On Wed, Oct 11, 2023 at 11:06:15PM +0300, Laurent Pinchart wrote: > On Wed, Oct 11, 2023 at 08:01:50PM +0000, Sakari Ailus wrote: > > On Wed, Oct 11, 2023 at 10:48:33PM +0300, Laurent Pinchart wrote: > > > On Wed, Oct 11, 2023 at 07:44:59PM +0000, Sakari Ailus wrote: > > > > On Fri, Oct 06, 2023 at 10:46:54AM +0300, Alexander Shiyan wrote: > > > > > Let's add a check for src_sd before using it. > > > > > The link may not be set, in which case the call to this function will fail. > > > > > > > > That would seem like an understatement. > > > > > > > > Any idea when this was introduced (and which patch did), Fixes: and Cc: > > > > stable should be added if this is already in a release. > > > > > > It's actually an issue in the pipeline validation code in the V4L2 core. > > > The link is marked as MUST_CONNECT, but that isn't handled properly :-( > > > It's been on my todo list for a while but I haven't had time to get to > > > it. Feel free to give it a go. > > > > What's wrong there? It used to work at least... > > It's called a regression :-) If I recall correctly, if the pad is not > connected, it's not added to the list of pads to check, and the > MUST_CONNECT flag is not checked. This has potential for other similar issues in a number of drivers. I suppose this broke with the graph traversal changes ~ a year ago? -- Sakari Ailus
