scsi_error.c 60.8 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5
/*
 *  scsi_error.c Copyright (C) 1997 Eric Youngdale
 *
 *  SCSI error/timeout handling
 *      Initial versions: Eric Youngdale.  Based upon conversations with
6
 *                        Leonard Zubkoff and David Miller at Linux Expo,
L
Linus Torvalds 已提交
7 8 9 10 11 12
 *                        ideas originating from all over the place.
 *
 *	Restructured scsi_unjam_host and associated functions.
 *	September 04, 2002 Mike Anderson (andmike@us.ibm.com)
 *
 *	Forward port of Russell King's (rmk@arm.linux.org.uk) changes and
13
 *	minor cleanups.
L
Linus Torvalds 已提交
14 15 16 17 18
 *	September 30, 2002 Mike Anderson (andmike@us.ibm.com)
 */

#include <linux/module.h>
#include <linux/sched.h>
19
#include <linux/gfp.h>
L
Linus Torvalds 已提交
20 21 22
#include <linux/timer.h>
#include <linux/string.h>
#include <linux/kernel.h>
23
#include <linux/freezer.h>
24
#include <linux/kthread.h>
L
Linus Torvalds 已提交
25 26 27
#include <linux/interrupt.h>
#include <linux/blkdev.h>
#include <linux/delay.h>
28
#include <linux/jiffies.h>
L
Linus Torvalds 已提交
29 30

#include <scsi/scsi.h>
31
#include <scsi/scsi_cmnd.h>
L
Linus Torvalds 已提交
32 33
#include <scsi/scsi_dbg.h>
#include <scsi/scsi_device.h>
34
#include <scsi/scsi_driver.h>
L
Linus Torvalds 已提交
35
#include <scsi/scsi_eh.h>
36
#include <scsi/scsi_transport.h>
L
Linus Torvalds 已提交
37 38 39 40 41
#include <scsi/scsi_host.h>
#include <scsi/scsi_ioctl.h>

#include "scsi_priv.h"
#include "scsi_logging.h"
42
#include "scsi_transport_api.h"
L
Linus Torvalds 已提交
43

44 45
#include <trace/events/scsi.h>

46 47
static void scsi_eh_done(struct scsi_cmnd *scmd);

L
Linus Torvalds 已提交
48 49 50 51 52 53 54
/*
 * These should *probably* be handled by the host itself.
 * Since it is allowed to sleep, it probably should.
 */
#define BUS_RESET_SETTLE_TIME   (10)
#define HOST_RESET_SETTLE_TIME  (10)

55 56
static int scsi_eh_try_stu(struct scsi_cmnd *scmd);

L
Linus Torvalds 已提交
57 58 59 60
/* called with shost->host_lock held */
void scsi_eh_wakeup(struct Scsi_Host *shost)
{
	if (shost->host_busy == shost->host_failed) {
61
		trace_scsi_eh_wakeup(shost);
62
		wake_up_process(shost->ehandler);
L
Linus Torvalds 已提交
63 64 65 66
		SCSI_LOG_ERROR_RECOVERY(5,
				printk("Waking error handler thread\n"));
	}
}
67 68 69 70 71 72

/**
 * scsi_schedule_eh - schedule EH for SCSI host
 * @shost:	SCSI host to invoke error handling on.
 *
 * Schedule SCSI EH without scmd.
73
 */
74 75 76 77 78 79 80 81 82 83 84 85 86 87 88
void scsi_schedule_eh(struct Scsi_Host *shost)
{
	unsigned long flags;

	spin_lock_irqsave(shost->host_lock, flags);

	if (scsi_host_set_state(shost, SHOST_RECOVERY) == 0 ||
	    scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY) == 0) {
		shost->host_eh_scheduled++;
		scsi_eh_wakeup(shost);
	}

	spin_unlock_irqrestore(shost->host_lock, flags);
}
EXPORT_SYMBOL_GPL(scsi_schedule_eh);
L
Linus Torvalds 已提交
89 90 91 92 93 94 95 96

