 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, goto error_free_root; } + if (pasid) { + unsigned long flags; + + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + r = idr_alloc(&adev->vm_manager.pasid_idr, vm, pasid, pasid + 1, + GFP_ATOMIC); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + if (r < 0) + goto error_free_root; + + vm->pasid = pasid; + } + return 0; error_free_root: @@ -2673,6 +2733,14 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; int i; + if (vm->pasid) { + unsigned long flags; + + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + idr_remove(&adev->vm_manager.pasid_idr, vm->pasid); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + } + amd_sched_entity_fini(vm->entity.sched, &vm->entity); if (!RB_EMPTY_ROOT(&vm->va)) { @@ -2752,6 +2820,8 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) adev->vm_manager.vm_update_mode = 0; #endif + idr_init(&adev->vm_manager.pasid_idr); + spin_lock_init(&adev->vm_manager.pasid_lock); } /** @@ -2765,6 +2835,9 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) { unsigned i, j; + WARN_ON(!idr_is_empty(&adev->vm_manager.pasid_idr)); + idr_destroy(&adev->vm_manager.pasid_idr); + for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) { struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[i]; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 48c58ae4bb3a..7873dfa8c0f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -25,6 +25,7 @@ #define __AMDGPU_VM_H__ #include +#include #include "gpu_scheduler.h" #include "amdgpu_sync.h" @@ -148,8 +149,9 @@ struct amdgpu_vm { /* Scheduler entity for page table updates */ struct amd_sched_entity entity; - /* client id */ + /* client id and PASID (TODO: replace client_id with PASID) */ u64 client_id; + unsigned int pasid; /* dedicated to vm */ struct amdgpu_vm_id *reserved_vmid[AMDGPU_MAX_VMHUBS]; @@ -220,12 +222,20 @@ struct amdgpu_vm_manager { * BIT1[= 0] Compute updated by SDMA [= 1] by CPU */ int vm_update_mode; + + /* PASID to VM mapping, will be used in interrupt context to + * look up VM of a page fault + */ + struct idr pasid_idr; + spinlock_t pasid_lock; }; +int amdgpu_vm_alloc_pasid(unsigned int bits); +void amdgpu_vm_free_pasid(unsigned int pasid); void amdgpu_vm_manager_init(struct amdgpu_device *adev); void amdgpu_vm_manager_fini(struct amdgpu_device *adev); int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, - int vm_context); + int vm_context, unsigned int pasid); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, -- cgit From 00ecd8a27c03b6dd463ab8755dd6d58751d76297 Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Sat, 26 Aug 2017 02:40:45 -0400 Subject: drm/amdgpu: Add prescreening stage in IH processing (v2) To filter out high-frequency interrupts that can be safely ignored. v2: squash in trivial typo fix for si (Alex) Signed-off-by: Felix Kuehling Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c | 6 ++++++ drivers/gpu/drm/amd/amdgpu/cik_ih.c | 14 ++++++++++++++ drivers/gpu/drm/amd/amdgpu/cz_ih.c | 14 ++++++++++++++ drivers/gpu/drm/amd/amdgpu/iceland_ih.c | 14 ++++++++++++++ drivers/gpu/drm/amd/amdgpu/si_ih.c | 14 ++++++++++++++ drivers/gpu/drm/amd/amdgpu/tonga_ih.c | 14 ++++++++++++++ drivers/gpu/drm/amd/amdgpu/vega10_ih.c | 14 ++++++++++++++ 8 files changed, 92 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 802fdc11944b..d62a35ef26be 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -334,6 +334,7 @@ struct amdgpu_gart_funcs { struct amdgpu_ih_funcs { /* ring read/write ptr handling, called from interrupt context */ u32 (*get_wptr)(struct amdgpu_device *adev); + bool (*prescreen_iv)(struct amdgpu_device *adev); void (*decode_iv)(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry); void (*set_rptr)(struct amdgpu_device *adev); @@ -1749,6 +1750,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r)) #define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o)) #define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev)) +#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev)) #define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv)) #define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev)) #define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 3ab4c65ecc8b..c834a40cfad6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -169,6 +169,12 @@ restart_ih: while (adev->irq.ih.rptr != wptr) { u32 ring_index = adev->irq.ih.rptr >> 2; + /* Prescreening of high-frequency interrupts */ + if (!amdgpu_ih_prescreen_iv(adev)) { + adev->irq.ih.rptr &= adev->irq.ih.ptr_mask; + continue; + } + /* Before dispatching irq to IP blocks, send it to amdkfd */ amdgpu_amdkfd_interrupt(adev, (const void *) &adev->irq.ih.ring[ring_index]); diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index b8918432c572..07d3d895da10 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -228,6 +228,19 @@ static u32 cik_ih_get_wptr(struct amdgpu_device *adev) * [127:96] - reserved */ +/** + * cik_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool cik_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* Process all interrupts */ + return true; +} + /** * cik_ih_decode_iv - decode an interrupt vector * @@ -433,6 +446,7 @@ static const struct amd_ip_funcs cik_ih_ip_funcs = { static const struct amdgpu_ih_funcs cik_ih_funcs = { .get_wptr = cik_ih_get_wptr, + .prescreen_iv = cik_ih_prescreen_iv, .decode_iv = cik_ih_decode_iv, .set_rptr = cik_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index 0c1209cdd1cb..b6cdf4afaf46 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -207,6 +207,19 @@ static u32 cz_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } +/** + * cz_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool cz_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* Process all interrupts */ + return true; +} + /** * cz_ih_decode_iv - decode an interrupt vector * @@ -414,6 +427,7 @@ static const struct amd_ip_funcs cz_ih_ip_funcs = { static const struct amdgpu_ih_funcs cz_ih_funcs = { .get_wptr = cz_ih_get_wptr, + .prescreen_iv = cz_ih_prescreen_iv, .decode_iv = cz_ih_decode_iv, .set_rptr = cz_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index 7a0ea27ac429..65ed6d3a8f05 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -207,6 +207,19 @@ static u32 iceland_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } +/** + * iceland_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool iceland_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* Process all interrupts */ + return true; +} + /** * iceland_ih_decode_iv - decode an interrupt vector * @@ -412,6 +425,7 @@ static const struct amd_ip_funcs iceland_ih_ip_funcs = { static const struct amdgpu_ih_funcs iceland_ih_funcs = { .get_wptr = iceland_ih_get_wptr, + .prescreen_iv = iceland_ih_prescreen_iv, .decode_iv = iceland_ih_decode_iv, .set_rptr = iceland_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c index ce25e03a077d..d2c6b80309c8 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c @@ -118,6 +118,19 @@ static u32 si_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } +/** + * si_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool si_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* Process all interrupts */ + return true; +} + static void si_ih_decode_iv(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { @@ -288,6 +301,7 @@ static const struct amd_ip_funcs si_ih_ip_funcs = { static const struct amdgpu_ih_funcs si_ih_funcs = { .get_wptr = si_ih_get_wptr, + .prescreen_iv = si_ih_prescreen_iv, .decode_iv = si_ih_decode_iv, .set_rptr = si_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index 923df2c0e535..5ed00692618e 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -218,6 +218,19 @@ static u32 tonga_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } +/** + * tonga_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool tonga_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* Process all interrupts */ + return true; +} + /** * tonga_ih_decode_iv - decode an interrupt vector * @@ -478,6 +491,7 @@ static const struct amd_ip_funcs tonga_ih_ip_funcs = { static const struct amdgpu_ih_funcs tonga_ih_funcs = { .get_wptr = tonga_ih_get_wptr, + .prescreen_iv = tonga_ih_prescreen_iv, .decode_iv = tonga_ih_decode_iv, .set_rptr = tonga_ih_set_rptr }; diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index 56150e8d1ed2..eda4771e273f 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -226,6 +226,19 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev) return (wptr & adev->irq.ih.ptr_mask); } +/** + * vega10_ih_prescreen_iv - prescreen an interrupt vector + * + * @adev: amdgpu_device pointer + * + * Returns true if the interrupt vector should be further processed. + */ +static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) +{ + /* TODO: Filter known pending page faults */ + return true; +} + /** * vega10_ih_decode_iv - decode an interrupt vector * @@ -410,6 +423,7 @@ const struct amd_ip_funcs vega10_ih_ip_funcs = { static const struct amdgpu_ih_funcs vega10_ih_funcs = { .get_wptr = vega10_ih_get_wptr, + .prescreen_iv = vega10_ih_prescreen_iv, .decode_iv = vega10_ih_decode_iv, .set_rptr = vega10_ih_set_rptr }; -- cgit From a2f14820e3493145c25095873d4a510a1b25efdc Mon Sep 17 00:00:00 2001 From: Felix Kuehling Date: Sat, 26 Aug 2017 02:43:06 -0400 Subject: drm/amdgpu: Track pending retry faults in IH and VM (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit IH tracks pending retry faults in a hash table for fast lookup in interrupt context. Each VM has a short FIFO of pending VM faults for processing in a bottom half. The IH prescreening stage adds retry faults and filters out repeated retry interrupts to minimize the impact of interrupt storms. It's the VM's responsibility remove pending faults once they are handled. For now this is only done when the VM is destroyed. v2: - Made the hash table smaller and the FIFO longer. I never want the FIFO to fill up, because that would make prescreen take longer. 128 pending page faults should be enough to keep migrations busy. Signed-off-by: Felix Kuehling Acked-by: Christian König (v1) Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c | 76 +++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h | 12 ++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 7 +++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 7 +++ drivers/gpu/drm/amd/amdgpu/vega10_ih.c | 78 +++++++++++++++++++++++++++++++++- 5 files changed, 179 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index c834a40cfad6..f5f27e4f0f7f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -196,3 +196,79 @@ restart_ih: return IRQ_HANDLED; } + +/** + * amdgpu_ih_add_fault - Add a page fault record + * + * @adev: amdgpu device pointer + * @key: 64-bit encoding of PASID and address + * + * This should be called when a retry page fault interrupt is + * received. If this is a new page fault, it will be added to a hash + * table. The return value indicates whether this is a new fault, or + * a fault that was already known and is already being handled. + * + * If there are too many pending page faults, this will fail. Retry + * interrupts should be ignored in this case until there is enough + * free space. + * + * Returns 0 if the fault was added, 1 if the fault was already known, + * -ENOSPC if there are too many pending faults. + */ +int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key) +{ + unsigned long flags; + int r = -ENOSPC; + + if (WARN_ON_ONCE(!adev->irq.ih.faults)) + /* Should be allocated in _ih_sw_init on GPUs that + * support retry faults and require retry filtering. + */ + return r; + + spin_lock_irqsave(&adev->irq.ih.faults->lock, flags); + + /* Only let the hash table fill up to 50% for best performance */ + if (adev->irq.ih.faults->count >= (1 << (AMDGPU_PAGEFAULT_HASH_BITS-1))) + goto unlock_out; + + r = chash_table_copy_in(&adev->irq.ih.faults->hash, key, NULL); + if (!r) + adev->irq.ih.faults->count++; + + /* chash_table_copy_in should never fail unless we're losing count */ + WARN_ON_ONCE(r < 0); + +unlock_out: + spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags); + return r; +} + +/** + * amdgpu_ih_clear_fault - Remove a page fault record + * + * @adev: amdgpu device pointer + * @key: 64-bit encoding of PASID and address + * + * This should be called when a page fault has been handled. Any + * future interrupt with