From: Matthew Brost <matthew.brost@intel.com>
To: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: <intel-xe@lists.freedesktop.org>
Subject: Re: [PATCH] drm/xe: Suppress reset log for destroyed queues
Date: Thu, 2 Apr 2026 14:37:08 -0700 [thread overview]
Message-ID: <ac7hhMY7sZWW3aT7@gsse-cloud1.jf.intel.com> (raw)
In-Reply-To: <20260402213029.3970116-2-daniele.ceraolospurio@intel.com>
On Thu, Apr 02, 2026 at 02:30:30PM -0700, Daniele Ceraolo Spurio wrote:
> When a queue is destroyed while still active on the HW (for example
> because the app owning it is exiting abruptly), the driver tells the GuC
> to preempt it off the HW immediately and to reset it if it doesn't
> preempt. This can cause a reset log to be printed to dmesg, which can
> be confusing to users as resets are commonly tied to errors, while in
> this case the reset is just done to speed up the cleanup.
> Given that a queue is only destroyed once all refs on it have been
> released (i.e., no one cares about it anymore), the log of it being
> reset is not useful and therefore we can simply suppress it to avoid
> confusion.
>
> Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
> Cc: Matthew Brost <matthew.brost@intel.com>
> ---
> drivers/gpu/drm/xe/xe_guc_submit.c | 7 ++++---
> 1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> index 10556156eaad..e6702bd99309 100644
> --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> @@ -2968,9 +2968,10 @@ int xe_guc_exec_queue_reset_handler(struct xe_guc *guc, u32 *msg, u32 len)
> if (unlikely(!q))
> return -EPROTO;
>
> - xe_gt_info(gt, "Engine reset: engine_class=%s, logical_mask: 0x%x, guc_id=%d, state=0x%0x",
> - xe_hw_engine_class_to_str(q->class), q->logical_mask, guc_id,
> - atomic_read(&q->guc->state));
> + if (!exec_queue_destroyed(q))
I think you want killed here—right? Destroyed is tied to the refcount,
whereas killed is tied to either a user closing an exec_queue or the DRM
render FD being closed. There may also be a multi-queue case that needs
slightly different logic. I forget the exact multi-queue teardown flow,
but IIRC it is slightly different. Niranjana would likely know that
offhand; otherwise, I’d have to reverse-engineer those flows again.
Matt
> + xe_gt_info(gt, "Engine reset: engine_class=%s, logical_mask: 0x%x, guc_id=%d, state=0x%0x",
> + xe_hw_engine_class_to_str(q->class), q->logical_mask, guc_id,
> + atomic_read(&q->guc->state));
>
> trace_xe_exec_queue_reset(q);
>
> --
> 2.43.0
>
next prev parent reply other threads:[~2026-04-02 21:37 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-04-02 21:30 [PATCH] drm/xe: Suppress reset log for destroyed queues Daniele Ceraolo Spurio
2026-04-02 21:37 ` Matthew Brost [this message]
2026-04-02 22:51 ` Daniele Ceraolo Spurio
2026-04-03 0:45 ` Matthew Brost
2026-04-03 17:33 ` Daniele Ceraolo Spurio
2026-04-06 20:08 ` Matthew Brost
2026-04-02 21:37 ` ✓ CI.KUnit: success for " Patchwork
2026-04-02 22:14 ` ✓ Xe.CI.BAT: " Patchwork
2026-04-03 11:29 ` ✗ Xe.CI.FULL: failure " 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=ac7hhMY7sZWW3aT7@gsse-cloud1.jf.intel.com \
--to=matthew.brost@intel.com \
--cc=daniele.ceraolospurio@intel.com \
--cc=intel-xe@lists.freedesktop.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox