[PATCH 0/5] drm/i915: Get rid of fence error propagation (v2)

All of lore.kernel.org
 help / color / mirror / Atom feed

* [PATCH 0/5] drm/i915: Get rid of fence error propagation (v2)
@ 2021-06-03 15:40 ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, Jon Bloomfield, Jason Ekstrand

Fence error propagation is sketchy at best.  Instead of explicitly handling
fences which might have errors set in the code which is aware of errors, we
just kick them down the line and hope that userspace knows what to do when
a wait eventually fails.  This is sketchy at best because most userspace
isn't prepared to handle errors in those places.  To make things worse, it
allows errors to propagate across processes in unpredictable ways.  This is
causing hangs in one client to kill X11.

Unfortunately, there's no quick path from here to there thanks to the fact
that we're now running the command parser asynchronously and relying on
fence errors for when it fails.  This series first gets rid of asynchronous
command parsing and then cleans up from there.  There was never any real
use-case for asynchronous parsing and the platforms that rely heavily on
the command parser are old enough (Gen7) that, when we changed the way the
command parser works, it wasn't really a change anyone was asking for
anyway.

I think we probably want this whole mess back-ported.  I'm happy to take
suggestions on the strategy there because the history there is a bit
annoying and I'm not 100% sure where the Linux release cuts land.  In any
case, I'm happy to make a version of this series per-release if needed for
Greg to back-port.

v2 (Daniel Vetter):
 - Re-order to put the reverts first
 - Add ACKs from Daniel
 - Add better CC and Fixes tags

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>

Jason Ekstrand (5):
  drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  Revert "drm/i915: Propagate errors on awaiting already signaled
    fences"
  drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
  drm/i915: Drop error handling from dma_fence_work
  Revert "drm/i915: Skip over MI_NOOP when parsing"

 drivers/gpu/drm/i915/gem/i915_gem_clflush.c   |   4 +-
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 drivers/gpu/drm/i915/gem/i915_gem_object.h    |  10 +-
 drivers/gpu/drm/i915/gem/i915_gem_pages.c     |  21 +-
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/gt/intel_ggtt.c          |   2 +-
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 199 ++++++++-------
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 drivers/gpu/drm/i915/i915_request.c           |   8 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.c     |   5 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.h     |   2 +-
 drivers/gpu/drm/i915/i915_vma.c               |   3 +-
 12 files changed, 141 insertions(+), 351 deletions(-)

-- 
2.31.1

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [Intel-gfx] [PATCH 0/5] drm/i915: Get rid of fence error propagation (v2)
@ 2021-06-03 15:40 ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter

Fence error propagation is sketchy at best.  Instead of explicitly handling
fences which might have errors set in the code which is aware of errors, we
just kick them down the line and hope that userspace knows what to do when
a wait eventually fails.  This is sketchy at best because most userspace
isn't prepared to handle errors in those places.  To make things worse, it
allows errors to propagate across processes in unpredictable ways.  This is
causing hangs in one client to kill X11.

Unfortunately, there's no quick path from here to there thanks to the fact
that we're now running the command parser asynchronously and relying on
fence errors for when it fails.  This series first gets rid of asynchronous
command parsing and then cleans up from there.  There was never any real
use-case for asynchronous parsing and the platforms that rely heavily on
the command parser are old enough (Gen7) that, when we changed the way the
command parser works, it wasn't really a change anyone was asking for
anyway.

I think we probably want this whole mess back-ported.  I'm happy to take
suggestions on the strategy there because the history there is a bit
annoying and I'm not 100% sure where the Linux release cuts land.  In any
case, I'm happy to make a version of this series per-release if needed for
Greg to back-port.

v2 (Daniel Vetter):
 - Re-order to put the reverts first
 - Add ACKs from Daniel
 - Add better CC and Fixes tags

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>

Jason Ekstrand (5):
  drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  Revert "drm/i915: Propagate errors on awaiting already signaled
    fences"
  drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
  drm/i915: Drop error handling from dma_fence_work
  Revert "drm/i915: Skip over MI_NOOP when parsing"

 drivers/gpu/drm/i915/gem/i915_gem_clflush.c   |   4 +-
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 drivers/gpu/drm/i915/gem/i915_gem_object.h    |  10 +-
 drivers/gpu/drm/i915/gem/i915_gem_pages.c     |  21 +-
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/gt/intel_ggtt.c          |   2 +-
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 199 ++++++++-------
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 drivers/gpu/drm/i915/i915_request.c           |   8 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.c     |   5 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.h     |   2 +-
 drivers/gpu/drm/i915/i915_vma.c               |   3 +-
 12 files changed, 141 insertions(+), 351 deletions(-)

-- 
2.31.1

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

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
  (?)
@ 2021-06-03 15:40   ` Jason Ekstrand
  -1 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Jason Ekstrand, stable, Maarten Lankhorst, Jon Bloomfield, Daniel Vetter

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 297143511f99b..a49da4b24d4d4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1456,6 +1454,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2372,217 +2374,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2672,7 +2463,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 5b4b2bd46e7c8..056a233f443b4 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 1cdd64116fcec..a92b05a18f056 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1923,17 +1923,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [Intel-gfx] [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, stable

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 297143511f99b..a49da4b24d4d4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1456,6 +1454,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2372,217 +2374,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2672,7 +2463,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 5b4b2bd46e7c8..056a233f443b4 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 1cdd64116fcec..a92b05a18f056 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1923,17 +1923,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1

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

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Daniel Vetter, Jon Bloomfield, stable, Jason Ekstrand

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 297143511f99b..a49da4b24d4d4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1456,6 +1454,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2372,217 +2374,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2672,7 +2463,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 5b4b2bd46e7c8..056a233f443b4 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 1cdd64116fcec..a92b05a18f056 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1923,17 +1923,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 2/5] Revert "drm/i915: Propagate errors on awaiting already signaled fences"
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
  (?)
@ 2021-06-03 15:40   ` Jason Ekstrand
  -1 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Jason Ekstrand, Marcin Slusarz, stable, Jason Ekstrand,
	Daniel Vetter, Jon Bloomfield

This reverts commit 9e31c1fe45d555a948ff66f1f0e3fe1f83ca63f7.  Ever
since that commit, we've been having issues where a hang in one client
can propagate to another.  In particular, a hang in an app can propagate
to the X server which causes the whole desktop to lock up.

Error propagation along fences sound like a good idea, but as your bug
shows, surprising consequences, since propagating errors across security
boundaries is not a good thing.

What we do have is track the hangs on the ctx, and report information to
userspace using RESET_STATS. That's how arb_robustness works. Also, if my
understanding is still correct, the EIO from execbuf is when your context
is banned (because not recoverable or too many hangs). And in all these
cases it's up to userspace to figure out what is all impacted and should
be reported to the application, that's not on the kernel to guess and
automatically propagate.

What's more, we're also building more features on top of ctx error
reporting with RESET_STATS ioctl: Encrypted buffers use the same, and the
userspace fence wait also relies on that mechanism. So it is the path
going forward for reporting gpu hangs and resets to userspace.

So all together that's why I think we should just bury this idea again as
not quite the direction we want to go to, hence why I think the revert is
the right option here.

For backporters: Please note that you _must_ have a backport of
https://lore.kernel.org/dri-devel/20210602164149.391653-2-jason@jlekstrand.net/
for otherwise backporting just this patch opens up a security bug.

v2: Augment commit message. Also restore Jason's sob that I
accidentally lost.

v3: Add a note for backporters

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reported-by: Marcin Slusarz <marcin.slusarz@intel.com>
Cc: <stable@vger.kernel.org> # v5.6+
Cc: Jason Ekstrand <jason.ekstrand@intel.com>
Cc: Marcin Slusarz <marcin.slusarz@intel.com>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/3080
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
---
 drivers/gpu/drm/i915/i915_request.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 970d8f4986bbe..b796197c07722 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -1426,10 +1426,8 @@ i915_request_await_execution(struct i915_request *rq,

 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
-			i915_sw_fence_set_error_once(&rq->submit, fence->error);
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 			continue;
-		}

 		if (fence->context == rq->fence.context)
 			continue;
@@ -1527,10 +1525,8 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)

 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
-			i915_sw_fence_set_error_once(&rq->submit, fence->error);
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 			continue;
-		}

 		/*
 		 * Requests on the same timeline are explicitly ordered, along
-- 
2.31.1

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [Intel-gfx] [PATCH 2/5] Revert "drm/i915: Propagate errors on awaiting already signaled fences"
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, stable, Jason Ekstrand

This reverts commit 9e31c1fe45d555a948ff66f1f0e3fe1f83ca63f7.  Ever
since that commit, we've been having issues where a hang in one client
can propagate to another.  In particular, a hang in an app can propagate
to the X server which causes the whole desktop to lock up.

Error propagation along fences sound like a good idea, but as your bug
shows, surprising consequences, since propagating errors across security
boundaries is not a good thing.

What we do have is track the hangs on the ctx, and report information to
userspace using RESET_STATS. That's how arb_robustness works. Also, if my
understanding is still correct, the EIO from execbuf is when your context
is banned (because not recoverable or too many hangs). And in all these
cases it's up to userspace to figure out what is all impacted and should
be reported to the application, that's not on the kernel to guess and
automatically propagate.

What's more, we're also building more features on top of ctx error
reporting with RESET_STATS ioctl: Encrypted buffers use the same, and the
userspace fence wait also relies on that mechanism. So it is the path
going forward for reporting gpu hangs and resets to userspace.

So all together that's why I think we should just bury this idea again as
not quite the direction we want to go to, hence why I think the revert is
the right option here.

For backporters: Please note that you _must_ have a backport of
https://lore.kernel.org/dri-devel/20210602164149.391653-2-jason@jlekstrand.net/
for otherwise backporting just this patch opens up a security bug.

v2: Augment commit message. Also restore Jason's sob that I
accidentally lost.

v3: Add a note for backporters

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reported-by: Marcin Slusarz <marcin.slusarz@intel.com>
Cc: <stable@vger.kernel.org> # v5.6+
Cc: Jason Ekstrand <jason.ekstrand@intel.com>
Cc: Marcin Slusarz <marcin.slusarz@intel.com>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/3080
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
---
 drivers/gpu/drm/i915/i915_request.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 970d8f4986bbe..b796197c07722 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -1426,10 +1426,8 @@ i915_request_await_execution(struct i915_request *rq,

 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
-			i915_sw_fence_set_error_once(&rq->submit, fence->error);
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 			continue;
-		}

 		if (fence->context == rq->fence.context)
 			continue;
@@ -1527,10 +1525,8 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)

 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
-			i915_sw_fence_set_error_once(&rq->submit, fence->error);
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 			continue;
-		}

 		/*
 		 * Requests on the same timeline are explicitly ordered, along
-- 
2.31.1

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

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 2/5] Revert "drm/i915: Propagate errors on awaiting already signaled fences"
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Daniel Vetter, stable, Jason Ekstrand, Jon Bloomfield,
	Jason Ekstrand, Marcin Slusarz

This reverts commit 9e31c1fe45d555a948ff66f1f0e3fe1f83ca63f7.  Ever
since that commit, we've been having issues where a hang in one client
can propagate to another.  In particular, a hang in an app can propagate
to the X server which causes the whole desktop to lock up.

Error propagation along fences sound like a good idea, but as your bug
shows, surprising consequences, since propagating errors across security
boundaries is not a good thing.

What we do have is track the hangs on the ctx, and report information to
userspace using RESET_STATS. That's how arb_robustness works. Also, if my
understanding is still correct, the EIO from execbuf is when your context
is banned (because not recoverable or too many hangs). And in all these
cases it's up to userspace to figure out what is all impacted and should
be reported to the application, that's not on the kernel to guess and
automatically propagate.

What's more, we're also building more features on top of ctx error
reporting with RESET_STATS ioctl: Encrypted buffers use the same, and the
userspace fence wait also relies on that mechanism. So it is the path
going forward for reporting gpu hangs and resets to userspace.

So all together that's why I think we should just bury this idea again as
not quite the direction we want to go to, hence why I think the revert is
the right option here.

