From: Hans Verkuil <hverkuil@xs4all.nl>
To: Mauro Carvalho Chehab <mchehab@osg.samsung.com>
Cc: linux-media@vger.kernel.org,
"Niklas Söderlund" <niklas.soderlund+renesas@ragnatech.se>,
"Hans Verkuil" <hans.verkuil@cisco.com>
Subject: Re: [PATCHv2 1/2] v4l2-ioctl: simplify code
Date: Wed, 13 Apr 2016 23:51:45 +0200 [thread overview]
Message-ID: <570EBF71.8050506@xs4all.nl> (raw)
In-Reply-To: <20160413165714.587c5e7c@recife.lan>
On 04/13/2016 09:57 PM, Mauro Carvalho Chehab wrote:
> Em Mon, 21 Mar 2016 09:47:59 +0100
> Hans Verkuil <hverkuil@xs4all.nl> escreveu:
>
>> From: Hans Verkuil <hans.verkuil@cisco.com>
>>
>> Instead of a big if at the beginning, just check if g_selection == NULL
>> and call the cropcap op immediately and return the result.
>>
>> No functional changes in this patch.
>
> Hmm... not true. See below.
>
>>
>> Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
>
> You forgot to add Miklas review here.
>
>> ---
>> drivers/media/v4l2-core/v4l2-ioctl.c | 51 ++++++++++++++++++++----------------
>> 1 file changed, 29 insertions(+), 22 deletions(-)
>>
>> diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c
>> index 6bf5a3e..3cf8d3a 100644
>> --- a/drivers/media/v4l2-core/v4l2-ioctl.c
>> +++ b/drivers/media/v4l2-core/v4l2-ioctl.c
>> @@ -2160,33 +2160,40 @@ static int v4l_cropcap(const struct v4l2_ioctl_ops *ops,
>> struct file *file, void *fh, void *arg)
>> {
>> struct v4l2_cropcap *p = arg;
>> + struct v4l2_selection s = { .type = p->type };
>> + int ret;
>>
>> - if (ops->vidioc_g_selection) {
>> - struct v4l2_selection s = { .type = p->type };
>> - int ret;
>> + if (ops->vidioc_g_selection == NULL) {
>> + /*
>> + * The determine_valid_ioctls() call already should ensure
>> + * that ops->vidioc_cropcap != NULL, but just in case...
>> + */
>> + if (ops->vidioc_cropcap)
>> + return ops->vidioc_cropcap(file, fh, p);
>> + return -ENOTTY;
>
> Actually, before this patch, the logic would be doing, instead:
>
> /* setting trivial pixelaspect */
> p->pixelaspect.numerator = 1;
> p->pixelaspect.denominator = 1;
>
> if (ops->vidioc_cropcap)
> return ops->vidioc_cropcap(file, fh, p);
>
> return 0;
>
> With is not the same as what's there. So, there are actually two
> differences:
>
> 1) it will now return -ENOTTY if !ops->vidioc_cropcap instead of 0.
>
> (returning -ENOTTY is probably the best, but this is a fixup change,
> and a functional change, so, should be in a separate patch, if needed)
>
> 2) pixel numerator/denominator is not initialized in the new code.
>
> Ok, maybe all drivers would be setting numerator and denominator to 1
> as default, so the code won't need to do it, but:
>
> a) This is not a trivial assumption by looking only on this patch;
>
> b) It sounds risky to not setup denominator, as, if a driver is not
> initializing it, it could cause a division by zero at userspace.
>
> So, I would prefer to keep initializing numerator/denominator here.
>
> As this is just a cleanup patch, I'm skipping it for now.
I'm confused, you say you're skipping it, yet it is merged. But patch 2/2
(v4l2-ioctl: improve cropcap handling) isn't merged.
I refrain from commenting on the points you raised until I know what's going
on here.
Regards,
Hans
>
> Btw, IMHO, the best here would be to split the logic that it is called
> if ops->vidioc_g_selection is not NULL on a separate function. That
> would make the logic even clearer, and would help to show that it will
> be handling the logic below "setting trivial pixelaspect" the same way.
>
> Regards,
> Mauro
>
>> + }
>>
>> - /* obtaining bounds */
>> - if (V4L2_TYPE_IS_OUTPUT(p->type))
>> - s.target = V4L2_SEL_TGT_COMPOSE_BOUNDS;
>> - else
>> - s.target = V4L2_SEL_TGT_CROP_BOUNDS;
>> + /* obtaining bounds */
>> + if (V4L2_TYPE_IS_OUTPUT(p->type))
>> + s.target = V4L2_SEL_TGT_COMPOSE_BOUNDS;
>> + else
>> + s.target = V4L2_SEL_TGT_CROP_BOUNDS;
>>
>> - ret = ops->vidioc_g_selection(file, fh, &s);
>> - if (ret)
>> - return ret;
>> - p->bounds = s.r;
>> + ret = ops->vidioc_g_selection(file, fh, &s);
>> + if (ret)
>> + return ret;
>> + p->bounds = s.r;
>>
>> - /* obtaining defrect */
>> - if (V4L2_TYPE_IS_OUTPUT(p->type))
>> - s.target = V4L2_SEL_TGT_COMPOSE_DEFAULT;
>> - else
>> - s.target = V4L2_SEL_TGT_CROP_DEFAULT;
>> + /* obtaining defrect */
>> + if (V4L2_TYPE_IS_OUTPUT(p->type))
>> + s.target = V4L2_SEL_TGT_COMPOSE_DEFAULT;
>> + else
>> + s.target = V4L2_SEL_TGT_CROP_DEFAULT;
>>
>> - ret = ops->vidioc_g_selection(file, fh, &s);
>> - if (ret)
>> - return ret;
>> - p->defrect = s.r;
>> - }
>> + ret = ops->vidioc_g_selection(file, fh, &s);
>> + if (ret)
>> + return ret;
>> + p->defrect = s.r;
>>
>> /* setting trivial pixelaspect */
>> p->pixelaspect.numerator = 1;
>
>
next prev parent reply other threads:[~2016-04-13 21:51 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-03-21 8:47 [PATCHv2 0/2] v4l2-ioctl: cropcap improvements Hans Verkuil
2016-03-21 8:47 ` [PATCHv2 1/2] v4l2-ioctl: simplify code Hans Verkuil
2016-04-13 19:57 ` Mauro Carvalho Chehab
2016-04-13 21:51 ` Hans Verkuil [this message]
2016-04-13 22:09 ` Mauro Carvalho Chehab
2016-03-21 8:48 ` [PATCHv2 2/2] v4l2-ioctl: improve cropcap handling Hans Verkuil
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=570EBF71.8050506@xs4all.nl \
--to=hverkuil@xs4all.nl \
--cc=hans.verkuil@cisco.com \
--cc=linux-media@vger.kernel.org \
--cc=mchehab@osg.samsung.com \
--cc=niklas.soderlund+renesas@ragnatech.se \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox