Hi,
On 06/18/2014 01:43 PM, Hans de Goede wrote:
> Hi,
>
> On 06/16/2014 05:00 PM, Antonio Ospite wrote:
>> Fix a regression introduced in commit
>> efc29f1764a30808ebf7b3e1d9bfa27b909bf641 (libv4lconvert: Reject too
>> short source buffer before accessing it).
>>
>> The old code:
>>
>> case V4L2_PIX_FMT_Y10BPACK:
>> ...
>> if (result == 0 && src_size < (width * height * 10 / 8)) {
>> V4LCONVERT_ERR("short y10b data frame\n");
>> errno = EPIPE;
>> result = -1;
>> }
>> ...
>>
>> meant to say "If the conversion was *successful* _but_ the frame size
>> was invalid, then take the error path", but in
>> efc29f1764a30808ebf7b3e1d9bfa27b909bf641 this (maybe weird) logic was
>> misunderstood and v4lconvert_convert_pixfmt() was made to return an
>> error even in the case of a successful conversion from Y10B.
>>
>> Fix the check, and now print only the message letting the errno and the
>> result from the conversion routines to be propagated to the caller.
>>
>> Signed-off-by: Antonio Ospite <ao2@xxxxxx>
>> Cc: Gregor Jasny <gjasny@xxxxxxxxxxxxxx>
>
> Thanks for the patch, but: ...
>
>> ---
>> lib/libv4lconvert/libv4lconvert.c | 5 +----
>> 1 file changed, 1 insertion(+), 4 deletions(-)
>>
>> diff --git a/lib/libv4lconvert/libv4lconvert.c b/lib/libv4lconvert/libv4lconvert.c
>> index c49d30d..50d6906 100644
>> --- a/lib/libv4lconvert/libv4lconvert.c
>> +++ b/lib/libv4lconvert/libv4lconvert.c
>> @@ -1052,11 +1052,8 @@ static int v4lconvert_convert_pixfmt(struct v4lconvert_data *data,
>> width, height);
>> break;
>> }
>> - if (result == 0) {
>> + if (result != 0)
>> V4LCONVERT_ERR("y10b conversion failed\n");
>> - errno = EPIPE;
>> - result = -1;
>> - }
>> break;
>>
>> case V4L2_PIX_FMT_RGB565:
>
> Why print a message here at all in the != 0 case? In the old code before commit
> efc29f1764 you did not print an error when v4lconvert_y10b_to_... failed, so
> I assume that that already does a V4LCONVERT_ERR in that case. So why do it a
> second time with a less precise error message here?
>
> So I believe that the proper fix would be to just remove the entire block instead
> of flipping the test and keeping the V4LCONVERT_ERR. Please send a new version
> with this fixed, then I'll merge it asap.
Scrap that, I decided I might just as well fix this bit myself, so I've just
pushed an updated patch completely removing the second check from the
V4L2_PIX_FMT_Y10BPACK case.
Regards,
Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
