Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * scsi_error.c Copyright (C) 1997 Eric Youngdale |
| 3 | * |
| 4 | * SCSI error/timeout handling |
| 5 | * Initial versions: Eric Youngdale. Based upon conversations with |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 6 | * Leonard Zubkoff and David Miller at Linux Expo, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | * ideas originating from all over the place. |
| 8 | * |
| 9 | * Restructured scsi_unjam_host and associated functions. |
| 10 | * September 04, 2002 Mike Anderson (andmike@us.ibm.com) |
| 11 | * |
| 12 | * Forward port of Russell King's (rmk@arm.linux.org.uk) changes and |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 13 | * minor cleanups. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 14 | * September 30, 2002 Mike Anderson (andmike@us.ibm.com) |
| 15 | */ |
| 16 | |
| 17 | #include <linux/module.h> |
| 18 | #include <linux/sched.h> |
Tejun Heo | 5a0e3ad | 2010-03-24 17:04:11 +0900 | [diff] [blame] | 19 | #include <linux/gfp.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 20 | #include <linux/timer.h> |
| 21 | #include <linux/string.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include <linux/kernel.h> |
Rafael J. Wysocki | 8314418 | 2007-07-17 04:03:35 -0700 | [diff] [blame] | 23 | #include <linux/freezer.h> |
Christoph Hellwig | c5478de | 2005-09-06 14:04:26 +0200 | [diff] [blame] | 24 | #include <linux/kthread.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 25 | #include <linux/interrupt.h> |
| 26 | #include <linux/blkdev.h> |
| 27 | #include <linux/delay.h> |
| 28 | |
| 29 | #include <scsi/scsi.h> |
Christoph Hellwig | beb4048 | 2006-06-10 18:01:03 +0200 | [diff] [blame] | 30 | #include <scsi/scsi_cmnd.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 31 | #include <scsi/scsi_dbg.h> |
| 32 | #include <scsi/scsi_device.h> |
| 33 | #include <scsi/scsi_eh.h> |
James Smart | c829c39 | 2006-03-13 08:28:57 -0500 | [diff] [blame] | 34 | #include <scsi/scsi_transport.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 35 | #include <scsi/scsi_host.h> |
| 36 | #include <scsi/scsi_ioctl.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 37 | |
| 38 | #include "scsi_priv.h" |
| 39 | #include "scsi_logging.h" |
Adrian Bunk | 79ee830 | 2007-08-10 14:50:42 -0700 | [diff] [blame] | 40 | #include "scsi_transport_api.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | |
Kei Tokunaga | bf81623 | 2010-04-01 20:41:40 +0900 | [diff] [blame] | 42 | #include <trace/events/scsi.h> |
| 43 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | #define SENSE_TIMEOUT (10*HZ) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 45 | |
| 46 | /* |
| 47 | * These should *probably* be handled by the host itself. |
| 48 | * Since it is allowed to sleep, it probably should. |
| 49 | */ |
| 50 | #define BUS_RESET_SETTLE_TIME (10) |
| 51 | #define HOST_RESET_SETTLE_TIME (10) |
| 52 | |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 53 | static int scsi_eh_try_stu(struct scsi_cmnd *scmd); |
| 54 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 55 | /* called with shost->host_lock held */ |
| 56 | void scsi_eh_wakeup(struct Scsi_Host *shost) |
| 57 | { |
| 58 | if (shost->host_busy == shost->host_failed) { |
Kei Tokunaga | bf81623 | 2010-04-01 20:41:40 +0900 | [diff] [blame] | 59 | trace_scsi_eh_wakeup(shost); |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 60 | wake_up_process(shost->ehandler); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 61 | SCSI_LOG_ERROR_RECOVERY(5, |
| 62 | printk("Waking error handler thread\n")); |
| 63 | } |
| 64 | } |
Tejun Heo | f8bbfc2 | 2006-05-19 21:07:05 +0900 | [diff] [blame] | 65 | |
| 66 | /** |
| 67 | * scsi_schedule_eh - schedule EH for SCSI host |
| 68 | * @shost: SCSI host to invoke error handling on. |
| 69 | * |
| 70 | * Schedule SCSI EH without scmd. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 71 | */ |
Tejun Heo | f8bbfc2 | 2006-05-19 21:07:05 +0900 | [diff] [blame] | 72 | void scsi_schedule_eh(struct Scsi_Host *shost) |
| 73 | { |
| 74 | unsigned long flags; |
| 75 | |
| 76 | spin_lock_irqsave(shost->host_lock, flags); |
| 77 | |
| 78 | if (scsi_host_set_state(shost, SHOST_RECOVERY) == 0 || |
| 79 | scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY) == 0) { |
| 80 | shost->host_eh_scheduled++; |
| 81 | scsi_eh_wakeup(shost); |
| 82 | } |
| 83 | |
| 84 | spin_unlock_irqrestore(shost->host_lock, flags); |
| 85 | } |
| 86 | EXPORT_SYMBOL_GPL(scsi_schedule_eh); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 87 | |
| 88 | /** |
| 89 | * scsi_eh_scmd_add - add scsi cmd to error handling. |
| 90 | * @scmd: scmd to run eh on. |
| 91 | * @eh_flag: optional SCSI_EH flag. |
| 92 | * |
| 93 | * Return value: |
| 94 | * 0 on failure. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 95 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 96 | int scsi_eh_scmd_add(struct scsi_cmnd *scmd, int eh_flag) |
| 97 | { |
| 98 | struct Scsi_Host *shost = scmd->device->host; |
| 99 | unsigned long flags; |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 100 | int ret = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 101 | |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 102 | if (!shost->ehandler) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 103 | return 0; |
| 104 | |
| 105 | spin_lock_irqsave(shost->host_lock, flags); |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 106 | if (scsi_host_set_state(shost, SHOST_RECOVERY)) |
| 107 | if (scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY)) |
| 108 | goto out_unlock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 110 | ret = 1; |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 111 | scmd->eh_eflags |= eh_flag; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 112 | list_add_tail(&scmd->eh_entry, &shost->eh_cmd_q); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 113 | shost->host_failed++; |
| 114 | scsi_eh_wakeup(shost); |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 115 | out_unlock: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 116 | spin_unlock_irqrestore(shost->host_lock, flags); |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 117 | return ret; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 118 | } |
| 119 | |
| 120 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 121 | * scsi_times_out - Timeout function for normal scsi commands. |
Jens Axboe | 242f9dc | 2008-09-14 05:55:09 -0700 | [diff] [blame] | 122 | * @req: request that is timing out. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 123 | * |
| 124 | * Notes: |
| 125 | * We do not need to lock this. There is the potential for a race |
| 126 | * only in that the normal completion handling might run, but if the |
| 127 | * normal completion function determines that the timer has already |
| 128 | * fired, then it mustn't do anything. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 129 | */ |
Jens Axboe | 242f9dc | 2008-09-14 05:55:09 -0700 | [diff] [blame] | 130 | enum blk_eh_timer_return scsi_times_out(struct request *req) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 131 | { |
Jens Axboe | 242f9dc | 2008-09-14 05:55:09 -0700 | [diff] [blame] | 132 | struct scsi_cmnd *scmd = req->special; |
Jens Axboe | 242f9dc | 2008-09-14 05:55:09 -0700 | [diff] [blame] | 133 | enum blk_eh_timer_return rtn = BLK_EH_NOT_HANDLED; |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 134 | struct Scsi_Host *host = scmd->device->host; |
James Bottomley | 6c5f8ce | 2007-03-16 17:44:41 -0500 | [diff] [blame] | 135 | |
Kei Tokunaga | bf81623 | 2010-04-01 20:41:40 +0900 | [diff] [blame] | 136 | trace_scsi_dispatch_cmd_timeout(scmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 137 | scsi_log_completion(scmd, TIMEOUT_ERROR); |
| 138 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 139 | if (host->transportt->eh_timed_out) |
| 140 | rtn = host->transportt->eh_timed_out(scmd); |
| 141 | else if (host->hostt->eh_timed_out) |
| 142 | rtn = host->hostt->eh_timed_out(scmd); |
James Bottomley | 6c5f8ce | 2007-03-16 17:44:41 -0500 | [diff] [blame] | 143 | |
Christoph Hellwig | fa99078 | 2008-11-05 15:03:15 +0100 | [diff] [blame] | 144 | if (unlikely(rtn == BLK_EH_NOT_HANDLED && |
| 145 | !scsi_eh_scmd_add(scmd, SCSI_EH_CANCEL_CMD))) { |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 146 | scmd->result |= DID_TIME_OUT << 16; |
Christoph Hellwig | fa99078 | 2008-11-05 15:03:15 +0100 | [diff] [blame] | 147 | rtn = BLK_EH_HANDLED; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 148 | } |
Jens Axboe | 242f9dc | 2008-09-14 05:55:09 -0700 | [diff] [blame] | 149 | |
Christoph Hellwig | fa99078 | 2008-11-05 15:03:15 +0100 | [diff] [blame] | 150 | return rtn; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 151 | } |
| 152 | |
| 153 | /** |
| 154 | * scsi_block_when_processing_errors - Prevent cmds from being queued. |
| 155 | * @sdev: Device on which we are performing recovery. |
| 156 | * |
| 157 | * Description: |
| 158 | * We block until the host is out of error recovery, and then check to |
| 159 | * see whether the host or the device is offline. |
| 160 | * |
| 161 | * Return value: |
| 162 | * 0 when dev was taken offline by error recovery. 1 OK to proceed. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 163 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 164 | int scsi_block_when_processing_errors(struct scsi_device *sdev) |
| 165 | { |
| 166 | int online; |
| 167 | |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 168 | wait_event(sdev->host->host_wait, !scsi_host_in_recovery(sdev->host)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 169 | |
| 170 | online = scsi_device_online(sdev); |
| 171 | |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 172 | SCSI_LOG_ERROR_RECOVERY(5, printk("%s: rtn: %d\n", __func__, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 173 | online)); |
| 174 | |
| 175 | return online; |
| 176 | } |
| 177 | EXPORT_SYMBOL(scsi_block_when_processing_errors); |
| 178 | |
| 179 | #ifdef CONFIG_SCSI_LOGGING |
| 180 | /** |
| 181 | * scsi_eh_prt_fail_stats - Log info on failures. |
| 182 | * @shost: scsi host being recovered. |
| 183 | * @work_q: Queue of scsi cmds to process. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 184 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 185 | static inline void scsi_eh_prt_fail_stats(struct Scsi_Host *shost, |
| 186 | struct list_head *work_q) |
| 187 | { |
| 188 | struct scsi_cmnd *scmd; |
| 189 | struct scsi_device *sdev; |
| 190 | int total_failures = 0; |
| 191 | int cmd_failed = 0; |
| 192 | int cmd_cancel = 0; |
| 193 | int devices_failed = 0; |
| 194 | |
| 195 | shost_for_each_device(sdev, shost) { |
| 196 | list_for_each_entry(scmd, work_q, eh_entry) { |
| 197 | if (scmd->device == sdev) { |
| 198 | ++total_failures; |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 199 | if (scmd->eh_eflags & SCSI_EH_CANCEL_CMD) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 200 | ++cmd_cancel; |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 201 | else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 202 | ++cmd_failed; |
| 203 | } |
| 204 | } |
| 205 | |
| 206 | if (cmd_cancel || cmd_failed) { |
| 207 | SCSI_LOG_ERROR_RECOVERY(3, |
James Bottomley | 9ccfc75 | 2005-10-02 11:45:08 -0500 | [diff] [blame] | 208 | sdev_printk(KERN_INFO, sdev, |
| 209 | "%s: cmds failed: %d, cancel: %d\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 210 | __func__, cmd_failed, |
James Bottomley | 9ccfc75 | 2005-10-02 11:45:08 -0500 | [diff] [blame] | 211 | cmd_cancel)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 212 | cmd_cancel = 0; |
| 213 | cmd_failed = 0; |
| 214 | ++devices_failed; |
| 215 | } |
| 216 | } |
| 217 | |
| 218 | SCSI_LOG_ERROR_RECOVERY(2, printk("Total of %d commands on %d" |
| 219 | " devices require eh work\n", |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 220 | total_failures, devices_failed)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 221 | } |
| 222 | #endif |
| 223 | |
| 224 | /** |
| 225 | * scsi_check_sense - Examine scsi cmd sense |
| 226 | * @scmd: Cmd to have sense checked. |
| 227 | * |
| 228 | * Return value: |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 229 | * SUCCESS or FAILED or NEEDS_RETRY or TARGET_ERROR |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 230 | * |
| 231 | * Notes: |
| 232 | * When a deferred error is detected the current command has |
| 233 | * not been executed and needs retrying. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 234 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 235 | static int scsi_check_sense(struct scsi_cmnd *scmd) |
| 236 | { |
Chandra Seetharaman | a6a8d9f | 2008-05-01 14:49:46 -0700 | [diff] [blame] | 237 | struct scsi_device *sdev = scmd->device; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 238 | struct scsi_sense_hdr sshdr; |
| 239 | |
| 240 | if (! scsi_command_normalize_sense(scmd, &sshdr)) |
| 241 | return FAILED; /* no valid sense data */ |
| 242 | |
| 243 | if (scsi_sense_is_deferred(&sshdr)) |
| 244 | return NEEDS_RETRY; |
| 245 | |
Chandra Seetharaman | a6a8d9f | 2008-05-01 14:49:46 -0700 | [diff] [blame] | 246 | if (sdev->scsi_dh_data && sdev->scsi_dh_data->scsi_dh && |
| 247 | sdev->scsi_dh_data->scsi_dh->check_sense) { |
| 248 | int rc; |
| 249 | |
| 250 | rc = sdev->scsi_dh_data->scsi_dh->check_sense(sdev, &sshdr); |
| 251 | if (rc != SCSI_RETURN_NOT_HANDLED) |
| 252 | return rc; |
| 253 | /* handler does not care. Drop down to default handling */ |
| 254 | } |
| 255 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 256 | /* |
| 257 | * Previous logic looked for FILEMARK, EOM or ILI which are |
| 258 | * mainly associated with tapes and returned SUCCESS. |
| 259 | */ |
| 260 | if (sshdr.response_code == 0x70) { |
| 261 | /* fixed format */ |
| 262 | if (scmd->sense_buffer[2] & 0xe0) |
| 263 | return SUCCESS; |
| 264 | } else { |
| 265 | /* |
