linux-kernel - Re: [PATCH] media: Media Controller fix to not let stream

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:	Wed, 3 Feb 2016 01:09:08 +0200
From:	Sakari Ailus <sakari.ailus@....fi>
To:	Shuah Khan <shuahkh@....samsung.com>
Cc:	mchehab@....samsung.com, linux-media@...r.kernel.org,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH] media: Media Controller fix to not let stream_count go
 negative

Hi Shuah,

On Tue, Feb 02, 2016 at 04:04:50PM -0700, Shuah Khan wrote:
> On 02/02/2016 03:53 PM, Sakari Ailus wrote:
> > Hi Shuah,
> > 
> > On Sat, Jan 30, 2016 at 01:10:52PM -0700, Shuah Khan wrote:
> >> Change media_entity_pipeline_stop() to not decrement
> >> stream_count of an inactive media pipeline. Doing so,
> >> results in preventing starting the pipeline.
> >>
> >> Signed-off-by: Shuah Khan <shuahkh@....samsung.com>
> >> ---
> >>  drivers/media/media-entity.c | 18 ++++++++++++------
> >>  1 file changed, 12 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/drivers/media/media-entity.c b/drivers/media/media-entity.c
> >> index e89d85a..f2e4360 100644
> >> --- a/drivers/media/media-entity.c
> >> +++ b/drivers/media/media-entity.c
> >> @@ -452,9 +452,12 @@ error:
> >>  	media_entity_graph_walk_start(graph, entity_err);
> >>  
> >>  	while ((entity_err = media_entity_graph_walk_next(graph))) {
> >> -		entity_err->stream_count--;
> >> -		if (entity_err->stream_count == 0)
> >> -			entity_err->pipe = NULL;
> >> +		/* don't let the stream_count go negative */
> >> +		if (entity->stream_count > 0) {
> >> +			entity_err->stream_count--;
> >> +			if (entity_err->stream_count == 0)
> >> +				entity_err->pipe = NULL;
> >> +		}
> >>  
> >>  		/*
> >>  		 * We haven't increased stream_count further than this
> >> @@ -486,9 +489,12 @@ void media_entity_pipeline_stop(struct media_entity *entity)
> >>  	media_entity_graph_walk_start(graph, entity);
> >>  
> >>  	while ((entity = media_entity_graph_walk_next(graph))) {
> >> -		entity->stream_count--;
> >> -		if (entity->stream_count == 0)
> >> -			entity->pipe = NULL;
> >> +		/* don't let the stream_count go negative */
> >> +		if (entity->stream_count > 0) {
> >> +			entity->stream_count--;
> >> +			if (entity->stream_count == 0)
> >> +				entity->pipe = NULL;
> >> +		}
> >>  	}
> >>  
> >>  	if (!--pipe->streaming_count)
> > 
> > Have you seen issues with a certain driver, for instance?
> > 
> > In the original design the streaming count is really a count --- streaming
> > starts when count becomes non-zero, and stops when it reaches zero again.
> > 
> > The calls to media_entity_pipeline_start() and media_entity_pipeline_stop()
> > should thus always be balanced. I'm fine with the patch, but the framework
> > should shout loud when the count would be decremented when it's zero.
> > 
> > That was some four or more years ago. I have to say I really haven't been
> > able to see good reasons for making this a count --- rather what's needed is
> > to mark the entity as busy so that its link configuration isn't touched. The
> > request API is a completely different matter then.
> > 
> > Such change would require more discussion IMHO.
> > 
> 
> Yes. I found problems with au0828 and ALSA
> media controller use-case. It got into a state
> where pipeline was essentially locked in a state.
> It took some work to debug and find that the
> stream_count was negative.
> 
> Granted the start and stop should be balanced, however,
> the media_entity_pipeline_stop() still needs to protect
> the stream_count from going negative.

Agreed; I'm fine with the patch, except I think it should complain when that
happens rather than silently ignoring it.

WARN_ON(1) perhaps? Or how about dev_err() loudly complaining about a driver
bug?

-- 
Regards,

Sakari Ailus
e-mail: sakari.ailus@....fi	XMPP: sailus@...iisi.org.uk