For backporters: Please note that you _must_ have a backport of
https://lore.kernel.org/dri-devel/20210602164149.391653-2-jason@jlekstrand.net/
for otherwise backporting just this patch opens up a security bug.

v2: Augment commit message. Also restore Jason's sob that I
accidentally lost.

v3: Add a note for backporters

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reported-by: Marcin Slusarz <marcin.slusarz@intel.com>
Cc: <stable@vger.kernel.org> # v5.6+
Cc: Jason Ekstrand <jason.ekstrand@intel.com>
Cc: Marcin Slusarz <marcin.slusarz@intel.com>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/3080
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
---
 drivers/gpu/drm/i915/i915_request.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 970d8f4986bbe..b796197c07722 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -1426,10 +1426,8 @@ i915_request_await_execution(struct i915_request *rq,

 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
-			i915_sw_fence_set_error_once(&rq->submit, fence->error);
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 			continue;
-		}

 		if (fence->context == rq->fence.context)
 			continue;
@@ -1527,10 +1525,8 @@ i915_request_await_dma_fence(struct i915_request *rq, struct dma_fence *fence)

 	do {
 		fence = *child++;
-		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags)) {
-			i915_sw_fence_set_error_once(&rq->submit, fence->error);
+		if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, &fence->flags))
 			continue;
-		}

 		/*
 		 * Requests on the same timeline are explicitly ordered, along
-- 
2.31.1

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 3/5] drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
@ 2021-06-03 15:40   ` Jason Ekstrand
  -1 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, Jon Bloomfield, Jason Ekstrand

This reverts the rest of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer").  Now that the only user of i915_gem_object_get_sg
without allow_alloc has been removed, we can drop the parameter.  This
portion of the revert was broken into its own patch to aid review.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/gem/i915_gem_object.h | 10 +++++-----
 drivers/gpu/drm/i915/gem/i915_gem_pages.c  | 21 ++++-----------------
 drivers/gpu/drm/i915/gt/intel_ggtt.c       |  2 +-
 3 files changed, 10 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 2ebd79537aea9..329d848f3ff6d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -339,22 +339,22 @@ struct scatterlist *
 __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 			 struct i915_gem_object_page_iter *iter,
 			 unsigned int n,
-			 unsigned int *offset, bool allow_alloc);
+			 unsigned int *offset);
 
 static inline struct scatterlist *
 i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 		       unsigned int n,
-		       unsigned int *offset, bool allow_alloc)
+		       unsigned int *offset)
 {
-	return __i915_gem_object_get_sg(obj, &obj->mm.get_page, n, offset, allow_alloc);
+	return __i915_gem_object_get_sg(obj, &obj->mm.get_page, n, offset);
 }
 
 static inline struct scatterlist *
 i915_gem_object_get_sg_dma(struct drm_i915_gem_object *obj,
 			   unsigned int n,
-			   unsigned int *offset, bool allow_alloc)
+			   unsigned int *offset)
 {
-	return __i915_gem_object_get_sg(obj, &obj->mm.get_dma_page, n, offset, allow_alloc);
+	return __i915_gem_object_get_sg(obj, &obj->mm.get_dma_page, n, offset);
 }
 
 struct page *
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
index 6444e097016da..f234bd5bf22fc 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
@@ -466,8 +466,7 @@ struct scatterlist *
 __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 			 struct i915_gem_object_page_iter *iter,
 			 unsigned int n,
-			 unsigned int *offset,
-			 bool allow_alloc)
+			 unsigned int *offset)
 {
 	const bool dma = iter == &obj->mm.get_dma_page;
 	struct scatterlist *sg;
@@ -490,9 +489,6 @@ __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 	if (n < READ_ONCE(iter->sg_idx))
 		goto lookup;
 
-	if (!allow_alloc)
-		goto manual_lookup;
-
 	mutex_lock(&iter->lock);
 
 	/* We prefer to reuse the last sg so that repeated lookup of this
@@ -542,16 +538,7 @@ __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 	if (unlikely(n < idx)) /* insertion completed by another thread */
 		goto lookup;
 
-	goto manual_walk;
-
-manual_lookup:
-	idx = 0;
-	sg = obj->mm.pages->sgl;
-	count = __sg_page_count(sg);
-
-manual_walk:
-	/*
-	 * In case we failed to insert the entry into the radixtree, we need
+	/* In case we failed to insert the entry into the radixtree, we need
 	 * to look beyond the current sg.
 	 */
 	while (idx + count <= n) {
@@ -598,7 +585,7 @@ i915_gem_object_get_page(struct drm_i915_gem_object *obj, unsigned int n)
 
 	GEM_BUG_ON(!i915_gem_object_has_struct_page(obj));
 
-	sg = i915_gem_object_get_sg(obj, n, &offset, true);
+	sg = i915_gem_object_get_sg(obj, n, &offset);
 	return nth_page(sg_page(sg), offset);
 }
 
@@ -624,7 +611,7 @@ i915_gem_object_get_dma_address_len(struct drm_i915_gem_object *obj,
 	struct scatterlist *sg;
 	unsigned int offset;
 
-	sg = i915_gem_object_get_sg_dma(obj, n, &offset, true);
+	sg = i915_gem_object_get_sg_dma(obj, n, &offset);
 
 	if (len)
 		*len = sg_dma_len(sg) - (offset << PAGE_SHIFT);
diff --git a/drivers/gpu/drm/i915/gt/intel_ggtt.c b/drivers/gpu/drm/i915/gt/intel_ggtt.c
index 10c23a749a950..46c98887f17e3 100644
--- a/drivers/gpu/drm/i915/gt/intel_ggtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_ggtt.c
@@ -1494,7 +1494,7 @@ intel_partial_pages(const struct i915_ggtt_view *view,
 	if (ret)
 		goto err_sg_alloc;
 
-	iter = i915_gem_object_get_sg_dma(obj, view->partial.offset, &offset, true);
+	iter = i915_gem_object_get_sg_dma(obj, view->partial.offset, &offset);
 	GEM_BUG_ON(!iter);
 
 	sg = st->sgl;
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [Intel-gfx] [PATCH 3/5] drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter

This reverts the rest of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer").  Now that the only user of i915_gem_object_get_sg
without allow_alloc has been removed, we can drop the parameter.  This
portion of the revert was broken into its own patch to aid review.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/gem/i915_gem_object.h | 10 +++++-----
 drivers/gpu/drm/i915/gem/i915_gem_pages.c  | 21 ++++-----------------
 drivers/gpu/drm/i915/gt/intel_ggtt.c       |  2 +-
 3 files changed, 10 insertions(+), 23 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h
index 2ebd79537aea9..329d848f3ff6d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_object.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h
@@ -339,22 +339,22 @@ struct scatterlist *
 __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 			 struct i915_gem_object_page_iter *iter,
 			 unsigned int n,
-			 unsigned int *offset, bool allow_alloc);
+			 unsigned int *offset);
 
 static inline struct scatterlist *
 i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 		       unsigned int n,
-		       unsigned int *offset, bool allow_alloc)
+		       unsigned int *offset)
 {
-	return __i915_gem_object_get_sg(obj, &obj->mm.get_page, n, offset, allow_alloc);
+	return __i915_gem_object_get_sg(obj, &obj->mm.get_page, n, offset);
 }
 
 static inline struct scatterlist *
 i915_gem_object_get_sg_dma(struct drm_i915_gem_object *obj,
 			   unsigned int n,
-			   unsigned int *offset, bool allow_alloc)
+			   unsigned int *offset)
 {
-	return __i915_gem_object_get_sg(obj, &obj->mm.get_dma_page, n, offset, allow_alloc);
+	return __i915_gem_object_get_sg(obj, &obj->mm.get_dma_page, n, offset);
 }
 
 struct page *
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
index 6444e097016da..f234bd5bf22fc 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c
@@ -466,8 +466,7 @@ struct scatterlist *
 __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 			 struct i915_gem_object_page_iter *iter,
 			 unsigned int n,
-			 unsigned int *offset,
-			 bool allow_alloc)
+			 unsigned int *offset)
 {
 	const bool dma = iter == &obj->mm.get_dma_page;
 	struct scatterlist *sg;
@@ -490,9 +489,6 @@ __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 	if (n < READ_ONCE(iter->sg_idx))
 		goto lookup;
 
-	if (!allow_alloc)
-		goto manual_lookup;
-
 	mutex_lock(&iter->lock);
 
 	/* We prefer to reuse the last sg so that repeated lookup of this
@@ -542,16 +538,7 @@ __i915_gem_object_get_sg(struct drm_i915_gem_object *obj,
 	if (unlikely(n < idx)) /* insertion completed by another thread */
 		goto lookup;
 
-	goto manual_walk;
-
-manual_lookup:
-	idx = 0;
-	sg = obj->mm.pages->sgl;
-	count = __sg_page_count(sg);
-
-manual_walk:
-	/*
-	 * In case we failed to insert the entry into the radixtree, we need
+	/* In case we failed to insert the entry into the radixtree, we need
 	 * to look beyond the current sg.
 	 */
 	while (idx + count <= n) {
@@ -598,7 +585,7 @@ i915_gem_object_get_page(struct drm_i915_gem_object *obj, unsigned int n)
 
 	GEM_BUG_ON(!i915_gem_object_has_struct_page(obj));
 
-	sg = i915_gem_object_get_sg(obj, n, &offset, true);
+	sg = i915_gem_object_get_sg(obj, n, &offset);
 	return nth_page(sg_page(sg), offset);
 }
 
@@ -624,7 +611,7 @@ i915_gem_object_get_dma_address_len(struct drm_i915_gem_object *obj,
 	struct scatterlist *sg;
 	unsigned int offset;
 
-	sg = i915_gem_object_get_sg_dma(obj, n, &offset, true);
+	sg = i915_gem_object_get_sg_dma(obj, n, &offset);
 
 	if (len)
 		*len = sg_dma_len(sg) - (offset << PAGE_SHIFT);
diff --git a/drivers/gpu/drm/i915/gt/intel_ggtt.c b/drivers/gpu/drm/i915/gt/intel_ggtt.c
index 10c23a749a950..46c98887f17e3 100644
--- a/drivers/gpu/drm/i915/gt/intel_ggtt.c
+++ b/drivers/gpu/drm/i915/gt/intel_ggtt.c
@@ -1494,7 +1494,7 @@ intel_partial_pages(const struct i915_ggtt_view *view,
 	if (ret)
 		goto err_sg_alloc;
 
-	iter = i915_gem_object_get_sg_dma(obj, view->partial.offset, &offset, true);
+	iter = i915_gem_object_get_sg_dma(obj, view->partial.offset, &offset);
 	GEM_BUG_ON(!iter);
 
 	sg = st->sgl;
-- 
2.31.1

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

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 4/5] drm/i915: Drop error handling from dma_fence_work
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
@ 2021-06-03 15:40   ` Jason Ekstrand
  -1 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, Jon Bloomfield, Jason Ekstrand

Asynchronous command parsing was the only thing which ever returned a
non-zero error.  With that gone, we can drop the error handling from
dma_fence_work.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/gem/i915_gem_clflush.c | 4 +---
 drivers/gpu/drm/i915/i915_sw_fence_work.c   | 5 +----
 drivers/gpu/drm/i915/i915_sw_fence_work.h   | 2 +-
 drivers/gpu/drm/i915/i915_vma.c             | 3 +--
 4 files changed, 4 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
index daf9284ef1f54..f0435c6feb68b 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
@@ -24,13 +24,11 @@ static void __do_clflush(struct drm_i915_gem_object *obj)
 	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 }
 
-static int clflush_work(struct dma_fence_work *base)
+static void clflush_work(struct dma_fence_work *base)
 {
 	struct clflush *clflush = container_of(base, typeof(*clflush), base);
 
 	__do_clflush(clflush->obj);
-
-	return 0;
 }
 
 static void clflush_release(struct dma_fence_work *base)
diff --git a/drivers/gpu/drm/i915/i915_sw_fence_work.c b/drivers/gpu/drm/i915/i915_sw_fence_work.c
index a3a81bb8f2c36..5b33ef23d54c9 100644
--- a/drivers/gpu/drm/i915/i915_sw_fence_work.c
+++ b/drivers/gpu/drm/i915/i915_sw_fence_work.c
@@ -16,11 +16,8 @@ static void fence_complete(struct dma_fence_work *f)
 static void fence_work(struct work_struct *work)
 {
 	struct dma_fence_work *f = container_of(work, typeof(*f), work);
-	int err;
 
-	err = f->ops->work(f);
-	if (err)
-		dma_fence_set_error(&f->dma, err);
+	f->ops->work(f);
 
 	fence_complete(f);
 	dma_fence_put(&f->dma);
diff --git a/drivers/gpu/drm/i915/i915_sw_fence_work.h b/drivers/gpu/drm/i915/i915_sw_fence_work.h
index 2c409f11c5c59..d56806918d131 100644
--- a/drivers/gpu/drm/i915/i915_sw_fence_work.h
+++ b/drivers/gpu/drm/i915/i915_sw_fence_work.h
@@ -17,7 +17,7 @@ struct dma_fence_work;
 
 struct dma_fence_work_ops {
 	const char *name;
-	int (*work)(struct dma_fence_work *f);
+	void (*work)(struct dma_fence_work *f);
 	void (*release)(struct dma_fence_work *f);
 };
 
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 0f227f28b2802..5b9dce0f443b0 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -300,14 +300,13 @@ struct i915_vma_work {
 	unsigned int flags;
 };
 
-static int __vma_bind(struct dma_fence_work *work)
+static void __vma_bind(struct dma_fence_work *work)
 {
 	struct i915_vma_work *vw = container_of(work, typeof(*vw), base);
 	struct i915_vma *vma = vw->vma;
 
 	vma->ops->bind_vma(vw->vm, &vw->stash,
 			   vma, vw->cache_level, vw->flags);
-	return 0;
 }
 
 static void __vma_release(struct dma_fence_work *work)
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [Intel-gfx] [PATCH 4/5] drm/i915: Drop error handling from dma_fence_work
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter

Asynchronous command parsing was the only thing which ever returned a
non-zero error.  With that gone, we can drop the error handling from
dma_fence_work.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/gem/i915_gem_clflush.c | 4 +---
 drivers/gpu/drm/i915/i915_sw_fence_work.c   | 5 +----
 drivers/gpu/drm/i915/i915_sw_fence_work.h   | 2 +-
 drivers/gpu/drm/i915/i915_vma.c             | 3 +--
 4 files changed, 4 insertions(+), 10 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
index daf9284ef1f54..f0435c6feb68b 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_clflush.c
@@ -24,13 +24,11 @@ static void __do_clflush(struct drm_i915_gem_object *obj)
 	i915_gem_object_flush_frontbuffer(obj, ORIGIN_CPU);
 }
 
-static int clflush_work(struct dma_fence_work *base)
+static void clflush_work(struct dma_fence_work *base)
 {
 	struct clflush *clflush = container_of(base, typeof(*clflush), base);
 
 	__do_clflush(clflush->obj);
-
-	return 0;
 }
 
 static void clflush_release(struct dma_fence_work *base)
diff --git a/drivers/gpu/drm/i915/i915_sw_fence_work.c b/drivers/gpu/drm/i915/i915_sw_fence_work.c
index a3a81bb8f2c36..5b33ef23d54c9 100644
--- a/drivers/gpu/drm/i915/i915_sw_fence_work.c
+++ b/drivers/gpu/drm/i915/i915_sw_fence_work.c
@@ -16,11 +16,8 @@ static void fence_complete(struct dma_fence_work *f)
 static void fence_work(struct work_struct *work)
 {
 	struct dma_fence_work *f = container_of(work, typeof(*f), work);
-	int err;
 