| 266 | * descriptor format: look for "stream commands sense data |
| 267 | * descriptor" (see SSC-3). Assume single sense data |
| 268 | * descriptor. Ignore ILI from SBC-2 READ LONG and WRITE LONG. |
| 269 | */ |
| 270 | if ((sshdr.additional_length > 3) && |
| 271 | (scmd->sense_buffer[8] == 0x4) && |
| 272 | (scmd->sense_buffer[11] & 0xe0)) |
| 273 | return SUCCESS; |
| 274 | } |
| 275 | |
| 276 | switch (sshdr.sense_key) { |
| 277 | case NO_SENSE: |
| 278 | return SUCCESS; |
| 279 | case RECOVERED_ERROR: |
| 280 | return /* soft_error */ SUCCESS; |
| 281 | |
| 282 | case ABORTED_COMMAND: |
Martin K. Petersen | 511e44f | 2008-07-17 04:28:33 -0400 | [diff] [blame] | 283 | if (sshdr.asc == 0x10) /* DIF */ |
| 284 | return SUCCESS; |
| 285 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 286 | return NEEDS_RETRY; |
| 287 | case NOT_READY: |
| 288 | case UNIT_ATTENTION: |
| 289 | /* |
| 290 | * if we are expecting a cc/ua because of a bus reset that we |
| 291 | * performed, treat this just as a retry. otherwise this is |
| 292 | * information that we should pass up to the upper-level driver |
| 293 | * so that we can deal with it there. |
| 294 | */ |
| 295 | if (scmd->device->expecting_cc_ua) { |
TARUISI Hiroaki | dfcf777 | 2011-08-11 20:25:20 +0900 | [diff] [blame] | 296 | /* |
| 297 | * Because some device does not queue unit |
| 298 | * attentions correctly, we carefully check |
| 299 | * additional sense code and qualifier so as |
| 300 | * not to squash media change unit attention. |
| 301 | */ |
| 302 | if (sshdr.asc != 0x28 || sshdr.ascq != 0x00) { |
| 303 | scmd->device->expecting_cc_ua = 0; |
| 304 | return NEEDS_RETRY; |
| 305 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 306 | } |
| 307 | /* |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 308 | * if the device is in the process of becoming ready, we |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 309 | * should retry. |
| 310 | */ |
| 311 | if ((sshdr.asc == 0x04) && (sshdr.ascq == 0x01)) |
| 312 | return NEEDS_RETRY; |
| 313 | /* |
| 314 | * if the device is not started, we need to wake |
| 315 | * the error handler to start the motor |
| 316 | */ |
| 317 | if (scmd->device->allow_restart && |
| 318 | (sshdr.asc == 0x04) && (sshdr.ascq == 0x02)) |
| 319 | return FAILED; |
James Bottomley | 77a4229 | 2010-05-04 16:51:40 -0400 | [diff] [blame] | 320 | |
Mike Christie | 6e49949 | 2010-04-25 07:03:57 -0500 | [diff] [blame] | 321 | if (sshdr.asc == 0x3f && sshdr.ascq == 0x0e) |
| 322 | scmd_printk(KERN_WARNING, scmd, |
| 323 | "Warning! Received an indication that the " |
| 324 | "LUN assignments on this target have " |
| 325 | "changed. The Linux SCSI layer does not " |
| 326 | "automatically remap LUN assignments.\n"); |
| 327 | else if (sshdr.asc == 0x3f) |
| 328 | scmd_printk(KERN_WARNING, scmd, |
| 329 | "Warning! Received an indication that the " |
| 330 | "operating parameters on this target have " |
| 331 | "changed. The Linux SCSI layer does not " |
| 332 | "automatically adjust these parameters.\n"); |
| 333 | |
Shyam Iyer | deb1cb6 | 2011-02-26 01:59:44 -0500 | [diff] [blame] | 334 | if (sshdr.asc == 0x38 && sshdr.ascq == 0x07) |
| 335 | scmd_printk(KERN_WARNING, scmd, |
| 336 | "Warning! Received an indication that the " |
| 337 | "LUN reached a thin provisioning soft " |
| 338 | "threshold.\n"); |
| 339 | |
Christoph Hellwig | 02e031c | 2010-11-10 14:54:09 +0100 | [diff] [blame] | 340 | /* |
| 341 | * Pass the UA upwards for a determination in the completion |
| 342 | * functions. |
| 343 | */ |
| 344 | return SUCCESS; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 345 | |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 346 | /* these are not supported */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 347 | case COPY_ABORTED: |
| 348 | case VOLUME_OVERFLOW: |
| 349 | case MISCOMPARE: |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 350 | case BLANK_CHECK: |
| 351 | case DATA_PROTECT: |
| 352 | return TARGET_ERROR; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 353 | |
| 354 | case MEDIUM_ERROR: |
Luben Tuikov | fd1b494 | 2006-11-29 19:45:23 -0800 | [diff] [blame] | 355 | if (sshdr.asc == 0x11 || /* UNRECOVERED READ ERR */ |
| 356 | sshdr.asc == 0x13 || /* AMNF DATA FIELD */ |
| 357 | sshdr.asc == 0x14) { /* RECORD NOT FOUND */ |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 358 | return TARGET_ERROR; |
Luben Tuikov | fd1b494 | 2006-11-29 19:45:23 -0800 | [diff] [blame] | 359 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 360 | return NEEDS_RETRY; |
| 361 | |
| 362 | case HARDWARE_ERROR: |
| 363 | if (scmd->device->retry_hwerror) |
Mike Anderson | bb0003c | 2008-08-12 12:11:58 -0700 | [diff] [blame] | 364 | return ADD_TO_MLQUEUE; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 365 | else |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 366 | return TARGET_ERROR; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 367 | |
| 368 | case ILLEGAL_REQUEST: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 369 | default: |
| 370 | return SUCCESS; |
| 371 | } |
| 372 | } |
| 373 | |
Vasu Dev | 4a84067 | 2009-10-22 15:46:33 -0700 | [diff] [blame] | 374 | static void scsi_handle_queue_ramp_up(struct scsi_device *sdev) |
| 375 | { |
| 376 | struct scsi_host_template *sht = sdev->host->hostt; |
| 377 | struct scsi_device *tmp_sdev; |
| 378 | |
| 379 | if (!sht->change_queue_depth || |
| 380 | sdev->queue_depth >= sdev->max_queue_depth) |
| 381 | return; |
| 382 | |
| 383 | if (time_before(jiffies, |
| 384 | sdev->last_queue_ramp_up + sdev->queue_ramp_up_period)) |
| 385 | return; |
| 386 | |
| 387 | if (time_before(jiffies, |
| 388 | sdev->last_queue_full_time + sdev->queue_ramp_up_period)) |
| 389 | return; |
| 390 | |
| 391 | /* |
| 392 | * Walk all devices of a target and do |
| 393 | * ramp up on them. |
| 394 | */ |
| 395 | shost_for_each_device(tmp_sdev, sdev->host) { |
| 396 | if (tmp_sdev->channel != sdev->channel || |
| 397 | tmp_sdev->id != sdev->id || |
| 398 | tmp_sdev->queue_depth == sdev->max_queue_depth) |
| 399 | continue; |
| 400 | /* |
| 401 | * call back into LLD to increase queue_depth by one |
| 402 | * with ramp up reason code. |
| 403 | */ |
| 404 | sht->change_queue_depth(tmp_sdev, tmp_sdev->queue_depth + 1, |
| 405 | SCSI_QDEPTH_RAMP_UP); |
| 406 | sdev->last_queue_ramp_up = jiffies; |
| 407 | } |
| 408 | } |
| 409 | |
Mike Christie | 42a6a91 | 2009-10-15 17:46:44 -0700 | [diff] [blame] | 410 | static void scsi_handle_queue_full(struct scsi_device *sdev) |
| 411 | { |
| 412 | struct scsi_host_template *sht = sdev->host->hostt; |
| 413 | struct scsi_device *tmp_sdev; |
| 414 | |
| 415 | if (!sht->change_queue_depth) |
| 416 | return; |
| 417 | |
| 418 | shost_for_each_device(tmp_sdev, sdev->host) { |
| 419 | if (tmp_sdev->channel != sdev->channel || |
| 420 | tmp_sdev->id != sdev->id) |
| 421 | continue; |
| 422 | /* |
| 423 | * We do not know the number of commands that were at |
| 424 | * the device when we got the queue full so we start |
| 425 | * from the highest possible value and work our way down. |
| 426 | */ |
| 427 | sht->change_queue_depth(tmp_sdev, tmp_sdev->queue_depth - 1, |
| 428 | SCSI_QDEPTH_QFULL); |
| 429 | } |
| 430 | } |
| 431 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 432 | /** |
| 433 | * scsi_eh_completed_normally - Disposition a eh cmd on return from LLD. |
| 434 | * @scmd: SCSI cmd to examine. |
| 435 | * |
| 436 | * Notes: |
| 437 | * This is *only* called when we are examining the status of commands |
| 438 | * queued during error recovery. the main difference here is that we |
| 439 | * don't allow for the possibility of retries here, and we are a lot |
| 440 | * more restrictive about what we consider acceptable. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 441 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 442 | static int scsi_eh_completed_normally(struct scsi_cmnd *scmd) |
| 443 | { |
| 444 | /* |
| 445 | * first check the host byte, to see if there is anything in there |
| 446 | * that would indicate what we need to do. |
| 447 | */ |
| 448 | if (host_byte(scmd->result) == DID_RESET) { |
| 449 | /* |
| 450 | * rats. we are already in the error handler, so we now |
| 451 | * get to try and figure out what to do next. if the sense |
| 452 | * is valid, we have a pretty good idea of what to do. |
| 453 | * if not, we mark it as FAILED. |
| 454 | */ |
| 455 | return scsi_check_sense(scmd); |
| 456 | } |
| 457 | if (host_byte(scmd->result) != DID_OK) |
| 458 | return FAILED; |
| 459 | |
| 460 | /* |
| 461 | * next, check the message byte. |
| 462 | */ |
| 463 | if (msg_byte(scmd->result) != COMMAND_COMPLETE) |
| 464 | return FAILED; |
| 465 | |
| 466 | /* |
| 467 | * now, check the status byte to see if this indicates |
| 468 | * anything special. |
| 469 | */ |
| 470 | switch (status_byte(scmd->result)) { |
| 471 | case GOOD: |
Vasu Dev | 4a84067 | 2009-10-22 15:46:33 -0700 | [diff] [blame] | 472 | scsi_handle_queue_ramp_up(scmd->device); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 473 | case COMMAND_TERMINATED: |
| 474 | return SUCCESS; |
| 475 | case CHECK_CONDITION: |
| 476 | return scsi_check_sense(scmd); |
| 477 | case CONDITION_GOOD: |
| 478 | case INTERMEDIATE_GOOD: |
| 479 | case INTERMEDIATE_C_GOOD: |
| 480 | /* |
| 481 | * who knows? FIXME(eric) |
| 482 | */ |
| 483 | return SUCCESS; |
Michael Reed | 5f91bb0 | 2009-08-10 11:59:28 -0500 | [diff] [blame] | 484 | case RESERVATION_CONFLICT: |
James Bottomley | 67110df | 2010-08-06 15:17:24 -0500 | [diff] [blame] | 485 | if (scmd->cmnd[0] == TEST_UNIT_READY) |
| 486 | /* it is a success, we probed the device and |
| 487 | * found it */ |
| 488 | return SUCCESS; |
| 489 | /* otherwise, we failed to send the command */ |
| 490 | return FAILED; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 491 | case QUEUE_FULL: |
Mike Christie | 42a6a91 | 2009-10-15 17:46:44 -0700 | [diff] [blame] | 492 | scsi_handle_queue_full(scmd->device); |
| 493 | /* fall through */ |
| 494 | case BUSY: |
Hannes Reinecke | 3eb3a92 | 2010-07-29 10:10:16 +0200 | [diff] [blame] | 495 | return NEEDS_RETRY; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 496 | default: |
| 497 | return FAILED; |
| 498 | } |
| 499 | return FAILED; |
| 500 | } |
| 501 | |
| 502 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 503 | * scsi_eh_done - Completion function for error handling. |
| 504 | * @scmd: Cmd that is done. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 505 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 506 | static void scsi_eh_done(struct scsi_cmnd *scmd) |
| 507 | { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 508 | struct completion *eh_action; |
Michael Reed | 8563167 | 2005-12-07 21:46:27 -0600 | [diff] [blame] | 509 | |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 510 | SCSI_LOG_ERROR_RECOVERY(3, |
| 511 | printk("%s scmd: %p result: %x\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 512 | __func__, scmd, scmd->result)); |
Michael Reed | 8563167 | 2005-12-07 21:46:27 -0600 | [diff] [blame] | 513 | |
| 514 | eh_action = scmd->device->host->eh_action; |
| 515 | if (eh_action) |
| 516 | complete(eh_action); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 517 | } |
| 518 | |
| 519 | /** |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 520 | * scsi_try_host_reset - ask host adapter to reset itself |
| 521 | * @scmd: SCSI cmd to send hsot reset. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 522 | */ |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 523 | static int scsi_try_host_reset(struct scsi_cmnd *scmd) |
| 524 | { |
| 525 | unsigned long flags; |
| 526 | int rtn; |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 527 | struct Scsi_Host *host = scmd->device->host; |
| 528 | struct scsi_host_template *hostt = host->hostt; |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 529 | |
| 530 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Snd Host RST\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 531 | __func__)); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 532 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 533 | if (!hostt->eh_host_reset_handler) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 534 | return FAILED; |
| 535 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 536 | rtn = hostt->eh_host_reset_handler(scmd); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 537 | |
| 538 | if (rtn == SUCCESS) { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 539 | if (!hostt->skip_settle_delay) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 540 | ssleep(HOST_RESET_SETTLE_TIME); |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 541 | spin_lock_irqsave(host->host_lock, flags); |
| 542 | scsi_report_bus_reset(host, scmd_channel(scmd)); |
| 543 | spin_unlock_irqrestore(host->host_lock, flags); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 544 | } |
| 545 | |
| 546 | return rtn; |
| 547 | } |
| 548 | |
| 549 | /** |
| 550 | * scsi_try_bus_reset - ask host to perform a bus reset |
| 551 | * @scmd: SCSI cmd to send bus reset. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 552 | */ |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 553 | static int scsi_try_bus_reset(struct scsi_cmnd *scmd) |
| 554 | { |
| 555 | unsigned long flags; |
| 556 | int rtn; |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 557 | struct Scsi_Host *host = scmd->device->host; |
| 558 | struct scsi_host_template *hostt = host->hostt; |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 559 | |
| 560 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Snd Bus RST\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 561 | __func__)); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 562 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 563 | if (!hostt->eh_bus_reset_handler) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 564 | return FAILED; |
| 565 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 566 | rtn = hostt->eh_bus_reset_handler(scmd); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 567 | |
