From 406bc5633c6b1c7e7a86230db312ee34e785a8f1 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 26 Jul 2018 11:47:59 +0100 Subject: drm/i915: Avoid computing tile_row_size() for untiled objects i915_gem_tile_height() asserts that the object is tiled, but inside the error printer for the selftest we computed the row size regardless of tiling, tripping over the assert. Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20180726104759.8684-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/i915_gem_object.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/selftests/i915_gem_object.c index c69cbd5aed52..d9eca1b02aee 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_object.c @@ -282,7 +282,7 @@ static int check_partial_mapping(struct drm_i915_gem_object *obj, view.partial.offset, view.partial.size, vma->size >> PAGE_SHIFT, - tile_row_pages(obj), + tile->tiling ? tile_row_pages(obj) : 0, vma->fence ? vma->fence->id : -1, tile->tiling, tile->stride, offset >> PAGE_SHIFT, (unsigned int)offset_in_page(offset), -- cgit From ab84a110490d38d40780113a1cdfce03b1cdec13 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 19 Jul 2018 20:47:45 +0100 Subject: drm/i915/selftests: Use a full emulation of a user ppgtt context To test eviction from a ppgtt, we just want a ppgtt i.e. something other than the Global GTT which is shared and used by the kernel for HW features like fencing and scanout. However, we also need it to pass !i915_is_ggtt() and the simplest way is to emulate a full user context rather than the internal kernel context that is used for the GGTT. Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20180719194746.19111-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/intel_hangcheck.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c index 65d66cdedd26..b2d6d15f025a 100644 --- a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c @@ -1144,19 +1144,27 @@ static int igt_reset_evict_ppgtt(void *arg) { struct drm_i915_private *i915 = arg; struct i915_gem_context *ctx; + struct drm_file *file; int err; + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + mutex_lock(&i915->drm.struct_mutex); - ctx = kernel_context(i915); + ctx = live_context(i915, file); mutex_unlock(&i915->drm.struct_mutex); - if (IS_ERR(ctx)) - return PTR_ERR(ctx); + if (IS_ERR(ctx)) { + err = PTR_ERR(ctx); + goto out; + } err = 0; if (ctx->ppgtt) /* aliasing == global gtt locking, covered above */ err = __igt_reset_evict_vma(i915, &ctx->ppgtt->vm); - kernel_context_close(ctx); +out: + mock_file_free(i915, file); return err; } -- cgit From 6dc17d69f83ec315157b76fbf47d4379e1266cef Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 19 Jul 2018 20:47:46 +0100 Subject: drm/i915/selftests: Exercise resetting in the middle of a wait-on-fence On older HW, gen2/3, fence registers are used for detiling GPU commands and as such changing those registers requires serialisation with the requests on the GPU. Anything running on the GPU is subject to a hang, and so we must be able to recover cleanly in the middle of a stuck wait on a fence register. We can simulate using the fence on the GPU simply by marking the fence as active on the request for this vma, the interface being common to all gen, thus broadening the test. Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20180719194746.19111-2-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/intel_hangcheck.c | 85 +++++++++++++++++++++--- 1 file changed, 77 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c index b2d6d15f025a..db378226ac10 100644 --- a/drivers/gpu/drm/i915/selftests/intel_hangcheck.c +++ b/drivers/gpu/drm/i915/selftests/intel_hangcheck.c @@ -1018,8 +1018,41 @@ static int evict_vma(void *data) return err; } +static int evict_fence(void *data) +{ + struct evict_vma *arg = data; + struct drm_i915_private *i915 = arg->vma->vm->i915; + int err; + + complete(&arg->completion); + + mutex_lock(&i915->drm.struct_mutex); + + /* Mark the fence register as dirty to force the mmio update. */ + err = i915_gem_object_set_tiling(arg->vma->obj, I915_TILING_Y, 512); + if (err) { + pr_err("Invalid Y-tiling settings; err:%d\n", err); + goto out_unlock; + } + + err = i915_vma_pin_fence(arg->vma); + if (err) { + pr_err("Unable to pin Y-tiled fence; err:%d\n", err); + goto out_unlock; + } + + i915_vma_unpin_fence(arg->vma); + +out_unlock: + mutex_unlock(&i915->drm.struct_mutex); + + return err; +} + static int __igt_reset_evict_vma(struct drm_i915_private *i915, - struct i915_address_space *vm) + struct i915_address_space *vm, + int (*fn)(void *), + unsigned int flags) { struct drm_i915_gem_object *obj; struct task_struct *tsk = NULL; @@ -1040,12 +1073,20 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, if (err) goto unlock; - obj = i915_gem_object_create_internal(i915, PAGE_SIZE); + obj = i915_gem_object_create_internal(i915, SZ_1M); if (IS_ERR(obj)) { err = PTR_ERR(obj); goto fini; } + if (flags & EXEC_OBJECT_NEEDS_FENCE) { + err = i915_gem_object_set_tiling(obj, I915_TILING_X, 512); + if (err) { + pr_err("Invalid X-tiling settings; err:%d\n", err); + goto out_obj; + } + } + arg.vma = i915_vma_instance(obj, vm, NULL); if (IS_ERR(arg.vma)) { err = PTR_ERR(arg.vma); @@ -1059,11 +1100,28 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, } err = i915_vma_pin(arg.vma, 0, 0, - i915_vma_is_ggtt(arg.vma) ? PIN_GLOBAL : PIN_USER); - if (err) + i915_vma_is_ggtt(arg.vma) ? + PIN_GLOBAL | PIN_MAPPABLE : + PIN_USER); + if (err) { + i915_request_add(rq); goto out_obj; + } + + if (flags & EXEC_OBJECT_NEEDS_FENCE) { + err = i915_vma_pin_fence(arg.vma); + if (err) { + pr_err("Unable to pin X-tiled fence; err:%d\n", err); + i915_vma_unpin(arg.vma); + i915_request_add(rq); + goto out_obj; + } + } - err = i915_vma_move_to_active(arg.vma, rq, EXEC_OBJECT_WRITE); + err = i915_vma_move_to_active(arg.vma, rq, flags); + + if (flags & EXEC_OBJECT_NEEDS_FENCE) + i915_vma_unpin_fence(arg.vma); i915_vma_unpin(arg.vma); i915_request_get(rq); @@ -1086,7 +1144,7 @@ static int __igt_reset_evict_vma(struct drm_i915_private *i915, init_completion(&arg.completion); - tsk = kthread_run(evict_vma, &arg, "igt/evict_vma"); + tsk = kthread_run(fn, &arg, "igt/evict_vma"); if (IS_ERR(tsk)) { err = PTR_ERR(tsk); tsk = NULL; @@ -1137,7 +1195,8 @@ static int igt_reset_evict_ggtt(void *arg) { struct drm_i915_private *i915 = arg; - return __igt_reset_evict_vma(i915, &i915->ggtt.vm); + return __igt_reset_evict_vma(i915, &i915->ggtt.vm, + evict_vma, EXEC_OBJECT_WRITE); } static int igt_reset_evict_ppgtt(void *arg) @@ -1161,13 +1220,22 @@ static int igt_reset_evict_ppgtt(void *arg) err = 0; if (ctx->ppgtt) /* aliasing == global gtt locking, covered above */ - err = __igt_reset_evict_vma(i915, &ctx->ppgtt->vm); + err = __igt_reset_evict_vma(i915, &ctx->ppgtt->vm, + evict_vma, EXEC_OBJECT_WRITE); out: mock_file_free(i915, file); return err; } +static int igt_reset_evict_fence(void *arg) +{ + struct drm_i915_private *i915 = arg; + + return __igt_reset_evict_vma(i915, &i915->ggtt.vm, + evict_fence, EXEC_OBJECT_NEEDS_FENCE); +} + static int wait_for_others(struct drm_i915_private *i915, struct intel_engine_cs *exclude) { @@ -1417,6 +1485,7 @@ int intel_hangcheck_live_selftests(struct drm_i915_private *i915) SUBTEST(igt_reset_wait), SUBTEST(igt_reset_evict_ggtt), SUBTEST(igt_reset_evict_ppgtt), + SUBTEST(igt_reset_evict_fence), SUBTEST(igt_handle_error), }; bool saved_hangcheck; -- cgit From f6844a85e0c96a55c61fa3e611f414999b11e4de Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 30 Jul 2018 08:53:51 +0100 Subject: drm/i915/selftests: Replace opencoded clflush with drm_clflush_virt_range We occasionally see that the clflush prior to a read of GPU data is returning stale data, reminiscent of much earlier bugs fixed by adding a second clflush for serialisation. As drm_clflush_virt_range() already supplies the workaround, use it rather than open code the clflush instruction. References: 396f5d62d1a5 ("drm: Restore double clflush on the last partial cacheline") Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20180730075351.15569-3-chris@chris-wilson.co.uk --- .../gpu/drm/i915/selftests/i915_gem_coherency.c | 38 ++++++++++------------ 1 file changed, 17 insertions(+), 21 deletions(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c index 3a095c37c120..4e6a221063ac 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_coherency.c @@ -33,7 +33,8 @@ static int cpu_set(struct drm_i915_gem_object *obj, { unsigned int needs_clflush; struct page *page; - u32 *map; + void *map; + u32 *cpu; int err; err = i915_gem_obj_prepare_shmem_write(obj, &needs_clflush); @@ -42,24 +43,19 @@ static int cpu_set(struct drm_i915_gem_object *obj, page = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT); map = kmap_atomic(page); + cpu = map + offset_in_page(offset); - if (needs_clflush & CLFLUSH_BEFORE) { - mb(); - clflush(map+offset_in_page(offset) / sizeof(*map)); - mb(); - } + if (needs_clflush & CLFLUSH_BEFORE) + drm_clflush_virt_range(cpu, sizeof(*cpu)); - map[offset_in_page(offset) / sizeof(*map)] = v; + *cpu = v; - if (needs_clflush & CLFLUSH_AFTER) { - mb(); - clflush(map+offset_in_page(offset) / sizeof(*map)); - mb(); - } + if (needs_clflush & CLFLUSH_AFTER) + drm_clflush_virt_range(cpu, sizeof(*cpu)); kunmap_atomic(map); - i915_gem_obj_finish_shmem_access(obj); + return 0; } @@ -69,7 +65,8 @@ static int cpu_get(struct drm_i915_gem_object *obj, { unsigned int needs_clflush; struct page *page; - u32 *map; + void *map; + u32 *cpu; int err; err = i915_gem_obj_prepare_shmem_read(obj, &needs_clflush); @@ -78,17 +75,16 @@ static int cpu_get(struct drm_i915_gem_object *obj, page = i915_gem_object_get_page(obj, offset >> PAGE_SHIFT); map = kmap_atomic(page); + cpu = map + offset_in_page(offset); - if (needs_clflush & CLFLUSH_BEFORE) { - mb(); - clflush(map+offset_in_page(offset) / sizeof(*map)); - mb(); - } + if (needs_clflush & CLFLUSH_BEFORE) + drm_clflush_virt_range(cpu, sizeof(*cpu)); - *v = map[offset_in_page(offset) / sizeof(*map)]; - kunmap_atomic(map); + *v = *cpu; + kunmap_atomic(map); i915_gem_obj_finish_shmem_access(obj); + return 0; } -- cgit From e6a59382924e2d007b554a2aebcd4445ebb01fef Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 6 Aug 2018 15:46:04 +0100 Subject: drm/i915/selftests: Unconditionally do a chipset flush before emit_bb_start Experience teaches us over and over again that coherency on Baytrail requires the odd heavy hammer, and in particular clflush alone is not enough to guarrantee that writes from the CPU are picked up by the CS. Do as we do elsewhere and ensure we have an unconditional i915_gem_chipset_flush() after writing to memory and submitting a batch to HW. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=107499 Signed-off-by: Chris Wilson Reviewed-by: Matthew Auld Link: https://patchwork.freedesktop.org/patch/msgid/20180806144604.8346-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/huge_pages.