this key will be processed as a new + * page fault. + */ +void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key) +{ + unsigned long flags; + int r; + + if (!adev->irq.ih.faults) + return; + + spin_lock_irqsave(&adev->irq.ih.faults->lock, flags); + + r = chash_table_remove(&adev->irq.ih.faults->hash, key, NULL); + if (!WARN_ON_ONCE(r < 0)) { + adev->irq.ih.faults->count--; + WARN_ON_ONCE(adev->irq.ih.faults->count < 0); + } + + spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h index 3de8e74e5b3a..ada89358e220 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h @@ -24,6 +24,8 @@ #ifndef __AMDGPU_IH_H__ #define __AMDGPU_IH_H__ +#include + struct amdgpu_device; /* * vega10+ IH clients @@ -69,6 +71,13 @@ enum amdgpu_ih_clientid #define AMDGPU_IH_CLIENTID_LEGACY 0 +#define AMDGPU_PAGEFAULT_HASH_BITS 8 +struct amdgpu_retryfault_hashtable { + DECLARE_CHASH_TABLE(hash, AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); + spinlock_t lock; + int count; +}; + /* * R6xx+ IH ring */ @@ -87,6 +96,7 @@ struct amdgpu_ih_ring { bool use_doorbell; bool use_bus_addr; dma_addr_t rb_dma_addr; /* only used when use_bus_addr = true */ + struct amdgpu_retryfault_hashtable *faults; }; #define AMDGPU_IH_SRC_DATA_MAX_SIZE_DW 4 @@ -109,5 +119,7 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size, bool use_bus_addr); void amdgpu_ih_ring_fini(struct amdgpu_device *adev); int amdgpu_ih_process(struct amdgpu_device *adev); +int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key); +void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 9b795915cab1..6c1133298b17 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -2680,6 +2680,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->pasid = pasid; } + INIT_KFIFO(vm->faults); + return 0; error_free_root: @@ -2731,8 +2733,13 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdgpu_bo_va_mapping *mapping, *tmp; bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; + u64 fault; int i; + /* Clear pending page faults from IH when the VM is destroyed */ + while (kfifo_get(&vm->faults, &fault)) + amdgpu_ih_clear_fault(adev, fault); + if (vm->pasid) { unsigned long flags; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 7873dfa8c0f9..447ed6e7e586 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -120,6 +120,10 @@ struct amdgpu_vm_pt { unsigned last_entry_used; }; +#define AMDGPU_VM_FAULT(pasid, addr) (((u64)(pasid) << 48) | (addr)) +#define AMDGPU_VM_FAULT_PASID(fault) ((u64)(fault) >> 48) +#define AMDGPU_VM_FAULT_ADDR(fault) ((u64)(fault) & 0xfffffffff000ULL) + struct amdgpu_vm { /* tree of virtual addresses mapped */ struct rb_root va; @@ -160,6 +164,9 @@ struct amdgpu_vm { /* Flag to indicate ATS support from PTE for GFX9 */ bool pte_support_ats; + + /* Up to 128 pending page faults */ + DECLARE_KFIFO(faults, u64, 128); }; struct amdgpu_vm_id { diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index eda4771e273f..dd6af2176d3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -235,8 +235,73 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev) */ static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) { - /* TODO: Filter known pending page faults */ + u32 ring_index = adev->irq.ih.rptr >> 2; + u32 dw0, dw3, dw4, dw5; + u16 pasid; + u64 addr, key; + struct amdgpu_vm *vm; + int r; + + dw0 = le32_to_cpu(adev->irq.ih.ring[ring_index + 0]); + dw3 = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); + dw4 = le32_to_cpu(adev->irq.ih.ring[ring_index + 4]); + dw5 = le32_to_cpu(adev->irq.ih.ring[ring_index + 5]); + + /* Filter retry page faults, let only the first one pass. If + * there are too many outstanding faults, ignore them until + * some faults get cleared. + */ + switch (dw0 & 0xff) { + case AMDGPU_IH_CLIENTID_VMC: + case AMDGPU_IH_CLIENTID_UTCL2: + break; + default: + /* Not a VM fault */ + return true; + } + + /* Not a retry fault */ + if (!(dw5 & 0x80)) + return true; + + pasid = dw3 & 0xffff; + /* No PASID, can't identify faulting process */ + if (!pasid) + return true; + + addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); + key = AMDGPU_VM_FAULT(pasid, addr); + r = amdgpu_ih_add_fault(adev, key); + + /* Hash table is full or the fault is already being processed, + * ignore further page faults + */ + if (r != 0) + goto ignore_iv; + + /* Track retry faults in per-VM fault FIFO. */ + spin_lock(&adev->vm_manager.pasid_lock); + vm = idr_find(&adev->vm_manager.pasid_idr, pasid); + spin_unlock(&adev->vm_manager.pasid_lock); + if (WARN_ON_ONCE(!vm)) { + /* VM not found, process it normally */ + amdgpu_ih_clear_fault(adev, key); + return true; + } + /* No locking required with single writer and single reader */ + r = kfifo_put(&vm->faults, key); + if (!r) { + /* FIFO is full. Ignore it until there is space */ + amdgpu_ih_clear_fault(adev, key); + goto ignore_iv; + } + + /* It's the first fault for this address, process it normally */ return true; + +ignore_iv: + adev->irq.ih.rptr += 32; + return false; } /** @@ -323,6 +388,14 @@ static int vega10_ih_sw_init(void *handle) adev->irq.ih.use_doorbell = true; adev->irq.ih.doorbell_index = AMDGPU_DOORBELL64_IH << 1; + adev->irq.ih.faults = kmalloc(sizeof(*adev->irq.ih.faults), GFP_KERNEL); + if (!adev->irq.ih.faults) + return -ENOMEM; + INIT_CHASH_TABLE(adev->irq.ih.faults->hash, + AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); + spin_lock_init(&adev->irq.ih.faults->lock); + adev->irq.ih.faults->count = 0; + r = amdgpu_irq_init(adev); return r; @@ -335,6 +408,9 @@ static int vega10_ih_sw_fini(void *handle) amdgpu_irq_fini(adev); amdgpu_ih_ring_fini(adev); + kfree(adev->irq.ih.faults); + adev->irq.ih.faults = NULL; + return 0; } -- cgit From e33dac39bcf9978061df95f1510b2ec6bfab07fc Mon Sep 17 00:00:00 2001 From: "Xiangliang.Yu" Date: Tue, 12 Sep 2017 17:31:46 +0800 Subject: drm/amdgpu/sdma3: Enable sdma wptr polling for SRIOV When hypervisor triggering FLR for one of VFs, need to enable sdma wptr polling to avoid missing wptr update if enabling doorbell. Signed-off-by: Xiangliang.Yu Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index b1de44f22824..b7d69ab9a94a 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -641,10 +641,11 @@ static void sdma_v3_0_enable(struct amdgpu_device *adev, bool enable) static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - u32 rb_cntl, ib_cntl; + u32 rb_cntl, ib_cntl, wptr_poll_cntl; u32 rb_bufsz; u32 wb_offset; u32 doorbell; + u64 wptr_gpu_addr; int i, j, r; for (i = 0; i < adev->sdma.num_instances; i++) { @@ -707,6 +708,20 @@ static int sdma_v3_0_gfx_resume(struct amdgpu_device *adev) } WREG32(mmSDMA0_GFX_DOORBELL + sdma_offsets[i], doorbell); + /* setup the wptr shadow polling */ + wptr_gpu_addr = adev->wb.gpu_addr + (ring->wptr_offs * 4); + + WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_LO + sdma_offsets[i], + lower_32_bits(wptr_gpu_addr)); + WREG32(mmSDMA0_GFX_RB_WPTR_POLL_ADDR_HI + sdma_offsets[i], + upper_32_bits(wptr_gpu_addr)); + wptr_poll_cntl = RREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i]); + if (amdgpu_sriov_vf(adev)) + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 1); + else + wptr_poll_cntl = REG_SET_FIELD(wptr_poll_cntl, SDMA0_GFX_RB_WPTR_POLL_CNTL, F32_POLL_ENABLE, 0); + WREG32(mmSDMA0_GFX_RB_WPTR_POLL_CNTL + sdma_offsets[i], wptr_poll_cntl); + /* enable DMA RB */ rb_cntl = REG_SET_FIELD(rb_cntl, SDMA0_GFX_RB_CNTL, RB_ENABLE, 1); WREG32(mmSDMA0_GFX_RB_CNTL + sdma_offsets[i], rb_cntl); -- cgit From 3e4b0bd96092bbf9c5a538916a61faa5eca8e9cb Mon Sep 17 00:00:00 2001 From: "Xiangliang.Yu" Date: Wed, 13 Sep 2017 10:58:19 +0800 Subject: drm/amdgpu/sdma3: set wptr shadow atomically Port it from sdma4 for wptr polling usage. Signed-off-by: Xiangliang.Yu Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index b7d69ab9a94a..728c0d8e849b 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -379,8 +379,10 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; if (ring->use_doorbell) { + u32 *wb = (u32 *)&adev->wb.wb[ring->wptr_offs]; + /* XXX check if swapping is necessary on BE */ - adev->wb.wb[ring->wptr_offs] = lower_32_bits(ring->wptr) << 2; + WRITE_ONCE(*wb, (lower_32_bits(ring->wptr) << 2)); WDOORBELL32(ring->doorbell_index, lower_32_bits(ring->wptr) << 2); } else { int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; -- cgit From f6e8b15af7c96a429b320eb8414791666c0bd2b7 Mon Sep 17 00:00:00 2001 From: Leo Liu Date: Fri, 3 Mar 2017 11:54:37 -0500 Subject: drm/amdgpu: remove the clearance of vce 4.0 interrupt mask Requested by SRIOV, the clearance of the bit moved into firmware Signed-off-by: Leo Liu Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vce_v4_0.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index 11134d5f7443..75745544600a 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -1011,10 +1011,6 @@ static int vce_v4_0_process_interrupt(struct amdgpu_device *adev, { DRM_DEBUG("IH: VCE\n"); - WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_SYS_INT_STATUS), - VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK, - ~VCE_SYS_INT_STATUS__VCE_SYS_INT_TRAP_INTERRUPT_INT_MASK); - switch (entry->src_data[0]) { case 0: case 1: -- cgit From 0b693f0b5612594d089e669804b1b6a62639779f Mon Sep 17 00:00:00 2001 From: Rex Zhu Date: Tue, 19 Sep 2017 14:36:08 +0800 Subject: drm/amdgpu: fix checkpatch.pl warning to amdgpu_drv.c MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit fix checkpatch.pl WARNING: Prefer 'unsigned int' to bare use of 'unsigned' Reviewed-by: Christian König Signed-off-by: Rex Zhu Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 14 +++++++------- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 14 +++++++------- 2 files changed, 14 insertions(+), 14 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index d62a35ef26be..ac026f5f56fb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -92,7 +92,7 @@ extern int amdgpu_dpm; extern int amdgpu_fw_load_type; extern int amdgpu_aspm; extern int amdgpu_runtime_pm; -extern unsigned amdgpu_ip_block_mask; +extern uint amdgpu_ip_block_mask; extern int amdgpu_bapm; extern int amdgpu_deep_color; extern int amdgpu_vm_size; @@ -105,14 +105,14 @@ extern int amdgpu_sched_jobs; extern int amdgpu_sched_hw_submission; extern int amdgpu_no_evict; extern int amdgpu_direct_gma_size; -extern unsigned amdgpu_pcie_gen_cap; -extern unsigned amdgpu_pcie_lane_cap; -extern unsigned amdgpu_cg_mask; -extern unsigned amdgpu_pg_mask; -extern unsigned amdgpu_sdma_phase_quantum; +extern uint amdgpu_pcie_gen_cap; +extern uint amdgpu_pcie_lane_cap; +extern uint amdgpu_cg_mask; +extern uint amdgpu_pg_mask; +extern uint amdgpu_sdma_phase_quantum; extern char *amdgpu_disable_cu; extern char *amdgpu_virtual_display; -extern unsigned amdgpu_pp_feature_mask; +extern uint amdgpu_pp_feature_mask; extern int amdgpu_vram_page_split; extern int amdgpu_ngg; extern int amdgpu_prim_buf_per_se; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 792b11795a81..dee35a9c5723 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -92,7 +92,7 @@ int amdgpu_dpm = -1; int amdgpu_fw_load_type = -1; int amdgpu_aspm = -1; int amdgpu_runtime_pm = -1; -unsigned amdgpu_ip_block_mask = 0xffffffff; +uint amdgpu_ip_block_mask = 0xffffffff; int amdgpu_bapm = -1; int amdgpu_deep_color = 0; int amdgpu_vm_size = -1; @@ -107,14 +107,14 @@ int amdgpu_sched_jobs = 32; int amdgpu_sched_hw_submission = 2; int amdgpu_no_evict = 0; int amdgpu_direct_gma_size = 0; -unsigned amdgpu_pcie_gen_cap = 0; -unsigned amdgpu_pcie_lane_cap = 0; -unsigned amdgpu_cg_mask = 0xffffffff; -unsigned amdgpu_pg_mask = 0xffffffff; -unsigned amdgpu_sdma_phase_quantum = 32; +uint amdgpu_pcie_gen_cap = 0; +uint amdgpu_pcie_lane_cap = 0; +uint amdgpu_cg_mask = 0xffffffff; +uint amdgpu_pg_mask = 0xffffffff; +uint amdgpu_sdma_phase_quantum = 32; char *amdgpu_disable_cu = NULL; char *amdgpu_virtual_display = NULL; -unsigned amdgpu_pp_feature_mask = 0xffffffff; +uint amdgpu_pp_feature_mask = 0xffffffff; int amdgpu_ngg = 0; int amdgpu_prim_buf_per_se = 0; int amdgpu_pos_buf_per_se = 0; -- cgit From a40cfa0bef9366ad156717078dae681305099a15 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Mon, 18 Sep 2017 07:14:56 -0400 Subject: drm/amd/amdgpu: Fold TTM debugfs entries into array (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Christian König Signed-off-by: Alex Deucher (v2): add domains and avoid strcmp --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 