/**
 * scsi_eh_scmd_add - add scsi cmd to error handling.
 * @scmd:	scmd to run eh on.
 * @eh_flag:	optional SCSI_EH flag.
 *
 * Return value:
 *	0 on failure.
97
 */
L
Linus Torvalds 已提交
98 99 100 101
int scsi_eh_scmd_add(struct scsi_cmnd *scmd, int eh_flag)
{
	struct Scsi_Host *shost = scmd->device->host;
	unsigned long flags;
102
	int ret = 0;
L
Linus Torvalds 已提交
103

104
	if (!shost->ehandler)
L
Linus Torvalds 已提交
105 106 107
		return 0;

	spin_lock_irqsave(shost->host_lock, flags);
108 109 110
	if (scsi_host_set_state(shost, SHOST_RECOVERY))
		if (scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY))
			goto out_unlock;
L
Linus Torvalds 已提交
111

112
	ret = 1;
113
	scmd->eh_eflags |= eh_flag;
L
Linus Torvalds 已提交
114 115 116
	list_add_tail(&scmd->eh_entry, &shost->eh_cmd_q);
	shost->host_failed++;
	scsi_eh_wakeup(shost);
117
 out_unlock:
L
Linus Torvalds 已提交
118
	spin_unlock_irqrestore(shost->host_lock, flags);
119
	return ret;
L
Linus Torvalds 已提交
120 121 122 123
}

/**
 * scsi_times_out - Timeout function for normal scsi commands.
J
Jens Axboe 已提交
124
 * @req:	request that is timing out.
L
Linus Torvalds 已提交
125 126 127 128 129 130
 *
 * Notes:
 *     We do not need to lock this.  There is the potential for a race
 *     only in that the normal completion handling might run, but if the
 *     normal completion function determines that the timer has already
 *     fired, then it mustn't do anything.
131
 */
J
Jens Axboe 已提交
132
enum blk_eh_timer_return scsi_times_out(struct request *req)
L
Linus Torvalds 已提交
133
{
J
Jens Axboe 已提交
134 135
	struct scsi_cmnd *scmd = req->special;
	enum blk_eh_timer_return rtn = BLK_EH_NOT_HANDLED;
136
	struct Scsi_Host *host = scmd->device->host;
137

138
	trace_scsi_dispatch_cmd_timeout(scmd);
L
Linus Torvalds 已提交
139 140
	scsi_log_completion(scmd, TIMEOUT_ERROR);

141 142 143 144
	if (host->transportt->eh_timed_out)
		rtn = host->transportt->eh_timed_out(scmd);
	else if (host->hostt->eh_timed_out)
		rtn = host->hostt->eh_timed_out(scmd);
L
Linus Torvalds 已提交
145

146 147
	scmd->result |= DID_TIME_OUT << 16;

148
	if (unlikely(rtn == BLK_EH_NOT_HANDLED &&
149
		     !scsi_eh_scmd_add(scmd, SCSI_EH_CANCEL_CMD)))
150
		rtn = BLK_EH_HANDLED;
J
Jens Axboe 已提交
151

152
	return rtn;
L
Linus Torvalds 已提交
153 154 155 156 157 158 159 160 161 162 163 164
}

/**
 * scsi_block_when_processing_errors - Prevent cmds from being queued.
 * @sdev:	Device on which we are performing recovery.
 *
 * Description:
 *     We block until the host is out of error recovery, and then check to
 *     see whether the host or the device is offline.
 *
 * Return value:
 *     0 when dev was taken offline by error recovery. 1 OK to proceed.
165
 */
L
Linus Torvalds 已提交
166 167 168 169
int scsi_block_when_processing_errors(struct scsi_device *sdev)
{
	int online;

170
	wait_event(sdev->host->host_wait, !scsi_host_in_recovery(sdev->host));
L
Linus Torvalds 已提交
171 172 173

	online = scsi_device_online(sdev);

174
	SCSI_LOG_ERROR_RECOVERY(5, printk("%s: rtn: %d\n", __func__,
L
Linus Torvalds 已提交
175 176 177 178 179 180 181 182 183 184 185
					  online));

	return online;
}
EXPORT_SYMBOL(scsi_block_when_processing_errors);

