All of lore.kernel.org
 help / color / mirror / Atom feed
From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Subject: [Intel-gfx] [PATCH] drm/i915: Neuter virtual rq->engine on retire
Date: Tue, 19 May 2020 15:51:31 +0100	[thread overview]
Message-ID: <20200519145131.2893-1-chris@chris-wilson.co.uk> (raw)

We do not hold a reference to rq->engine, and so if it is a virtual
engine it may have already been freed by the time we free the request.
The last reference we hold on the virtual engine is via rq->context,
and that is released on request retirement. So if we find ourselves
retiring a virtual request, redirect it to a real sibling.

Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/1906
Fixes: 43acd6516ca9 ("drm/i915: Keep a per-engine request pool")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
 drivers/gpu/drm/i915/i915_request.c | 17 +++++++++++++++++
 1 file changed, 17 insertions(+)

diff --git a/drivers/gpu/drm/i915/i915_request.c b/drivers/gpu/drm/i915/i915_request.c
index 31ef683d27b4..a816218cc693 100644
--- a/drivers/gpu/drm/i915/i915_request.c
+++ b/drivers/gpu/drm/i915/i915_request.c
@@ -242,9 +242,26 @@ static void remove_from_engine(struct i915_request *rq)
 		spin_lock(&engine->active.lock);
 		locked = engine;
 	}
+
 	list_del_init(&rq->sched.link);
 	clear_bit(I915_FENCE_FLAG_PQUEUE, &rq->fence.flags);
 	clear_bit(I915_FENCE_FLAG_HOLD, &rq->fence.flags);
+
+	/*
+	 * During i915_fence_release we stash one request on the
+	 * rq->engine for use as an emergency reserve. However, we
+	 * neither want to keep a request on a virtual engine, nor do
+	 * we hold a reference to a virtual engine at that point. So
+	 * if rq->engine is virtual, replace it with a real one. Which
+	 * one is immaterial at this point as the request has been
+	 * retired, and if it was a virtual engine will not have any
+	 * signaling or other related paraphernalia.
+	 *
+	 * However, it would be nice if we didn't have to...
+	 */
+	if (intel_engine_is_virtual(rq->engine))
+		rq->engine = intel_virtual_engine_get_sibling(rq->engine, 0);
+
 	spin_unlock_irq(&locked->active.lock);
 }
 
-- 
2.20.1

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

             reply	other threads:[~2020-05-19 14:51 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-19 14:51 Chris Wilson [this message]
2020-05-19 17:00 ` [Intel-gfx] [PATCH] drm/i915: Neuter virtual rq->engine on retire Chris Wilson
2020-05-19 17:21   ` Chris Wilson
2020-05-19 20:54 ` [Intel-gfx] ✓ Fi.CI.BAT: success for " Patchwork
2020-05-20  7:52 ` [Intel-gfx] ✓ Fi.CI.IGT: " 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=20200519145131.2893-1-chris@chris-wilson.co.uk \
    --to=chris@chris-wilson.co.uk \
    --cc=intel-gfx@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 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.