Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * scsi_error.c Copyright (C) 1997 Eric Youngdale | |
4 | * | |
5 | * SCSI error/timeout handling | |
6 | * Initial versions: Eric Youngdale. Based upon conversations with | |
0bf8c869 | 7 | * Leonard Zubkoff and David Miller at Linux Expo, |
1da177e4 LT |
8 | * ideas originating from all over the place. |
9 | * | |
10 | * Restructured scsi_unjam_host and associated functions. | |
11 | * September 04, 2002 Mike Anderson (andmike@us.ibm.com) | |
12 | * | |
13 | * Forward port of Russell King's (rmk@arm.linux.org.uk) changes and | |
0bf8c869 | 14 | * minor cleanups. |
1da177e4 LT |
15 | * September 30, 2002 Mike Anderson (andmike@us.ibm.com) |
16 | */ | |
17 | ||
18 | #include <linux/module.h> | |
19 | #include <linux/sched.h> | |
5a0e3ad6 | 20 | #include <linux/gfp.h> |
1da177e4 LT |
21 | #include <linux/timer.h> |
22 | #include <linux/string.h> | |
1da177e4 | 23 | #include <linux/kernel.h> |
83144186 | 24 | #include <linux/freezer.h> |
c5478def | 25 | #include <linux/kthread.h> |
1da177e4 LT |
26 | #include <linux/interrupt.h> |
27 | #include <linux/blkdev.h> | |
28 | #include <linux/delay.h> | |
fc73648a | 29 | #include <linux/jiffies.h> |
1da177e4 LT |
30 | |
31 | #include <scsi/scsi.h> | |
beb40487 | 32 | #include <scsi/scsi_cmnd.h> |
1da177e4 LT |
33 | #include <scsi/scsi_dbg.h> |
34 | #include <scsi/scsi_device.h> | |
18a4d0a2 | 35 | #include <scsi/scsi_driver.h> |
1da177e4 | 36 | #include <scsi/scsi_eh.h> |
7708c165 | 37 | #include <scsi/scsi_common.h> |
c829c394 | 38 | #include <scsi/scsi_transport.h> |
1da177e4 LT |
39 | #include <scsi/scsi_host.h> |
40 | #include <scsi/scsi_ioctl.h> | |
ee14c674 | 41 | #include <scsi/scsi_dh.h> |
29cfc2ab | 42 | #include <scsi/scsi_devinfo.h> |
176aa9d6 | 43 | #include <scsi/sg.h> |
1da177e4 LT |
44 | |
45 | #include "scsi_priv.h" | |
46 | #include "scsi_logging.h" | |
79ee8304 | 47 | #include "scsi_transport_api.h" |
1da177e4 | 48 | |
bf816235 KT |
49 | #include <trace/events/scsi.h> |
50 | ||
2908769c DLM |
51 | #include <asm/unaligned.h> |
52 | ||
1da177e4 LT |
53 | /* |
54 | * These should *probably* be handled by the host itself. | |
55 | * Since it is allowed to sleep, it probably should. | |
56 | */ | |
57 | #define BUS_RESET_SETTLE_TIME (10) | |
58 | #define HOST_RESET_SETTLE_TIME (10) | |
59 | ||
3eef6257 | 60 | static int scsi_eh_try_stu(struct scsi_cmnd *scmd); |
b8e162f9 BVA |
61 | static enum scsi_disposition scsi_try_to_abort_cmd(struct scsi_host_template *, |
62 | struct scsi_cmnd *); | |
3eef6257 | 63 | |
1da177e4 LT |
64 | void scsi_eh_wakeup(struct Scsi_Host *shost) |
65 | { | |
f0317e88 BVA |
66 | lockdep_assert_held(shost->host_lock); |
67 | ||
c84b023a | 68 | if (scsi_host_busy(shost) == shost->host_failed) { |
bf816235 | 69 | trace_scsi_eh_wakeup(shost); |
3ed7a470 | 70 | wake_up_process(shost->ehandler); |
91921e01 HR |
71 | SCSI_LOG_ERROR_RECOVERY(5, shost_printk(KERN_INFO, shost, |
72 | "Waking error handler thread\n")); | |
1da177e4 LT |
73 | } |
74 | } | |
f8bbfc24 TH |
75 | |
76 | /** | |
77 | * scsi_schedule_eh - schedule EH for SCSI host | |
78 | * @shost: SCSI host to invoke error handling on. | |
79 | * | |
80 | * Schedule SCSI EH without scmd. | |
dc8875e1 | 81 | */ |
f8bbfc24 TH |
82 | void scsi_schedule_eh(struct Scsi_Host *shost) |
83 | { | |
84 | unsigned long flags; | |
85 | ||
86 | spin_lock_irqsave(shost->host_lock, flags); | |
87 | ||
88 | if (scsi_host_set_state(shost, SHOST_RECOVERY) == 0 || | |
89 | scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY) == 0) { | |
90 | shost->host_eh_scheduled++; | |
91 | scsi_eh_wakeup(shost); | |
92 | } | |
93 | ||
94 | spin_unlock_irqrestore(shost->host_lock, flags); | |
95 | } | |
96 | EXPORT_SYMBOL_GPL(scsi_schedule_eh); | |
1da177e4 | 97 | |
b4562022 HR |
98 | static int scsi_host_eh_past_deadline(struct Scsi_Host *shost) |
99 | { | |
bb3b621a | 100 | if (!shost->last_reset || shost->eh_deadline == -1) |
b4562022 HR |
101 | return 0; |
102 | ||
76ad3e59 HR |
103 | /* |
104 | * 32bit accesses are guaranteed to be atomic | |
105 | * (on all supported architectures), so instead | |
106 | * of using a spinlock we can as well double check | |
bb3b621a | 107 | * if eh_deadline has been set to 'off' during the |
76ad3e59 HR |
108 | * time_before call. |
109 | */ | |
110 | if (time_before(jiffies, shost->last_reset + shost->eh_deadline) && | |
bb3b621a | 111 | shost->eh_deadline > -1) |
b4562022 HR |
112 | return 0; |
113 | ||
114 | return 1; | |
115 | } | |
116 | ||
2a242d59 MC |
117 | static bool scsi_cmd_retry_allowed(struct scsi_cmnd *cmd) |
118 | { | |
119 | if (cmd->allowed == SCSI_CMD_RETRIES_NO_LIMIT) | |
120 | return true; | |
121 | ||
122 | return ++cmd->retries <= cmd->allowed; | |
123 | } | |
124 | ||
60bee27b MK |
125 | static bool scsi_eh_should_retry_cmd(struct scsi_cmnd *cmd) |
126 | { | |
127 | struct scsi_device *sdev = cmd->device; | |
128 | struct Scsi_Host *host = sdev->host; | |
129 | ||
130 | if (host->hostt->eh_should_retry_cmd) | |
131 | return host->hostt->eh_should_retry_cmd(cmd); | |
132 | ||
133 | return true; | |
134 | } | |
135 | ||
e494f6a7 HR |
136 | /** |
137 | * scmd_eh_abort_handler - Handle command aborts | |
138 | * @work: command to be aborted. | |
923f46f9 BVA |
139 | * |
140 | * Note: this function must be called only for a command that has timed out. | |
141 | * Because the block layer marks a request as complete before it calls | |
deef1be1 | 142 | * scsi_timeout(), a .scsi_done() call from the LLD for a command that has |
923f46f9 BVA |
143 | * timed out do not have any effect. Hence it is safe to call |
144 | * scsi_finish_command() from this function. | |
e494f6a7 HR |
145 | */ |
146 | void | |
147 | scmd_eh_abort_handler(struct work_struct *work) | |
148 | { | |
149 | struct scsi_cmnd *scmd = | |
150 | container_of(work, struct scsi_cmnd, abort_work.work); | |
151 | struct scsi_device *sdev = scmd->device; | |
54d816d3 | 152 | struct Scsi_Host *shost = sdev->host; |
b8e162f9 | 153 | enum scsi_disposition rtn; |
5ae17501 | 154 | unsigned long flags; |
e494f6a7 | 155 | |
54d816d3 | 156 | if (scsi_host_eh_past_deadline(shost)) { |
e494f6a7 HR |
157 | SCSI_LOG_ERROR_RECOVERY(3, |
158 | scmd_printk(KERN_INFO, scmd, | |
470613b4 | 159 | "eh timeout, not aborting\n")); |
54d816d3 EM |
160 | goto out; |
161 | } | |
162 | ||
163 | SCSI_LOG_ERROR_RECOVERY(3, | |
e494f6a7 | 164 | scmd_printk(KERN_INFO, scmd, |
470613b4 | 165 | "aborting command\n")); |
54d816d3 EM |
166 | rtn = scsi_try_to_abort_cmd(shost->hostt, scmd); |
167 | if (rtn != SUCCESS) { | |
168 | SCSI_LOG_ERROR_RECOVERY(3, | |
169 | scmd_printk(KERN_INFO, scmd, | |
170 | "cmd abort %s\n", | |
171 | (rtn == FAST_IO_FAIL) ? | |
172 | "not send" : "failed")); | |
173 | goto out; | |
174 | } | |
175 | set_host_byte(scmd, DID_TIME_OUT); | |
176 | if (scsi_host_eh_past_deadline(shost)) { | |
177 | SCSI_LOG_ERROR_RECOVERY(3, | |
178 | scmd_printk(KERN_INFO, scmd, | |
179 | "eh timeout, not retrying " | |
180 | "aborted command\n")); | |
181 | goto out; | |
e494f6a7 HR |
182 | } |
183 | ||
54d816d3 | 184 | spin_lock_irqsave(shost->host_lock, flags); |
5ae17501 | 185 | list_del_init(&scmd->eh_entry); |
54d816d3 EM |
186 | |
187 | /* | |
188 | * If the abort succeeds, and there is no further | |
189 | * EH action, clear the ->last_reset time. | |
190 | */ | |
191 | if (list_empty(&shost->eh_abort_list) && | |
192 | list_empty(&shost->eh_cmd_q)) | |
193 | if (shost->eh_deadline != -1) | |
194 | shost->last_reset = 0; | |
195 | ||
196 | spin_unlock_irqrestore(shost->host_lock, flags); | |
197 | ||
198 | if (!scsi_noretry_cmd(scmd) && | |
199 | scsi_cmd_retry_allowed(scmd) && | |
200 | scsi_eh_should_retry_cmd(scmd)) { | |
201 | SCSI_LOG_ERROR_RECOVERY(3, | |
202 | scmd_printk(KERN_WARNING, scmd, | |
203 | "retry aborted command\n")); | |
204 | scsi_queue_insert(scmd, SCSI_MLQUEUE_EH_RETRY); | |
205 | } else { | |
206 | SCSI_LOG_ERROR_RECOVERY(3, | |
207 | scmd_printk(KERN_WARNING, scmd, | |
208 | "finish aborted command\n")); | |
209 | scsi_finish_command(scmd); | |
210 | } | |
211 | return; | |
212 | ||
213 | out: | |
214 | spin_lock_irqsave(shost->host_lock, flags); | |
215 | list_del_init(&scmd->eh_entry); | |
216 | spin_unlock_irqrestore(shost->host_lock, flags); | |
217 | ||
a0658632 | 218 | scsi_eh_scmd_add(scmd); |
e494f6a7 HR |
219 | } |
220 | ||
221 | /** | |
222 | * scsi_abort_command - schedule a command abort | |
223 | * @scmd: scmd to abort. | |
224 | * | |
225 | * We only need to abort commands after a command timeout | |
226 | */ | |
227 | static int | |
228 | scsi_abort_command(struct scsi_cmnd *scmd) | |
229 | { | |
230 | struct scsi_device *sdev = scmd->device; | |
231 | struct Scsi_Host *shost = sdev->host; | |
232 | unsigned long flags; | |
233 | ||
234 | if (scmd->eh_eflags & SCSI_EH_ABORT_SCHEDULED) { | |
235 | /* | |
236 | * Retry after abort failed, escalate to next level. | |
237 | */ | |
238 | SCSI_LOG_ERROR_RECOVERY(3, | |
239 | scmd_printk(KERN_INFO, scmd, | |
470613b4 | 240 | "previous abort failed\n")); |
fcc95a76 | 241 | BUG_ON(delayed_work_pending(&scmd->abort_work)); |
e494f6a7 HR |
242 | return FAILED; |
243 | } | |
244 | ||
e494f6a7 | 245 | spin_lock_irqsave(shost->host_lock, flags); |
bb3b621a | 246 | if (shost->eh_deadline != -1 && !shost->last_reset) |
e494f6a7 | 247 | shost->last_reset = jiffies; |
5ae17501 EM |
248 | BUG_ON(!list_empty(&scmd->eh_entry)); |
249 | list_add_tail(&scmd->eh_entry, &shost->eh_abort_list); | |
e494f6a7 HR |
250 | spin_unlock_irqrestore(shost->host_lock, flags); |
251 | ||
252 | scmd->eh_eflags |= SCSI_EH_ABORT_SCHEDULED; | |
253 | SCSI_LOG_ERROR_RECOVERY(3, | |
470613b4 | 254 | scmd_printk(KERN_INFO, scmd, "abort scheduled\n")); |
e494f6a7 HR |
255 | queue_delayed_work(shost->tmf_work_q, &scmd->abort_work, HZ / 100); |
256 | return SUCCESS; | |
257 | } | |
258 | ||
1da177e4 | 259 | /** |
7a38dc0b | 260 | * scsi_eh_reset - call into ->eh_action to reset internal counters |
1da177e4 | 261 | * @scmd: scmd to run eh on. |
1da177e4 | 262 | * |
7a38dc0b HR |
263 | * The scsi driver might be carrying internal state about the |
264 | * devices, so we need to call into the driver to reset the | |
265 | * internal state once the error handler is started. | |
dc8875e1 | 266 | */ |
7a38dc0b HR |
267 | static void scsi_eh_reset(struct scsi_cmnd *scmd) |
268 | { | |
aa8e25e5 | 269 | if (!blk_rq_is_passthrough(scsi_cmd_to_rq(scmd))) { |
7a38dc0b HR |
270 | struct scsi_driver *sdrv = scsi_cmd_to_driver(scmd); |
271 | if (sdrv->eh_reset) | |
272 | sdrv->eh_reset(scmd); | |
273 | } | |
274 | } | |
275 | ||
3bd6f43f BVA |
276 | static void scsi_eh_inc_host_failed(struct rcu_head *head) |
277 | { | |
3be8828f BVA |
278 | struct scsi_cmnd *scmd = container_of(head, typeof(*scmd), rcu); |
279 | struct Scsi_Host *shost = scmd->device->host; | |
3bd6f43f BVA |
280 | unsigned long flags; |
281 | ||
282 | spin_lock_irqsave(shost->host_lock, flags); | |
283 | shost->host_failed++; | |
284 | scsi_eh_wakeup(shost); | |
285 | spin_unlock_irqrestore(shost->host_lock, flags); | |
286 | } | |
287 | ||
1da177e4 LT |
288 | /** |
289 | * scsi_eh_scmd_add - add scsi cmd to error handling. | |
290 | * @scmd: scmd to run eh on. | |
dc8875e1 | 291 | */ |
a0658632 | 292 | void scsi_eh_scmd_add(struct scsi_cmnd *scmd) |
1da177e4 LT |
293 | { |
294 | struct Scsi_Host *shost = scmd->device->host; | |
295 | unsigned long flags; | |
2171b6d0 | 296 | int ret; |
1da177e4 | 297 | |
2171b6d0 | 298 | WARN_ON_ONCE(!shost->ehandler); |
1da177e4 LT |
299 | |
300 | spin_lock_irqsave(shost->host_lock, flags); | |
2171b6d0 HR |
301 | if (scsi_host_set_state(shost, SHOST_RECOVERY)) { |
302 | ret = scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY); | |
303 | WARN_ON_ONCE(ret); | |
304 | } | |
bb3b621a | 305 | if (shost->eh_deadline != -1 && !shost->last_reset) |
b4562022 HR |
306 | shost->last_reset = jiffies; |
307 | ||
7a38dc0b | 308 | scsi_eh_reset(scmd); |
1da177e4 | 309 | list_add_tail(&scmd->eh_entry, &shost->eh_cmd_q); |
1da177e4 | 310 | spin_unlock_irqrestore(shost->host_lock, flags); |
3bd6f43f BVA |
311 | /* |
312 | * Ensure that all tasks observe the host state change before the | |
313 | * host_failed change. | |
314 | */ | |
54d87b0a | 315 | call_rcu_hurry(&scmd->rcu, scsi_eh_inc_host_failed); |
1da177e4 LT |
316 | } |
317 | ||
1da177e4 | 318 | /** |
deef1be1 | 319 | * scsi_timeout - Timeout function for normal scsi commands. |
242f9dcb | 320 | * @req: request that is timing out. |
1da177e4 LT |
321 | * |
322 | * Notes: | |
323 | * We do not need to lock this. There is the potential for a race | |
324 | * only in that the normal completion handling might run, but if the | |
325 | * normal completion function determines that the timer has already | |
326 | * fired, then it mustn't do anything. | |
dc8875e1 | 327 | */ |
9bdb4833 | 328 | enum blk_eh_timer_return scsi_timeout(struct request *req) |
