On Wed, 2005-09-14 at 11:49 -0400, Alan Stern wrote: > (James, I see a possible problem with scsi_insert_special_req. It adds to > the queue a request with REQ_DONTPREP set. How can such a request, with > no associated scsi_cmnd, ever work? Also, won't scsi_end_request and > __scsi_release_request end up putting the same scsi_command twice?) It's a historical anomaly which will hopefully die when we finally manage to get sg and st converted to the generic request infrastructure. Then scsi_request can be killed and this along with it. What used to happen (as the comment implies) is that drivers would allocate a single request and then reuse it for multiple independent commands. Since they weren't too picky about cleaning it up after each use, we had to reset the DONTPREP flag to ensure each new invocation was actually correctly prepared. 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