| 568 | if (rtn == SUCCESS) { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 569 | if (!hostt->skip_settle_delay) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 570 | ssleep(BUS_RESET_SETTLE_TIME); |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 571 | spin_lock_irqsave(host->host_lock, flags); |
| 572 | scsi_report_bus_reset(host, scmd_channel(scmd)); |
| 573 | spin_unlock_irqrestore(host->host_lock, flags); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 574 | } |
| 575 | |
| 576 | return rtn; |
| 577 | } |
| 578 | |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 579 | static void __scsi_report_device_reset(struct scsi_device *sdev, void *data) |
| 580 | { |
| 581 | sdev->was_reset = 1; |
| 582 | sdev->expecting_cc_ua = 1; |
| 583 | } |
| 584 | |
| 585 | /** |
| 586 | * scsi_try_target_reset - Ask host to perform a target reset |
| 587 | * @scmd: SCSI cmd used to send a target reset |
| 588 | * |
| 589 | * Notes: |
| 590 | * There is no timeout for this operation. if this operation is |
| 591 | * unreliable for a given host, then the host itself needs to put a |
| 592 | * timer on it, and set the host back to a consistent state prior to |
| 593 | * returning. |
| 594 | */ |
| 595 | static int scsi_try_target_reset(struct scsi_cmnd *scmd) |
| 596 | { |
| 597 | unsigned long flags; |
| 598 | int rtn; |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 599 | struct Scsi_Host *host = scmd->device->host; |
| 600 | struct scsi_host_template *hostt = host->hostt; |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 601 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 602 | if (!hostt->eh_target_reset_handler) |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 603 | return FAILED; |
| 604 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 605 | rtn = hostt->eh_target_reset_handler(scmd); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 606 | if (rtn == SUCCESS) { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 607 | spin_lock_irqsave(host->host_lock, flags); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 608 | __starget_for_each_device(scsi_target(scmd->device), NULL, |
| 609 | __scsi_report_device_reset); |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 610 | spin_unlock_irqrestore(host->host_lock, flags); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 611 | } |
| 612 | |
| 613 | return rtn; |
| 614 | } |
| 615 | |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 616 | /** |
| 617 | * scsi_try_bus_device_reset - Ask host to perform a BDR on a dev |
| 618 | * @scmd: SCSI cmd used to send BDR |
| 619 | * |
| 620 | * Notes: |
| 621 | * There is no timeout for this operation. if this operation is |
| 622 | * unreliable for a given host, then the host itself needs to put a |
| 623 | * timer on it, and set the host back to a consistent state prior to |
| 624 | * returning. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 625 | */ |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 626 | static int scsi_try_bus_device_reset(struct scsi_cmnd *scmd) |
| 627 | { |
| 628 | int rtn; |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 629 | struct scsi_host_template *hostt = scmd->device->host->hostt; |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 630 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 631 | if (!hostt->eh_device_reset_handler) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 632 | return FAILED; |
| 633 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 634 | rtn = hostt->eh_device_reset_handler(scmd); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 635 | if (rtn == SUCCESS) |
| 636 | __scsi_report_device_reset(scmd->device, NULL); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 637 | return rtn; |
| 638 | } |
| 639 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 640 | static int scsi_try_to_abort_cmd(struct scsi_host_template *hostt, struct scsi_cmnd *scmd) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 641 | { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 642 | if (!hostt->eh_abort_handler) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 643 | return FAILED; |
| 644 | |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 645 | return hostt->eh_abort_handler(scmd); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 646 | } |
| 647 | |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 648 | static void scsi_abort_eh_cmnd(struct scsi_cmnd *scmd) |
| 649 | { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 650 | if (scsi_try_to_abort_cmd(scmd->device->host->hostt, scmd) != SUCCESS) |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 651 | if (scsi_try_bus_device_reset(scmd) != SUCCESS) |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 652 | if (scsi_try_target_reset(scmd) != SUCCESS) |
| 653 | if (scsi_try_bus_reset(scmd) != SUCCESS) |
| 654 | scsi_try_host_reset(scmd); |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 655 | } |
| 656 | |
| 657 | /** |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 658 | * scsi_eh_prep_cmnd - Save a scsi command info as part of error recory |
Christoph Hellwig | 2dc611d | 2006-11-04 20:04:21 +0100 | [diff] [blame] | 659 | * @scmd: SCSI command structure to hijack |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 660 | * @ses: structure to save restore information |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 661 | * @cmnd: CDB to send. Can be NULL if no new cmnd is needed |
Boaz Harrosh | 64a87b2 | 2008-04-30 11:19:47 +0300 | [diff] [blame] | 662 | * @cmnd_size: size in bytes of @cmnd (must be <= BLK_MAX_CDB) |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 663 | * @sense_bytes: size of sense data to copy. or 0 (if != 0 @cmnd is ignored) |
Christoph Hellwig | 2dc611d | 2006-11-04 20:04:21 +0100 | [diff] [blame] | 664 | * |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 665 | * This function is used to save a scsi command information before re-execution |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 666 | * as part of the error recovery process. If @sense_bytes is 0 the command |
| 667 | * sent must be one that does not transfer any data. If @sense_bytes != 0 |
| 668 | * @cmnd is ignored and this functions sets up a REQUEST_SENSE command |
| 669 | * and cmnd buffers to read @sense_bytes into @scmd->sense_buffer. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 670 | */ |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 671 | void scsi_eh_prep_cmnd(struct scsi_cmnd *scmd, struct scsi_eh_save *ses, |
| 672 | unsigned char *cmnd, int cmnd_size, unsigned sense_bytes) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 673 | { |
| f59114b | 2005-04-17 15:00:23 -0500 | [diff] [blame] | 674 | struct scsi_device *sdev = scmd->device; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 675 | |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 676 | /* |
| 677 | * We need saved copies of a number of fields - this is because |
| 678 | * error handling may need to overwrite these with different values |
| 679 | * to run different commands, and once error handling is complete, |
| 680 | * we will need to restore these values prior to running the actual |
| 681 | * command. |
| 682 | */ |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 683 | ses->cmd_len = scmd->cmd_len; |
Boaz Harrosh | 64a87b2 | 2008-04-30 11:19:47 +0300 | [diff] [blame] | 684 | ses->cmnd = scmd->cmnd; |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 685 | ses->data_direction = scmd->sc_data_direction; |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 686 | ses->sdb = scmd->sdb; |
Boaz Harrosh | 6f9a35e | 2007-12-13 13:50:53 +0200 | [diff] [blame] | 687 | ses->next_rq = scmd->request->next_rq; |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 688 | ses->result = scmd->result; |
Alan Stern | 1226570 | 2008-07-21 10:25:52 -0400 | [diff] [blame] | 689 | ses->underflow = scmd->underflow; |
Martin K. Petersen | db007fc | 2008-07-17 04:28:31 -0400 | [diff] [blame] | 690 | ses->prot_op = scmd->prot_op; |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 691 | |
Martin K. Petersen | db007fc | 2008-07-17 04:28:31 -0400 | [diff] [blame] | 692 | scmd->prot_op = SCSI_PROT_NORMAL; |
Boaz Harrosh | 64a87b2 | 2008-04-30 11:19:47 +0300 | [diff] [blame] | 693 | scmd->cmnd = ses->eh_cmnd; |
| 694 | memset(scmd->cmnd, 0, BLK_MAX_CDB); |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 695 | memset(&scmd->sdb, 0, sizeof(scmd->sdb)); |
Boaz Harrosh | 6f9a35e | 2007-12-13 13:50:53 +0200 | [diff] [blame] | 696 | scmd->request->next_rq = NULL; |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 697 | |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 698 | if (sense_bytes) { |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 699 | scmd->sdb.length = min_t(unsigned, SCSI_SENSE_BUFFERSIZE, |
| 700 | sense_bytes); |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 701 | sg_init_one(&ses->sense_sgl, scmd->sense_buffer, |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 702 | scmd->sdb.length); |
| 703 | scmd->sdb.table.sgl = &ses->sense_sgl; |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 704 | scmd->sc_data_direction = DMA_FROM_DEVICE; |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 705 | scmd->sdb.table.nents = 1; |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 706 | scmd->cmnd[0] = REQUEST_SENSE; |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 707 | scmd->cmnd[4] = scmd->sdb.length; |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 708 | scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]); |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 709 | } else { |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 710 | scmd->sc_data_direction = DMA_NONE; |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 711 | if (cmnd) { |
Boaz Harrosh | 64a87b2 | 2008-04-30 11:19:47 +0300 | [diff] [blame] | 712 | BUG_ON(cmnd_size > BLK_MAX_CDB); |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 713 | memcpy(scmd->cmnd, cmnd, cmnd_size); |
| 714 | scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]); |
| 715 | } |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 716 | } |
| 717 | |
| 718 | scmd->underflow = 0; |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 719 | |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 720 | if (sdev->scsi_level <= SCSI_2 && sdev->scsi_level != SCSI_UNKNOWN) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 721 | scmd->cmnd[1] = (scmd->cmnd[1] & 0x1f) | |
| f59114b | 2005-04-17 15:00:23 -0500 | [diff] [blame] | 722 | (sdev->lun << 5 & 0xe0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 723 | |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 724 | /* |
| 725 | * Zero the sense buffer. The scsi spec mandates that any |
| 726 | * untransferred sense data should be interpreted as being zero. |
| 727 | */ |
FUJITA Tomonori | b80ca4f | 2008-01-13 15:46:13 +0900 | [diff] [blame] | 728 | memset(scmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE); |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 729 | } |
| 730 | EXPORT_SYMBOL(scsi_eh_prep_cmnd); |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 731 | |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 732 | /** |
| 733 | * scsi_eh_restore_cmnd - Restore a scsi command info as part of error recory |
| 734 | * @scmd: SCSI command structure to restore |
Bartlomiej Zolnierkiewicz | 477e608 | 2009-04-27 20:54:22 +0200 | [diff] [blame] | 735 | * @ses: saved information from a coresponding call to scsi_eh_prep_cmnd |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 736 | * |
Bartlomiej Zolnierkiewicz | 477e608 | 2009-04-27 20:54:22 +0200 | [diff] [blame] | 737 | * Undo any damage done by above scsi_eh_prep_cmnd(). |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 738 | */ |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 739 | void scsi_eh_restore_cmnd(struct scsi_cmnd* scmd, struct scsi_eh_save *ses) |
| 740 | { |
| 741 | /* |
| 742 | * Restore original data |
| 743 | */ |
| 744 | scmd->cmd_len = ses->cmd_len; |
Boaz Harrosh | 64a87b2 | 2008-04-30 11:19:47 +0300 | [diff] [blame] | 745 | scmd->cmnd = ses->cmnd; |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 746 | scmd->sc_data_direction = ses->data_direction; |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 747 | scmd->sdb = ses->sdb; |
Boaz Harrosh | 6f9a35e | 2007-12-13 13:50:53 +0200 | [diff] [blame] | 748 | scmd->request->next_rq = ses->next_rq; |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 749 | scmd->result = ses->result; |
Alan Stern | 1226570 | 2008-07-21 10:25:52 -0400 | [diff] [blame] | 750 | scmd->underflow = ses->underflow; |
Martin K. Petersen | db007fc | 2008-07-17 04:28:31 -0400 | [diff] [blame] | 751 | scmd->prot_op = ses->prot_op; |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 752 | } |
| 753 | EXPORT_SYMBOL(scsi_eh_restore_cmnd); |
| 754 | |
| 755 | /** |
| 756 | * scsi_send_eh_cmnd - submit a scsi command as part of error recory |
| 757 | * @scmd: SCSI command structure to hijack |
| 758 | * @cmnd: CDB to send |
| 759 | * @cmnd_size: size in bytes of @cmnd |
| 760 | * @timeout: timeout for this request |
| 761 | * @sense_bytes: size of sense data to copy or 0 |
| 762 | * |
| 763 | * This function is used to send a scsi command down to a target device |
| 764 | * as part of the error recovery process. See also scsi_eh_prep_cmnd() above. |
| 765 | * |
| 766 | * Return value: |
| 767 | * SUCCESS or FAILED or NEEDS_RETRY |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 768 | */ |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 769 | static int scsi_send_eh_cmnd(struct scsi_cmnd *scmd, unsigned char *cmnd, |
| 770 | int cmnd_size, int timeout, unsigned sense_bytes) |
| 771 | { |
| 772 | struct scsi_device *sdev = scmd->device; |
| 773 | struct Scsi_Host *shost = sdev->host; |
| 774 | DECLARE_COMPLETION_ONSTACK(done); |
