James Bottomley wrote: > On Sun, 2005-05-15 at 10:15 +0900, Tejun Heo wrote: > >> I've made two new versions of the same patch. The first one just >>BUG() such cases, and the second one makes scsi_prep_fn() tell >>scsi_request_fn() to kill requests instead of doing itself w/ >>BLKPREP_KILL. In both cases, I made req->flags error case a BUG(). >>If you don't like it, feel free to drop that part. >> >> Oh... one more thing. I forgot to mention the scsi_kill_requests() >>path. As it's a temporary fix, I just left it as it is (terminate >>commands w/ end_that_*). I guess this patch should be pushed after >>removal of that kludge. But with or without this patch, that path >>will leak resources. > > > I suppose it's not surprising that I don't like either. > > You remove the code that handles the BLKPREP_KILL case and then contort > the request functions to try not to do it. We have to handle this case, > it's not optional, so just leave the code that does it in. IMHO, as special request BLKPREP_KILL path is a kernel bug, code that try to handle it normally is unnecessary, but it's your call. Thanks. -- tejun - : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