#ifdef CONFIG_SCSI_LOGGING
/**
 * scsi_eh_prt_fail_stats - Log info on failures.
 * @shost:	scsi host being recovered.
 * @work_q:	Queue of scsi cmds to process.
186
 */
L
Linus Torvalds 已提交
187 188 189 190 191 192 193 194 195 196 197 198 199 200
static inline void scsi_eh_prt_fail_stats(struct Scsi_Host *shost,
					  struct list_head *work_q)
{
	struct scsi_cmnd *scmd;
	struct scsi_device *sdev;
	int total_failures = 0;
	int cmd_failed = 0;
	int cmd_cancel = 0;
	int devices_failed = 0;

	shost_for_each_device(sdev, shost) {
		list_for_each_entry(scmd, work_q, eh_entry) {
			if (scmd->device == sdev) {
				++total_failures;
201
				if (scmd->eh_eflags & SCSI_EH_CANCEL_CMD)
L
Linus Torvalds 已提交
202
					++cmd_cancel;
203
				else
L
Linus Torvalds 已提交
204 205 206 207 208 209
					++cmd_failed;
			}
		}

		if (cmd_cancel || cmd_failed) {
			SCSI_LOG_ERROR_RECOVERY(3,
210 211
				sdev_printk(KERN_INFO, sdev,
					    "%s: cmds failed: %d, cancel: %d\n",
212
					    __func__, cmd_failed,
213
					    cmd_cancel));
L
Linus Torvalds 已提交
214 215 216 217 218 219 220 221
			cmd_cancel = 0;
			cmd_failed = 0;
			++devices_failed;
		}
	}

	SCSI_LOG_ERROR_RECOVERY(2, printk("Total of %d commands on %d"
					  " devices require eh work\n",
222
				   total_failures, devices_failed));
L
Linus Torvalds 已提交
223 224 225 226 227 228 229 230
}
#endif

/**
 * scsi_check_sense - Examine scsi cmd sense
 * @scmd:	Cmd to have sense checked.
 *
 * Return value:
231
 *	SUCCESS or FAILED or NEEDS_RETRY or ADD_TO_MLQUEUE
L
Linus Torvalds 已提交
232 233 234 235
 *
 * Notes:
 *	When a deferred error is detected the current command has
 *	not been executed and needs retrying.
236
 */
L
Linus Torvalds 已提交
237 238
static int scsi_check_sense(struct scsi_cmnd *scmd)
{
239
	struct scsi_device *sdev = scmd->device;
L
Linus Torvalds 已提交
240 241 242 243 244
	struct scsi_sense_hdr sshdr;

	if (! scsi_command_normalize_sense(scmd, &sshdr))
		return FAILED;	/* no valid sense data */

245 246 247 248 249 250 251 252
	if (scmd->cmnd[0] == TEST_UNIT_READY && scmd->scsi_done != scsi_eh_done)
		/*
		 * nasty: for mid-layer issued TURs, we need to return the
		 * actual sense data without any recovery attempt.  For eh
		 * issued ones, we need to try to recover and interpret
		 */
		return SUCCESS;

L
Linus Torvalds 已提交
253 254 255
	if (scsi_sense_is_deferred(&sshdr))
		return NEEDS_RETRY;

256 257 258 259 260 261 262 263 264 265
	if (sdev->scsi_dh_data && sdev->scsi_dh_data->scsi_dh &&
			sdev->scsi_dh_data->scsi_dh->check_sense) {
		int rc;

		rc = sdev->scsi_dh_data->scsi_dh->check_sense(sdev, &sshdr);
		if (rc != SCSI_RETURN_NOT_HANDLED)
			return rc;
		/* handler does not care. Drop down to default handling */
	}

