| Message ID | 20220204121514.2762676-5-alexander.stein@ew.tq-group.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | imx7/imx8mm media / csi patches | expand |
Hi Alexander and Dorota, Thank you for the patch. On Fri, Feb 04, 2022 at 01:15:10PM +0100, Alexander Stein wrote: > From: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > imx_media_find_mbus_format has NULL as a valid return value, > therefore the caller should take it into account. > > Signed-off-by: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com> > --- > drivers/staging/media/imx/imx-media-utils.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/drivers/staging/media/imx/imx-media-utils.c b/drivers/staging/media/imx/imx-media-utils.c > index 32aaa2e81bea..e0a256a08c3b 100644 > --- a/drivers/staging/media/imx/imx-media-utils.c > +++ b/drivers/staging/media/imx/imx-media-utils.c > @@ -544,6 +544,9 @@ static int imx56_media_mbus_fmt_to_pix_fmt(struct v4l2_pix_format *pix, > cc = imx_media_find_mbus_format(code, PIXFMT_SEL_YUV); The code passed to the function comes from the previous line: imx_media_enum_mbus_formats(&code, 0, PIXFMT_SEL_YUV); As far as I can tell, this is guaranteed to return a code that will result in imx_media_find_mbus_format() returning a non-NULL pointer. > } > > + if (!cc) > + return -EINVAL; > + > /* Round up width for minimum burst size */ > width = round_up(mbus->width, 8); >
Hi Laurent, On Sat, 5 Feb 2022 06:07:37 +0200 Laurent Pinchart <laurent.pinchart@ideasonboard.com> wrote: > Hi Alexander and Dorota, > > Thank you for the patch. > > On Fri, Feb 04, 2022 at 01:15:10PM +0100, Alexander Stein wrote: > > From: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > > > imx_media_find_mbus_format has NULL as a valid return value, > > therefore the caller should take it into account. > > > > Signed-off-by: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com> > > --- > > drivers/staging/media/imx/imx-media-utils.c | 3 +++ > > 1 file changed, 3 insertions(+) > > > > diff --git a/drivers/staging/media/imx/imx-media-utils.c b/drivers/staging/media/imx/imx-media-utils.c > > index 32aaa2e81bea..e0a256a08c3b 100644 > > --- a/drivers/staging/media/imx/imx-media-utils.c > > +++ b/drivers/staging/media/imx/imx-media-utils.c > > @@ -544,6 +544,9 @@ static int imx56_media_mbus_fmt_to_pix_fmt(struct v4l2_pix_format *pix, > > cc = imx_media_find_mbus_format(code, PIXFMT_SEL_YUV); > > The code passed to the function comes from the previous line: > > imx_media_enum_mbus_formats(&code, 0, PIXFMT_SEL_YUV); > > As far as I can tell, this is guaranteed to return a code that will > result in imx_media_find_mbus_format() returning a non-NULL pointer. > While I am not well-versed in the implicit code style of the kernel, I decided to leave it in because it makes the code more locally legible. With a check here, even a non-functional one, there's no need to understand the internals of `imx_media_find_mbus_format` that are only implicit. That makes the code less surprising when interested only in the outer function. The other advantage of a check is becoming robust against future changes to `imx_media_find_mbus_format` itself. I don't have a strong preference about keeping or leaving this patch, but if this check was there in the first place, I wouldn't have spent time trying to figure out whether there's a bug here. Cheers, Dorota > > } > > > > + if (!cc) > > + return -EINVAL; > > + > > /* Round up width for minimum burst size */ > > width = round_up(mbus->width, 8); > > >
Hi Laurent and Dorota, Am Samstag, 5. Februar 2022, 08:51:51 CET schrieb Dorota Czaplejewicz: > * PGP Signed by an unknown key > > Hi Laurent, > > On Sat, 5 Feb 2022 06:07:37 +0200 > > Laurent Pinchart <laurent.pinchart@ideasonboard.com> wrote: > > Hi Alexander and Dorota, > > > > Thank you for the patch. > > > > On Fri, Feb 04, 2022 at 01:15:10PM +0100, Alexander Stein wrote: > > > From: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > > > > > imx_media_find_mbus_format has NULL as a valid return value, > > > therefore the caller should take it into account. > > > > > > Signed-off-by: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > > Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com> > > > --- > > > > > > drivers/staging/media/imx/imx-media-utils.c | 3 +++ > > > 1 file changed, 3 insertions(+) > > > > > > diff --git a/drivers/staging/media/imx/imx-media-utils.c > > > b/drivers/staging/media/imx/imx-media-utils.c index > > > 32aaa2e81bea..e0a256a08c3b 100644 > > > --- a/drivers/staging/media/imx/imx-media-utils.c > > > +++ b/drivers/staging/media/imx/imx-media-utils.c > > > @@ -544,6 +544,9 @@ static int