54 ++++++++++++++++++--------------- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h | 4 +-- 2 files changed, 32 insertions(+), 26 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 8ee16dfdb8af..50d20903de4f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1809,6 +1809,19 @@ static const struct file_operations amdgpu_ttm_gtt_fops = { #endif + + +static const struct { + char *name; + const struct file_operations *fops; + int domain; +} ttm_debugfs_entries[] = { + { "amdgpu_vram", &amdgpu_ttm_vram_fops, TTM_PL_VRAM }, +#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS + { "amdgpu_gtt", &amdgpu_ttm_gtt_fops, TTM_PL_TT }, +#endif +}; + #endif static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) @@ -1819,22 +1832,21 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) struct drm_minor *minor = adev->ddev->primary; struct dentry *ent, *root = minor->debugfs_root; - ent = debugfs_create_file("amdgpu_vram", S_IFREG | S_IRUGO, root, - adev, &amdgpu_ttm_vram_fops); - if (IS_ERR(ent)) - return PTR_ERR(ent); - i_size_write(ent->d_inode, adev->mc.mc_vram_size); - adev->mman.vram = ent; - -#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS - ent = debugfs_create_file("amdgpu_gtt", S_IFREG | S_IRUGO, root, - adev, &amdgpu_ttm_gtt_fops); - if (IS_ERR(ent)) - return PTR_ERR(ent); - i_size_write(ent->d_inode, adev->mc.gart_size); - adev->mman.gtt = ent; + for (count = 0; count < ARRAY_SIZE(ttm_debugfs_entries); count++) { + ent = debugfs_create_file( + ttm_debugfs_entries[count].name, + S_IFREG | S_IRUGO, root, + adev, + ttm_debugfs_entries[count].fops); + if (IS_ERR(ent)) + return PTR_ERR(ent); + if (ttm_debugfs_entries[count].domain == TTM_PL_VRAM) + i_size_write(ent->d_inode, adev->mc.mc_vram_size); + else if (ttm_debugfs_entries[count].domain == TTM_PL_TT) + i_size_write(ent->d_inode, adev->mc.gart_size); + adev->mman.debugfs_entries[count] = ent; + } -#endif count = ARRAY_SIZE(amdgpu_ttm_debugfs_list); #ifdef CONFIG_SWIOTLB @@ -1844,7 +1856,6 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) return amdgpu_debugfs_add_files(adev, amdgpu_ttm_debugfs_list, count); #else - return 0; #endif } @@ -1852,14 +1863,9 @@ static int amdgpu_ttm_debugfs_init(struct amdgpu_device *adev) static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev) { #if defined(CONFIG_DEBUG_FS) + unsigned i; - debugfs_remove(adev->mman.vram); - adev->mman.vram = NULL; - -#ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS - debugfs_remove(adev->mman.gtt); - adev->mman.gtt = NULL; -#endif - + for (i = 0; i < ARRAY_SIZE(ttm_debugfs_entries); i++) + debugfs_remove(adev->mman.debugfs_entries[i]); #endif } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 64709e041d5b..7abae6867339 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -24,6 +24,7 @@ #ifndef __AMDGPU_TTM_H__ #define __AMDGPU_TTM_H__ +#include "amdgpu.h" #include "gpu_scheduler.h" #define AMDGPU_PL_GDS (TTM_PL_PRIV + 0) @@ -45,8 +46,7 @@ struct amdgpu_mman { bool initialized; #if defined(CONFIG_DEBUG_FS) - struct dentry *vram; - struct dentry *gtt; + struct dentry *debugfs_entries[8]; #endif /* buffer handling */ -- cgit From 38290b2c456ade70db7cedf489332af5e4263fb0 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Mon, 18 Sep 2017 07:28:14 -0400 Subject: drm/amd/amdgpu: add support for iova_to_phys to replace TTM trace (v5) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Christian König Signed-off-by: Alex Deucher (v2): Add domain to iova debugfs (v3): Add true read/write methods to access system memory of pages mapped to the device (v4): Move get_domain call out of loop and return on error (v5): Just use kmap/kunmap --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 99 +++++++++++++++++++++++++++++++++ 1 file changed, 99 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 50d20903de4f..c7f8e081a772 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -43,6 +43,7 @@ #include #include #include +#include #include "amdgpu.h" #include "amdgpu_trace.h" #include "bif/bif_4_1_d.h" @@ -1809,7 +1810,104 @@ static const struct file_operations amdgpu_ttm_gtt_fops = { #endif +static ssize_t amdgpu_iova_to_phys_read(struct file *f, char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result, n; + int r; + uint64_t phys; + void *ptr; + struct iommu_domain *dom; + + dom = iommu_get_domain_for_dev(adev->dev); + if (!dom) + return -EFAULT; + + result = 0; + while (size) { + // get physical address and map + phys = iommu_iova_to_phys(dom, *pos); + + // copy upto one page + if (size > PAGE_SIZE) + n = PAGE_SIZE; + else + n = size; + + // to end of the page + if (((*pos & (PAGE_SIZE - 1)) + n) >= PAGE_SIZE) + n = PAGE_SIZE - (*pos & (PAGE_SIZE - 1)); + + ptr = kmap(pfn_to_page(PFN_DOWN(phys))); + if (!ptr) + return -EFAULT; + + r = copy_to_user(buf, ptr, n); + kunmap(pfn_to_page(PFN_DOWN(phys))); + if (r) + return -EFAULT; + + *pos += n; + size -= n; + result += n; + } + + return result; +} + +static ssize_t amdgpu_iova_to_phys_write(struct file *f, const char __user *buf, + size_t size, loff_t *pos) +{ + struct amdgpu_device *adev = file_inode(f)->i_private; + ssize_t result, n; + int r; + uint64_t phys; + void *ptr; + struct iommu_domain *dom; + + dom = iommu_get_domain_for_dev(adev->dev); + if (!dom) + return -EFAULT; + + result = 0; + while (size) { + // get physical address and map + phys = iommu_iova_to_phys(dom, *pos); + // copy upto one page + if (size > PAGE_SIZE) + n = PAGE_SIZE; + else + n = size; + + // to end of the page + if (((*pos & (PAGE_SIZE - 1)) + n) >= PAGE_SIZE) + n = PAGE_SIZE - (*pos & (PAGE_SIZE - 1)); + + ptr = kmap(pfn_to_page(PFN_DOWN(phys))); + if (!ptr) + return -EFAULT; + + r = copy_from_user(ptr, buf, n); + kunmap(pfn_to_page(PFN_DOWN(phys))); + if (r) + return -EFAULT; + + *pos += n; + size -= n; + result += n; + } + + return result; +} + +static const struct file_operations amdgpu_ttm_iova_fops = { + .owner = THIS_MODULE, + .read = amdgpu_iova_to_phys_read, + .write = amdgpu_iova_to_phys_write, + .llseek = default_llseek +}; static const struct { char *name; @@ -1820,6 +1918,7 @@ static const struct { #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS { "amdgpu_gtt", &amdgpu_ttm_gtt_fops, TTM_PL_TT }, #endif + { "amdgpu_iova", &amdgpu_ttm_iova_fops, TTM_PL_SYSTEM }, }; #endif -- cgit From 79ba2800662bcfaef41269dd9389d0e9ed729a28 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Mon, 18 Sep 2017 08:10:00 -0400 Subject: drm/amd/amdgpu: remove usage of ttm trace MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Tom St Denis Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 36 +++------------------------------ 1 file changed, 3 insertions(+), 33 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index c7f8e081a772..0e5f78f3a97e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -34,7 +34,6 @@ #include #include #include -#include #include #include #include @@ -704,22 +703,6 @@ void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) } } -static void amdgpu_trace_dma_map(struct ttm_tt *ttm) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); - struct amdgpu_ttm_tt *gtt = (void *)ttm; - - ttm_trace_dma_map(adev->dev, >t->ttm); -} - -static void amdgpu_trace_dma_unmap(struct ttm_tt *ttm) -{ - struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); - struct amdgpu_ttm_tt *gtt = (void *)ttm; - - ttm_trace_dma_unmap(adev->dev, >t->ttm); -} - /* prepare the sg table with the user pages */ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) { @@ -746,8 +729,6 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages, gtt->ttm.dma_address, ttm->num_pages); - amdgpu_trace_dma_map(ttm); - return 0; release_sg: @@ -773,8 +754,6 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) amdgpu_ttm_tt_mark_user_pages(ttm); - amdgpu_trace_dma_unmap(ttm); - sg_free_table(ttm->sg); } @@ -958,7 +937,6 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; - int r; bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); if (ttm->state != tt_unpopulated) @@ -978,22 +956,16 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm) drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages, gtt->ttm.dma_address, ttm->num_pages); ttm->state = tt_unbound; - r = 0; - goto trace_mappings; + return 0; } #ifdef CONFIG_SWIOTLB if (swiotlb_nr_tbl()) { - r = ttm_dma_populate(>t->ttm, adev->dev); - goto trace_mappings; + return ttm_dma_populate(>t->ttm, adev->dev); } #endif - r = ttm_populate_and_map_pages(adev->dev, >t->ttm); -trace_mappings: - if (likely(!r)) - amdgpu_trace_dma_map(ttm); - return r; + return ttm_populate_and_map_pages(adev->dev, >t->ttm); } static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) @@ -1014,8 +986,6 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) adev = amdgpu_ttm_adev(ttm->bdev); - amdgpu_trace_dma_unmap(ttm); - #ifdef CONFIG_SWIOTLB if (swiotlb_nr_tbl()) { ttm_dma_unpopulate(>t->ttm, adev->dev); -- cgit From 98512bb8c241f67f3a5cf0a121624b28a852db45 Mon Sep 17 00:00:00 2001 From: Ken Wang Date: Thu, 14 Sep 2017 16:25:19 +0800 Subject: drm/amdgpu: Add GPU reset functionality for Vega10 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit V2 Signed-off-by: Ken Wang Reviewed-by: Alex Deucher Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 3 ++- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 21 +++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h | 3 +++ drivers/gpu/drm/amd/amdgpu/psp_v10_0.c | 7 ++++++ drivers/gpu/drm/amd/amdgpu/psp_v10_0.h | 2 ++ drivers/gpu/drm/amd/amdgpu/psp_v3_1.c | 34 ++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/psp_v3_1.h | 1 + drivers/gpu/drm/amd/amdgpu/soc15.c | 27 ++++++++++++------------ 8 files changed, 83 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index b0109ebe0a1b..36979e1d7b5f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2581,7 +2581,8 @@ static bool amdgpu_need_full_reset(struct amdgpu_device *adev) if ((adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) || (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) || (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_ACP) || - (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE)) { + (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) || + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) { if (adev->ip_blocks[i].status.hang) { DRM_INFO("Some block need full reset!\n"); return true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 84bd6ed7a641..a5b595a75d91 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -62,6 +62,7 @@ static int psp_sw_init(void *handle) psp->cmd_submit = psp_v3_1_cmd_submit; psp->compare_sram_data = psp_v3_1_compare_sram_data; psp->smu_reload_quirk = psp_v3_1_smu_reload_quirk; + psp->mode1_reset = psp_v3_1_mode1_reset; break; case CHIP_RAVEN: psp->init_microcode = psp_v10_0_init_microcode; @@ -72,6 +73,7 @@ static int psp_sw_init(void *handle) psp->ring_destroy = psp_v10_0_ring_destroy; psp->cmd_submit = psp_v10_0_cmd_submit; psp->compare_sram_data = psp_v10_0_compare_sram_data; + psp->mode1_reset = psp_v10_0_mode1_reset; break; default: return -EINVAL; @@ -497,6 +499,22 @@ failed: return ret; } +static bool psp_check_reset(void* handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (adev->flags & AMD_IS_APU) + return true; + + return false; +} + +static int psp_reset(void* handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + return psp_mode1_reset(&adev->psp); +} + static bool psp_check_fw_loading_status(struct amdgpu_device *adev, enum AMDGPU_UCODE_ID ucode_type) { @@ -540,8 +558,9 @@ const struct amd_ip_funcs psp_ip_funcs = { .suspend = psp_suspend, .resume = psp_resume, .is_idle = NULL, + .check_soft_reset = psp_check_reset, .wait_for_idle = NULL, - .soft_reset = NULL, + .soft_reset = psp_reset, .set_clockgating_state = psp_set_clockgating_state, .set_powergating_state = psp_set_powergating_state, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 1b7d12d88720..ce4654550416 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -76,6 +76,7 @@ struct psp_context struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); bool (*smu_reload_quirk)(struct psp_context *psp); + int (*mode1_reset)(struct psp_context *psp); /* fence buffer */ struct amdgpu_bo *fw_pri_bo; @@ -139,6 +140,8 @@ struct amdgpu_psp_funcs { ((psp)->bootloader_load_sos ? (psp)->bootloader_load_sos((psp)) : 0) #define psp_smu_reload_quirk(psp) \ ((psp)->smu_reload_quirk ? (psp)->smu_reload_quirk((psp)) : false) +#define psp_mode1_reset(psp) \ + ((psp)->mode1_reset ? (psp)->mode1_reset((psp)) : false) extern const struct amd_ip_funcs psp_ip_funcs; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index 77cab1ff0254..dea7c909ca5f 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -407,3 +407,10 @@ bool psp_v10_0_compare_sram_data(struct psp_context *psp, return true; } + + +int psp_v10_0_mode1_reset(struct psp_context *psp) +{ + DRM_INFO("psp mode 1 reset not supported now! \n"); + return -EINVAL; +} diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h index 3af3ad1320ff..451e8308303f 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.h @@ -45,4 +45,6 @@ extern int psp_v10_0_cmd_submit(struct psp_context *psp, extern bool psp_v10_0_compare_sram_data(struct psp_context *psp, struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); + +extern int psp_v10_0_mode1_reset(struct psp_context *psp); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index bcbe30dfff39..cee5c396b277 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -530,3 +530,37 @@ bool psp_v3_1_smu_reload_quirk(struct psp_context *psp) reg = RREG32_SOC15(NBIO, 0, mmPCIE_DATA2); return (reg & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) ? true : false; } + +int psp_v3_1_mode1_reset(struct psp_context *psp) +{ + int ret; + uint32_t offset; + struct amdgpu_device *adev = psp->adev; + + offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64); + + ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false); + + if (ret) { + DRM_INFO("psp is not working correctly before mode1 reset!