1da177e4 | 329 | { |
bed2213d | 330 | struct scsi_cmnd *scmd = blk_mq_rq_to_pdu(req); |
0bf8c869 | 331 | struct Scsi_Host *host = scmd->device->host; |
6c5f8ce1 | 332 | |
bf816235 | 333 | trace_scsi_dispatch_cmd_timeout(scmd); |
1da177e4 LT |
334 | scsi_log_completion(scmd, TIMEOUT_ERROR); |
335 | ||
48517eef | 336 | atomic_inc(&scmd->device->iotmo_cnt); |
bb3b621a | 337 | if (host->eh_deadline != -1 && !host->last_reset) |
b4562022 HR |
338 | host->last_reset = jiffies; |
339 | ||
dee7121e BVA |
340 | if (host->hostt->eh_timed_out) { |
341 | switch (host->hostt->eh_timed_out(scmd)) { | |
342 | case SCSI_EH_DONE: | |
978b7922 | 343 | return BLK_EH_DONE; |
dee7121e | 344 | case SCSI_EH_RESET_TIMER: |
f1342709 | 345 | return BLK_EH_RESET_TIMER; |
dee7121e BVA |
346 | case SCSI_EH_NOT_HANDLED: |
347 | break; | |
2171b6d0 | 348 | } |
a33c070b | 349 | } |
242f9dcb | 350 | |
dee7121e BVA |
351 | /* |
352 | * If scsi_done() has already set SCMD_STATE_COMPLETE, do not modify | |
353 | * *scmd. | |
354 | */ | |
355 | if (test_and_set_bit(SCMD_STATE_COMPLETE, &scmd->state)) | |
356 | return BLK_EH_DONE; | |
ec9780e4 | 357 | atomic_inc(&scmd->device->iodone_cnt); |
dee7121e BVA |
358 | if (scsi_abort_command(scmd) != SUCCESS) { |
359 | set_host_byte(scmd, DID_TIME_OUT); | |
360 | scsi_eh_scmd_add(scmd); | |
361 | } | |
362 | ||
363 | return BLK_EH_DONE; | |
1da177e4 LT |
364 | } |
365 | ||
366 | /** | |
367 | * scsi_block_when_processing_errors - Prevent cmds from being queued. | |
368 | * @sdev: Device on which we are performing recovery. | |
369 | * | |
370 | * Description: | |
371 | * We block until the host is out of error recovery, and then check to | |
372 | * see whether the host or the device is offline. | |
373 | * | |
374 | * Return value: | |
375 | * 0 when dev was taken offline by error recovery. 1 OK to proceed. | |
dc8875e1 | 376 | */ |
1da177e4 LT |
377 | int scsi_block_when_processing_errors(struct scsi_device *sdev) |
378 | { | |
379 | int online; | |
380 | ||
939647ee | 381 | wait_event(sdev->host->host_wait, !scsi_host_in_recovery(sdev->host)); |
1da177e4 LT |
382 | |
383 | online = scsi_device_online(sdev); | |
384 | ||
1da177e4 LT |
385 | return online; |
386 | } | |
387 | EXPORT_SYMBOL(scsi_block_when_processing_errors); | |
388 | ||
389 | #ifdef CONFIG_SCSI_LOGGING | |
390 | /** | |
391 | * scsi_eh_prt_fail_stats - Log info on failures. | |
392 | * @shost: scsi host being recovered. | |
393 | * @work_q: Queue of scsi cmds to process. | |
dc8875e1 | 394 | */ |
1da177e4 LT |
395 | static inline void scsi_eh_prt_fail_stats(struct Scsi_Host *shost, |
396 | struct list_head *work_q) | |
397 | { | |
398 | struct scsi_cmnd *scmd; | |
399 | struct scsi_device *sdev; | |
400 | int total_failures = 0; | |
401 | int cmd_failed = 0; | |
402 | int cmd_cancel = 0; | |
403 | int devices_failed = 0; | |
404 | ||
405 | shost_for_each_device(sdev, shost) { | |
406 | list_for_each_entry(scmd, work_q, eh_entry) { | |
407 | if (scmd->device == sdev) { | |
408 | ++total_failures; | |
a0658632 | 409 | if (scmd->eh_eflags & SCSI_EH_ABORT_SCHEDULED) |
1da177e4 | 410 | ++cmd_cancel; |
0bf8c869 | 411 | else |
1da177e4 LT |
412 | ++cmd_failed; |
413 | } | |
414 | } | |
415 | ||
416 | if (cmd_cancel || cmd_failed) { | |
417 | SCSI_LOG_ERROR_RECOVERY(3, | |
a3a790dc | 418 | shost_printk(KERN_INFO, shost, |
9ccfc756 | 419 | "%s: cmds failed: %d, cancel: %d\n", |
cadbd4a5 | 420 | __func__, cmd_failed, |
9ccfc756 | 421 | cmd_cancel)); |
1da177e4 LT |
422 | cmd_cancel = 0; |
423 | cmd_failed = 0; | |
424 | ++devices_failed; | |
425 | } | |
426 | } | |
427 | ||
91921e01 HR |
428 | SCSI_LOG_ERROR_RECOVERY(2, shost_printk(KERN_INFO, shost, |
429 | "Total of %d commands on %d" | |
430 | " devices require eh work\n", | |
0bf8c869 | 431 | total_failures, devices_failed)); |
1da177e4 LT |
432 | } |
433 | #endif | |
434 | ||
279afdfe EM |
435 | /** |
436 | * scsi_report_lun_change - Set flag on all *other* devices on the same target | |
437 | * to indicate that a UNIT ATTENTION is expected. | |
438 | * @sdev: Device reporting the UNIT ATTENTION | |
439 | */ | |
440 | static void scsi_report_lun_change(struct scsi_device *sdev) | |
441 | { | |
442 | sdev->sdev_target->expecting_lun_change = 1; | |
443 | } | |
444 | ||
445 | /** | |
446 | * scsi_report_sense - Examine scsi sense information and log messages for | |
447 | * certain conditions, also issue uevents for some of them. | |
448 | * @sdev: Device reporting the sense code | |
449 | * @sshdr: sshdr to be examined | |
450 | */ | |
451 | static void scsi_report_sense(struct scsi_device *sdev, | |
452 | struct scsi_sense_hdr *sshdr) | |
453 | { | |
454 | enum scsi_device_event evt_type = SDEV_EVT_MAXBITS; /* i.e. none */ | |
455 | ||
456 | if (sshdr->sense_key == UNIT_ATTENTION) { | |
457 | if (sshdr->asc == 0x3f && sshdr->ascq == 0x03) { | |
458 | evt_type = SDEV_EVT_INQUIRY_CHANGE_REPORTED; | |
459 | sdev_printk(KERN_WARNING, sdev, | |
460 | "Inquiry data has changed"); | |
461 | } else if (sshdr->asc == 0x3f && sshdr->ascq == 0x0e) { | |
462 | evt_type = SDEV_EVT_LUN_CHANGE_REPORTED; | |
463 | scsi_report_lun_change(sdev); | |
464 | sdev_printk(KERN_WARNING, sdev, | |
279afdfe EM |
465 | "LUN assignments on this target have " |
466 | "changed. The Linux SCSI layer does not " | |
467 | "automatically remap LUN assignments.\n"); | |
468 | } else if (sshdr->asc == 0x3f) | |
469 | sdev_printk(KERN_WARNING, sdev, | |
a2417db3 | 470 | "Operating parameters on this target have " |
279afdfe EM |
471 | "changed. The Linux SCSI layer does not " |
472 | "automatically adjust these parameters.\n"); | |
473 | ||
474 | if (sshdr->asc == 0x38 && sshdr->ascq == 0x07) { | |
475 | evt_type = SDEV_EVT_SOFT_THRESHOLD_REACHED_REPORTED; | |
476 | sdev_printk(KERN_WARNING, sdev, | |
477 | "Warning! Received an indication that the " | |
478 | "LUN reached a thin provisioning soft " | |
479 | "threshold.\n"); | |
480 | } | |
481 | ||
cf3431bb HR |
482 | if (sshdr->asc == 0x29) { |
483 | evt_type = SDEV_EVT_POWER_ON_RESET_OCCURRED; | |
af4edb1d AH |
484 | /* |
485 | * Do not print message if it is an expected side-effect | |
486 | * of runtime PM. | |
487 | */ | |
488 | if (!sdev->silence_suspend) | |
489 | sdev_printk(KERN_WARNING, sdev, | |
490 | "Power-on or device reset occurred\n"); | |
cf3431bb HR |
491 | } |
492 | ||
279afdfe EM |
493 | if (sshdr->asc == 0x2a && sshdr->ascq == 0x01) { |
494 | evt_type = SDEV_EVT_MODE_PARAMETER_CHANGE_REPORTED; | |
495 | sdev_printk(KERN_WARNING, sdev, | |
496 | "Mode parameters changed"); | |
14c3e677 HR |
497 | } else if (sshdr->asc == 0x2a && sshdr->ascq == 0x06) { |
498 | evt_type = SDEV_EVT_ALUA_STATE_CHANGE_REPORTED; | |
499 | sdev_printk(KERN_WARNING, sdev, | |
500 | "Asymmetric access state changed"); | |
279afdfe EM |
501 | } else if (sshdr->asc == 0x2a && sshdr->ascq == 0x09) { |
502 | evt_type = SDEV_EVT_CAPACITY_CHANGE_REPORTED; | |
503 | sdev_printk(KERN_WARNING, sdev, | |
504 | "Capacity data has changed"); | |
505 | } else if (sshdr->asc == 0x2a) | |
506 | sdev_printk(KERN_WARNING, sdev, | |
507 | "Parameters changed"); | |
508 | } | |
509 | ||
510 | if (evt_type != SDEV_EVT_MAXBITS) { | |
511 | set_bit(evt_type, sdev->pending_events); | |
512 | schedule_work(&sdev->event_work); | |
513 | } | |
514 | } | |
515 | ||
36ebf1e2 MC |
516 | static inline void set_scsi_ml_byte(struct scsi_cmnd *cmd, u8 status) |
517 | { | |
518 | cmd->result = (cmd->result & 0xffff00ff) | (status << 8); | |
519 | } | |
520 | ||
1da177e4 LT |
521 | /** |
522 | * scsi_check_sense - Examine scsi cmd sense | |
523 | * @scmd: Cmd to have sense checked. | |
524 | * | |
525 | * Return value: | |
87f14e65 | 526 | * SUCCESS or FAILED or NEEDS_RETRY or ADD_TO_MLQUEUE |
1da177e4 LT |
527 | * |
528 | * Notes: | |
529 | * When a deferred error is detected the current command has | |
530 | * not been executed and needs retrying. | |
dc8875e1 | 531 | */ |
b8e162f9 | 532 | enum scsi_disposition scsi_check_sense(struct scsi_cmnd *scmd) |
1da177e4 | 533 | { |
a6a8d9f8 | 534 | struct scsi_device *sdev = scmd->device; |
1da177e4 LT |
535 | struct scsi_sense_hdr sshdr; |
536 | ||
537 | if (! scsi_command_normalize_sense(scmd, &sshdr)) | |
538 | return FAILED; /* no valid sense data */ | |
539 | ||
279afdfe EM |
540 | scsi_report_sense(sdev, &sshdr); |
541 | ||
1da177e4 LT |
542 | if (scsi_sense_is_deferred(&sshdr)) |
543 | return NEEDS_RETRY; | |
544 | ||
ee14c674 | 545 | if (sdev->handler && sdev->handler->check_sense) { |
b8e162f9 | 546 | enum scsi_disposition rc; |
a6a8d9f8 | 547 | |
ee14c674 | 548 | rc = sdev->handler->check_sense(sdev, &sshdr); |
a6a8d9f8 CS |
549 | if (rc != SCSI_RETURN_NOT_HANDLED) |
550 | return rc; | |
551 | /* handler does not care. Drop down to default handling */ | |
552 | } | |
553 | ||
bf23e619 BVA |
554 | if (scmd->cmnd[0] == TEST_UNIT_READY && |
555 | scmd->submitter != SUBMITTED_BY_SCSI_ERROR_HANDLER) | |
e925cc43 CH |
556 | /* |
557 | * nasty: for mid-layer issued TURs, we need to return the | |
558 | * actual sense data without any recovery attempt. For eh | |
559 | * issued ones, we need to try to recover and interpret | |
560 | */ | |
561 | return SUCCESS; | |
562 | ||
1da177e4 LT |
563 | /* |
564 | * Previous logic looked for FILEMARK, EOM or ILI which are | |
565 | * mainly associated with tapes and returned SUCCESS. | |
566 | */ | |
567 | if (sshdr.response_code == 0x70) { | |
568 | /* fixed format */ | |
569 | if (scmd->sense_buffer[2] & 0xe0) | |
570 | return SUCCESS; | |
571 | } else { | |
572 | /* | |
573 | * descriptor format: look for "stream commands sense data | |
574 | * descriptor" (see SSC-3). Assume single sense data | |
575 | * descriptor. Ignore ILI from SBC-2 READ LONG and WRITE LONG. | |
576 | */ | |
577 | if ((sshdr.additional_length > 3) && | |
578 | (scmd->sense_buffer[8] == 0x4) && | |
579 | (scmd->sense_buffer[11] & 0xe0)) | |
580 | return SUCCESS; | |
581 | } | |
582 | ||
583 | switch (sshdr.sense_key) { | |
584 | case NO_SENSE: | |
585 | return SUCCESS; | |
586 | case RECOVERED_ERROR: | |
587 | return /* soft_error */ SUCCESS; | |
588 | ||
589 | case ABORTED_COMMAND: | |
511e44f4 MP |
590 | if (sshdr.asc == 0x10) /* DIF */ |
591 | return SUCCESS; | |
592 | ||
29cfc2ab MW |
593 | if (sshdr.asc == 0x44 && sdev->sdev_bflags & BLIST_RETRY_ITF) |
594 | return ADD_TO_MLQUEUE; | |
c3606520 MW |
595 | if (sshdr.asc == 0xc1 && sshdr.ascq == 0x01 && |
596 | sdev->sdev_bflags & BLIST_RETRY_ASC_C1) | |
597 | return ADD_TO_MLQUEUE; | |
29cfc2ab | 598 | |
1da177e4 LT |
599 | return NEEDS_RETRY; |
600 | case NOT_READY: | |
601 | case UNIT_ATTENTION: | |
602 | /* | |
603 | * if we are expecting a cc/ua because of a bus reset that we | |
604 | * performed, treat this just as a retry. otherwise this is | |
605 | * information that we should pass up to the upper-level driver | |
606 | * so that we can deal with it there. | |
607 | */ | |
608 | if (scmd->device->expecting_cc_ua) { | |
dfcf7775 TH |
609 | /* |
610 | * Because some device does not queue unit | |
611 | * attentions correctly, we carefully check | |
612 | * additional sense code and qualifier so as | |
613 | * not to squash media change unit attention. | |
614 | */ | |
615 | if (sshdr.asc != 0x28 || sshdr.ascq != 0x00) { | |
616 | scmd->device->expecting_cc_ua = 0; | |
617 | return NEEDS_RETRY; | |
618 | } | |
1da177e4 | 619 | } |
279afdfe EM |
620 | /* |
621 | * we might also expect a cc/ua if another LUN on the target | |
622 | * reported a UA with an ASC/ASCQ of 3F 0E - | |
623 | * REPORTED LUNS DATA HAS CHANGED. | |
624 | */ | |
625 | if (scmd->device->sdev_target->expecting_lun_change && | |
626 | sshdr.asc == 0x3f && sshdr.ascq == 0x0e) | |
627 | return NEEDS_RETRY; | |
1da177e4 | 628 | /* |
0bf8c869 | 629 | * if the device is in the process of becoming ready, we |
1da177e4 LT |
630 | * should retry. |
631 | */ | |
632 | if ((sshdr.asc == 0x04) && (sshdr.ascq == 0x01)) | |
633 | return NEEDS_RETRY; | |
634 | /* | |
635 | * if the device is not started, we need to wake | |
636 | * the error handler to start the motor | |
637 | */ | |
638 | if (scmd->device->allow_restart && | |
639 | (sshdr.asc == 0x04) && (sshdr.ascq == 0x02)) | |
640 | return FAILED; | |
02e031cb CH |
641 | /* |
642 | * Pass the UA upwards for a determination in the completion | |
643 | * functions. | |
644 | */ | |
645 | return SUCCESS; | |
1da177e4 | 646 | |
63583cca | 647 | /* these are not supported */ |
a9d6ceb8 HR |
648 | case DATA_PROTECT: |
649 | if (sshdr.asc == 0x27 && sshdr.ascq == 0x07) { | |
650 | /* Thin provisioning hard threshold reached */ | |
7dfaae6a | 651 | set_scsi_ml_byte(scmd, SCSIML_STAT_NOSPC); |
a9d6ceb8 HR |
652 | return SUCCESS; |
653 | } | |
df561f66 | 654 | fallthrough; |
1da177e4 LT |
655 | case COPY_ABORTED: |
656 | case VOLUME_OVERFLOW: | |
657 | case MISCOMPARE: | |
63583cca | 658 | case BLANK_CHECK: |
7dfaae6a | 659 | set_scsi_ml_byte(scmd, SCSIML_STAT_TGT_FAILURE); |
87f14e65 | 660 | return SUCCESS; |
1da177e4 LT |
661 | |
662 | case MEDIUM_ERROR: | |
fd1b494d LT |
663 | if (sshdr.asc == 0x11 || /* UNRECOVERED READ ERR */ |
664 | sshdr.asc == 0x13 || /* AMNF DATA FIELD */ | |
665 | sshdr.asc == 0x14) { /* RECORD NOT FOUND */ | |
7dfaae6a | 666 | set_scsi_ml_byte(scmd, SCSIML_STAT_MED_ERROR); |
87f14e65 | 667 | return SUCCESS; |
fd1b494d | 668 | } |
1da177e4 LT |
669 | return NEEDS_RETRY; |
670 | ||
671 | case HARDWARE_ERROR: | |
672 | if (scmd->device->retry_hwerror) | |
bb0003c1 | 673 | return ADD_TO_MLQUEUE; |
1da177e4 | 674 | else |
7dfaae6a | 675 | set_scsi_ml_byte(scmd, SCSIML_STAT_TGT_FAILURE); |
df561f66 | 676 | fallthrough; |
1da177e4 LT |
677 | |
678 | case ILLEGAL_REQUEST: | |
47ac56db MS |
679 | if (sshdr.asc == 0x20 || /* Invalid command operation code */ |
680 | sshdr.asc == 0x21 || /* Logical block address out of range */ | |
a8bbb2ab | 681 | sshdr.asc == 0x22 || /* Invalid function */ |
47ac56db | 682 | sshdr.asc == 0x24 || /* Invalid field in cdb */ |
d0b7a909 MW |
683 | sshdr.asc == 0x26 || /* Parameter value invalid */ |
684 | sshdr.asc == 0x27) { /* Write protected */ | |
7dfaae6a | 685 | set_scsi_ml_byte(scmd, SCSIML_STAT_TGT_FAILURE); |
47ac56db MS |
686 | } |
687 | return SUCCESS; | |
688 | ||
1da177e4 LT |
689 | default: |
690 | return SUCCESS; | |
691 | } | |
692 | } | |
3852e373 | 693 | EXPORT_SYMBOL_GPL(scsi_check_sense); |
1da177e4 | 694 | |
4a84067d VD |
695 | static void scsi_handle_queue_ramp_up(struct scsi_device *sdev) |
696 | { | |
697 | struct scsi_host_template *sht = sdev->host->hostt; | |
698 | struct scsi_device *tmp_sdev; | |
699 | ||
c40ecc12 | 700 | if (!sht->track_queue_depth || |
4a84067d VD |
701 | sdev->queue_depth >= sdev->max_queue_depth) |
702 | return; | |
703 | ||
704 | if (time_before(jiffies, | |
705 | sdev->last_queue_ramp_up + sdev->queue_ramp_up_period)) | |
706 | return; | |
707 | ||
708 | if (time_before(jiffies, | |
709 | sdev->last_queue_full_time + sdev->queue_ramp_up_period)) | |
710 | return; | |
711 | ||
712 | /* | |
713 | * Walk all devices of a target and do | |
714 | * ramp up on them. | |
715 | */ | |
716 | shost_for_each_device(tmp_sdev, sdev->host) { | |
717 | if (tmp_sdev->channel != sdev->channel || | |
718 | tmp_sdev->id != sdev->id || | |
719 | tmp_sdev->queue_depth == sdev->max_queue_depth) | |
720 | continue; | |
c40ecc12 | 721 | |
db5ed4df | 722 | scsi_change_queue_depth(tmp_sdev, tmp_sdev->queue_depth + 1); |
4a84067d VD |
723 | sdev->last_queue_ramp_up = jiffies; |
724 | } | |
725 | } | |
726 | ||
42a6a918 MC |
727 | static void scsi_handle_queue_full(struct scsi_device *sdev) |
728 | { | |
729 | struct scsi_host_template *sht = sdev->host->hostt; | |
730 | struct scsi_device *tmp_sdev; | |
731 | ||
c40ecc12 | 732 | if (!sht->track_queue_depth) |
42a6a918 MC |
733 | return; |
734 | ||
735 | shost_for_each_device(tmp_sdev, sdev->host) { | |
736 | if (tmp_sdev->channel != sdev->channel || | |
737 | tmp_sdev->id != sdev->id) | |
738 | continue; | |
739 | /* | |
740 | * We do not know the number of commands that were at | |
741 | * the device when we got the queue full so we start | |
742 | * from the highest possible value and work our way down. | |
743 | */ | |
c40ecc12 | 744 | scsi_track_queue_full(tmp_sdev, tmp_sdev->queue_depth - 1); |
42a6a918 MC |
745 | } |
746 | } | |
747 | ||
1da177e4 LT |
748 | /** |
749 | * scsi_eh_completed_normally - Disposition a eh cmd on return from LLD. | |
750 | * @scmd: SCSI cmd to examine. | |
751 | * | |
752 | * Notes: | |
753 | * This is *only* called when we are examining the status of commands | |
754 | * queued during error recovery. the main difference here is that we | |
755 | * don't allow for the possibility of retries here, and we are a lot | |
756 | * more restrictive about what we consider acceptable. | |
dc8875e1 | 757 | */ |
b8e162f9 | 758 | static enum scsi_disposition scsi_eh_completed_normally(struct scsi_cmnd *scmd) |
1da177e4 LT |
759 | { |
760 | /* | |
761 | * first check the host byte, to see if there is anything in there | |
762 | * that would indicate what we need to do. | |
763 | */ | |
764 | if (host_byte(scmd->result) == DID_RESET) { | |
765 | /* | |
766 | * rats. we are already in the error handler, so we now | |
767 | * get to try and figure out what to do next. if the sense | |
768 | * is valid, we have a pretty good idea of what to do. | |
769 | * if not, we mark it as FAILED. | |
770 | */ | |
771 | return scsi_check_sense(scmd); | |
772 | } | |
773 | if (host_byte(scmd->result) != DID_OK) | |
774 | return FAILED; | |
775 | ||
1da177e4 LT |
776 | /* |
777 | * now, check the status byte to see if this indicates | |
778 | * anything special. | |
779 | */ | |
3d45cefc HR |
780 | switch (get_status_byte(scmd)) { |
781 | case SAM_STAT_GOOD: | |
4a84067d | 782 | scsi_handle_queue_ramp_up(scmd->device); |
df561f66 | 783 | fallthrough; |
3d45cefc | 784 | case SAM_STAT_COMMAND_TERMINATED: |
1da177e4 | 785 | return SUCCESS; |
3d45cefc | 786 | case SAM_STAT_CHECK_CONDITION: |
1da177e4 | 787 | return scsi_check_sense(scmd); |
3d45cefc HR |
788 | case SAM_STAT_CONDITION_MET: |
789 | case SAM_STAT_INTERMEDIATE: | |
790 | case SAM_STAT_INTERMEDIATE_CONDITION_MET: | |
1da177e4 LT |
791 | /* |
792 | * who knows? FIXME(eric) | |
793 | */ | |
794 | return SUCCESS; | |
3d45cefc | 795 | case SAM_STAT_RESERVATION_CONFLICT: |
67110dfd JB |
796 | if (scmd->cmnd[0] == TEST_UNIT_READY) |
797 | /* it is a success, we probed the device and | |
798 | * found it */ | |
799 | return SUCCESS; | |
800 | /* otherwise, we failed to send the command */ | |
801 | return FAILED; | |
3d45cefc | 802 | case SAM_STAT_TASK_SET_FULL: |
42a6a918 | 803 | scsi_handle_queue_full(scmd->device); |
df561f66 | 804 | fallthrough; |
3d45cefc | 805 | case SAM_STAT_BUSY: |
3eb3a928 | 806 | return NEEDS_RETRY; |
1da177e4 LT |
807 | default: |
808 | return FAILED; | |
809 | } | |
810 | return FAILED; | |
811 | } | |
812 | ||
1da177e4 LT |
813 | /** |
814 | * scsi_eh_done - Completion function for error handling. | |
815 | * @scmd: Cmd that is done. | |
dc8875e1 | 816 | */ |
bf23e619 | 817 | void scsi_eh_done(struct scsi_cmnd *scmd) |
1da177e4 | 818 | { |
0bf8c869 | 819 | struct completion *eh_action; |
85631672 | 820 | |
91921e01 | 821 | SCSI_LOG_ERROR_RECOVERY(3, scmd_printk(KERN_INFO, scmd, |
470613b4 | 822 | "%s result: %x\n", __func__, scmd->result)); |
85631672 MR |
823 | |
824 | eh_action = scmd->device->host->eh_action; | |
825 | if (eh_action) | |
826 | complete(eh_action); | |
1da177e4 LT |
827 | } |
828 | ||
292148f8 BK |
829 | /** |
830 | * scsi_try_host_reset - ask host adapter to reset itself | |
c2b3ebd0 | 831 | * @scmd: SCSI cmd to send host reset. |
dc8875e1 | 832 | */ |
b8e162f9 | 833 | static enum scsi_disposition scsi_try_host_reset(struct scsi_cmnd *scmd) |
292148f8 BK |
834 | { |
835 | unsigned long flags; | |
b8e162f9 | 836 | enum scsi_disposition rtn; |
0bf8c869 JJ |
837 | struct Scsi_Host *host = scmd->device->host; |
838 | struct scsi_host_template *hostt = host->hostt; | |
292148f8 | 839 | |
91921e01 HR |
840 | SCSI_LOG_ERROR_RECOVERY(3, |
841 | shost_printk(KERN_INFO, host, "Snd Host RST\n")); | |
292148f8 | 842 | |
0bf8c869 | 843 | if (!hostt->eh_host_reset_handler) |
292148f8 BK |
844 | return FAILED; |
845 | ||
0bf8c869 | 846 | rtn = hostt->eh_host_reset_handler(scmd); |
292148f8 BK |
847 | |
848 | if (rtn == SUCCESS) { | |
0bf8c869 | 849 | if (!hostt->skip_settle_delay) |
292148f8 | 850 | ssleep(HOST_RESET_SETTLE_TIME); |
0bf8c869 JJ |
851 | spin_lock_irqsave(host->host_lock, flags); |
852 | scsi_report_bus_reset(host, scmd_channel(scmd)); | |
853 | spin_unlock_irqrestore(host->host_lock, flags); | |
292148f8 BK |
854 | } |
855 | ||
856 | return rtn; | |
857 | } | |
858 | ||
859 | /** | |
860 | * scsi_try_bus_reset - ask host to perform a bus reset | |
861 | * @scmd: SCSI cmd to send bus reset. | |
dc8875e1 | 862 | */ |
b8e162f9 | 863 | static enum scsi_disposition scsi_try_bus_reset(struct scsi_cmnd *scmd) |
292148f8 BK |
864 | { |
865 | unsigned long flags; | |
b8e162f9 | 866 | enum scsi_disposition rtn; |
0bf8c869 JJ |
867 | struct Scsi_Host *host = scmd->device->host; |
868 | struct scsi_host_template *hostt = host->hostt; | |
292148f8 | 869 | |
91921e01 HR |
870 | SCSI_LOG_ERROR_RECOVERY(3, scmd_printk(KERN_INFO, scmd, |
871 | "%s: Snd Bus RST\n", __func__)); | |
292148f8 | 872 | |
0bf8c869 | 873 | if (!hostt->eh_bus_reset_handler) |
292148f8 BK |
874 | return FAILED; |
875 | ||
0bf8c869 | 876 | rtn = hostt->eh_bus_reset_handler(scmd); |
292148f8 BK |
877 | |
878 | if (rtn == SUCCESS) { | |
0bf8c869 | 879 | if (!hostt->skip_settle_delay) |
292148f8 | 880 | ssleep(BUS_RESET_SETTLE_TIME); |
0bf8c869 JJ |
881 | spin_lock_irqsave(host->host_lock, flags); |
882 | scsi_report_bus_reset(host, scmd_channel(scmd)); | |
883 | spin_unlock_irqrestore(host->host_lock, flags); | |
292148f8 BK |
884 | } |
885 | ||
886 | return rtn; | |
887 | } | |
888 | ||
30bd7df8 MC |
889 | static void __scsi_report_device_reset(struct scsi_device *sdev, void *data) |
890 | { | |
891 | sdev->was_reset = 1; | |
892 | sdev->expecting_cc_ua = 1; | |
893 | } | |
894 | ||
895 | /** | |
896 | * scsi_try_target_reset - Ask host to perform a target reset | |
897 | * @scmd: SCSI cmd used to send a target reset | |
898 | * | |
899 | * Notes: | |
900 | * There is no timeout for this operation. if this operation is | |
901 | * unreliable for a given host, then the host itself needs to put a | |
902 | * timer on it, and set the host back to a consistent state prior to | |
903 | * returning. | |
904 | */ | |
b8e162f9 | 905 | static enum scsi_disposition scsi_try_target_reset(struct scsi_cmnd *scmd) |
30bd7df8 MC |
906 | { |
907 | unsigned long flags; | |
b8e162f9 | 908 | enum scsi_disposition rtn; |
0bf8c869 JJ |
909 | struct Scsi_Host *host = scmd->device->host; |
910 | struct scsi_host_template *hostt = host->hostt; | |
30bd7df8 | 911 | |
0bf8c869 | 912 | if (!hostt->eh_target_reset_handler) |
30bd7df8 MC |
913 | return FAILED; |
914 | ||
0bf8c869 | 915 | rtn = hostt->eh_target_reset_handler(scmd); |
30bd7df8 | 916 | if (rtn == SUCCESS) { |
0bf8c869 | 917 | spin_lock_irqsave(host->host_lock, flags); |
30bd7df8 MC |
918 | __starget_for_each_device(scsi_target(scmd->device), NULL, |
919 | __scsi_report_device_reset); | |
0bf8c869 | 920 | spin_unlock_irqrestore(host->host_lock, flags); |
30bd7df8 MC |
921 | } |
922 | ||
923 | return rtn; | |
924 | } | |
925 | ||
292148f8 BK |
926 | /** |
927 | * scsi_try_bus_device_reset - Ask host to perform a BDR on a dev | |
928 | * @scmd: SCSI cmd used to send BDR | |
929 | * | |
930 | * Notes: | |
931 | * There is no timeout for this operation. if this operation is | |
932 | * unreliable for a given host, then the host itself needs to put a | |
933 | * timer on it, and set the host back to a consistent state prior to | |
934 | * returning. | |
dc8875e1 | 935 | */ |
b8e162f9 | 936 | static enum scsi_disposition scsi_try_bus_device_reset(struct scsi_cmnd *scmd) |
292148f8 | 937 | { |
b8e162f9 | 938 | enum scsi_disposition rtn; |
0bf8c869 | 939 | struct scsi_host_template *hostt = scmd->device->host->hostt; |
292148f8 | 940 | |
0bf8c869 | 941 | if (!hostt->eh_device_reset_handler) |
292148f8 BK |
942 | return FAILED; |
943 | ||
0bf8c869 | 944 | rtn = hostt->eh_device_reset_handler(scmd); |
30bd7df8 MC |
945 | if (rtn == SUCCESS) |
946 | __scsi_report_device_reset(scmd->device, NULL); | |
292148f8 BK |
947 | return rtn; |
948 | } | |
949 | ||
883a030f HR |
950 | /** |
951 | * scsi_try_to_abort_cmd - Ask host to abort a SCSI command | |
6583f6fb | 952 | * @hostt: SCSI driver host template |
883a030f HR |
953 | * @scmd: SCSI cmd used to send a target reset |
954 | * | |
955 | * Return value: | |
956 | * SUCCESS, FAILED, or FAST_IO_FAIL | |
957 | * | |
958 | * Notes: | |
959 | * SUCCESS does not necessarily indicate that the command | |
960 | * has been aborted; it only indicates that the LLDDs | |
961 | * has cleared all references to that command. | |
962 | * LLDDs should return FAILED only if an abort was required | |
963 | * but could not be executed. LLDDs should return FAST_IO_FAIL | |
964 | * if the device is temporarily unavailable (eg due to a | |
965 | * link down on FibreChannel) | |
966 | */ | |
b8e162f9 BVA |
967 | static enum scsi_disposition |
968 | scsi_try_to_abort_cmd(struct scsi_host_template *hostt, struct scsi_cmnd *scmd) | |
292148f8 | 969 | { |
0bf8c869 | 970 | if (!hostt->eh_abort_handler) |
292148f8 BK |
971 | return FAILED; |
972 | ||
0bf8c869 | 973 | return hostt->eh_abort_handler(scmd); |
292148f8 BK |
974 | } |
975 | ||
292148f8 BK |
976 | static void scsi_abort_eh_cmnd(struct scsi_cmnd *scmd) |
977 | { | |
0bf8c869 | 978 | if (scsi_try_to_abort_cmd(scmd->device->host->hostt, scmd) != SUCCESS) |
