AMD-GFX Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: "Christian König" <ckoenig.leichtzumerken@gmail.com>
To: Alexander.Deucher@amd.com, amd-gfx@lists.freedesktop.org
Cc: vprosyak@amd.com
Subject: [PATCH 2/4] drm/amdgpu: fix calling VM invalidation in amdgpu_hmm_invalidate_gfx
Date: Tue, 12 May 2026 18:27:09 +0200	[thread overview]
Message-ID: <20260512162711.51118-2-christian.koenig@amd.com> (raw)
In-Reply-To: <20260512162711.51118-1-christian.koenig@amd.com>

Otherwise we don't invalidate page tables on next CS.

Signed-off-by: Christian König <christian.koenig@amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c | 1 +
 drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c  | 7 +++++--
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
index eb470e252399..5d72878c8fe9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_hmm.c
@@ -76,6 +76,7 @@ static bool amdgpu_hmm_invalidate_gfx(struct mmu_interval_notifier *mni,
 
 	mmu_interval_set_seq(mni, cur_seq);
 
+	amdgpu_vm_bo_invalidate(bo, false);
 	r = dma_resv_wait_timeout(bo->tbo.base.resv, DMA_RESV_USAGE_BOOKKEEP,
 				  false, MAX_SCHEDULE_TIMEOUT);
 	mutex_unlock(&adev->notifier_lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index 82a1c19350ee..cd82541ad5d7 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -1631,6 +1631,7 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 {
 	struct amdgpu_bo_va *bo_va;
 	struct dma_resv *resv;
+	struct amdgpu_bo *bo;
 	bool clear, unlock;
 	int r;
 
@@ -1650,11 +1651,13 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 	while (!list_empty(&vm->invalidated)) {
 		bo_va = list_first_entry(&vm->invalidated, struct amdgpu_bo_va,
 					 base.vm_status);
-		resv = bo_va->base.bo->tbo.base.resv;
+		bo = bo_va->base.bo;
+		resv = bo->tbo.base.resv;
 		spin_unlock(&vm->status_lock);
 
 		/* Try to reserve the BO to avoid clearing its ptes */
-		if (!adev->debug_vm && dma_resv_trylock(resv)) {
+		if (!adev->debug_vm && !amdgpu_ttm_tt_get_usermm(bo->tbo.ttm) &&
+		    dma_resv_trylock(resv)) {
 			clear = false;
 			unlock = true;
 		/* The caller is already holding the reservation lock */
-- 
2.43.0


  reply	other threads:[~2026-05-12 16:27 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-05-12 16:27 [PATCH 1/4] drm/amdgpu: fix amdgpu_hmm_range_get_pages Christian König
2026-05-12 16:27 ` Christian König [this message]
2026-05-12 16:27 ` [PATCH 3/4] drm/amdgpu: fix waiting for all submissions for userptrs Christian König
2026-05-12 16:27 ` [PATCH 4/4] drm/amdgpu: restart the CS if some parts of the VM are still invalidated Christian König
2026-05-12 18:29   ` Alex Deucher
2026-05-12 18:26 ` [PATCH 1/4] drm/amdgpu: fix amdgpu_hmm_range_get_pages vitaly prosyak

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=20260512162711.51118-2-christian.koenig@amd.com \
    --to=ckoenig.leichtzumerken@gmail.com \
    --cc=Alexander.Deucher@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=christian.koenig@amd.com \
    --cc=vprosyak@amd.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