-	err = f->ops->work(f);
-	if (err)
-		dma_fence_set_error(&f->dma, err);
+	f->ops->work(f);
 
 	fence_complete(f);
 	dma_fence_put(&f->dma);
diff --git a/drivers/gpu/drm/i915/i915_sw_fence_work.h b/drivers/gpu/drm/i915/i915_sw_fence_work.h
index 2c409f11c5c59..d56806918d131 100644
--- a/drivers/gpu/drm/i915/i915_sw_fence_work.h
+++ b/drivers/gpu/drm/i915/i915_sw_fence_work.h
@@ -17,7 +17,7 @@ struct dma_fence_work;
 
 struct dma_fence_work_ops {
 	const char *name;
-	int (*work)(struct dma_fence_work *f);
+	void (*work)(struct dma_fence_work *f);
 	void (*release)(struct dma_fence_work *f);
 };
 
diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
index 0f227f28b2802..5b9dce0f443b0 100644
--- a/drivers/gpu/drm/i915/i915_vma.c
+++ b/drivers/gpu/drm/i915/i915_vma.c
@@ -300,14 +300,13 @@ struct i915_vma_work {
 	unsigned int flags;
 };
 
-static int __vma_bind(struct dma_fence_work *work)
+static void __vma_bind(struct dma_fence_work *work)
 {
 	struct i915_vma_work *vw = container_of(work, typeof(*vw), base);
 	struct i915_vma *vma = vw->vma;
 
 	vma->ops->bind_vma(vw->vm, &vw->stash,
 			   vma, vw->cache_level, vw->flags);
-	return 0;
 }
 
 static void __vma_release(struct dma_fence_work *work)
-- 
2.31.1

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

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 5/5] Revert "drm/i915: Skip over MI_NOOP when parsing"
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
@ 2021-06-03 15:40   ` Jason Ekstrand
  -1 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, Jon Bloomfield, Jason Ekstrand

This reverts a6c5e2aea704 ("drm/i915: Skip over MI_NOOP when parsing").
It complicates the batch parsing code a bit and increases indentation
for no reason other than fast-skipping a command that userspace uses
only rarely.  Sure, there may be IGT tests that fill batches with NOOPs
but that's not a case we should optimize for in the kernel.  We should
optimize for code clarity instead.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_cmd_parser.c | 67 +++++++++++++-------------
 1 file changed, 34 insertions(+), 33 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 056a233f443b4..8d34f05d22b75 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1470,42 +1470,43 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 	 * space. Parsing should be faster in some cases this way.
 	 */
 	batch_end = cmd + batch_length / sizeof(*batch_end);
-	while (*cmd != MI_BATCH_BUFFER_END) {
-		u32 length = 1;
-
-		if (*cmd != MI_NOOP) { /* MI_NOOP == 0 */
-			desc = find_cmd(engine, *cmd, desc, &default_desc);
-			if (!desc) {
-				DRM_DEBUG("CMD: Unrecognized command: 0x%08X\n", *cmd);
-				ret = -EINVAL;
-				break;
-			}
+	do {
+		u32 length;
 
-			if (desc->flags & CMD_DESC_FIXED)
-				length = desc->length.fixed;
-			else
-				length = (*cmd & desc->length.mask) + LENGTH_BIAS;
+		if (*cmd == MI_BATCH_BUFFER_END)
+			break;
 
-			if ((batch_end - cmd) < length) {
-				DRM_DEBUG("CMD: Command length exceeds batch length: 0x%08X length=%u batchlen=%td\n",
-					  *cmd,
-					  length,
-					  batch_end - cmd);
-				ret = -EINVAL;
-				break;
-			}
+		desc = find_cmd(engine, *cmd, desc, &default_desc);
+		if (!desc) {
+			DRM_DEBUG("CMD: Unrecognized command: 0x%08X\n", *cmd);
+			ret = -EINVAL;
+			break;
+		}
 
-			if (!check_cmd(engine, desc, cmd, length)) {
-				ret = -EACCES;
-				break;
-			}
+		if (desc->flags & CMD_DESC_FIXED)
+			length = desc->length.fixed;
+		else
+			length = (*cmd & desc->length.mask) + LENGTH_BIAS;
 
-			if (cmd_desc_is(desc, MI_BATCH_BUFFER_START)) {
-				ret = check_bbstart(cmd, offset, length, batch_length,
-						    batch_addr, shadow_addr,
-						    jump_whitelist);
-				break;
-			}
+		if ((batch_end - cmd) < length) {
+			DRM_DEBUG("CMD: Command length exceeds batch length: 0x%08X length=%u batchlen=%td\n",
+				  *cmd,
+				  length,
+				  batch_end - cmd);
+			ret = -EINVAL;
+			break;
+		}
+
+		if (!check_cmd(engine, desc, cmd, length)) {
+			ret = -EACCES;
+			break;
+		}
+
+		if (cmd_desc_is(desc, MI_BATCH_BUFFER_START)) {
+			ret = check_bbstart(cmd, offset, length, batch_length,
+					    batch_addr, shadow_addr,
+					    jump_whitelist);
+			break;
 		}
 
 		if (!IS_ERR_OR_NULL(jump_whitelist))
@@ -1518,7 +1519,7 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			ret = -EINVAL;
 			break;
 		}
-	}
+	} while (1);
 
 	if (trampoline) {
 		/*
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [Intel-gfx] [PATCH 5/5] Revert "drm/i915: Skip over MI_NOOP when parsing"
@ 2021-06-03 15:40   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-03 15:40 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter

This reverts a6c5e2aea704 ("drm/i915: Skip over MI_NOOP when parsing").
It complicates the batch parsing code a bit and increases indentation
for no reason other than fast-skipping a command that userspace uses
only rarely.  Sure, there may be IGT tests that fill batches with NOOPs
but that's not a case we should optimize for in the kernel.  We should
optimize for code clarity instead.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 drivers/gpu/drm/i915/i915_cmd_parser.c | 67 +++++++++++++-------------
 1 file changed, 34 insertions(+), 33 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 056a233f443b4..8d34f05d22b75 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1470,42 +1470,43 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 	 * space. Parsing should be faster in some cases this way.
 	 */
 	batch_end = cmd + batch_length / sizeof(*batch_end);
-	while (*cmd != MI_BATCH_BUFFER_END) {
-		u32 length = 1;
-
-		if (*cmd != MI_NOOP) { /* MI_NOOP == 0 */
-			desc = find_cmd(engine, *cmd, desc, &default_desc);
-			if (!desc) {
-				DRM_DEBUG("CMD: Unrecognized command: 0x%08X\n", *cmd);
-				ret = -EINVAL;
-				break;
-			}
+	do {
+		u32 length;
 
-			if (desc->flags & CMD_DESC_FIXED)
-				length = desc->length.fixed;
-			else
-				length = (*cmd & desc->length.mask) + LENGTH_BIAS;
+		if (*cmd == MI_BATCH_BUFFER_END)
+			break;
 
-			if ((batch_end - cmd) < length) {
-				DRM_DEBUG("CMD: Command length exceeds batch length: 0x%08X length=%u batchlen=%td\n",
-					  *cmd,
-					  length,
-					  batch_end - cmd);
-				ret = -EINVAL;
-				break;
-			}
+		desc = find_cmd(engine, *cmd, desc, &default_desc);
+		if (!desc) {
+			DRM_DEBUG("CMD: Unrecognized command: 0x%08X\n", *cmd);
+			ret = -EINVAL;
+			break;
+		}
 
-			if (!check_cmd(engine, desc, cmd, length)) {
-				ret = -EACCES;
-				break;
-			}
+		if (desc->flags & CMD_DESC_FIXED)
+			length = desc->length.fixed;
+		else
+			length = (*cmd & desc->length.mask) + LENGTH_BIAS;
 
-			if (cmd_desc_is(desc, MI_BATCH_BUFFER_START)) {
-				ret = check_bbstart(cmd, offset, length, batch_length,
-						    batch_addr, shadow_addr,
-						    jump_whitelist);
-				break;
-			}
+		if ((batch_end - cmd) < length) {
+			DRM_DEBUG("CMD: Command length exceeds batch length: 0x%08X length=%u batchlen=%td\n",
+				  *cmd,
+				  length,
+				  batch_end - cmd);
+			ret = -EINVAL;
+			break;
+		}
+
+		if (!check_cmd(engine, desc, cmd, length)) {
+			ret = -EACCES;
+			break;
+		}
+
+		if (cmd_desc_is(desc, MI_BATCH_BUFFER_START)) {
+			ret = check_bbstart(cmd, offset, length, batch_length,
+					    batch_addr, shadow_addr,
+					    jump_whitelist);
+			break;
 		}
 
 		if (!IS_ERR_OR_NULL(jump_whitelist))
@@ -1518,7 +1519,7 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			ret = -EINVAL;
 			break;
 		}
-	}
+	} while (1);
 
 	if (trampoline) {
 		/*
-- 
2.31.1

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

^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Get rid of fence error propagation (rev2)
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
                   ` (5 preceding siblings ...)
  (?)
@ 2021-06-03 16:32 ` Patchwork
  -1 siblings, 0 replies; 24+ messages in thread
From: Patchwork @ 2021-06-03 16:32 UTC (permalink / raw)
  To: Jason Ekstrand; +Cc: intel-gfx

== Series Details ==

Series: drm/i915: Get rid of fence error propagation (rev2)
URL   : https://patchwork.freedesktop.org/series/90891/
State : warning

== Summary ==

$ dim checkpatch origin/drm-tip
20301afd39c9 drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
-:6: ERROR:GIT_COMMIT_ID: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser")'
#6: 
This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The

-:23: ERROR:GIT_COMMIT_ID: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit 0edbb9ba1bfe ("drm/i915: Move cmd parser pinning to execbuffer")'
#23: 
This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser

total: 2 errors, 0 warnings, 0 checks, 482 lines checked
b6fda25b1d9f Revert "drm/i915: Propagate errors on awaiting already signaled fences"
-:34: WARNING:COMMIT_LOG_LONG_LINE: Possible unwrapped commit description (prefer a maximum 75 chars per line)
#34: 
https://lore.kernel.org/dri-devel/20210602164149.391653-2-jason@jlekstrand.net/

total: 0 errors, 1 warnings, 0 checks, 22 lines checked
7b00404854f1 drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
-:6: ERROR:GIT_COMMIT_ID: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit 0edbb9ba1bfe ("drm/i915: Move cmd parser pinning to execbuffer")'
#6: 
This reverts the rest of 0edbb9ba1bfe ("drm/i915: Move cmd parser

total: 1 errors, 0 warnings, 0 checks, 86 lines checked
8e472c11e36c drm/i915: Drop error handling from dma_fence_work
1013d3982bcb Revert "drm/i915: Skip over MI_NOOP when parsing"
-:6: ERROR:GIT_COMMIT_ID: Please use git commit description style 'commit <12+ chars of sha1> ("<title line>")' - ie: 'commit a6c5e2aea704 ("drm/i915: Skip over MI_NOOP when parsing")'
#6: 
This reverts a6c5e2aea704 ("drm/i915: Skip over MI_NOOP when parsing").

total: 1 errors, 0 warnings, 0 checks, 83 lines checked


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

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [Intel-gfx] ✗ Fi.CI.SPARSE: warning for drm/i915: Get rid of fence error propagation (rev2)
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
                   ` (6 preceding siblings ...)
  (?)
@ 2021-06-03 16:33 ` Patchwork
  -1 siblings, 0 replies; 24+ messages in thread
From: Patchwork @ 2021-06-03 16:33 UTC (permalink / raw)
  To: Jason Ekstrand; +Cc: intel-gfx

== Series Details ==

Series: drm/i915: Get rid of fence error propagation (rev2)
URL   : https://patchwork.freedesktop.org/series/90891/
State : warning

== Summary ==

$ dim sparse --fast origin/drm-tip
Sparse version: v0.6.2
Fast mode used, each commit won't be checked separately.
-
+drivers/gpu/drm/i915/display/intel_display.c:1893:21:    expected struct i915_vma *[assigned] vma
+drivers/gpu/drm/i915/display/intel_display.c:1893:21:    got void [noderef] __iomem *[assigned] iomem
+drivers/gpu/drm/i915/display/intel_display.c:1893:21: warning: incorrect type in assignment (different address spaces)
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:27:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:27:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:27:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:32:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:32:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:49:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:49:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:49:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:56:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_engine_stats.h:56:9: warning: trying to copy expression type 31
+drivers/gpu/drm/i915/gt/intel_reset.c:1396:5: warning: context imbalance in 'intel_gt_reset_trylock' - different lock contexts for basic block
+drivers/gpu/drm/i915/gt/intel_ring_submission.c:1207:24: warning: Using plain integer as NULL pointer
+drivers/gpu/drm/i915/i915_perf.c:1434:15: warning: memset with byte count of 16777216
+drivers/gpu/drm/i915/i915_perf.c:1488:15: warning: memset with byte count of 16777216
+./include/asm-generic/bitops/find.h:112:45: warning: shift count is negative (-262080)
+./include/asm-generic/bitops/find.h:32:31: warning: shift count is negative (-262080)
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_read16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_read32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_read64' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_read8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_write16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_write32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'fwtable_write8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_read16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_read32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_read64' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_read8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_write16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_write32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen11_fwtable_write8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_read16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_read32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_read64' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_read8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_write16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_write32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen12_fwtable_write8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_read16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_read32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_read64' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_read8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_write16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_write32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen6_write8' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen8_write16' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen8_write32' - different lock contexts for basic block
+./include/linux/spinlock.h:409:9: warning: context imbalance in 'gen8_write8' - different lock contexts for basic block


