Hi Frank, On Thu, 2010-12-09 at 11:22 +0100, Frank Wille wrote: > Was the driver ever tested on big-endian systems? Could there be a bug? I don't believe uvideo has been used on anything other than i386/amd64, so it's quite possible that there's an endian fault. > I just skimmed over the source of video.c and uvideo.c without > understanding it, but maybe there has to be an endian-conversion in > buf32tobuf() and buftobuf32() of video.c? As I understand it, those are related to COMPAT32 and handle the situation where a 32-bit process is running under a 64-bit kernel (i.e. some ABI incompatibilities). The most likely place for a problem to occur is in uvideo.c, where control data gets packed/unpacked to/from structs that go to USB. Could you recompile with options UVIDEO_DEBUG and VIDEO_DEBUG, then send the relevant portions of dmesg? Those tend to dump a lot of descriptor data, which might reveal any fields that haven't been correctly swapped. -- Thanks, Jeremy
Attachment:
signature.asc
Description: This is a digitally signed message part