public inbox for drm-ai-reviews@public-inbox.freedesktop.org
 help / color / mirror / Atom feed
From: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
To: Alex Deucher <alexander.deucher@amd.com>,
	Christian König <christian.koenig@amd.com>
Cc: Eric Huang <jinhuieric.huang@amd.com>,
	David Airlie <airlied@gmail.com>, Simona Vetter <simona@ffwll.ch>,
	amd-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org,
	stable@vger.kernel.org,
	Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Subject: [PATCH v2 2/2] drm/amdgpu: use spin_lock_irqsave for PASID IDR lock
Date: Mon, 30 Mar 2026 10:30:25 +0500	[thread overview]
Message-ID: <20260330053025.19203-3-mikhail.v.gavrilov@gmail.com> (raw)
In-Reply-To: <20260330053025.19203-1-mikhail.v.gavrilov@gmail.com>

amdgpu_pasid_free() can be called from hardirq context via the fence
signal path:

  sdma_v6_0_process_trap_irq
   -> amdgpu_fence_process
    -> dma_fence_signal
     -> drm_sched_job_done
      -> dma_fence_signal
       -> amdgpu_pasid_free_cb
        -> amdgpu_pasid_free
         -> spin_lock(&amdgpu_pasid_idr_lock)  <- hardirq context

But the lock was originally taken with plain spin_lock() in process
context (amdgpu_pasid_alloc), creating an inconsistent
{HARDIRQ-ON-W} -> {IN-HARDIRQ-W} lock state that can deadlock if an
interrupt arrives while the lock is held on the same CPU.

Use spin_lock_irqsave/spin_unlock_irqrestore for all call sites of
amdgpu_pasid_idr_lock to prevent the deadlock.

This patch applies on top of "drm/amdgpu: fix sleeping allocation
under spinlock in PASID IDR".

Fixes: 8f1de51f49be ("drm/amdgpu: prevent immediate PASID reuse case")
Cc: stable@vger.kernel.org
Signed-off-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
index 515775eab2ef..762ceb3c708a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c
@@ -62,16 +62,17 @@ struct amdgpu_pasid_cb {
  */
 int amdgpu_pasid_alloc(unsigned int bits)
 {
+	unsigned long flags;
 	int pasid;
 
 	if (bits == 0)
 		return -EINVAL;
 
 	idr_preload(GFP_KERNEL);
-	spin_lock(&amdgpu_pasid_idr_lock);
+	spin_lock_irqsave(&amdgpu_pasid_idr_lock, flags);
 	pasid = idr_alloc_cyclic(&amdgpu_pasid_idr, NULL, 1,
 				 1U << bits, GFP_NOWAIT);
-	spin_unlock(&amdgpu_pasid_idr_lock);
+	spin_unlock_irqrestore(&amdgpu_pasid_idr_lock, flags);
 	idr_preload_end();
 
 	if (pasid >= 0)
@@ -86,11 +87,12 @@ int amdgpu_pasid_alloc(unsigned int bits)
  */
 void amdgpu_pasid_free(u32 pasid)
 {
+	unsigned long flags;
 	trace_amdgpu_pasid_freed(pasid);
 
-	spin_lock(&amdgpu_pasid_idr_lock);
+	spin_lock_irqsave(&amdgpu_pasid_idr_lock, flags);
 	idr_remove(&amdgpu_pasid_idr, pasid);
-	spin_unlock(&amdgpu_pasid_idr_lock);
+	spin_unlock_irqrestore(&amdgpu_pasid_idr_lock, flags);
 }
 
 static void amdgpu_pasid_free_cb(struct dma_fence *fence,
@@ -633,7 +635,9 @@ void amdgpu_vmid_mgr_fini(struct amdgpu_device *adev)
  */
 void amdgpu_pasid_mgr_cleanup(void)
 {
-	spin_lock(&amdgpu_pasid_idr_lock);
+	unsigned long flags;
+
+	spin_lock_irqsave(&amdgpu_pasid_idr_lock, flags);
 	idr_destroy(&amdgpu_pasid_idr);
-	spin_unlock(&amdgpu_pasid_idr_lock);
+	spin_unlock_irqrestore(&amdgpu_pasid_idr_lock, flags);
 }
-- 
2.53.0


  parent reply	other threads:[~2026-03-30  5:30 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-30  5:30 [PATCH v2 0/2] drm/amdgpu: fix locking issues in PASID IDR management Mikhail Gavrilov
2026-03-30  5:30 ` [PATCH v2 1/2] drm/amdgpu: fix sleeping allocation under spinlock in PASID IDR Mikhail Gavrilov
2026-03-31  7:36   ` Claude review: " Claude Code Review Bot
2026-03-30  5:30 ` Mikhail Gavrilov [this message]
2026-03-31  7:36   ` Claude review: drm/amdgpu: use spin_lock_irqsave for PASID IDR lock Claude Code Review Bot
2026-03-31  7:36 ` Claude review: drm/amdgpu: fix locking issues in PASID IDR management Claude Code Review Bot

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=20260330053025.19203-3-mikhail.v.gavrilov@gmail.com \
    --to=mikhail.v.gavrilov@gmail.com \
    --cc=airlied@gmail.com \
    --cc=alexander.deucher@amd.com \
    --cc=amd-gfx@lists.freedesktop.org \
    --cc=christian.koenig@amd.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=jinhuieric.huang@amd.com \
    --cc=simona@ffwll.ch \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

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

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