Received-SPF: pass (google.com: domain of linux-kernel+bounces-32028-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240117031212.1104034-1-nunes.erico@gmail.com>
 <20240117031212.1104034-5-nunes.erico@gmail.com> <20240121095604.2368-1-hdanton@sina.com>
In-Reply-To: <20240121095604.2368-1-hdanton@sina.com>
From: Qiang Yu <yuq825@gmail.com>
Date: Sun, 21 Jan 2024 19:20:14 +0800
Message-ID: <CAKGbVbtww65w+sr7OiccawRkGt-p3MRWaWNU=nkQTaqdyMxqVA@mail.gmail.com>
Subject: Re: [PATCH v1 4/6] drm/lima: handle spurious timeouts due to high irq latency
To: Hillf Danton <hdanton@sina.com>
Cc: Erico Nunes <nunes.erico@gmail.com>, dri-devel@lists.freedesktop.org, 
	lima@lists.freedesktop.org, David Airlie <airlied@gmail.com>, 
	Daniel Vetter <daniel@ffwll.ch>, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sun, Jan 21, 2024 at 5:56=E2=80=AFPM Hillf Danton <hdanton@sina.com> wro=
te:
>
> On Wed, 17 Jan 2024 04:12:10 +0100 Erico Nunes <nunes.erico@gmail.com>
> >
> > @@ -401,9 +399,33 @@ static enum drm_gpu_sched_stat lima_sched_timedout=
_job(struct drm_sched_job *job
> >       struct lima_sched_pipe *pipe =3D to_lima_pipe(job->sched);
> >       struct lima_sched_task *task =3D to_lima_task(job);
> >       struct lima_device *ldev =3D pipe->ldev;
> > +     struct lima_ip *ip =3D pipe->processor[0];
> > +
> > +     /*
> > +      * If the GPU managed to complete this jobs fence, the timeout is
> > +      * spurious. Bail out.
> > +      */
> > +     if (dma_fence_is_signaled(task->done_fence)) {
> > +             DRM_WARN("%s spurious timeout\n", lima_ip_name(ip));
> > +             return DRM_GPU_SCHED_STAT_NOMINAL;
> > +     }
>
> Given 500ms in lima_sched_pipe_init(), no timeout is spurious by define,
> and stop selling bandaid like this because you have options like locating
> the reasons behind timeout.

This chang do look like to aim for 2FPS apps. Maybe 500ms is too short
for week mali4x0 gpus (2FPS apps appear more likely). AMD/NV GPU uses
10s timeout. So increasing the timeout seems to be an equivalent and better
way?

> > +
> > +     /*
> > +      * Lima IRQ handler may take a long time to process an interrupt
> > +      * if there is another IRQ handler hogging the processing.
> > +      * In order to catch such cases and not report spurious Lima job
> > +      * timeouts, synchronize the IRQ handler and re-check the fence
> > +      * status.
> > +      */
> > +     synchronize_irq(ip->irq);
> > +
> > +     if (dma_fence_is_signaled(task->done_fence)) {
> > +             DRM_WARN("%s unexpectedly high interrupt latency\n", lima=
_ip_name(ip));
> > +             return DRM_GPU_SCHED_STAT_NOMINAL;
> > +     }
> >
> >       if (!pipe->error)
> > -             DRM_ERROR("lima job timeout\n");
> > +             DRM_ERROR("%s lima job timeout\n", lima_ip_name(ip));
> >
> >       drm_sched_stop(&pipe->base, &task->base);
> >