292148f8 | 979 | if (scsi_try_bus_device_reset(scmd) != SUCCESS) |
30bd7df8 MC |
980 | if (scsi_try_target_reset(scmd) != SUCCESS) |
981 | if (scsi_try_bus_reset(scmd) != SUCCESS) | |
982 | scsi_try_host_reset(scmd); | |
292148f8 BK |
983 | } |
984 | ||
1da177e4 | 985 | /** |
3b729f76 | 986 | * scsi_eh_prep_cmnd - Save a scsi command info as part of error recovery |
2dc611de | 987 | * @scmd: SCSI command structure to hijack |
e1c23468 | 988 | * @ses: structure to save restore information |
55db6c1b | 989 | * @cmnd: CDB to send. Can be NULL if no new cmnd is needed |
ce70fd9a | 990 | * @cmnd_size: size in bytes of @cmnd (must be <= MAX_COMMAND_SIZE) |
55db6c1b | 991 | * @sense_bytes: size of sense data to copy. or 0 (if != 0 @cmnd is ignored) |
2dc611de | 992 | * |
e1c23468 | 993 | * This function is used to save a scsi command information before re-execution |
55db6c1b BH |
994 | * as part of the error recovery process. If @sense_bytes is 0 the command |
995 | * sent must be one that does not transfer any data. If @sense_bytes != 0 | |
996 | * @cmnd is ignored and this functions sets up a REQUEST_SENSE command | |
997 | * and cmnd buffers to read @sense_bytes into @scmd->sense_buffer. | |
dc8875e1 | 998 | */ |
e1c23468 BH |
999 | void scsi_eh_prep_cmnd(struct scsi_cmnd *scmd, struct scsi_eh_save *ses, |
1000 | unsigned char *cmnd, int cmnd_size, unsigned sense_bytes) | |
1da177e4 | 1001 | { |
f59114b7 | 1002 | struct scsi_device *sdev = scmd->device; |
1da177e4 | 1003 | |
631c228c CH |
1004 | /* |
1005 | * We need saved copies of a number of fields - this is because | |
1006 | * error handling may need to overwrite these with different values | |
1007 | * to run different commands, and once error handling is complete, | |
1008 | * we will need to restore these values prior to running the actual | |
1009 | * command. | |
1010 | */ | |
e1c23468 | 1011 | ses->cmd_len = scmd->cmd_len; |
e1c23468 | 1012 | ses->data_direction = scmd->sc_data_direction; |
30b0c37b | 1013 | ses->sdb = scmd->sdb; |
e1c23468 | 1014 | ses->result = scmd->result; |
a9a4ea11 | 1015 | ses->resid_len = scmd->resid_len; |
12265709 | 1016 | ses->underflow = scmd->underflow; |
db007fc5 | 1017 | ses->prot_op = scmd->prot_op; |
8e8c9d01 | 1018 | ses->eh_eflags = scmd->eh_eflags; |
631c228c | 1019 | |
db007fc5 | 1020 | scmd->prot_op = SCSI_PROT_NORMAL; |
c69e6f81 | 1021 | scmd->eh_eflags = 0; |
ce70fd9a CH |
1022 | memcpy(ses->cmnd, scmd->cmnd, sizeof(ses->cmnd)); |
1023 | memset(scmd->cmnd, 0, sizeof(scmd->cmnd)); | |
30b0c37b | 1024 | memset(&scmd->sdb, 0, sizeof(scmd->sdb)); |
644373a4 | 1025 | scmd->result = 0; |
a9a4ea11 | 1026 | scmd->resid_len = 0; |
30b0c37b | 1027 | |
55db6c1b | 1028 | if (sense_bytes) { |
30b0c37b BH |
1029 | scmd->sdb.length = min_t(unsigned, SCSI_SENSE_BUFFERSIZE, |
1030 | sense_bytes); | |
e1c23468 | 1031 | sg_init_one(&ses->sense_sgl, scmd->sense_buffer, |
30b0c37b BH |
1032 | scmd->sdb.length); |
1033 | scmd->sdb.table.sgl = &ses->sense_sgl; | |
55db6c1b | 1034 | scmd->sc_data_direction = DMA_FROM_DEVICE; |
0c958ecc | 1035 | scmd->sdb.table.nents = scmd->sdb.table.orig_nents = 1; |
55db6c1b | 1036 | scmd->cmnd[0] = REQUEST_SENSE; |
30b0c37b | 1037 | scmd->cmnd[4] = scmd->sdb.length; |
55db6c1b | 1038 | scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]); |
631c228c | 1039 | } else { |
631c228c | 1040 | scmd->sc_data_direction = DMA_NONE; |
55db6c1b | 1041 | if (cmnd) { |
ce70fd9a | 1042 | BUG_ON(cmnd_size > sizeof(scmd->cmnd)); |
55db6c1b BH |
1043 | memcpy(scmd->cmnd, cmnd, cmnd_size); |
1044 | scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]); | |
1045 | } | |
631c228c CH |
1046 | } |
1047 | ||
1048 | scmd->underflow = 0; | |
631c228c | 1049 | |
55db6c1b | 1050 | if (sdev->scsi_level <= SCSI_2 && sdev->scsi_level != SCSI_UNKNOWN) |
1da177e4 | 1051 | scmd->cmnd[1] = (scmd->cmnd[1] & 0x1f) | |
f59114b7 | 1052 | (sdev->lun << 5 & 0xe0); |
1da177e4 | 1053 | |
631c228c CH |
1054 | /* |
1055 | * Zero the sense buffer. The scsi spec mandates that any | |
1056 | * untransferred sense data should be interpreted as being zero. | |
1057 | */ | |
b80ca4f7 | 1058 | memset(scmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE); |
e1c23468 BH |
1059 | } |
1060 | EXPORT_SYMBOL(scsi_eh_prep_cmnd); | |
1061 | ||
1062 | /** | |
3b729f76 | 1063 | * scsi_eh_restore_cmnd - Restore a scsi command info as part of error recovery |
e1c23468 | 1064 | * @scmd: SCSI command structure to restore |
477e608c | 1065 | * @ses: saved information from a coresponding call to scsi_eh_prep_cmnd |
e1c23468 | 1066 | * |
477e608c | 1067 | * Undo any damage done by above scsi_eh_prep_cmnd(). |
dc8875e1 | 1068 | */ |
e1c23468 BH |
1069 | void scsi_eh_restore_cmnd(struct scsi_cmnd* scmd, struct scsi_eh_save *ses) |
1070 | { | |
1071 | /* | |
1072 | * Restore original data | |
1073 | */ | |
1074 | scmd->cmd_len = ses->cmd_len; | |
ce70fd9a | 1075 | memcpy(scmd->cmnd, ses->cmnd, sizeof(ses->cmnd)); |
e1c23468 | 1076 | scmd->sc_data_direction = ses->data_direction; |
30b0c37b | 1077 | scmd->sdb = ses->sdb; |
e1c23468 | 1078 | scmd->result = ses->result; |
a9a4ea11 | 1079 | scmd->resid_len = ses->resid_len; |
12265709 | 1080 | scmd->underflow = ses->underflow; |
db007fc5 | 1081 | scmd->prot_op = ses->prot_op; |
8e8c9d01 | 1082 | scmd->eh_eflags = ses->eh_eflags; |
e1c23468 BH |
1083 | } |
1084 | EXPORT_SYMBOL(scsi_eh_restore_cmnd); | |
631c228c | 1085 | |
e1c23468 | 1086 | /** |
3b729f76 | 1087 | * scsi_send_eh_cmnd - submit a scsi command as part of error recovery |
e1c23468 BH |
1088 | * @scmd: SCSI command structure to hijack |
1089 | * @cmnd: CDB to send | |
1090 | * @cmnd_size: size in bytes of @cmnd | |
1091 | * @timeout: timeout for this request | |
1092 | * @sense_bytes: size of sense data to copy or 0 | |
1093 | * | |
1094 | * This function is used to send a scsi command down to a target device | |
1095 | * as part of the error recovery process. See also scsi_eh_prep_cmnd() above. | |
1096 | * | |
1097 | * Return value: | |
1098 | * SUCCESS or FAILED or NEEDS_RETRY | |
dc8875e1 | 1099 | */ |
b8e162f9 BVA |
1100 | static enum scsi_disposition scsi_send_eh_cmnd(struct scsi_cmnd *scmd, |
1101 | unsigned char *cmnd, int cmnd_size, int timeout, unsigned sense_bytes) | |
e1c23468 BH |
1102 | { |
1103 | struct scsi_device *sdev = scmd->device; | |
1104 | struct Scsi_Host *shost = sdev->host; | |
1105 | DECLARE_COMPLETION_ONSTACK(done); | |
bbe9fb0d | 1106 | unsigned long timeleft = timeout, delay; |
e1c23468 | 1107 | struct scsi_eh_save ses; |
fc73648a | 1108 | const unsigned long stall_for = msecs_to_jiffies(100); |
e1c23468 BH |
1109 | int rtn; |
1110 | ||
fc73648a | 1111 | retry: |
e1c23468 | 1112 | scsi_eh_prep_cmnd(scmd, &ses, cmnd, cmnd_size, sense_bytes); |
7dfdc9a5 | 1113 | shost->eh_action = &done; |
1da177e4 | 1114 | |
1da177e4 | 1115 | scsi_log_send(scmd); |
bf23e619 | 1116 | scmd->submitter = SUBMITTED_BY_SCSI_ERROR_HANDLER; |
bbe9fb0d BVA |
1117 | |
1118 | /* | |
1119 | * Lock sdev->state_mutex to avoid that scsi_device_quiesce() can | |
1120 | * change the SCSI device state after we have examined it and before | |
1121 | * .queuecommand() is called. | |
1122 | */ | |
1123 | mutex_lock(&sdev->state_mutex); | |
1124 | while (sdev->sdev_state == SDEV_BLOCK && timeleft > 0) { | |
1125 | mutex_unlock(&sdev->state_mutex); | |
1126 | SCSI_LOG_ERROR_RECOVERY(5, sdev_printk(KERN_DEBUG, sdev, | |
1127 | "%s: state %d <> %d\n", __func__, sdev->sdev_state, | |
1128 | SDEV_BLOCK)); | |
1129 | delay = min(timeleft, stall_for); | |
1130 | timeleft -= delay; | |
1131 | msleep(jiffies_to_msecs(delay)); | |
1132 | mutex_lock(&sdev->state_mutex); | |
1133 | } | |
1134 | if (sdev->sdev_state != SDEV_BLOCK) | |
1135 | rtn = shost->hostt->queuecommand(shost, scmd); | |
1136 | else | |
280e91b0 | 1137 | rtn = FAILED; |
bbe9fb0d BVA |
1138 | mutex_unlock(&sdev->state_mutex); |
1139 | ||
fc73648a HR |
1140 | if (rtn) { |
1141 | if (timeleft > stall_for) { | |
1142 | scsi_eh_restore_cmnd(scmd, &ses); | |
bf23e619 | 1143 | |
fc73648a HR |
1144 | timeleft -= stall_for; |
1145 | msleep(jiffies_to_msecs(stall_for)); | |
1146 | goto retry; | |
1147 | } | |
1148 | /* signal not to enter either branch of the if () below */ | |
1149 | timeleft = 0; | |
511833ac | 1150 | rtn = FAILED; |
fc73648a HR |
1151 | } else { |
1152 | timeleft = wait_for_completion_timeout(&done, timeout); | |
ac61d195 | 1153 | rtn = SUCCESS; |
fc73648a | 1154 | } |
1da177e4 | 1155 | |
f59114b7 | 1156 | shost->eh_action = NULL; |
1da177e4 | 1157 | |
fc73648a | 1158 | scsi_log_completion(scmd, rtn); |
1da177e4 | 1159 | |
91921e01 | 1160 | SCSI_LOG_ERROR_RECOVERY(3, scmd_printk(KERN_INFO, scmd, |
470613b4 HR |
1161 | "%s timeleft: %ld\n", |
1162 | __func__, timeleft)); | |
1da177e4 LT |
1163 | |
1164 | /* | |
fc73648a HR |
1165 | * If there is time left scsi_eh_done got called, and we will examine |
1166 | * the actual status codes to see whether the command actually did | |
1167 | * complete normally, else if we have a zero return and no time left, | |
1168 | * the command must still be pending, so abort it and return FAILED. | |
1169 | * If we never actually managed to issue the command, because | |
1170 | * ->queuecommand() kept returning non zero, use the rtn = FAILED | |
1171 | * value above (so don't execute either branch of the if) | |
1da177e4 | 1172 | */ |
7dfdc9a5 | 1173 | if (timeleft) { |
1da177e4 | 1174 | rtn = scsi_eh_completed_normally(scmd); |
91921e01 HR |
1175 | SCSI_LOG_ERROR_RECOVERY(3, scmd_printk(KERN_INFO, scmd, |
1176 | "%s: scsi_eh_completed_normally %x\n", __func__, rtn)); | |
7dfdc9a5 | 1177 | |
1da177e4 LT |
1178 | switch (rtn) { |
1179 | case SUCCESS: | |
1180 | case NEEDS_RETRY: | |
1181 | case FAILED: | |
1182 | break; | |
6e883b0e HR |
1183 | case ADD_TO_MLQUEUE: |
1184 | rtn = NEEDS_RETRY; | |
1185 | break; | |
1da177e4 LT |
1186 | default: |
1187 | rtn = FAILED; | |
1188 | break; | |
1189 | } | |
511833ac | 1190 | } else if (rtn != FAILED) { |
292148f8 | 1191 | scsi_abort_eh_cmnd(scmd); |
7dfdc9a5 | 1192 | rtn = FAILED; |
1da177e4 LT |
1193 | } |
1194 | ||
e1c23468 | 1195 | scsi_eh_restore_cmnd(scmd, &ses); |
18a4d0a2 | 1196 | |
1da177e4 LT |
1197 | return rtn; |
1198 | } | |
1199 | ||
1200 | /** | |
1201 | * scsi_request_sense - Request sense data from a particular target. | |
1202 | * @scmd: SCSI cmd for request sense. | |
1203 | * | |
1204 | * Notes: | |
1205 | * Some hosts automatically obtain this information, others require | |
1206 | * that we obtain it on our own. This function will *not* return until | |
1207 | * the command either times out, or it completes. | |
dc8875e1 | 1208 | */ |
b8e162f9 | 1209 | static enum scsi_disposition scsi_request_sense(struct scsi_cmnd *scmd) |
1da177e4 | 1210 | { |
0816c925 | 1211 | return scsi_send_eh_cmnd(scmd, NULL, 0, scmd->device->eh_timeout, ~0); |
1da177e4 LT |
1212 | } |
1213 | ||
b8e162f9 BVA |
1214 | static enum scsi_disposition |
1215 | scsi_eh_action(struct scsi_cmnd *scmd, enum scsi_disposition rtn) | |
2451079b | 1216 | { |
aa8e25e5 | 1217 | if (!blk_rq_is_passthrough(scsi_cmd_to_rq(scmd))) { |
2451079b JB |
1218 | struct scsi_driver *sdrv = scsi_cmd_to_driver(scmd); |
1219 | if (sdrv->eh_action) | |
1220 | rtn = sdrv->eh_action(scmd, rtn); | |
1221 | } | |
1222 | return rtn; | |
1223 | } | |
1224 | ||
1da177e4 LT |
1225 | /** |
1226 | * scsi_eh_finish_cmd - Handle a cmd that eh is finished with. | |
1227 | * @scmd: Original SCSI cmd that eh has finished. | |
1228 | * @done_q: Queue for processed commands. | |
1229 | * | |
1230 | * Notes: | |
1231 | * We don't want to use the normal command completion while we are are | |
1232 | * still handling errors - it may cause other commands to be queued, | |
eb44820c | 1233 | * and that would disturb what we are doing. Thus we really want to |
1da177e4 LT |
1234 | * keep a list of pending commands for final completion, and once we |
1235 | * are ready to leave error handling we handle completion for real. | |
dc8875e1 | 1236 | */ |
041c5fc3 | 1237 | void scsi_eh_finish_cmd(struct scsi_cmnd *scmd, struct list_head *done_q) |
1da177e4 | 1238 | { |
1da177e4 LT |
1239 | list_move_tail(&scmd->eh_entry, done_q); |
1240 | } | |
041c5fc3 | 1241 | EXPORT_SYMBOL(scsi_eh_finish_cmd); |
1da177e4 LT |
1242 | |
1243 | /** | |
1244 | * scsi_eh_get_sense - Get device sense data. | |
1245 | * @work_q: Queue of commands to process. | |
eb44820c | 1246 | * @done_q: Queue of processed commands. |
1da177e4 LT |
1247 | * |
1248 | * Description: | |
1249 | * See if we need to request sense information. if so, then get it | |
0bf8c869 | 1250 | * now, so we have a better idea of what to do. |
1da177e4 LT |
1251 | * |
1252 | * Notes: | |
1253 | * This has the unfortunate side effect that if a shost adapter does | |
eb44820c | 1254 | * not automatically request sense information, we end up shutting |
1da177e4 LT |
1255 | * it down before we request it. |
1256 | * | |
1257 | * All drivers should request sense information internally these days, | |
1258 | * so for now all I have to say is tough noogies if you end up in here. | |
1259 | * | |
1260 | * XXX: Long term this code should go away, but that needs an audit of | |
1261 | * all LLDDs first. | |
dc8875e1 | 1262 | */ |
dca84e46 DW |
1263 | int scsi_eh_get_sense(struct list_head *work_q, |
1264 | struct list_head *done_q) | |
1da177e4 | 1265 | { |
937abeaa | 1266 | struct scsi_cmnd *scmd, *next; |
b4562022 | 1267 | struct Scsi_Host *shost; |
b8e162f9 | 1268 | enum scsi_disposition rtn; |
1da177e4 | 1269 | |
709c75b5 | 1270 | /* |
1271 | * If SCSI_EH_ABORT_SCHEDULED has been set, it is timeout IO, | |
1272 | * should not get sense. | |
1273 | */ | |
937abeaa | 1274 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
a0658632 | 1275 | if ((scmd->eh_eflags & SCSI_EH_ABORT_SCHEDULED) || |
1da177e4 LT |
1276 | SCSI_SENSE_VALID(scmd)) |
1277 | continue; | |
1278 | ||
b4562022 | 1279 | shost = scmd->device->host; |
b4562022 | 1280 | if (scsi_host_eh_past_deadline(shost)) { |
b4562022 | 1281 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1282 | scmd_printk(KERN_INFO, scmd, |
1283 | "%s: skip request sense, past eh deadline\n", | |
1284 | current->comm)); | |
b4562022 HR |
1285 | break; |
1286 | } | |
d0672a03 | 1287 | if (!scsi_status_is_check_condition(scmd->result)) |
d555a2ab JB |
1288 | /* |
1289 | * don't request sense if there's no check condition | |
1290 | * status because the error we're processing isn't one | |
1291 | * that has a sense code (and some devices get | |
1292 | * confused by sense requests out of the blue) | |
1293 | */ | |
1294 | continue; | |
1295 | ||
3bf743e7 JG |
1296 | SCSI_LOG_ERROR_RECOVERY(2, scmd_printk(KERN_INFO, scmd, |
1297 | "%s: requesting sense\n", | |
1298 | current->comm)); | |
1da177e4 LT |
1299 | rtn = scsi_request_sense(scmd); |
1300 | if (rtn != SUCCESS) | |
1301 | continue; | |
1302 | ||
91921e01 | 1303 | SCSI_LOG_ERROR_RECOVERY(3, scmd_printk(KERN_INFO, scmd, |
470613b4 | 1304 | "sense requested, result %x\n", scmd->result)); |
d811b848 | 1305 | SCSI_LOG_ERROR_RECOVERY(3, scsi_print_sense(scmd)); |
1da177e4 LT |
1306 | |
1307 | rtn = scsi_decide_disposition(scmd); | |
1308 | ||
1309 | /* | |
1310 | * if the result was normal, then just pass it along to the | |
1311 | * upper level. | |
1312 | */ | |
1313 | if (rtn == SUCCESS) | |
2a242d59 MC |
1314 | /* |
1315 | * We don't want this command reissued, just finished | |
1316 | * with the sense data, so set retries to the max | |
1317 | * allowed to ensure it won't get reissued. If the user | |
1318 | * has requested infinite retries, we also want to | |
1319 | * finish this command, so force completion by setting | |
1320 | * retries and allowed to the same value. | |
1321 | */ | |
1322 | if (scmd->allowed == SCSI_CMD_RETRIES_NO_LIMIT) | |
1323 | scmd->retries = scmd->allowed = 1; | |
1324 | else | |
1325 | scmd->retries = scmd->allowed; | |
1da177e4 LT |
1326 | else if (rtn != NEEDS_RETRY) |
1327 | continue; | |
1328 | ||
1329 | scsi_eh_finish_cmd(scmd, done_q); | |
1330 | } | |
1331 | ||
1332 | return list_empty(work_q); | |
1333 | } | |
dca84e46 | 1334 | EXPORT_SYMBOL_GPL(scsi_eh_get_sense); |
1da177e4 | 1335 | |
1da177e4 LT |
1336 | /** |
1337 | * scsi_eh_tur - Send TUR to device. | |
eb44820c | 1338 | * @scmd: &scsi_cmnd to send TUR |
1da177e4 LT |
1339 | * |
1340 | * Return value: | |
1341 | * 0 - Device is ready. 1 - Device NOT ready. | |
dc8875e1 | 1342 | */ |
1da177e4 LT |
1343 | static int scsi_eh_tur(struct scsi_cmnd *scmd) |
1344 | { | |
1345 | static unsigned char tur_command[6] = {TEST_UNIT_READY, 0, 0, 0, 0, 0}; | |
b8e162f9 BVA |
1346 | int retry_cnt = 1; |
1347 | enum scsi_disposition rtn; | |
1da177e4 LT |
1348 | |
1349 | retry_tur: | |
0816c925 MP |
1350 | rtn = scsi_send_eh_cmnd(scmd, tur_command, 6, |
1351 | scmd->device->eh_timeout, 0); | |
1da177e4 | 1352 | |
91921e01 | 1353 | SCSI_LOG_ERROR_RECOVERY(3, scmd_printk(KERN_INFO, scmd, |
470613b4 | 1354 | "%s return: %x\n", __func__, rtn)); |
631c228c CH |
1355 | |
1356 | switch (rtn) { | |
1357 | case NEEDS_RETRY: | |
1da177e4 LT |
1358 | if (retry_cnt--) |
1359 | goto retry_tur; | |
df561f66 | 1360 | fallthrough; |
631c228c | 1361 | case SUCCESS: |
e47373ec | 1362 | return 0; |
631c228c CH |
1363 | default: |
1364 | return 1; | |
e47373ec | 1365 | } |
1da177e4 LT |
1366 | } |
1367 | ||
3eef6257 DJ |
1368 | /** |
1369 | * scsi_eh_test_devices - check if devices are responding from error recovery. | |
1370 | * @cmd_list: scsi commands in error recovery. | |
74cf298f RD |
1371 | * @work_q: queue for commands which still need more error recovery |
1372 | * @done_q: queue for commands which are finished | |
1373 | * @try_stu: boolean on if a STU command should be tried in addition to TUR. | |
3eef6257 DJ |
1374 | * |
1375 | * Decription: | |
1376 | * Tests if devices are in a working state. Commands to devices now in | |
1377 | * a working state are sent to the done_q while commands to devices which | |
1378 | * are still failing to respond are returned to the work_q for more | |
1379 | * processing. | |
1380 | **/ | |
1381 | static int scsi_eh_test_devices(struct list_head *cmd_list, | |
1382 | struct list_head *work_q, | |
1383 | struct list_head *done_q, int try_stu) | |
1384 | { | |
1385 | struct scsi_cmnd *scmd, *next; | |
1386 | struct scsi_device *sdev; | |
1387 | int finish_cmds; | |
1388 | ||
1389 | while (!list_empty(cmd_list)) { | |
1390 | scmd = list_entry(cmd_list->next, struct scsi_cmnd, eh_entry); | |
1391 | sdev = scmd->device; | |
1392 | ||
b4562022 | 1393 | if (!try_stu) { |
b4562022 HR |
1394 | if (scsi_host_eh_past_deadline(sdev->host)) { |
1395 | /* Push items back onto work_q */ | |
1396 | list_splice_init(cmd_list, work_q); | |
b4562022 | 1397 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1398 | sdev_printk(KERN_INFO, sdev, |
1399 | "%s: skip test device, past eh deadline", | |
1400 | current->comm)); | |
b4562022 HR |
1401 | break; |
1402 | } | |
b4562022 HR |
1403 | } |
1404 | ||
3eef6257 DJ |
1405 | finish_cmds = !scsi_device_online(scmd->device) || |
1406 | (try_stu && !scsi_eh_try_stu(scmd) && | |
1407 | !scsi_eh_tur(scmd)) || | |
1408 | !scsi_eh_tur(scmd); | |
1409 | ||
1410 | list_for_each_entry_safe(scmd, next, cmd_list, eh_entry) | |
1411 | if (scmd->device == sdev) { | |
2451079b JB |
1412 | if (finish_cmds && |
1413 | (try_stu || | |
1414 | scsi_eh_action(scmd, SUCCESS) == SUCCESS)) | |
3eef6257 DJ |
1415 | scsi_eh_finish_cmd(scmd, done_q); |
1416 | else | |
1417 | list_move_tail(&scmd->eh_entry, work_q); | |
1418 | } | |
1419 | } | |
1420 | return list_empty(work_q); | |
1421 | } | |
1422 | ||
1da177e4 LT |
1423 | /** |
1424 | * scsi_eh_try_stu - Send START_UNIT to device. | |
eb44820c | 1425 | * @scmd: &scsi_cmnd to send START_UNIT |
1da177e4 LT |
1426 | * |
1427 | * Return value: | |
1428 | * 0 - Device is ready. 1 - Device NOT ready. | |
dc8875e1 | 1429 | */ |
1da177e4 LT |
1430 | static int scsi_eh_try_stu(struct scsi_cmnd *scmd) |
1431 | { | |
1432 | static unsigned char stu_command[6] = {START_STOP, 0, 0, 0, 1, 0}; | |
1da177e4 | 1433 | |
631c228c | 1434 | if (scmd->device->allow_restart) { |
b8e162f9 BVA |
1435 | int i; |
1436 | enum scsi_disposition rtn = NEEDS_RETRY; | |
ed773e66 BK |
1437 | |
1438 | for (i = 0; rtn == NEEDS_RETRY && i < 2; i++) | |
adcc796b CX |
1439 | rtn = scsi_send_eh_cmnd(scmd, stu_command, 6, |
1440 | scmd->device->eh_timeout, 0); | |
1da177e4 | 1441 | |
631c228c CH |
1442 | if (rtn == SUCCESS) |
1443 | return 0; | |
1444 | } | |
1da177e4 | 1445 | |
1da177e4 LT |
1446 | return 1; |
1447 | } | |
1448 | ||
1449 | /** | |
1450 | * scsi_eh_stu - send START_UNIT if needed | |
eb44820c | 1451 | * @shost: &scsi host being recovered. |
74cf298f | 1452 | * @work_q: &list_head for pending commands. |
eb44820c | 1453 | * @done_q: &list_head for processed commands. |
1da177e4 LT |
1454 | * |
1455 | * Notes: | |
1456 | * If commands are failing due to not ready, initializing command required, | |
0bf8c869 | 1457 | * try revalidating the device, which will end up sending a start unit. |
dc8875e1 | 1458 | */ |
1da177e4 LT |
1459 | static int scsi_eh_stu(struct Scsi_Host *shost, |
1460 | struct list_head *work_q, | |
1461 | struct list_head *done_q) | |
1462 | { | |
937abeaa | 1463 | struct scsi_cmnd *scmd, *stu_scmd, *next; |
1da177e4 LT |
1464 | struct scsi_device *sdev; |
1465 | ||
1466 | shost_for_each_device(sdev, shost) { | |
b4562022 | 1467 | if (scsi_host_eh_past_deadline(shost)) { |
b4562022 | 1468 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1469 | sdev_printk(KERN_INFO, sdev, |
1470 | "%s: skip START_UNIT, past eh deadline\n", | |
1471 | current->comm)); | |
4dea170f | 1472 | scsi_device_put(sdev); |
b4562022 HR |
1473 | break; |
1474 | } | |
1da177e4 LT |
1475 | stu_scmd = NULL; |
1476 | list_for_each_entry(scmd, work_q, eh_entry) | |
1477 | if (scmd->device == sdev && SCSI_SENSE_VALID(scmd) && | |
1478 | scsi_check_sense(scmd) == FAILED ) { | |
1479 | stu_scmd = scmd; | |
1480 | break; | |
1481 | } | |
1482 | ||
1483 | if (!stu_scmd) | |
1484 | continue; | |
1485 | ||
91921e01 | 1486 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1487 | sdev_printk(KERN_INFO, sdev, |
1488 | "%s: Sending START_UNIT\n", | |
1489 | current->comm)); | |
1da177e4 LT |
1490 | |
1491 | if (!scsi_eh_try_stu(stu_scmd)) { | |
1492 | if (!scsi_device_online(sdev) || | |
1493 | !scsi_eh_tur(stu_scmd)) { | |
937abeaa CH |
1494 | list_for_each_entry_safe(scmd, next, |
1495 | work_q, eh_entry) { | |
2451079b JB |
1496 | if (scmd->device == sdev && |
1497 | scsi_eh_action(scmd, SUCCESS) == SUCCESS) | |
1da177e4 LT |
1498 | scsi_eh_finish_cmd(scmd, done_q); |
1499 | } | |
1500 | } | |
1501 | } else { | |
1502 | SCSI_LOG_ERROR_RECOVERY(3, | |
a222b1e2 HR |
1503 | sdev_printk(KERN_INFO, sdev, |
1504 | "%s: START_UNIT failed\n", | |
1505 | current->comm)); | |
1da177e4 LT |
1506 | } |
1507 | } | |
1508 | ||
1509 | return list_empty(work_q); | |
1510 | } | |
1511 | ||
1512 | ||
1513 | /** | |
1514 | * scsi_eh_bus_device_reset - send bdr if needed | |
1515 | * @shost: scsi host being recovered. | |
74cf298f | 1516 | * @work_q: &list_head for pending commands. |
eb44820c | 1517 | * @done_q: &list_head for processed commands. |
1da177e4 LT |
1518 | * |
1519 | * Notes: | |
eb44820c | 1520 | * Try a bus device reset. Still, look to see whether we have multiple |
1da177e4 LT |
1521 | * devices that are jammed or not - if we have multiple devices, it |
1522 | * makes no sense to try bus_device_reset - we really would need to try | |
0bf8c869 | 1523 | * a bus_reset instead. |
dc8875e1 | 1524 | */ |
1da177e4 LT |
1525 | static int scsi_eh_bus_device_reset(struct Scsi_Host *shost, |
1526 | struct list_head *work_q, | |
1527 | struct list_head *done_q) | |
1528 | { | |
937abeaa | 1529 | struct scsi_cmnd *scmd, *bdr_scmd, *next; |
1da177e4 | 1530 | struct scsi_device *sdev; |
b8e162f9 | 1531 | enum scsi_disposition rtn; |
1da177e4 LT |
1532 | |
1533 | shost_for_each_device(sdev, shost) { | |
b4562022 | 1534 | if (scsi_host_eh_past_deadline(shost)) { |
b4562022 | 1535 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1536 | sdev_printk(KERN_INFO, sdev, |
1537 | "%s: skip BDR, past eh deadline\n", | |
1538 | current->comm)); | |
4dea170f | 1539 | scsi_device_put(sdev); |
b4562022 HR |
1540 | break; |
1541 | } | |
1da177e4 LT |
1542 | bdr_scmd = NULL; |
1543 | list_for_each_entry(scmd, work_q, eh_entry) | |
1544 | if (scmd->device == sdev) { | |
1545 | bdr_scmd = scmd; | |
1546 | break; | |
1547 | } | |
1548 | ||
1549 | if (!bdr_scmd) | |
1550 | continue; | |
1551 | ||
91921e01 | 1552 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1553 | sdev_printk(KERN_INFO, sdev, |
1554 | "%s: Sending BDR\n", current->comm)); | |
1da177e4 | 1555 | rtn = scsi_try_bus_device_reset(bdr_scmd); |
2f2eb587 | 1556 | if (rtn == SUCCESS || rtn == FAST_IO_FAIL) { |
1da177e4 | 1557 | if (!scsi_device_online(sdev) || |
2f2eb587 | 1558 | rtn == FAST_IO_FAIL || |
1da177e4 | 1559 | !scsi_eh_tur(bdr_scmd)) { |
937abeaa CH |
1560 | list_for_each_entry_safe(scmd, next, |
1561 | work_q, eh_entry) { | |
2451079b JB |
1562 | if (scmd->device == sdev && |
1563 | scsi_eh_action(scmd, rtn) != FAILED) | |
1da177e4 LT |
1564 | scsi_eh_finish_cmd(scmd, |
1565 | done_q); | |
1566 | } | |
1567 | } | |
1568 | } else { | |
91921e01 | 1569 | SCSI_LOG_ERROR_RECOVERY(3, |
a222b1e2 HR |
1570 | sdev_printk(KERN_INFO, sdev, |
1571 | "%s: BDR failed\n", current->comm)); | |
1da177e4 LT |
1572 | } |
1573 | } | |
1574 | ||
1575 | return list_empty(work_q); | |
1576 | } | |
1577 | ||
30bd7df8 MC |
1578 | /** |
1579 | * scsi_eh_target_reset - send target reset if needed | |
1580 | * @shost: scsi host being recovered. | |
74cf298f | 1581 | * @work_q: &list_head for pending commands. |
30bd7df8 MC |