L
Linus Torvalds 已提交
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292
	/*
	 * Previous logic looked for FILEMARK, EOM or ILI which are
	 * mainly associated with tapes and returned SUCCESS.
	 */
	if (sshdr.response_code == 0x70) {
		/* fixed format */
		if (scmd->sense_buffer[2] & 0xe0)
			return SUCCESS;
	} else {
		/*
		 * descriptor format: look for "stream commands sense data
		 * descriptor" (see SSC-3). Assume single sense data
		 * descriptor. Ignore ILI from SBC-2 READ LONG and WRITE LONG.
		 */
		if ((sshdr.additional_length > 3) &&
		    (scmd->sense_buffer[8] == 0x4) &&
		    (scmd->sense_buffer[11] & 0xe0))
			return SUCCESS;
	}

	switch (sshdr.sense_key) {
	case NO_SENSE:
		return SUCCESS;
	case RECOVERED_ERROR:
		return /* soft_error */ SUCCESS;

	case ABORTED_COMMAND:
293 294 295
		if (sshdr.asc == 0x10) /* DIF */
			return SUCCESS;

L
Linus Torvalds 已提交
296 297 298 299 300 301 302 303 304 305
		return NEEDS_RETRY;
	case NOT_READY:
	case UNIT_ATTENTION:
		/*
		 * if we are expecting a cc/ua because of a bus reset that we
		 * performed, treat this just as a retry.  otherwise this is
		 * information that we should pass up to the upper-level driver
		 * so that we can deal with it there.
		 */
		if (scmd->device->expecting_cc_ua) {
306 307 308 309 310 311 312 313 314 315
			/*
			 * Because some device does not queue unit
			 * attentions correctly, we carefully check
			 * additional sense code and qualifier so as
			 * not to squash media change unit attention.
			 */
			if (sshdr.asc != 0x28 || sshdr.ascq != 0x00) {
				scmd->device->expecting_cc_ua = 0;
				return NEEDS_RETRY;
			}
L
Linus Torvalds 已提交
316 317
		}
		/*
318
		 * if the device is in the process of becoming ready, we
L
Linus Torvalds 已提交
319 320 321 322 323 324 325 326 327 328 329
		 * should retry.
		 */
		if ((sshdr.asc == 0x04) && (sshdr.ascq == 0x01))
			return NEEDS_RETRY;
		/*
		 * if the device is not started, we need to wake
		 * the error handler to start the motor
		 */
		if (scmd->device->allow_restart &&
		    (sshdr.asc == 0x04) && (sshdr.ascq == 0x02))
			return FAILED;
330

331 332 333 334 335 336 337 338 339 340 341 342 343
		if (sshdr.asc == 0x3f && sshdr.ascq == 0x0e)
			scmd_printk(KERN_WARNING, scmd,
				    "Warning! Received an indication that the "
				    "LUN assignments on this target have "
				    "changed. The Linux SCSI layer does not "
				    "automatically remap LUN assignments.\n");
		else if (sshdr.asc == 0x3f)
			scmd_printk(KERN_WARNING, scmd,
				    "Warning! Received an indication that the "
				    "operating parameters on this target have "
				    "changed. The Linux SCSI layer does not "
				    "automatically adjust these parameters.\n");

344 345 346 347 348 349
		if (sshdr.asc == 0x38 && sshdr.ascq == 0x07)
			scmd_printk(KERN_WARNING, scmd,
				    "Warning! Received an indication that the "
				    "LUN reached a thin provisioning soft "
				    "threshold.\n");

C
Christoph Hellwig 已提交
350 351 352 353 354
		/*
		 * Pass the UA upwards for a determination in the completion
		 * functions.
		 */
		return SUCCESS;
L
Linus Torvalds 已提交
355

