1 .. SPDX-License-Identifier: GPL-2.0 1 .. SPDX-License-Identifier: GPL-2.0
2 2
3 ======= 3 =======
4 SCSI EH 4 SCSI EH
5 ======= 5 =======
6 6
7 This document describes SCSI midlayer error ha 7 This document describes SCSI midlayer error handling infrastructure.
8 Please refer to Documentation/scsi/scsi_mid_lo 8 Please refer to Documentation/scsi/scsi_mid_low_api.rst for more
9 information regarding SCSI midlayer. 9 information regarding SCSI midlayer.
10 10
11 .. TABLE OF CONTENTS 11 .. TABLE OF CONTENTS
12 12
13 [1] How SCSI commands travel through the mi 13 [1] How SCSI commands travel through the midlayer and to EH
14 [1-1] struct scsi_cmnd 14 [1-1] struct scsi_cmnd
15 [1-2] How do scmd's get completed? 15 [1-2] How do scmd's get completed?
16 [1-2-1] Completing a scmd w/ scsi_done 16 [1-2-1] Completing a scmd w/ scsi_done
17 [1-2-2] Completing a scmd w/ timeout 17 [1-2-2] Completing a scmd w/ timeout
18 [1-3] How EH takes over 18 [1-3] How EH takes over
19 [2] How SCSI EH works 19 [2] How SCSI EH works
20 [2-1] EH through fine-grained callbacks 20 [2-1] EH through fine-grained callbacks
21 [2-1-1] Overview 21 [2-1-1] Overview
22 [2-1-2] Flow of scmds through EH 22 [2-1-2] Flow of scmds through EH
23 [2-1-3] Flow of control 23 [2-1-3] Flow of control
24 [2-2] EH through transportt->eh_strateg 24 [2-2] EH through transportt->eh_strategy_handler()
25 [2-2-1] Pre transportt->eh_strategy_ha 25 [2-2-1] Pre transportt->eh_strategy_handler() SCSI midlayer conditions
26 [2-2-2] Post transportt->eh_strategy_h 26 [2-2-2] Post transportt->eh_strategy_handler() SCSI midlayer conditions
27 [2-2-3] Things to consider 27 [2-2-3] Things to consider
28 28
29 29
30 1. How SCSI commands travel through the midlay 30 1. How SCSI commands travel through the midlayer and to EH
31 ============================================== 31 ==========================================================
32 32
33 1.1 struct scsi_cmnd 33 1.1 struct scsi_cmnd
34 -------------------- 34 --------------------
35 35
36 Each SCSI command is represented with struct s 36 Each SCSI command is represented with struct scsi_cmnd (== scmd). A
37 scmd has two list_head's to link itself into l 37 scmd has two list_head's to link itself into lists. The two are
38 scmd->list and scmd->eh_entry. The former is 38 scmd->list and scmd->eh_entry. The former is used for free list or
39 per-device allocated scmd list and not of much 39 per-device allocated scmd list and not of much interest to this EH
40 discussion. The latter is used for completion 40 discussion. The latter is used for completion and EH lists and unless
41 otherwise stated scmds are always linked using 41 otherwise stated scmds are always linked using scmd->eh_entry in this
42 discussion. 42 discussion.
43 43
44 44
45 1.2 How do scmd's get completed? 45 1.2 How do scmd's get completed?
46 -------------------------------- 46 --------------------------------
47 47
48 Once LLDD gets hold of a scmd, either the LLDD 48 Once LLDD gets hold of a scmd, either the LLDD will complete the
49 command by calling scsi_done callback passed f 49 command by calling scsi_done callback passed from midlayer when
50 invoking hostt->queuecommand() or the block la 50 invoking hostt->queuecommand() or the block layer will time it out.
51 51
52 52
53 1.2.1 Completing a scmd w/ scsi_done 53 1.2.1 Completing a scmd w/ scsi_done
54 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 54 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
55 55
56 For all non-EH commands, scsi_done() is the co 56 For all non-EH commands, scsi_done() is the completion callback. It
57 just calls blk_complete_request() to delete th 57 just calls blk_complete_request() to delete the block layer timer and
58 raise SCSI_SOFTIRQ 58 raise SCSI_SOFTIRQ
59 59
60 SCSI_SOFTIRQ handler scsi_softirq calls scsi_d 60 SCSI_SOFTIRQ handler scsi_softirq calls scsi_decide_disposition() to
61 determine what to do with the command. scsi_d 61 determine what to do with the command. scsi_decide_disposition()
62 looks at the scmd->result value and sense data 62 looks at the scmd->result value and sense data to determine what to do
63 with the command. 63 with the command.