1582 | * @done_q: &list_head for processed commands. |
1583 | * | |
1584 | * Notes: | |
1585 | * Try a target reset. | |
1586 | */ | |
1587 | static int scsi_eh_target_reset(struct Scsi_Host *shost, | |
1588 | struct list_head *work_q, | |
1589 | struct list_head *done_q) | |
1590 | { | |
98db5195 | 1591 | LIST_HEAD(tmp_list); |
3eef6257 | 1592 | LIST_HEAD(check_list); |
30bd7df8 | 1593 | |
98db5195 JB |
1594 | list_splice_init(work_q, &tmp_list); |
1595 | ||
1596 | while (!list_empty(&tmp_list)) { | |
1597 | struct scsi_cmnd *next, *scmd; | |
b8e162f9 | 1598 | enum scsi_disposition rtn; |
98db5195 | 1599 | unsigned int id; |
b4562022 | 1600 | |
b4562022 | 1601 | if (scsi_host_eh_past_deadline(shost)) { |
b4562022 HR |
1602 | /* push back on work queue for further processing */ |
1603 | list_splice_init(&check_list, work_q); | |
1604 | list_splice_init(&tmp_list, work_q); | |
1605 | SCSI_LOG_ERROR_RECOVERY(3, | |
1606 | shost_printk(KERN_INFO, shost, | |
a222b1e2 HR |
1607 | "%s: Skip target reset, past eh deadline\n", |
1608 | current->comm)); | |
b4562022 HR |
1609 | return list_empty(work_q); |
1610 | } | |
98db5195 JB |
1611 | |
1612 | scmd = list_entry(tmp_list.next, struct scsi_cmnd, eh_entry); | |
1613 | id = scmd_id(scmd); | |
30bd7df8 | 1614 | |
91921e01 HR |
1615 | SCSI_LOG_ERROR_RECOVERY(3, |
1616 | shost_printk(KERN_INFO, shost, | |
1617 | "%s: Sending target reset to target %d\n", | |
1618 | current->comm, id)); | |
98db5195 JB |
1619 | rtn = scsi_try_target_reset(scmd); |
1620 | if (rtn != SUCCESS && rtn != FAST_IO_FAIL) | |
91921e01 HR |
1621 | SCSI_LOG_ERROR_RECOVERY(3, |
1622 | shost_printk(KERN_INFO, shost, | |
1623 | "%s: Target reset failed" | |
1624 | " target: %d\n", | |
1625 | current->comm, id)); | |
98db5195 JB |
1626 | list_for_each_entry_safe(scmd, next, &tmp_list, eh_entry) { |
1627 | if (scmd_id(scmd) != id) | |
1628 | continue; | |
1629 | ||
3eef6257 DJ |
1630 | if (rtn == SUCCESS) |
1631 | list_move_tail(&scmd->eh_entry, &check_list); | |
1632 | else if (rtn == FAST_IO_FAIL) | |
98db5195 JB |
1633 | scsi_eh_finish_cmd(scmd, done_q); |
1634 | else | |
1635 | /* push back on work queue for further processing */ | |
1636 | list_move(&scmd->eh_entry, work_q); | |
1637 | } | |
1638 | } | |
30bd7df8 | 1639 | |
3eef6257 | 1640 | return scsi_eh_test_devices(&check_list, work_q, done_q, 0); |
30bd7df8 MC |
1641 | } |
1642 | ||
1da177e4 | 1643 | /** |
0bf8c869 | 1644 | * scsi_eh_bus_reset - send a bus reset |
eb44820c | 1645 | * @shost: &scsi host being recovered. |
74cf298f | 1646 | * @work_q: &list_head for pending commands. |
eb44820c | 1647 | * @done_q: &list_head for processed commands. |
dc8875e1 | 1648 | */ |
1da177e4 LT |
1649 | static int scsi_eh_bus_reset(struct Scsi_Host *shost, |
1650 | struct list_head *work_q, | |
1651 | struct list_head *done_q) | |
1652 | { | |
937abeaa | 1653 | struct scsi_cmnd *scmd, *chan_scmd, *next; |
3eef6257 | 1654 | LIST_HEAD(check_list); |
1da177e4 | 1655 | unsigned int channel; |
b8e162f9 | 1656 | enum scsi_disposition rtn; |
1da177e4 LT |
1657 | |
1658 | /* | |
1659 | * we really want to loop over the various channels, and do this on | |
1660 | * a channel by channel basis. we should also check to see if any | |
1661 | * of the failed commands are on soft_reset devices, and if so, skip | |
0bf8c869 | 1662 | * the reset. |
1da177e4 LT |
1663 | */ |
1664 | ||
1665 | for (channel = 0; channel <= shost->max_channel; channel++) { | |
b4562022 | 1666 | if (scsi_host_eh_past_deadline(shost)) { |
b4562022 HR |
1667 | list_splice_init(&check_list, work_q); |
1668 | SCSI_LOG_ERROR_RECOVERY(3, | |
1669 | shost_printk(KERN_INFO, shost, | |
a222b1e2 HR |
1670 | "%s: skip BRST, past eh deadline\n", |
1671 | current->comm)); | |
b4562022 HR |
1672 | return list_empty(work_q); |
1673 | } | |
b4562022 | 1674 | |
1da177e4 LT |
1675 | chan_scmd = NULL; |
1676 | list_for_each_entry(scmd, work_q, eh_entry) { | |
422c0d61 | 1677 | if (channel == scmd_channel(scmd)) { |
1da177e4 LT |
1678 | chan_scmd = scmd; |
1679 | break; | |
1680 | /* | |
1681 | * FIXME add back in some support for | |
1682 | * soft_reset devices. | |
1683 | */ | |
1684 | } | |
1685 | } | |
1686 | ||
1687 | if (!chan_scmd) | |
1688 | continue; | |
91921e01 HR |
1689 | SCSI_LOG_ERROR_RECOVERY(3, |
1690 | shost_printk(KERN_INFO, shost, | |
1691 | "%s: Sending BRST chan: %d\n", | |
1692 | current->comm, channel)); | |
1da177e4 | 1693 | rtn = scsi_try_bus_reset(chan_scmd); |
2f2eb587 | 1694 | if (rtn == SUCCESS || rtn == FAST_IO_FAIL) { |
937abeaa | 1695 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
3eef6257 DJ |
1696 | if (channel == scmd_channel(scmd)) { |
1697 | if (rtn == FAST_IO_FAIL) | |
1da177e4 LT |
1698 | scsi_eh_finish_cmd(scmd, |
1699 | done_q); | |
3eef6257 DJ |
1700 | else |
1701 | list_move_tail(&scmd->eh_entry, | |
1702 | &check_list); | |
1703 | } | |
1da177e4 LT |
1704 | } |
1705 | } else { | |
91921e01 HR |
1706 | SCSI_LOG_ERROR_RECOVERY(3, |
1707 | shost_printk(KERN_INFO, shost, | |
1708 | "%s: BRST failed chan: %d\n", | |
1709 | current->comm, channel)); | |
1da177e4 LT |
1710 | } |
1711 | } | |
3eef6257 | 1712 | return scsi_eh_test_devices(&check_list, work_q, done_q, 0); |
1da177e4 LT |
1713 | } |
1714 | ||
1715 | /** | |
0bf8c869 | 1716 | * scsi_eh_host_reset - send a host reset |
74cf298f RD |
1717 | * @shost: host to be reset. |
1718 | * @work_q: &list_head for pending commands. | |
1719 | * @done_q: &list_head for processed commands. | |
dc8875e1 | 1720 | */ |
91921e01 HR |
1721 | static int scsi_eh_host_reset(struct Scsi_Host *shost, |
1722 | struct list_head *work_q, | |
1da177e4 LT |
1723 | struct list_head *done_q) |
1724 | { | |
937abeaa | 1725 | struct scsi_cmnd *scmd, *next; |
3eef6257 | 1726 | LIST_HEAD(check_list); |
b8e162f9 | 1727 | enum scsi_disposition rtn; |
1da177e4 LT |
1728 | |
1729 | if (!list_empty(work_q)) { | |
1730 | scmd = list_entry(work_q->next, | |
1731 | struct scsi_cmnd, eh_entry); | |
1732 | ||
91921e01 HR |
1733 | SCSI_LOG_ERROR_RECOVERY(3, |
1734 | shost_printk(KERN_INFO, shost, | |
1735 | "%s: Sending HRST\n", | |
1736 | current->comm)); | |
1da177e4 LT |
1737 | |
1738 | rtn = scsi_try_host_reset(scmd); | |
3eef6257 DJ |
1739 | if (rtn == SUCCESS) { |
1740 | list_splice_init(work_q, &check_list); | |
1741 | } else if (rtn == FAST_IO_FAIL) { | |
937abeaa | 1742 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
1da177e4 LT |
1743 | scsi_eh_finish_cmd(scmd, done_q); |
1744 | } | |
1745 | } else { | |
91921e01 HR |
1746 | SCSI_LOG_ERROR_RECOVERY(3, |
1747 | shost_printk(KERN_INFO, shost, | |
1748 | "%s: HRST failed\n", | |
1749 | current->comm)); | |
1da177e4 LT |
1750 | } |
1751 | } | |
3eef6257 | 1752 | return scsi_eh_test_devices(&check_list, work_q, done_q, 1); |
1da177e4 LT |
1753 | } |
1754 | ||
1755 | /** | |
1756 | * scsi_eh_offline_sdevs - offline scsi devices that fail to recover | |
74cf298f RD |
1757 | * @work_q: &list_head for pending commands. |
1758 | * @done_q: &list_head for processed commands. | |
dc8875e1 | 1759 | */ |
1da177e4 LT |
1760 | static void scsi_eh_offline_sdevs(struct list_head *work_q, |
1761 | struct list_head *done_q) | |
1762 | { | |
937abeaa | 1763 | struct scsi_cmnd *scmd, *next; |
0db6ca8a | 1764 | struct scsi_device *sdev; |
1da177e4 | 1765 | |
937abeaa | 1766 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
31765d7d MW |
1767 | sdev_printk(KERN_INFO, scmd->device, "Device offlined - " |
1768 | "not ready after error recovery\n"); | |
0db6ca8a BVA |
1769 | sdev = scmd->device; |
1770 | ||
1771 | mutex_lock(&sdev->state_mutex); | |
1772 | scsi_device_set_state(sdev, SDEV_OFFLINE); | |
1773 | mutex_unlock(&sdev->state_mutex); | |
1774 | ||
1da177e4 LT |
1775 | scsi_eh_finish_cmd(scmd, done_q); |
1776 | } | |
1777 | return; | |
1778 | } | |
1779 | ||
4a27446f | 1780 | /** |
e494f6a7 | 1781 | * scsi_noretry_cmd - determine if command should be failed fast |
4a27446f MC |
1782 | * @scmd: SCSI cmd to examine. |
1783 | */ | |
88b32c3c | 1784 | bool scsi_noretry_cmd(struct scsi_cmnd *scmd) |
4a27446f | 1785 | { |
aa8e25e5 BVA |
1786 | struct request *req = scsi_cmd_to_rq(scmd); |
1787 | ||
4a27446f MC |
1788 | switch (host_byte(scmd->result)) { |
1789 | case DID_OK: | |
1790 | break; | |
e494f6a7 HR |
1791 | case DID_TIME_OUT: |
1792 | goto check_type; | |
4a27446f | 1793 | case DID_BUS_BUSY: |
88b32c3c | 1794 | return !!(req->cmd_flags & REQ_FAILFAST_TRANSPORT); |
4a27446f | 1795 | case DID_PARITY: |
88b32c3c | 1796 | return !!(req->cmd_flags & REQ_FAILFAST_DEV); |
4a27446f | 1797 | case DID_ERROR: |
3d45cefc | 1798 | if (get_status_byte(scmd) == SAM_STAT_RESERVATION_CONFLICT) |
88b32c3c | 1799 | return false; |
df561f66 | 1800 | fallthrough; |
4a27446f | 1801 | case DID_SOFT_ERROR: |
88b32c3c | 1802 | return !!(req->cmd_flags & REQ_FAILFAST_DRIVER); |
4a27446f MC |
1803 | } |
1804 | ||
d0672a03 | 1805 | if (!scsi_status_is_check_condition(scmd->result)) |
88b32c3c | 1806 | return false; |
4a27446f | 1807 | |
e494f6a7 HR |
1808 | check_type: |
1809 | /* | |
1810 | * assume caller has checked sense and determined | |
1811 | * the check condition was retryable. | |
1812 | */ | |
aa8e25e5 | 1813 | if (req->cmd_flags & REQ_FAILFAST_DEV || blk_rq_is_passthrough(req)) |
88b32c3c | 1814 | return true; |
342c81ee | 1815 | |
88b32c3c | 1816 | return false; |
4a27446f MC |
1817 | } |
1818 | ||
1da177e4 LT |
1819 | /** |
1820 | * scsi_decide_disposition - Disposition a cmd on return from LLD. | |
1821 | * @scmd: SCSI cmd to examine. | |
1822 | * | |
1823 | * Notes: | |
1824 | * This is *only* called when we are examining the status after sending | |
1825 | * out the actual data command. any commands that are queued for error | |
1826 | * recovery (e.g. test_unit_ready) do *not* come through here. | |
1827 | * | |
1828 | * When this routine returns failed, it means the error handler thread | |
1829 | * is woken. In cases where the error code indicates an error that | |
1830 | * doesn't require the error handler read (i.e. we don't need to | |
1831 | * abort/reset), this function should return SUCCESS. | |
dc8875e1 | 1832 | */ |
b8e162f9 | 1833 | enum scsi_disposition scsi_decide_disposition(struct scsi_cmnd *scmd) |
1da177e4 | 1834 | { |
b8e162f9 | 1835 | enum scsi_disposition rtn; |
1da177e4 LT |
1836 | |
1837 | /* | |
1838 | * if the device is offline, then we clearly just pass the result back | |
1839 | * up to the top level. | |
1840 | */ | |
1841 | if (!scsi_device_online(scmd->device)) { | |
91921e01 HR |
1842 | SCSI_LOG_ERROR_RECOVERY(5, scmd_printk(KERN_INFO, scmd, |
1843 | "%s: device offline - report as SUCCESS\n", __func__)); | |
1da177e4 LT |
1844 | return SUCCESS; |
1845 | } | |
1846 | ||
1847 | /* | |
1848 | * first check the host byte, to see if there is anything in there | |
1849 | * that would indicate what we need to do. | |
1850 | */ | |
1851 | switch (host_byte(scmd->result)) { | |
1852 | case DID_PASSTHROUGH: | |
1853 | /* | |
1854 | * no matter what, pass this through to the upper layer. | |
1855 | * nuke this special code so that it looks like we are saying | |
1856 | * did_ok. | |
1857 | */ | |
1858 | scmd->result &= 0xff00ffff; | |
1859 | return SUCCESS; | |
1860 | case DID_OK: | |
1861 | /* | |
1862 | * looks good. drop through, and check the next byte. | |
1863 | */ | |
1864 | break; | |
e494f6a7 HR |
1865 | case DID_ABORT: |
1866 | if (scmd->eh_eflags & SCSI_EH_ABORT_SCHEDULED) { | |
8922a908 | 1867 | set_host_byte(scmd, DID_TIME_OUT); |
e494f6a7 HR |
1868 | return SUCCESS; |
1869 | } | |
df561f66 | 1870 | fallthrough; |
1da177e4 LT |
1871 | case DID_NO_CONNECT: |
1872 | case DID_BAD_TARGET: | |
1da177e4 LT |
1873 | /* |
1874 | * note - this means that we just report the status back | |
1875 | * to the top level driver, not that we actually think | |
1876 | * that it indicates SUCCESS. | |
1877 | */ | |
1878 | return SUCCESS; | |
ad95028a | 1879 | case DID_SOFT_ERROR: |
1da177e4 LT |
1880 | /* |
1881 | * when the low level driver returns did_soft_error, | |
0bf8c869 | 1882 | * it is responsible for keeping an internal retry counter |
1da177e4 | 1883 | * in order to avoid endless loops (db) |
1da177e4 | 1884 | */ |
1da177e4 LT |
1885 | goto maybe_retry; |
1886 | case DID_IMM_RETRY: | |
1887 | return NEEDS_RETRY; | |
1888 | ||
bf341919 | 1889 | case DID_REQUEUE: |
1890 | return ADD_TO_MLQUEUE; | |
a4dfaa6f MC |
1891 | case DID_TRANSPORT_DISRUPTED: |
1892 | /* | |
1893 | * LLD/transport was disrupted during processing of the IO. | |
1894 | * The transport class is now blocked/blocking, | |
1895 | * and the transport will decide what to do with the IO | |
939c2288 MC |
1896 | * based on its timers and recovery capablilities if |
1897 | * there are enough retries. | |
a4dfaa6f | 1898 | */ |
939c2288 | 1899 | goto maybe_retry; |
a4dfaa6f MC |
1900 | case DID_TRANSPORT_FAILFAST: |
1901 | /* | |
1902 | * The transport decided to failfast the IO (most likely | |
