[PATCH] drm/i915: get/put runtime PM at i915_semaphore_status

[PATCH] drm/i915: get/put runtime PM at i915_semaphore_status

[Paulo Zanoni]

From: Paulo Zanoni <paulo.r.zanoni@intel.com>

Otherwise we will print some WARNs when we read registers and the
machine is suspended.

Testcase: igt/pm_rpm/debugfs-read
Cc: Ben Widawsky <ben@bwidawsk.net>
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
---
 drivers/gpu/drm/i915/i915_debugfs.c | 2 ++
 1 file changed, 2 insertions(+)

I still dream with the day we will have a system that will run IGT on our
patches/series _before_ they are applied to the tree, and tell us in case some
testcase fails or gives us a WARN.

diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 0c665df..1ff56ec 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -2316,6 +2316,7 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)
 	ret = mutex_lock_interruptible(&dev->struct_mutex);
 	if (ret)
 		return ret;
+	intel_runtime_pm_get(dev_priv);
 
 	if (IS_BROADWELL(dev)) {
 		struct page *page;
@@ -2365,6 +2366,7 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)
 	}
 	seq_putc(m, '\n');
 
+	intel_runtime_pm_put(dev_priv);
 	mutex_unlock(&dev->struct_mutex);
 	return 0;
 }
-- 
2.0.0

Re: [PATCH] drm/i915: get/put runtime PM at i915_semaphore_status

[Ben Widawsky]

On Wed, Jul 09, 2014 at 02:31:57PM -0300, Paulo Zanoni wrote:
> From: Paulo Zanoni <paulo.r.zanoni@intel.com>
> 
> Otherwise we will print some WARNs when we read registers and the
> machine is suspended.

Do all register reads really require this? If so, shouldn't it work
similarly to forcewake? I didn't follow any of the PM mails, so feel
free to ignore if this has been discussed and addressed already.

> 
> Testcase: igt/pm_rpm/debugfs-read
> Cc: Ben Widawsky <ben@bwidawsk.net>
> Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>

Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

> ---
>  drivers/gpu/drm/i915/i915_debugfs.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> I still dream with the day we will have a system that will run IGT on our
> patches/series _before_ they are applied to the tree, and tell us in case some
> testcase fails or gives us a WARN.
> 
> diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> index 0c665df..1ff56ec 100644
> --- a/drivers/gpu/drm/i915/i915_debugfs.c
> +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> @@ -2316,6 +2316,7 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)
>  	ret = mutex_lock_interruptible(&dev->struct_mutex);
>  	if (ret)
>  		return ret;
> +	intel_runtime_pm_get(dev_priv);
>  
>  	if (IS_BROADWELL(dev)) {
>  		struct page *page;
> @@ -2365,6 +2366,7 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)
>  	}
>  	seq_putc(m, '\n');
>  
> +	intel_runtime_pm_put(dev_priv);
>  	mutex_unlock(&dev->struct_mutex);
>  	return 0;
>  }
> -- 
> 2.0.0
> 

-- 
Ben Widawsky, Intel Open Source Technology Center

Re: [PATCH] drm/i915: get/put runtime PM at i915_semaphore_status

[Daniel Vetter]

On Wed, Jul 09, 2014 at 02:02:27PM -0700, Ben Widawsky wrote:
> On Wed, Jul 09, 2014 at 02:31:57PM -0300, Paulo Zanoni wrote:
> > From: Paulo Zanoni <paulo.r.zanoni@intel.com>
> > 
> > Otherwise we will print some WARNs when we read registers and the
> > machine is suspended.
> 
> Do all register reads really require this? If so, shouldn't it work
> similarly to forcewake? I didn't follow any of the PM mails, so feel
> free to ignore if this has been discussed and addressed already.

All need them, but wake-up can potentially sleep. Which means we can't
wake-up from the interrupt handler. Due to that I much prefer explicit
management of the runtime pm reference - it makes it clear that something
is going on and it's not just magic.

We've talked about grabbing a runtime pm ref unconditionally for all
debugfs files but for the sw state files (e.g. gem stuff) it's not
required. And the _get might hang and so could hamper debuggability. In
any case we have nice testcases so can automatically hit any issues like
this one here.
> 
> > 
> > Testcase: igt/pm_rpm/debugfs-read
> > Cc: Ben Widawsky <ben@bwidawsk.net>
> > Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
> 
> Reviewed-by: Ben Widawsky <ben@bwidawsk.net>

Queued for -next, thanks for the patch.
-Daniel

> 
> > ---
> >  drivers/gpu/drm/i915/i915_debugfs.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > I still dream with the day we will have a system that will run IGT on our
> > patches/series _before_ they are applied to the tree, and tell us in case some
> > testcase fails or gives us a WARN.
> > 
> > diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
> > index 0c665df..1ff56ec 100644
> > --- a/drivers/gpu/drm/i915/i915_debugfs.c
> > +++ b/drivers/gpu/drm/i915/i915_debugfs.c
> > @@ -2316,6 +2316,7 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)
> >  	ret = mutex_lock_interruptible(&dev->struct_mutex);
> >  	if (ret)
> >  		return ret;
> > +	intel_runtime_pm_get(dev_priv);
> >  
> >  	if (IS_BROADWELL(dev)) {
> >  		struct page *page;
> > @@ -2365,6 +2366,7 @@ static int i915_semaphore_status(struct seq_file *m, void *unused)
> >  	}
> >  	seq_putc(m, '\n');
> >  
> > +	intel_runtime_pm_put(dev_priv);
> >  	mutex_unlock(&dev->struct_mutex);
> >  	return 0;
> >  }
> > -- 
> > 2.0.0
> > 
> 
> -- 
> Ben Widawsky, Intel Open Source Technology Center
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch