Re: [PATCH RFC 2/7] mem2mem_testdev: Use mem-to-mem ioctl and vb2 helpers

[Philipp Zabel <p.zabel@pengutronix.de>]

Hi Sylwester,

Am Sonntag, den 15.09.2013, 22:58 +0200 schrieb Sylwester Nawrocki:
> Hi Philipp,
> 
> On 09/13/2013 03:08 PM, Philipp Zabel wrote:
> > Am Freitag, den 13.09.2013, 14:56 +0200 schrieb Sylwester Nawrocki:
> [...]
> >> @@ -865,6 +793,7 @@ static int queue_init(void *priv, struct vb2_queue *src_vq, struct vb2_queue *ds
> >>   	dst_vq->ops =&m2mtest_qops;
> >>   	dst_vq->mem_ops =&vb2_vmalloc_memops;
> >>   	dst_vq->timestamp_type = V4L2_BUF_FLAG_TIMESTAMP_COPY;
> >> +	dst_vq->lock =&ctx->dev->dev_mutex;
> >>
> >>   	return vb2_queue_init(dst_vq);
> >>   }
> >> @@ -945,6 +874,7 @@ static int m2mtest_open(struct file *file)
> >>   		kfree(ctx);
> >>   		goto open_unlock;
> >>   	}
> >> +	ctx->fh.m2m_ctx = ctx->m2m_ctx;
> >
> > Since you added m2m_ctx to v4l2_fh, why not drop ctx->m2m_ctx altogether
> > and always use ctx->fh.m2m_ctx instead?
> 
> Yes, that might make it a bit cleaner. I guess I wanted to minimize
> the necessary changes.

You are right. Converting the coda driver, I notice that depending on
how many occurences of ctx->m2m_ctx there are, it might be better to do
this in a separate patch.

> I'll amend that in all drivers in this series
> for the next iteration, as they all look very similar. Thanks for the
> suggestion.

regards
Philipp