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/huge_pages.c b/drivers/gpu/drm/i915/selftests/huge_pages.c index 7efb326badcd..e272127783fe 100644 --- a/drivers/gpu/drm/i915/selftests/huge_pages.c +++ b/drivers/gpu/drm/i915/selftests/huge_pages.c @@ -906,7 +906,11 @@ gpu_write_dw(struct i915_vma *vma, u64 offset, u32 val) if (IS_ERR(obj)) return ERR_CAST(obj); - cmd = i915_gem_object_pin_map(obj, I915_MAP_WB); + err = i915_gem_object_set_to_wc_domain(obj, true); + if (err) + goto err; + + cmd = i915_gem_object_pin_map(obj, I915_MAP_WC); if (IS_ERR(cmd)) { err = PTR_ERR(cmd); goto err; @@ -936,13 +940,10 @@ gpu_write_dw(struct i915_vma *vma, u64 offset, u32 val) } *cmd = MI_BATCH_BUFFER_END; + i915_gem_chipset_flush(i915); i915_gem_object_unpin_map(obj); - err = i915_gem_object_set_to_gtt_domain(obj, false); - if (err) - goto err; - batch = i915_vma_instance(obj, vma->vm, NULL); if (IS_ERR(batch)) { err = PTR_ERR(batch); -- cgit From d60996ab430c8a6033a0944c068edc5ec5becb9b Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 8 Aug 2018 22:08:42 +0100 Subject: drm/i915: Restore user forcewake domains across suspend On suspend, we cancel the automatic forcewake and clear all other sources of forcewake so the machine can sleep before we do suspend. However, we expose the forcewake to userspace (only via debugfs, but nevertheless we do) and want to restore that upon resume or else our accounting will be off and we may not acquire the forcewake before we use it. So record which domains we cleared on suspend and reacquire them early on resume. v2: Hold the spinlock to appease our sanitychecks v3: s/fw_domains_user/fw_domains_saved/ to convey intent more clearly Reported-by: Imre Deak Fixes: b8473050805f ("drm/i915: Fix forcewake active domain tracking") Signed-off-by: Chris Wilson Cc: Tvrtko Ursulin Cc: Mika Kuoppala Cc: Imre Deak Reviewed-by: Imre Deak Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20180808210842.3555-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/intel_uncore.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/intel_uncore.c b/drivers/gpu/drm/i915/selftests/intel_uncore.c index 47bc5b2ddb56..81d9d31042a9 100644 --- a/drivers/gpu/drm/i915/selftests/intel_uncore.c +++ b/drivers/gpu/drm/i915/selftests/intel_uncore.c @@ -160,7 +160,7 @@ static int intel_uncore_check_forcewake_domains(struct drm_i915_private *dev_pri i915_reg_t reg = { offset }; iosf_mbi_punit_acquire(); - intel_uncore_forcewake_reset(dev_priv, false); + intel_uncore_forcewake_reset(dev_priv); iosf_mbi_punit_release(); check_for_unclaimed_mmio(dev_priv); -- cgit From 7b5ee80a5da3ea44c5abff48e3621135ae9d8177 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 9 Aug 2018 07:34:49 +0100 Subject: drm/i915/selftests: Hold rpm for unparking The call to i915_gem_unpark() checks that we hold a rpm wakeref before taking a long term wakeref for i915->gt.awake. We should therefore make sure we do hold the wakeref when directly calling unpark to disable the retire worker. Fixes: 932cac10c8fb ("drm/i915/selftests: Prevent background reaping of active objects") Signed-off-by: Chris Wilson Cc: Mika Kuoppala Cc: Matthew Auld Reviewed-by: Mika Kuoppala Link: https://patchwork.freedesktop.org/patch/msgid/20180809063449.4474-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/i915_gem_object.