(Sorry for the double post, the previous one did not include the patch)
In cx88-mpeg.c, there is code that sets core->input to CX88_VMUX_DVB.
However, this may be incorrect, since core->input is actually an
index to core->board.input[], which has not enough elements to be
indexed by the value of CX88_VMUX_DVB. So, the modified code searches
core->board.input[] for an input with a type of CX88_VMUX_DVB, and if
it does not find one, the index is simply set to zero.
The change may not have much effect, though, since it appears the only
case when core->input is actually used is when the current input is
being queried.
Signed-off-by: Istvan Varga <istvanv@xxxxxxxxxxxxxxxxxxxxx>
diff -r -d -N -U4 v4l-dvb-a79dd2ae4d0e.old/linux/drivers/media/video/cx88/cx88-mpeg.c v4l-dvb-a79dd2ae4d0e/linux/drivers/media/video/cx88/cx88-mpeg.c
--- v4l-dvb-a79dd2ae4d0e.old/linux/drivers/media/video/cx88/cx88-mpeg.c 2010-03-23 03:39:52.000000000 +0100
+++ v4l-dvb-a79dd2ae4d0e/linux/drivers/media/video/cx88/cx88-mpeg.c 2010-03-23 19:07:26.000000000 +0100
@@ -637,15 +637,24 @@
/* Driver asked for hardware access. */
static int cx8802_request_acquire(struct cx8802_driver *drv)
{
struct cx88_core *core = drv->core;
+ unsigned int i;
/* Fail a request for hardware if the device is busy. */
if (core->active_type_id != CX88_BOARD_NONE &&
core->active_type_id != drv->type_id)
return -EBUSY;
- core->input = CX88_VMUX_DVB;
+ core->input = 0;
+ for (i = 0;
+ i < (sizeof(core->board.input) / sizeof(struct cx88_input));
+ i++) {
+ if (core->board.input[i].type == CX88_VMUX_DVB) {
+ core->input = i;
+ break;
+ }
+ }
if (drv->advise_acquire)
{
mutex_lock(&drv->core->lock);