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

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [Intel-gfx] ✗ Fi.CI.DOCS: warning for drm/i915: Get rid of fence error propagation (rev2)
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
                   ` (7 preceding siblings ...)
  (?)
@ 2021-06-03 16:37 ` Patchwork
  -1 siblings, 0 replies; 24+ messages in thread
From: Patchwork @ 2021-06-03 16:37 UTC (permalink / raw)
  To: Jason Ekstrand; +Cc: intel-gfx

== Series Details ==

Series: drm/i915: Get rid of fence error propagation (rev2)
URL   : https://patchwork.freedesktop.org/series/90891/
State : warning

== Summary ==

$ make htmldocs 2>&1 > /dev/null | grep i915
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Excess function parameter 'jump_whitelist' description in 'intel_engine_cmd_parser'
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Excess function parameter 'shadow_map' description in 'intel_engine_cmd_parser'
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Excess function parameter 'batch_map' description in 'intel_engine_cmd_parser'
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Function parameter or member 'trampoline' not described in 'intel_engine_cmd_parser'
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Excess function parameter 'jump_whitelist' description in 'intel_engine_cmd_parser'
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Excess function parameter 'shadow_map' description in 'intel_engine_cmd_parser'
./drivers/gpu/drm/i915/i915_cmd_parser.c:1436: warning: Excess function parameter 'batch_map' description in 'intel_engine_cmd_parser'


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

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [Intel-gfx] ✓ Fi.CI.BAT: success for drm/i915: Get rid of fence error propagation (rev2)
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
                   ` (8 preceding siblings ...)
  (?)
@ 2021-06-03 17:04 ` Patchwork
  -1 siblings, 0 replies; 24+ messages in thread
From: Patchwork @ 2021-06-03 17:04 UTC (permalink / raw)
  To: Jason Ekstrand; +Cc: intel-gfx


[-- Attachment #1.1: Type: text/plain, Size: 6801 bytes --]

== Series Details ==

Series: drm/i915: Get rid of fence error propagation (rev2)
URL   : https://patchwork.freedesktop.org/series/90891/
State : success

== Summary ==

CI Bug Log - changes from CI_DRM_10165 -> Patchwork_20274
====================================================

Summary
-------

  **SUCCESS**

  No regressions found.

  External URL: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/index.html

Known issues
------------

  Here are the changes found in Patchwork_20274 that come from known issues:

### IGT changes ###

#### Issues hit ####

  * igt@kms_frontbuffer_tracking@basic:
    - fi-icl-u2:          [PASS][1] -> [FAIL][2] ([i915#49])
   [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-icl-u2/igt@kms_frontbuffer_tracking@basic.html
   [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-icl-u2/igt@kms_frontbuffer_tracking@basic.html

  
#### Possible fixes ####

  * igt@i915_selftest@live@hangcheck:
    - {fi-hsw-gt1}:       [DMESG-WARN][3] ([i915#3303]) -> [PASS][4]
   [3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-hsw-gt1/igt@i915_selftest@live@hangcheck.html
   [4]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-hsw-gt1/igt@i915_selftest@live@hangcheck.html

  * igt@kms_frontbuffer_tracking@basic:
    - fi-tgl-u2:          [FAIL][5] ([i915#2416]) -> [PASS][6]
   [5]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-tgl-u2/igt@kms_frontbuffer_tracking@basic.html
   [6]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-tgl-u2/igt@kms_frontbuffer_tracking@basic.html

  
#### Warnings ####

  * igt@i915_selftest@live@execlists:
    - fi-bsw-nick:        [DMESG-FAIL][7] ([i915#3462]) -> [INCOMPLETE][8] ([i915#2782] / [i915#2940] / [i915#3462])
   [7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-bsw-nick/igt@i915_selftest@live@execlists.html
   [8]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-bsw-nick/igt@i915_selftest@live@execlists.html
    - fi-tgl-u2:          [INCOMPLETE][9] ([i915#3462]) -> [DMESG-FAIL][10] ([i915#3462])
   [9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-tgl-u2/igt@i915_selftest@live@execlists.html
   [10]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-tgl-u2/igt@i915_selftest@live@execlists.html

  * igt@runner@aborted:
    - fi-skl-6600u:       [FAIL][11] ([i915#1436] / [i915#2426] / [i915#3363]) -> [FAIL][12] ([i915#1436] / [i915#3363])
   [11]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-skl-6600u/igt@runner@aborted.html
   [12]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-skl-6600u/igt@runner@aborted.html
    - fi-glk-dsi:         [FAIL][13] ([i915#2426] / [i915#3363] / [k.org#202321]) -> [FAIL][14] ([i915#3363] / [k.org#202321])
   [13]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-glk-dsi/igt@runner@aborted.html
   [14]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-glk-dsi/igt@runner@aborted.html
    - fi-kbl-soraka:      [FAIL][15] ([i915#1436] / [i915#3363]) -> [FAIL][16] ([i915#1436] / [i915#2426] / [i915#3363])
   [15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-kbl-soraka/igt@runner@aborted.html
   [16]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-kbl-soraka/igt@runner@aborted.html
    - fi-cml-u2:          [FAIL][17] ([i915#2082] / [i915#2426] / [i915#3363] / [i915#3462]) -> [FAIL][18] ([i915#3363] / [i915#3462])
   [17]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-cml-u2/igt@runner@aborted.html
   [18]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-cml-u2/igt@runner@aborted.html
    - fi-cml-s:           [FAIL][19] ([i915#3363] / [i915#3462]) -> [FAIL][20] ([i915#2082] / [i915#2426] / [i915#3363] / [i915#3462])
   [19]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-cml-s/igt@runner@aborted.html
   [20]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-cml-s/igt@runner@aborted.html
    - fi-cfl-guc:         [FAIL][21] ([i915#3363]) -> [FAIL][22] ([i915#2426] / [i915#3363])
   [21]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-cfl-guc/igt@runner@aborted.html
   [22]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-cfl-guc/igt@runner@aborted.html
    - fi-kbl-7567u:       [FAIL][23] ([i915#1436] / [i915#2426] / [i915#3363]) -> [FAIL][24] ([i915#1436] / [i915#3363])
   [23]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/fi-kbl-7567u/igt@runner@aborted.html
   [24]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/fi-kbl-7567u/igt@runner@aborted.html

  
  {name}: This element is suppressed. This means it is ignored when computing
          the status of the difference (SUCCESS, WARNING, or FAILURE).

  [i915#1436]: https://gitlab.freedesktop.org/drm/intel/issues/1436
  [i915#2082]: https://gitlab.freedesktop.org/drm/intel/issues/2082
  [i915#2416]: https://gitlab.freedesktop.org/drm/intel/issues/2416
  [i915#2426]: https://gitlab.freedesktop.org/drm/intel/issues/2426
  [i915#2782]: https://gitlab.freedesktop.org/drm/intel/issues/2782
  [i915#2932]: https://gitlab.freedesktop.org/drm/intel/issues/2932
  [i915#2940]: https://gitlab.freedesktop.org/drm/intel/issues/2940
  [i915#2966]: https://gitlab.freedesktop.org/drm/intel/issues/2966
  [i915#3277]: https://gitlab.freedesktop.org/drm/intel/issues/3277
  [i915#3283]: https://gitlab.freedesktop.org/drm/intel/issues/3283
  [i915#3303]: https://gitlab.freedesktop.org/drm/intel/issues/3303
  [i915#3363]: https://gitlab.freedesktop.org/drm/intel/issues/3363
  [i915#3462]: https://gitlab.freedesktop.org/drm/intel/issues/3462
  [i915#49]: https://gitlab.freedesktop.org/drm/intel/issues/49
  [k.org#202321]: https://bugzilla.kernel.org/show_bug.cgi?id=202321


Participating hosts (46 -> 41)
------------------------------

  Missing    (5): fi-ilk-m540 fi-hsw-4200u fi-skl-guc fi-bsw-cyan fi-bdw-samus 


Build changes
-------------

  * Linux: CI_DRM_10165 -> Patchwork_20274

  CI-20190529: 20190529
  CI_DRM_10165: 91b8d0406bd51d9bca01127e0bcb585312fd2950 @ git://anongit.freedesktop.org/gfx-ci/linux
  IGT_6098: 1fbc1e7d602f96a7f4e2b95057eef994656b8e74 @ git://anongit.freedesktop.org/xorg/app/intel-gpu-tools
  Patchwork_20274: 1013d3982bcbf87d919298c36a244a87eff5e085 @ git://anongit.freedesktop.org/gfx-ci/linux


== Linux commits ==

1013d3982bcb Revert "drm/i915: Skip over MI_NOOP when parsing"
8e472c11e36c drm/i915: Drop error handling from dma_fence_work
7b00404854f1 drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
b6fda25b1d9f Revert "drm/i915: Propagate errors on awaiting already signaled fences"
20301afd39c9 drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/index.html

[-- Attachment #1.2: Type: text/html, Size: 9514 bytes --]

[-- Attachment #2: Type: text/plain, Size: 160 bytes --]

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

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [Intel-gfx] ✗ Fi.CI.IGT: failure for drm/i915: Get rid of fence error propagation (rev2)
  2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
                   ` (9 preceding siblings ...)
  (?)
@ 2021-06-03 18:09 ` Patchwork
  -1 siblings, 0 replies; 24+ messages in thread
From: Patchwork @ 2021-06-03 18:09 UTC (permalink / raw)
  To: Jason Ekstrand; +Cc: intel-gfx


[-- Attachment #1.1: Type: text/plain, Size: 30274 bytes --]

== Series Details ==

Series: drm/i915: Get rid of fence error propagation (rev2)
URL   : https://patchwork.freedesktop.org/series/90891/
State : failure

== Summary ==

CI Bug Log - changes from CI_DRM_10165_full -> Patchwork_20274_full
====================================================

Summary
-------

  **FAILURE**

  Serious unknown changes coming with Patchwork_20274_full absolutely need to be
  verified manually.
  
  If you think the reported changes have nothing to do with the changes
  introduced in Patchwork_20274_full, please notify your bug team to allow them
  to document this new failure mode, which will reduce false positives in CI.

  

Possible new issues
-------------------

  Here are the unknown changes that may have been introduced in Patchwork_20274_full:

### IGT changes ###

#### Possible regressions ####

  * igt@gem_exec_reloc@basic-spin@bcs0:
    - shard-kbl:          [PASS][1] -> [FAIL][2] +1 similar issue
   [1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl4/igt@gem_exec_reloc@basic-spin@bcs0.html
   [2]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@gem_exec_reloc@basic-spin@bcs0.html
    - shard-apl:          NOTRUN -> [FAIL][3]
   [3]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@gem_exec_reloc@basic-spin@bcs0.html

  * igt@gem_exec_schedule@deep@bcs0:
    - shard-skl:          [PASS][4] -> [FAIL][5] +1 similar issue
   [4]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl5/igt@gem_exec_schedule@deep@bcs0.html
   [5]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl4/igt@gem_exec_schedule@deep@bcs0.html
    - shard-glk:          [PASS][6] -> [FAIL][7] +1 similar issue
   [6]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk2/igt@gem_exec_schedule@deep@bcs0.html
   [7]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk5/igt@gem_exec_schedule@deep@bcs0.html
    - shard-apl:          [PASS][8] -> [FAIL][9]
   [8]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl3/igt@gem_exec_schedule@deep@bcs0.html
   [9]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl6/igt@gem_exec_schedule@deep@bcs0.html

  * igt@kms_big_fb@linear-16bpp-rotate-180:
    - shard-iclb:         [PASS][10] -> [FAIL][11]
   [10]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb7/igt@kms_big_fb@linear-16bpp-rotate-180.html
   [11]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb2/igt@kms_big_fb@linear-16bpp-rotate-180.html

  
Known issues
------------

  Here are the changes found in Patchwork_20274_full that come from known issues:

### IGT changes ###

#### Issues hit ####

  * igt@gem_create@create-clear:
    - shard-glk:          [PASS][12] -> [FAIL][13] ([i915#3160])
   [12]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk2/igt@gem_create@create-clear.html
   [13]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk8/igt@gem_create@create-clear.html

  * igt@gem_create@create-massive:
    - shard-apl:          NOTRUN -> [DMESG-WARN][14] ([i915#3002])
   [14]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl3/igt@gem_create@create-massive.html

  * igt@gem_ctx_persistence@legacy-engines-cleanup:
    - shard-snb:          NOTRUN -> [SKIP][15] ([fdo#109271] / [i915#1099])
   [15]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-snb7/igt@gem_ctx_persistence@legacy-engines-cleanup.html

  * igt@gem_eio@in-flight-contexts-10ms:
    - shard-iclb:         [PASS][16] -> [TIMEOUT][17] ([i915#3070])
   [16]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb2/igt@gem_eio@in-flight-contexts-10ms.html
   [17]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb8/igt@gem_eio@in-flight-contexts-10ms.html

  * igt@gem_exec_fair@basic-none@vcs1:
    - shard-kbl:          [PASS][18] -> [FAIL][19] ([i915#2842])
   [18]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl7/igt@gem_exec_fair@basic-none@vcs1.html
   [19]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl7/igt@gem_exec_fair@basic-none@vcs1.html

  * igt@gem_exec_reloc@basic-many-active@bcs0:
    - shard-apl:          [PASS][20] -> [FAIL][21] ([i915#2389])
   [20]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl8/igt@gem_exec_reloc@basic-many-active@bcs0.html
   [21]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl3/igt@gem_exec_reloc@basic-many-active@bcs0.html
    - shard-skl:          NOTRUN -> [FAIL][22] ([i915#2389])
   [22]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@gem_exec_reloc@basic-many-active@bcs0.html
    - shard-kbl:          [PASS][23] -> [FAIL][24] ([i915#2389])
   [23]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl3/igt@gem_exec_reloc@basic-many-active@bcs0.html
   [24]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl7/igt@gem_exec_reloc@basic-many-active@bcs0.html
    - shard-glk:          [PASS][25] -> [FAIL][26] ([i915#2389])
   [25]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk9/igt@gem_exec_reloc@basic-many-active@bcs0.html
   [26]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk6/igt@gem_exec_reloc@basic-many-active@bcs0.html

  * igt@gem_exec_reloc@basic-parallel:
    - shard-kbl:          [PASS][27] -> [TIMEOUT][28] ([i915#3183])
   [27]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl2/igt@gem_exec_reloc@basic-parallel.html
   [28]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl4/igt@gem_exec_reloc@basic-parallel.html
    - shard-apl:          [PASS][29] -> [TIMEOUT][30] ([i915#3183])
   [29]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl8/igt@gem_exec_reloc@basic-parallel.html
   [30]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl1/igt@gem_exec_reloc@basic-parallel.html

  * igt@gem_huc_copy@huc-copy:
    - shard-apl:          NOTRUN -> [SKIP][31] ([fdo#109271] / [i915#2190])
   [31]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@gem_huc_copy@huc-copy.html

  * igt@gem_mmap_gtt@cpuset-big-copy:
    - shard-iclb:         [PASS][32] -> [FAIL][33] ([i915#307])
   [32]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb7/igt@gem_mmap_gtt@cpuset-big-copy.html
   [33]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb3/igt@gem_mmap_gtt@cpuset-big-copy.html

  * igt@gem_userptr_blits@vma-merge:
    - shard-kbl:          NOTRUN -> [FAIL][34] ([i915#3318])
   [34]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@gem_userptr_blits@vma-merge.html

  * igt@i915_pm_dc@dc6-dpms:
    - shard-skl:          NOTRUN -> [FAIL][35] ([i915#454])
   [35]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@i915_pm_dc@dc6-dpms.html

  * igt@i915_selftest@live@execlists:
    - shard-apl:          NOTRUN -> [DMESG-FAIL][36] ([i915#3462])
   [36]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@i915_selftest@live@execlists.html

  * igt@kms_ccs@pipe-c-random-ccs-data:
    - shard-skl:          NOTRUN -> [SKIP][37] ([fdo#109271] / [fdo#111304])
   [37]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_ccs@pipe-c-random-ccs-data.html

  * igt@kms_chamelium@dp-hpd-for-each-pipe:
    - shard-kbl:          NOTRUN -> [SKIP][38] ([fdo#109271] / [fdo#111827]) +1 similar issue
   [38]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@kms_chamelium@dp-hpd-for-each-pipe.html

  * igt@kms_chamelium@hdmi-edid-change-during-suspend:
    - shard-apl:          NOTRUN -> [SKIP][39] ([fdo#109271] / [fdo#111827]) +16 similar issues
   [39]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@kms_chamelium@hdmi-edid-change-during-suspend.html

  * igt@kms_chamelium@vga-hpd:
    - shard-skl:          NOTRUN -> [SKIP][40] ([fdo#109271] / [fdo#111827]) +3 similar issues
   [40]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_chamelium@vga-hpd.html

  * igt@kms_color_chamelium@pipe-d-ctm-0-5:
    - shard-snb:          NOTRUN -> [SKIP][41] ([fdo#109271] / [fdo#111827]) +4 similar issues
   [41]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-snb7/igt@kms_color_chamelium@pipe-d-ctm-0-5.html

  * igt@kms_content_protection@atomic-dpms:
    - shard-apl:          NOTRUN -> [TIMEOUT][42] ([i915#1319])
   [42]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@kms_content_protection@atomic-dpms.html

  * igt@kms_cursor_crc@pipe-b-cursor-256x256-sliding:
    - shard-skl:          [PASS][43] -> [FAIL][44] ([i915#3444])
   [43]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl8/igt@kms_cursor_crc@pipe-b-cursor-256x256-sliding.html
   [44]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl9/igt@kms_cursor_crc@pipe-b-cursor-256x256-sliding.html

  * igt@kms_cursor_legacy@pipe-d-single-bo:
    - shard-kbl:          NOTRUN -> [SKIP][45] ([fdo#109271] / [i915#533])
   [45]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@kms_cursor_legacy@pipe-d-single-bo.html

  * igt@kms_fbcon_fbt@fbc-suspend:
    - shard-apl:          [PASS][46] -> [INCOMPLETE][47] ([i915#180] / [i915#1982])
   [46]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl8/igt@kms_fbcon_fbt@fbc-suspend.html
   [47]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl6/igt@kms_fbcon_fbt@fbc-suspend.html
    - shard-kbl:          [PASS][48] -> [INCOMPLETE][49] ([i915#155] / [i915#180] / [i915#636])
   [48]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl4/igt@kms_fbcon_fbt@fbc-suspend.html
   [49]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@kms_fbcon_fbt@fbc-suspend.html

  * igt@kms_flip@absolute-wf_vblank-interruptible@a-edp1:
    - shard-skl:          NOTRUN -> [DMESG-WARN][50] ([i915#1982])
   [50]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_flip@absolute-wf_vblank-interruptible@a-edp1.html

  * igt@kms_flip@plain-flip-ts-check@a-edp1:
    - shard-skl:          [PASS][51] -> [FAIL][52] ([i915#2122])
   [51]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl10/igt@kms_flip@plain-flip-ts-check@a-edp1.html
   [52]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl7/igt@kms_flip@plain-flip-ts-check@a-edp1.html

  * igt@kms_flip_scaled_crc@flip-64bpp-ytile-to-32bpp-ytilercccs:
    - shard-apl:          NOTRUN -> [SKIP][53] ([fdo#109271] / [i915#2672])
   [53]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@kms_flip_scaled_crc@flip-64bpp-ytile-to-32bpp-ytilercccs.html

  * igt@kms_frontbuffer_tracking@fbcpsr-1p-primscrn-indfb-msflip-blt:
    - shard-skl:          NOTRUN -> [SKIP][54] ([fdo#109271]) +33 similar issues
   [54]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_frontbuffer_tracking@fbcpsr-1p-primscrn-indfb-msflip-blt.html

  * igt@kms_hdr@bpc-switch:
    - shard-skl:          [PASS][55] -> [FAIL][56] ([i915#1188])
   [55]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl10/igt@kms_hdr@bpc-switch.html
   [56]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl2/igt@kms_hdr@bpc-switch.html

  * igt@kms_hdr@bpc-switch-suspend:
    - shard-kbl:          [PASS][57] -> [DMESG-WARN][58] ([i915#180]) +5 similar issues
   [57]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl4/igt@kms_hdr@bpc-switch-suspend.html
   [58]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@kms_hdr@bpc-switch-suspend.html

  * igt@kms_pipe_crc_basic@disable-crc-after-crtc-pipe-d:
    - shard-apl:          NOTRUN -> [SKIP][59] ([fdo#109271] / [i915#533])
   [59]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@kms_pipe_crc_basic@disable-crc-after-crtc-pipe-d.html

  * igt@kms_plane_alpha_blend@pipe-b-alpha-basic:
    - shard-apl:          NOTRUN -> [FAIL][60] ([fdo#108145] / [i915#265])
   [60]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@kms_plane_alpha_blend@pipe-b-alpha-basic.html

  * igt@kms_plane_alpha_blend@pipe-b-alpha-transparent-fb:
    - shard-apl:          NOTRUN -> [FAIL][61] ([i915#265])
   [61]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl2/igt@kms_plane_alpha_blend@pipe-b-alpha-transparent-fb.html

  * igt@kms_plane_alpha_blend@pipe-b-constant-alpha-max:
    - shard-skl:          NOTRUN -> [FAIL][62] ([fdo#108145] / [i915#265])
   [62]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_plane_alpha_blend@pipe-b-constant-alpha-max.html

  * igt@kms_plane_alpha_blend@pipe-b-constant-alpha-min:
    - shard-skl:          [PASS][63] -> [FAIL][64] ([fdo#108145] / [i915#265])
   [63]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl8/igt@kms_plane_alpha_blend@pipe-b-constant-alpha-min.html
   [64]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl9/igt@kms_plane_alpha_blend@pipe-b-constant-alpha-min.html

  * igt@kms_plane_alpha_blend@pipe-c-constant-alpha-max:
    - shard-kbl:          NOTRUN -> [FAIL][65] ([fdo#108145] / [i915#265])
   [65]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@kms_plane_alpha_blend@pipe-c-constant-alpha-max.html

  * igt@kms_plane_alpha_blend@pipe-d-constant-alpha-mid:
    - shard-kbl:          NOTRUN -> [SKIP][66] ([fdo#109271]) +18 similar issues
   [66]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@kms_plane_alpha_blend@pipe-d-constant-alpha-mid.html

  * igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-2:
    - shard-apl:          NOTRUN -> [SKIP][67] ([fdo#109271] / [i915#658]) +3 similar issues
   [67]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl1/igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-2.html

  * igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-3:
    - shard-skl:          NOTRUN -> [SKIP][68] ([fdo#109271] / [i915#658])
   [68]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-3.html

  * igt@kms_psr2_su@frontbuffer:
    - shard-iclb:         [PASS][69] -> [SKIP][70] ([fdo#109642] / [fdo#111068] / [i915#658]) +1 similar issue
   [69]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb2/igt@kms_psr2_su@frontbuffer.html
   [70]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb4/igt@kms_psr2_su@frontbuffer.html
    - shard-kbl:          NOTRUN -> [SKIP][71] ([fdo#109271] / [i915#658])
   [71]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@kms_psr2_su@frontbuffer.html

  * igt@kms_psr@psr2_cursor_plane_onoff:
    - shard-iclb:         [PASS][72] -> [SKIP][73] ([fdo#109441]) +1 similar issue
   [72]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb2/igt@kms_psr@psr2_cursor_plane_onoff.html
   [73]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb8/igt@kms_psr@psr2_cursor_plane_onoff.html

  * igt@kms_vblank@pipe-b-accuracy-idle:
    - shard-glk:          [PASS][74] -> [FAIL][75] ([i915#43])
   [74]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk4/igt@kms_vblank@pipe-b-accuracy-idle.html
   [75]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk6/igt@kms_vblank@pipe-b-accuracy-idle.html

  * igt@kms_vblank@pipe-d-query-forked-hang:
    - shard-snb:          NOTRUN -> [SKIP][76] ([fdo#109271]) +81 similar issues
   [76]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-snb7/igt@kms_vblank@pipe-d-query-forked-hang.html

  * igt@kms_vblank@pipe-d-wait-forked-hang:
    - shard-apl:          NOTRUN -> [SKIP][77] ([fdo#109271]) +135 similar issues
   [77]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@kms_vblank@pipe-d-wait-forked-hang.html

  * igt@kms_writeback@writeback-invalid-parameters:
    - shard-skl:          NOTRUN -> [SKIP][78] ([fdo#109271] / [i915#2437])
   [78]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@kms_writeback@writeback-invalid-parameters.html

  * igt@kms_writeback@writeback-pixel-formats:
    - shard-kbl:          NOTRUN -> [SKIP][79] ([fdo#109271] / [i915#2437])
   [79]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@kms_writeback@writeback-pixel-formats.html

  * igt@perf@polling:
    - shard-skl:          [PASS][80] -> [FAIL][81] ([i915#1542])
   [80]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl3/igt@perf@polling.html
   [81]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl1/igt@perf@polling.html

  * igt@perf@polling-small-buf:
    - shard-skl:          [PASS][82] -> [FAIL][83] ([i915#1722])
   [82]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl10/igt@perf@polling-small-buf.html
   [83]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl2/igt@perf@polling-small-buf.html

  * igt@sysfs_clients@fair-0:
    - shard-apl:          NOTRUN -> [SKIP][84] ([fdo#109271] / [i915#2994])
   [84]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@sysfs_clients@fair-0.html

  * igt@sysfs_clients@fair-7:
    - shard-skl:          NOTRUN -> [SKIP][85] ([fdo#109271] / [i915#2994])
   [85]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl6/igt@sysfs_clients@fair-7.html

  * igt@sysfs_clients@recycle:
    - shard-kbl:          NOTRUN -> [SKIP][86] ([fdo#109271] / [i915#2994])
   [86]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl4/igt@sysfs_clients@recycle.html

  
#### Possible fixes ####

  * igt@gem_ctx_isolation@preservation-s3@vcs0:
    - shard-kbl:          [DMESG-WARN][87] ([i915#180]) -> [PASS][88] +3 similar issues
   [87]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl1/igt@gem_ctx_isolation@preservation-s3@vcs0.html
   [88]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@gem_ctx_isolation@preservation-s3@vcs0.html

  * igt@gem_eio@in-flight-contexts-10ms:
    - shard-snb:          [INCOMPLETE][89] ([i915#3426]) -> [PASS][90]
   [89]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-snb7/igt@gem_eio@in-flight-contexts-10ms.html
   [90]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-snb6/igt@gem_eio@in-flight-contexts-10ms.html

  * igt@gem_exec_fair@basic-none-share@rcs0:
    - shard-iclb:         [FAIL][91] ([i915#2842]) -> [PASS][92]
   [91]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb5/igt@gem_exec_fair@basic-none-share@rcs0.html
   [92]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb6/igt@gem_exec_fair@basic-none-share@rcs0.html
    - shard-tglb:         [FAIL][93] ([i915#2842]) -> [PASS][94]
   [93]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-tglb1/igt@gem_exec_fair@basic-none-share@rcs0.html
   [94]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-tglb6/igt@gem_exec_fair@basic-none-share@rcs0.html

  * igt@gem_exec_fair@basic-pace@vecs0:
    - shard-kbl:          [SKIP][95] ([fdo#109271]) -> [PASS][96]
   [95]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl1/igt@gem_exec_fair@basic-pace@vecs0.html
   [96]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl7/igt@gem_exec_fair@basic-pace@vecs0.html