64 64
65 - SUCCESS 65 - SUCCESS
66 66
67 scsi_finish_command() is invoked for t 67 scsi_finish_command() is invoked for the command. The
68 function does some maintenance chores 68 function does some maintenance chores and then calls
69 scsi_io_completion() to finish the I/O 69 scsi_io_completion() to finish the I/O.
70 scsi_io_completion() then notifies the 70 scsi_io_completion() then notifies the block layer on
71 the completed request by calling blk_e 71 the completed request by calling blk_end_request and
72 friends or figures out what to do with 72 friends or figures out what to do with the remainder
73 of the data in case of an error. 73 of the data in case of an error.
74 74
75 - NEEDS_RETRY 75 - NEEDS_RETRY
76 76
77 - ADD_TO_MLQUEUE 77 - ADD_TO_MLQUEUE
78 78
79 scmd is requeued to blk queue. 79 scmd is requeued to blk queue.
80 80
81 - otherwise 81 - otherwise
82 82
83 scsi_eh_scmd_add(scmd) is invoked for 83 scsi_eh_scmd_add(scmd) is invoked for the command. See
84 [1-3] for details of this function. 84 [1-3] for details of this function.
85 85
86 86
87 1.2.2 Completing a scmd w/ timeout 87 1.2.2 Completing a scmd w/ timeout
88 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 88 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
89 89
90 The timeout handler is scsi_timeout(). When a !! 90 The timeout handler is scsi_times_out(). When a timeout occurs, this
>> 91 function
91 92
92 1. invokes optional hostt->eh_timed_out() cal 93 1. invokes optional hostt->eh_timed_out() callback. Return value can
93 be one of 94 be one of
94 95
95 - SCSI_EH_RESET_TIMER !! 96 - BLK_EH_RESET_TIMER
96 This indicates that more time is requi 97 This indicates that more time is required to finish the
97 command. Timer is restarted. 98 command. Timer is restarted.
98 99
99 - SCSI_EH_NOT_HANDLED !! 100 - BLK_EH_DONE
100 eh_timed_out() callback did not handle 101 eh_timed_out() callback did not handle the command.
101 Step #2 is taken. 102 Step #2 is taken.
102 <<
103 - SCSI_EH_DONE <<
104 eh_timed_out() completed the command. <<
105 103
106 2. scsi_abort_command() is invoked to schedul 104 2. scsi_abort_command() is invoked to schedule an asynchronous abort which may
107 issue a retry scmd->allowed + 1 times. As 105 issue a retry scmd->allowed + 1 times. Asynchronous aborts are not invoked
108 for commands for which the SCSI_EH_ABORT_S 106 for commands for which the SCSI_EH_ABORT_SCHEDULED flag is set (this
109 indicates that the command already had bee 107 indicates that the command already had been aborted once, and this is a
110 retry which failed), when retries are exce 108 retry which failed), when retries are exceeded, or when the EH deadline is
111 expired. In these cases Step #3 is taken. 109 expired. In these cases Step #3 is taken.
112 110
113 3. scsi_eh_scmd_add(scmd, SCSI_EH_CANCEL_CMD) 111 3. scsi_eh_scmd_add(scmd, SCSI_EH_CANCEL_CMD) is invoked for the
114 command. See [1-4] for more information. 112 command. See [1-4] for more information.
115 113
116 1.3 Asynchronous command aborts 114 1.3 Asynchronous command aborts
117 ------------------------------- 115 -------------------------------
118 116
119 After a timeout occurs a command abort is sch 117 After a timeout occurs a command abort is scheduled from
120 scsi_abort_command(). If the abort is success 118 scsi_abort_command(). If the abort is successful the command
121 will either be retried (if the number of retr 119 will either be retried (if the number of retries is not exhausted)
122 or terminated with DID_TIME_OUT. 120 or terminated with DID_TIME_OUT.
123 121
124 Otherwise scsi_eh_scmd_add() is invoked for t 122 Otherwise scsi_eh_scmd_add() is invoked for the command.
125 See [1-4] for more information. 123 See [1-4] for more information.
126 124
127 1.4 How EH takes over 125 1.4 How EH takes over
128 --------------------- 126 ---------------------
129 127
130 scmds enter EH via scsi_eh_scmd_add(), which d 128 scmds enter EH via scsi_eh_scmd_add(), which does the following.
