Re: [PATCH v2 3/8] drm/panfrost: Handle job HW submit errors

[Boris Brezillon <boris.brezillon@collabora.com>]

On Mon, 2 Dec 2024 10:21:51 +0100
Boris Brezillon <boris.brezillon@collabora.com> wrote:

> On Thu, 28 Nov 2024 21:06:18 +0000
> Adrián Larumbe <adrian.larumbe@collabora.com> wrote:
> 
> > Avoid waiting for the DRM scheduler job timedout handler, and instead, let
> > the DRM scheduler core signal the error fence immediately when HW job
> > submission fails.
> > 
> > That means we must also decrement the runtime-PM refcnt for the device,
> > because the job will never be enqueued or inflight.
> > 
> > Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
> > ---
> >  drivers/gpu/drm/panfrost/panfrost_job.c | 20 ++++++++++++++++----
> >  1 file changed, 16 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
> > index f640d211cc3a..3f4f0682d69d 100644
> > --- a/drivers/gpu/drm/panfrost/panfrost_job.c
> > +++ b/drivers/gpu/drm/panfrost/panfrost_job.c
> > @@ -195,7 +195,7 @@ panfrost_enqueue_job(struct panfrost_device *pfdev, int slot,
> >  	return 1;
> >  }
> >  
> > -static void panfrost_job_hw_submit(struct panfrost_job *job, int js)
> > +static int panfrost_job_hw_submit(struct panfrost_job *job, int js)
> >  {
> >  	struct panfrost_device *pfdev = job->pfdev;
> >  	unsigned int subslot;
> > @@ -207,10 +207,11 @@ static void panfrost_job_hw_submit(struct panfrost_job *job, int js)
> >  
> >  	ret = pm_runtime_get_sync(pfdev->base.dev);
> >  	if (ret < 0)
> > -		return;
> > +		goto err_hwsubmit;
> >  
> >  	if (WARN_ON(job_read(pfdev, JS_COMMAND_NEXT(js)))) {
> > -		return;
> > +		ret = -EINVAL;
> > +		goto err_hwsubmit;
> >  	}
> >  
> >  	cfg = panfrost_mmu_as_get(pfdev, job->mmu);
> > @@ -261,6 +262,12 @@ static void panfrost_job_hw_submit(struct panfrost_job *job, int js)
> >  			job, js, subslot, jc_head, cfg & 0xf);
> >  	}
> >  	spin_unlock(&pfdev->js->job_lock);
> > +
> > +	return 0;
> > +
> > +err_hwsubmit:
> > +	pm_runtime_put_autosuspend(pfdev->base.dev);
> > +	return ret;
> >  }
> >  
> >  static int panfrost_acquire_object_fences(struct drm_gem_object **bos,
> > @@ -382,6 +389,7 @@ static struct dma_fence *panfrost_job_run(struct drm_sched_job *sched_job)
> >  	struct panfrost_device *pfdev = job->pfdev;
> >  	int slot = panfrost_job_get_slot(job);
> >  	struct dma_fence *fence = NULL;
> > +	int ret;
> >  
> >  	if (unlikely(job->base.s_fence->finished.error))
> >  		return NULL;
> > @@ -400,7 +408,11 @@ static struct dma_fence *panfrost_job_run(struct drm_sched_job *sched_job)
> >  		dma_fence_put(job->done_fence);
> >  	job->done_fence = dma_fence_get(fence);
> >  
> > -	panfrost_job_hw_submit(job, slot);
> > +	ret = panfrost_job_hw_submit(job, slot);
> > +	if (ret) {
> > +		dma_fence_put(job->done_fence);  
> 
> If you call dma_fence_put() here, you need to set job->done_fence to
> NULL, otherwise dma_fence_put() will be called again on an already
> freed fence in panfrost_job_cleanup(). Question is, do we really need
> to call dma_fence_put(job->done_fence) here? Can't we let the job
> destructor take care of that?

My bad, it's the ref you've taken to return to drm_sched you're
releasing here. I'd go for `dma_fence_put(fence)` to make that clear.

With this minor change, the patch is

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>

> 
> > +		return ERR_PTR(ret);
> > +	}
> >  
> >  	return fence;
> >  }  
>