  * igt@gem_exec_fair@basic-throttle@rcs0:
    - shard-glk:          [FAIL][97] ([i915#2842]) -> [PASS][98]
   [97]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk4/igt@gem_exec_fair@basic-throttle@rcs0.html
   [98]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk6/igt@gem_exec_fair@basic-throttle@rcs0.html

  * igt@gem_exec_whisper@basic-contexts-all:
    - shard-glk:          [DMESG-WARN][99] ([i915#118] / [i915#95]) -> [PASS][100]
   [99]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk9/igt@gem_exec_whisper@basic-contexts-all.html
   [100]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk6/igt@gem_exec_whisper@basic-contexts-all.html

  * igt@gem_mmap_gtt@big-copy-odd:
    - shard-glk:          [FAIL][101] ([i915#307]) -> [PASS][102]
   [101]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk3/igt@gem_mmap_gtt@big-copy-odd.html
   [102]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk2/igt@gem_mmap_gtt@big-copy-odd.html

  * igt@gem_mmap_gtt@cpuset-medium-copy:
    - shard-iclb:         [FAIL][103] ([i915#307]) -> [PASS][104]
   [103]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb1/igt@gem_mmap_gtt@cpuset-medium-copy.html
   [104]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb1/igt@gem_mmap_gtt@cpuset-medium-copy.html

  * igt@gen9_exec_parse@bb-large:
    - shard-glk:          [FAIL][105] ([i915#3296]) -> [PASS][106]
   [105]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-glk2/igt@gen9_exec_parse@bb-large.html
   [106]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-glk7/igt@gen9_exec_parse@bb-large.html
    - shard-kbl:          [FAIL][107] ([i915#3296]) -> [PASS][108]
   [107]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl2/igt@gen9_exec_parse@bb-large.html
   [108]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl2/igt@gen9_exec_parse@bb-large.html

  * igt@kms_color@pipe-c-ctm-0-25:
    - shard-skl:          [DMESG-WARN][109] ([i915#1982]) -> [PASS][110] +1 similar issue
   [109]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl7/igt@kms_color@pipe-c-ctm-0-25.html
   [110]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl9/igt@kms_color@pipe-c-ctm-0-25.html

  * igt@kms_cursor_legacy@flip-vs-cursor-atomic-transitions:
    - shard-skl:          [FAIL][111] ([i915#2346]) -> [PASS][112]
   [111]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl10/igt@kms_cursor_legacy@flip-vs-cursor-atomic-transitions.html
   [112]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl7/igt@kms_cursor_legacy@flip-vs-cursor-atomic-transitions.html

  * igt@kms_flip@flip-vs-expired-vblank@c-edp1:
    - shard-skl:          [FAIL][113] ([i915#79]) -> [PASS][114]
   [113]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl1/igt@kms_flip@flip-vs-expired-vblank@c-edp1.html
   [114]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl5/igt@kms_flip@flip-vs-expired-vblank@c-edp1.html

  * igt@kms_hdr@bpc-switch-suspend:
    - shard-apl:          [DMESG-WARN][115] ([i915#180]) -> [PASS][116]
   [115]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl1/igt@kms_hdr@bpc-switch-suspend.html
   [116]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl2/igt@kms_hdr@bpc-switch-suspend.html

  * igt@kms_psr@psr2_sprite_blt:
    - shard-iclb:         [SKIP][117] ([fdo#109441]) -> [PASS][118] +1 similar issue
   [117]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb7/igt@kms_psr@psr2_sprite_blt.html
   [118]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb2/igt@kms_psr@psr2_sprite_blt.html

  * igt@kms_vblank@pipe-a-ts-continuation-suspend:
    - shard-kbl:          [DMESG-WARN][119] ([i915#180] / [i915#295]) -> [PASS][120]
   [119]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl1/igt@kms_vblank@pipe-a-ts-continuation-suspend.html
   [120]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl4/igt@kms_vblank@pipe-a-ts-continuation-suspend.html

  
#### Warnings ####

  * igt@gen9_exec_parse@bb-large:
    - shard-skl:          [FAIL][121] ([i915#3296]) -> [SKIP][122] ([fdo#109271])
   [121]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl8/igt@gen9_exec_parse@bb-large.html
   [122]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl3/igt@gen9_exec_parse@bb-large.html

  * igt@i915_pm_rc6_residency@rc6-idle:
    - shard-iclb:         [WARN][123] ([i915#2684]) -> [WARN][124] ([i915#1804] / [i915#2684])
   [123]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb1/igt@i915_pm_rc6_residency@rc6-idle.html
   [124]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb6/igt@i915_pm_rc6_residency@rc6-idle.html

  * igt@i915_selftest@live@execlists:
    - shard-tglb:         [DMESG-FAIL][125] ([i915#3462]) -> [INCOMPLETE][126] ([i915#3462])
   [125]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-tglb1/igt@i915_selftest@live@execlists.html
   [126]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-tglb7/igt@i915_selftest@live@execlists.html
    - shard-iclb:         [INCOMPLETE][127] ([i915#2782] / [i915#3462]) -> [DMESG-FAIL][128] ([i915#3462])
   [127]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb4/igt@i915_selftest@live@execlists.html
   [128]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb7/igt@i915_selftest@live@execlists.html

  * igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-2:
    - shard-iclb:         [SKIP][129] ([i915#2920]) -> [SKIP][130] ([i915#658]) +1 similar issue
   [129]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb2/igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-2.html
   [130]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb5/igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-2.html

  * igt@kms_psr2_sf@overlay-primary-update-sf-dmg-area-4:
    - shard-iclb:         [SKIP][131] ([i915#658]) -> [SKIP][132] ([i915#2920])
   [131]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-iclb3/igt@kms_psr2_sf@overlay-primary-update-sf-dmg-area-4.html
   [132]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-iclb2/igt@kms_psr2_sf@overlay-primary-update-sf-dmg-area-4.html

  * igt@runner@aborted:
    - shard-kbl:          ([FAIL][133], [FAIL][134], [FAIL][135], [FAIL][136], [FAIL][137], [FAIL][138]) ([i915#1436] / [i915#180] / [i915#1814] / [i915#3002] / [i915#3363] / [i915#602]) -> ([FAIL][139], [FAIL][140], [FAIL][141], [FAIL][142], [FAIL][143], [FAIL][144], [FAIL][145], [FAIL][146]) ([i915#1436] / [i915#180] / [i915#1814] / [i915#3002] / [i915#3363] / [i915#602] / [i915#92])
   [133]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl7/igt@runner@aborted.html
   [134]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl3/igt@runner@aborted.html
   [135]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl1/igt@runner@aborted.html
   [136]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl1/igt@runner@aborted.html
   [137]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl1/igt@runner@aborted.html
   [138]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-kbl2/igt@runner@aborted.html
   [139]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@runner@aborted.html
   [140]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@runner@aborted.html
   [141]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@runner@aborted.html
   [142]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@runner@aborted.html
   [143]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@runner@aborted.html
   [144]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl1/igt@runner@aborted.html
   [145]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl7/igt@runner@aborted.html
   [146]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-kbl3/igt@runner@aborted.html
    - shard-apl:          ([FAIL][147], [FAIL][148]) ([i915#180] / [i915#1814] / [i915#3363]) -> ([FAIL][149], [FAIL][150], [FAIL][151]) ([fdo#109271] / [i915#180] / [i915#3002] / [i915#3363])
   [147]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl2/igt@runner@aborted.html
   [148]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-apl1/igt@runner@aborted.html
   [149]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl3/igt@runner@aborted.html
   [150]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl7/igt@runner@aborted.html
   [151]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-apl6/igt@runner@aborted.html
    - shard-tglb:         ([FAIL][152], [FAIL][153], [FAIL][154]) ([i915#1436] / [i915#2426] / [i915#2966] / [i915#3002]) -> ([FAIL][155], [FAIL][156], [FAIL][157]) ([i915#1436] / [i915#2966] / [i915#3002])
   [152]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-tglb8/igt@runner@aborted.html
   [153]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-tglb1/igt@runner@aborted.html
   [154]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-tglb7/igt@runner@aborted.html
   [155]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-tglb5/igt@runner@aborted.html
   [156]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-tglb2/igt@runner@aborted.html
   [157]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-tglb7/igt@runner@aborted.html
    - shard-skl:          ([FAIL][158], [FAIL][159], [FAIL][160], [FAIL][161]) ([i915#1436] / [i915#2029] / [i915#3002] / [i915#3363]) -> ([FAIL][162], [FAIL][163], [FAIL][164]) ([i915#1436] / [i915#3002] / [i915#3363])
   [158]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl2/igt@runner@aborted.html
   [159]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl6/igt@runner@aborted.html
   [160]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl3/igt@runner@aborted.html
   [161]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10165/shard-skl5/igt@runner@aborted.html
   [162]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl7/igt@runner@aborted.html
   [163]: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/shard-skl8/igt@runne

== Logs ==

For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_20274/index.html

[-- Attachment #1.2: Type: text/html, Size: 33597 bytes --]

[-- Attachment #2: Type: text/plain, Size: 160 bytes --]

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

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [PATCH 0/5] drm/i915: Get rid of fence error propagation (v4)
@ 2021-07-14 19:34 Jason Ekstrand
  2021-07-14 19:34   ` Jason Ekstrand
  0 siblings, 1 reply; 24+ messages in thread
From: Jason Ekstrand @ 2021-07-14 19:34 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Jason Ekstrand

Fence error propagation is sketchy at best.  Instead of explicitly handling
fences which might have errors set in the code which is aware of errors, we
just kick them down the line and hope that userspace knows what to do when
a wait eventually fails.  This is sketchy at best because most userspace
isn't prepared to handle errors in those places.  To make things worse, it
allows errors to propagate across processes in unpredictable ways.  This is
causing hangs in one client to kill X11.

Unfortunately, there's no quick path from here to there thanks to the fact
that we're now running the command parser asynchronously and relying on
fence errors for when it fails.  This series first gets rid of asynchronous
command parsing and then cleans up from there.  There was never any real
use-case for asynchronous parsing and the platforms that rely heavily on
the command parser are old enough (Gen7) that, when we changed the way the
command parser works, it wasn't really a change anyone was asking for
anyway.

I think we probably want this whole mess back-ported.  I'm happy to take
suggestions on the strategy there because the history there is a bit
annoying and I'm not 100% sure where the Linux release cuts land.  In any
case, I'm happy to make a version of this series per-release if needed for
Greg to back-port.

v2 (Daniel Vetter):
 - Re-order to put the reverts first
 - Add ACKs from Daniel
 - Add better CC and Fixes tags

v3 (Jason Ekstrand):
 - Rebase on drm-tip

v4 (Jason Ekstrand):
 - Rebase on drm-tip

Test-with: 20210714173141.1381686-1-jason@jlekstrand.net

Jason Ekstrand (5):
  drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  Revert "drm/i915: Propagate errors on awaiting already signaled
    fences"
  drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
  drm/i915: Drop error handling from dma_fence_work
  Revert "drm/i915: Skip over MI_NOOP when parsing"

 drivers/gpu/drm/i915/gem/i915_gem_clflush.c   |   4 +-
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 drivers/gpu/drm/i915/gem/i915_gem_object.h    |  10 +-
 drivers/gpu/drm/i915/gem/i915_gem_pages.c     |  20 +-
 drivers/gpu/drm/i915/gem/i915_gem_ttm.c       |   2 +-
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/gt/intel_ggtt.c          |   2 +-
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 199 ++++++++-------
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 drivers/gpu/drm/i915/i915_request.c           |   8 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.c     |   5 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.h     |   2 +-
 drivers/gpu/drm/i915/i915_vma.c               |   3 +-
 13 files changed, 142 insertions(+), 351 deletions(-)

-- 
2.31.1

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  2021-07-14 19:34 [PATCH 0/5] drm/i915: Get rid of fence error propagation (v4) Jason Ekstrand
@ 2021-07-14 19:34   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-07-14 19:34 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Jason Ekstrand, stable, Maarten Lankhorst, Jon Bloomfield, Daniel Vetter

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 5ea8b4e23e428..1ed7475de454d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1471,6 +1469,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2385,217 +2387,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2685,7 +2476,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 3992c25a191da..00ec618d01590 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index c4747f4407ef1..d38167a77ec00 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1957,17 +1957,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
@ 2021-07-14 19:34   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-07-14 19:34 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Daniel Vetter, Jon Bloomfield, stable, Jason Ekstrand

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 5ea8b4e23e428..1ed7475de454d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1471,6 +1469,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2385,217 +2387,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2685,7 +2476,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 3992c25a191da..00ec618d01590 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index c4747f4407ef1..d38167a77ec00 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1957,17 +1957,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 0/5] drm/i915: Get rid of fence error propagation (v2)
@ 2021-07-11  3:53 Jason Ekstrand
  2021-07-11  3:53   ` Jason Ekstrand
  0 siblings, 1 reply; 24+ messages in thread
From: Jason Ekstrand @ 2021-07-11  3:53 UTC (permalink / raw)
  To: intel-gfx, dri-devel; +Cc: Daniel Vetter, Jon Bloomfield, Jason Ekstrand

Fence error propagation is sketchy at best.  Instead of explicitly handling
fences which might have errors set in the code which is aware of errors, we
just kick them down the line and hope that userspace knows what to do when
a wait eventually fails.  This is sketchy at best because most userspace
isn't prepared to handle errors in those places.  To make things worse, it
allows errors to propagate across processes in unpredictable ways.  This is
causing hangs in one client to kill X11.

Unfortunately, there's no quick path from here to there thanks to the fact
that we're now running the command parser asynchronously and relying on
fence errors for when it fails.  This series first gets rid of asynchronous
command parsing and then cleans up from there.  There was never any real
use-case for asynchronous parsing and the platforms that rely heavily on
the command parser are old enough (Gen7) that, when we changed the way the
command parser works, it wasn't really a change anyone was asking for
anyway.

