All of lore.kernel.org
 help / color / mirror / Atom feed
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
To: intel-gfx@lists.freedesktop.org
Subject: [Intel-gfx] [PATCH 03/21] drm/i915: Remove locking from i915_gem_object_prepare_read/write
Date: Tue, 24 Mar 2020 17:51:28 +0100	[thread overview]
Message-ID: <20200324165146.1032624-3-maarten.lankhorst@linux.intel.com> (raw)
In-Reply-To: <20200324165146.1032624-1-maarten.lankhorst@linux.intel.com>

Execbuffer submission will perform its own WW locking, and we
cannot rely on the implicit lock there.

This also makes it clear that the GVT code will get a lockdep splat when
multiple batchbuffer shadows need to be performed in the same instance,
fix that up.

Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_domain.c    | 20 ++++++-------------
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 13 ++++++++++--
 drivers/gpu/drm/i915/gem/i915_gem_object.h    |  1 -
 .../gpu/drm/i915/gem/selftests/huge_pages.c   |  5 ++++-
 .../i915/gem/selftests/i915_gem_coherency.c   | 14 +++++++++----
 .../drm/i915/gem/selftests/i915_gem_context.c | 12 ++++++++---
 drivers/gpu/drm/i915/gt/intel_renderstate.c   |  5 ++++-
 drivers/gpu/drm/i915/gvt/cmd_parser.c         |  9 ++++++++-
 drivers/gpu/drm/i915/i915_gem.c               | 20 +++++++++++++++++--
 9 files changed, 70 insertions(+), 29 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
index f4602faa8db9..e9d3b587f562 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c
@@ -581,19 +581,17 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
 	if (!i915_gem_object_has_struct_page(obj))
 		return -ENODEV;
 
-	ret = i915_gem_object_lock_interruptible(obj, NULL);
-	if (ret)
-		return ret;
+	assert_object_held(obj);
 
 	ret = i915_gem_object_wait(obj,
 				   I915_WAIT_INTERRUPTIBLE,
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
-		goto err_unlock;
+		return ret;
 
 	ret = i915_gem_object_pin_pages(obj);
 	if (ret)
-		goto err_unlock;
+		return ret;
 
 	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ ||
 	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
@@ -621,8 +619,6 @@ int i915_gem_object_prepare_read(struct drm_i915_gem_object *obj,
 
 err_unpin:
 	i915_gem_object_unpin_pages(obj);
-err_unlock:
-	i915_gem_object_unlock(obj);
 	return ret;
 }
 
@@ -635,20 +631,18 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
 	if (!i915_gem_object_has_struct_page(obj))
 		return -ENODEV;
 
-	ret = i915_gem_object_lock_interruptible(obj, NULL);
-	if (ret)
-		return ret;
+	assert_object_held(obj);
 
 	ret = i915_gem_object_wait(obj,
 				   I915_WAIT_INTERRUPTIBLE |
 				   I915_WAIT_ALL,
 				   MAX_SCHEDULE_TIMEOUT);
 	if (ret)
-		goto err_unlock;
+		return ret;
 
 	ret = i915_gem_object_pin_pages(obj);
 	if (ret)
-		goto err_unlock;
+		return ret;
 
 	if (obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_WRITE ||
 	    !static_cpu_has(X86_FEATURE_CLFLUSH)) {
@@ -685,7 +679,5 @@ int i915_gem_object_prepare_write(struct drm_i915_gem_object *obj,
 
 err_unpin:
 	i915_gem_object_unpin_pages(obj);
-err_unlock:
-	i915_gem_object_unlock(obj);
 	return ret;
 }
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 28d5a0ce79e8..6e865a034d11 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -960,11 +960,14 @@ static void reloc_cache_reset(struct reloc_cache *cache)
 
 	vaddr = unmask_page(cache->vaddr);
 	if (cache->vaddr & KMAP) {
+		struct drm_i915_gem_object *obj =
+			(struct drm_i915_gem_object *)cache->node.mm;
 		if (cache->vaddr & CLFLUSH_AFTER)
 			mb();
 
 		kunmap_atomic(vaddr);
-		i915_gem_object_finish_access((struct drm_i915_gem_object *)cache->node.mm);
+		i915_gem_object_finish_access(obj);
+		i915_gem_object_unlock(obj);
 	} else {
 		struct i915_ggtt *ggtt = cache_to_ggtt(cache);
 
@@ -999,10 +1002,16 @@ static void *reloc_kmap(struct drm_i915_gem_object *obj,
 		unsigned int flushes;
 		int err;
 
-		err = i915_gem_object_prepare_write(obj, &flushes);
+		err = i915_gem_object_lock_interruptible(obj, NULL);
 		if (err)
 			return ERR_PTR(err);
 
+		err = i915_gem_object_prepare_write(obj, &flushes);
+		if (err) {
+			i915_gem_object_unlock(obj);
+			return ERR_PTR(err);
+		}
+
 		BUILD_BUG_ON(KMAP & CLFLUSH_FLAGS);
 		BUILD_BUG_ON((KMAP | CLFLUSH_FLAGS) & PAGE_MASK);
 
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 5103067269b0..11b8e2735071 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -434,7 +434,6 @@ static inline void
 i915_gem_object_finish_access(struct drm_i915_gem_object *obj)
 {
 	i915_gem_object_unpin_pages(obj);
-	i915_gem_object_unlock(obj);
 }
 
 static inline struct intel_engine_cs *
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 2a2f33228b3d..36db96f902cc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -991,9 +991,10 @@ __cpu_check_shmem(struct drm_i915_gem_object *obj, u32 dword, u32 val)
 	unsigned long n;
 	int err;
 
+	i915_gem_object_lock(obj, NULL);
 	err = i915_gem_object_prepare_read(obj, &needs_flush);
 	if (err)
-		return err;
+		goto err_unlock;
 
 	for (n = 0; n < obj->base.size >> PAGE_SHIFT; ++n) {
 		u32 *ptr = kmap_atomic(i915_gem_object_get_page(obj, n));
@@ -1013,6 +1014,8 @@ __cpu_check_shmem(struct drm_i915_gem_object *obj, u32 dword, u32 val)
 	}
 
 	i915_gem_object_finish_access(obj);
+err_unlock:
+	i915_gem_object_unlock(obj);
 
 	return err;
 }
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
index f5a0c7b1e113..99f8466a108a 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_coherency.c
@@ -27,9 +27,10 @@ static int cpu_set(struct context *ctx, unsigned long offset, u32 v)
 	u32 *cpu;
 	int err;
 
+	i915_gem_object_lock(ctx->obj, NULL);
 	err = i915_gem_object_prepare_write(ctx->obj, &needs_clflush);
 	if (err)
-		return err;
+		goto out;
 
 	page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT);
 	map = kmap_atomic(page);
@@ -46,7 +47,9 @@ static int cpu_set(struct context *ctx, unsigned long offset, u32 v)
 	kunmap_atomic(map);
 	i915_gem_object_finish_access(ctx->obj);
 
-	return 0;
+out:
+	i915_gem_object_unlock(ctx->obj);
+	return err;
 }
 
 static int cpu_get(struct context *ctx, unsigned long offset, u32 *v)
@@ -57,9 +60,10 @@ static int cpu_get(struct context *ctx, unsigned long offset, u32 *v)
 	u32 *cpu;
 	int err;
 
+	i915_gem_object_lock(ctx->obj, NULL);
 	err = i915_gem_object_prepare_read(ctx->obj, &needs_clflush);
 	if (err)
-		return err;
+		goto out;
 
 	page = i915_gem_object_get_page(ctx->obj, offset >> PAGE_SHIFT);
 	map = kmap_atomic(page);
@@ -73,7 +77,9 @@ static int cpu_get(struct context *ctx, unsigned long offset, u32 *v)
 	kunmap_atomic(map);
 	i915_gem_object_finish_access(ctx->obj);
 
-	return 0;
+out:
+	i915_gem_object_unlock(ctx->obj);
+	return err;
 }
 
 static int gtt_set(struct context *ctx, unsigned long offset, u32 v)
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index e0ea3e48f2b0..a9b149d60737 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -461,9 +461,10 @@ static int cpu_fill(struct drm_i915_gem_object *obj, u32 value)
 	unsigned int n, m, need_flush;
 	int err;
 