\n"); + return -EINVAL; + } + + /*send the mode 1 reset command*/ + WREG32(offset, 0x70000); + + mdelay(1000); + + offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); + + ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false); + + if (ret) { + DRM_INFO("psp mode 1 reset failed!\n"); + return -EINVAL; + } + + DRM_INFO("psp mode1 reset succeed \n"); + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h index 5af2231b7099..b05dbada7751 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.h @@ -53,4 +53,5 @@ extern bool psp_v3_1_compare_sram_data(struct psp_context *psp, struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); extern bool psp_v3_1_smu_reload_quirk(struct psp_context *psp); +extern int psp_v3_1_mode1_reset(struct psp_context *psp); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index f2c3a49f73a0..245a18aeb389 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -407,18 +407,27 @@ static int soc15_read_register(struct amdgpu_device *adev, u32 se_num, return -EINVAL; } -static void soc15_gpu_pci_config_reset(struct amdgpu_device *adev) +static int soc15_asic_reset(struct amdgpu_device *adev) { u32 i; - dev_info(adev->dev, "GPU pci config reset\n"); + amdgpu_atombios_scratch_regs_engine_hung(adev, true); + + dev_info(adev->dev, "GPU reset\n"); /* disable BM */ pci_clear_master(adev->pdev); - /* reset */ - amdgpu_pci_config_reset(adev); - udelay(100); + pci_save_state(adev->pdev); + + for (i = 0; i < AMDGPU_MAX_IP_NUM; i++) { + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP){ + adev->ip_blocks[i].version->funcs->soft_reset((void *)adev); + break; + } + } + + pci_restore_state(adev->pdev); /* wait for asic to come out of reset */ for (i = 0; i < adev->usec_timeout; i++) { @@ -430,14 +439,6 @@ static void soc15_gpu_pci_config_reset(struct amdgpu_device *adev) udelay(1); } -} - -static int soc15_asic_reset(struct amdgpu_device *adev) -{ - amdgpu_atombios_scratch_regs_engine_hung(adev, true); - - soc15_gpu_pci_config_reset(adev); - amdgpu_atombios_scratch_regs_engine_hung(adev, false); return 0; -- cgit From ab5d6227b78be29acf0340da2673f662c0df9b2d Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Tue, 12 Sep 2017 14:33:29 +0800 Subject: drm/amdgpu/sriov:fix missing error handling MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 79d9ab43d42c..4fd06f8d9768 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -841,8 +841,11 @@ int amdgpu_driver_open_kms(struct drm_device *dev, struct drm_file *file_priv) if (amdgpu_sriov_vf(adev)) { r = amdgpu_map_static_csa(adev, &fpriv->vm, &fpriv->csa_va); - if (r) + if (r) { + amdgpu_vm_fini(adev, &fpriv->vm); + kfree(fpriv); goto out_suspend; + } } mutex_init(&fpriv->bo_list_lock); -- cgit From 7c3f2167b4b66a0994a643c8a3bcf01ec3433b8b Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Thu, 14 Sep 2017 19:45:33 +0800 Subject: drm/amdgpu:no kiq in IH MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/vega10_ih.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index dd6af2176d3e..a3b30d84dbb3 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -219,9 +219,9 @@ static u32 vega10_ih_get_wptr(struct amdgpu_device *adev) wptr, adev->irq.ih.rptr, tmp); adev->irq.ih.rptr = tmp; - tmp = RREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL)); + tmp = RREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL)); tmp = REG_SET_FIELD(tmp, IH_RB_CNTL, WPTR_OVERFLOW_CLEAR, 1); - WREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp); + WREG32_NO_KIQ(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_CNTL), tmp); } return (wptr & adev->irq.ih.ptr_mask); } -- cgit From 3224a12b90f6bee789d2051b18a8249a82ad92d4 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 18:57:12 +0800 Subject: drm/amdgpu/sriov:move in_reset to adev and rename MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit currently in_reset is only used in sriov gpu reset, and it will be used for other non-gfx hw component later, like PSP, so move it from gfx to adev and rename to in_sriov_reset make more sense. Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 2 +- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 ++-- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 6 +++--- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 6 +++--- 4 files changed, 9 insertions(+), 9 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index ac026f5f56fb..33e1d619d060 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1018,7 +1018,6 @@ struct amdgpu_gfx { /* reset mask */ uint32_t grbm_soft_reset; uint32_t srbm_soft_reset; - bool in_reset; /* s3/s4 mask */ bool in_suspend; /* NGG */ @@ -1583,6 +1582,7 @@ struct amdgpu_device { /* record last mm index being written through WREG32*/ unsigned long last_mm_index; + bool in_sriov_reset; }; static inline struct amdgpu_device *amdgpu_ttm_adev(struct ttm_bo_device *bdev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 36979e1d7b5f..506a26793ce6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2690,7 +2690,7 @@ int amdgpu_sriov_gpu_reset(struct amdgpu_device *adev, struct amdgpu_job *job) mutex_lock(&adev->virt.lock_reset); atomic_inc(&adev->gpu_reset_counter); - adev->gfx.in_reset = true; + adev->in_sriov_reset = true; /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); @@ -2801,7 +2801,7 @@ give_up_reset: dev_info(adev->dev, "GPU reset successed!\n"); } - adev->gfx.in_reset = false; + adev->in_sriov_reset = false; mutex_unlock(&adev->virt.lock_reset); return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 666a1545e949..fe603cc2c4a9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -4811,7 +4811,7 @@ static int gfx_v8_0_kiq_init_queue(struct amdgpu_ring *ring) gfx_v8_0_kiq_setting(ring); - if (adev->gfx.in_reset) { /* for GPU_RESET case */ + if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation)); @@ -4848,7 +4848,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) struct vi_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { + if (!adev->in_sriov_reset && !adev->gfx.in_suspend) { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; @@ -4860,7 +4860,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct vi_mqd_allocation)); - } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ + } else if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct vi_mqd_allocation)); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 72dbf890c65b..ecb9674bc803 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2698,7 +2698,7 @@ static int gfx_v9_0_kiq_init_queue(struct amdgpu_ring *ring) gfx_v9_0_kiq_setting(ring); - if (adev->gfx.in_reset) { /* for GPU_RESET case */ + if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); @@ -2736,7 +2736,7 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) struct v9_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->gfx.in_reset && !adev->gfx.in_suspend) { + if (!adev->in_sriov_reset && !adev->gfx.in_suspend) { memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; @@ -2748,7 +2748,7 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(adev->gfx.mec.mqd_backup[mqd_idx], mqd, sizeof(struct v9_mqd_allocation)); - } else if (adev->gfx.in_reset) { /* for GPU_RESET case */ + } else if (adev->in_sriov_reset) { /* for GPU_RESET case */ /* reset MQD to a clean status */ if (adev->gfx.mec.mqd_backup[mqd_idx]) memcpy(mqd, adev->gfx.mec.mqd_backup[mqd_idx], sizeof(struct v9_mqd_allocation)); -- cgit From 55981bd2e88a0b6e572a3997098886e2f9b2121e Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 18:42:12 +0800 Subject: drm/amdgpu/sriov:don't load psp fw during gpu reset MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit At least for SRIOV we found reload PSP fw during gpu reset cause PSP hang. Signed-off-by: Monk Liu Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index a5b595a75d91..e02828665b14 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -255,15 +255,18 @@ static int psp_asd_load(struct psp_context *psp) static int psp_hw_start(struct psp_context *psp) { + struct amdgpu_device *adev = psp->adev; int ret; - ret = psp_bootloader_load_sysdrv(psp); - if (ret) - return ret; + if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) { + ret = psp_bootloader_load_sysdrv(psp); + if (ret) + return ret; - ret = psp_bootloader_load_sos(psp); - if (ret) - return ret; + ret = psp_bootloader_load_sos(psp); + if (ret) + return ret; + } ret = psp_ring_create(psp, PSP_RING_TYPE__KM); if (ret) -- cgit From 2ea6ab2741fc4caf9fd4a48de1b4946f09c365e0 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 9 Jun 2017 15:04:49 +0800 Subject: drm/amdgpu:insert TMZ_BEGIN MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit FRAME_CONTROL(begin) is needed for vega10 due to ucode logic change, it can fix some CTS random fail under gfx preemption enabled mode. Signed-off-by: Monk Liu Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index ecb9674bc803..0a8072f73a3b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3764,6 +3764,12 @@ static void gfx_v9_0_ring_emit_de_meta(struct amdgpu_ring *ring) amdgpu_ring_write_multiple(ring, (void *)&de_payload, sizeof(de_payload) >> 2); } +static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); + amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ +} + static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) { uint32_t dw2 = 0; @@ -3771,6 +3777,8 @@ static void gfx_v9_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) if (amdgpu_sriov_vf(ring->adev)) gfx_v9_0_ring_emit_ce_meta(ring); + gfx_v9_0_ring_emit_tmz(ring, true); + dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { /* set load_global_config & load_global_uconfig */ @@ -3821,12 +3829,6 @@ static void gfx_v9_0_ring_emit_patch_cond_exec(struct amdgpu_ring *ring, unsigne ring->ring[offset] = (ring->ring_size>>2) - offset + cur; } -static void gfx_v9_0_ring_emit_tmz(struct amdgpu_ring *ring, bool start) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_FRAME_CONTROL, 0)); - amdgpu_ring_write(ring, FRAME_CMD(start ? 0 : 1)); /* frame_end */ -} - static void gfx_v9_0_ring_emit_rreg(struct amdgpu_ring *ring, uint32_t reg) { struct amdgpu_device *adev = ring->adev; -- cgit From 1d4e0a8c4f12acbc0767f8f9fd75005b9125ada6 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 15:03:24 +0800 Subject: drm/amdgpu:hdp flush should be put it initialized MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Monk Liu Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 7ca9cbec3004..99147f576e76 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -696,12 +696,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) if (r) return r; - /* After HDP is initialized, flush HDP.