diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c | 526 |
1 files changed, 439 insertions, 87 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c index 82eab49be82b..9c169112a5e7 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_4_2.c @@ -30,6 +30,7 @@ #include "amdgpu_xcp.h" #include "amdgpu_ucode.h" #include "amdgpu_trace.h" +#include "amdgpu_reset.h" #include "sdma/sdma_4_4_2_offset.h" #include "sdma/sdma_4_4_2_sh_mask.h" @@ -44,6 +45,54 @@ #include "amdgpu_ras.h" MODULE_FIRMWARE("amdgpu/sdma_4_4_2.bin"); +MODULE_FIRMWARE("amdgpu/sdma_4_4_5.bin"); + +static const struct amdgpu_hwip_reg_entry sdma_reg_list_4_4_2[] = { + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS1_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS2_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_STATUS3_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UCODE_CHECKSUM), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RB_RPTR_FETCH_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RB_RPTR_FETCH), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_RD_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_WR_STATUS), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_RD_XNACK0), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_RD_XNACK1), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_WR_XNACK0), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_UTCL1_WR_XNACK1), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_RPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_RB_WPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_IB_SUB_REMAIN), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_GFX_DUMMY_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_RPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_RB_WPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_IB_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_IB_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_IB_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_PAGE_DUMMY_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_CNTL), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_RPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_RPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_WPTR), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_RB_WPTR_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_IB_OFFSET), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_IB_BASE_LO), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_IB_BASE_HI), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_RLC0_DUMMY_REG), + SOC15_REG_ENTRY_STR(GC, 0, regSDMA_VM_CNTL) +}; #define mmSMNAID_AID0_MCA_SMU 0x03b30400 @@ -57,6 +106,9 @@ static void sdma_v4_4_2_set_buffer_funcs(struct amdgpu_device *adev); static void sdma_v4_4_2_set_vm_pte_funcs(struct amdgpu_device *adev); static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev); static void sdma_v4_4_2_set_ras_funcs(struct amdgpu_device *adev); +static void sdma_v4_4_2_update_reset_mask(struct amdgpu_device *adev); +static int sdma_v4_4_2_stop_queue(struct amdgpu_ring *ring); +static int sdma_v4_4_2_restore_queue(struct amdgpu_ring *ring); static u32 sdma_v4_4_2_get_reg_offset(struct amdgpu_device *adev, u32 instance, u32 offset) @@ -82,7 +134,7 @@ static unsigned sdma_v4_4_2_seq_to_irq_id(int seq_num) } } -static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id) +static int sdma_v4_4_2_irq_id_to_seq(struct amdgpu_device *adev, unsigned client_id) { switch (client_id) { case SOC15_IH_CLIENTID_SDMA0: @@ -90,9 +142,15 @@ static int sdma_v4_4_2_irq_id_to_seq(unsigned client_id) case SOC15_IH_CLIENTID_SDMA1: return 1; case SOC15_IH_CLIENTID_SDMA2: - return 2; + if (amdgpu_sriov_vf(adev) && (adev->gfx.xcc_mask == 0x1)) + return 0; + else + return 2; case SOC15_IH_CLIENTID_SDMA3: - return 3; + if (amdgpu_sriov_vf(adev) && (adev->gfx.xcc_mask == 0x1)) + return 1; + else + return 3; default: return -EINVAL; } @@ -134,8 +192,9 @@ static int sdma_v4_4_2_init_microcode(struct amdgpu_device *adev) int ret, i; for (i = 0; i < adev->sdma.num_instances; i++) { - if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == - IP_VERSION(4, 4, 2)) { + if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 2) || + amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 4) || + amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 5)) { ret = amdgpu_sdma_init_microcode(adev, 0, true); break; } else { @@ -368,7 +427,8 @@ static void sdma_v4_4_2_ring_emit_hdp_flush(struct amdgpu_ring *ring) u32 ref_and_mask = 0; const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio.hdp_flush_reg; - ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 << ring->me; + ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0 + << (ring->me % adev->sdma.num_inst_per_aid); sdma_v4_4_2_wait_reg_mem(ring, 0, 1, adev->nbio.funcs->get_hdp_flush_done_offset(adev), @@ -612,11 +672,12 @@ static uint32_t sdma_v4_4_2_rb_cntl(struct amdgpu_ring *ring, uint32_t rb_cntl) * * @adev: amdgpu_device pointer * @i: instance to resume + * @restore: used to restore wptr when restart * * Set up the gfx DMA ring buffers and enable them. * Returns 0 for success, error for failure. */ -static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i) +static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i, bool restore) { struct amdgpu_ring *ring = &adev->sdma.instance[i].ring; u32 rb_cntl, ib_cntl, wptr_poll_cntl; @@ -624,6 +685,7 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i) u32 doorbell; u32 doorbell_offset; u64 wptr_gpu_addr; + u64 rwptr; wb_offset = (ring->rptr_offs * 4); @@ -643,16 +705,32 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i) WREG32_SDMA(i, regSDMA_GFX_RB_BASE, ring->gpu_addr >> 8); WREG32_SDMA(i, regSDMA_GFX_RB_BASE_HI, ring->gpu_addr >> 40); - ring->wptr = 0; + if (!restore) + ring->wptr = 0; /* before programing wptr to a less value, need set minor_ptr_update first */ WREG32_SDMA(i, regSDMA_GFX_MINOR_PTR_UPDATE, 1); + /* For the guilty queue, set RPTR to the current wptr to skip bad commands, + * It is not a guilty queue, restore cache_rptr and continue execution. + */ + if (adev->sdma.instance[i].gfx_guilty) + rwptr = ring->wptr; + else + rwptr = ring->cached_rptr; + /* Initialize the ring buffer's read and write pointers */ - WREG32_SDMA(i, regSDMA_GFX_RB_RPTR, 0); - WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_HI, 0); - WREG32_SDMA(i, regSDMA_GFX_RB_WPTR, 0); - WREG32_SDMA(i, regSDMA_GFX_RB_WPTR_HI, 0); + if (restore) { + WREG32_SDMA(i, regSDMA_GFX_RB_RPTR, lower_32_bits(rwptr << 2)); + WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_HI, upper_32_bits(rwptr << 2)); + WREG32_SDMA(i, regSDMA_GFX_RB_WPTR, lower_32_bits(rwptr << 2)); + WREG32_SDMA(i, regSDMA_GFX_RB_WPTR_HI, upper_32_bits(rwptr << 2)); + } else { + WREG32_SDMA(i, regSDMA_GFX_RB_RPTR, 0); + WREG32_SDMA(i, regSDMA_GFX_RB_RPTR_HI, 0); + WREG32_SDMA(i, regSDMA_GFX_RB_WPTR, 0); + WREG32_SDMA(i, regSDMA_GFX_RB_WPTR_HI, 0); + } doorbell = RREG32_SDMA(i, regSDMA_GFX_DOORBELL); doorbell_offset = RREG32_SDMA(i, regSDMA_GFX_DOORBELL_OFFSET); @@ -700,11 +778,12 @@ static void sdma_v4_4_2_gfx_resume(struct amdgpu_device *adev, unsigned int i) * * @adev: amdgpu_device pointer * @i: instance to resume + * @restore: boolean to say restore needed or not * * Set up the page DMA ring buffers and enable them. * Returns 0 for success, error for failure. */ -static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i) +static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i, bool restore) { struct amdgpu_ring *ring = &adev->sdma.instance[i].page; u32 rb_cntl, ib_cntl, wptr_poll_cntl; @@ -712,6 +791,7 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i) u32 doorbell; u32 doorbell_offset; u64 wptr_gpu_addr; + u64 rwptr; wb_offset = (ring->rptr_offs * 4); @@ -719,11 +799,26 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i) rb_cntl = sdma_v4_4_2_rb_cntl(ring, rb_cntl); WREG32_SDMA(i, regSDMA_PAGE_RB_CNTL, rb_cntl); + /* For the guilty queue, set RPTR to the current wptr to skip bad commands, + * It is not a guilty queue, restore cache_rptr and continue execution. + */ + if (adev->sdma.instance[i].page_guilty) + rwptr = ring->wptr; + else + rwptr = ring->cached_rptr; + /* Initialize the ring buffer's read and write