imx56_media_mbus_fmt_to_pix_fmt(struct > > > v4l2_pix_format *pix,> > > > > cc = imx_media_find_mbus_format(code, PIXFMT_SEL_YUV); > > > > The code passed to the function comes from the previous line: > > imx_media_enum_mbus_formats(&code, 0, PIXFMT_SEL_YUV); > > > > As far as I can tell, this is guaranteed to return a code that will > > result in imx_media_find_mbus_format() returning a non-NULL pointer. > > While I am not well-versed in the implicit code style of the kernel, I > decided to leave it in because it makes the code more locally legible. With > a check here, even a non-functional one, there's no need to understand the > internals of `imx_media_find_mbus_format` that are only implicit. That > makes the code less surprising when interested only in the outer function. > > The other advantage of a check is becoming robust against future changes to > `imx_media_find_mbus_format` itself. > > I don't have a strong preference about keeping or leaving this patch, but if > this check was there in the first place, I wouldn't have spent time trying > to figure out whether there's a bug here. Laurent, thanks for your feedback. I'm on Dorota's side here. While you are right that the code from imx_media_enum_mbus_formats(&code, 0, PIXFMT_SEL_YUV); is guaranteed to return something. I don't lnek this implicit assumption over 2 function calls. Better be safe than sorry. For that reason code should be initialized as well. Will create a patch for that. Regards, Alexander
Hello, On Mon, Feb 07, 2022 at 10:52:00AM +0100, Alexander Stein wrote: > Am Samstag, 5. Februar 2022, 08:51:51 CET schrieb Dorota Czaplejewicz: > > On Sat, 5 Feb 2022 06:07:37 +0200 Laurent Pinchart wrote: > > > On Fri, Feb 04, 2022 at 01:15:10PM +0100, Alexander Stein wrote: > > > > From: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > > > > > > > imx_media_find_mbus_format has NULL as a valid return value, > > > > therefore the caller should take it into account. > > > > > > > > Signed-off-by: Dorota Czaplejewicz <dorota.czaplejewicz@puri.sm> > > > > Signed-off-by: Alexander Stein <alexander.stein@ew.tq-group.com> > > > > --- > > > > > > > > drivers/staging/media/imx/imx-media-utils.c | 3 +++ > > > > 1 file changed, 3 insertions(+) > > > > > > > > diff --git a/drivers/staging/media/imx/imx-media-utils.c > > > > b/drivers/staging/media/imx/imx-media-utils.c index > > > > 32aaa2e81bea..e0a256a08c3b 100644 > > > > --- a/drivers/staging/media/imx/imx-media-utils.c > > > > +++ b/drivers/staging/media/imx/imx-media-utils.c > > > > @@ -544,6 +544,9 @@ static int imx56_media_mbus_fmt_to_pix_fmt(struct > > > > v4l2_pix_format *pix,> > > > > > cc = imx_media_find_mbus_format(code, PIXFMT_SEL_YUV); > > > > > > The code passed to the function comes from the previous line: > > > imx_media_enum_mbus_formats(&code, 0, PIXFMT_SEL_YUV); > > > > > > As far as I can tell, this is guaranteed to return a code that will > > > result in imx_media_find_mbus_format() returning a non-NULL pointer. > > > > While I am not well-versed in the implicit code style of the kernel, I > > decided to leave it in because it makes the code more locally legible. With > > a check here, even a non-functional one, there's no need to understand the > > internals of `imx_media_find_mbus_format` that are only implicit. That > > makes the code less surprising when interested only in the outer function. > > > > The other advantage of a check is becoming robust against future changes to > > `imx_media_find_mbus_format` itself. > > > > I don't have a strong preference about keeping or leaving this patch, but if > > this check was there in the first place, I wouldn't have spent time trying > > to figure out whether there's a bug here. > > Laurent, thanks for your feedback. > I'm on Dorota's side here. While you are right that the code from > imx_media_enum_mbus_formats(&code, 0, PIXFMT_SEL_YUV); > is guaranteed to return something. I don't lnek this implicit assumption over > 2 function calls. Better be safe than sorry. For that reason code should be > initialized as well. Will create a patch for that. I'm fine with hardening the code even if the issue can't occur at the moment.
diff --git a/drivers/staging/media/imx/imx-media-utils.c b/drivers/staging/media/imx/imx-media-utils.c index 32aaa2e81bea..e0a256a08c3b 100644 --- a/drivers/staging/media/imx/imx-media-utils.c +++ b/drivers/staging/media/imx/imx-media-utils.c @@ -544,6 +544,9 @@ static int imx56_media_mbus_fmt_to_pix_fmt(struct v4l2_pix_format *pix, cc = imx_media_find_mbus_format(code, PIXFMT_SEL_YUV); } + if (!cc) + return -EINVAL; + /* Round up width for minimum burst size */ width = round_up(mbus->width, 8);