356
		/* these are not supported */
357 358 359 360 361 362
	case DATA_PROTECT:
		if (sshdr.asc == 0x27 && sshdr.ascq == 0x07) {
			/* Thin provisioning hard threshold reached */
			set_host_byte(scmd, DID_ALLOC_FAILURE);
			return SUCCESS;
		}
L
Linus Torvalds 已提交
363 364 365
	case COPY_ABORTED:
	case VOLUME_OVERFLOW:
	case MISCOMPARE:
366
	case BLANK_CHECK:
367 368
		set_host_byte(scmd, DID_TARGET_FAILURE);
		return SUCCESS;
L
Linus Torvalds 已提交
369 370

	case MEDIUM_ERROR:
371 372 373
		if (sshdr.asc == 0x11 || /* UNRECOVERED READ ERR */
		    sshdr.asc == 0x13 || /* AMNF DATA FIELD */
		    sshdr.asc == 0x14) { /* RECORD NOT FOUND */
374 375
			set_host_byte(scmd, DID_TARGET_FAILURE);
			return SUCCESS;
376
		}
L
Linus Torvalds 已提交
377 378 379 380
		return NEEDS_RETRY;

	case HARDWARE_ERROR:
		if (scmd->device->retry_hwerror)
381
			return ADD_TO_MLQUEUE;
L
Linus Torvalds 已提交
382
		else
383
			set_host_byte(scmd, DID_TARGET_FAILURE);
L
Linus Torvalds 已提交
384 385

	case ILLEGAL_REQUEST:
386 387 388 389
		if (sshdr.asc == 0x20 || /* Invalid command operation code */
		    sshdr.asc == 0x21 || /* Logical block address out of range */
		    sshdr.asc == 0x24 || /* Invalid field in cdb */
		    sshdr.asc == 0x26) { /* Parameter value invalid */
390
			set_host_byte(scmd, DID_TARGET_FAILURE);
391 392 393
		}
		return SUCCESS;

L
Linus Torvalds 已提交
394 395 396 397 398
	default:
		return SUCCESS;
	}
}

V
Vasu Dev 已提交
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
static void scsi_handle_queue_ramp_up(struct scsi_device *sdev)
{
	struct scsi_host_template *sht = sdev->host->hostt;
	struct scsi_device *tmp_sdev;

	if (!sht->change_queue_depth ||
	    sdev->queue_depth >= sdev->max_queue_depth)
		return;

	if (time_before(jiffies,
	    sdev->last_queue_ramp_up + sdev->queue_ramp_up_period))
		return;

	if (time_before(jiffies,
	    sdev->last_queue_full_time + sdev->queue_ramp_up_period))
		return;

	/*
	 * Walk all devices of a target and do
	 * ramp up on them.
	 */
	shost_for_each_device(tmp_sdev, sdev->host) {
		if (tmp_sdev->channel != sdev->channel ||
		    tmp_sdev->id != sdev->id ||
		    tmp_sdev->queue_depth == sdev->max_queue_depth)
			continue;
		/*
		 * call back into LLD to increase queue_depth by one
		 * with ramp up reason code.
		 */
		sht->change_queue_depth(tmp_sdev, tmp_sdev->queue_depth + 1,
					SCSI_QDEPTH_RAMP_UP);
		sdev->last_queue_ramp_up = jiffies;
	}
}

435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456
static void scsi_handle_queue_full(struct scsi_device *sdev)
{
	struct scsi_host_template *sht = sdev->host->hostt;
	struct scsi_device *tmp_sdev;

	if (!sht->change_queue_depth)
		return;

	shost_for_each_device(tmp_sdev, sdev->host) {
		if (tmp_sdev->channel != sdev->channel ||
		    tmp_sdev->id != sdev->id)
			continue;
		/*
		 * We do not know the number of commands that were at
		 * the device when we got the queue full so we start
		 * from the highest possible value and work our way down.
		 */
		sht->change_queue_depth(tmp_sdev, tmp_sdev->queue_depth - 1,
					SCSI_QDEPTH_QFULL);
	}
}

L
Linus Torvalds 已提交
457 458 459 460 461 462 463 464 465
/**
 * scsi_eh_completed_normally - Disposition a eh cmd on return from LLD.
 * @scmd:	SCSI cmd to examine.
 *
 * Notes:
 *    This is *only* called when we are examining the status of commands
 *    queued during error recovery.  the main difference here is that we
 *    don't allow for the possibility of retries here, and we are a lot
 *    more restrictive about what we consider acceptable.
466
 */
L
Linus Torvalds 已提交
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496
static int scsi_eh_completed_normally(struct scsi_cmnd *scmd)
{
	/*
	 * first check the host byte, to see if there is anything in there
	 * that would indicate what we need to do.
	 */
	if (host_byte(scmd->result) == DID_RESET) {
		/*
		 * rats.  we are already in the error handler, so we now
		 * get to try and figure out what to do next.  if the sense
		 * is valid, we have a pretty good idea of what to do.
		 * if not, we mark it as FAILED.
		 */
		return scsi_check_sense(scmd);
	}
	if (host_byte(scmd->result) != DID_OK)
		return FAILED;