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_object.c b/drivers/gpu/drm/i915/selftests/i915_gem_object.c index d9eca1b02aee..6d3516d5bff9 100644 --- a/drivers/gpu/drm/i915/selftests/i915_gem_object.c +++ b/drivers/gpu/drm/i915/selftests/i915_gem_object.c @@ -499,6 +499,19 @@ static bool assert_mmap_offset(struct drm_i915_private *i915, return err == expected; } +static void disable_retire_worker(struct drm_i915_private *i915) +{ + mutex_lock(&i915->drm.struct_mutex); + if (!i915->gt.active_requests++) { + intel_runtime_pm_get(i915); + i915_gem_unpark(i915); + intel_runtime_pm_put(i915); + } + mutex_unlock(&i915->drm.struct_mutex); + cancel_delayed_work_sync(&i915->gt.retire_work); + cancel_delayed_work_sync(&i915->gt.idle_work); +} + static int igt_mmap_offset_exhaustion(void *arg) { struct drm_i915_private *i915 = arg; @@ -509,12 +522,7 @@ static int igt_mmap_offset_exhaustion(void *arg) int loop, err; /* Disable background reaper */ - mutex_lock(&i915->drm.struct_mutex); - if (!i915->gt.active_requests++) - i915_gem_unpark(i915); - mutex_unlock(&i915->drm.struct_mutex); - cancel_delayed_work_sync(&i915->gt.retire_work); - cancel_delayed_work_sync(&i915->gt.idle_work); + disable_retire_worker(i915); GEM_BUG_ON(!i915->gt.awake); /* Trim the device mmap space to only a page */ -- cgit From 5382bed38f09636330fd119ca2c83d738a551540 Mon Sep 17 00:00:00 2001 From: Daniele Ceraolo Spurio Date: Mon, 27 Aug 2018 15:36:14 -0700 Subject: drm/i915/selftests: ring all doorbells in igt_guc_doorbells We currently verify that all doorbells can be registered with GuC and HW but don't check that all works as expected after a db ring. Do a nop ring of all doorbells to make sure we haven't misprogrammed any WQ or stage descriptor data. This will also help validating upcoming changes in the db programming flow. Cc: Michel Thierry Cc: Michal Wajdeczko Signed-off-by: Daniele Ceraolo Spurio Reviewed-by: Michel Thierry Acked-by: Katarzyna Dec Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20180827223614.22789-1-daniele.ceraolospurio@intel.com --- drivers/gpu/drm/i915/selftests/intel_guc.c | 38 ++++++++++++++++++++++++++++++ 1 file changed, 38 insertions(+) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/intel_guc.c b/drivers/gpu/drm/i915/selftests/intel_guc.c index 407c98fb9170..90ba88c972cf 100644 --- a/drivers/gpu/drm/i915/selftests/intel_guc.c +++ b/drivers/gpu/drm/i915/selftests/intel_guc.c @@ -65,6 +65,40 @@ static int check_all_doorbells(struct intel_guc *guc) return 0; } +static int ring_doorbell_nop(struct intel_guc_client *client) +{ + struct guc_process_desc *desc = __get_process_desc(client); + int err; + + client->use_nop_wqi = true; + + spin_lock_irq(&client->wq_lock); + + guc_wq_item_append(client, 0, 0, 0, 0); + guc_ring_doorbell(client); + + spin_unlock_irq(&client->wq_lock); + + client->use_nop_wqi = false; + + /* if there are no issues GuC will update the WQ head and keep the + * WQ in active status + */ + err = wait_for(READ_ONCE(desc->head) == READ_ONCE(desc->tail), 10); + if (err) { + pr_err("doorbell %u ring failed!\n", client->doorbell_id); + return -EIO; + } + + if (desc->wq_status != WQ_STATUS_ACTIVE) { + pr_err("doorbell %u ring put WQ in bad state (%u)!\n", + client->doorbell_id, desc->wq_status); + return -EIO; + } + + return 0; +} + /* * Basic client sanity check, handy to validate create_clients. */ @@ -332,6 +366,10 @@ static int igt_guc_doorbells(void *arg) err = check_all_doorbells(guc); if (err) goto out; + + err = ring_doorbell_nop(clients[i]); + if (err) + goto out; } out: -- cgit From 3f51b7e1f36a37cfc6ed281a231485e4e6b511c3 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 30 Aug 2018 14:48:06 +0100 Subject: drm/i915/selftests: Add a simple exerciser for suspend/hibernate MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Although we cannot do a full