*/ - if (adev->flags & AMD_IS_APU) - nbio_v7_0_hdp_flush(adev); - else - nbio_v6_1_hdp_flush(adev); - switch (adev->asic_type) { case CHIP_RAVEN: mmhub_v1_0_initialize_power_gating(adev); @@ -724,6 +718,12 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) tmp = RREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL); WREG32_SOC15(HDP, 0, mmHDP_HOST_PATH_CNTL, tmp); + /* After HDP is initialized, flush HDP.*/ + if (adev->flags & AMD_IS_APU) + nbio_v7_0_hdp_flush(adev); + else + nbio_v6_1_hdp_flush(adev); + if (amdgpu_vm_fault_stop == AMDGPU_VM_FAULT_STOP_ALWAYS) value = false; else -- cgit From 6e2e216fadd80b4280783bb78e543593ebf2cb69 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Tue, 4 Jul 2017 15:43:38 +0800 Subject: drm/amdgpu:use formal register to trigger hdp invalidate MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c | 2 +- 4 files changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 0a8072f73a3b..31891149e6fc 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -3535,7 +3535,7 @@ static void gfx_v9_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) static void gfx_v9_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { gfx_v9_0_write_data_to_reg(ring, 0, true, - SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 1); + SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); } static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index fd7c72aaafa6..d5f3848af26a 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -398,7 +398,7 @@ static void sdma_v4_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_SRBM_WRITE) | SDMA_PKT_SRBM_WRITE_HEADER_BYTE_EN(0xf)); - amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0)); + amdgpu_ring_write(ring, SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE)); amdgpu_ring_write(ring, 1); } diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 23a85750edd6..b8ed8faf2003 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -1161,7 +1161,7 @@ static void uvd_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) */ static void uvd_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0)); + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0)); amdgpu_ring_write(ring, 1); } diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 21e7b88401e1..1eb4d79d6e30 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -812,7 +812,7 @@ static void vcn_v1_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 */ static void vcn_v1_0_dec_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) { - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_DEBUG0), 0)); + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 0)); amdgpu_ring_write(ring, 1); } -- cgit From 030308fcbd944348e5be079f4c2d53d2bda00d53 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 15:34:52 +0800 Subject: drm/amdgpu/sriov:fix page fault issue of driver unload MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit bo_free on csa is too late to put in amdgpu_fini because that time ttm is already finished, Move it earlier to avoid the page fault. Signed-off-by: Monk Liu Signed-off-by: Horace Chen Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 4 +--- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 1 + drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 1 + 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 506a26793ce6..3e84ddf9e3b5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -1790,10 +1790,8 @@ static int amdgpu_fini(struct amdgpu_device *adev) adev->ip_blocks[i].status.late_initialized = false; } - if (amdgpu_sriov_vf(adev)) { - amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); + if (amdgpu_sriov_vf(adev)) amdgpu_virt_release_full_gpu(adev, false); - } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index fe603cc2c4a9..0c4a3b8e8596 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -2113,6 +2113,7 @@ static int gfx_v8_0_sw_fini(void *handle) amdgpu_gfx_compute_mqd_sw_fini(adev); amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq); amdgpu_gfx_kiq_fini(adev); + amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); gfx_v8_0_mec_fini(adev); gfx_v8_0_rlc_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 31891149e6fc..f1d7c5d0f385 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -1425,6 +1425,7 @@ static int gfx_v9_0_sw_fini(void *handle) amdgpu_gfx_compute_mqd_sw_fini(adev); amdgpu_gfx_kiq_free_ring(&adev->gfx.kiq.ring, &adev->gfx.kiq.irq); amdgpu_gfx_kiq_fini(adev); + amdgpu_bo_free_kernel(&adev->virt.csa_obj, &adev->virt.csa_vmid0_addr, NULL); gfx_v9_0_mec_fini(adev); gfx_v9_0_ngg_fini(adev); -- cgit From 6e132ca0bb62b30c0eb053d99f75bb827f7876f5 Mon Sep 17 00:00:00 2001 From: Horace Chen Date: Wed, 28 Jun 2017 17:51:50 +0800 Subject: drm/amdgpu/sriov:increate mailbox polling timeout MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit increase timeout to 12 seconds,because there may have multiple FLR waiting for done, the waiting time of events may be long, increase to 12s to reduce timeout failure. Signed-off-by: Horace Chen Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h | 2 +- drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h index 1e91b9a1c591..67e78576a9eb 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.h @@ -24,7 +24,7 @@ #ifndef __MXGPU_AI_H__ #define __MXGPU_AI_H__ -#define AI_MAILBOX_TIMEDOUT 5000 +#define AI_MAILBOX_TIMEDOUT 12000 enum idh_request { IDH_REQ_GPU_INIT_ACCESS = 1, diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h index c791d73d2d54..f13dc6cc158f 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.h @@ -23,7 +23,7 @@ #ifndef __MXGPU_VI_H__ #define __MXGPU_VI_H__ -#define VI_MAILBOX_TIMEDOUT 5000 +#define VI_MAILBOX_TIMEDOUT 12000 #define VI_MAILBOX_RESET_TIME 12 /* VI mailbox messages request */ -- cgit From f840cc5f8447db7efff447a25bcddbf084bd3e2e Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 16:58:08 +0800 Subject: drm/amdgpu/sriov:init csb for gfxv9 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit RLC need CSB registers initiated under SRIOV during world switch otherwise the clear state buffer behav will not be recovered to current VF scheme after switch back Signed-off-by: Monk Liu Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index f1d7c5d0f385..e2ae00df1d52 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2045,8 +2045,10 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) { int r; - if (amdgpu_sriov_vf(adev)) + if (amdgpu_sriov_vf(adev)) { + gfx_v9_0_init_csb(adev); return 0; + } gfx_v9_0_rlc_stop(adev); -- cgit From eb01abc7c4fd1faa26d0787f410894d9c704eb60 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 13:40:31 +0800 Subject: drm/amdgpu:make ctx_add_fence interruptible(v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit otherwise a gpu hang will make application couldn't be killed under timedout=0 mode v2: Fix memoryleak job/job->s_fence issue unlock mn remove the ERROR msg after waiting being interrupted Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 ++-- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 16 ++++++++++++++-- drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 12 +++++++----- 3 files changed, 23 insertions(+), 9 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 33e1d619d060..9cce59f6ada5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -735,8 +735,8 @@ struct amdgpu_ctx_mgr { struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id); int amdgpu_ctx_put(struct amdgpu_ctx *ctx); -uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, - struct dma_fence *fence); +int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, + struct dma_fence *fence, uint64_t *seq); struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, uint64_t seq); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 9f1202a4182f..c6a214f1e991 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1129,6 +1129,8 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, struct amd_sched_entity *entity = &p->ctx->rings[ring->idx].entity; struct amdgpu_job *job; unsigned i; + uint64_t seq; + int r; amdgpu_mn_lock(p->mn); @@ -1158,10 +1160,20 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, job->fence_ctx = entity->fence_context; p->fence = dma_fence_get(&job->base.s_fence->finished); + r = amdgpu_ctx_add_fence(p->ctx, ring, p->fence, &seq); + if (r) { + dma_fence_put(p->fence); + dma_fence_put(&job->base.s_fence->finished); + amdgpu_job_free(job); + amdgpu_mn_unlock(p->mn); + return r; + } + amdgpu_cs_post_dependencies(p); - cs->out.handle = amdgpu_ctx_add_fence(p->ctx, ring, p->fence); - job->uf_sequence = cs->out.handle; + cs->out.handle = seq; + job->uf_sequence = seq; + amdgpu_job_free_resources(job); trace_amdgpu_cs_ioctl(job); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index a11e44340b23..75c933b1a432 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -246,8 +246,8 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx) return 0; } -uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, - struct dma_fence *fence) +int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, + struct dma_fence *fence, uint64_t* handler) { struct amdgpu_ctx_ring *cring = & ctx->rings[ring->idx]; uint64_t seq = cring->sequence; @@ -258,9 +258,9 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, other = cring->fences[idx]; if (other) { signed long r; - r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT); + r = dma_fence_wait_timeout(other, true, MAX_SCHEDULE_TIMEOUT); if (r < 0) - DRM_ERROR("Error (%ld) waiting for fence!\n", r); + return r; } dma_fence_get(fence); @@ -271,8 +271,10 @@ uint64_t amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, spin_unlock(&ctx->ring_lock); dma_fence_put(other); + if (handler) + *handler = seq; - return seq; + return 0; } struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, -- cgit From d59c026b7be1dd9ae00b54d1916a90775fe3bdda Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 14:35:09 +0800 Subject: drm/amdgpu/sriov:fix memory leak after gpu reset MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit GPU reset will require all hw doing hw_init thus ucode_init_bo will be invoked again, which lead to memory leak skip the fw_buf allocation during sriov gpu reset to avoid memory leak. Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 3 ++ drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 64 +++++++++++++++---------------- 2 files changed, 35 insertions(+), 32 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 9cce59f6ada5..f085c8c9f267 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -1187,6 +1187,9 @@ struct amdgpu_firmware { /* gpu info firmware data pointer */ const struct firmware *gpu_info_fw; + + void *fw_buf_ptr; + uint64_t fw_buf_mc; }; /* diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index f306374ff654..65649026b836 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -360,8 +360,6 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode, int amdgpu_ucode_init_bo(struct amdgpu_device *adev) { struct amdgpu_bo **bo = &adev->firmware.fw_buf; - uint64_t fw_mc_addr; - void *fw_buf_ptr = NULL; uint64_t fw_offset = 0; int i, err; struct amdgpu_firmware_info *ucode = NULL; @@ -372,37 +370,39 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev) return 0; } - err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true, - amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - NULL, NULL, 0, bo); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err); - goto failed; - } + if (!amdgpu_sriov_vf(adev) || !adev->in_sriov_reset) { + err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true, + amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, + NULL, NULL, 0, bo); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err); + goto failed; + } - err = amdgpu_bo_reserve(*bo, false); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err); - goto failed_reserve; - } + err = amdgpu_bo_reserve(*bo, false); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer reserve failed\n", err); + goto failed_reserve; + } - err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, - &fw_mc_addr); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err); - goto failed_pin; - } + err = amdgpu_bo_pin(*bo, amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + &adev->firmware.fw_buf_mc); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer pin failed\n", err); + goto failed_pin; + } - err = amdgpu_bo_kmap(*bo, &fw_buf_ptr); - if (err) { - dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err); - goto failed_kmap; - } + err = amdgpu_bo_kmap(*bo, &adev->firmware.fw_buf_ptr); + if (err) { + dev_err(adev->dev, "(%d) Firmware buffer kmap failed\n", err); + goto failed_kmap; + } - amdgpu_bo_unreserve(*bo); + amdgpu_bo_unreserve(*bo); + } - memset(fw_buf_ptr, 0, adev->firmware.fw_size); + memset(adev->firmware.fw_buf_ptr, 0, adev->firmware.fw_size); /* * if SMU loaded firmware, it needn't add SMC, UVD, and VCE @@ -421,14 +421,14 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev) ucode = &adev->firmware.ucode[i]; if (ucode->fw) { header = (const struct common_firmware_header *)ucode->fw->data; - amdgpu_ucode_init_single_fw(adev, ucode, fw_mc_addr + fw_offset, - (void *)((uint8_t *)fw_buf_ptr + fw_offset)); + amdgpu_ucode_init_single_fw(adev, ucode, adev->firmware.fw_buf_mc + fw_offset, + adev->firmware.fw_buf_ptr + fw_offset); if (i == AMDGPU_UCODE_ID_CP_MEC1 && adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { const struct gfx_firmware_header_v1_0 *cp_hdr; cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data; - amdgpu_ucode_patch_jt(ucode, fw_mc_addr + fw_offset, - fw_buf_ptr + fw_offset); + amdgpu_ucode_patch_jt(ucode, adev->firmware.fw_buf_mc + fw_offset, + adev->firmware.fw_buf_ptr + fw_offset); fw_offset += ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE); } fw_offset += ALIGN(ucode->ucode_size, PAGE_SIZE); -- cgit From 6fe85429575c4d1f468e3b1715bd8643991570b1 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Wed, 20 Sep 2017 10:53:39 +0800 Subject: drm/amd/amgpu: update raven sdma golden setting Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index d5f3848af26a..9206d39cb41b 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -89,7 +89,7 @@ static const u32 golden_settings_sdma_vg10[] = { static const u32 golden_settings_sdma_4_1[] = { - SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07, + SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831d07, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xffffffff, 0x3f000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0111, 0x00000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), 0xfffffff7, 0x00403000, -- cgit From a49ccdbd1d70005049647ac1f8e8989c54b41e63 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Wed, 20 Sep 2017 10:55:44 +0800 Subject: drm/amd/amgpu: update vega10 sdma golden setting Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 9206d39cb41b..3524060f8480 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -54,7 +54,7 @@ static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev); static void sdma_v4_0_set_irq_funcs(struct amdgpu_device *adev); static const u32 golden_settings_sdma_4[] = { - SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831f07, + SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CHICKEN_BITS), 0xfe931f07, 0x02831d07, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CLK_CTRL), 0xff000ff0, 0x3f000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_IB_CNTL), 0x800f0100, 0x00000100, SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL), 0xfffffff7, 0x00403000, -- cgit From 10cfafd62af4295abc73cbf4531654d4f335ff15 Mon Sep 17 00:00:00 2001 From: Tom St Denis Date: Tue, 19 Sep 2017 11:29:04 -0400 Subject: drm/amd/amdgpu: Partial revert of iova debugfs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We discovered that on some devices even with iommu enabled you can access all of system memory through the iommu translation. Therefore, we revert the read method to the translation only service and drop the write method completely. Signed-off-by: Tom St Denis Reviewed-by: Christan König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 90 +++++---------------------------- 1 file changed, 13 insertions(+), 77 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 0e5f78f3a97e..ce435dbbb398 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1784,98 +1784,34 @@ static ssize_t amdgpu_iova_to_phys_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { struct amdgpu_device *adev = file_inode(f)->i_private; - ssize_t result, n; int r; uint64_t phys; - void *ptr; struct iommu_domain *dom; - dom = iommu_get_domain_for_dev(adev->dev); - if (!dom) - return -EFAULT; - - result = 0; - while (size) { - // get physical address and map - phys = iommu_iova_to_phys(dom, *pos); - - // copy upto one page - if (size > PAGE_SIZE) - n = PAGE_SIZE; - else - n = size; - - // to end of the page - if (((*pos & (PAGE_SIZE - 1)) + n) >= PAGE_SIZE) - n = PAGE_SIZE - (*pos & (PAGE_SIZE - 1)); - - ptr = kmap(pfn_to_page(PFN_DOWN(phys))); - if (!ptr) - return -EFAULT; - - r = copy_to_user(buf, ptr, n); - kunmap(pfn_to_page(PFN_DOWN(phys))); - if (r) - return -EFAULT; - - *pos += n; - size -= n; - result += n; - } - - return result; -} + // always return 8 bytes + if (size != 8) + return -EINVAL; -static ssize_t amdgpu_iova_to_phys_write(struct file *f, const char __user *buf, - size_t size, loff_t *pos) -{ - struct amdgpu_device *adev = file_inode(f)->i_private; - ssize_t result, n; - int r; - uint64_t phys; - void *ptr; - struct iommu_domain *dom; + // only accept page addresses + if (*pos & 0xFFF) + return -EINVAL; dom = iommu_get_domain_for_dev(adev->dev); - if (!dom) - return -EFAULT; - - result = 0; - while (size) { - // get physical address and map + if (dom) phys = iommu_iova_to_phys(dom, *pos); + else + phys = *pos; - // copy upto one page - if (size > PAGE_SIZE) - n = PAGE_SIZE; - else - n = size; - - // to end of the page - if (((*pos & (PAGE_SIZE - 1)) + n) >= PAGE_SIZE) - n = PAGE_SIZE - (*pos & (PAGE_SIZE - 1)); - - ptr = kmap(pfn_to_page(PFN_DOWN(phys))); - if (!ptr) - return -EFAULT; - - r = copy_from_user(ptr, buf, n); - kunmap(pfn_to_page(PFN_DOWN(phys))); - if (r) - return -EFAULT; - - *pos += n; - size -= n; - result += n; - } + r = copy_to_user(buf, &phys, 8); + if (r) + return -EFAULT; - return result; + return 8; } static const struct file_operations amdgpu_ttm_iova_fops = { .owner = THIS_MODULE, .read = amdgpu_iova_to_phys_read, - .write = amdgpu_iova_to_phys_write, .llseek = default_llseek }; -- cgit From 4d1f9fb721fdfa7789515167d1bd2e42cf87e12e Mon Sep 17 00:00:00 2001 From: Eric Huang Date: Fri, 15 Sep 2017 16:33:38 -0400 Subject: drm/amdgpu: add cgs query info of pci bus devfn Signed-off-by: Eric Huang Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index e52192051507..383204e911a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -875,6 +875,9 @@ static int amdgpu_cgs_query_system_info(struct cgs_device *cgs_device, case CGS_SYSTEM_INFO_PCIE_SUB_SYS_VENDOR_ID: sys_info->value = adev->pdev->subsystem_vendor; break; + case CGS_SYSTEM_INFO_PCIE_BUS_DEVFN: + sys_info->value = adev->pdev->devfn; + break; default: return -ENODEV; } -- cgit From 5c58301856e69223b85546fac319363eef1ac1b9 Mon Sep 17 00:00:00 2001 From: Evan Quan Date: Wed, 20 Sep 2017 16:25:40 +0800 Subject: drm/amd/amdgpu: add vega10/raven mmhub/athub golden settings Signed-off-by: Evan Quan Reviewed-by: Junwei Zhang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 99147f576e76..621699331e09 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -32,6 +32,8 @@ #include "vega10/DC/dce_12_0_offset.h" #include "vega10/DC/dce_12_0_sh_mask.h" #include "vega10/vega10_enum.h" +#include "vega10/MMHUB/mmhub_1_0_offset.h" +#include "vega10/ATHUB/athub_1_0_offset.h" #include "soc15_common.h" @@ -71,6 +73,18 @@ static const u32 golden_settings_vega10_hdp[] = 0xf6e, 0x0fffffff, 0x00000000, }; +static const u32 golden_settings_mmhub_1_0_0[] = +{ + SOC15_REG_OFFSET(MMHUB, 0, mmDAGB1_WRCLI2), 0x00000007, 0xfe5fe0fa, + SOC15_REG_OFFSET(MMHUB, 0, mmMMEA1_DRAM_WR_CLI2GRP_MAP0), 0x00000030, 0x55555565 +}; + +static const u32 golden_settings_athub_1_0_0[] = +{ + SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL), 0x0000ff00, 0x00000800, + SOC15_REG_OFFSET(ATHUB, 0, mmRPB_ARB_CNTL2), 0x00ff00ff, 0x00080008 +}; + static int gmc_v9_0_vm_fault_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type, @@ -665,8 +679,17 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev) { switch (adev->asic_type) { case CHIP_VEGA10: + amdgpu_program_register_sequence(adev, + golden_settings_mmhub_1_0_0, + (const u32)ARRAY_SIZE(golden_settings_mmhub_1_0_0)); + amdgpu_program_register_sequence(adev, + golden_settings_athub_1_0_0, + (const u32)ARRAY_SIZE(golden_settings_athub_1_0_0)); break; case CHIP_RAVEN: + amdgpu_program_register_sequence(adev, + golden_settings_athub_1_0_0, + (const u32)ARRAY_SIZE(golden_settings_athub_1_0_0)); break; default: break; -- cgit From fd4495e57cd2594098493a0e5b6beccae8cbf47b Mon Sep 17 00:00:00 2001 From: "Xiangliang.Yu" Date: Thu, 21 Sep 2017 10:19:49 +0800 Subject: drm/amdgpu: Fix driver reloading failure SRIOV doesn't implement PMC capability of PCIe, so it can't update power state by reading PMC register. Currently, amdgpu driver doesn't disable pci device when removing driver, the enable_cnt of pci device will not be decrease to 0. When reloading driver, pci_enable_device will do nothing as enable_cnt is not zero. And power state will not be updated as PMC is not support. So current_state of pci device is not D0 state and pci_enable_msi return fail. Add pci_disable_device when remmoving driver to fix the issue. Signed-off-by: Xiangliang.Yu Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index dee35a9c5723..91e42b60d66b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -609,6 +609,8 @@ amdgpu_pci_remove(struct pci_dev *pdev) drm_dev_unregister(dev); drm_dev_unref(dev); + pci_disable_device(pdev); + pci_set_drvdata(pdev, NULL); } static void -- cgit From a8ffeac96daa66132d44a624c1cf3cbcc2598a34 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 18 Sep 2017 14:32:38 +0200 Subject: drm/amdgpu: use 2MB fragment size for GFX6,7 and 8 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use 2MB fragment size by default for older hardware generations as well. Signed-off-by: Christian König Acked-by: John Bridgman Reviewed-by: Roger He Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 2 +- drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 5be9c83dfcf7..2d1f3f651e1f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -831,7 +831,7 @@ static int gmc_v6_0_sw_init(void *handle) if (r) return r; - amdgpu_vm_adjust_size(adev, 64, 4); + amdgpu_vm_adjust_size(adev, 64, 9); adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; adev->mc.mc_mask = 0xffffffffffULL; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index eace9e7182c8..2256277d102f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -970,7 +970,7 @@ static int gmc_v7_0_sw_init(void *handle) * Currently set to 4GB ((1 << 20) 4k pages). * Max GPUVM size for cayman and SI is 40 bits. */ - amdgpu_vm_adjust_size(adev, 64, 4); + amdgpu_vm_adjust_size(adev, 64, 9); adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; /* Set the internal MC address mask diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 3b3326daf32b..114671b57004 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -1067,7 +1067,7 @@ static int gmc_v8_0_sw_init(void *handle) * Currently set to 4GB ((1 << 20) 4k pages). * Max GPUVM size for cayman and SI is 40 bits. */ - amdgpu_vm_adjust_size(adev, 64, 4); + amdgpu_vm_adjust_size(adev, 64, 9); adev->vm_manager.max_pfn = adev->vm_manager.vm_size << 18; /* Set the internal MC address mask -- cgit From dfe5c2b76b2a32cd37283809737e55f9208f8346 Mon Sep 17 00:00:00 2001 From: Yong Zhao Date: Mon, 18 Sep 2017 14:25:31 -0400 Subject: drm/amdgpu: Correct bytes limit for SDMA 3.0 copy and fill MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Yong Zhao Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 728c0d8e849b..4858c9974c86 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -1730,11 +1730,11 @@ static void sdma_v3_0_emit_fill_buffer(struct amdgpu_ib *ib, } static const struct amdgpu_buffer_funcs sdma_v3_0_buffer_funcs = { - .copy_max_bytes = 0x1fffff, + .copy_max_bytes = 0x3fffe0, /* not 0x3fffff due to HW limitation */ .copy_num_dw = 7, .emit_copy_buffer = sdma_v3_0_emit_copy_buffer, - .fill_max_bytes = 0x1fffff, + .fill_max_bytes = 0x3fffe0, /* not 0x3fffff due to HW limitation */ .fill_num_dw = 5, .emit_fill_buffer = sdma_v3_0_emit_fill_buffer, }; -- cgit From 7bdc53f925af085ffa0580f10489f82b36cc2f1c Mon Sep 17 00:00:00 2001 From: Yong Zhao Date: Fri, 15 Sep 2017 18:20:37 -0400 Subject: drm/amdgpu: Fix a bug in amdgpu_fill_buffer() MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When