131 129
132 1. Links scmd->eh_entry to shost->eh_cmd_q 130 1. Links scmd->eh_entry to shost->eh_cmd_q
133 131
134 2. Sets SHOST_RECOVERY bit in shost->shost_st 132 2. Sets SHOST_RECOVERY bit in shost->shost_state
135 133
136 3. Increments shost->host_failed 134 3. Increments shost->host_failed
137 135
138 4. Wakes up SCSI EH thread if shost->host_bus 136 4. Wakes up SCSI EH thread if shost->host_busy == shost->host_failed
139 137
140 As can be seen above, once any scmd is added t 138 As can be seen above, once any scmd is added to shost->eh_cmd_q,
141 SHOST_RECOVERY shost_state bit is turned on. 139 SHOST_RECOVERY shost_state bit is turned on. This prevents any new
142 scmd to be issued from blk queue to the host; 140 scmd to be issued from blk queue to the host; eventually, all scmds on
143 the host either complete normally, fail and ge 141 the host either complete normally, fail and get added to eh_cmd_q, or
144 time out and get added to shost->eh_cmd_q. 142 time out and get added to shost->eh_cmd_q.
145 143
146 If all scmds either complete or fail, the numb 144 If all scmds either complete or fail, the number of in-flight scmds
147 becomes equal to the number of failed scmds - 145 becomes equal to the number of failed scmds - i.e. shost->host_busy ==
148 shost->host_failed. This wakes up SCSI EH thr 146 shost->host_failed. This wakes up SCSI EH thread. So, once woken up,
149 SCSI EH thread can expect that all in-flight c 147 SCSI EH thread can expect that all in-flight commands have failed and
150 are linked on shost->eh_cmd_q. 148 are linked on shost->eh_cmd_q.
151 149
152 Note that this does not mean lower layers are 150 Note that this does not mean lower layers are quiescent. If a LLDD
153 completed a scmd with error status, the LLDD a 151 completed a scmd with error status, the LLDD and lower layers are
154 assumed to forget about the scmd at that point 152 assumed to forget about the scmd at that point. However, if a scmd
155 has timed out, unless hostt->eh_timed_out() ma 153 has timed out, unless hostt->eh_timed_out() made lower layers forget
156 about the scmd, which currently no LLDD does, 154 about the scmd, which currently no LLDD does, the command is still
157 active as long as lower layers are concerned a 155 active as long as lower layers are concerned and completion could
158 occur at any time. Of course, all such comple 156 occur at any time. Of course, all such completions are ignored as the
159 timer has already expired. 157 timer has already expired.
160 158
161 We'll talk about how SCSI EH takes actions to 159 We'll talk about how SCSI EH takes actions to abort - make LLDD
162 forget about - timed out scmds later. 160 forget about - timed out scmds later.
163 161
164 162
165 2. How SCSI EH works 163 2. How SCSI EH works
166 ==================== 164 ====================
167 165
168 LLDD's can implement SCSI EH actions in one of 166 LLDD's can implement SCSI EH actions in one of the following two
169 ways. 167 ways.
170 168
171 - Fine-grained EH callbacks 169 - Fine-grained EH callbacks
172 LLDD can implement fine-grained EH cal 170 LLDD can implement fine-grained EH callbacks and let SCSI
173 midlayer drive error handling and call 171 midlayer drive error handling and call appropriate callbacks.
174 This will be discussed further in [2-1 172 This will be discussed further in [2-1].
175 173
176 - eh_strategy_handler() callback 174 - eh_strategy_handler() callback
177 This is one big callback which should 175 This is one big callback which should perform whole error
178 handling. As such, it should do all c 176 handling. As such, it should do all chores the SCSI midlayer
179 performs during recovery. This will b 177 performs during recovery. This will be discussed in [2-2].
180 178
181 Once recovery is complete, SCSI EH resumes nor 179 Once recovery is complete, SCSI EH resumes normal operation by
182 calling scsi_restart_operations(), which 180 calling scsi_restart_operations(), which
183 181
184 1. Checks if door locking is needed and locks 182 1. Checks if door locking is needed and locks door.
185 183
186 2. Clears SHOST_RECOVERY shost_state bit 184 2. Clears SHOST_RECOVERY shost_state bit
187 185
188 3. Wakes up waiters on shost->host_wait. Thi 186 3. Wakes up waiters on shost->host_wait. This occurs if someone
189 calls scsi_block_when_processing_errors() 187 calls scsi_block_when_processing_errors() on the host.
