From mboxrd@z Thu Jan 1 00:00:00 1970 From: Boaz Harrosh Subject: Re: [PATCH 3/3 version 3] scsi_lib: Collapse scsi_end_request into only user Date: Tue, 05 Jan 2010 10:49:18 +0200 Message-ID: <4B42FD0E.8030203@panasas.com> References: <4B41DADD.4000102@panasas.com> <4B41DC42.1090706@panasas.com> <4B41F637.8070809@panasas.com> <4B42EFF4.30409@panasas.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Return-path: Received: from daytona.panasas.com ([67.152.220.89]:15164 "EHLO daytona.int.panasas.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1753636Ab0AEItW (ORCPT ); Tue, 5 Jan 2010 03:49:22 -0500 In-Reply-To: <4B42EFF4.30409@panasas.com> Sender: linux-scsi-owner@vger.kernel.org List-Id: linux-scsi@vger.kernel.org To: James Bottomley , Alan Stern , Mike Christie , Hannes Reinecke , linux-scsi On 01/05/2010 09:53 AM, Boaz Harrosh wrote: > > Embedding scsi_end_request() into scsi_io_completion actually simplifies > the code and makes it clearer what's going on. > > There is absolutely no functional and/or side effects changes after this > patch. > > Patch was inspired by Alan Stern. > > CC: Alan Stern > CC: Matthew Wilcox > Signed-off-by: Boaz Harrosh > --- > drivers/scsi/scsi_lib.c | 92 ++++++++++------------------------------------- > 1 files changed, 19 insertions(+), 73 deletions(-) > > diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c > index 8d8b4eb..41df4e8 100644 > --- a/drivers/scsi/scsi_lib.c > +++ b/drivers/scsi/scsi_lib.c > @@ -512,66 +512,6 @@ void scsi_run_host_queues(struct Scsi_Host *shost) > scsi_run_queue(sdev->request_queue); > } > > -/* > - * Function: scsi_end_request() > - * > - * Purpose: Post-processing of completed commands (usually invoked at end > - * of upper level post-processing and scsi_io_completion). > - * > - * Arguments: cmd - command that is complete. > - * error - 0 if I/O indicates success, < 0 for I/O error. > - * bytes - number of bytes of completed I/O > - * requeue - indicates whether we should requeue leftovers. > - * > - * Lock status: Assumed that lock is not held upon entry. > - * > - * Returns: cmd if requeue required, NULL otherwise. > - * > - * Notes: This is called for block device requests in order to > - * mark some number of sectors as complete. > - * > - * We are guaranteeing that the request queue will be goosed > - * at some point during this call. > - * Notes: If cmd was requeued, upon return it will be a stale pointer. > - */ > -static struct scsi_cmnd *scsi_end_request(struct scsi_cmnd *cmd, int error, > - int bytes, int requeue) > -{ > - struct request *req = cmd->request; > - > - /* > - * If there are blocks left over at the end, set up the command > - * to queue the remainder of them. > - */ > - if (blk_end_request(req, error, bytes)) { > - /* kill remainder if no retrys */ > - if (error && scsi_noretry_cmd(cmd)) > - blk_end_request_all(req, error); > - else { > - if (requeue) { > - /* > - * Bleah. Leftovers again. Stick the > - * leftovers in the front of the > - * queue, and goose the queue again. > - */ > - scsi_release_buffers(cmd); > - scsi_requeue_command(cmd); > - cmd = NULL; > - } > - return cmd; > - } > - } > - > - cmd->request = NULL; > - /* > - * This will goose the queue request function at the end, so we don't > - * need to worry about launching another command. > - */ > - scsi_release_buffers(cmd); > - scsi_next_command(cmd); > - return NULL; > -} > - > static inline unsigned int scsi_sgtable_index(unsigned short nents) > { > unsigned int index; > @@ -704,7 +644,7 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes) > struct scsi_sense_hdr sshdr; > int sense_valid = 0; > int sense_deferred = 0; > - enum {ACTION_FAIL, ACTION_REPREP, ACTION_RETRY, > + enum {ACTION_FAIL, ACTION_REPREP, ACTION_NEXT_CMND, ACTION_RETRY, > ACTION_DELAYED_RETRY} action; > char *description = NULL; > > @@ -773,17 +713,19 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes) > error = 0; > } > > - /* > - * A number of bytes were successfully read. If there > - * are leftovers and there is some kind of error > - * (result != 0), retry the rest. > - */ > - if (scsi_end_request(cmd, error, good_bytes, result == 0) == NULL) > - return; > - > - error = -EIO; > - > - if (host_byte(result) == DID_RESET) { > + if (likely(0 == blk_end_request(req, error, good_bytes))) { > + /* All is done and good move to next command */ > + cmd->request = NULL; > + action = ACTION_NEXT_CMND; > + } else if (result == 0) { > + /* Wrote some bytes but request was split */ > + action = ACTION_REPREP; > + } else if (error && scsi_noretry_cmd(cmd)) { > + /* kill remainder if no retries */ > + blk_end_request_all(req, error); > + cmd->request = NULL; > + action = ACTION_NEXT_CMND; > + } else if (host_byte(result) == DID_RESET) { > /* Third party bus reset or reset for error recovery > * reasons. Just retry the command and see what > * happens. > @@ -891,11 +833,15 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned int good_bytes) > scsi_print_sense("", cmd); > scsi_print_command(cmd); > } > - if (blk_end_request_err(req, error)) > + if (blk_end_request_err(req, error ? error : -EIO)) > scsi_requeue_command(cmd); > else > scsi_next_command(cmd); > break; > + case ACTION_NEXT_CMND: > + scsi_release_buffers(cmd); > + scsi_next_command(cmd); > + break; In a switch expression, Does the first case have an advantage? I would think that even if the branch predictor does it's job the cache proximity to the head of the switch would matter. Is there a likely() for case statements? In anyway I thinks I'll send a 4th version with this case at the top of the switch Boaz > case ACTION_REPREP: > /* Unprep the request and put it back at the head of the queue. > * A new command will be prepared and issued.