pointers */ - WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR, 0); - WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_HI, 0); - WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR, 0); - WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR_HI, 0); + if (restore) { + WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR, lower_32_bits(rwptr << 2)); + WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_HI, upper_32_bits(rwptr << 2)); + WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR, lower_32_bits(rwptr << 2)); + WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR_HI, upper_32_bits(rwptr << 2)); + } else { + WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR, 0); + WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_HI, 0); + WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR, 0); + WREG32_SDMA(i, regSDMA_PAGE_RB_WPTR_HI, 0); + } /* set the wb address whether it's enabled or not */ WREG32_SDMA(i, regSDMA_PAGE_RB_RPTR_ADDR_HI, @@ -737,7 +832,8 @@ static void sdma_v4_4_2_page_resume(struct amdgpu_device *adev, unsigned int i) WREG32_SDMA(i, regSDMA_PAGE_RB_BASE, ring->gpu_addr >> 8); WREG32_SDMA(i, regSDMA_PAGE_RB_BASE_HI, ring->gpu_addr >> 40); - ring->wptr = 0; + if (!restore) + ring->wptr = 0; /* before programing wptr to a less value, need set minor_ptr_update first */ WREG32_SDMA(i, regSDMA_PAGE_MINOR_PTR_UPDATE, 1); @@ -856,12 +952,13 @@ static int sdma_v4_4_2_inst_load_microcode(struct amdgpu_device *adev, * * @adev: amdgpu_device pointer * @inst_mask: mask of dma engine instances to be enabled + * @restore: boolean to say restore needed or not * * Set up the DMA engines and enable them. * Returns 0 for success, error for failure. */ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev, - uint32_t inst_mask) + uint32_t inst_mask, bool restore) { struct amdgpu_ring *ring; uint32_t tmp_mask; @@ -872,7 +969,7 @@ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev, sdma_v4_4_2_inst_enable(adev, false, inst_mask); } else { /* bypass sdma microcode loading on Gopher */ - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP && + if (!restore && adev->firmware.load_type != AMDGPU_FW_LOAD_PSP && adev->sdma.instance[0].fw) { r = sdma_v4_4_2_inst_load_microcode(adev, inst_mask); if (r) @@ -891,17 +988,19 @@ static int sdma_v4_4_2_inst_start(struct amdgpu_device *adev, uint32_t temp; WREG32_SDMA(i, regSDMA_SEM_WAIT_FAIL_TIMER_CNTL, 0); - sdma_v4_4_2_gfx_resume(adev, i); + sdma_v4_4_2_gfx_resume(adev, i, restore); if (adev->sdma.has_page_queue) - sdma_v4_4_2_page_resume(adev, i); + sdma_v4_4_2_page_resume(adev, i, restore); /* set utc l1 enable flag always to 1 */ temp = RREG32_SDMA(i, regSDMA_CNTL); temp = REG_SET_FIELD(temp, SDMA_CNTL, UTC_L1_ENABLE, 1); - /* enable context empty interrupt during initialization */ - temp = REG_SET_FIELD(temp, SDMA_CNTL, CTXEMPTY_INT_ENABLE, 1); - WREG32_SDMA(i, regSDMA_CNTL, temp); + if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) < IP_VERSION(4, 4, 5)) { + /* enable context empty interrupt during initialization */ + temp = REG_SET_FIELD(temp, SDMA_CNTL, CTXEMPTY_INT_ENABLE, 1); + WREG32_SDMA(i, regSDMA_CNTL, temp); + } if (!amdgpu_sriov_vf(adev)) { if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { /* unhalt engine */ @@ -1055,7 +1154,7 @@ static int sdma_v4_4_2_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = -EINVAL; err1: - amdgpu_ib_free(adev, &ib, NULL); + amdgpu_ib_free(&ib, NULL); dma_fence_put(f); err0: amdgpu_device_wb_free(adev, index); @@ -1228,15 +1327,21 @@ static bool sdma_v4_4_2_fw_support_paging_queue(struct amdgpu_device *adev) { switch (amdgpu_ip_version(adev, SDMA0_HWIP, 0)) { case IP_VERSION(4, 4, 2): + case IP_VERSION(4, 4, 5): return false; default: return false; } } -static int sdma_v4_4_2_early_init(void *handle) +static const struct amdgpu_sdma_funcs sdma_v4_4_2_sdma_funcs = { + .stop_kernel_queue = &sdma_v4_4_2_stop_queue, + .start_kernel_queue = &sdma_v4_4_2_restore_queue, +}; + +static int sdma_v4_4_2_early_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int r; r = sdma_v4_4_2_init_microcode(adev); @@ -1252,7 +1357,6 @@ static int sdma_v4_4_2_early_init(void *handle) sdma_v4_4_2_set_vm_pte_funcs(adev); sdma_v4_4_2_set_irq_funcs(adev); sdma_v4_4_2_set_ras_funcs(adev); - return 0; } @@ -1262,9 +1366,9 @@ static int sdma_v4_4_2_process_ras_data_cb(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry); #endif -static int sdma_v4_4_2_late_init(void *handle) +static int sdma_v4_4_2_late_init(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; #if 0 struct ras_ih_if ih_info = { .cb = sdma_v4_4_2_process_ras_data_cb, @@ -1273,15 +1377,23 @@ static int sdma_v4_4_2_late_init(void *handle) if (!amdgpu_persistent_edc_harvesting_supported(adev)) amdgpu_ras_reset_error_count(adev, AMDGPU_RAS_BLOCK__SDMA); + /* The initialization is done in the late_init stage to ensure that the SMU + * initialization and capability setup are completed before we check the SDMA + * reset capability + */ + sdma_v4_4_2_update_reset_mask(adev); + return 0; } -static int sdma_v4_4_2_sw_init(void *handle) +static int sdma_v4_4_2_sw_init(struct amdgpu_ip_block *ip_block) { struct amdgpu_ring *ring; int r, i; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; u32 aid_id; + uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_4_2); + uint32_t *ptr; /* SDMA trap event */ for (i = 0; i < adev->sdma.num_inst_per_aid; i++) { @@ -1326,9 +1438,21 @@ static int sdma_v4_4_2_sw_init(void *handle) &adev->sdma.srbm_write_irq); if (r) return r; + + r = amdgpu_irq_add_id(adev, sdma_v4_4_2_seq_to_irq_id(i), + SDMA0_4_0__SRCID__SDMA_CTXEMPTY, + &adev->sdma.ctxt_empty_irq); + if (r) + return r; } for (i = 0; i < adev->sdma.num_instances; i++) { + mutex_init(&adev->sdma.instance[i].engine_reset_mutex); + /* Initialize guilty flags for GFX and PAGE queues */ + adev->sdma.instance[i].gfx_guilty = false; + adev->sdma.instance[i].page_guilty = false; + adev->sdma.instance[i].funcs = &sdma_v4_4_2_sdma_funcs; + ring = &adev->sdma.instance[i].ring; ring->ring_obj = NULL; ring->use_doorbell = true; @@ -1372,17 +1496,31 @@ static int sdma_v4_4_2_sw_init(void *handle) } } + adev->sdma.supported_reset = + amdgpu_get_soft_full_reset_mask(&adev->sdma.instance[0].ring); + if (amdgpu_sdma_ras_sw_init(adev)) { dev_err(adev->dev, "fail to initialize sdma ras block\n"); return -EINVAL; } + /* Allocate memory for SDMA IP Dump buffer */ + ptr = kcalloc(adev->sdma.num_instances * reg_count, sizeof(uint32_t), GFP_KERNEL); + if (ptr) + adev->sdma.ip_dump = ptr; + else + DRM_ERROR("Failed to allocated memory for SDMA IP Dump\n"); + + r = amdgpu_sdma_sysfs_reset_mask_init(adev); + if (r) + return r; + return r; } -static int sdma_v4_4_2_sw_fini(void *handle) +static int sdma_v4_4_2_sw_fini(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int i; for (i = 0; i < adev->sdma.num_instances; i++) { @@ -1391,32 +1529,37 @@ static int sdma_v4_4_2_sw_fini(void *handle) amdgpu_ring_fini(&adev->sdma.instance[i].page); } - if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 2)) + amdgpu_sdma_sysfs_reset_mask_fini(adev); + if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 2) || + amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 4) || + amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(4, 4, 5)) amdgpu_sdma_destroy_inst_ctx(adev, true); else amdgpu_sdma_destroy_inst_ctx(adev, false); + kfree(adev->sdma.ip_dump); + return 0; } -static int sdma_v4_4_2_hw_init(void *handle) +static int sdma_v4_4_2_hw_init(struct amdgpu_ip_block *ip_block) { int r; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; uint32_t inst_mask; inst_mask = GENMASK(adev->sdma.num_instances - 1, 0); if (!amdgpu_sriov_vf(adev)) sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask); - r = sdma_v4_4_2_inst_start(adev, inst_mask); + r = sdma_v4_4_2_inst_start(adev, inst_mask, false); return r; } -static int sdma_v4_4_2_hw_fini(void *handle) +static int sdma_v4_4_2_hw_fini(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; uint32_t inst_mask; int i; @@ -1437,29 +1580,27 @@ static int sdma_v4_4_2_hw_fini(void *handle) return 0; } -static int sdma_v4_4_2_set_clockgating_state(void *handle, +static int sdma_v4_4_2_set_clockgating_state(struct amdgpu_ip_block *ip_block, enum