max_bytes is not 8 bytes aligned and bo size is larger than max_bytes, the last 8 bytes in a ttm node may be left unchanged. For example, on pre SDMA 4.0, max_bytes = 0x1fffff, and the bo size is 0x200000, the problem will happen. In order to fix the problem, we separately store the max nums of PTEs/PDEs a single operation can set in amdgpu_vm_pte_funcs structure, rather than inferring it from bytes limit of SDMA constant fill, i.e. fill_max_bytes. Together with the fix, we replace the hard code value "10" in amdgpu_vm_bo_update_mapping() with the corresponding values from structure amdgpu_vm_pte_funcs. Signed-off-by: Yong Zhao Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 7 +++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c | 8 ++++---- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 5 +++-- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 3 +++ drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 3 +++ drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 4 ++++ drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 3 +++ drivers/gpu/drm/amd/amdgpu/si_dma.c | 3 +++ 8 files changed, 30 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index f085c8c9f267..1bf42a8ef23e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -302,6 +302,13 @@ struct amdgpu_vm_pte_funcs { void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, uint64_t value, unsigned count, uint32_t incr); + + /* maximum nums of PTEs/PDEs in a single operation */ + uint32_t set_max_nums_pte_pde; + + /* number of dw to reserve per operation */ + unsigned set_pte_pde_num_dw; + /* for linear pte/pde updates without addr mapping */ void (*set_pte_pde)(struct amdgpu_ib *ib, uint64_t pe, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index ce435dbbb398..1086f039d8e3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -1527,8 +1527,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, struct dma_fence **fence) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - /* max_bytes applies to SDMA_OP_PTEPDE as well as SDMA_OP_CONST_FILL*/ - uint32_t max_bytes = adev->mman.buffer_funcs->fill_max_bytes; + uint32_t max_bytes = 8 * + adev->vm_manager.vm_pte_funcs->set_max_nums_pte_pde; struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; struct drm_mm_node *mm_node; @@ -1560,8 +1560,8 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, ++mm_node; } - /* 10 double words for each SDMA_OP_PTEPDE cmd */ - num_dw = num_loops * 10; + /* num of dwords for each SDMA_OP_PTEPDE cmd */ + num_dw = num_loops * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw; /* for IB padding */ num_dw += 64; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 6c1133298b17..28d16781377f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1606,10 +1606,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, } else { /* set page commands needed */ - ndw += ncmds * 10; + ndw += ncmds * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw; /* extra commands for begin/end fragments */ - ndw += 2 * 10 * adev->vm_manager.fragment_size; + ndw += 2 * adev->vm_manager.vm_pte_funcs->set_pte_pde_num_dw + * adev->vm_manager.fragment_size; params.func = amdgpu_vm_do_set_ptes; } diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index f508f4d01e4a..c64dcd1883b5 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -1389,6 +1389,9 @@ static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev) static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = { .copy_pte = cik_sdma_vm_copy_pte, .write_pte = cik_sdma_vm_write_pte, + + .set_max_nums_pte_pde = 0x1fffff >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = cik_sdma_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index f2d0710258cb..c05eb74d3404 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -1326,6 +1326,9 @@ static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev) static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = { .copy_pte = sdma_v2_4_vm_copy_pte, .write_pte = sdma_v2_4_vm_write_pte, + + .set_max_nums_pte_pde = 0x1fffff >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = sdma_v2_4_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 4858c9974c86..2079340656d2 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -1750,6 +1750,10 @@ static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev) static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = { .copy_pte = sdma_v3_0_vm_copy_pte, .write_pte = sdma_v3_0_vm_write_pte, + + /* not 0x3fffff due to HW limitation */ + .set_max_nums_pte_pde = 0x3fffe0 >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = sdma_v3_0_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 3524060f8480..2605faf56dff 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1716,6 +1716,9 @@ static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev) static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = { .copy_pte = sdma_v4_0_vm_copy_pte, .write_pte = sdma_v4_0_vm_write_pte, + + .set_max_nums_pte_pde = 0x400000 >> 3, + .set_pte_pde_num_dw = 10, .set_pte_pde = sdma_v4_0_vm_set_pte_pde, }; diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index 112969f3301a..adb6ae7d63ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -889,6 +889,9 @@ static void si_dma_set_buffer_funcs(struct amdgpu_device *adev) static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = { .copy_pte = si_dma_vm_copy_pte, .write_pte = si_dma_vm_write_pte, + + .set_max_nums_pte_pde = 0xffff8 >> 3, + .set_pte_pde_num_dw = 9, .set_pte_pde = si_dma_vm_set_pte_pde, }; -- cgit From e6d921974a51e607515b39baa8d1c3f1a27d008b Mon Sep 17 00:00:00 2001 From: Yong Zhao Date: Tue, 19 Sep 2017 12:58:15 -0400 Subject: drm/amdgpu: Add copy_pte_num_dw member in amdgpu_vm_pte_funcs MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Use it to replace the hard coded value in amdgpu_vm_bo_update_mapping(). Signed-off-by: Yong Zhao Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 4 ++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 2 +- drivers/gpu/drm/amd/amdgpu/cik_sdma.c | 2 ++ drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c | 2 ++ drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c | 2 ++ drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 2 ++ drivers/gpu/drm/amd/amdgpu/si_dma.c | 2 ++ 7 files changed, 15 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 1bf42a8ef23e..7c43add4e444 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -294,10 +294,14 @@ struct amdgpu_buffer_funcs { /* provided by hw blocks that can write ptes, e.g., sdma */ struct amdgpu_vm_pte_funcs { + /* number of dw to reserve per operation */ + unsigned copy_pte_num_dw; + /* copy pte entries from GART */ void (*copy_pte)(struct amdgpu_ib *ib, uint64_t pe, uint64_t src, unsigned count); + /* write pte one entry at a time with addr mapping */ void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, uint64_t value, unsigned count, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index 28d16781377f..8fcc743dfa86 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -1597,7 +1597,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (pages_addr) { /* copy commands needed */ - ndw += ncmds * 7; + ndw += ncmds * adev->vm_manager.vm_pte_funcs->copy_pte_num_dw; /* and also PTEs */ ndw += nptes * 2; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index c64dcd1883b5..60cecd117705 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -1387,7 +1387,9 @@ static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = cik_sdma_vm_copy_pte, + .write_pte = cik_sdma_vm_write_pte, .set_max_nums_pte_pde = 0x1fffff >> 3, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index c05eb74d3404..acdee3a4602c 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -1324,7 +1324,9 @@ static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = sdma_v2_4_vm_copy_pte, + .write_pte = sdma_v2_4_vm_write_pte, .set_max_nums_pte_pde = 0x1fffff >> 3, diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index 2079340656d2..72f31cc7df00 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -1748,7 +1748,9 @@ static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = sdma_v3_0_vm_copy_pte, + .write_pte = sdma_v3_0_vm_write_pte, /* not 0x3fffff due to HW limitation */ diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 2605faf56dff..61572c5d19ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1714,7 +1714,9 @@ static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = { + .copy_pte_num_dw = 7, .copy_pte = sdma_v4_0_vm_copy_pte, + .write_pte = sdma_v4_0_vm_write_pte, .set_max_nums_pte_pde = 0x400000 >> 3, diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index adb6ae7d63ef..3fa2fbf8c9a1 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -887,7 +887,9 @@ static void si_dma_set_buffer_funcs(struct amdgpu_device *adev) } static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = { + .copy_pte_num_dw = 5, .copy_pte = si_dma_vm_copy_pte, + .write_pte = si_dma_vm_write_pte, .set_max_nums_pte_pde = 0xffff8 >> 3, -- cgit From 4bd9a67e17b9a2c1b0ca55e7dfc5a711c161373d Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Tue, 4 Jul 2017 16:40:58 +0800 Subject: drm/amdgpu:halt when vm fault MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit only with this way we can debug the VMC page fault issue Signed-off-by: Monk Liu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c | 6 ++++++ drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 7 +++++++ 2 files changed, 13 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index 6c8040e616c4..c17996e18086 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -319,6 +319,12 @@ void gfxhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } WREG32_SOC15(GC, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp); } diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 7ff70762cfc8..cc21c4bdec27 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -561,6 +561,13 @@ void mmhub_v1_0_set_fault_enable_default(struct amdgpu_device *adev, bool value) WRITE_PROTECTION_FAULT_ENABLE_DEFAULT, value); tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, EXECUTE_PROTECTION_FAULT_ENABLE_DEFAULT, value); + if (!value) { + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_NO_RETRY_FAULT, 1); + tmp = REG_SET_FIELD(tmp, VM_L2_PROTECTION_FAULT_CNTL, + CRASH_ON_RETRY_FAULT, 1); + } + WREG32_SOC15(MMHUB, 0, mmVM_L2_PROTECTION_FAULT_CNTL, tmp); } -- cgit From 85f95ad629558b65ab27cce583c683fb9e3da35c Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Thu, 21 Sep 2017 14:59:40 +0800 Subject: drm/amdgpu:unmap KCQ in gfx hw_fini(v2) v2: move kcq_disable out of SRIOV, make it genearal Signed-off-by: Monk Liu Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 57 ++++++++++++++++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 58 +++++++++++++++++++++++++++++++++++ 2 files changed, 115 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 0c4a3b8e8596..dfc10b1baea0 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -5034,12 +5034,69 @@ static int gfx_v8_0_hw_init(void *handle) return r; } +static int gfx_v8_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint32_t scratch, tmp = 0; + int r, i; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("Failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + + r = amdgpu_ring_alloc(kiq_ring, 10); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + amdgpu_gfx_scratch_free(adev, scratch); + return r; + } + + /* unmap queues */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + /* write to scratch for completion */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(kiq_ring, 0xDEADBEEF); + amdgpu_ring_commit(kiq_ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i >= adev->usec_timeout) { + DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); + r = -EINVAL; + } + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + static int gfx_v8_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) + gfx_v8_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); + if (amdgpu_sriov_vf(adev)) { pr_debug("For SRIOV client, shouldn't do anything.\n"); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index e2ae00df1d52..9945218a5489 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -2895,12 +2895,70 @@ static int gfx_v9_0_hw_init(void *handle) return r; } +static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = kiq_ring->adev; + uint32_t scratch, tmp = 0; + int r, i; + + r = amdgpu_gfx_scratch_get(adev, &scratch); + if (r) { + DRM_ERROR("Failed to get scratch reg (%d).