190 (*QUESTION* why is it needed? All operati 188 (*QUESTION* why is it needed? All operations will be blocked
191 anyway after it reaches blk queue.) 189 anyway after it reaches blk queue.)
192 190
193 4. Kicks queues in all devices on the host in 191 4. Kicks queues in all devices on the host in the asses
194 192
195 193
196 2.1 EH through fine-grained callbacks 194 2.1 EH through fine-grained callbacks
197 ------------------------------------- 195 -------------------------------------
198 196
199 2.1.1 Overview 197 2.1.1 Overview
200 ^^^^^^^^^^^^^^ 198 ^^^^^^^^^^^^^^
201 199
202 If eh_strategy_handler() is not present, SCSI 200 If eh_strategy_handler() is not present, SCSI midlayer takes charge
203 of driving error handling. EH's goals are two 201 of driving error handling. EH's goals are two - make LLDD, host and
204 device forget about timed out scmds and make t 202 device forget about timed out scmds and make them ready for new
205 commands. A scmd is said to be recovered if t 203 commands. A scmd is said to be recovered if the scmd is forgotten by
206 lower layers and lower layers are ready to pro 204 lower layers and lower layers are ready to process or fail the scmd
207 again. 205 again.
208 206
209 To achieve these goals, EH performs recovery a 207 To achieve these goals, EH performs recovery actions with increasing
210 severity. Some actions are performed by issui 208 severity. Some actions are performed by issuing SCSI commands and
211 others are performed by invoking one of the fo 209 others are performed by invoking one of the following fine-grained
212 hostt EH callbacks. Callbacks may be omitted 210 hostt EH callbacks. Callbacks may be omitted and omitted ones are
213 considered to fail always. 211 considered to fail always.
214 212
215 :: 213 ::
216 214
217 int (* eh_abort_handler)(struct scsi_cmnd 215 int (* eh_abort_handler)(struct scsi_cmnd *);
218 int (* eh_device_reset_handler)(struct scs 216 int (* eh_device_reset_handler)(struct scsi_cmnd *);
219 int (* eh_bus_reset_handler)(struct scsi_c 217 int (* eh_bus_reset_handler)(struct scsi_cmnd *);
220 int (* eh_host_reset_handler)(struct scsi_ 218 int (* eh_host_reset_handler)(struct scsi_cmnd *);
221 219
222 Higher-severity actions are taken only when lo 220 Higher-severity actions are taken only when lower-severity actions
223 cannot recover some of failed scmds. Also, no 221 cannot recover some of failed scmds. Also, note that failure of the
224 highest-severity action means EH failure and r 222 highest-severity action means EH failure and results in offlining of
225 all unrecovered devices. 223 all unrecovered devices.
226 224
227 During recovery, the following rules are follo 225 During recovery, the following rules are followed
228 226
229 - Recovery actions are performed on failed sc 227 - Recovery actions are performed on failed scmds on the to do list,
230 eh_work_q. If a recovery action succeeds f 228 eh_work_q. If a recovery action succeeds for a scmd, recovered
231 scmds are removed from eh_work_q. 229 scmds are removed from eh_work_q.
232 230
233 Note that single recovery action on a scmd 231 Note that single recovery action on a scmd can recover multiple
234 scmds. e.g. resetting a device recovers al 232 scmds. e.g. resetting a device recovers all failed scmds on the
235 device. 233 device.
236 234
237 - Higher severity actions are taken iff eh_wo 235 - Higher severity actions are taken iff eh_work_q is not empty after
238 lower severity actions are complete. 236 lower severity actions are complete.
239 237
240 - EH reuses failed scmds to issue commands fo 238 - EH reuses failed scmds to issue commands for recovery. For
241 timed-out scmds, SCSI EH ensures that LLDD 239 timed-out scmds, SCSI EH ensures that LLDD forgets about a scmd
242 before reusing it for EH commands. 240 before reusing it for EH commands.
243 241
244 When a scmd is recovered, the scmd is moved fr 242 When a scmd is recovered, the scmd is moved from eh_work_q to EH
245 local eh_done_q using scsi_eh_finish_cmd(). A 243 local eh_done_q using scsi_eh_finish_cmd(). After all scmds are
246 recovered (eh_work_q is empty), scsi_eh_flush_ 244 recovered (eh_work_q is empty), scsi_eh_flush_done_q() is invoked to
247 either retry or error-finish (notify upper lay 245 either retry or error-finish (notify upper layer of failure) recovered
248 scmds. 246 scmds.