I think we probably want this whole mess back-ported.  I'm happy to take
suggestions on the strategy there because the history there is a bit
annoying and I'm not 100% sure where the Linux release cuts land.  In any
case, I'm happy to make a version of this series per-release if needed for
Greg to back-port.

v2 (Daniel Vetter):
 - Re-order to put the reverts first
 - Add ACKs from Daniel
 - Add better CC and Fixes tags

v3 (Daniel Vetter):
 - Rebase on drm-tip

Test-with: 20210711035204.802908-1-jason@jlekstrand.net
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>

Jason Ekstrand (5):
  drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  Revert "drm/i915: Propagate errors on awaiting already signaled
    fences"
  drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
  drm/i915: Drop error handling from dma_fence_work
  Revert "drm/i915: Skip over MI_NOOP when parsing"

 drivers/gpu/drm/i915/gem/i915_gem_clflush.c   |   4 +-
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 drivers/gpu/drm/i915/gem/i915_gem_object.h    |  10 +-
 drivers/gpu/drm/i915/gem/i915_gem_pages.c     |  20 +-
 drivers/gpu/drm/i915/gem/i915_gem_ttm.c       |   2 +-
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/gt/intel_ggtt.c          |   2 +-
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 199 ++++++++-------
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 drivers/gpu/drm/i915/i915_request.c           |   8 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.c     |   5 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.h     |   2 +-
 drivers/gpu/drm/i915/i915_vma.c               |   3 +-
 13 files changed, 142 insertions(+), 351 deletions(-)

-- 
2.31.1

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  2021-07-11  3:53 [PATCH 0/5] drm/i915: Get rid of fence error propagation (v2) Jason Ekstrand
@ 2021-07-11  3:53   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-07-11  3:53 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Jason Ekstrand, stable, Maarten Lankhorst, Jon Bloomfield, Daniel Vetter

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 5ea8b4e23e428..1ed7475de454d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1471,6 +1469,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2385,217 +2387,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2685,7 +2476,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 3992c25a191da..00ec618d01590 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index d14cda2ff9230..5351376333b42 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1951,17 +1951,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
@ 2021-07-11  3:53   ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-07-11  3:53 UTC (permalink / raw)
  To: intel-gfx, dri-devel
  Cc: Daniel Vetter, Jon Bloomfield, stable, Jason Ekstrand

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

v2 (Daniel Vetter):
 - Add stabel Cc and Fixes tag

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: <stable@vger.kernel.org> # v5.6+
Fixes: 9e31c1fe45d5 ("drm/i915: Propagate errors on awaiting already signaled fences")
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 5ea8b4e23e428..1ed7475de454d 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1471,6 +1469,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2385,217 +2387,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2685,7 +2476,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 3992c25a191da..00ec618d01590 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index d14cda2ff9230..5351376333b42 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1951,17 +1951,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

* [PATCH 0/5] drm/i915: Get rid of fence error propagation
@ 2021-06-02 16:41 Jason Ekstrand
  2021-06-02 16:41 ` [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser" Jason Ekstrand
  0 siblings, 1 reply; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-02 16:41 UTC (permalink / raw)
  To: dri-devel, intel-gfx; +Cc: Daniel Vetter, Jon Bloomfield, Jason Ekstrand

Fence error propagation is sketchy at best.  Instead of explicitly handling
fences which might have errors set in the code which is aware of errors, we
just kick them down the line and hope that userspace knows what to do when
a wait eventually fails.  This is sketchy at best because most userspace
isn't prepared to handle errors in those places.  To make things worse, it
allows errors to propagate across processes in unpredictable ways.  This is
causing hangs in one client to kill X11.

Unfortunately, there's no quick path from here to there thanks to the fact
that we're now running the command parser asynchronously and relying on
fence errors for when it fails.  This series first gets rid of asynchronous
command parsing and then cleans up from there.  There was never any real
use-case for asynchronous parsing and the platforms that rely heavily on
the command parser are old enough (Gen7) that, when we changed the way the
command parser works, it wasn't really a change anyone was asking for
anyway.

I think we probably want this whole mess back-ported.  I'm happy to take
suggestions on the strategy there because the history there is a bit
annoying and I'm not 100% sure where the Linux release cuts land.  In any
case, I'm happy to make a version of this series per-release if needed for
Greg to back-port.

Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jon Bloomfield <jon.bloomfield@intel.com>

Jason Ekstrand (5):
  drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  drm/i915: Remove allow_alloc from i915_gem_object_get_sg*
  drm/i915: Drop error handling from dma_fence_work
  Revert "drm/i915: Propagate errors on awaiting already signaled
    fences"
  Revert "drm/i915: Skip over MI_NOOP when parsing"

 drivers/gpu/drm/i915/gem/i915_gem_clflush.c   |   4 +-
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 drivers/gpu/drm/i915/gem/i915_gem_object.h    |  10 +-
 drivers/gpu/drm/i915/gem/i915_gem_pages.c     |  21 +-
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/gt/intel_ggtt.c          |   2 +-
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 199 ++++++++-------
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 drivers/gpu/drm/i915/i915_request.c           |   8 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.c     |   5 +-
 drivers/gpu/drm/i915/i915_sw_fence_work.h     |   2 +-
 drivers/gpu/drm/i915/i915_vma.c               |   3 +-
 12 files changed, 141 insertions(+), 351 deletions(-)

-- 
2.31.1

^ permalink raw reply	[flat|nested] 24+ messages in thread

* [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser"
  2021-06-02 16:41 [PATCH 0/5] drm/i915: Get rid of fence error propagation Jason Ekstrand
@ 2021-06-02 16:41 ` Jason Ekstrand
  0 siblings, 0 replies; 24+ messages in thread
From: Jason Ekstrand @ 2021-06-02 16:41 UTC (permalink / raw)
  To: dri-devel, intel-gfx; +Cc: Jon Bloomfield, Jason Ekstrand

This reverts 686c7c35abc2 ("drm/i915/gem: Asynchronous cmdparser").  The
justification for this commit in the git history was a vague comment
about getting it out from under the struct_mutex.  While this may
improve perf for some workloads on Gen7 platforms where we rely on the
command parser for features such as indirect rendering, no numbers were
provided to prove such an improvement.  It claims to closed two
gitlab/bugzilla issues but with no explanation whatsoever as to why or
what bug it's fixing.

Meanwhile, by moving command parsing off to an async callback, it leaves
us with a problem of what to do on error.  When things were synchronous,
EXECBUFFER2 would fail with an error code if parsing failed.  When
moving it to async, we needed another way to handle that error and the
solution employed was to set an error on the dma_fence and then trust
that said error gets propagated to the client eventually.  Moving back
to synchronous will help us untangle the fence error propagation mess.

This also reverts most of 0edbb9ba1bfe ("drm/i915: Move cmd parser
pinning to execbuffer") which is a refactor of some of our allocation
paths for asynchronous parsing.  Now that everything is synchronous, we
don't need it.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Reviewed-by: Jon Bloomfield <jon.bloomfield@intel.com>
---
 .../gpu/drm/i915/gem/i915_gem_execbuffer.c    | 227 +-----------------
 .../i915/gem/selftests/i915_gem_execbuffer.c  |   4 +
 drivers/gpu/drm/i915/i915_cmd_parser.c        | 132 +++++-----
 drivers/gpu/drm/i915/i915_drv.h               |   7 +-
 4 files changed, 91 insertions(+), 279 deletions(-)

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 297143511f99b..a49da4b24d4d4 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -25,10 +25,8 @@
 #include "i915_gem_clflush.h"
 #include "i915_gem_context.h"
 #include "i915_gem_ioctls.h"
-#include "i915_sw_fence_work.h"
 #include "i915_trace.h"
 #include "i915_user_extensions.h"
-#include "i915_memcpy.h"
 
 struct eb_vma {
 	struct i915_vma *vma;
@@ -1456,6 +1454,10 @@ static u32 *reloc_gpu(struct i915_execbuffer *eb,
 		int err;
 		struct intel_engine_cs *engine = eb->engine;
 
+		/* If we need to copy for the cmdparser, we will stall anyway */
+		if (eb_use_cmdparser(eb))
+			return ERR_PTR(-EWOULDBLOCK);
+
 		if (!reloc_can_use_engine(engine)) {
 			engine = engine->gt->engine_class[COPY_ENGINE_CLASS][0];
 			if (!engine)
@@ -2372,217 +2374,6 @@ shadow_batch_pin(struct i915_execbuffer *eb,
 	return vma;
 }
 
-struct eb_parse_work {
-	struct dma_fence_work base;
-	struct intel_engine_cs *engine;
-	struct i915_vma *batch;
-	struct i915_vma *shadow;
-	struct i915_vma *trampoline;
-	unsigned long batch_offset;
-	unsigned long batch_length;
-	unsigned long *jump_whitelist;
-	const void *batch_map;
-	void *shadow_map;
-};
-
-static int __eb_parse(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-	int ret;
-	bool cookie;
-
-	cookie = dma_fence_begin_signalling();
-	ret = intel_engine_cmd_parser(pw->engine,
-				      pw->batch,
-				      pw->batch_offset,
-				      pw->batch_length,
-				      pw->shadow,
-				      pw->jump_whitelist,
-				      pw->shadow_map,
-				      pw->batch_map);
-	dma_fence_end_signalling(cookie);
-
-	return ret;
-}
-
-static void __eb_parse_release(struct dma_fence_work *work)
-{
-	struct eb_parse_work *pw = container_of(work, typeof(*pw), base);
-
-	if (!IS_ERR_OR_NULL(pw->jump_whitelist))
-		kfree(pw->jump_whitelist);
-
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(pw->batch->obj);
-	else
-		i915_gem_object_unpin_pages(pw->batch->obj);
-
-	i915_gem_object_unpin_map(pw->shadow->obj);
-
-	if (pw->trampoline)
-		i915_active_release(&pw->trampoline->active);
-	i915_active_release(&pw->shadow->active);
-	i915_active_release(&pw->batch->active);
-}
-
-static const struct dma_fence_work_ops eb_parse_ops = {
-	.name = "eb_parse",
-	.work = __eb_parse,
-	.release = __eb_parse_release,
-};
-
-static inline int
-__parser_mark_active(struct i915_vma *vma,
-		     struct intel_timeline *tl,
-		     struct dma_fence *fence)
-{
-	struct intel_gt_buffer_pool_node *node = vma->private;
-
-	return i915_active_ref(&node->active, tl->fence_context, fence);
-}
-
-static int
-parser_mark_active(struct eb_parse_work *pw, struct intel_timeline *tl)
-{
-	int err;
-
-	mutex_lock(&tl->mutex);
-
-	err = __parser_mark_active(pw->shadow, tl, &pw->base.dma);
-	if (err)
-		goto unlock;
-
-	if (pw->trampoline) {
-		err = __parser_mark_active(pw->trampoline, tl, &pw->base.dma);
-		if (err)
-			goto unlock;
-	}
-
-unlock:
-	mutex_unlock(&tl->mutex);
-	return err;
-}
-
-static int eb_parse_pipeline(struct i915_execbuffer *eb,
-			     struct i915_vma *shadow,
-			     struct i915_vma *trampoline)
-{
-	struct eb_parse_work *pw;
-	struct drm_i915_gem_object *batch = eb->batch->vma->obj;
-	bool needs_clflush;
-	int err;
-
-	GEM_BUG_ON(overflows_type(eb->batch_start_offset, pw->batch_offset));
-	GEM_BUG_ON(overflows_type(eb->batch_len, pw->batch_length));
-
-	pw = kzalloc(sizeof(*pw), GFP_KERNEL);
-	if (!pw)
-		return -ENOMEM;
-
-	err = i915_active_acquire(&eb->batch->vma->active);
-	if (err)
-		goto err_free;
-
-	err = i915_active_acquire(&shadow->active);
-	if (err)
-		goto err_batch;
-
-	if (trampoline) {
-		err = i915_active_acquire(&trampoline->active);
-		if (err)
-			goto err_shadow;
-	}
-
-	pw->shadow_map = i915_gem_object_pin_map(shadow->obj, I915_MAP_WB);
-	if (IS_ERR(pw->shadow_map)) {
-		err = PTR_ERR(pw->shadow_map);
-		goto err_trampoline;
-	}
-
-	needs_clflush =
-		!(batch->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	pw->batch_map = ERR_PTR(-ENODEV);
-	if (needs_clflush && i915_has_memcpy_from_wc())
-		pw->batch_map = i915_gem_object_pin_map(batch, I915_MAP_WC);
-
-	if (IS_ERR(pw->batch_map)) {
-		err = i915_gem_object_pin_pages(batch);
-		if (err)
-			goto err_unmap_shadow;
-		pw->batch_map = NULL;
-	}
-
-	pw->jump_whitelist =
-		intel_engine_cmd_parser_alloc_jump_whitelist(eb->batch_len,
-							     trampoline);
-	if (IS_ERR(pw->jump_whitelist)) {
-		err = PTR_ERR(pw->jump_whitelist);
-		goto err_unmap_batch;
-	}
-
-	dma_fence_work_init(&pw->base, &eb_parse_ops);
-
-	pw->engine = eb->engine;
-	pw->batch = eb->batch->vma;
-	pw->batch_offset = eb->batch_start_offset;
-	pw->batch_length = eb->batch_len;
-	pw->shadow = shadow;
-	pw->trampoline = trampoline;
-
-	/* Mark active refs early for this worker, in case we get interrupted */
-	err = parser_mark_active(pw, eb->context->timeline);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(pw->batch->resv, 1);
-	if (err)
-		goto err_commit;
-
-	err = dma_resv_reserve_shared(shadow->resv, 1);
-	if (err)
-		goto err_commit;
-
-	/* Wait for all writes (and relocs) into the batch to complete */
-	err = i915_sw_fence_await_reservation(&pw->base.chain,
-					      pw->batch->resv, NULL, false,
-					      0, I915_FENCE_GFP);
-	if (err < 0)
-		goto err_commit;
-
-	/* Keep the batch alive and unwritten as we parse */
-	dma_resv_add_shared_fence(pw->batch->resv, &pw->base.dma);
-
-	/* Force execution to wait for completion of the parser */
-	dma_resv_add_excl_fence(shadow->resv, &pw->base.dma);
-
-	dma_fence_work_commit_imm(&pw->base);
-	return 0;
-
-err_commit:
-	i915_sw_fence_set_error_once(&pw->base.chain, err);
-	dma_fence_work_commit_imm(&pw->base);
-	return err;
-
-err_unmap_batch:
-	if (pw->batch_map)
-		i915_gem_object_unpin_map(batch);
-	else
-		i915_gem_object_unpin_pages(batch);
-err_unmap_shadow:
-	i915_gem_object_unpin_map(shadow->obj);
-err_trampoline:
-	if (trampoline)
-		i915_active_release(&trampoline->active);
-err_shadow:
-	i915_active_release(&shadow->active);
-err_batch:
-	i915_active_release(&eb->batch->vma->active);
-err_free:
-	kfree(pw);
-	return err;
-}
-
 static struct i915_vma *eb_dispatch_secure(struct i915_execbuffer *eb, struct i915_vma *vma)
 {
 	/*
@@ -2672,7 +2463,15 @@ static int eb_parse(struct i915_execbuffer *eb)
 		goto err_trampoline;
 	}
 