| 775 | unsigned long timeleft; |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 776 | struct scsi_eh_save ses; |
| 777 | int rtn; |
| 778 | |
| 779 | scsi_eh_prep_cmnd(scmd, &ses, cmnd, cmnd_size, sense_bytes); |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 780 | shost->eh_action = &done; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 781 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 782 | scsi_log_send(scmd); |
Jeff Garzik | f281233 | 2010-11-16 02:10:29 -0500 | [diff] [blame] | 783 | scmd->scsi_done = scsi_eh_done; |
| 784 | shost->hostt->queuecommand(shost, scmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 785 | |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 786 | timeleft = wait_for_completion_timeout(&done, timeout); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 787 | |
| f59114b | 2005-04-17 15:00:23 -0500 | [diff] [blame] | 788 | shost->eh_action = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 789 | |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 790 | scsi_log_completion(scmd, SUCCESS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 791 | |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 792 | SCSI_LOG_ERROR_RECOVERY(3, |
| 793 | printk("%s: scmd: %p, timeleft: %ld\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 794 | __func__, scmd, timeleft)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 795 | |
| 796 | /* |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 797 | * If there is time left scsi_eh_done got called, and we will |
| 798 | * examine the actual status codes to see whether the command |
| 799 | * actually did complete normally, else tell the host to forget |
| 800 | * about this command. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 801 | */ |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 802 | if (timeleft) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 803 | rtn = scsi_eh_completed_normally(scmd); |
| 804 | SCSI_LOG_ERROR_RECOVERY(3, |
| 805 | printk("%s: scsi_eh_completed_normally %x\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 806 | __func__, rtn)); |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 807 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 808 | switch (rtn) { |
| 809 | case SUCCESS: |
| 810 | case NEEDS_RETRY: |
| 811 | case FAILED: |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 812 | case TARGET_ERROR: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 813 | break; |
Hannes Reinecke | 6e883b0 | 2009-09-17 17:00:26 +0200 | [diff] [blame] | 814 | case ADD_TO_MLQUEUE: |
| 815 | rtn = NEEDS_RETRY; |
| 816 | break; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 817 | default: |
| 818 | rtn = FAILED; |
| 819 | break; |
| 820 | } |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 821 | } else { |
Brian King | 292148f | 2007-01-30 17:51:17 -0600 | [diff] [blame] | 822 | scsi_abort_eh_cmnd(scmd); |
Christoph Hellwig | 7dfdc9a | 2005-10-31 18:49:52 +0100 | [diff] [blame] | 823 | rtn = FAILED; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 824 | } |
| 825 | |
Boaz Harrosh | e1c2346 | 2007-10-08 16:36:45 +0200 | [diff] [blame] | 826 | scsi_eh_restore_cmnd(scmd, &ses); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 827 | return rtn; |
| 828 | } |
| 829 | |
| 830 | /** |
| 831 | * scsi_request_sense - Request sense data from a particular target. |
| 832 | * @scmd: SCSI cmd for request sense. |
| 833 | * |
| 834 | * Notes: |
| 835 | * Some hosts automatically obtain this information, others require |
| 836 | * that we obtain it on our own. This function will *not* return until |
| 837 | * the command either times out, or it completes. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 838 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 839 | static int scsi_request_sense(struct scsi_cmnd *scmd) |
| 840 | { |
Boaz Harrosh | 55db6c1 | 2007-10-08 16:35:19 +0200 | [diff] [blame] | 841 | return scsi_send_eh_cmnd(scmd, NULL, 0, SENSE_TIMEOUT, ~0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 842 | } |
| 843 | |
| 844 | /** |
| 845 | * scsi_eh_finish_cmd - Handle a cmd that eh is finished with. |
| 846 | * @scmd: Original SCSI cmd that eh has finished. |
| 847 | * @done_q: Queue for processed commands. |
| 848 | * |
| 849 | * Notes: |
| 850 | * We don't want to use the normal command completion while we are are |
| 851 | * still handling errors - it may cause other commands to be queued, |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 852 | * and that would disturb what we are doing. Thus we really want to |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 853 | * keep a list of pending commands for final completion, and once we |
| 854 | * are ready to leave error handling we handle completion for real. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 855 | */ |
Tejun Heo | 041c5fc | 2006-01-23 13:09:36 +0900 | [diff] [blame] | 856 | void scsi_eh_finish_cmd(struct scsi_cmnd *scmd, struct list_head *done_q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 857 | { |
| 858 | scmd->device->host->host_failed--; |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 859 | scmd->eh_eflags = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 860 | list_move_tail(&scmd->eh_entry, done_q); |
| 861 | } |
Tejun Heo | 041c5fc | 2006-01-23 13:09:36 +0900 | [diff] [blame] | 862 | EXPORT_SYMBOL(scsi_eh_finish_cmd); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 863 | |
| 864 | /** |
| 865 | * scsi_eh_get_sense - Get device sense data. |
| 866 | * @work_q: Queue of commands to process. |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 867 | * @done_q: Queue of processed commands. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 868 | * |
| 869 | * Description: |
| 870 | * See if we need to request sense information. if so, then get it |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 871 | * now, so we have a better idea of what to do. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 872 | * |
| 873 | * Notes: |
| 874 | * This has the unfortunate side effect that if a shost adapter does |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 875 | * not automatically request sense information, we end up shutting |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 876 | * it down before we request it. |
| 877 | * |
| 878 | * All drivers should request sense information internally these days, |
| 879 | * so for now all I have to say is tough noogies if you end up in here. |
| 880 | * |
| 881 | * XXX: Long term this code should go away, but that needs an audit of |
| 882 | * all LLDDs first. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 883 | */ |
Darrick J. Wong | dca84e4 | 2007-01-26 14:08:49 -0800 | [diff] [blame] | 884 | int scsi_eh_get_sense(struct list_head *work_q, |
| 885 | struct list_head *done_q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 886 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 887 | struct scsi_cmnd *scmd, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 888 | int rtn; |
| 889 | |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 890 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 891 | if ((scmd->eh_eflags & SCSI_EH_CANCEL_CMD) || |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 892 | SCSI_SENSE_VALID(scmd)) |
| 893 | continue; |
| 894 | |
Jeff Garzik | 3bf743e | 2005-10-24 18:04:06 -0400 | [diff] [blame] | 895 | SCSI_LOG_ERROR_RECOVERY(2, scmd_printk(KERN_INFO, scmd, |
| 896 | "%s: requesting sense\n", |
| 897 | current->comm)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 898 | rtn = scsi_request_sense(scmd); |
| 899 | if (rtn != SUCCESS) |
| 900 | continue; |
| 901 | |
| 902 | SCSI_LOG_ERROR_RECOVERY(3, printk("sense requested for %p" |
| 903 | " result %x\n", scmd, |
| 904 | scmd->result)); |
| 905 | SCSI_LOG_ERROR_RECOVERY(3, scsi_print_sense("bh", scmd)); |
| 906 | |
| 907 | rtn = scsi_decide_disposition(scmd); |
| 908 | |
| 909 | /* |
| 910 | * if the result was normal, then just pass it along to the |
| 911 | * upper level. |
| 912 | */ |
| 913 | if (rtn == SUCCESS) |
| 914 | /* we don't want this command reissued, just |
| 915 | * finished with the sense data, so set |
| 916 | * retries to the max allowed to ensure it |
| 917 | * won't get reissued */ |
| 918 | scmd->retries = scmd->allowed; |
| 919 | else if (rtn != NEEDS_RETRY) |
| 920 | continue; |
| 921 | |
| 922 | scsi_eh_finish_cmd(scmd, done_q); |
| 923 | } |
| 924 | |
| 925 | return list_empty(work_q); |
| 926 | } |
Darrick J. Wong | dca84e4 | 2007-01-26 14:08:49 -0800 | [diff] [blame] | 927 | EXPORT_SYMBOL_GPL(scsi_eh_get_sense); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 928 | |
| 929 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 930 | * scsi_eh_tur - Send TUR to device. |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 931 | * @scmd: &scsi_cmnd to send TUR |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 932 | * |
| 933 | * Return value: |
| 934 | * 0 - Device is ready. 1 - Device NOT ready. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 935 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 936 | static int scsi_eh_tur(struct scsi_cmnd *scmd) |
| 937 | { |
| 938 | static unsigned char tur_command[6] = {TEST_UNIT_READY, 0, 0, 0, 0, 0}; |
| 939 | int retry_cnt = 1, rtn; |
| 940 | |
| 941 | retry_tur: |
Mike Christie | 0db99e3 | 2006-08-26 03:00:22 -0400 | [diff] [blame] | 942 | rtn = scsi_send_eh_cmnd(scmd, tur_command, 6, SENSE_TIMEOUT, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 943 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 944 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: scmd %p rtn %x\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 945 | __func__, scmd, rtn)); |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 946 | |
| 947 | switch (rtn) { |
| 948 | case NEEDS_RETRY: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 949 | if (retry_cnt--) |
| 950 | goto retry_tur; |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 951 | /*FALLTHRU*/ |
| 952 | case SUCCESS: |
Alan Stern | e47373e | 2005-03-30 15:05:45 -0500 | [diff] [blame] | 953 | return 0; |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 954 | default: |
| 955 | return 1; |
Alan Stern | e47373e | 2005-03-30 15:05:45 -0500 | [diff] [blame] | 956 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 957 | } |
| 958 | |
| 959 | /** |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 960 | * scsi_eh_test_devices - check if devices are responding from error recovery. |
| 961 | * @cmd_list: scsi commands in error recovery. |
| 962 | * @work_q: queue for commands which still need more error recovery |
| 963 | * @done_q: queue for commands which are finished |
| 964 | * @try_stu: boolean on if a STU command should be tried in addition to TUR. |
| 965 | * |
| 966 | * Decription: |
| 967 | * Tests if devices are in a working state. Commands to devices now in |
| 968 | * a working state are sent to the done_q while commands to devices which |
| 969 | * are still failing to respond are returned to the work_q for more |
| 970 | * processing. |
| 971 | **/ |
| 972 | static int scsi_eh_test_devices(struct list_head *cmd_list, |
| 973 | struct list_head *work_q, |
| 974 | struct list_head *done_q, int try_stu) |
| 975 | { |
| 976 | struct scsi_cmnd *scmd, *next; |
| 977 | struct scsi_device *sdev; |
| 978 | int finish_cmds; |
| 979 | |
| 980 | while (!list_empty(cmd_list)) { |
| 981 | scmd = list_entry(cmd_list->next, struct scsi_cmnd, eh_entry); |
| 982 | sdev = scmd->device; |
| 983 | |
| 984 | finish_cmds = !scsi_device_online(scmd->device) || |
| 985 | (try_stu && !scsi_eh_try_stu(scmd) && |
| 986 | !scsi_eh_tur(scmd)) || |
| 987 | !scsi_eh_tur(scmd); |
| 988 | |
| 989 | list_for_each_entry_safe(scmd, next, cmd_list, eh_entry) |
| 990 | if (scmd->device == sdev) { |
| 991 | if (finish_cmds) |
| 992 | scsi_eh_finish_cmd(scmd, done_q); |
| 993 | else |
| 994 | list_move_tail(&scmd->eh_entry, work_q); |
| 995 | } |
| 996 | } |
| 997 | return list_empty(work_q); |
| 998 | } |
| 999 | |
| 1000 | |
| 1001 | /** |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1002 | * scsi_eh_abort_cmds - abort pending commands. |
| 1003 | * @work_q: &list_head for pending commands. |
| 1004 | * @done_q: &list_head for processed commands. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1005 | * |
| 1006 | * Decription: |
| 1007 | * Try and see whether or not it makes sense to try and abort the |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1008 | * running command. This only works out to be the case if we have one |
| 1009 | * command that has timed out. If the command simply failed, it makes |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1010 | * no sense to try and abort the command, since as far as the shost |
| 1011 | * adapter is concerned, it isn't running. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1012 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1013 | static int scsi_eh_abort_cmds(struct list_head *work_q, |
| 1014 | struct list_head *done_q) |
| 1015 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1016 | struct scsi_cmnd *scmd, *next; |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1017 | LIST_HEAD(check_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1018 | int rtn; |
| 1019 | |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1020 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 1021 | if (!(scmd->eh_eflags & SCSI_EH_CANCEL_CMD)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1022 | continue; |
| 1023 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: aborting cmd:" |
| 1024 | "0x%p\n", current->comm, |
| 1025 | scmd)); |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1026 | rtn = scsi_try_to_abort_cmd(scmd->device->host->hostt, scmd); |
