From: Matthew Auld <matthew.auld@intel.com>
To: intel-xe@lists.freedesktop.org
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Subject: [Intel-xe] [PATCH v5 4/7] drm/xe: keep pulling mem_access_get further back
Date: Wed, 17 May 2023 16:22:41 +0100 [thread overview]
Message-ID: <20230517152244.348171-4-matthew.auld@intel.com> (raw)
In-Reply-To: <20230517152244.348171-1-matthew.auld@intel.com>
Lockdep is unhappy about ggtt->lock -> mem_access.lock, where it seems
to think this can somehow get inverted. The ggtt->lock looks like a
potentially sensitive driver lock, so likely a sensible move to never
call the runtime_pm routines while holding it. Actually it looks like
d3cold wants to grab this, so perhaps this can indeed deadlock.
Signed-off-by: Matthew Auld <matthew.auld@intel.com>
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
---
drivers/gpu/drm/xe/display/xe_fb_pin.c | 12 +++++++-----
drivers/gpu/drm/xe/xe_ggtt.c | 6 ++++++
drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c | 2 --
3 files changed, 13 insertions(+), 7 deletions(-)
diff --git a/drivers/gpu/drm/xe/display/xe_fb_pin.c b/drivers/gpu/drm/xe/display/xe_fb_pin.c
index ed691d28b34d..a2ed04ca2421 100644
--- a/drivers/gpu/drm/xe/display/xe_fb_pin.c
+++ b/drivers/gpu/drm/xe/display/xe_fb_pin.c
@@ -130,9 +130,10 @@ static int __xe_pin_fb_vma_ggtt(struct intel_framebuffer *fb,
/* TODO: Consider sharing framebuffer mapping?
* embed i915_vma inside intel_framebuffer
*/
+ xe_device_mem_access_get(ggtt->gt->xe);
ret = mutex_lock_interruptible(&ggtt->lock);
if (ret)
- return ret;
+ goto out;
align = XE_PAGE_SIZE;
if (xe_bo_is_vram(bo) && ggtt->flags & XE_GGTT_FLAGS_64K)
@@ -146,7 +147,7 @@ static int __xe_pin_fb_vma_ggtt(struct intel_framebuffer *fb,
ret = xe_ggtt_insert_special_node_locked(ggtt, &vma->node, size,
align, 0);
if (ret)
- goto out;
+ goto out_unlock;
for (x = 0; x < size; x += XE_PAGE_SIZE)
xe_ggtt_set_pte(ggtt, vma->node.start + x, xe_ggtt_pte_encode(bo, x));
@@ -160,7 +161,7 @@ static int __xe_pin_fb_vma_ggtt(struct intel_framebuffer *fb,
ret = xe_ggtt_insert_special_node_locked(ggtt, &vma->node, size,
align, 0);
if (ret)
- goto out;
+ goto out_unlock;
ggtt_ofs = vma->node.start;
@@ -174,9 +175,10 @@ static int __xe_pin_fb_vma_ggtt(struct intel_framebuffer *fb,
}
xe_ggtt_invalidate(to_gt(xe));
-
-out:
+out_unlock:
mutex_unlock(&ggtt->lock);
+out:
+ xe_device_mem_access_put(ggtt->gt->xe);
return ret;
}
diff --git a/drivers/gpu/drm/xe/xe_ggtt.c b/drivers/gpu/drm/xe/xe_ggtt.c
index 546240261e0a..f986e8218820 100644
--- a/drivers/gpu/drm/xe/xe_ggtt.c
+++ b/drivers/gpu/drm/xe/xe_ggtt.c
@@ -142,12 +142,14 @@ static void xe_ggtt_initial_clear(struct xe_ggtt *ggtt)
u64 start, end;
/* Display may have allocated inside ggtt, so be careful with clearing here */
+ xe_device_mem_access_get(ggtt->gt->xe);
mutex_lock(&ggtt->lock);
drm_mm_for_each_hole(hole, &ggtt->mm, start, end)
xe_ggtt_clear(ggtt, start, end - start);
xe_ggtt_invalidate(ggtt->gt);
mutex_unlock(&ggtt->lock);
+ xe_device_mem_access_put(ggtt->gt->xe);
}
int xe_ggtt_init(struct xe_gt *gt, struct xe_ggtt *ggtt)
@@ -288,12 +290,14 @@ static int __xe_ggtt_insert_bo_at(struct xe_ggtt *ggtt, struct xe_bo *bo,
if (err)
return err;
+ xe_device_mem_access_get(ggtt->gt->xe);
mutex_lock(&ggtt->lock);
err = drm_mm_insert_node_in_range(&ggtt->mm, &bo->ggtt_node, bo->size,
alignment, 0, start, end, 0);
if (!err)
xe_ggtt_map_bo(ggtt, bo);
mutex_unlock(&ggtt->lock);
+ xe_device_mem_access_put(ggtt->gt->xe);
return err;
}
@@ -311,6 +315,7 @@ int xe_ggtt_insert_bo(struct xe_ggtt *ggtt, struct xe_bo *bo)
void xe_ggtt_remove_node(struct xe_ggtt *ggtt, struct drm_mm_node *node)
{
+ xe_device_mem_access_get(ggtt->gt->xe);
mutex_lock(&ggtt->lock);
xe_ggtt_clear(ggtt, node->start, node->size);
@@ -320,6 +325,7 @@ void xe_ggtt_remove_node(struct xe_ggtt *ggtt, struct drm_mm_node *node)
xe_ggtt_invalidate(ggtt->gt);
mutex_unlock(&ggtt->lock);
+ xe_device_mem_access_put(ggtt->gt->xe);
}
void xe_ggtt_remove_bo(struct xe_ggtt *ggtt, struct xe_bo *bo)
diff --git a/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c b/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c
index 20f8f0aae6b4..1961e9bc79f1 100644
--- a/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c
+++ b/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c
@@ -114,7 +114,6 @@ static int send_tlb_invalidation(struct xe_guc *guc,
* need to be updated.