+	i915_gem_object_lock(obj, NULL);
 	err = i915_gem_object_prepare_write(obj, &need_flush);
 	if (err)
-		return err;
+		goto out;
 
 	for (n = 0; n < real_page_count(obj); n++) {
 		u32 *map;
@@ -479,7 +480,9 @@ static int cpu_fill(struct drm_i915_gem_object *obj, u32 value)
 	i915_gem_object_finish_access(obj);
 	obj->read_domains = I915_GEM_DOMAIN_GTT | I915_GEM_DOMAIN_CPU;
 	obj->write_domain = 0;
-	return 0;
+out:
+	i915_gem_object_unlock(obj);
+	return err;
 }
 
 static noinline int cpu_check(struct drm_i915_gem_object *obj,
@@ -488,9 +491,10 @@ static noinline int cpu_check(struct drm_i915_gem_object *obj,
 	unsigned int n, m, needs_flush;
 	int err;
 
+	i915_gem_object_lock(obj, NULL);
 	err = i915_gem_object_prepare_read(obj, &needs_flush);
 	if (err)
-		return err;
+		goto out_unlock;
 
 	for (n = 0; n < real_page_count(obj); n++) {
 		u32 *map;
@@ -527,6 +531,8 @@ static noinline int cpu_check(struct drm_i915_gem_object *obj,
 	}
 
 	i915_gem_object_finish_access(obj);
+out_unlock:
+	i915_gem_object_unlock(obj);
 	return err;
 }
 
diff --git a/drivers/gpu/drm/i915/gt/intel_renderstate.c b/drivers/gpu/drm/i915/gt/intel_renderstate.c
index 26e78db33675..ca533d98d14d 100644
--- a/drivers/gpu/drm/i915/gt/intel_renderstate.c
+++ b/drivers/gpu/drm/i915/gt/intel_renderstate.c
@@ -74,9 +74,10 @@ static int render_state_setup(struct intel_renderstate *so,
 	u32 *d;
 	int ret;
 
+	i915_gem_object_lock(so->vma->obj, NULL);
 	ret = i915_gem_object_prepare_write(so->vma->obj, &needs_clflush);
 	if (ret)
-		return ret;
+		goto out_unlock;
 
 	d = kmap_atomic(i915_gem_object_get_dirty_page(so->vma->obj, 0));
 
@@ -157,6 +158,8 @@ static int render_state_setup(struct intel_renderstate *so,
 	ret = 0;
 out:
 	i915_gem_object_finish_access(so->vma->obj);
+out_unlock:
+	i915_gem_object_unlock(so->vma->obj);
 	return ret;
 
 err:
diff --git a/drivers/gpu/drm/i915/gvt/cmd_parser.c b/drivers/gpu/drm/i915/gvt/cmd_parser.c
index dd2961715fe1..322aff9e0668 100644
--- a/drivers/gpu/drm/i915/gvt/cmd_parser.c
+++ b/drivers/gpu/drm/i915/gvt/cmd_parser.c
@@ -1870,10 +1870,14 @@ static int perform_bb_shadow(struct parser_exec_state *s)
 		goto err_free_bb;
 	}
 
-	ret = i915_gem_object_prepare_write(bb->obj, &bb->clflush);
+	ret = i915_gem_object_lock_interruptible(bb->obj, NULL);
 	if (ret)
 		goto err_free_obj;
 
+	ret = i915_gem_object_prepare_write(bb->obj, &bb->clflush);
+	if (ret)
+		goto err_unlock;
+
 	bb->va = i915_gem_object_pin_map(bb->obj, I915_MAP_WB);
 	if (IS_ERR(bb->va)) {
 		ret = PTR_ERR(bb->va);
@@ -1898,6 +1902,7 @@ static int perform_bb_shadow(struct parser_exec_state *s)
 	if (ret)
 		goto err_unmap;
 
+	i915_gem_object_unlock(bb->obj);
 	INIT_LIST_HEAD(&bb->list);
 	list_add(&bb->list, &s->workload->shadow_bb);
 
@@ -1924,6 +1929,8 @@ static int perform_bb_shadow(struct parser_exec_state *s)
 	i915_gem_object_unpin_map(bb->obj);
 err_finish_shmem_access:
 	i915_gem_object_finish_access(bb->obj);
+err_unlock:
+	i915_gem_object_unlock(bb->obj);
 err_free_obj:
 	i915_gem_object_put(bb->obj);
 err_free_bb:
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index cb311a6349cc..1effd4b96459 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -330,12 +330,20 @@ i915_gem_shmem_pread(struct drm_i915_gem_object *obj,
 	u64 remain;
 	int ret;
 
