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 <stern@xxxxxxxxxxxxxxxxxxx>
> CC: Matthew Wilcox <matthew@xxxxxx>
> Signed-off-by: Boaz Harrosh <bharrosh@xxxxxxxxxxx>
> ---
> 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.
--
To unsubscribe from this list: 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