\n", r); + return r; + } + WREG32(scratch, 0xCAFEDEAD); + + r = amdgpu_ring_alloc(kiq_ring, 10); + if (r) { + DRM_ERROR("Failed to lock KIQ (%d).\n", r); + amdgpu_gfx_scratch_free(adev, scratch); + return r; + } + + /* unmap queues */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ + PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | + PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | + PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + /* write to scratch for completion */ + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); + amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); + amdgpu_ring_write(kiq_ring, 0xDEADBEEF); + amdgpu_ring_commit(kiq_ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(scratch); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + if (i >= adev->usec_timeout) { + DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); + r = -EINVAL; + } + amdgpu_gfx_scratch_free(adev, scratch); + return r; +} + + static int gfx_v9_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int i; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) + gfx_v9_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); + if (amdgpu_sriov_vf(adev)) { pr_debug("For SRIOV client, shouldn't do anything.\n"); return 0; -- cgit From beb841028480a13a664f868688102f3cee762a6b Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Thu, 21 Sep 2017 15:10:06 +0800 Subject: drm/amdgpu/sriov:alloc KIQ MQD in VRAM(v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit this way after KIQ MQD released in drv unloading, CPC can still let KIQ access this MQD thus RLCV SAVE_VF will not fail v2: always use VRAM domain for KIQ MQD no matter BM or SRIOV Signed-off-by: Monk Liu Acked-by: Alex Deucher Acked-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 4f6c68fc1dd9..4fcd98e65998 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -260,8 +260,13 @@ int amdgpu_gfx_compute_mqd_sw_init(struct amdgpu_device *adev, /* create MQD for KIQ */ ring = &adev->gfx.kiq.ring; if (!ring->mqd_obj) { + /* originaly the KIQ MQD is put in GTT domain, but for SRIOV VRAM domain is a must + * otherwise hypervisor trigger SAVE_VF fail after driver unloaded which mean MQD + * deallocated and gart_unbind, to strict diverage we decide to use VRAM domain for + * KIQ MQD no matter SRIOV or Bare-metal + */ r = amdgpu_bo_create_kernel(adev, mqd_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, &ring->mqd_obj, + AMDGPU_GEM_DOMAIN_VRAM, &ring->mqd_obj, &ring->mqd_gpu_addr, &ring->mqd_ptr); if (r) { dev_warn(adev->dev, "failed to create ring mqd ob (%d)", r); -- cgit From 4ff184d70e1dd85cc39f3b8ed0d98b728d6d9b6c Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Fri, 15 Sep 2017 16:43:01 +0800 Subject: drm/amdgpu:fix uvd ring fini routine(v2) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit fix missing finish uvd enc_ring. v2: since the adev pointer check in already in ring_fini so drop the check outsider Signed-off-by: Monk Liu Reviewed-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 331e34ac61fd..e8bd50cf9785 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -269,6 +269,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) { + int i; kfree(adev->uvd.saved_bo); amd_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity); @@ -279,6 +280,9 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) amdgpu_ring_fini(&adev->uvd.ring); + for (i = 0; i < AMDGPU_MAX_UVD_ENC_RINGS; ++i) + amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + release_firmware(adev->uvd.fw); return 0; -- cgit From c833d8aa4d829e858f1be8f4bd82a1503b611013 Mon Sep 17 00:00:00 2001 From: Monk Liu Date: Tue, 19 Sep 2017 16:09:53 +0800 Subject: drm/amdgpu:fix firmware memoryleak(v2) this fix memory leak due to request_firmware after driver unloaded v2: release gmc firmware for gmc6/7/8 as well Signed-off-by: Monk Liu Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 6 ++++++ drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c | 20 ++++++++++++++++++++ drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c | 2 ++ drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c | 2 ++ drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 2 ++ drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c | 5 +++++ 6 files changed, 37 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index e02828665b14..447d446b5015 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -92,6 +92,12 @@ static int psp_sw_init(void *handle) static int psp_sw_fini(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + release_firmware(adev->psp.sos_fw); + adev->psp.sos_fw = NULL; + release_firmware(adev->psp.asd_fw); + adev->psp.asd_fw = NULL; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 9945218a5489..deeaee1457ef 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -384,6 +384,25 @@ err1: return r; } + +static void gfx_v9_0_free_microcode(struct amdgpu_device *adev) +{ + release_firmware(adev->gfx.pfp_fw); + adev->gfx.pfp_fw = NULL; + release_firmware(adev->gfx.me_fw); + adev->gfx.me_fw = NULL; + release_firmware(adev->gfx.ce_fw); + adev->gfx.ce_fw = NULL; + release_firmware(adev->gfx.rlc_fw); + adev->gfx.rlc_fw = NULL; + release_firmware(adev->gfx.mec_fw); + adev->gfx.mec_fw = NULL; + release_firmware(adev->gfx.mec2_fw); + adev->gfx.mec2_fw = NULL; + + kfree(adev->gfx.rlc.register_list_format); +} + static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; @@ -1429,6 +1448,7 @@ static int gfx_v9_0_sw_fini(void *handle) gfx_v9_0_mec_fini(adev); gfx_v9_0_ngg_fini(adev); + gfx_v9_0_free_microcode(adev); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 2d1f3f651e1f..f4603a7c8ef3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -901,6 +901,8 @@ static int gmc_v6_0_sw_fini(void *handle) gmc_v6_0_gart_fini(adev); amdgpu_gem_force_release(adev); amdgpu_bo_fini(adev); + release_firmware(adev->mc.fw); + adev->mc.fw = NULL; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 2256277d102f..b0528ca9207b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -1050,6 +1050,8 @@ static int gmc_v7_0_sw_fini(void *handle) gmc_v7_0_gart_fini(adev); amdgpu_gem_force_release(adev); amdgpu_bo_fini(adev); + release_firmware(adev->mc.fw); + adev->mc.fw = NULL; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 114671b57004..f368cfe2f585 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -1147,6 +1147,8 @@ static int gmc_v8_0_sw_fini(void *handle) gmc_v8_0_gart_fini(adev); amdgpu_gem_force_release(adev); amdgpu_bo_fini(adev); + release_firmware(adev->mc.fw); + adev->mc.fw = NULL; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 61572c5d19ef..c26d205ff3bf 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -1264,6 +1264,11 @@ static int sdma_v4_0_sw_fini(void *handle) for (i = 0; i < adev->sdma.num_instances; i++) amdgpu_ring_fini(&adev->sdma.instance[i].ring); + for (i = 0; i < adev->sdma.num_instances; i++) { + release_firmware(adev->sdma.instance[i].fw); + adev->sdma.instance[i].fw = NULL; + } + return 0; } -- cgit From e9c7577c09b5062a4f1464b8ef4472c14bdfd2e7 Mon Sep 17 00:00:00 2001 From: Christian König Date: Mon, 11 Sep 2017 17:29:26 +0200 Subject: drm/amdgpu: simplify pinning into visible VRAM MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Just set the CPU access required flag when we pin it. Signed-off-by: Christian König Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_object.c | 23 ++++++++--------------- 1 file changed, 8 insertions(+), 15 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 726a662f43f4..6982baeccd14 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -635,7 +635,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); int r, i; - unsigned fpfn, lpfn; if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) return -EPERM; @@ -667,22 +666,16 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, } bo->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + /* force to pin into visible video ram */ + if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) + bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; amdgpu_ttm_placement_from_domain(bo, domain); for (i = 0; i < bo->placement.num_placement; i++) { - /* force to pin into visible video ram */ - if ((bo->placements[i].flags & TTM_PL_FLAG_VRAM) && - !(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS) && - (!max_offset || max_offset > - adev->mc.visible_vram_size)) { - if (WARN_ON_ONCE(min_offset > - adev->mc.visible_vram_size)) - return -EINVAL; - fpfn = min_offset >> PAGE_SHIFT; - lpfn = adev->mc.visible_vram_size >> PAGE_SHIFT; - } else { - fpfn = min_offset >> PAGE_SHIFT; - lpfn = max_offset >> PAGE_SHIFT; - } + unsigned fpfn, lpfn; + + fpfn = min_offset >> PAGE_SHIFT; + lpfn = max_offset >> PAGE_SHIFT; + if (fpfn > bo->placements[i].fpfn) bo->placements[i].fpfn = fpfn; if (!bo->placements[i].lpfn || -- cgit From dfced2e4bcbc8d6bd9bad11cc6b4643ba36ed35a Mon Sep 17 00:00:00 2001 From: Samuel Li Date: Tue, 22 Aug 2017 15:25:33 -0400 Subject: drm/amdgpu: Add gem_prime_mmap support MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit v2: drop hdp invalidate/flush. v3: honor pgoff during prime mmap. Add a barrier after cpu access. v4: drop begin/end_cpu_access() for now, revisit later. Signed-off-by: Samuel Li Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c | 34 +++++++++++++++++++++++++++++++ 3 files changed, 36 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 7c43add4e444..ebfc267467ee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -413,6 +413,7 @@ void amdgpu_gem_prime_unpin(struct drm_gem_object *obj); struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *); void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj); void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); +int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); int amdgpu_gem_debugfs_init(struct amdgpu_device *adev); /* sub-allocation manager, it has to be protected by another lock. diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 91e42b60d66b..4f98960e47f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -855,6 +855,7 @@ static struct drm_driver kms_driver = { .gem_prime_import_sg_table = amdgpu_gem_prime_import_sg_table, .gem_prime_vmap = amdgpu_gem_prime_vmap, .gem_prime_vunmap = amdgpu_gem_prime_vunmap, + .gem_prime_mmap = amdgpu_gem_prime_mmap, .name = DRIVER_NAME, .desc = DRIVER_DESC, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 7e0826469b5e..90af8e82b16a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -57,6 +57,40 @@ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) ttm_bo_kunmap(&bo->dma_buf_vmap); } +int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) +{ + struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + unsigned asize = amdgpu_bo_size(bo); + int ret; + + if (!vma->vm_file) + return -ENODEV; + + if (adev == NULL) + return -ENODEV; + + /* Check for valid size. */ + if (asize < vma->vm_end - vma->vm_start) + return -EINVAL; + + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) || + (bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) { + return -EPERM; + } + vma->vm_pgoff += amdgpu_bo_mmap_offset(bo) >> PAGE_SHIFT; + + /* prime mmap does not need to check access, so allow here */ + ret = drm_vma_node_allow(&obj->vma_node, vma->vm_file->private_data); + if (ret) + return ret; + + ret = ttm_bo_mmap(vma->vm_file, vma, &adev->mman.bdev); + drm_vma_node_revoke(&obj->vma_node, vma->vm_file->private_data); + + return ret; +} + struct drm_gem_object * amdgpu_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, -- cgit From 6f87a895709eecc1542fe947e349364ad061ac00 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 19 Sep 2017 10:20:27 -0400 Subject: drm/amdgpu: clarify license in amdgpu_trace_points.c MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit It was not clear. The rest of the driver is MIT/X11. Reviewed-by: Christian König Acked-by: Dave Airlie Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) (limited to 'drivers/gpu/drm/amd/amdgpu') diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c index 385b7e1d72f9..9ec96b9e85d1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c @@ -1,4 +1,23 @@ /* Copyright Red Hat Inc 2010. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * * Author : Dave Airlie */ #include -- cgit