1903 | * the fast io fail tmo fired), so send IO directly upwards. | |
1904 | */ | |
1905 | return SUCCESS; | |
962c8dcd MK |
1906 | case DID_TRANSPORT_MARGINAL: |
1907 | /* | |
1908 | * caller has decided not to do retries on | |
1909 | * abort success, so send IO directly upwards | |
1910 | */ | |
1911 | return SUCCESS; | |
1da177e4 | 1912 | case DID_ERROR: |
3d45cefc | 1913 | if (get_status_byte(scmd) == SAM_STAT_RESERVATION_CONFLICT) |
1da177e4 LT |
1914 | /* |
1915 | * execute reservation conflict processing code | |
1916 | * lower down | |
1917 | */ | |
1918 | break; | |
df561f66 | 1919 | fallthrough; |
1da177e4 LT |
1920 | case DID_BUS_BUSY: |
1921 | case DID_PARITY: | |
1922 | goto maybe_retry; | |
1923 | case DID_TIME_OUT: | |
1924 | /* | |
1925 | * when we scan the bus, we get timeout messages for | |
1926 | * these commands if there is no device available. | |
1927 | * other hosts report did_no_connect for the same thing. | |
1928 | */ | |
1929 | if ((scmd->cmnd[0] == TEST_UNIT_READY || | |
1930 | scmd->cmnd[0] == INQUIRY)) { | |
1931 | return SUCCESS; | |
1932 | } else { | |
1933 | return FAILED; | |
1934 | } | |
1935 | case DID_RESET: | |
1936 | return SUCCESS; | |
1937 | default: | |
1938 | return FAILED; | |
1939 | } | |
1940 | ||
1da177e4 LT |
1941 | /* |
1942 | * check the status byte to see if this indicates anything special. | |
1943 | */ | |
3d45cefc HR |
1944 | switch (get_status_byte(scmd)) { |
1945 | case SAM_STAT_TASK_SET_FULL: | |
42a6a918 | 1946 | scsi_handle_queue_full(scmd->device); |
1da177e4 LT |
1947 | /* |
1948 | * the case of trying to send too many commands to a | |
1949 | * tagged queueing device. | |
1950 | */ | |
df561f66 | 1951 | fallthrough; |
3d45cefc | 1952 | case SAM_STAT_BUSY: |
1da177e4 LT |
1953 | /* |
1954 | * device can't talk to us at the moment. Should only | |
1955 | * occur (SAM-3) when the task queue is empty, so will cause | |
1956 | * the empty queue handling to trigger a stall in the | |
1957 | * device. | |
1958 | */ | |
1959 | return ADD_TO_MLQUEUE; | |
3d45cefc | 1960 | case SAM_STAT_GOOD: |
279afdfe EM |
1961 | if (scmd->cmnd[0] == REPORT_LUNS) |
1962 | scmd->device->sdev_target->expecting_lun_change = 0; | |
4a84067d | 1963 | scsi_handle_queue_ramp_up(scmd->device); |
df561f66 | 1964 | fallthrough; |
3d45cefc | 1965 | case SAM_STAT_COMMAND_TERMINATED: |
1da177e4 | 1966 | return SUCCESS; |
3d45cefc | 1967 | case SAM_STAT_TASK_ABORTED: |
a9b589d9 | 1968 | goto maybe_retry; |
3d45cefc | 1969 | case SAM_STAT_CHECK_CONDITION: |
1da177e4 LT |
1970 | rtn = scsi_check_sense(scmd); |
1971 | if (rtn == NEEDS_RETRY) | |
1972 | goto maybe_retry; | |
1973 | /* if rtn == FAILED, we have no sense information; | |
1974 | * returning FAILED will wake the error handler thread | |
1975 | * to collect the sense and redo the decide | |
1976 | * disposition */ | |
1977 | return rtn; | |
3d45cefc HR |
1978 | case SAM_STAT_CONDITION_MET: |
1979 | case SAM_STAT_INTERMEDIATE: | |
1980 | case SAM_STAT_INTERMEDIATE_CONDITION_MET: | |
1981 | case SAM_STAT_ACA_ACTIVE: | |
1da177e4 LT |
1982 | /* |
1983 | * who knows? FIXME(eric) | |
1984 | */ | |
1985 | return SUCCESS; | |
1986 | ||
3d45cefc | 1987 | case SAM_STAT_RESERVATION_CONFLICT: |
9ccfc756 JB |
1988 | sdev_printk(KERN_INFO, scmd->device, |
1989 | "reservation conflict\n"); | |
7dfaae6a | 1990 | set_scsi_ml_byte(scmd, SCSIML_STAT_RESV_CONFLICT); |
1da177e4 | 1991 | return SUCCESS; /* causes immediate i/o error */ |
1da177e4 LT |
1992 | } |
1993 | return FAILED; | |
1994 | ||
8ef7fe4b | 1995 | maybe_retry: |
1da177e4 LT |
1996 | |
1997 | /* we requeue for retry because the error was retryable, and | |
1998 | * the request was not marked fast fail. Note that above, | |
1999 | * even if the request is marked fast fail, we still requeue | |
2000 | * for queue congestion conditions (QUEUE_FULL or BUSY) */ | |
2a242d59 | 2001 | if (scsi_cmd_retry_allowed(scmd) && !scsi_noretry_cmd(scmd)) { |
1da177e4 LT |
2002 | return NEEDS_RETRY; |
2003 | } else { | |
2004 | /* | |
2005 | * no more retries - report this one back to upper level. | |
2006 | */ | |
2007 | return SUCCESS; | |
2008 | } | |
2009 | } | |
2010 | ||
de671d61 JA |
2011 | static enum rq_end_io_ret eh_lock_door_done(struct request *req, |
2012 | blk_status_t status) | |
f078727b | 2013 | { |
0bf6d96c | 2014 | blk_mq_free_request(req); |
de671d61 | 2015 | return RQ_END_IO_NONE; |
f078727b FT |
2016 | } |
2017 | ||
1da177e4 LT |
2018 | /** |
2019 | * scsi_eh_lock_door - Prevent medium removal for the specified device | |
2020 | * @sdev: SCSI device to prevent medium removal | |
2021 | * | |
2022 | * Locking: | |
91bc31fb | 2023 | * We must be called from process context. |
1da177e4 LT |
2024 | * |
2025 | * Notes: | |
2026 | * We queue up an asynchronous "ALLOW MEDIUM REMOVAL" request on the | |
2027 | * head of the devices request queue, and continue. | |
dc8875e1 | 2028 | */ |
1da177e4 LT |
2029 | static void scsi_eh_lock_door(struct scsi_device *sdev) |
2030 | { | |
ce70fd9a | 2031 | struct scsi_cmnd *scmd; |
f078727b | 2032 | struct request *req; |
1da177e4 | 2033 | |
68ec3b81 | 2034 | req = scsi_alloc_request(sdev->request_queue, REQ_OP_DRV_IN, 0); |
a492f075 | 2035 | if (IS_ERR(req)) |
eb571eea | 2036 | return; |
ce70fd9a | 2037 | scmd = blk_mq_rq_to_pdu(req); |
1da177e4 | 2038 | |
ce70fd9a CH |
2039 | scmd->cmnd[0] = ALLOW_MEDIUM_REMOVAL; |
2040 | scmd->cmnd[1] = 0; | |
2041 | scmd->cmnd[2] = 0; | |
2042 | scmd->cmnd[3] = 0; | |
2043 | scmd->cmnd[4] = SCSI_REMOVAL_PREVENT; | |
2044 | scmd->cmnd[5] = 0; | |
2045 | scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]); | |
e2e53086 | 2046 | scmd->allowed = 5; |
f078727b | 2047 | |
e8064021 | 2048 | req->rq_flags |= RQF_QUIET; |
f078727b | 2049 | req->timeout = 10 * HZ; |
e2e53086 | 2050 | req->end_io = eh_lock_door_done; |
f078727b | 2051 | |
e2e53086 | 2052 | blk_execute_rq_nowait(req, true); |
f078727b | 2053 | } |
1da177e4 LT |
2054 | |
2055 | /** | |
2056 | * scsi_restart_operations - restart io operations to the specified host. | |
2057 | * @shost: Host we are restarting. | |
2058 | * | |
2059 | * Notes: | |
2060 | * When we entered the error handler, we blocked all further i/o to | |
2061 | * this device. we need to 'reverse' this process. | |
dc8875e1 | 2062 | */ |
1da177e4 LT |
2063 | static void scsi_restart_operations(struct Scsi_Host *shost) |
2064 | { | |
2065 | struct scsi_device *sdev; | |
939647ee | 2066 | unsigned long flags; |
1da177e4 LT |
2067 | |
2068 | /* | |
2069 | * If the door was locked, we need to insert a door lock request | |
2070 | * onto the head of the SCSI request queue for the device. There | |
2071 | * is no point trying to lock the door of an off-line device. | |
2072 | */ | |
2073 | shost_for_each_device(sdev, shost) { | |
48379270 | 2074 | if (scsi_device_online(sdev) && sdev->was_reset && sdev->locked) { |
1da177e4 | 2075 | scsi_eh_lock_door(sdev); |
48379270 CH |
2076 | sdev->was_reset = 0; |
2077 | } | |
1da177e4 LT |
2078 | } |
2079 | ||
2080 | /* | |
2081 | * next free up anything directly waiting upon the host. this | |
2082 | * will be requests for character device operations, and also for | |
2083 | * ioctls to queued block devices. | |
2084 | */ | |
b4562022 | 2085 | SCSI_LOG_ERROR_RECOVERY(3, |
91921e01 | 2086 | shost_printk(KERN_INFO, shost, "waking up host to restart\n")); |
1da177e4 | 2087 | |
939647ee JB |
2088 | spin_lock_irqsave(shost->host_lock, flags); |
2089 | if (scsi_host_set_state(shost, SHOST_RUNNING)) | |
2090 | if (scsi_host_set_state(shost, SHOST_CANCEL)) | |
2091 | BUG_ON(scsi_host_set_state(shost, SHOST_DEL)); | |
2092 | spin_unlock_irqrestore(shost->host_lock, flags); | |
1da177e4 LT |
2093 | |
2094 | wake_up(&shost->host_wait); | |
2095 | ||
2096 | /* | |
2097 | * finally we need to re-initiate requests that may be pending. we will | |
2098 | * have had everything blocked while error handling is taking place, and | |
2099 | * now that error recovery is done, we will need to ensure that these | |
2100 | * requests are started. | |
2101 | */ | |
2102 | scsi_run_host_queues(shost); | |
57fc2e33 DW |
2103 | |
2104 | /* | |
2105 | * if eh is active and host_eh_scheduled is pending we need to re-run | |
2106 | * recovery. we do this check after scsi_run_host_queues() to allow | |
2107 | * everything pent up since the last eh run a chance to make forward | |
2108 | * progress before we sync again. Either we'll immediately re-run | |
2109 | * recovery or scsi_device_unbusy() will wake us again when these | |
2110 | * pending commands complete. | |
2111 | */ | |
2112 | spin_lock_irqsave(shost->host_lock, flags); | |
2113 | if (shost->host_eh_scheduled) | |
2114 | if (scsi_host_set_state(shost, SHOST_RECOVERY)) | |
2115 | WARN_ON(scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY)); | |
2116 | spin_unlock_irqrestore(shost->host_lock, flags); | |
1da177e4 LT |
2117 | } |
2118 | ||
2119 | /** | |
2120 | * scsi_eh_ready_devs - check device ready state and recover if not. | |
74cf298f RD |
2121 | * @shost: host to be recovered. |
2122 | * @work_q: &list_head for pending commands. | |
eb44820c | 2123 | * @done_q: &list_head for processed commands. |
dc8875e1 | 2124 | */ |
dca84e46 DW |
2125 | void scsi_eh_ready_devs(struct Scsi_Host *shost, |
2126 | struct list_head *work_q, | |
2127 | struct list_head *done_q) | |
1da177e4 LT |
2128 | { |
2129 | if (!scsi_eh_stu(shost, work_q, done_q)) | |
2130 | if (!scsi_eh_bus_device_reset(shost, work_q, done_q)) | |
30bd7df8 MC |
2131 | if (!scsi_eh_target_reset(shost, work_q, done_q)) |
2132 | if (!scsi_eh_bus_reset(shost, work_q, done_q)) | |
91921e01 | 2133 | if (!scsi_eh_host_reset(shost, work_q, done_q)) |
30bd7df8 MC |
2134 | scsi_eh_offline_sdevs(work_q, |
2135 | done_q); | |
1da177e4 | 2136 | } |
dca84e46 | 2137 | EXPORT_SYMBOL_GPL(scsi_eh_ready_devs); |
1da177e4 LT |
2138 | |
2139 | /** | |
2140 | * scsi_eh_flush_done_q - finish processed commands or retry them. | |
2141 | * @done_q: list_head of processed commands. | |
dc8875e1 | 2142 | */ |
041c5fc3 | 2143 | void scsi_eh_flush_done_q(struct list_head *done_q) |
1da177e4 | 2144 | { |
937abeaa | 2145 | struct scsi_cmnd *scmd, *next; |
1da177e4 | 2146 | |
937abeaa CH |
2147 | list_for_each_entry_safe(scmd, next, done_q, eh_entry) { |
2148 | list_del_init(&scmd->eh_entry); | |
1da177e4 | 2149 | if (scsi_device_online(scmd->device) && |
60bee27b MK |
2150 | !scsi_noretry_cmd(scmd) && scsi_cmd_retry_allowed(scmd) && |
2151 | scsi_eh_should_retry_cmd(scmd)) { | |
91921e01 HR |
2152 | SCSI_LOG_ERROR_RECOVERY(3, |
2153 | scmd_printk(KERN_INFO, scmd, | |
470613b4 HR |
2154 | "%s: flush retry cmd\n", |
2155 | current->comm)); | |
1da177e4 LT |
2156 | scsi_queue_insert(scmd, SCSI_MLQUEUE_EH_RETRY); |
2157 | } else { | |
793698ce PM |
2158 | /* |
2159 | * If just we got sense for the device (called | |
2160 | * scsi_eh_get_sense), scmd->result is already | |
4bd51e54 | 2161 | * set, do not set DID_TIME_OUT. |
793698ce | 2162 | */ |
1da177e4 | 2163 | if (!scmd->result) |
4bd51e54 | 2164 | scmd->result |= (DID_TIME_OUT << 16); |
91921e01 HR |
2165 | SCSI_LOG_ERROR_RECOVERY(3, |
2166 | scmd_printk(KERN_INFO, scmd, | |
470613b4 HR |
2167 | "%s: flush finish cmd\n", |
2168 | current->comm)); | |
1da177e4 LT |
2169 | scsi_finish_command(scmd); |
2170 | } | |
2171 | } | |
2172 | } | |
041c5fc3 | 2173 | EXPORT_SYMBOL(scsi_eh_flush_done_q); |
1da177e4 LT |
2174 | |
2175 | /** | |
2176 | * scsi_unjam_host - Attempt to fix a host which has a cmd that failed. | |
2177 | * @shost: Host to unjam. | |
2178 | * | |
2179 | * Notes: | |
2180 | * When we come in here, we *know* that all commands on the bus have | |
2181 | * either completed, failed or timed out. we also know that no further | |
2182 | * commands are being sent to the host, so things are relatively quiet | |
2183 | * and we have freedom to fiddle with things as we wish. | |
2184 | * | |
2185 | * This is only the *default* implementation. it is possible for | |
2186 | * individual drivers to supply their own version of this function, and | |
2187 | * if the maintainer wishes to do this, it is strongly suggested that | |
2188 | * this function be taken as a template and modified. this function | |
2189 | * was designed to correctly handle problems for about 95% of the | |
2190 | * different cases out there, and it should always provide at least a | |
2191 | * reasonable amount of error recovery. | |
2192 | * | |
2193 | * Any command marked 'failed' or 'timeout' must eventually have | |
2194 | * scsi_finish_cmd() called for it. we do all of the retry stuff | |
2195 | * here, so when we restart the host after we return it should have an | |
2196 | * empty queue. | |
dc8875e1 | 2197 | */ |
1da177e4 LT |
2198 | static void scsi_unjam_host(struct Scsi_Host *shost) |
2199 | { | |
2200 | unsigned long flags; | |
2201 | LIST_HEAD(eh_work_q); | |