system-level test of suspend/hibernate from deep with the kernel selftests, we can exercise the GEM subsystem in isolation and simulate the external effects (such as losing stolen contents and trashing the register state). v2: Don't forget to hold rpm v3: Suspend the GTT mappings, and more rpm! References: https://bugs.freedesktop.org/show_bug.cgi?id=96526 References: 5ab57c702069 ("drm/i915: Flush logical context image out to memory upon suspend") Signed-off-by: Chris Wilson Cc: Jakub Bartmiński Cc: Matthew Auld Cc: Joonas Lahtinen Reviewed-by: Jakub Bartmiński Link: https://patchwork.freedesktop.org/patch/msgid/20180830134806.21939-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/i915_gem.c | 221 +++++++++++++++++++++ .../gpu/drm/i915/selftests/i915_live_selftests.h | 1 + 2 files changed, 222 insertions(+) create mode 100644 drivers/gpu/drm/i915/selftests/i915_gem.c (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c new file mode 100644 index 000000000000..d0aa19d17653 --- /dev/null +++ b/drivers/gpu/drm/i915/selftests/i915_gem.c @@ -0,0 +1,221 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2018 Intel Corporation + */ + +#include + +#include "../i915_selftest.h" + +#include "mock_context.h" +#include "igt_flush_test.h" + +static int switch_to_context(struct drm_i915_private *i915, + struct i915_gem_context *ctx) +{ + struct intel_engine_cs *engine; + enum intel_engine_id id; + int err = 0; + + intel_runtime_pm_get(i915); + + for_each_engine(engine, i915, id) { + struct i915_request *rq; + + rq = i915_request_alloc(engine, ctx); + if (IS_ERR(rq)) { + err = PTR_ERR(rq); + break; + } + + i915_request_add(rq); + } + + intel_runtime_pm_put(i915); + + return err; +} + +static void trash_stolen(struct drm_i915_private *i915) +{ + struct i915_ggtt *ggtt = &i915->ggtt; + const u64 slot = ggtt->error_capture.start; + const resource_size_t size = resource_size(&i915->dsm); + unsigned long page; + u32 prng = 0x12345678; + + for (page = 0; page < size; page += PAGE_SIZE) { + const dma_addr_t dma = i915->dsm.start + page; + u32 __iomem *s; + int x; + + ggtt->vm.insert_page(&ggtt->vm, dma, slot, I915_CACHE_NONE, 0); + + s = io_mapping_map_atomic_wc(&ggtt->iomap, slot); + for (x = 0; x < PAGE_SIZE / sizeof(u32); x++) { + prng = next_pseudo_random32(prng); + iowrite32(prng, &s[x]); + } + io_mapping_unmap_atomic(s); + } + + ggtt->vm.clear_range(&ggtt->vm, slot, PAGE_SIZE); +} + +static void simulate_hibernate(struct drm_i915_private *i915) +{ + intel_runtime_pm_get(i915); + + /* + * As a final sting in the tail, invalidate stolen. Under a real S4, + * stolen is lost and needs to be refilled on resume. However, under + * CI we merely do S4-device testing (as full S4 is too unreliable + * for automated testing across a cluster), so to simulate the effect + * of stolen being trashed across S4, we trash it ourselves. + */ + trash_stolen(i915); + + intel_runtime_pm_put(i915); +} + +static int pm_prepare(struct drm_i915_private *i915) +{ + int err = 0; + + if (i915_gem_suspend(i915)) { + pr_err("i915_gem_suspend failed\n"); + err = -EINVAL; + } + + return err; +} + +static void pm_suspend(struct drm_i915_private *i915) +{ + intel_runtime_pm_get(i915); + + i915_gem_suspend_gtt_mappings(i915); + i915_gem_suspend_late(i915); + + intel_runtime_pm_put(i915); +} + +static void pm_hibernate(struct drm_i915_private *i915) +{ + intel_runtime_pm_get(i915); + + i915_gem_suspend_gtt_mappings(i915); + + i915_gem_freeze(i915); + i915_gem_freeze_late(i915); + + intel_runtime_pm_put(i915); +} + +static void pm_resume(struct drm_i915_private *i915) +{ + /* + * Both suspend and