249 247
250 scmds are retried iff its sdev is still online 248 scmds are retried iff its sdev is still online (not offlined during
251 EH), REQ_FAILFAST is not set and ++scmd->retri 249 EH), REQ_FAILFAST is not set and ++scmd->retries is less than
252 scmd->allowed. 250 scmd->allowed.
253 251
254 252
255 2.1.2 Flow of scmds through EH 253 2.1.2 Flow of scmds through EH
256 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 254 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
257 255
258 1. Error completion / time out 256 1. Error completion / time out
259 257
260 :ACTION: scsi_eh_scmd_add() is invoked for 258 :ACTION: scsi_eh_scmd_add() is invoked for scmd
261 259
262 - add scmd to shost->eh_cmd_q 260 - add scmd to shost->eh_cmd_q
263 - set SHOST_RECOVERY 261 - set SHOST_RECOVERY
264 - shost->host_failed++ 262 - shost->host_failed++
265 263
266 :LOCKING: shost->host_lock 264 :LOCKING: shost->host_lock
267 265
268 2. EH starts 266 2. EH starts
269 267
270 :ACTION: move all scmds to EH's local eh_w 268 :ACTION: move all scmds to EH's local eh_work_q. shost->eh_cmd_q
271 is cleared. 269 is cleared.
272 270
273 :LOCKING: shost->host_lock (not strictly n 271 :LOCKING: shost->host_lock (not strictly necessary, just for
274 consistency) 272 consistency)
275 273
276 3. scmd recovered 274 3. scmd recovered
277 275
278 :ACTION: scsi_eh_finish_cmd() is invoked t 276 :ACTION: scsi_eh_finish_cmd() is invoked to EH-finish scmd
279 277
280 - scsi_setup_cmd_retry() 278 - scsi_setup_cmd_retry()
281 - move from local eh_work_q to local e 279 - move from local eh_work_q to local eh_done_q
282 280
283 :LOCKING: none 281 :LOCKING: none
284 282
285 :CONCURRENCY: at most one thread per separ 283 :CONCURRENCY: at most one thread per separate eh_work_q to
286 keep queue manipulation lock 284 keep queue manipulation lockless
287 285
288 4. EH completes 286 4. EH completes
289 287
290 :ACTION: scsi_eh_flush_done_q() retries sc 288 :ACTION: scsi_eh_flush_done_q() retries scmds or notifies upper
291 layer of failure. May be called c 289 layer of failure. May be called concurrently but must have
292 a no more than one thread per sep 290 a no more than one thread per separate eh_work_q to
293 manipulate the queue locklessly 291 manipulate the queue locklessly
294 292
295 - scmd is removed from eh_done_q 293 - scmd is removed from eh_done_q and scmd->eh_entry is cleared
296 - if retry is necessary, scmd is 294 - if retry is necessary, scmd is requeued using
297 scsi_queue_insert() 295 scsi_queue_insert()
298 - otherwise, scsi_finish_command( 296 - otherwise, scsi_finish_command() is invoked for scmd
299 - zero shost->host_failed 297 - zero shost->host_failed
300 298
301 :LOCKING: queue or finish function perform 299 :LOCKING: queue or finish function performs appropriate locking
302 300
303 301
304 2.1.3 Flow of control 302 2.1.3 Flow of control
305 ^^^^^^^^^^^^^^^^^^^^^^ 303 ^^^^^^^^^^^^^^^^^^^^^^
306 304
307 EH through fine-grained callbacks start from 305 EH through fine-grained callbacks start from scsi_unjam_host().
308 306
309 ``scsi_unjam_host`` 307 ``scsi_unjam_host``
310 308
311 1. Lock shost->host_lock, splice_init shos 309 1. Lock shost->host_lock, splice_init shost->eh_cmd_q into local
312 eh_work_q and unlock host_lock. Note t 310 eh_work_q and unlock host_lock. Note that shost->eh_cmd_q is
313 cleared by this action. 311 cleared by this action.
314 312
315 2. Invoke scsi_eh_get_sense. 313 2. Invoke scsi_eh_get_sense.
