[PATCH 3/8] drm/etnaviv: move runtime PM handling to events

Christian Gmeiner christian.gmeiner at gmail.com
Mon Jun 19 13:09:41 UTC 2023


Hi Lucas,

Am Do., 15. Juni 2023 um 11:37 Uhr schrieb Lucas Stach <l.stach at pengutronix.de>:
>
> Hi Christian,
>
> Am Mittwoch, dem 14.06.2023 um 20:41 +0200 schrieb Christian Gmeiner:
> > Hi Lucas
> >
> > >
> > > Conceptually events are the right abstraction to handle the GPU
> > > runtime PM state: as long as any event is pending the GPU can not
> > > be idle. Events are also properly freed and reallocated when the
> > > GPU has been reset after a hang.
> > >
> > > Signed-off-by: Lucas Stach <l.stach at pengutronix.de>
> > > ---
> > >  drivers/gpu/drm/etnaviv/etnaviv_gem.h        |  1 -
> > >  drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c |  3 ---
> > >  drivers/gpu/drm/etnaviv/etnaviv_gpu.c        | 27 ++++++++++++--------
> > >  3 files changed, 16 insertions(+), 15 deletions(-)
> > >
> > > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem.h b/drivers/gpu/drm/etnaviv/etnaviv_gem.h
> > > index baa81cbf701a..a42d260cac2c 100644
> > > --- a/drivers/gpu/drm/etnaviv/etnaviv_gem.h
> > > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem.h
> > > @@ -97,7 +97,6 @@ struct etnaviv_gem_submit {
> > >         struct list_head node; /* GPU active submit list */
> > >         struct etnaviv_cmdbuf cmdbuf;
> > >         struct pid *pid;       /* submitting process */
> > > -       bool runtime_resumed;
> > >         u32 exec_state;
> > >         u32 flags;
> > >         unsigned int nr_pmrs;
> > > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c b/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c
> > > index 45403ea38906..2416c526f9b0 100644
> > > --- a/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c
> > > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gem_submit.c
> > > @@ -362,9 +362,6 @@ static void submit_cleanup(struct kref *kref)
> > >                         container_of(kref, struct etnaviv_gem_submit, refcount);
> > >         unsigned i;
> > >
> > > -       if (submit->runtime_resumed)
> > > -               pm_runtime_put_autosuspend(submit->gpu->dev);
> > > -
> > >         if (submit->cmdbuf.suballoc)
> > >                 etnaviv_cmdbuf_free(&submit->cmdbuf);
> > >
> > > diff --git a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c
> > > index 4e18aa8566c6..54a1249c5bca 100644
> > > --- a/drivers/gpu/drm/etnaviv/etnaviv_gpu.c
> > > +++ b/drivers/gpu/drm/etnaviv/etnaviv_gpu.c
> > > @@ -1139,7 +1139,8 @@ static int event_alloc(struct etnaviv_gpu *gpu, unsigned nr_events,
> > >         unsigned int *events)
> > >  {
> > >         unsigned long timeout = msecs_to_jiffies(10 * 10000);
> > > -       unsigned i, acquired = 0;
> > > +       unsigned i, acquired = 0, rpm_count = 0;
> >
> > rpm is the short form of runtime power management?
> >
> Yes, it's used in this way in multiple places in the kernel. Do you
> think it's clear enough from the code what's going on to keep it that
> way or should I change it to a longer name?
>

Yes it is clear what is going on - even with short variable name :)

Reviewed-by: Christian Gmeiner <cgmeiner at igalia.com>

> Regards,
> Lucas
>
> > > +       int ret;
> > >
> > >         for (i = 0; i < nr_events; i++) {
> > >                 unsigned long ret;
> > > @@ -1148,6 +1149,7 @@ static int event_alloc(struct etnaviv_gpu *gpu, unsigned nr_events,
> > >
> > >                 if (!ret) {
> > >                         dev_err(gpu->dev, "wait_for_completion_timeout failed");
> > > +                       ret = -EBUSY;
> > >                         goto out;
> > >                 }
> > >
> > > @@ -1167,13 +1169,23 @@ static int event_alloc(struct etnaviv_gpu *gpu, unsigned nr_events,
> > >
> > >         spin_unlock(&gpu->event_spinlock);
> > >
> > > +       for (i = 0; i < nr_events; i++) {
> > > +               ret = pm_runtime_resume_and_get(gpu->dev);
> > > +               if (ret)
> > > +                       goto out_rpm;
> > > +               rpm_count++;
> > > +       }
> > > +
> > >         return 0;
> > >
> > > +out_rpm:
> > > +       for (i = 0; i < rpm_count; i++)
> > > +               pm_runtime_put_autosuspend(gpu->dev);
> > >  out:
> > >         for (i = 0; i < acquired; i++)
> > >                 complete(&gpu->event_free);
> > >
> > > -       return -EBUSY;
> > > +       return ret;
> > >  }
> > >
> > >  static void event_free(struct etnaviv_gpu *gpu, unsigned int event)
> > > @@ -1185,6 +1197,8 @@ static void event_free(struct etnaviv_gpu *gpu, unsigned int event)
> > >                 clear_bit(event, gpu->event_bitmap);
> > >                 complete(&gpu->event_free);
> > >         }
> > > +
> > > +       pm_runtime_put_autosuspend(gpu->dev);
> > >  }
> > >
> > >  /*
> > > @@ -1327,15 +1341,6 @@ struct dma_fence *etnaviv_gpu_submit(struct etnaviv_gem_submit *submit)
> > >         unsigned int i, nr_events = 1, event[3];
> > >         int ret;
> > >
> > > -       if (!submit->runtime_resumed) {
> > > -               ret = pm_runtime_get_sync(gpu->dev);
> > > -               if (ret < 0) {
> > > -                       pm_runtime_put_noidle(gpu->dev);
> > > -                       return NULL;
> > > -               }
> > > -               submit->runtime_resumed = true;
> > > -       }
> > > -
> > >         /*
> > >          * if there are performance monitor requests we need to have
> > >          * - a sync point to re-configure gpu and process ETNA_PM_PROCESS_PRE
> > > --
> > > 2.39.2
> > >
> >
> >
>


-- 
greets
--
Christian Gmeiner, MSc

https://christian-gmeiner.info/privacypolicy


More information about the dri-devel mailing list