On 10/08/2014 10:41 PM, Elliott, Robert (Server Storage) wrote:
-----Original Message-----
From: Hannes Reinecke [mailto:hare@xxxxxxx]
Sent: Tuesday, 07 October, 2014 4:03 AM
To: James Bottomley
Cc: Christoph Hellwig; linux-scsi@xxxxxxxxxxxxxxx; Elliott, Robert (Server
Storage); Hannes Reinecke
Subject: [PATCH 21/26] scsi: simplify scsi_log_(send|completion)
Simplify scsi_log_(send|completion) by externalizing
scsi_mlreturn_string() and always print the command address.
...
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index b853659..2df485f 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -831,7 +831,7 @@ void scsi_io_completion(struct scsi_cmnd *cmd, unsigned
int good_bytes)
int error = 0;
struct scsi_sense_hdr sshdr;
bool sense_valid = false;
- int sense_deferred = 0;
+ int sense_deferred = 0, level = 0;
enum {ACTION_FAIL, ACTION_REPREP, ACTION_RETRY,
ACTION_DELAYED_RETRY} action;
unsigned long wait_for = (cmd->allowed + 1) * req->timeout;
@@ -1037,8 +1037,15 @@ void scsi_io_completion(struct scsi_cmnd *cmd,
unsigned int good_bytes)
switch (action) {
case ACTION_FAIL:
/* Give up and fail the remainder of the request */
- if (!(req->cmd_flags & REQ_QUIET)) {
- scsi_print_result(cmd);
+ if (unlikely(scsi_logging_level))
+ level = SCSI_LOG_LEVEL(SCSI_LOG_MLQUEUE_SHIFT,
+ SCSI_LOG_MLQUEUE_BITS);
scsi_log_completion prints based on SCSI_LOG_MLCOMPLETE,
not SCSI_LOG_MLQUEUE, so this is not accomplishing what
the next comment says.
+ /*
+ * if logging is enabled the failure will be printed
+ * in scsi_log_completion(), so avoid duplicate messages
+ */
+ if (!level && !(req->cmd_flags & REQ_QUIET)) {
+ scsi_print_result(cmd, NULL, FAILED);
if (driver_byte(result) & DRIVER_SENSE)
scsi_print_sense(cmd);
scsi_print_command(cmd);
This still results in prints on every completion with ACTION_FAIL.
Since they're not ratelimited, the massive number of prints after
a device failure under heavy load causes other timeouts. Since
the block layer also prints errors, with ratelimiting, these are
not really necessary (although they're more informative than the
block layer prints). There needs to be a way to turn these off,
both in scsi_log_completion and here - maybe different MLCOMPLETE
levels?
The issue will be mitigated if the next patch series makes these
ratelimited.
Okay, I'll be adding a printk_ratelimited here.
Cheers,
Hannes
--
Dr. Hannes Reinecke zSeries & Storage
hare@xxxxxxx +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: J. Hawn, J. Guild, F. Imendörffer, HRB 16746 (AG Nürnberg)
--
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