316 314
317 ``scsi_eh_get_sense`` 315 ``scsi_eh_get_sense``
318 316
319 This action is taken for each error-co 317 This action is taken for each error-completed
320 (!SCSI_EH_CANCEL_CMD) commands without 318 (!SCSI_EH_CANCEL_CMD) commands without valid sense data. Most
321 SCSI transports/LLDDs automatically ac 319 SCSI transports/LLDDs automatically acquire sense data on
322 command failures (autosense). Autosen 320 command failures (autosense). Autosense is recommended for
323 performance reasons and as sense infor 321 performance reasons and as sense information could get out of
324 sync between occurrence of CHECK CONDI 322 sync between occurrence of CHECK CONDITION and this action.
325 323
326 Note that if autosense is not supporte 324 Note that if autosense is not supported, scmd->sense_buffer
327 contains invalid sense data when error 325 contains invalid sense data when error-completing the scmd
328 with scsi_done(). scsi_decide_disposi 326 with scsi_done(). scsi_decide_disposition() always returns
329 FAILED in such cases thus invoking SCS 327 FAILED in such cases thus invoking SCSI EH. When the scmd
330 reaches here, sense data is acquired a 328 reaches here, sense data is acquired and
331 scsi_decide_disposition() is called ag 329 scsi_decide_disposition() is called again.
332 330
333 1. Invoke scsi_request_sense() which i 331 1. Invoke scsi_request_sense() which issues REQUEST_SENSE
334 command. If fails, no action. Not 332 command. If fails, no action. Note that taking no action
335 causes higher-severity recovery to 333 causes higher-severity recovery to be taken for the scmd.
336 334
337 2. Invoke scsi_decide_disposition() on 335 2. Invoke scsi_decide_disposition() on the scmd
338 336
339 - SUCCESS 337 - SUCCESS
340 scmd->retries is set to scmd-> 338 scmd->retries is set to scmd->allowed preventing
341 scsi_eh_flush_done_q() from re 339 scsi_eh_flush_done_q() from retrying the scmd and
342 scsi_eh_finish_cmd() is invoke 340 scsi_eh_finish_cmd() is invoked.
343 341
344 - NEEDS_RETRY 342 - NEEDS_RETRY
345 scsi_eh_finish_cmd() invoked 343 scsi_eh_finish_cmd() invoked
346 344
347 - otherwise 345 - otherwise
348 No action. 346 No action.
349 347
350 3. If !list_empty(&eh_work_q), invoke scsi 348 3. If !list_empty(&eh_work_q), invoke scsi_eh_abort_cmds().
351 349
352 ``scsi_eh_abort_cmds`` 350 ``scsi_eh_abort_cmds``
353 351
354 This action is taken for each timed ou 352 This action is taken for each timed out command when
355 no_async_abort is enabled in the host 353 no_async_abort is enabled in the host template.
356 hostt->eh_abort_handler() is invoked f 354 hostt->eh_abort_handler() is invoked for each scmd. The
357 handler returns SUCCESS if it has succ 355 handler returns SUCCESS if it has succeeded to make LLDD and
358 all related hardware forget about the 356 all related hardware forget about the scmd.
359 357
360 If a timedout scmd is successfully abo 358 If a timedout scmd is successfully aborted and the sdev is
361 either offline or ready, scsi_eh_finis 359 either offline or ready, scsi_eh_finish_cmd() is invoked for
362 the scmd. Otherwise, the scmd is left 360 the scmd. Otherwise, the scmd is left in eh_work_q for
363 higher-severity actions. 361 higher-severity actions.
364 362
365 Note that both offline and ready statu 363 Note that both offline and ready status mean that the sdev is
366 ready to process new scmds, where proc 364 ready to process new scmds, where processing also implies
367 immediate failing; thus, if a sdev is 365 immediate failing; thus, if a sdev is in one of the two
368 states, no further recovery action is 366 states, no further recovery action is needed.
369 367
370 Device readiness is tested using scsi_ 368 Device readiness is tested using scsi_eh_tur() which issues
371 TEST_UNIT_READY command. Note that th 369 TEST_UNIT_READY command. Note that the scmd must have been
372 aborted successfully before reusing it 370 aborted successfully before reusing it for TEST_UNIT_READY.
373 371
374 4. If !list_empty(&eh_work_q), invoke scsi 372 4. If !list_empty(&eh_work_q), invoke scsi_eh_ready_devs()
375 373
376 ``scsi_eh_ready_devs`` 374 ``scsi_eh_ready_devs``
377 375
378 This function takes four increasingly 376 This function takes four increasingly more severe measures to
379 make failed sdevs ready for new comman 377 make failed sdevs ready for new commands.
