On Sun, 2005-05-15 at 00:47 +0900, Tejun Heo wrote: > BLKPREP_KILL is only used to kill illegal (unpreparable, way-off) > requests. Actually, for special requests, the only tests performed > are req->flags and CDB_SIZE tests. I don't think anyone does/will > submit that illegal requests via scsi_wait_req(). And if so, it will > be a bug. True, but without the code you're removing it will simply hang the system, which isn't a correct response to a detected bug. And if I had a shilling for every time someone's predicated a code change on "oh, users will never do this" ... I'd be reasonably rich. This also leads naturally into the next observation: Checking in the request function should be done. However, it makes little sense wasting resources preparing requests we know are going to be killed, so the correct thing to do seems to be to abstract the checks and do them in both prep_fn and request_fn. James - : 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