*/
- xe_device_mem_access_get(gt->xe);
mutex_lock(&guc->ct.lock);
seqno = gt->tlb_invalidation.seqno;
if (fence) {
@@ -143,7 +142,6 @@ static int send_tlb_invalidation(struct xe_guc *guc,
if (ret < 0 && fence)
invalidation_fence_signal(fence);
mutex_unlock(&guc->ct.lock);
- xe_device_mem_access_put(gt->xe);
return ret;
}
--
2.40.1
next prev parent reply other threads:[~2023-05-17 15:23 UTC|newest]
Thread overview: 31+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-05-17 15:22 [Intel-xe] [PATCH v5 1/7] drm: fix drmm_mutex_init() Matthew Auld
2023-05-17 15:22 ` [Intel-xe] [PATCH v5 2/7] Revert "drm/xe: Use atomic instead of mutex for xe_device_mem_access_ongoing" Matthew Auld
2023-05-17 15:22 ` [Intel-xe] [PATCH v5 3/7] drm/xe: don't allocate under ct->lock Matthew Auld
2023-05-17 15:50 ` Rodrigo Vivi
2023-05-17 15:22 ` Matthew Auld [this message]
2023-05-17 15:53 ` [Intel-xe] [PATCH v5 4/7] drm/xe: keep pulling mem_access_get further back Rodrigo Vivi
2023-05-17 15:22 ` [Intel-xe] [PATCH v5 5/7] drm/xe/ggtt: prime ggtt->lock against FS_RECLAIM Matthew Auld
2023-05-17 15:48 ` Rodrigo Vivi
2023-05-17 16:21 ` Matthew Auld
2023-05-17 15:22 ` [Intel-xe] [PATCH v5 6/7] drm/xe: fix xe_device_mem_access_get() races Matthew Auld
2023-05-19 20:25 ` Rodrigo Vivi
2023-05-22 9:49 ` Matthew Auld
2023-05-17 15:22 ` [Intel-xe] [PATCH v5 7/7] drm/xe: Use atomic for mem_access.ref Matthew Auld
2023-05-17 15:25 ` [Intel-xe] ✓ CI.Patch_applied: success for series starting with [v5,1/7] drm: fix drmm_mutex_init() Patchwork
2023-05-17 15:27 ` [Intel-xe] ✓ CI.KUnit: " Patchwork
2023-05-17 15:31 ` [Intel-xe] ✓ CI.Build: " Patchwork
2023-05-17 15:49 ` [Intel-xe] ○ CI.BAT: info " Patchwork
2023-05-17 16:05 ` [Intel-xe] [PATCH v5 1/7] " Stanislaw Gruszka
2023-05-17 16:29 ` Matthew Auld
2023-05-17 17:03 ` Thomas Hellström
2023-05-17 17:47 ` Thomas Zimmermann
2023-05-17 16:21 ` Thomas Zimmermann
2023-05-17 17:04 ` Matthew Auld
2023-05-17 17:43 ` Thomas Zimmermann
2023-05-18 9:51 ` [Intel-xe] ✓ CI.Patch_applied: success for series starting with [v5,1/7] drm: fix drmm_mutex_init() (rev2) Patchwork
2023-05-18 9:52 ` [Intel-xe] ✓ CI.KUnit: " Patchwork
2023-05-18 9:56 ` [Intel-xe] ✓ CI.Build: " Patchwork
2023-05-18 10:16 ` [Intel-xe] ○ CI.BAT: info " Patchwork
2023-05-19 15:49 ` [Intel-xe] ✓ CI.Patch_applied: success for series starting with [v5,1/7] drm: fix drmm_mutex_init() (rev3) Patchwork
2023-05-19 15:50 ` [Intel-xe] ✓ CI.KUnit: " Patchwork
2023-05-19 15:54 ` [Intel-xe] ✓ CI.Build: " Patchwork
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230517152244.348171-4-matthew.auld@intel.com \
--to=matthew.auld@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=rodrigo.vivi@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox