On 7/24/19 3:22 PM, Philipp Zabel wrote: > On Wed, 2019-07-24 at 13:05 +0200, Hans Verkuil wrote: >> If a driver sets a FMT flag in the enum_fmt op, then that will be >> ignored since v4l_fill_fmtdesc() overwrites it again. >> >> v4l_fill_fmtdesc() should OR its flag, not overwrite it. >> >> Signed-off-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx> >> --- >> drivers/media/v4l2-core/v4l2-ioctl.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/drivers/media/v4l2-core/v4l2-ioctl.c b/drivers/media/v4l2-core/v4l2-ioctl.c >> index 80efc581e3f9..911a20f915c5 100644 >> --- a/drivers/media/v4l2-core/v4l2-ioctl.c >> +++ b/drivers/media/v4l2-core/v4l2-ioctl.c >> @@ -1390,7 +1390,7 @@ static void v4l_fill_fmtdesc(struct v4l2_fmtdesc *fmt) >> >> if (descr) >> WARN_ON(strscpy(fmt->description, descr, sz) < 0); >> - fmt->flags = flags; >> + fmt->flags |= flags; >> } >> >> static int v4l_enum_fmt(const struct v4l2_ioctl_ops *ops, > > If the enum_fmt op does not write fmt->flags, will it not contain the > value provided by userspace at this point? I think p->flags must be > cleared in v4l2_enum_fmt() with this change, before the enum_fmt op is > called. All fields after 'type' in struct v4l2_fmtdesc are cleared by the core: search for INFO_FL_CLEAR(v4l2_fmtdesc, type) in v4l2-ioctl.c. So 'flags' is already zeroed when this function is called. Regards, Hans