380 378
381 1. Invoke scsi_eh_stu() 379 1. Invoke scsi_eh_stu()
382 380
383 ``scsi_eh_stu`` 381 ``scsi_eh_stu``
384 382
385 For each sdev which has failed scm 383 For each sdev which has failed scmds with valid sense data
386 of which scsi_check_sense()'s verd 384 of which scsi_check_sense()'s verdict is FAILED,
387 START_STOP_UNIT command is issued 385 START_STOP_UNIT command is issued w/ start=1. Note that
388 as we explicitly choose error-comp 386 as we explicitly choose error-completed scmds, it is known
389 that lower layers have forgotten a 387 that lower layers have forgotten about the scmd and we can
390 reuse it for STU. 388 reuse it for STU.
391 389
392 If STU succeeds and the sdev is ei 390 If STU succeeds and the sdev is either offline or ready,
393 all failed scmds on the sdev are E 391 all failed scmds on the sdev are EH-finished with
394 scsi_eh_finish_cmd(). 392 scsi_eh_finish_cmd().
395 393
396 *NOTE* If hostt->eh_abort_handler( 394 *NOTE* If hostt->eh_abort_handler() isn't implemented or
397 failed, we may still have timed ou 395 failed, we may still have timed out scmds at this point
398 and STU doesn't make lower layers 396 and STU doesn't make lower layers forget about those
399 scmds. Yet, this function EH-fini 397 scmds. Yet, this function EH-finish all scmds on the sdev
400 if STU succeeds leaving lower laye 398 if STU succeeds leaving lower layers in an inconsistent
401 state. It seems that STU action s 399 state. It seems that STU action should be taken only when
402 a sdev has no timed out scmd. 400 a sdev has no timed out scmd.
403 401
404 2. If !list_empty(&eh_work_q), invoke 402 2. If !list_empty(&eh_work_q), invoke scsi_eh_bus_device_reset().
405 403
406 ``scsi_eh_bus_device_reset`` 404 ``scsi_eh_bus_device_reset``
407 405
408 This action is very similar to scs 406 This action is very similar to scsi_eh_stu() except that,
409 instead of issuing STU, hostt->eh_ 407 instead of issuing STU, hostt->eh_device_reset_handler()
410 is used. Also, as we're not issui 408 is used. Also, as we're not issuing SCSI commands and
411 resetting clears all scmds on the 409 resetting clears all scmds on the sdev, there is no need
412 to choose error-completed scmds. 410 to choose error-completed scmds.
413 411
414 3. If !list_empty(&eh_work_q), invoke 412 3. If !list_empty(&eh_work_q), invoke scsi_eh_bus_reset()
415 413
416 ``scsi_eh_bus_reset`` 414 ``scsi_eh_bus_reset``
417 415
418 hostt->eh_bus_reset_handler() is i 416 hostt->eh_bus_reset_handler() is invoked for each channel
419 with failed scmds. If bus reset s 417 with failed scmds. If bus reset succeeds, all failed
420 scmds on all ready or offline sdev 418 scmds on all ready or offline sdevs on the channel are
421 EH-finished. 419 EH-finished.
422 420
423 4. If !list_empty(&eh_work_q), invoke 421 4. If !list_empty(&eh_work_q), invoke scsi_eh_host_reset()
424 422
425 ``scsi_eh_host_reset`` 423 ``scsi_eh_host_reset``
426 424
427 This is the last resort. hostt->e 425 This is the last resort. hostt->eh_host_reset_handler()
428 is invoked. If host reset succeed 426 is invoked. If host reset succeeds, all failed scmds on
429 all ready or offline sdevs on the 427 all ready or offline sdevs on the host are EH-finished.
430 428
431 5. If !list_empty(&eh_work_q), invoke 429 5. If !list_empty(&eh_work_q), invoke scsi_eh_offline_sdevs()
432 430
433 ``scsi_eh_offline_sdevs`` 431 ``scsi_eh_offline_sdevs``
434 432
435 Take all sdevs which still have un 433 Take all sdevs which still have unrecovered scmds offline
436 and EH-finish the scmds. 434 and EH-finish the scmds.
437 435
438 5. Invoke scsi_eh_flush_done_q(). 436 5. Invoke scsi_eh_flush_done_q().