-	err = eb_parse_pipeline(eb, shadow, trampoline);
+	err = dma_resv_reserve_shared(shadow->resv, 1);
+	if (err)
+		goto err_trampoline;
+
+	err = intel_engine_cmd_parser(eb->engine,
+				      eb->batch->vma,
+				      eb->batch_start_offset,
+				      eb->batch_len,
+				      shadow, trampoline);
 	if (err)
 		goto err_unpin_batch;
 
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
index 4df505e4c53ae..16162fc2782dc 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_execbuffer.c
@@ -125,6 +125,10 @@ static int igt_gpu_reloc(void *arg)
 	intel_gt_pm_get(&eb.i915->gt);
 
 	for_each_uabi_engine(eb.engine, eb.i915) {
+		if (intel_engine_requires_cmd_parser(eb.engine) ||
+		    intel_engine_using_cmd_parser(eb.engine))
+			continue;
+
 		reloc_cache_init(&eb.reloc_cache, eb.i915);
 		memset(map, POISON_INUSE, 4096);
 
diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 5b4b2bd46e7c8..056a233f443b4 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -1145,19 +1145,41 @@ find_reg(const struct intel_engine_cs *engine, u32 addr)
 static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		       struct drm_i915_gem_object *src_obj,
 		       unsigned long offset, unsigned long length,
-		       void *dst, const void *src)
+		       bool *needs_clflush_after)
 {
-	bool needs_clflush =
-		!(src_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ);
-
-	if (src) {
-		GEM_BUG_ON(!needs_clflush);
-		i915_unaligned_memcpy_from_wc(dst, src + offset, length);
-	} else {
-		struct scatterlist *sg;
+	unsigned int src_needs_clflush;
+	unsigned int dst_needs_clflush;
+	void *dst, *src;
+	int ret;
+
+	ret = i915_gem_object_prepare_write(dst_obj, &dst_needs_clflush);
+	if (ret)
+		return ERR_PTR(ret);
+
+	dst = i915_gem_object_pin_map(dst_obj, I915_MAP_WB);
+	i915_gem_object_finish_access(dst_obj);
+	if (IS_ERR(dst))
+		return dst;
+
+	ret = i915_gem_object_prepare_read(src_obj, &src_needs_clflush);
+	if (ret) {
+		i915_gem_object_unpin_map(dst_obj);
+		return ERR_PTR(ret);
+	}
+
+	src = ERR_PTR(-ENODEV);
+	if (src_needs_clflush && i915_has_memcpy_from_wc()) {
+		src = i915_gem_object_pin_map(src_obj, I915_MAP_WC);
+		if (!IS_ERR(src)) {
+			i915_unaligned_memcpy_from_wc(dst,
+						      src + offset,
+						      length);
+			i915_gem_object_unpin_map(src_obj);
+		}
+	}
+	if (IS_ERR(src)) {
+		unsigned long x, n, remain;
 		void *ptr;
-		unsigned int x, sg_ofs;
-		unsigned long remain;
 
 		/*
 		 * We can avoid clflushing partial cachelines before the write
@@ -1168,40 +1190,34 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
 		 * validate up to the end of the batch.
 		 */
 		remain = length;
-		if (!(dst_obj->cache_coherent & I915_BO_CACHE_COHERENT_FOR_READ))
+		if (dst_needs_clflush & CLFLUSH_BEFORE)
 			remain = round_up(remain,
 					  boot_cpu_data.x86_clflush_size);
 
 		ptr = dst;
 		x = offset_in_page(offset);
-		sg = i915_gem_object_get_sg(src_obj, offset >> PAGE_SHIFT, &sg_ofs, false);
-
-		while (remain) {
-			unsigned long sg_max = sg->length >> PAGE_SHIFT;
-
-			for (; remain && sg_ofs < sg_max; sg_ofs++) {
-				unsigned long len = min(remain, PAGE_SIZE - x);
-				void *map;
-
-				map = kmap_atomic(nth_page(sg_page(sg), sg_ofs));
-				if (needs_clflush)
-					drm_clflush_virt_range(map + x, len);
-				memcpy(ptr, map + x, len);
-				kunmap_atomic(map);
-
-				ptr += len;
-				remain -= len;
-				x = 0;
-			}
-
-			sg_ofs = 0;
-			sg = sg_next(sg);
+		for (n = offset >> PAGE_SHIFT; remain; n++) {
+			int len = min(remain, PAGE_SIZE - x);
+
+			src = kmap_atomic(i915_gem_object_get_page(src_obj, n));
+			if (src_needs_clflush)
+				drm_clflush_virt_range(src + x, len);
+			memcpy(ptr, src + x, len);
+			kunmap_atomic(src);
+
+			ptr += len;
+			remain -= len;
+			x = 0;
 		}
 	}
 
+	i915_gem_object_finish_access(src_obj);
+
 	memset32(dst + length, 0, (dst_obj->base.size - length) / sizeof(u32));
 
 	/* dst_obj is returned with vmap pinned */
+	*needs_clflush_after = dst_needs_clflush & CLFLUSH_AFTER;
+
 	return dst;
 }
 
@@ -1360,6 +1376,9 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	if (target_cmd_index == offset)
 		return 0;
 
+	if (IS_ERR(jump_whitelist))
+		return PTR_ERR(jump_whitelist);
+
 	if (!test_bit(target_cmd_index, jump_whitelist)) {
 		DRM_DEBUG("CMD: BB_START to 0x%llx not a previously executed cmd\n",
 			  jump_target);
@@ -1369,28 +1388,10 @@ static int check_bbstart(u32 *cmd, u32 offset, u32 length,
 	return 0;
 }
 
-/**
- * intel_engine_cmd_parser_alloc_jump_whitelist() - preallocate jump whitelist for intel_engine_cmd_parser()
- * @batch_length: length of the commands in batch_obj
- * @trampoline: Whether jump trampolines are used.
- *
- * Preallocates a jump whitelist for parsing the cmd buffer in intel_engine_cmd_parser().
- * This has to be preallocated, because the command parser runs in signaling context,
- * and may not allocate any memory.
- *
- * Return: NULL or pointer to a jump whitelist, or ERR_PTR() on failure. Use
- * IS_ERR() to check for errors. Must bre freed() with kfree().
- *
- * NULL is a valid value, meaning no allocation was required.
- */
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline)
+static unsigned long *alloc_whitelist(u32 batch_length)
 {
 	unsigned long *jmp;
 
-	if (trampoline)
-		return NULL;
-
 	/*
 	 * We expect batch_length to be less than 256KiB for known users,
 	 * i.e. we need at most an 8KiB bitmap allocation which should be
@@ -1425,21 +1426,21 @@ unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
  * Return: non-zero if the parser finds violations or otherwise fails; -EACCES
  * if the batch appears legal but should use hardware parsing
  */
+
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map)
+			    bool trampoline)
 {
 	u32 *cmd, *batch_end, offset = 0;
 	struct drm_i915_cmd_descriptor default_desc = noop_desc;
 	const struct drm_i915_cmd_descriptor *desc = &default_desc;
+	bool needs_clflush_after = false;
+	unsigned long *jump_whitelist;
 	u64 batch_addr, shadow_addr;
 	int ret = 0;
-	bool trampoline = !jump_whitelist;
 
 	GEM_BUG_ON(!IS_ALIGNED(batch_offset, sizeof(*cmd)));
 	GEM_BUG_ON(!IS_ALIGNED(batch_length, sizeof(*cmd)));
@@ -1447,8 +1448,18 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 				     batch->size));
 	GEM_BUG_ON(!batch_length);
 
-	cmd = copy_batch(shadow->obj, batch->obj, batch_offset, batch_length,
-			 shadow_map, batch_map);
+	cmd = copy_batch(shadow->obj, batch->obj,
+			 batch_offset, batch_length,
+			 &needs_clflush_after);
+	if (IS_ERR(cmd)) {
+		DRM_DEBUG("CMD: Failed to copy batch\n");
+		return PTR_ERR(cmd);
+	}
+
+	jump_whitelist = NULL;
+	if (!trampoline)
+		/* Defer failure until attempted use */
+		jump_whitelist = alloc_whitelist(batch_length);
 
 	shadow_addr = gen8_canonical_addr(shadow->node.start);
 	batch_addr = gen8_canonical_addr(batch->node.start + batch_offset);
@@ -1549,6 +1560,9 @@ int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 
 	i915_gem_object_flush_map(shadow->obj);
 
+	if (!IS_ERR_OR_NULL(jump_whitelist))
+		kfree(jump_whitelist);
+	i915_gem_object_unpin_map(shadow->obj);
 	return ret;
 }
 
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 39b5e019c1a5b..92003970253e5 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1913,17 +1913,12 @@ const char *i915_cache_level_str(struct drm_i915_private *i915, int type);
 int i915_cmd_parser_get_version(struct drm_i915_private *dev_priv);
 int intel_engine_init_cmd_parser(struct intel_engine_cs *engine);
 void intel_engine_cleanup_cmd_parser(struct intel_engine_cs *engine);
-unsigned long *intel_engine_cmd_parser_alloc_jump_whitelist(u32 batch_length,
-							    bool trampoline);
-
 int intel_engine_cmd_parser(struct intel_engine_cs *engine,
 			    struct i915_vma *batch,
 			    unsigned long batch_offset,
 			    unsigned long batch_length,
 			    struct i915_vma *shadow,
-			    unsigned long *jump_whitelist,
-			    void *shadow_map,
-			    const void *batch_map);
+			    bool trampoline);
 #define I915_CMD_PARSER_TRAMPOLINE_SIZE 8
 
 /* intel_device_info.c */
-- 
2.31.1


^ permalink raw reply related	[flat|nested] 24+ messages in thread

end of thread, other threads:[~2021-07-14 19:34 UTC | newest]

Thread overview: 24+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-03 15:40 [PATCH 0/5] drm/i915: Get rid of fence error propagation (v2) Jason Ekstrand
2021-06-03 15:40 ` [Intel-gfx] " Jason Ekstrand
2021-06-03 15:40 ` [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser" Jason Ekstrand
2021-06-03 15:40   ` [Intel-gfx] " Jason Ekstrand
2021-06-03 15:40   ` Jason Ekstrand
2021-06-03 15:40 ` [PATCH 2/5] Revert "drm/i915: Propagate errors on awaiting already signaled fences" Jason Ekstrand
2021-06-03 15:40   ` [Intel-gfx] " Jason Ekstrand
2021-06-03 15:40   ` Jason Ekstrand
2021-06-03 15:40 ` [PATCH 3/5] drm/i915: Remove allow_alloc from i915_gem_object_get_sg* Jason Ekstrand
2021-06-03 15:40   ` [Intel-gfx] " Jason Ekstrand
2021-06-03 15:40 ` [PATCH 4/5] drm/i915: Drop error handling from dma_fence_work Jason Ekstrand
2021-06-03 15:40   ` [Intel-gfx] " Jason Ekstrand
2021-06-03 15:40 ` [PATCH 5/5] Revert "drm/i915: Skip over MI_NOOP when parsing" Jason Ekstrand
2021-06-03 15:40   ` [Intel-gfx] " Jason Ekstrand
2021-06-03 16:32 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Get rid of fence error propagation (rev2) Patchwork
2021-06-03 16:33 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-06-03 16:37 ` [Intel-gfx] ✗ Fi.CI.DOCS: " Patchwork
2021-06-03 17:04 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-06-03 18:09 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2021-07-14 19:34 [PATCH 0/5] drm/i915: Get rid of fence error propagation (v4) Jason Ekstrand
2021-07-14 19:34 ` [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser" Jason Ekstrand
2021-07-14 19:34   ` Jason Ekstrand
2021-07-11  3:53 [PATCH 0/5] drm/i915: Get rid of fence error propagation (v2) Jason Ekstrand
2021-07-11  3:53 ` [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser" Jason Ekstrand
2021-07-11  3:53   ` Jason Ekstrand
2021-06-02 16:41 [PATCH 0/5] drm/i915: Get rid of fence error propagation Jason Ekstrand
2021-06-02 16:41 ` [PATCH 1/5] drm/i915: Revert "drm/i915/gem: Asynchronous cmdparser" Jason Ekstrand

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.