hibernate follow the same wakeup path and assume + * that runtime-pm just works. + */ + intel_runtime_pm_get(i915); + + intel_engines_sanitize(i915); + i915_gem_sanitize(i915); + i915_gem_resume(i915); + + intel_runtime_pm_put(i915); +} + +static int igt_gem_suspend(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct i915_gem_context *ctx; + struct drm_file *file; + int err; + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + + err = -ENOMEM; + mutex_lock(&i915->drm.struct_mutex); + ctx = live_context(i915, file); + if (!IS_ERR(ctx)) + err = switch_to_context(i915, ctx); + mutex_unlock(&i915->drm.struct_mutex); + if (err) + goto out; + + err = pm_prepare(i915); + if (err) + goto out; + + pm_suspend(i915); + + /* Here be dragons! Note that with S3RST any S3 may become S4! */ + simulate_hibernate(i915); + + pm_resume(i915); + + mutex_lock(&i915->drm.struct_mutex); + err = switch_to_context(i915, ctx); + if (igt_flush_test(i915, I915_WAIT_LOCKED)) + err = -EIO; + mutex_unlock(&i915->drm.struct_mutex); +out: + mock_file_free(i915, file); + return err; +} + +static int igt_gem_hibernate(void *arg) +{ + struct drm_i915_private *i915 = arg; + struct i915_gem_context *ctx; + struct drm_file *file; + int err; + + file = mock_file(i915); + if (IS_ERR(file)) + return PTR_ERR(file); + + err = -ENOMEM; + mutex_lock(&i915->drm.struct_mutex); + ctx = live_context(i915, file); + if (!IS_ERR(ctx)) + err = switch_to_context(i915, ctx); + mutex_unlock(&i915->drm.struct_mutex); + if (err) + goto out; + + err = pm_prepare(i915); + if (err) + goto out; + + pm_hibernate(i915); + + /* Here be dragons! */ + simulate_hibernate(i915); + + pm_resume(i915); + + mutex_lock(&i915->drm.struct_mutex); + err = switch_to_context(i915, ctx); + if (igt_flush_test(i915, I915_WAIT_LOCKED)) + err = -EIO; + mutex_unlock(&i915->drm.struct_mutex); +out: + mock_file_free(i915, file); + return err; +} + +int i915_gem_live_selftests(struct drm_i915_private *i915) +{ + static const struct i915_subtest tests[] = { + SUBTEST(igt_gem_suspend), + SUBTEST(igt_gem_hibernate), + }; + + return i915_subtests(tests, i915); +} diff --git a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h index a00e2bd08bce..a15713cae3b3 100644 --- a/drivers/gpu/drm/i915/selftests/i915_live_selftests.h +++ b/drivers/gpu/drm/i915/selftests/i915_live_selftests.h @@ -17,6 +17,7 @@ selftest(objects, i915_gem_object_live_selftests) selftest(dmabuf, i915_gem_dmabuf_live_selftests) selftest(coherency, i915_gem_coherency_live_selftests) selftest(gtt, i915_gem_gtt_live_selftests) +selftest(gem, i915_gem_live_selftests) selftest(evict, i915_gem_evict_live_selftests) selftest(hugepages, i915_gem_huge_page_live_selftests) selftest(contexts, i915_gem_context_live_selftests) -- cgit From 48e905048f39ae97bd08dbbbc78a848d1d555d80 Mon Sep 17 00:00:00 2001 From: Tvrtko Ursulin Date: Fri, 31 Aug 2018 15:36:43 +0100 Subject: drm/i915: Explicitly mark Global GTT address spaces So far we have been relying on vm->file pointer being NULL to declare something GGTT. This has the unfortunate consequence that the default kernel context is also declared GGTT and interferes with the following patch which wants to instantiate VMA's and execute requests against the kernel context. Change the is_ggtt test to use an explicit flag in struct address_space to solve this issue. Note that the bit used is free since there is an alignment hole in the struct. v2: * Mark mock ggtt. Signed-off-by: Tvrtko Ursulin Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Chris Wilson Link: https://patchwork.freedesktop.org/patch/msgid/20180831143643.12366-1-tvrtko.ursulin@linux.intel.com --- drivers/gpu/drm/i915/selftests/mock_gtt.