439 437
440 ``scsi_eh_flush_done_q`` 438 ``scsi_eh_flush_done_q``
441 439
442 At this point all scmds are recove 440 At this point all scmds are recovered (or given up) and
443 put on eh_done_q by scsi_eh_finish 441 put on eh_done_q by scsi_eh_finish_cmd(). This function
444 flushes eh_done_q by either retryi 442 flushes eh_done_q by either retrying or notifying upper
445 layer of failure of the scmds. 443 layer of failure of the scmds.
446 444
447 445
448 2.2 EH through transportt->eh_strategy_handler 446 2.2 EH through transportt->eh_strategy_handler()
449 ---------------------------------------------- 447 ------------------------------------------------
450 448
451 transportt->eh_strategy_handler() is invoked i 449 transportt->eh_strategy_handler() is invoked in the place of
452 scsi_unjam_host() and it is responsible for wh 450 scsi_unjam_host() and it is responsible for whole recovery process.
453 On completion, the handler should have made lo 451 On completion, the handler should have made lower layers forget about
454 all failed scmds and either ready for new comm 452 all failed scmds and either ready for new commands or offline. Also,
455 it should perform SCSI EH maintenance chores t 453 it should perform SCSI EH maintenance chores to maintain integrity of
456 SCSI midlayer. IOW, of the steps described in 454 SCSI midlayer. IOW, of the steps described in [2-1-2], all steps
457 except for #1 must be implemented by eh_strate 455 except for #1 must be implemented by eh_strategy_handler().
458 456
459 457
460 2.2.1 Pre transportt->eh_strategy_handler() SC 458 2.2.1 Pre transportt->eh_strategy_handler() SCSI midlayer conditions
461 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 459 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
462 460
463 The following conditions are true on entry to 461 The following conditions are true on entry to the handler.
464 462
465 - Each failed scmd's eh_flags field is set ap 463 - Each failed scmd's eh_flags field is set appropriately.
466 464
467 - Each failed scmd is linked on scmd->eh_cmd_ 465 - Each failed scmd is linked on scmd->eh_cmd_q by scmd->eh_entry.
468 466
469 - SHOST_RECOVERY is set. 467 - SHOST_RECOVERY is set.
470 468
471 - shost->host_failed == shost->host_busy 469 - shost->host_failed == shost->host_busy
472 470
473 471
474 2.2.2 Post transportt->eh_strategy_handler() S 472 2.2.2 Post transportt->eh_strategy_handler() SCSI midlayer conditions
475 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 473 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
476 474
477 The following conditions must be true on exit 475 The following conditions must be true on exit from the handler.
478 476
479 - shost->host_failed is zero. 477 - shost->host_failed is zero.
480 478
481 - Each scmd is in such a state that scsi_setu 479 - Each scmd is in such a state that scsi_setup_cmd_retry() on the
482 scmd doesn't make any difference. 480 scmd doesn't make any difference.
483 481
484 - shost->eh_cmd_q is cleared. 482 - shost->eh_cmd_q is cleared.
485 483
486 - Each scmd->eh_entry is cleared. 484 - Each scmd->eh_entry is cleared.
487 485
488 - Either scsi_queue_insert() or scsi_finish_c 486 - Either scsi_queue_insert() or scsi_finish_command() is called on
489 each scmd. Note that the handler is free t 487 each scmd. Note that the handler is free to use scmd->retries and
490 ->allowed to limit the number of retries. 488 ->allowed to limit the number of retries.
491 489
492 490
493 2.2.3 Things to consider 491 2.2.3 Things to consider
494 ^^^^^^^^^^^^^^^^^^^^^^^^ 492 ^^^^^^^^^^^^^^^^^^^^^^^^
495 493
496 - Know that timed out scmds are still active 494 - Know that timed out scmds are still active on lower layers. Make
497 lower layers forget about them before doing 495 lower layers forget about them before doing anything else with
498 those scmds. 496 those scmds.
499 497
500 - For consistency, when accessing/modifying s 498 - For consistency, when accessing/modifying shost data structure,
501 grab shost->host_lock. 499 grab shost->host_lock.
502 500
503 - On completion, each failed sdev must have f 501 - On completion, each failed sdev must have forgotten about all
504 active scmds. 502 active scmds.
505 503
506 - On completion, each failed sdev must be rea 504 - On completion, each failed sdev must be ready for new commands or
507 offline. 505 offline.
508 506
509 507
510 Tejun Heo 508 Tejun Heo
511 htejun@gmail.com 509 htejun@gmail.com
512 510
513 11th September 2005 511 11th September 2005
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.