-	ret = i915_gem_object_prepare_read(obj, &needs_clflush);
+	ret = i915_gem_object_lock_interruptible(obj, NULL);
 	if (ret)
 		return ret;
 
+	ret = i915_gem_object_prepare_read(obj, &needs_clflush);
+	if (ret) {
+		i915_gem_object_unlock(obj);
+		return ret;
+	}
+
 	fence = i915_gem_object_lock_fence(obj);
 	i915_gem_object_finish_access(obj);
+	i915_gem_object_unlock(obj);
+
 	if (!fence)
 		return -ENOMEM;
 
@@ -729,12 +737,20 @@ i915_gem_shmem_pwrite(struct drm_i915_gem_object *obj,
 	u64 remain;
 	int ret;
 
-	ret = i915_gem_object_prepare_write(obj, &needs_clflush);
+	ret = i915_gem_object_lock_interruptible(obj, NULL);
 	if (ret)
 		return ret;
 
+	ret = i915_gem_object_prepare_write(obj, &needs_clflush);
+	if (ret) {
+		i915_gem_object_unlock(obj);
+		return ret;
+	}
+
 	fence = i915_gem_object_lock_fence(obj);
 	i915_gem_object_finish_access(obj);
+	i915_gem_object_unlock(obj);
+
 	if (!fence)
 		return -ENOMEM;
 
-- 
2.25.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  parent reply	other threads:[~2020-03-24 16:51 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-24 16:51 [Intel-gfx] [PATCH 01/21] Revert "drm/i915/gem: Drop relocation slowpath" Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 02/21] drm/i915: Add an implementation for i915_gem_ww_ctx locking, v2 Maarten Lankhorst
2020-03-24 16:51 ` Maarten Lankhorst [this message]
2020-03-24 16:51 ` [Intel-gfx] [PATCH 04/21] drm/i915: Parse command buffer earlier in eb_relocate(slow) Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 05/21] drm/i915: Use per object locking in execbuf, v6 Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 06/21] drm/i915: Use ww locking in intel_renderstate Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 07/21] drm/i915: Add ww context handling to context_barrier_task Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 08/21] drm/i915: Nuke arguments to eb_pin_engine Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 09/21] drm/i915: Pin engine before pinning all objects, v3 Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 10/21] drm/i915: Rework intel_context pinning to do everything outside of pin_mutex Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 11/21] drm/i915: Make sure execbuffer always passes ww state to i915_vma_pin Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 12/21] drm/i915: Convert i915_gem_object/client_blt.c to use ww locking as well, v2 Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 13/21] drm/i915: Kill last user of intel_context_create_request outside of selftests Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 14/21] drm/i915: Convert i915_perf to ww locking as well Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 15/21] drm/i915: Dirty hack to fix selftests locking inversion Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 16/21] drm/i915/selftests: Fix locking inversion in lrc selftest Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 17/21] drm/i915: Use ww pinning for intel_context_create_request() Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 18/21] drm/i915: Move i915_vma_lock in the selftests to avoid lock inversion Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 19/21] drm/i915: Add ww locking to vm_fault_gtt Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 20/21] drm/i915: Add ww locking to pin_to_display_plane Maarten Lankhorst
2020-03-24 16:51 ` [Intel-gfx] [PATCH 21/21] drm/i915: Ensure we hold the pin mutex Maarten Lankhorst
2020-03-24 17:02 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/21] Revert "drm/i915/gem: Drop relocation slowpath" Patchwork
2020-03-24 17:17 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork
2020-03-25  9:09 ` [Intel-gfx] ✓ Fi.CI.IGT: success " Patchwork
2020-03-26 11:49 [Intel-gfx] [PATCH 01/21] " Maarten Lankhorst
2020-03-26 11:49 ` [Intel-gfx] [PATCH 03/21] drm/i915: Remove locking from i915_gem_object_prepare_read/write Maarten Lankhorst

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200324165146.1032624-3-maarten.lankhorst@linux.intel.com \
    --to=maarten.lankhorst@linux.intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.