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/mock_gtt.c b/drivers/gpu/drm/i915/selftests/mock_gtt.c index a140ea5c3a7c..6ae418c76015 100644 --- a/drivers/gpu/drm/i915/selftests/mock_gtt.c +++ b/drivers/gpu/drm/i915/selftests/mock_gtt.c @@ -118,6 +118,8 @@ void mock_init_ggtt(struct drm_i915_private *i915) ggtt->vm.vma_ops.clear_pages = clear_pages; i915_address_space_init(&ggtt->vm, i915); + + ggtt->vm.is_ggtt = true; } void mock_fini_ggtt(struct drm_i915_private *i915) -- cgit From 288f1ced5e24abe3e768224f701a205c3a7e16f9 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Tue, 4 Sep 2018 16:31:17 +0100 Subject: drm/i915: Reduce context HW ID lifetime Future gen reduce the number of bits we will have available to differentiate between contexts, so reduce the lifetime of the ID assignment from that of the context to its current active cycle (i.e. only while it is pinned for use by the HW, will it have a constant ID). This means that instead of a max of 2k allocated contexts (worst case before fun with bit twiddling), we instead have a limit of 2k in flight contexts (minus a few that have been pinned by the kernel or by perf). To reduce the number of contexts id we require, we allocate a context id on first and mark it as pinned for as long as the GEM context itself is, that is we keep it pinned it while active on each engine. If we exhaust our context id space, then we try to reclaim an id from an idle context. In the extreme case where all context ids are pinned by active contexts, we force the system to idle in order to recover ids. We cannot reduce the scope of an HW-ID to an engine (allowing the same gem_context to have different ids on each engine) as in the future we will need to preassign an id before we know which engine the context is being executed on. v2: Improved commentary (Tvrtko) [I tried at least] References: https://bugs.freedesktop.org/show_bug.cgi?id=107788 Signed-off-by: Chris Wilson Cc: Lionel Landwerlin Cc: Tvrtko Ursulin Cc: Mika Kuoppala Cc: Michel Thierry Cc: Michal Wajdeczko Cc: Daniele Ceraolo Spurio Reviewed-by: Tvrtko Ursulin Link: https://patchwork.freedesktop.org/patch/msgid/20180904153117.3907-1-chris@chris-wilson.co.uk --- drivers/gpu/drm/i915/selftests/mock_context.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) (limited to 'drivers/gpu/drm/i915/selftests') diff --git a/drivers/gpu/drm/i915/selftests/mock_context.c b/drivers/gpu/drm/i915/selftests/mock_context.c index 8904f1ce64e3..d937bdff26f9 100644 --- a/drivers/gpu/drm/i915/selftests/mock_context.c +++ b/drivers/gpu/drm/i915/selftests/mock_context.c @@ -43,6 +43,7 @@ mock_context(struct drm_i915_private *i915, INIT_RADIX_TREE(&ctx->handles_vma, GFP_KERNEL); INIT_LIST_HEAD(&ctx->handles_list); + INIT_LIST_HEAD(&ctx->hw_id_link); for (n = 0; n < ARRAY_SIZE(ctx->__engine); n++) { struct intel_context *ce = &ctx->__engine[n]; @@ -50,11 +51,9 @@ mock_context(struct drm_i915_private *i915, ce->gem_context = ctx; } - ret = ida_simple_get(&i915->contexts.hw_ida, - 0, MAX_CONTEXT_HW_ID, GFP_KERNEL); + ret = i915_gem_context_pin_hw_id(ctx); if (ret < 0) goto err_handles; - ctx->hw_id = ret; if (name) { ctx->name = kstrdup(name, GFP_KERNEL); @@ -85,11 +84,7 @@ void mock_context_close(struct i915_gem_context *ctx) void mock_init_contexts(struct drm_i915_private *i915) { - INIT_LIST_HEAD(&i915->contexts.list); - ida_init(&i915->contexts.hw_ida); - - INIT_WORK(&i915->contexts.free_work, contexts_free_worker); - init_llist_head(&i915->contexts.free_list); + init_contexts(i915); } struct i915_gem_context * -- cgit