amd_clockgating_state state); -static int sdma_v4_4_2_suspend(void *handle) +static int sdma_v4_4_2_suspend(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; if (amdgpu_in_reset(adev)) - sdma_v4_4_2_set_clockgating_state(adev, AMD_CG_STATE_UNGATE); + sdma_v4_4_2_set_clockgating_state(ip_block, AMD_CG_STATE_UNGATE); - return sdma_v4_4_2_hw_fini(adev); + return sdma_v4_4_2_hw_fini(ip_block); } -static int sdma_v4_4_2_resume(void *handle) +static int sdma_v4_4_2_resume(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - return sdma_v4_4_2_hw_init(adev); + return sdma_v4_4_2_hw_init(ip_block); } -static bool sdma_v4_4_2_is_idle(void *handle) +static bool sdma_v4_4_2_is_idle(struct amdgpu_ip_block *ip_block) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; u32 i; for (i = 0; i < adev->sdma.num_instances; i++) { @@ -1472,11 +1613,11 @@ static bool sdma_v4_4_2_is_idle(void *handle) return true; } -static int sdma_v4_4_2_wait_for_idle(void *handle) +static int sdma_v4_4_2_wait_for_idle(struct amdgpu_ip_block *ip_block) { unsigned i, j; u32 sdma[AMDGPU_MAX_SDMA_INSTANCES]; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; for (i = 0; i < adev->usec_timeout; i++) { for (j = 0; j < adev->sdma.num_instances; j++) { @@ -1491,13 +1632,119 @@ static int sdma_v4_4_2_wait_for_idle(void *handle) return -ETIMEDOUT; } -static int sdma_v4_4_2_soft_reset(void *handle) +static int sdma_v4_4_2_soft_reset(struct amdgpu_ip_block *ip_block) { /* todo */ return 0; } +static bool sdma_v4_4_2_is_queue_selected(struct amdgpu_device *adev, uint32_t instance_id, bool is_page_queue) +{ + uint32_t reg_offset = is_page_queue ? regSDMA_PAGE_CONTEXT_STATUS : regSDMA_GFX_CONTEXT_STATUS; + uint32_t context_status = RREG32(sdma_v4_4_2_get_reg_offset(adev, instance_id, reg_offset)); + + /* Check if the SELECTED bit is set */ + return (context_status & SDMA_GFX_CONTEXT_STATUS__SELECTED_MASK) != 0; +} + +static bool sdma_v4_4_2_ring_is_guilty(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t instance_id = ring->me; + + return sdma_v4_4_2_is_queue_selected(adev, instance_id, false); +} + +static bool sdma_v4_4_2_page_ring_is_guilty(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t instance_id = ring->me; + + if (!adev->sdma.has_page_queue) + return false; + + return sdma_v4_4_2_is_queue_selected(adev, instance_id, true); +} + +static int sdma_v4_4_2_reset_queue(struct amdgpu_ring *ring, unsigned int vmid) +{ + struct amdgpu_device *adev = ring->adev; + u32 id = GET_INST(SDMA0, ring->me); + int r; + + if (!(adev->sdma.supported_reset & AMDGPU_RESET_TYPE_PER_QUEUE)) + return -EOPNOTSUPP; + + amdgpu_amdkfd_suspend(adev, false); + r = amdgpu_sdma_reset_engine(adev, id); + amdgpu_amdkfd_resume(adev, false); + + return r; +} + +static int sdma_v4_4_2_stop_queue(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + u32 instance_id = GET_INST(SDMA0, ring->me); + u32 inst_mask; + uint64_t rptr; + + if (amdgpu_sriov_vf(adev)) + return -EINVAL; + + /* Check if this queue is the guilty one */ + adev->sdma.instance[instance_id].gfx_guilty = + sdma_v4_4_2_is_queue_selected(adev, instance_id, false); + if (adev->sdma.has_page_queue) + adev->sdma.instance[instance_id].page_guilty = + sdma_v4_4_2_is_queue_selected(adev, instance_id, true); + + /* Cache the rptr before reset, after the reset, + * all of the registers will be reset to 0 + */ + rptr = amdgpu_ring_get_rptr(ring); + ring->cached_rptr = rptr; + /* Cache the rptr for the page queue if it exists */ + if (adev->sdma.has_page_queue) { + struct amdgpu_ring *page_ring = &adev->sdma.instance[instance_id].page; + rptr = amdgpu_ring_get_rptr(page_ring); + page_ring->cached_rptr = rptr; + } + + /* stop queue */ + inst_mask = 1 << ring->me; + sdma_v4_4_2_inst_gfx_stop(adev, inst_mask); + if (adev->sdma.has_page_queue) + sdma_v4_4_2_inst_page_stop(adev, inst_mask); + + return 0; +} + +static int sdma_v4_4_2_restore_queue(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + u32 inst_mask; + int i; + + inst_mask = 1 << ring->me; + udelay(50); + + for (i = 0; i < adev->usec_timeout; i++) { + if (!REG_GET_FIELD(RREG32_SDMA(ring->me, regSDMA_F32_CNTL), SDMA_F32_CNTL, HALT)) + break; + udelay(1); + } + + if (i == adev->usec_timeout) { + dev_err(adev->dev, "timed out waiting for SDMA%d unhalt after reset\n", + ring->me); + return -ETIMEDOUT; + } + + return sdma_v4_4_2_inst_start(adev, inst_mask, true); +} + static int sdma_v4_4_2_set_trap_irq_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, unsigned type, @@ -1520,7 +1767,7 @@ static int sdma_v4_4_2_process_trap_irq(struct amdgpu_device *adev, uint32_t instance, i; DRM_DEBUG("IH: SDMA trap\n"); - instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id); + instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id); /* Client id gives the SDMA instance in AID. To know the exact SDMA * instance, interrupt entry gives the node id which corresponds to the AID instance. @@ -1543,6 +1790,9 @@ static int sdma_v4_4_2_process_trap_irq(struct amdgpu_device *adev, case 0: amdgpu_fence_process(&adev->sdma.instance[i].ring); break; + case 1: + amdgpu_fence_process(&adev->sdma.instance[i].page); + break; default: break; } @@ -1563,7 +1813,7 @@ static int sdma_v4_4_2_process_ras_data_cb(struct amdgpu_device *adev, if (amdgpu_ras_is_supported(adev, AMDGPU_RAS_BLOCK__SDMA)) goto out; - instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id); + instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id); if (instance < 0) goto out; @@ -1582,7 +1832,7 @@ static int sdma_v4_4_2_process_illegal_inst_irq(struct amdgpu_device *adev, DRM_ERROR("Illegal instruction in SDMA command stream\n"); - instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id); + instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id); if (instance < 0) return 0; @@ -1616,7 +1866,7 @@ static int sdma_v4_4_2_print_iv_entry(struct amdgpu_device *adev, struct amdgpu_task_info *task_info; u64 addr; - instance = sdma_v4_4_2_irq_id_to_seq(entry->client_id); + instance = sdma_v4_4_2_irq_id_to_seq(adev, entry->client_id); if (instance < 0 || instance >= adev->sdma.num_instances) { dev_err(adev->dev, "sdma instance invalid %d\n", instance); return -EINVAL; @@ -1680,6 +1930,16 @@ static int sdma_v4_4_2_process_srbm_write_irq(struct amdgpu_device *adev, return 0; } +static int sdma_v4_4_2_process_ctxt_empty_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + /* There is nothing useful to be done here, only kept for debug */ + dev_dbg_ratelimited(adev->dev, "SDMA context empty interrupt"); + sdma_v4_4_2_print_iv_entry(adev, entry); + return 0; +} + static void sdma_v4_4_2_inst_update_medium_grain_light_sleep( struct amdgpu_device *adev, bool enable, uint32_t inst_mask) { @@ -1746,10 +2006,10 @@ static void sdma_v4_4_2_inst_update_medium_grain_clock_gating( } } -static int sdma_v4_4_2_set_clockgating_state(void *handle, +static int sdma_v4_4_2_set_clockgating_state(struct amdgpu_ip_block *ip_block, enum amd_clockgating_state state) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; uint32_t inst_mask; if (amdgpu_sriov_vf(adev)) @@ -1764,15 +2024,15 @@ static int sdma_v4_4_2_set_clockgating_state(void *handle, return 0; } -static int sdma_v4_4_2_set_powergating_state(void *handle, +static int sdma_v4_4_2_set_powergating_state(struct amdgpu_ip_block *ip_block, enum amd_powergating_state state) { return 0; } -static void sdma_v4_4_2_get_clockgating_state(void *handle, u64 *flags) +static void sdma_v4_4_2_get_clockgating_state(struct amdgpu_ip_block *ip_block, u64 *flags) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; + struct amdgpu_device *adev = ip_block->adev; int data; if (amdgpu_sriov_vf(adev)) @@ -1789,6 +2049,46 @@ static void sdma_v4_4_2_get_clockgating_state(void *handle, u64 *flags) *flags |= AMD_CG_SUPPORT_SDMA_LS; } +static void sdma_v4_4_2_print_ip_state(struct amdgpu_ip_block *ip_block, struct drm_printer *p) +{ + struct amdgpu_device *adev = ip_block->adev; + int i, j; + uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_4_2); + uint32_t instance_offset; + + if (!adev->sdma.ip_dump) + return; + + drm_printf(p, "num_instances:%d\n", adev->sdma.num_instances); + for (i = 0; i < adev->sdma.num_instances; i++) { + instance_offset = i * reg_count; + drm_printf(p, "\nInstance:%d\n", i); + + for (j = 0; j < reg_count; j++) + drm_printf(p, "%-50s \t 0x%08x\n", sdma_reg_list_4_4_2[j].reg_name, + adev->sdma.ip_dump[instance_offset + j]); + } +} + +static void sdma_v4_4_2_dump_ip_state(struct amdgpu_ip_block *ip_block) +{ + struct amdgpu_device *adev = ip_block->adev; + int i, j; + uint32_t instance_offset; + uint32_t reg_count = ARRAY_SIZE(sdma_reg_list_4_4_2); + + if (!adev->sdma.ip_dump) + return; + + for (i = 0; i < adev->sdma.num_instances; i++) { + instance_offset = i * reg_count; + for (j = 0; j < reg_count; j++) + adev->sdma.ip_dump[instance_offset + j] = + RREG32(sdma_v4_4_2_get_reg_offset(adev, i, + sdma_reg_list_4_4_2[j].reg_offset)); + } +} + const struct amd_ip_funcs sdma_v4_4_2_ip_funcs = { .name = "sdma_v4_4_2", .early_init = sdma_v4_4_2_early_init, @@ -1805,6 +2105,8 @@ const struct amd_ip_funcs sdma_v4_4_2_ip_funcs = { .set_clockgating_state = sdma_v4_4_2_set_clockgating_state, .set_powergating_state = sdma_v4_4_2_set_powergating_state, .get_clockgating_state = sdma_v4_4_2_get_clockgating_state, + .dump_ip_state = sdma_v4_4_2_dump_ip_state, + .print_ip_state = sdma_v4_4_2_print_ip_state, }; static const struct amdgpu_ring_funcs sdma_v4_4_2_ring_funcs = { @@ -1836,6 +2138,8 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_ring_funcs = { .emit_wreg = sdma_v4_4_2_ring_emit_wreg, .emit_reg_wait = sdma_v4_4_2_ring_emit_reg_wait, .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, + .reset = sdma_v4_4_2_reset_queue, + .is_guilty = sdma_v4_4_2_ring_is_guilty, }; static const struct amdgpu_ring_funcs sdma_v4_4_2_page_ring_funcs = { @@ -1867,6 +2171,8 @@ static const struct amdgpu_ring_funcs sdma_v4_4_2_page_ring_funcs = { .emit_wreg = sdma_v4_4_2_ring_emit_wreg, .emit_reg_wait = sdma_v4_4_2_ring_emit_reg_wait, .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, + .reset = sdma_v4_4_2_reset_queue, + .is_guilty = sdma_v4_4_2_page_ring_is_guilty, }; static void sdma_v4_4_2_set_ring_funcs(struct amdgpu_device *adev) @@ -1919,6 +2225,10 @@ static const struct amdgpu_irq_src_funcs sdma_v4_4_2_srbm_write_irq_funcs = { .process = sdma_v4_4_2_process_srbm_write_irq, }; +static const struct amdgpu_irq_src_funcs sdma_v4_4_2_ctxt_empty_irq_funcs = { + .process = sdma_v4_4_2_process_ctxt_empty_irq, +}; + static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev) { adev->sdma.trap_irq.num_types = adev->sdma.num_instances; @@ -1927,6 +2237,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev) adev->sdma.doorbell_invalid_irq.num_types = adev->sdma.num_instances; adev->sdma.pool_timeout_irq.num_types = adev->sdma.num_instances; adev->sdma.srbm_write_irq.num_types = adev->sdma.num_instances; + adev->sdma.ctxt_empty_irq.num_types = adev->sdma.num_instances; adev->sdma.trap_irq.funcs = &sdma_v4_4_2_trap_irq_funcs; adev->sdma.illegal_inst_irq.funcs = &sdma_v4_4_2_illegal_inst_irq_funcs; @@ -1935,6 +2246,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev) adev->sdma.doorbell_invalid_irq.funcs = &sdma_v4_4_2_doorbell_invalid_irq_funcs; adev->sdma.pool_timeout_irq.funcs = &sdma_v4_4_2_pool_timeout_irq_funcs; adev->sdma.srbm_write_irq.funcs = &sdma_v4_4_2_srbm_write_irq_funcs; + adev->sdma.ctxt_empty_irq.funcs = &sdma_v4_4_2_ctxt_empty_irq_funcs; } /** @@ -1944,7 +2256,7 @@ static void sdma_v4_4_2_set_irq_funcs(struct amdgpu_device *adev) * @src_offset: src GPU address * @dst_offset: dst GPU address * @byte_count: number of bytes to xfer - * @tmz: if a secure copy should be used + * @copy_flags: copy flags for the buffers * * Copy GPU buffers using the DMA engine. * Used by the amdgpu ttm implementation to move pages if @@ -1954,11 +2266,11 @@ static void sdma_v4_4_2_emit_copy_buffer(struct amdgpu_ib *ib, uint64_t src_offset, uint64_t dst_offset, uint32_t byte_count, - bool tmz) + uint32_t copy_flags) { ib->ptr[ib->length_dw++] = SDMA_PKT_HEADER_OP(SDMA_OP_COPY) | SDMA_PKT_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) | - SDMA_PKT_COPY_LINEAR_HEADER_TMZ(tmz ? 