Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Pull SCSI fixes from James Bottomley:
"Two straightforward fixes.
One is a concurrency issue only affecting SAS connected SATA drives,
but which could hang the storage subsystem if it triggers (because the
outstanding command count on error never goes back to zero) and the
other is a NO_TAG fallout from the switch to hostwide tags which
causes the system to crash on module insertion (we've checked
carefully and only the 53c700 family of drivers is vulnerable to this
issue)"
* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
53c700: fix BUG on untagged commands
scsi: fix race between simultaneous decrements of ->host_failed
+12
-6
+6
-2
Documentation/scsi/scsi_eh.txt
+6
-2
Documentation/scsi/scsi_eh.txt
···
263
263
264
264
3. scmd recovered
265
265
ACTION: scsi_eh_finish_cmd() is invoked to EH-finish scmd
266
-
- shost->host_failed--
267
266
- clear scmd->eh_eflags
268
267
- scsi_setup_cmd_retry()
269
268
- move from local eh_work_q to local eh_done_q
270
269
LOCKING: none
270
+
CONCURRENCY: at most one thread per separate eh_work_q to
271
+
keep queue manipulation lockless
271
272
272
273
4. EH completes
273
274
ACTION: scsi_eh_flush_done_q() retries scmds or notifies upper
274
-
layer of failure.
275
+
layer of failure. May be called concurrently but must have
276
+
a no more than one thread per separate eh_work_q to
277
+
manipulate the queue locklessly
275
278
- scmd is removed from eh_done_q and scmd->eh_entry is cleared
276
279
- if retry is necessary, scmd is requeued using
277
280
scsi_queue_insert()
278
281
- otherwise, scsi_finish_command() is invoked for scmd
282
+
- zero shost->host_failed
279
283
LOCKING: queue or finish function performs appropriate locking
280
284
281
285
+1
-1
drivers/ata/libata-eh.c
+1
-1
drivers/ata/libata-eh.c
+2
-2
drivers/scsi/53c700.c
+2
-2
drivers/scsi/53c700.c
···
1122
1122
} else {
1123
1123
struct scsi_cmnd *SCp;
1124
1124
1125
-
SCp = scsi_host_find_tag(SDp->host, SCSI_NO_TAG);
1125
+
SCp = SDp->current_cmnd;
1126
1126
if(unlikely(SCp == NULL)) {
1127
1127
sdev_printk(KERN_ERR, SDp,
1128
1128
"no saved request for untagged cmd\n");
···
1826
1826
slot->tag, slot);
1827
1827
} else {
1828
1828
slot->tag = SCSI_NO_TAG;
1829
-
/* must populate current_cmnd for scsi_host_find_tag to work */
1829
+
/* save current command for reselection */
1830
1830
SCp->device->current_cmnd = SCp;
1831
1831
}
1832
1832
/* sanity check: some of the commands generated by the mid-layer
+3
-1
drivers/scsi/scsi_error.c
+3
-1
drivers/scsi/scsi_error.c
···
1128
1128
*/
1129
1129
void scsi_eh_finish_cmd(struct scsi_cmnd *scmd, struct list_head *done_q)
1130
1130
{
1131
-
scmd->device->host->host_failed--;
1132
1131
scmd->eh_eflags = 0;
1133
1132
list_move_tail(&scmd->eh_entry, done_q);
1134
1133
}
···
2225
2226
shost->transportt->eh_strategy_handler(shost);
2226
2227
else
2227
2228
scsi_unjam_host(shost);
2229
+
2230
+
/* All scmds have been handled */
2231
+
shost->host_failed = 0;
2228
2232
2229
2233
/*
2230
2234
* Note - if the above fails completely, the action is to take