Christof Schmitt | 2f2eb58 | 2010-03-24 16:50:30 +0100 | [diff] [blame] | 1027 | if (rtn == SUCCESS || rtn == FAST_IO_FAIL) { |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 1028 | scmd->eh_eflags &= ~SCSI_EH_CANCEL_CMD; |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1029 | if (rtn == FAST_IO_FAIL) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1030 | scsi_eh_finish_cmd(scmd, done_q); |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1031 | else |
| 1032 | list_move_tail(&scmd->eh_entry, &check_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1033 | } else |
| 1034 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: aborting" |
| 1035 | " cmd failed:" |
| 1036 | "0x%p\n", |
| 1037 | current->comm, |
| 1038 | scmd)); |
| 1039 | } |
| 1040 | |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1041 | return scsi_eh_test_devices(&check_list, work_q, done_q, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1042 | } |
| 1043 | |
| 1044 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1045 | * scsi_eh_try_stu - Send START_UNIT to device. |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1046 | * @scmd: &scsi_cmnd to send START_UNIT |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1047 | * |
| 1048 | * Return value: |
| 1049 | * 0 - Device is ready. 1 - Device NOT ready. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1050 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1051 | static int scsi_eh_try_stu(struct scsi_cmnd *scmd) |
| 1052 | { |
| 1053 | static unsigned char stu_command[6] = {START_STOP, 0, 0, 0, 1, 0}; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1054 | |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 1055 | if (scmd->device->allow_restart) { |
Brian King | ed773e6 | 2007-03-29 15:25:52 -0500 | [diff] [blame] | 1056 | int i, rtn = NEEDS_RETRY; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1057 | |
Brian King | ed773e6 | 2007-03-29 15:25:52 -0500 | [diff] [blame] | 1058 | for (i = 0; rtn == NEEDS_RETRY && i < 2; i++) |
James Bottomley | 9728c08 | 2008-11-30 10:32:26 -0600 | [diff] [blame] | 1059 | rtn = scsi_send_eh_cmnd(scmd, stu_command, 6, scmd->device->request_queue->rq_timeout, 0); |
Brian King | ed773e6 | 2007-03-29 15:25:52 -0500 | [diff] [blame] | 1060 | |
Christoph Hellwig | 631c228 | 2006-07-08 20:42:15 +0200 | [diff] [blame] | 1061 | if (rtn == SUCCESS) |
| 1062 | return 0; |
| 1063 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1064 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1065 | return 1; |
| 1066 | } |
| 1067 | |
| 1068 | /** |
| 1069 | * scsi_eh_stu - send START_UNIT if needed |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1070 | * @shost: &scsi host being recovered. |
| 1071 | * @work_q: &list_head for pending commands. |
| 1072 | * @done_q: &list_head for processed commands. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1073 | * |
| 1074 | * Notes: |
| 1075 | * If commands are failing due to not ready, initializing command required, |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1076 | * try revalidating the device, which will end up sending a start unit. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1077 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1078 | static int scsi_eh_stu(struct Scsi_Host *shost, |
| 1079 | struct list_head *work_q, |
| 1080 | struct list_head *done_q) |
| 1081 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1082 | struct scsi_cmnd *scmd, *stu_scmd, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1083 | struct scsi_device *sdev; |
| 1084 | |
| 1085 | shost_for_each_device(sdev, shost) { |
| 1086 | stu_scmd = NULL; |
| 1087 | list_for_each_entry(scmd, work_q, eh_entry) |
| 1088 | if (scmd->device == sdev && SCSI_SENSE_VALID(scmd) && |
| 1089 | scsi_check_sense(scmd) == FAILED ) { |
| 1090 | stu_scmd = scmd; |
| 1091 | break; |
| 1092 | } |
| 1093 | |
| 1094 | if (!stu_scmd) |
| 1095 | continue; |
| 1096 | |
| 1097 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending START_UNIT to sdev:" |
| 1098 | " 0x%p\n", current->comm, sdev)); |
| 1099 | |
| 1100 | if (!scsi_eh_try_stu(stu_scmd)) { |
| 1101 | if (!scsi_device_online(sdev) || |
| 1102 | !scsi_eh_tur(stu_scmd)) { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1103 | list_for_each_entry_safe(scmd, next, |
| 1104 | work_q, eh_entry) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1105 | if (scmd->device == sdev) |
| 1106 | scsi_eh_finish_cmd(scmd, done_q); |
| 1107 | } |
| 1108 | } |
| 1109 | } else { |
| 1110 | SCSI_LOG_ERROR_RECOVERY(3, |
| 1111 | printk("%s: START_UNIT failed to sdev:" |
| 1112 | " 0x%p\n", current->comm, sdev)); |
| 1113 | } |
| 1114 | } |
| 1115 | |
| 1116 | return list_empty(work_q); |
| 1117 | } |
| 1118 | |
| 1119 | |
| 1120 | /** |
| 1121 | * scsi_eh_bus_device_reset - send bdr if needed |
| 1122 | * @shost: scsi host being recovered. |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1123 | * @work_q: &list_head for pending commands. |
| 1124 | * @done_q: &list_head for processed commands. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1125 | * |
| 1126 | * Notes: |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1127 | * Try a bus device reset. Still, look to see whether we have multiple |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1128 | * devices that are jammed or not - if we have multiple devices, it |
| 1129 | * makes no sense to try bus_device_reset - we really would need to try |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1130 | * a bus_reset instead. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1131 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1132 | static int scsi_eh_bus_device_reset(struct Scsi_Host *shost, |
| 1133 | struct list_head *work_q, |
| 1134 | struct list_head *done_q) |
| 1135 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1136 | struct scsi_cmnd *scmd, *bdr_scmd, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1137 | struct scsi_device *sdev; |
| 1138 | int rtn; |
| 1139 | |
| 1140 | shost_for_each_device(sdev, shost) { |
| 1141 | bdr_scmd = NULL; |
| 1142 | list_for_each_entry(scmd, work_q, eh_entry) |
| 1143 | if (scmd->device == sdev) { |
| 1144 | bdr_scmd = scmd; |
| 1145 | break; |
| 1146 | } |
| 1147 | |
| 1148 | if (!bdr_scmd) |
| 1149 | continue; |
| 1150 | |
| 1151 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending BDR sdev:" |
| 1152 | " 0x%p\n", current->comm, |
| 1153 | sdev)); |
| 1154 | rtn = scsi_try_bus_device_reset(bdr_scmd); |
Christof Schmitt | 2f2eb58 | 2010-03-24 16:50:30 +0100 | [diff] [blame] | 1155 | if (rtn == SUCCESS || rtn == FAST_IO_FAIL) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1156 | if (!scsi_device_online(sdev) || |
Christof Schmitt | 2f2eb58 | 2010-03-24 16:50:30 +0100 | [diff] [blame] | 1157 | rtn == FAST_IO_FAIL || |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1158 | !scsi_eh_tur(bdr_scmd)) { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1159 | list_for_each_entry_safe(scmd, next, |
| 1160 | work_q, eh_entry) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1161 | if (scmd->device == sdev) |
| 1162 | scsi_eh_finish_cmd(scmd, |
| 1163 | done_q); |
| 1164 | } |
| 1165 | } |
| 1166 | } else { |
| 1167 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: BDR" |
| 1168 | " failed sdev:" |
| 1169 | "0x%p\n", |
| 1170 | current->comm, |
| 1171 | sdev)); |
| 1172 | } |
| 1173 | } |
| 1174 | |
| 1175 | return list_empty(work_q); |
| 1176 | } |
| 1177 | |
| 1178 | /** |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1179 | * scsi_eh_target_reset - send target reset if needed |
| 1180 | * @shost: scsi host being recovered. |
| 1181 | * @work_q: &list_head for pending commands. |
| 1182 | * @done_q: &list_head for processed commands. |
| 1183 | * |
| 1184 | * Notes: |
| 1185 | * Try a target reset. |
| 1186 | */ |
| 1187 | static int scsi_eh_target_reset(struct Scsi_Host *shost, |
| 1188 | struct list_head *work_q, |
| 1189 | struct list_head *done_q) |
| 1190 | { |
James Bottomley | 98db519 | 2010-10-25 15:53:41 -0500 | [diff] [blame] | 1191 | LIST_HEAD(tmp_list); |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1192 | LIST_HEAD(check_list); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1193 | |
James Bottomley | 98db519 | 2010-10-25 15:53:41 -0500 | [diff] [blame] | 1194 | list_splice_init(work_q, &tmp_list); |
| 1195 | |
| 1196 | while (!list_empty(&tmp_list)) { |
| 1197 | struct scsi_cmnd *next, *scmd; |
| 1198 | int rtn; |
| 1199 | unsigned int id; |
| 1200 | |
| 1201 | scmd = list_entry(tmp_list.next, struct scsi_cmnd, eh_entry); |
| 1202 | id = scmd_id(scmd); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1203 | |
| 1204 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending target reset " |
| 1205 | "to target %d\n", |
| 1206 | current->comm, id)); |
James Bottomley | 98db519 | 2010-10-25 15:53:41 -0500 | [diff] [blame] | 1207 | rtn = scsi_try_target_reset(scmd); |
| 1208 | if (rtn != SUCCESS && rtn != FAST_IO_FAIL) |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1209 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Target reset" |
| 1210 | " failed target: " |
| 1211 | "%d\n", |
| 1212 | current->comm, id)); |
James Bottomley | 98db519 | 2010-10-25 15:53:41 -0500 | [diff] [blame] | 1213 | list_for_each_entry_safe(scmd, next, &tmp_list, eh_entry) { |
| 1214 | if (scmd_id(scmd) != id) |
| 1215 | continue; |
| 1216 | |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1217 | if (rtn == SUCCESS) |
| 1218 | list_move_tail(&scmd->eh_entry, &check_list); |
| 1219 | else if (rtn == FAST_IO_FAIL) |
James Bottomley | 98db519 | 2010-10-25 15:53:41 -0500 | [diff] [blame] | 1220 | scsi_eh_finish_cmd(scmd, done_q); |
| 1221 | else |
| 1222 | /* push back on work queue for further processing */ |
| 1223 | list_move(&scmd->eh_entry, work_q); |
| 1224 | } |
| 1225 | } |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1226 | |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1227 | return scsi_eh_test_devices(&check_list, work_q, done_q, 0); |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1228 | } |
| 1229 | |
| 1230 | /** |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1231 | * scsi_eh_bus_reset - send a bus reset |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1232 | * @shost: &scsi host being recovered. |
| 1233 | * @work_q: &list_head for pending commands. |
| 1234 | * @done_q: &list_head for processed commands. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1235 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1236 | static int scsi_eh_bus_reset(struct Scsi_Host *shost, |
| 1237 | struct list_head *work_q, |
| 1238 | struct list_head *done_q) |
| 1239 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1240 | struct scsi_cmnd *scmd, *chan_scmd, *next; |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1241 | LIST_HEAD(check_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1242 | unsigned int channel; |
| 1243 | int rtn; |
| 1244 | |
| 1245 | /* |
| 1246 | * we really want to loop over the various channels, and do this on |
| 1247 | * a channel by channel basis. we should also check to see if any |
| 1248 | * of the failed commands are on soft_reset devices, and if so, skip |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1249 | * the reset. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1250 | */ |
| 1251 | |
| 1252 | for (channel = 0; channel <= shost->max_channel; channel++) { |
| 1253 | chan_scmd = NULL; |
| 1254 | list_for_each_entry(scmd, work_q, eh_entry) { |
Jeff Garzik | 422c0d6 | 2005-10-24 18:05:09 -0400 | [diff] [blame] | 1255 | if (channel == scmd_channel(scmd)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1256 | chan_scmd = scmd; |
| 1257 | break; |
| 1258 | /* |
| 1259 | * FIXME add back in some support for |
| 1260 | * soft_reset devices. |
| 1261 | */ |
| 1262 | } |
| 1263 | } |
| 1264 | |
| 1265 | if (!chan_scmd) |
| 1266 | continue; |
| 1267 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending BRST chan:" |
| 1268 | " %d\n", current->comm, |
| 1269 | channel)); |
| 1270 | rtn = scsi_try_bus_reset(chan_scmd); |
Christof Schmitt | 2f2eb58 | 2010-03-24 16:50:30 +0100 | [diff] [blame] | 1271 | if (rtn == SUCCESS || rtn == FAST_IO_FAIL) { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1272 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1273 | if (channel == scmd_channel(scmd)) { |
| 1274 | if (rtn == FAST_IO_FAIL) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1275 | scsi_eh_finish_cmd(scmd, |
| 1276 | done_q); |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1277 | else |
| 1278 | list_move_tail(&scmd->eh_entry, |
| 1279 | &check_list); |
| 1280 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1281 | } |
| 1282 | } else { |
| 1283 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: BRST" |
| 1284 | " failed chan: %d\n", |
| 1285 | current->comm, |
| 1286 | channel)); |
| 1287 | } |
| 1288 | } |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1289 | return scsi_eh_test_devices(&check_list, work_q, done_q, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1290 | } |
| 1291 | |
| 1292 | /** |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1293 | * scsi_eh_host_reset - send a host reset |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1294 | * @work_q: list_head for processed commands. |
| 1295 | * @done_q: list_head for processed commands. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1296 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1297 | static int scsi_eh_host_reset(struct list_head *work_q, |
| 1298 | struct list_head *done_q) |
| 1299 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1300 | struct scsi_cmnd *scmd, *next; |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1301 | LIST_HEAD(check_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1302 | int rtn; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1303 | |
| 1304 | if (!list_empty(work_q)) { |
| 1305 | scmd = list_entry(work_q->next, |
| 1306 | struct scsi_cmnd, eh_entry); |
| 1307 | |
| 1308 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending HRST\n" |
| 1309 | , current->comm)); |
| 1310 | |