2202 | LIST_HEAD(eh_done_q); | |
2203 | ||
2204 | spin_lock_irqsave(shost->host_lock, flags); | |
2205 | list_splice_init(&shost->eh_cmd_q, &eh_work_q); | |
2206 | spin_unlock_irqrestore(shost->host_lock, flags); | |
2207 | ||
2208 | SCSI_LOG_ERROR_RECOVERY(1, scsi_eh_prt_fail_stats(shost, &eh_work_q)); | |
2209 | ||
2210 | if (!scsi_eh_get_sense(&eh_work_q, &eh_done_q)) | |
a0658632 | 2211 | scsi_eh_ready_devs(shost, &eh_work_q, &eh_done_q); |
1da177e4 | 2212 | |
b4562022 | 2213 | spin_lock_irqsave(shost->host_lock, flags); |
bb3b621a | 2214 | if (shost->eh_deadline != -1) |
b4562022 HR |
2215 | shost->last_reset = 0; |
2216 | spin_unlock_irqrestore(shost->host_lock, flags); | |
1da177e4 LT |
2217 | scsi_eh_flush_done_q(&eh_done_q); |
2218 | } | |
2219 | ||
2220 | /** | |
ad42eb1b | 2221 | * scsi_error_handler - SCSI error handler thread |
1da177e4 LT |
2222 | * @data: Host for which we are running. |
2223 | * | |
2224 | * Notes: | |
ad42eb1b CH |
2225 | * This is the main error handling loop. This is run as a kernel thread |
2226 | * for every SCSI host and handles all error handling activity. | |
dc8875e1 | 2227 | */ |
1da177e4 LT |
2228 | int scsi_error_handler(void *data) |
2229 | { | |
ad42eb1b | 2230 | struct Scsi_Host *shost = data; |
1da177e4 | 2231 | |
1da177e4 | 2232 | /* |
ad42eb1b CH |
2233 | * We use TASK_INTERRUPTIBLE so that the thread is not |
2234 | * counted against the load average as a running process. | |
2235 | * We never actually get interrupted because kthread_run | |
c03264a7 | 2236 | * disables signal delivery for the created thread. |
1da177e4 | 2237 | */ |
537b604c MH |
2238 | while (true) { |
2239 | /* | |
2240 | * The sequence in kthread_stop() sets the stop flag first | |
2241 | * then wakes the process. To avoid missed wakeups, the task | |
2242 | * should always be in a non running state before the stop | |
2243 | * flag is checked | |
2244 | */ | |
b9d5c6b7 | 2245 | set_current_state(TASK_INTERRUPTIBLE); |
537b604c MH |
2246 | if (kthread_should_stop()) |
2247 | break; | |
2248 | ||
ee7863bc | 2249 | if ((shost->host_failed == 0 && shost->host_eh_scheduled == 0) || |
c84b023a | 2250 | shost->host_failed != scsi_host_busy(shost)) { |
ad42eb1b | 2251 | SCSI_LOG_ERROR_RECOVERY(1, |
91921e01 HR |
2252 | shost_printk(KERN_INFO, shost, |
2253 | "scsi_eh_%d: sleeping\n", | |
2254 | shost->host_no)); | |
3ed7a470 | 2255 | schedule(); |
3ed7a470 JB |
2256 | continue; |
2257 | } | |
1da177e4 | 2258 | |
3ed7a470 | 2259 | __set_current_state(TASK_RUNNING); |
ad42eb1b | 2260 | SCSI_LOG_ERROR_RECOVERY(1, |
91921e01 HR |
2261 | shost_printk(KERN_INFO, shost, |
2262 | "scsi_eh_%d: waking up %d/%d/%d\n", | |
2263 | shost->host_no, shost->host_eh_scheduled, | |
74665016 | 2264 | shost->host_failed, |
c84b023a | 2265 | scsi_host_busy(shost))); |
1da177e4 | 2266 | |
1da177e4 LT |
2267 | /* |
2268 | * We have a host that is failing for some reason. Figure out | |
2269 | * what we need to do to get it up and online again (if we can). | |
2270 | * If we fail, we end up taking the thing offline. | |
2271 | */ | |
ae0751ff | 2272 | if (!shost->eh_noresume && scsi_autopm_get_host(shost) != 0) { |
bc4f2401 | 2273 | SCSI_LOG_ERROR_RECOVERY(1, |
a222b1e2 HR |
2274 | shost_printk(KERN_ERR, shost, |
2275 | "scsi_eh_%d: unable to autoresume\n", | |
2276 | shost->host_no)); | |
bc4f2401 AS |
2277 | continue; |
2278 | } | |
2279 | ||
9227c33d CH |
2280 | if (shost->transportt->eh_strategy_handler) |
2281 | shost->transportt->eh_strategy_handler(shost); | |
1da177e4 LT |
2282 | else |
2283 | scsi_unjam_host(shost); | |
2284 | ||
72d8c36e WF |
2285 | /* All scmds have been handled */ |
2286 | shost->host_failed = 0; | |
2287 | ||
1da177e4 LT |
2288 | /* |
2289 | * Note - if the above fails completely, the action is to take | |
2290 | * individual devices offline and flush the queue of any | |
2291 | * outstanding requests that may have been pending. When we | |
2292 | * restart, we restart any I/O to any other devices on the bus | |
2293 | * which are still online. | |
2294 | */ | |
2295 | scsi_restart_operations(shost); | |
ae0751ff LM |
2296 | if (!shost->eh_noresume) |
2297 | scsi_autopm_put_host(shost); | |
1da177e4 | 2298 | } |
461a0ffb SR |
2299 | __set_current_state(TASK_RUNNING); |
2300 | ||
ad42eb1b | 2301 | SCSI_LOG_ERROR_RECOVERY(1, |
91921e01 HR |
2302 | shost_printk(KERN_INFO, shost, |
2303 | "Error handler scsi_eh_%d exiting\n", | |
2304 | shost->host_no)); | |
3ed7a470 | 2305 | shost->ehandler = NULL; |
1da177e4 LT |
2306 | return 0; |
2307 | } | |
2308 | ||
2309 | /* | |
2310 | * Function: scsi_report_bus_reset() | |
2311 | * | |
2312 | * Purpose: Utility function used by low-level drivers to report that | |
2313 | * they have observed a bus reset on the bus being handled. | |
2314 | * | |
2315 | * Arguments: shost - Host in question | |
2316 | * channel - channel on which reset was observed. | |
2317 | * | |
2318 | * Returns: Nothing | |
2319 | * | |
2320 | * Lock status: Host lock must be held. | |
2321 | * | |
2322 | * Notes: This only needs to be called if the reset is one which | |
2323 | * originates from an unknown location. Resets originated | |
2324 | * by the mid-level itself don't need to call this, but there | |
2325 | * should be no harm. | |
2326 | * | |
2327 | * The main purpose of this is to make sure that a CHECK_CONDITION | |
2328 | * is properly treated. | |
2329 | */ | |
2330 | void scsi_report_bus_reset(struct Scsi_Host *shost, int channel) | |
2331 | { | |
2332 | struct scsi_device *sdev; | |
2333 | ||
2334 | __shost_for_each_device(sdev, shost) { | |
30bd7df8 MC |
2335 | if (channel == sdev_channel(sdev)) |
2336 | __scsi_report_device_reset(sdev, NULL); | |
1da177e4 LT |
2337 | } |
2338 | } | |
2339 | EXPORT_SYMBOL(scsi_report_bus_reset); | |
2340 | ||
2341 | /* | |
2342 | * Function: scsi_report_device_reset() | |
2343 | * | |
2344 | * Purpose: Utility function used by low-level drivers to report that | |
2345 | * they have observed a device reset on the device being handled. | |
2346 | * | |
2347 | * Arguments: shost - Host in question | |
2348 | * channel - channel on which reset was observed | |
2349 | * target - target on which reset was observed | |
2350 | * | |
2351 | * Returns: Nothing | |
2352 | * | |
2353 | * Lock status: Host lock must be held | |
2354 | * | |
2355 | * Notes: This only needs to be called if the reset is one which | |
2356 | * originates from an unknown location. Resets originated | |
2357 | * by the mid-level itself don't need to call this, but there | |
2358 | * should be no harm. | |
2359 | * | |
2360 | * The main purpose of this is to make sure that a CHECK_CONDITION | |
2361 | * is properly treated. | |
2362 | */ | |
2363 | void scsi_report_device_reset(struct Scsi_Host *shost, int channel, int target) | |
2364 | { | |
2365 | struct scsi_device *sdev; | |
2366 | ||
2367 | __shost_for_each_device(sdev, shost) { | |
422c0d61 | 2368 | if (channel == sdev_channel(sdev) && |
30bd7df8 MC |
2369 | target == sdev_id(sdev)) |
2370 | __scsi_report_device_reset(sdev, NULL); | |
1da177e4 LT |
2371 | } |
2372 | } | |
2373 | EXPORT_SYMBOL(scsi_report_device_reset); | |
2374 | ||
176aa9d6 CH |
2375 | /** |
2376 | * scsi_ioctl_reset: explicitly reset a host/bus/target/device | |
2377 | * @dev: scsi_device to operate on | |
2378 | * @arg: reset type (see sg.h) | |
1da177e4 LT |
2379 | */ |
2380 | int | |
176aa9d6 | 2381 | scsi_ioctl_reset(struct scsi_device *dev, int __user *arg) |
1da177e4 | 2382 | { |
bc4f2401 | 2383 | struct scsi_cmnd *scmd; |
d7a1bb0a | 2384 | struct Scsi_Host *shost = dev->host; |
e9c787e6 | 2385 | struct request *rq; |
d7a1bb0a | 2386 | unsigned long flags; |
b8e162f9 BVA |
2387 | int error = 0, val; |
2388 | enum scsi_disposition rtn; | |
176aa9d6 CH |
2389 | |
2390 | if (!capable(CAP_SYS_ADMIN) || !capable(CAP_SYS_RAWIO)) | |
2391 | return -EACCES; | |
2392 | ||
2393 | error = get_user(val, arg); | |
2394 | if (error) | |
2395 | return error; | |
1da177e4 | 2396 | |
bc4f2401 | 2397 | if (scsi_autopm_get_host(shost) < 0) |
176aa9d6 | 2398 | return -EIO; |
bc4f2401 | 2399 | |
176aa9d6 | 2400 | error = -EIO; |
e9c787e6 CH |
2401 | rq = kzalloc(sizeof(struct request) + sizeof(struct scsi_cmnd) + |
2402 | shost->hostt->cmd_size, GFP_KERNEL); | |
2403 | if (!rq) | |
95eeb5f5 | 2404 | goto out_put_autopm_host; |
e9c787e6 | 2405 | blk_rq_init(NULL, rq); |
95eeb5f5 | 2406 | |
e9c787e6 CH |
2407 | scmd = (struct scsi_cmnd *)(rq + 1); |
2408 | scsi_init_command(dev, scmd); | |
64a87b24 | 2409 | |
bf23e619 | 2410 | scmd->submitter = SUBMITTED_BY_SCSI_RESET_IOCTL; |
30b0c37b | 2411 | memset(&scmd->sdb, 0, sizeof(scmd->sdb)); |
1da177e4 LT |
2412 | |
2413 | scmd->cmd_len = 0; | |
2414 | ||
2415 | scmd->sc_data_direction = DMA_BIDIRECTIONAL; | |
1da177e4 | 2416 | |
d7a1bb0a JS |
2417 | spin_lock_irqsave(shost->host_lock, flags); |
2418 | shost->tmf_in_progress = 1; | |
2419 | spin_unlock_irqrestore(shost->host_lock, flags); | |
2420 | ||
176aa9d6 CH |
2421 | switch (val & ~SG_SCSI_RESET_NO_ESCALATE) { |
2422 | case SG_SCSI_RESET_NOTHING: | |
2423 | rtn = SUCCESS; | |
2424 | break; | |
2425 | case SG_SCSI_RESET_DEVICE: | |
1da177e4 | 2426 | rtn = scsi_try_bus_device_reset(scmd); |
176aa9d6 | 2427 | if (rtn == SUCCESS || (val & SG_SCSI_RESET_NO_ESCALATE)) |
1da177e4 | 2428 | break; |
df561f66 | 2429 | fallthrough; |
176aa9d6 | 2430 | case SG_SCSI_RESET_TARGET: |
30bd7df8 | 2431 | rtn = scsi_try_target_reset(scmd); |
176aa9d6 | 2432 | if (rtn == SUCCESS || (val & SG_SCSI_RESET_NO_ESCALATE)) |
30bd7df8 | 2433 | break; |
df561f66 | 2434 | fallthrough; |
176aa9d6 | 2435 | case SG_SCSI_RESET_BUS: |
1da177e4 | 2436 | rtn = scsi_try_bus_reset(scmd); |
176aa9d6 | 2437 | if (rtn == SUCCESS || (val & SG_SCSI_RESET_NO_ESCALATE)) |
1da177e4 | 2438 | break; |
df561f66 | 2439 | fallthrough; |
176aa9d6 | 2440 | case SG_SCSI_RESET_HOST: |
1da177e4 | 2441 | rtn = scsi_try_host_reset(scmd); |
176aa9d6 CH |
2442 | if (rtn == SUCCESS) |
2443 | break; | |
df561f66 | 2444 | fallthrough; |
3bf2ff67 | 2445 | default: |
1da177e4 | 2446 | rtn = FAILED; |
176aa9d6 | 2447 | break; |
1da177e4 LT |
2448 | } |
2449 | ||
176aa9d6 CH |
2450 | error = (rtn == SUCCESS) ? 0 : -EIO; |
2451 | ||
d7a1bb0a JS |
2452 | spin_lock_irqsave(shost->host_lock, flags); |
2453 | shost->tmf_in_progress = 0; | |
2454 | spin_unlock_irqrestore(shost->host_lock, flags); | |
2455 | ||
2456 | /* | |
2457 | * be sure to wake up anyone who was sleeping or had their queue | |
2458 | * suspended while we performed the TMF. | |
2459 | */ | |
2460 | SCSI_LOG_ERROR_RECOVERY(3, | |
91921e01 HR |
2461 | shost_printk(KERN_INFO, shost, |
2462 | "waking up host to restart after TMF\n")); | |
d7a1bb0a JS |
2463 | |
2464 | wake_up(&shost->host_wait); | |
d7a1bb0a JS |
2465 | scsi_run_host_queues(shost); |
2466 | ||
e9c787e6 | 2467 | kfree(rq); |
0f121dd8 | 2468 | |
04796336 | 2469 | out_put_autopm_host: |
bc4f2401 | 2470 | scsi_autopm_put_host(shost); |
176aa9d6 | 2471 | return error; |
1da177e4 | 2472 | } |
1da177e4 | 2473 | |
4753cbc0 HR |
2474 | bool scsi_command_normalize_sense(const struct scsi_cmnd *cmd, |
2475 | struct scsi_sense_hdr *sshdr) | |
1da177e4 LT |
2476 | { |
2477 | return scsi_normalize_sense(cmd->sense_buffer, | |
b80ca4f7 | 2478 | SCSI_SENSE_BUFFERSIZE, sshdr); |
1da177e4 LT |
2479 | } |
2480 | EXPORT_SYMBOL(scsi_command_normalize_sense); | |
2481 | ||
1da177e4 | 2482 | /** |
eb44820c | 2483 | * scsi_get_sense_info_fld - get information field from sense data (either fixed or descriptor format) |
1da177e4 LT |
2484 | * @sense_buffer: byte array of sense data |
2485 | * @sb_len: number of valid bytes in sense_buffer | |
2486 | * @info_out: pointer to 64 integer where 8 or 4 byte information | |
2487 | * field will be placed if found. | |
2488 | * | |
2489 | * Return value: | |
2908769c | 2490 | * true if information field found, false if not found. |
dc8875e1 | 2491 | */ |
2908769c DLM |
2492 | bool scsi_get_sense_info_fld(const u8 *sense_buffer, int sb_len, |
2493 | u64 *info_out) | |
1da177e4 | 2494 | { |
1da177e4 | 2495 | const u8 * ucp; |
1da177e4 LT |
2496 | |
2497 | if (sb_len < 7) | |
2908769c | 2498 | return false; |
1da177e4 LT |
2499 | switch (sense_buffer[0] & 0x7f) { |
2500 | case 0x70: | |
2501 | case 0x71: | |
2502 | if (sense_buffer[0] & 0x80) { | |
2908769c DLM |
2503 | *info_out = get_unaligned_be32(&sense_buffer[3]); |
2504 | return true; | |
2505 | } | |
2506 | return false; | |
1da177e4 LT |
2507 | case 0x72: |
2508 | case 0x73: | |
2509 | ucp = scsi_sense_desc_find(sense_buffer, sb_len, | |
2510 | 0 /* info desc */); | |
2511 | if (ucp && (0xa == ucp[1])) { | |
2908769c DLM |
2512 | *info_out = get_unaligned_be64(&ucp[4]); |
2513 | return true; | |
2514 | } | |
2515 | return false; | |
1da177e4 | 2516 | default: |
2908769c | 2517 | return false; |
1da177e4 LT |
2518 | } |
2519 | } | |
2520 | EXPORT_SYMBOL(scsi_get_sense_info_fld); |