1 : 0); + SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0); ib->ptr[ib->length_dw++] = byte_count - 1; ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ ib->ptr[ib->length_dw++] = lower_32_bits(src_offset); @@ -2032,6 +2344,40 @@ static void sdma_v4_4_2_set_vm_pte_funcs(struct amdgpu_device *adev) adev->vm_manager.vm_pte_num_scheds = adev->sdma.num_instances; } +/** + * sdma_v4_4_2_update_reset_mask - update reset mask for SDMA + * @adev: Pointer to the AMDGPU device structure + * + * This function update reset mask for SDMA and sets the supported + * reset types based on the IP version and firmware versions. + * + */ +static void sdma_v4_4_2_update_reset_mask(struct amdgpu_device *adev) +{ + /* per queue reset not supported for SRIOV */ + if (amdgpu_sriov_vf(adev)) + return; + + /* + * the user queue relies on MEC fw and pmfw when the sdma queue do reset. + * it needs to check both of them at here to skip old mec and pmfw. + */ + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(9, 4, 3): + case IP_VERSION(9, 4, 4): + if ((adev->gfx.mec_fw_version >= 0xb0) && amdgpu_dpm_reset_sdma_is_supported(adev)) + adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE; + break; + case IP_VERSION(9, 5, 0): + if ((adev->gfx.mec_fw_version >= 0xf) && amdgpu_dpm_reset_sdma_is_supported(adev)) + adev->sdma.supported_reset |= AMDGPU_RESET_TYPE_PER_QUEUE; + break; + default: + break; + } + +} + const struct amdgpu_ip_block_version sdma_v4_4_2_ip_block = { .type = AMD_IP_BLOCK_TYPE_SDMA, .major = 4, @@ -2048,7 +2394,7 @@ static int sdma_v4_4_2_xcp_resume(void *handle, uint32_t inst_mask) if (!amdgpu_sriov_vf(adev)) sdma_v4_4_2_inst_init_golden_registers(adev, inst_mask); - r = sdma_v4_4_2_inst_start(adev, inst_mask); + r = sdma_v4_4_2_inst_start(adev, inst_mask, false); return r; } @@ -2131,7 +2477,7 @@ static void sdma_v4_4_2_inst_query_ras_error_count(struct amdgpu_device *adev, AMDGPU_RAS_ERROR__MULTI_UNCORRECTABLE, &ue_count); - amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, NULL, ue_count); + amdgpu_ras_error_statistic_ue_count(err_data, &mcm_info, ue_count); } static void sdma_v4_4_2_query_ras_error_count(struct amdgpu_device *adev, @@ -2179,35 +2525,41 @@ static const struct amdgpu_ras_block_hw_ops sdma_v4_4_2_ras_hw_ops = { .reset_ras_error_count = sdma_v4_4_2_reset_ras_error_count, }; -static int sdma_v4_4_2_aca_bank_generate_report(struct aca_handle *handle, - struct aca_bank *bank, enum aca_error_type type, - struct aca_bank_report *report, void *data) +static int sdma_v4_4_2_aca_bank_parser(struct aca_handle *handle, struct aca_bank *bank, + enum aca_smu_type type, void *data) { - u64 status, misc0; + struct aca_bank_info info; + u64 misc0; int ret; - status = bank->regs[ACA_REG_IDX_STATUS]; - if ((type == ACA_ERROR_TYPE_UE && - ACA_REG__STATUS__ERRORCODEEXT(status) == ACA_EXTERROR_CODE_FAULT) || - (type == ACA_ERROR_TYPE_CE && - ACA_REG__STATUS__ERRORCODEEXT(status) == ACA_EXTERROR_CODE_CE)) { - - ret = aca_bank_info_decode(bank, &report->info); - if (ret) - return ret; + ret = aca_bank_info_decode(bank, &info); + if (ret) + return ret; - misc0 = bank->regs[ACA_REG_IDX_MISC0]; - report->count[type] = ACA_REG__MISC0__ERRCNT(misc0); + misc0 = bank->regs[ACA_REG_IDX_MISC0]; + switch (type) { + case ACA_SMU_TYPE_UE: + bank->aca_err_type = ACA_ERROR_TYPE_UE; + ret = aca_error_cache_log_bank_error(handle, &info, ACA_ERROR_TYPE_UE, + 1ULL); + break; + case ACA_SMU_TYPE_CE: + bank->aca_err_type = ACA_ERROR_TYPE_CE; + ret = aca_error_cache_log_bank_error(handle, &info, bank->aca_err_type, + ACA_REG__MISC0__ERRCNT(misc0)); + break; + default: + return -EINVAL; } - return 0; + return ret; } /* CODE_SDMA0 - CODE_SDMA4, reference to smu driver if header file */ static int sdma_v4_4_2_err_codes[] = { 33, 34, 35, 36 }; static bool sdma_v4_4_2_aca_bank_is_valid(struct aca_handle *handle, struct aca_bank *bank, - enum aca_error_type type, void *data) + enum aca_smu_type type, void *data) { u32 instlo; @@ -2226,7 +2578,7 @@ static bool sdma_v4_4_2_aca_bank_is_valid(struct aca_handle *handle, struct aca_ } static const struct aca_bank_ops sdma_v4_4_2_aca_bank_ops = { - .aca_bank_generate_report = sdma_v4_4_2_aca_bank_generate_report, + .aca_bank_parser = sdma_v4_4_2_aca_bank_parser, .aca_bank_is_valid = sdma_v4_4_2_aca_bank_is_valid, }; |