| 1311 | rtn = scsi_try_host_reset(scmd); |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1312 | if (rtn == SUCCESS) { |
| 1313 | list_splice_init(work_q, &check_list); |
| 1314 | } else if (rtn == FAST_IO_FAIL) { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1315 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1316 | scsi_eh_finish_cmd(scmd, done_q); |
| 1317 | } |
| 1318 | } else { |
| 1319 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: HRST" |
| 1320 | " failed\n", |
| 1321 | current->comm)); |
| 1322 | } |
| 1323 | } |
David Jeffery | 3eef625 | 2011-05-19 14:41:12 -0400 | [diff] [blame] | 1324 | return scsi_eh_test_devices(&check_list, work_q, done_q, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1325 | } |
| 1326 | |
| 1327 | /** |
| 1328 | * scsi_eh_offline_sdevs - offline scsi devices that fail to recover |
| 1329 | * @work_q: list_head for processed commands. |
| 1330 | * @done_q: list_head for processed commands. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1331 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1332 | static void scsi_eh_offline_sdevs(struct list_head *work_q, |
| 1333 | struct list_head *done_q) |
| 1334 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1335 | struct scsi_cmnd *scmd, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1336 | |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1337 | list_for_each_entry_safe(scmd, next, work_q, eh_entry) { |
Matthew Wilcox | 31765d7 | 2007-08-17 11:02:10 -0600 | [diff] [blame] | 1338 | sdev_printk(KERN_INFO, scmd->device, "Device offlined - " |
| 1339 | "not ready after error recovery\n"); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1340 | scsi_device_set_state(scmd->device, SDEV_OFFLINE); |
Christoph Hellwig | 3111b0d | 2005-06-19 13:43:26 +0200 | [diff] [blame] | 1341 | if (scmd->eh_eflags & SCSI_EH_CANCEL_CMD) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1342 | /* |
| 1343 | * FIXME: Handle lost cmds. |
| 1344 | */ |
| 1345 | } |
| 1346 | scsi_eh_finish_cmd(scmd, done_q); |
| 1347 | } |
| 1348 | return; |
| 1349 | } |
| 1350 | |
| 1351 | /** |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1352 | * scsi_noretry_cmd - determinte if command should be failed fast |
| 1353 | * @scmd: SCSI cmd to examine. |
| 1354 | */ |
| 1355 | int scsi_noretry_cmd(struct scsi_cmnd *scmd) |
| 1356 | { |
| 1357 | switch (host_byte(scmd->result)) { |
| 1358 | case DID_OK: |
| 1359 | break; |
| 1360 | case DID_BUS_BUSY: |
Christoph Hellwig | 33659eb | 2010-08-07 18:17:56 +0200 | [diff] [blame] | 1361 | return (scmd->request->cmd_flags & REQ_FAILFAST_TRANSPORT); |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1362 | case DID_PARITY: |
Christoph Hellwig | 33659eb | 2010-08-07 18:17:56 +0200 | [diff] [blame] | 1363 | return (scmd->request->cmd_flags & REQ_FAILFAST_DEV); |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1364 | case DID_ERROR: |
| 1365 | if (msg_byte(scmd->result) == COMMAND_COMPLETE && |
| 1366 | status_byte(scmd->result) == RESERVATION_CONFLICT) |
| 1367 | return 0; |
| 1368 | /* fall through */ |
| 1369 | case DID_SOFT_ERROR: |
Christoph Hellwig | 33659eb | 2010-08-07 18:17:56 +0200 | [diff] [blame] | 1370 | return (scmd->request->cmd_flags & REQ_FAILFAST_DRIVER); |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1371 | } |
| 1372 | |
| 1373 | switch (status_byte(scmd->result)) { |
| 1374 | case CHECK_CONDITION: |
| 1375 | /* |
| 1376 | * assume caller has checked sense and determinted |
| 1377 | * the check condition was retryable. |
| 1378 | */ |
FUJITA Tomonori | e96f6ab | 2010-07-09 09:38:26 +0900 | [diff] [blame] | 1379 | if (scmd->request->cmd_flags & REQ_FAILFAST_DEV || |
| 1380 | scmd->request->cmd_type == REQ_TYPE_BLOCK_PC) |
| 1381 | return 1; |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1382 | } |
| 1383 | |
| 1384 | return 0; |
| 1385 | } |
| 1386 | |
| 1387 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1388 | * scsi_decide_disposition - Disposition a cmd on return from LLD. |
| 1389 | * @scmd: SCSI cmd to examine. |
| 1390 | * |
| 1391 | * Notes: |
| 1392 | * This is *only* called when we are examining the status after sending |
| 1393 | * out the actual data command. any commands that are queued for error |
| 1394 | * recovery (e.g. test_unit_ready) do *not* come through here. |
| 1395 | * |
| 1396 | * When this routine returns failed, it means the error handler thread |
| 1397 | * is woken. In cases where the error code indicates an error that |
| 1398 | * doesn't require the error handler read (i.e. we don't need to |
| 1399 | * abort/reset), this function should return SUCCESS. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1400 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1401 | int scsi_decide_disposition(struct scsi_cmnd *scmd) |
| 1402 | { |
| 1403 | int rtn; |
| 1404 | |
| 1405 | /* |
| 1406 | * if the device is offline, then we clearly just pass the result back |
| 1407 | * up to the top level. |
| 1408 | */ |
| 1409 | if (!scsi_device_online(scmd->device)) { |
| 1410 | SCSI_LOG_ERROR_RECOVERY(5, printk("%s: device offline - report" |
| 1411 | " as SUCCESS\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 1412 | __func__)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1413 | return SUCCESS; |
| 1414 | } |
| 1415 | |
| 1416 | /* |
| 1417 | * first check the host byte, to see if there is anything in there |
| 1418 | * that would indicate what we need to do. |
| 1419 | */ |
| 1420 | switch (host_byte(scmd->result)) { |
| 1421 | case DID_PASSTHROUGH: |
| 1422 | /* |
| 1423 | * no matter what, pass this through to the upper layer. |
| 1424 | * nuke this special code so that it looks like we are saying |
| 1425 | * did_ok. |
| 1426 | */ |
| 1427 | scmd->result &= 0xff00ffff; |
| 1428 | return SUCCESS; |
| 1429 | case DID_OK: |
| 1430 | /* |
| 1431 | * looks good. drop through, and check the next byte. |
| 1432 | */ |
| 1433 | break; |
| 1434 | case DID_NO_CONNECT: |
| 1435 | case DID_BAD_TARGET: |
| 1436 | case DID_ABORT: |
| 1437 | /* |
| 1438 | * note - this means that we just report the status back |
| 1439 | * to the top level driver, not that we actually think |
| 1440 | * that it indicates SUCCESS. |
| 1441 | */ |
| 1442 | return SUCCESS; |
| 1443 | /* |
| 1444 | * when the low level driver returns did_soft_error, |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 1445 | * it is responsible for keeping an internal retry counter |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1446 | * in order to avoid endless loops (db) |
| 1447 | * |
| 1448 | * actually this is a bug in this function here. we should |
| 1449 | * be mindful of the maximum number of retries specified |
| 1450 | * and not get stuck in a loop. |
| 1451 | */ |
| 1452 | case DID_SOFT_ERROR: |
| 1453 | goto maybe_retry; |
| 1454 | case DID_IMM_RETRY: |
| 1455 | return NEEDS_RETRY; |
| 1456 | |
| bf34191 | 2005-04-12 17:49:09 -0500 | [diff] [blame] | 1457 | case DID_REQUEUE: |
| 1458 | return ADD_TO_MLQUEUE; |
Mike Christie | a4dfaa6 | 2008-08-19 18:45:25 -0500 | [diff] [blame] | 1459 | case DID_TRANSPORT_DISRUPTED: |
| 1460 | /* |
| 1461 | * LLD/transport was disrupted during processing of the IO. |
| 1462 | * The transport class is now blocked/blocking, |
| 1463 | * and the transport will decide what to do with the IO |
Mike Christie | 939c228 | 2008-11-04 19:47:19 -0600 | [diff] [blame] | 1464 | * based on its timers and recovery capablilities if |
| 1465 | * there are enough retries. |
Mike Christie | a4dfaa6 | 2008-08-19 18:45:25 -0500 | [diff] [blame] | 1466 | */ |
Mike Christie | 939c228 | 2008-11-04 19:47:19 -0600 | [diff] [blame] | 1467 | goto maybe_retry; |
Mike Christie | a4dfaa6 | 2008-08-19 18:45:25 -0500 | [diff] [blame] | 1468 | case DID_TRANSPORT_FAILFAST: |
| 1469 | /* |
| 1470 | * The transport decided to failfast the IO (most likely |
| 1471 | * the fast io fail tmo fired), so send IO directly upwards. |
| 1472 | */ |
| 1473 | return SUCCESS; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1474 | case DID_ERROR: |
| 1475 | if (msg_byte(scmd->result) == COMMAND_COMPLETE && |
| 1476 | status_byte(scmd->result) == RESERVATION_CONFLICT) |
| 1477 | /* |
| 1478 | * execute reservation conflict processing code |
| 1479 | * lower down |
| 1480 | */ |
| 1481 | break; |
| 1482 | /* fallthrough */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1483 | case DID_BUS_BUSY: |
| 1484 | case DID_PARITY: |
| 1485 | goto maybe_retry; |
| 1486 | case DID_TIME_OUT: |
| 1487 | /* |
| 1488 | * when we scan the bus, we get timeout messages for |
| 1489 | * these commands if there is no device available. |
| 1490 | * other hosts report did_no_connect for the same thing. |
| 1491 | */ |
| 1492 | if ((scmd->cmnd[0] == TEST_UNIT_READY || |
| 1493 | scmd->cmnd[0] == INQUIRY)) { |
| 1494 | return SUCCESS; |
| 1495 | } else { |
| 1496 | return FAILED; |
| 1497 | } |
| 1498 | case DID_RESET: |
| 1499 | return SUCCESS; |
| 1500 | default: |
| 1501 | return FAILED; |
| 1502 | } |
| 1503 | |
| 1504 | /* |
| 1505 | * next, check the message byte. |
| 1506 | */ |
| 1507 | if (msg_byte(scmd->result) != COMMAND_COMPLETE) |
| 1508 | return FAILED; |
| 1509 | |
| 1510 | /* |
| 1511 | * check the status byte to see if this indicates anything special. |
| 1512 | */ |
| 1513 | switch (status_byte(scmd->result)) { |
| 1514 | case QUEUE_FULL: |
Mike Christie | 42a6a91 | 2009-10-15 17:46:44 -0700 | [diff] [blame] | 1515 | scsi_handle_queue_full(scmd->device); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1516 | /* |
| 1517 | * the case of trying to send too many commands to a |
| 1518 | * tagged queueing device. |
| 1519 | */ |
| 1520 | case BUSY: |
| 1521 | /* |
| 1522 | * device can't talk to us at the moment. Should only |
| 1523 | * occur (SAM-3) when the task queue is empty, so will cause |
| 1524 | * the empty queue handling to trigger a stall in the |
| 1525 | * device. |
| 1526 | */ |
| 1527 | return ADD_TO_MLQUEUE; |
| 1528 | case GOOD: |
Vasu Dev | 4a84067 | 2009-10-22 15:46:33 -0700 | [diff] [blame] | 1529 | scsi_handle_queue_ramp_up(scmd->device); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1530 | case COMMAND_TERMINATED: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1531 | return SUCCESS; |
Vladislav Bolkhovitin | a9b589d | 2008-11-06 13:57:52 +0300 | [diff] [blame] | 1532 | case TASK_ABORTED: |
| 1533 | goto maybe_retry; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1534 | case CHECK_CONDITION: |
| 1535 | rtn = scsi_check_sense(scmd); |
| 1536 | if (rtn == NEEDS_RETRY) |
| 1537 | goto maybe_retry; |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 1538 | else if (rtn == TARGET_ERROR) { |
| 1539 | /* |
| 1540 | * Need to modify host byte to signal a |
| 1541 | * permanent target failure |
| 1542 | */ |
| 1543 | scmd->result |= (DID_TARGET_FAILURE << 16); |
| 1544 | rtn = SUCCESS; |
| 1545 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1546 | /* if rtn == FAILED, we have no sense information; |
| 1547 | * returning FAILED will wake the error handler thread |
| 1548 | * to collect the sense and redo the decide |
| 1549 | * disposition */ |
| 1550 | return rtn; |
| 1551 | case CONDITION_GOOD: |
| 1552 | case INTERMEDIATE_GOOD: |
| 1553 | case INTERMEDIATE_C_GOOD: |
| 1554 | case ACA_ACTIVE: |
| 1555 | /* |
| 1556 | * who knows? FIXME(eric) |
| 1557 | */ |
| 1558 | return SUCCESS; |
| 1559 | |
| 1560 | case RESERVATION_CONFLICT: |
James Bottomley | 9ccfc75 | 2005-10-02 11:45:08 -0500 | [diff] [blame] | 1561 | sdev_printk(KERN_INFO, scmd->device, |
| 1562 | "reservation conflict\n"); |
Hannes Reinecke | 63583cc | 2011-01-18 10:13:11 +0100 | [diff] [blame] | 1563 | scmd->result |= (DID_NEXUS_FAILURE << 16); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1564 | return SUCCESS; /* causes immediate i/o error */ |
| 1565 | default: |
| 1566 | return FAILED; |
| 1567 | } |
| 1568 | return FAILED; |
| 1569 | |
| 1570 | maybe_retry: |
| 1571 | |
| 1572 | /* we requeue for retry because the error was retryable, and |
| 1573 | * the request was not marked fast fail. Note that above, |
| 1574 | * even if the request is marked fast fail, we still requeue |
| 1575 | * for queue congestion conditions (QUEUE_FULL or BUSY) */ |
Brian King | 8884efa | 2006-02-24 17:10:04 -0600 | [diff] [blame] | 1576 | if ((++scmd->retries) <= scmd->allowed |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1577 | && !scsi_noretry_cmd(scmd)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1578 | return NEEDS_RETRY; |
| 1579 | } else { |
| 1580 | /* |
| 1581 | * no more retries - report this one back to upper level. |
| 1582 | */ |
| 1583 | return SUCCESS; |
| 1584 | } |
| 1585 | } |
| 1586 | |
FUJITA Tomonori | f078727 | 2008-12-14 01:23:45 +0900 | [diff] [blame] | 1587 | static void eh_lock_door_done(struct request *req, int uptodate) |
| 1588 | { |
| 1589 | __blk_put_request(req->q, req); |
| 1590 | } |
| 1591 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1592 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1593 | * scsi_eh_lock_door - Prevent medium removal for the specified device |
| 1594 | * @sdev: SCSI device to prevent medium removal |
| 1595 | * |
| 1596 | * Locking: |
James Bottomley | 91bc31f | 2009-05-17 09:30:48 -0500 | [diff] [blame] | 1597 | * We must be called from process context. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1598 | * |
| 1599 | * Notes: |
| 1600 | * We queue up an asynchronous "ALLOW MEDIUM REMOVAL" request on the |
| 1601 | * head of the devices request queue, and continue. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1602 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1603 | static void scsi_eh_lock_door(struct scsi_device *sdev) |
| 1604 | { |
FUJITA Tomonori | f078727 | 2008-12-14 01:23:45 +0900 | [diff] [blame] | 1605 | struct request *req; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1606 | |
James Bottomley | 91bc31f | 2009-05-17 09:30:48 -0500 | [diff] [blame] | 1607 | /* |
| 1608 | * blk_get_request with GFP_KERNEL (__GFP_WAIT) sleeps until a |
