Hi,
Thanks for the review,
On Thu, 2019-09-26 at 08:32 -0600, Shuah Khan wrote:
> On 9/19/19 2:32 PM, Dafna Hirschfeld wrote:
> > Userspace can disable links and create pipelines that
> > do not start with a source entity. Trying to stream
> > from such a pipeline should fail with -EPIPE
> > currently this is not handled and cause kernel crash.
> >
>
> Minor: Can you make these 75 long. Makes it easier to read.
>
> > Reproducing the crash:
> > media-ctl -d0 -l "5:1->21:0[0]" -v
> > v4l2-ctl -z platform:vimc -d "RGB/YUV Capture" -v width=1920,height=1440
> > v4l2-ctl --stream-mmap --stream-count=100 -d /dev/video2
>
> I really would like to see the panic message so it can checked during
> testing.
>
> If you are fixing a panic, please include the panic info. in the future.
>
> > Signed-off-by: Dafna Hirschfeld <[email protected]>
> > ---
> > drivers/media/platform/vimc/vimc-streamer.c | 39 +++++++++++++++------
> > 1 file changed, 28 insertions(+), 11 deletions(-)
> >
> > diff --git a/drivers/media/platform/vimc/vimc-streamer.c
> > b/drivers/media/platform/vimc/vimc-streamer.c
> > index faa2879c25df..d0a9f8a0f26a 100644
> > --- a/drivers/media/platform/vimc/vimc-streamer.c
> > +++ b/drivers/media/platform/vimc/vimc-streamer.c
> > @@ -12,6 +12,19 @@
> >
> > #include "vimc-streamer.h"
> >
> > +/**
> > + * Check if the entity has only source pads
> > + */
> > +static bool vimc_is_source(struct media_entity *ent)
> > +{
> > + int i;
> > +
> > + for (i = 0; i < ent->num_pads; i++)
> > + if (ent->pads[i].flags & MEDIA_PAD_FL_SINK)
> > + return false;
> > + return true;
> > +}
> > +
>
> Why not make this a common routine and add it to vimc-common.c?
>
> > /**
> > * vimc_get_source_entity - get the entity connected with the first sink
> > pad
> > *
> > @@ -82,14 +95,12 @@ static int vimc_streamer_pipeline_init(struct
> > vimc_stream *stream,
> > struct media_entity *entity;
> > struct video_device *vdev;
> > struct v4l2_subdev *sd;
> > - int ret = 0;
> > + int ret = -EINVAL;
> >
> > stream->pipe_size = 0;
> > while (stream->pipe_size < VIMC_STREAMER_PIPELINE_MAX_SIZE) {
> > - if (!ved) {
> > - vimc_streamer_pipeline_terminate(stream);
> > - return -EINVAL;
> > - }
> > + if (!ved)
> > + break;
> > stream->ved_pipeline[stream->pipe_size++] = ved;
> >
> > if (is_media_entity_v4l2_subdev(ved->ent)) {
> > @@ -98,15 +109,22 @@ static int vimc_streamer_pipeline_init(struct
> > vimc_stream *stream,
> > if (ret && ret != -ENOIOCTLCMD) {
> > pr_err("subdev_call error %s\n",
> > ved->ent->name);
>
> While you are at it, can you make this a dev_err() instead? I think we
> have access to dev here.
Actually, there is no access to dev here. It can be sent to the function as
an argument from the vimc-capture.c code, but maybe a better solution is to
move the dev
pointer of every vimc entity to the common `vimc_ent_dev ved` field, since all
entities have a pointer to it.
Thanks,
Dafna
>
> > - vimc_streamer_pipeline_terminate(stream);
> > - return ret;
> > + break;
> > }
> > }
> >
> > entity = vimc_get_source_entity(ved->ent);
> > - /* Check if the end of the pipeline was reached*/
> > - if (!entity)
> > + /* Check if the end of the pipeline was reached */
> > + if (!entity) {
> > + /* the first entity of the pipe should be source only */
> > + if (!vimc_is_source(ved->ent)) {
> > + pr_err("first entity in the pipe '%s' is not a
> > source\n",
> > + ved->ent->name);
>
> Same commnet about dev_err() here.
>
> > + ret = -EPIPE;
> > + break;
> > + }
> > return 0;
> > + }
> >
> > /* Get the next device in the pipeline */
> > if (is_media_entity_v4l2_subdev(entity)) {
> > @@ -119,9 +137,8 @@ static int vimc_streamer_pipeline_init(struct
> > vimc_stream *stream,
> > ved = video_get_drvdata(vdev);
> > }
> > }
> > -
> > vimc_streamer_pipeline_terminate(stream);
> > - return -EINVAL;
> > + return ret;
> > }
> >
> > /**
> >
>
> thanks,
> -- Shuah
