Set bh->b_count to 1 in dispatch_bh(). I'm not entirely sure this is necessary. RAID-1 sets this value to 1, RAID-5 sets it to zero, and loop doesn't set it at all. I guess it all really depends on if any of the device drivers are going to access that field. The only obvious place I see it being accessed is in ide-floppy and ide-tape. This patch sets it to one. May not be required, but shouldn't hurt anything. Anyone have any insights? --- linux-2.4.20a/drivers/md/kcopyd.c 2003/01/09 17:11:50 +++ linux-2.4.20b/drivers/md/kcopyd.c 2003/01/09 17:34:29 @@ -338,6 +338,7 @@ init_buffer(bh, end_bh, job); bh->b_dev = job->disk.dev; + atomic_set(&bh->b_count, 1); set_bit(BH_Uptodate, &bh->b_state); set_bit(BH_Mapped, &bh->b_state);