| 1609 | * request becomes available |
| 1610 | */ |
FUJITA Tomonori | f078727 | 2008-12-14 01:23:45 +0900 | [diff] [blame] | 1611 | req = blk_get_request(sdev->request_queue, READ, GFP_KERNEL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1612 | |
FUJITA Tomonori | f078727 | 2008-12-14 01:23:45 +0900 | [diff] [blame] | 1613 | req->cmd[0] = ALLOW_MEDIUM_REMOVAL; |
| 1614 | req->cmd[1] = 0; |
| 1615 | req->cmd[2] = 0; |
| 1616 | req->cmd[3] = 0; |
| 1617 | req->cmd[4] = SCSI_REMOVAL_PREVENT; |
| 1618 | req->cmd[5] = 0; |
| 1619 | |
| 1620 | req->cmd_len = COMMAND_SIZE(req->cmd[0]); |
| 1621 | |
| 1622 | req->cmd_type = REQ_TYPE_BLOCK_PC; |
| 1623 | req->cmd_flags |= REQ_QUIET; |
| 1624 | req->timeout = 10 * HZ; |
| 1625 | req->retries = 5; |
| 1626 | |
| 1627 | blk_execute_rq_nowait(req->q, NULL, req, 1, eh_lock_door_done); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1628 | } |
| 1629 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1630 | /** |
| 1631 | * scsi_restart_operations - restart io operations to the specified host. |
| 1632 | * @shost: Host we are restarting. |
| 1633 | * |
| 1634 | * Notes: |
| 1635 | * When we entered the error handler, we blocked all further i/o to |
| 1636 | * this device. we need to 'reverse' this process. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1637 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1638 | static void scsi_restart_operations(struct Scsi_Host *shost) |
| 1639 | { |
| 1640 | struct scsi_device *sdev; |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 1641 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1642 | |
| 1643 | /* |
| 1644 | * If the door was locked, we need to insert a door lock request |
| 1645 | * onto the head of the SCSI request queue for the device. There |
| 1646 | * is no point trying to lock the door of an off-line device. |
| 1647 | */ |
| 1648 | shost_for_each_device(sdev, shost) { |
| 1649 | if (scsi_device_online(sdev) && sdev->locked) |
| 1650 | scsi_eh_lock_door(sdev); |
| 1651 | } |
| 1652 | |
| 1653 | /* |
| 1654 | * next free up anything directly waiting upon the host. this |
| 1655 | * will be requests for character device operations, and also for |
| 1656 | * ioctls to queued block devices. |
| 1657 | */ |
| 1658 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: waking up host to restart\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 1659 | __func__)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1660 | |
James Bottomley | 939647e | 2005-09-18 15:05:20 -0500 | [diff] [blame] | 1661 | spin_lock_irqsave(shost->host_lock, flags); |
| 1662 | if (scsi_host_set_state(shost, SHOST_RUNNING)) |
| 1663 | if (scsi_host_set_state(shost, SHOST_CANCEL)) |
| 1664 | BUG_ON(scsi_host_set_state(shost, SHOST_DEL)); |
| 1665 | spin_unlock_irqrestore(shost->host_lock, flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1666 | |
| 1667 | wake_up(&shost->host_wait); |
| 1668 | |
| 1669 | /* |
| 1670 | * finally we need to re-initiate requests that may be pending. we will |
| 1671 | * have had everything blocked while error handling is taking place, and |
| 1672 | * now that error recovery is done, we will need to ensure that these |
| 1673 | * requests are started. |
| 1674 | */ |
| 1675 | scsi_run_host_queues(shost); |
| 1676 | } |
| 1677 | |
| 1678 | /** |
| 1679 | * scsi_eh_ready_devs - check device ready state and recover if not. |
| 1680 | * @shost: host to be recovered. |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 1681 | * @work_q: &list_head for pending commands. |
| 1682 | * @done_q: &list_head for processed commands. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1683 | */ |
Darrick J. Wong | dca84e4 | 2007-01-26 14:08:49 -0800 | [diff] [blame] | 1684 | void scsi_eh_ready_devs(struct Scsi_Host *shost, |
| 1685 | struct list_head *work_q, |
| 1686 | struct list_head *done_q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1687 | { |
| 1688 | if (!scsi_eh_stu(shost, work_q, done_q)) |
| 1689 | if (!scsi_eh_bus_device_reset(shost, work_q, done_q)) |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1690 | if (!scsi_eh_target_reset(shost, work_q, done_q)) |
| 1691 | if (!scsi_eh_bus_reset(shost, work_q, done_q)) |
| 1692 | if (!scsi_eh_host_reset(work_q, done_q)) |
| 1693 | scsi_eh_offline_sdevs(work_q, |
| 1694 | done_q); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1695 | } |
Darrick J. Wong | dca84e4 | 2007-01-26 14:08:49 -0800 | [diff] [blame] | 1696 | EXPORT_SYMBOL_GPL(scsi_eh_ready_devs); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1697 | |
| 1698 | /** |
| 1699 | * scsi_eh_flush_done_q - finish processed commands or retry them. |
| 1700 | * @done_q: list_head of processed commands. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1701 | */ |
Tejun Heo | 041c5fc | 2006-01-23 13:09:36 +0900 | [diff] [blame] | 1702 | void scsi_eh_flush_done_q(struct list_head *done_q) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1703 | { |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1704 | struct scsi_cmnd *scmd, *next; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1705 | |
Christoph Hellwig | 937abeaa | 2005-06-19 13:43:56 +0200 | [diff] [blame] | 1706 | list_for_each_entry_safe(scmd, next, done_q, eh_entry) { |
| 1707 | list_del_init(&scmd->eh_entry); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1708 | if (scsi_device_online(scmd->device) && |
Mike Christie | 4a27446 | 2008-08-19 18:45:31 -0500 | [diff] [blame] | 1709 | !scsi_noretry_cmd(scmd) && |
Brian King | 8884efa | 2006-02-24 17:10:04 -0600 | [diff] [blame] | 1710 | (++scmd->retries <= scmd->allowed)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1711 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: flush" |
| 1712 | " retry cmd: %p\n", |
| 1713 | current->comm, |
| 1714 | scmd)); |
| 1715 | scsi_queue_insert(scmd, SCSI_MLQUEUE_EH_RETRY); |
| 1716 | } else { |
Patrick Mansfield | 793698c | 2005-05-16 17:42:15 -0700 | [diff] [blame] | 1717 | /* |
| 1718 | * If just we got sense for the device (called |
| 1719 | * scsi_eh_get_sense), scmd->result is already |
| 1720 | * set, do not set DRIVER_TIMEOUT. |
| 1721 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1722 | if (!scmd->result) |
| 1723 | scmd->result |= (DRIVER_TIMEOUT << 24); |
| 1724 | SCSI_LOG_ERROR_RECOVERY(3, printk("%s: flush finish" |
| 1725 | " cmd: %p\n", |
| 1726 | current->comm, scmd)); |
| 1727 | scsi_finish_command(scmd); |
| 1728 | } |
| 1729 | } |
| 1730 | } |
Tejun Heo | 041c5fc | 2006-01-23 13:09:36 +0900 | [diff] [blame] | 1731 | EXPORT_SYMBOL(scsi_eh_flush_done_q); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1732 | |
| 1733 | /** |
| 1734 | * scsi_unjam_host - Attempt to fix a host which has a cmd that failed. |
| 1735 | * @shost: Host to unjam. |
| 1736 | * |
| 1737 | * Notes: |
| 1738 | * When we come in here, we *know* that all commands on the bus have |
| 1739 | * either completed, failed or timed out. we also know that no further |
| 1740 | * commands are being sent to the host, so things are relatively quiet |
| 1741 | * and we have freedom to fiddle with things as we wish. |
| 1742 | * |
| 1743 | * This is only the *default* implementation. it is possible for |
| 1744 | * individual drivers to supply their own version of this function, and |
| 1745 | * if the maintainer wishes to do this, it is strongly suggested that |
| 1746 | * this function be taken as a template and modified. this function |
| 1747 | * was designed to correctly handle problems for about 95% of the |
| 1748 | * different cases out there, and it should always provide at least a |
| 1749 | * reasonable amount of error recovery. |
| 1750 | * |
| 1751 | * Any command marked 'failed' or 'timeout' must eventually have |
| 1752 | * scsi_finish_cmd() called for it. we do all of the retry stuff |
| 1753 | * here, so when we restart the host after we return it should have an |
| 1754 | * empty queue. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1755 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1756 | static void scsi_unjam_host(struct Scsi_Host *shost) |
| 1757 | { |
| 1758 | unsigned long flags; |
| 1759 | LIST_HEAD(eh_work_q); |
| 1760 | LIST_HEAD(eh_done_q); |
| 1761 | |
| 1762 | spin_lock_irqsave(shost->host_lock, flags); |
| 1763 | list_splice_init(&shost->eh_cmd_q, &eh_work_q); |
| 1764 | spin_unlock_irqrestore(shost->host_lock, flags); |
| 1765 | |
| 1766 | SCSI_LOG_ERROR_RECOVERY(1, scsi_eh_prt_fail_stats(shost, &eh_work_q)); |
| 1767 | |
| 1768 | if (!scsi_eh_get_sense(&eh_work_q, &eh_done_q)) |
| 1769 | if (!scsi_eh_abort_cmds(&eh_work_q, &eh_done_q)) |
| 1770 | scsi_eh_ready_devs(shost, &eh_work_q, &eh_done_q); |
| 1771 | |
| 1772 | scsi_eh_flush_done_q(&eh_done_q); |
| 1773 | } |
| 1774 | |
| 1775 | /** |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1776 | * scsi_error_handler - SCSI error handler thread |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1777 | * @data: Host for which we are running. |
| 1778 | * |
| 1779 | * Notes: |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1780 | * This is the main error handling loop. This is run as a kernel thread |
| 1781 | * for every SCSI host and handles all error handling activity. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 1782 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1783 | int scsi_error_handler(void *data) |
| 1784 | { |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1785 | struct Scsi_Host *shost = data; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1786 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1787 | /* |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1788 | * We use TASK_INTERRUPTIBLE so that the thread is not |
| 1789 | * counted against the load average as a running process. |
| 1790 | * We never actually get interrupted because kthread_run |
Frederik Schwarzer | c03264a | 2008-12-23 00:54:00 +0100 | [diff] [blame] | 1791 | * disables signal delivery for the created thread. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1792 | */ |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 1793 | set_current_state(TASK_INTERRUPTIBLE); |
| 1794 | while (!kthread_should_stop()) { |
Tejun Heo | ee7863b | 2006-05-15 20:57:20 +0900 | [diff] [blame] | 1795 | if ((shost->host_failed == 0 && shost->host_eh_scheduled == 0) || |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 1796 | shost->host_failed != shost->host_busy) { |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1797 | SCSI_LOG_ERROR_RECOVERY(1, |
| 1798 | printk("Error handler scsi_eh_%d sleeping\n", |
| 1799 | shost->host_no)); |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 1800 | schedule(); |
| 1801 | set_current_state(TASK_INTERRUPTIBLE); |
| 1802 | continue; |
| 1803 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1804 | |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 1805 | __set_current_state(TASK_RUNNING); |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1806 | SCSI_LOG_ERROR_RECOVERY(1, |
| 1807 | printk("Error handler scsi_eh_%d waking up\n", |
| 1808 | shost->host_no)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1809 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1810 | /* |
| 1811 | * We have a host that is failing for some reason. Figure out |
| 1812 | * what we need to do to get it up and online again (if we can). |
| 1813 | * If we fail, we end up taking the thing offline. |
| 1814 | */ |
Alan Stern | bc4f240 | 2010-06-17 10:41:42 -0400 | [diff] [blame] | 1815 | if (scsi_autopm_get_host(shost) != 0) { |
| 1816 | SCSI_LOG_ERROR_RECOVERY(1, |
| 1817 | printk(KERN_ERR "Error handler scsi_eh_%d " |
| 1818 | "unable to autoresume\n", |
| 1819 | shost->host_no)); |
| 1820 | continue; |
| 1821 | } |
| 1822 | |
Christoph Hellwig | 9227c33 | 2006-04-01 19:21:04 +0200 | [diff] [blame] | 1823 | if (shost->transportt->eh_strategy_handler) |
| 1824 | shost->transportt->eh_strategy_handler(shost); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1825 | else |
| 1826 | scsi_unjam_host(shost); |
| 1827 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1828 | /* |
| 1829 | * Note - if the above fails completely, the action is to take |
| 1830 | * individual devices offline and flush the queue of any |
| 1831 | * outstanding requests that may have been pending. When we |
| 1832 | * restart, we restart any I/O to any other devices on the bus |
| 1833 | * which are still online. |
| 1834 | */ |
| 1835 | scsi_restart_operations(shost); |
Alan Stern | bc4f240 | 2010-06-17 10:41:42 -0400 | [diff] [blame] | 1836 | scsi_autopm_put_host(shost); |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 1837 | set_current_state(TASK_INTERRUPTIBLE); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1838 | } |
Steven Rostedt | 461a0ff | 2005-10-19 08:22:13 -0400 | [diff] [blame] | 1839 | __set_current_state(TASK_RUNNING); |
| 1840 | |
Christoph Hellwig | ad42eb1 | 2005-10-29 01:01:55 +0200 | [diff] [blame] | 1841 | SCSI_LOG_ERROR_RECOVERY(1, |
| 1842 | printk("Error handler scsi_eh_%d exiting\n", shost->host_no)); |
James Bottomley | 3ed7a47 | 2005-09-19 09:50:04 -0500 | [diff] [blame] | 1843 | shost->ehandler = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1844 | return 0; |
| 1845 | } |
| 1846 | |
| 1847 | /* |
| 1848 | * Function: scsi_report_bus_reset() |
| 1849 | * |
| 1850 | * Purpose: Utility function used by low-level drivers to report that |
| 1851 | * they have observed a bus reset on the bus being handled. |
| 1852 | * |
| 1853 | * Arguments: shost - Host in question |
| 1854 | * channel - channel on which reset was observed. |
| 1855 | * |
| 1856 | * Returns: Nothing |
| 1857 | * |
| 1858 | * Lock status: Host lock must be held. |
| 1859 | * |
| 1860 | * Notes: This only needs to be called if the reset is one which |
| 1861 | * originates from an unknown location. Resets originated |
| 1862 | * by the mid-level itself don't need to call this, but there |
| 1863 | * should be no harm. |
| 1864 | * |
