diff options
Diffstat (limited to 'drivers')
-rw-r--r-- | drivers/media/video/bt8xx/bttv-driver.c | 31 |
1 files changed, 27 insertions, 4 deletions
diff --git a/drivers/media/video/bt8xx/bttv-driver.c b/drivers/media/video/bt8xx/bttv-driver.c index 939d1e51297..3ff84c8185a 100644 --- a/drivers/media/video/bt8xx/bttv-driver.c +++ b/drivers/media/video/bt8xx/bttv-driver.c @@ -3800,11 +3800,34 @@ bttv_irq_next_video(struct bttv *btv, struct bttv_buffer_set *set) if (!V4L2_FIELD_HAS_BOTH(item->vb.field) && (item->vb.queue.next != &btv->capture)) { item = list_entry(item->vb.queue.next, struct bttv_buffer, vb.queue); + /* Mike Isely <isely@pobox.com> - Only check + * and set up the bottom field in the logic + * below. Don't ever do the top field. This + * of course means that if we set up the + * bottom field in the above code that we'll + * actually skip a field. But that's OK. + * Having processed only a single buffer this + * time, then the next time around the first + * available buffer should be for a top field. + * That will then cause us here to set up a + * top then a bottom field in the normal way. + * The alternative to this understanding is + * that we set up the second available buffer + * as a top field, but that's out of order + * since this driver always processes the top + * field first - the effect will be the two + * buffers being returned in the wrong order, + * with the second buffer also being delayed + * by one field time (owing to the fifo nature + * of videobuf). Worse still, we'll be stuck + * doing fields out of order now every time + * until something else causes a field to be + * dropped. By effectively forcing a field to + * drop this way then we always get back into + * sync within a single frame time. (Out of + * order fields can screw up deinterlacing + * algorithms.) */ if (!V4L2_FIELD_HAS_BOTH(item->vb.field)) { - if (NULL == set->top && - V4L2_FIELD_TOP == item->vb.field) { - set->top = item; - } if (NULL == set->bottom && V4L2_FIELD_BOTTOM == item->vb.field) { set->bottom = item; |