| 1865 | * The main purpose of this is to make sure that a CHECK_CONDITION |
| 1866 | * is properly treated. |
| 1867 | */ |
| 1868 | void scsi_report_bus_reset(struct Scsi_Host *shost, int channel) |
| 1869 | { |
| 1870 | struct scsi_device *sdev; |
| 1871 | |
| 1872 | __shost_for_each_device(sdev, shost) { |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1873 | if (channel == sdev_channel(sdev)) |
| 1874 | __scsi_report_device_reset(sdev, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1875 | } |
| 1876 | } |
| 1877 | EXPORT_SYMBOL(scsi_report_bus_reset); |
| 1878 | |
| 1879 | /* |
| 1880 | * Function: scsi_report_device_reset() |
| 1881 | * |
| 1882 | * Purpose: Utility function used by low-level drivers to report that |
| 1883 | * they have observed a device reset on the device being handled. |
| 1884 | * |
| 1885 | * Arguments: shost - Host in question |
| 1886 | * channel - channel on which reset was observed |
| 1887 | * target - target on which reset was observed |
| 1888 | * |
| 1889 | * Returns: Nothing |
| 1890 | * |
| 1891 | * Lock status: Host lock must be held |
| 1892 | * |
| 1893 | * Notes: This only needs to be called if the reset is one which |
| 1894 | * originates from an unknown location. Resets originated |
| 1895 | * by the mid-level itself don't need to call this, but there |
| 1896 | * should be no harm. |
| 1897 | * |
| 1898 | * The main purpose of this is to make sure that a CHECK_CONDITION |
| 1899 | * is properly treated. |
| 1900 | */ |
| 1901 | void scsi_report_device_reset(struct Scsi_Host *shost, int channel, int target) |
| 1902 | { |
| 1903 | struct scsi_device *sdev; |
| 1904 | |
| 1905 | __shost_for_each_device(sdev, shost) { |
Jeff Garzik | 422c0d6 | 2005-10-24 18:05:09 -0400 | [diff] [blame] | 1906 | if (channel == sdev_channel(sdev) && |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1907 | target == sdev_id(sdev)) |
| 1908 | __scsi_report_device_reset(sdev, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1909 | } |
| 1910 | } |
| 1911 | EXPORT_SYMBOL(scsi_report_device_reset); |
| 1912 | |
| 1913 | static void |
| 1914 | scsi_reset_provider_done_command(struct scsi_cmnd *scmd) |
| 1915 | { |
| 1916 | } |
| 1917 | |
| 1918 | /* |
| 1919 | * Function: scsi_reset_provider |
| 1920 | * |
| 1921 | * Purpose: Send requested reset to a bus or device at any phase. |
| 1922 | * |
| 1923 | * Arguments: device - device to send reset to |
| 1924 | * flag - reset type (see scsi.h) |
| 1925 | * |
| 1926 | * Returns: SUCCESS/FAILURE. |
| 1927 | * |
| 1928 | * Notes: This is used by the SCSI Generic driver to provide |
| 1929 | * Bus/Device reset capability. |
| 1930 | */ |
| 1931 | int |
| 1932 | scsi_reset_provider(struct scsi_device *dev, int flag) |
| 1933 | { |
Alan Stern | bc4f240 | 2010-06-17 10:41:42 -0400 | [diff] [blame] | 1934 | struct scsi_cmnd *scmd; |
James Smart | d7a1bb0 | 2006-03-08 14:50:12 -0500 | [diff] [blame] | 1935 | struct Scsi_Host *shost = dev->host; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1936 | struct request req; |
James Smart | d7a1bb0 | 2006-03-08 14:50:12 -0500 | [diff] [blame] | 1937 | unsigned long flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1938 | int rtn; |
| 1939 | |
Alan Stern | bc4f240 | 2010-06-17 10:41:42 -0400 | [diff] [blame] | 1940 | if (scsi_autopm_get_host(shost) < 0) |
| 1941 | return FAILED; |
| 1942 | |
| 1943 | scmd = scsi_get_command(dev, GFP_KERNEL); |
FUJITA Tomonori | 4f54eec | 2008-04-29 09:54:37 +0200 | [diff] [blame] | 1944 | blk_rq_init(NULL, &req); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1945 | scmd->request = &req; |
Christoph Hellwig | b4edcbc | 2005-06-19 13:40:52 +0200 | [diff] [blame] | 1946 | |
Boaz Harrosh | 64a87b2 | 2008-04-30 11:19:47 +0300 | [diff] [blame] | 1947 | scmd->cmnd = req.cmd; |
| 1948 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1949 | scmd->scsi_done = scsi_reset_provider_done_command; |
Boaz Harrosh | 30b0c37 | 2007-12-13 13:47:40 +0200 | [diff] [blame] | 1950 | memset(&scmd->sdb, 0, sizeof(scmd->sdb)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1951 | |
| 1952 | scmd->cmd_len = 0; |
| 1953 | |
| 1954 | scmd->sc_data_direction = DMA_BIDIRECTIONAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1955 | |
James Smart | d7a1bb0 | 2006-03-08 14:50:12 -0500 | [diff] [blame] | 1956 | spin_lock_irqsave(shost->host_lock, flags); |
| 1957 | shost->tmf_in_progress = 1; |
| 1958 | spin_unlock_irqrestore(shost->host_lock, flags); |
| 1959 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1960 | switch (flag) { |
| 1961 | case SCSI_TRY_RESET_DEVICE: |
| 1962 | rtn = scsi_try_bus_device_reset(scmd); |
| 1963 | if (rtn == SUCCESS) |
| 1964 | break; |
| 1965 | /* FALLTHROUGH */ |
Mike Christie | 30bd7df | 2008-02-29 18:25:19 -0600 | [diff] [blame] | 1966 | case SCSI_TRY_RESET_TARGET: |
| 1967 | rtn = scsi_try_target_reset(scmd); |
| 1968 | if (rtn == SUCCESS) |
| 1969 | break; |
| 1970 | /* FALLTHROUGH */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1971 | case SCSI_TRY_RESET_BUS: |
| 1972 | rtn = scsi_try_bus_reset(scmd); |
| 1973 | if (rtn == SUCCESS) |
| 1974 | break; |
| 1975 | /* FALLTHROUGH */ |
| 1976 | case SCSI_TRY_RESET_HOST: |
| 1977 | rtn = scsi_try_host_reset(scmd); |
| 1978 | break; |
| 1979 | default: |
| 1980 | rtn = FAILED; |
| 1981 | } |
| 1982 | |
James Smart | d7a1bb0 | 2006-03-08 14:50:12 -0500 | [diff] [blame] | 1983 | spin_lock_irqsave(shost->host_lock, flags); |
| 1984 | shost->tmf_in_progress = 0; |
| 1985 | spin_unlock_irqrestore(shost->host_lock, flags); |
| 1986 | |
| 1987 | /* |
| 1988 | * be sure to wake up anyone who was sleeping or had their queue |
| 1989 | * suspended while we performed the TMF. |
| 1990 | */ |
| 1991 | SCSI_LOG_ERROR_RECOVERY(3, |
| 1992 | printk("%s: waking up host to restart after TMF\n", |
Harvey Harrison | cadbd4a | 2008-07-03 23:47:27 -0700 | [diff] [blame] | 1993 | __func__)); |
James Smart | d7a1bb0 | 2006-03-08 14:50:12 -0500 | [diff] [blame] | 1994 | |
| 1995 | wake_up(&shost->host_wait); |
| 1996 | |
| 1997 | scsi_run_host_queues(shost); |
| 1998 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1999 | scsi_next_command(scmd); |
Alan Stern | bc4f240 | 2010-06-17 10:41:42 -0400 | [diff] [blame] | 2000 | scsi_autopm_put_host(shost); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2001 | return rtn; |
| 2002 | } |
| 2003 | EXPORT_SYMBOL(scsi_reset_provider); |
| 2004 | |
| 2005 | /** |
| 2006 | * scsi_normalize_sense - normalize main elements from either fixed or |
| 2007 | * descriptor sense data format into a common format. |
| 2008 | * |
| 2009 | * @sense_buffer: byte array containing sense data returned by device |
| 2010 | * @sb_len: number of valid bytes in sense_buffer |
| 2011 | * @sshdr: pointer to instance of structure that common |
| 2012 | * elements are written to. |
| 2013 | * |
| 2014 | * Notes: |
| 2015 | * The "main elements" from sense data are: response_code, sense_key, |
| 2016 | * asc, ascq and additional_length (only for descriptor format). |
| 2017 | * |
| 2018 | * Typically this function can be called after a device has |
| 2019 | * responded to a SCSI command with the CHECK_CONDITION status. |
| 2020 | * |
| 2021 | * Return value: |
| 2022 | * 1 if valid sense data information found, else 0; |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 2023 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2024 | int scsi_normalize_sense(const u8 *sense_buffer, int sb_len, |
| 2025 | struct scsi_sense_hdr *sshdr) |
| 2026 | { |
James Bottomley | 33aa687 | 2005-08-28 11:31:14 -0500 | [diff] [blame] | 2027 | if (!sense_buffer || !sb_len) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2028 | return 0; |
| 2029 | |
| 2030 | memset(sshdr, 0, sizeof(struct scsi_sense_hdr)); |
| 2031 | |
| 2032 | sshdr->response_code = (sense_buffer[0] & 0x7f); |
James Bottomley | 33aa687 | 2005-08-28 11:31:14 -0500 | [diff] [blame] | 2033 | |
| 2034 | if (!scsi_sense_valid(sshdr)) |
| 2035 | return 0; |
| 2036 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2037 | if (sshdr->response_code >= 0x72) { |
| 2038 | /* |
| 2039 | * descriptor format |
| 2040 | */ |
| 2041 | if (sb_len > 1) |
| 2042 | sshdr->sense_key = (sense_buffer[1] & 0xf); |
| 2043 | if (sb_len > 2) |
| 2044 | sshdr->asc = sense_buffer[2]; |
| 2045 | if (sb_len > 3) |
| 2046 | sshdr->ascq = sense_buffer[3]; |
| 2047 | if (sb_len > 7) |
| 2048 | sshdr->additional_length = sense_buffer[7]; |
| 2049 | } else { |
Jesper Juhl | 0bf8c86 | 2011-03-21 20:47:31 +0100 | [diff] [blame] | 2050 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2051 | * fixed format |
| 2052 | */ |
| 2053 | if (sb_len > 2) |
| 2054 | sshdr->sense_key = (sense_buffer[2] & 0xf); |
| 2055 | if (sb_len > 7) { |
| 2056 | sb_len = (sb_len < (sense_buffer[7] + 8)) ? |
| 2057 | sb_len : (sense_buffer[7] + 8); |
| 2058 | if (sb_len > 12) |
| 2059 | sshdr->asc = sense_buffer[12]; |
| 2060 | if (sb_len > 13) |
| 2061 | sshdr->ascq = sense_buffer[13]; |
| 2062 | } |
| 2063 | } |
| 2064 | |
| 2065 | return 1; |
| 2066 | } |
| 2067 | EXPORT_SYMBOL(scsi_normalize_sense); |
| 2068 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2069 | int scsi_command_normalize_sense(struct scsi_cmnd *cmd, |
| 2070 | struct scsi_sense_hdr *sshdr) |
| 2071 | { |
| 2072 | return scsi_normalize_sense(cmd->sense_buffer, |
FUJITA Tomonori | b80ca4f | 2008-01-13 15:46:13 +0900 | [diff] [blame] | 2073 | SCSI_SENSE_BUFFERSIZE, sshdr); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2074 | } |
| 2075 | EXPORT_SYMBOL(scsi_command_normalize_sense); |
| 2076 | |
| 2077 | /** |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 2078 | * scsi_sense_desc_find - search for a given descriptor type in descriptor sense data format. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2079 | * @sense_buffer: byte array of descriptor format sense data |
| 2080 | * @sb_len: number of valid bytes in sense_buffer |
| 2081 | * @desc_type: value of descriptor type to find |
| 2082 | * (e.g. 0 -> information) |
| 2083 | * |
| 2084 | * Notes: |
| 2085 | * only valid when sense data is in descriptor format |
| 2086 | * |
| 2087 | * Return value: |
| 2088 | * pointer to start of (first) descriptor if found else NULL |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 2089 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2090 | const u8 * scsi_sense_desc_find(const u8 * sense_buffer, int sb_len, |
| 2091 | int desc_type) |
| 2092 | { |
| 2093 | int add_sen_len, add_len, desc_len, k; |
| 2094 | const u8 * descp; |
| 2095 | |
| 2096 | if ((sb_len < 8) || (0 == (add_sen_len = sense_buffer[7]))) |
| 2097 | return NULL; |
| 2098 | if ((sense_buffer[0] < 0x72) || (sense_buffer[0] > 0x73)) |
| 2099 | return NULL; |
| 2100 | add_sen_len = (add_sen_len < (sb_len - 8)) ? |
| 2101 | add_sen_len : (sb_len - 8); |
| 2102 | descp = &sense_buffer[8]; |
| 2103 | for (desc_len = 0, k = 0; k < add_sen_len; k += desc_len) { |
| 2104 | descp += desc_len; |
| 2105 | add_len = (k < (add_sen_len - 1)) ? descp[1]: -1; |
| 2106 | desc_len = add_len + 2; |
| 2107 | if (descp[0] == desc_type) |
| 2108 | return descp; |
| 2109 | if (add_len < 0) // short descriptor ?? |
| 2110 | break; |
| 2111 | } |
| 2112 | return NULL; |
| 2113 | } |
| 2114 | EXPORT_SYMBOL(scsi_sense_desc_find); |
| 2115 | |
| 2116 | /** |
Rob Landley | eb44820 | 2007-11-03 13:30:39 -0500 | [diff] [blame] | 2117 | * scsi_get_sense_info_fld - get information field from sense data (either fixed or descriptor format) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2118 | * @sense_buffer: byte array of sense data |
| 2119 | * @sb_len: number of valid bytes in sense_buffer |
| 2120 | * @info_out: pointer to 64 integer where 8 or 4 byte information |
| 2121 | * field will be placed if found. |
| 2122 | * |
| 2123 | * Return value: |
| 2124 | * 1 if information field found, 0 if not found. |
Randy Dunlap | dc8875e | 2007-11-15 15:42:30 -0800 | [diff] [blame] | 2125 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2126 | int scsi_get_sense_info_fld(const u8 * sense_buffer, int sb_len, |
| 2127 | u64 * info_out) |
| 2128 | { |
| 2129 | int j; |
| 2130 | const u8 * ucp; |
| 2131 | u64 ull; |
| 2132 | |
| 2133 | if (sb_len < 7) |
| 2134 | return 0; |
| 2135 | switch (sense_buffer[0] & 0x7f) { |
| 2136 | case 0x70: |
| 2137 | case 0x71: |
| 2138 | if (sense_buffer[0] & 0x80) { |
| 2139 | *info_out = (sense_buffer[3] << 24) + |
| 2140 | (sense_buffer[4] << 16) + |
| 2141 | (sense_buffer[5] << 8) + sense_buffer[6]; |
| 2142 | return 1; |
| 2143 | } else |
| 2144 | return 0; |
| 2145 | case 0x72: |
| 2146 | case 0x73: |
| 2147 | ucp = scsi_sense_desc_find(sense_buffer, sb_len, |
| 2148 | 0 /* info desc */); |
| 2149 | if (ucp && (0xa == ucp[1])) { |
| 2150 | ull = 0; |
| 2151 | for (j = 0; j < 8; ++j) { |
| 2152 | if (j > 0) |
| 2153 | ull <<= 8; |
| 2154 | ull |= ucp[4 + j]; |
| 2155 | } |
| 2156 | *info_out = ull; |
| 2157 | return 1; |
| 2158 | } else |
| 2159 | return 0; |
| 2160 | default: |
| 2161 | return 0; |
| 2162 | } |
| 2163 | } |
| 2164 | EXPORT_SYMBOL(scsi_get_sense_info_fld); |
FUJITA Tomonori | 3bc6a26 | 2008-03-25 09:26:49 +0900 | [diff] [blame] | 2165 | |
| 2166 | /** |
| 2167 | * scsi_build_sense_buffer - build sense data in a buffer |
| 2168 | * @desc: Sense format (non zero == descriptor format, |
| 2169 | * 0 == fixed format) |
| 2170 | * @buf: Where to build sense data |
| 2171 | * @key: Sense key |
| 2172 | * @asc: Additional sense code |
| 2173 | * @ascq: Additional sense code qualifier |
| 2174 | * |
| 2175 | **/ |
| 2176 | void scsi_build_sense_buffer(int desc, u8 *buf, u8 key, u8 asc, u8 ascq) |
| 2177 | { |
| 2178 | if (desc) { |
| 2179 | buf[0] = 0x72; /* descriptor, current */ |
| 2180 | buf[1] = key; |
| 2181 | buf[2] = asc; |
| 2182 | buf[3] = ascq; |
| 2183 | buf[7] = 0; |
| 2184 | } else { |
| 2185 | buf[0] = 0x70; /* fixed, current */ |
| 2186 | buf[2] = key; |
| 2187 | buf[7] = 0xa; |
| 2188 | buf[12] = asc; |
| 2189 | buf[13] = ascq; |
| 2190 | } |
| 2191 | } |
| 2192 | EXPORT_SYMBOL(scsi_build_sense_buffer); |