	/*
	 * next, check the message byte.
	 */
	if (msg_byte(scmd->result) != COMMAND_COMPLETE)
		return FAILED;

	/*
	 * now, check the status byte to see if this indicates
	 * anything special.
	 */
	switch (status_byte(scmd->result)) {
	case GOOD:
V
Vasu Dev 已提交
497
		scsi_handle_queue_ramp_up(scmd->device);
L
Linus Torvalds 已提交
498 499 500 501 502 503 504 505 506 507 508
	case COMMAND_TERMINATED:
		return SUCCESS;
	case CHECK_CONDITION:
		return scsi_check_sense(scmd);
	case CONDITION_GOOD:
	case INTERMEDIATE_GOOD:
	case INTERMEDIATE_C_GOOD:
		/*
		 * who knows?  FIXME(eric)
		 */
		return SUCCESS;
509
	case RESERVATION_CONFLICT:
510 511 512 513 514 515
		if (scmd->cmnd[0] == TEST_UNIT_READY)
			/* it is a success, we probed the device and
			 * found it */
			return SUCCESS;
		/* otherwise, we failed to send the command */
		return FAILED;
L
Linus Torvalds 已提交
516
	case QUEUE_FULL:
517 518 519
		scsi_handle_queue_full(scmd->device);
		/* fall through */
	case BUSY:
520
		return NEEDS_RETRY;
L
Linus Torvalds 已提交
521 522 523 524 525 526 527 528 529
	default:
		return FAILED;
	}
	return FAILED;
}

/**
 * scsi_eh_done - Completion function for error handling.
 * @scmd:	Cmd that is done.
530
 */
L
Linus Torvalds 已提交
531 532
static void scsi_eh_done(struct scsi_cmnd *scmd)
{
533
	struct completion *eh_action;
534

535 536
	SCSI_LOG_ERROR_RECOVERY(3,
		printk("%s scmd: %p result: %x\n",
537
			__func__, scmd, scmd->result));
538 539 540 541

	eh_action = scmd->device->host->eh_action;
	if (eh_action)
		complete(eh_action);
L
Linus Torvalds 已提交
542 543
}

544 545
/**
 * scsi_try_host_reset - ask host adapter to reset itself
G
Geert Uytterhoeven 已提交
546
 * @scmd:	SCSI cmd to send host reset.
547
 */
548 549 550 551
static int scsi_try_host_reset(struct scsi_cmnd *scmd)
{
	unsigned long flags;
	int rtn;
552 553
	struct Scsi_Host *host = scmd->device->host;
	struct scsi_host_template *hostt = host->hostt;
554 555

	SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Snd Host RST\n",
556
					  __func__));
557

558
	if (!hostt->eh_host_reset_handler)
559 560
		return FAILED;

561
	rtn = hostt->eh_host_reset_handler(scmd);
562 563

	if (rtn == SUCCESS) {
564
		if (!hostt->skip_settle_delay)
565
			ssleep(HOST_RESET_SETTLE_TIME);
566 567 568
		spin_lock_irqsave(host->host_lock, flags);
		scsi_report_bus_reset(host, scmd_channel(scmd));
		spin_unlock_irqrestore(host->host_lock, flags);
569 570 571 572 573 574 575 576
	}

	return rtn;
}

/**
 * scsi_try_bus_reset - ask host to perform a bus reset
 * @scmd:	SCSI cmd to send bus reset.
577
 */
578 579 580 581
static int scsi_try_bus_reset(struct scsi_cmnd *scmd)
{
	unsigned long flags;
	int rtn;
582 583
	struct Scsi_Host *host = scmd->device->host;
	struct scsi_host_template *hostt = host->hostt;
584 585

	SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Snd Bus RST\n",
586
					  __func__));
587

588
	if (!hostt->eh_bus_reset_handler)
589 590
		return FAILED;

591
	rtn = hostt->eh_bus_reset_handler(scmd);
592 593

	if (rtn == SUCCESS) {
594
		if (!hostt->skip_settle_delay)
595
			ssleep(BUS_RESET_SETTLE_TIME);
596 597 598
		spin_lock_irqsave(host->host_lock, flags);
		scsi_report_bus_reset(host, scmd_channel(scmd));
		spin_unlock_irqrestore(host->host_lock, flags);
599 600 601 602 603
	}

	return rtn;
}

604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
static void __scsi_report_device_reset(struct scsi_device *sdev, void *data)
{
	sdev->was_reset = 1;
	sdev->expecting_cc_ua = 1;
}

/**
 * scsi_try_target_reset - Ask host to perform a target reset
 * @scmd:	SCSI cmd used to send a target reset
 *
 * Notes:
 *    There is no timeout for this operation.  if this operation is
 *    unreliable for a given host, then the host itself needs to put a
 *    timer on it, and set the host back to a consistent state prior to
 *    returning.
 */
static int scsi_try_target_reset(struct scsi_cmnd *scmd)
{
	unsigned long flags;
	int rtn;
624 625
	struct Scsi_Host *host = scmd->device->host;
	struct scsi_host_template *hostt = host->hostt;
626

627
	if (!hostt->eh_target_reset_handler)
628 629
		return FAILED;

630
	rtn = hostt->eh_target_reset_handler(scmd);
631
	if (rtn == SUCCESS) {
632
		spin_lock_irqsave(host->host_lock, flags);
633 634
		__starget_for_each_device(scsi_target(scmd->device), NULL,
					  __scsi_report_device_reset);
635
		spin_unlock_irqrestore(host->host_lock, flags);
636 637 638 639 640
	}

	return rtn;
}

641 642 643 644 645 646 647 648 649
/**
 * scsi_try_bus_device_reset - Ask host to perform a BDR on a dev
 * @scmd:	SCSI cmd used to send BDR
 *
 * Notes:
 *    There is no timeout for this operation.  if this operation is
 *    unreliable for a given host, then the host itself needs to put a
 *    timer on it, and set the host back to a consistent state prior to
 *    returning.
650
 */
651 652 653
static int scsi_try_bus_device_reset(struct scsi_cmnd *scmd)
{
	int rtn;
654
	struct scsi_host_template *hostt = scmd->device->host->hostt;
655

656
	if (!hostt->eh_device_reset_handler)
657 658
		return FAILED;

659
	rtn = hostt->eh_device_reset_handler(scmd);
660 661
	if (rtn == SUCCESS)
		__scsi_report_device_reset(scmd->device, NULL);
662 663 664
	return rtn;
}

665
static int scsi_try_to_abort_cmd(struct scsi_host_template *hostt, struct scsi_cmnd *scmd)
666
{
667
	if (!hostt->eh_abort_handler)
668 669
		return FAILED;

670
	return hostt->eh_abort_handler(scmd);
671 672 673 674
}

static void scsi_abort_eh_cmnd(struct scsi_cmnd *scmd)
{
675
	if (scsi_try_to_abort_cmd(scmd->device->host->hostt, scmd) != SUCCESS)
676
		if (scsi_try_bus_device_reset(scmd) != SUCCESS)
677 678 679
			if (scsi_try_target_reset(scmd) != SUCCESS)
				if (scsi_try_bus_reset(scmd) != SUCCESS)
					scsi_try_host_reset(scmd);
680 681
}

L
Linus Torvalds 已提交
682
/**
683
 * scsi_eh_prep_cmnd  - Save a scsi command info as part of error recovery
684
 * @scmd:       SCSI command structure to hijack
685
 * @ses:        structure to save restore information
686
 * @cmnd:       CDB to send. Can be NULL if no new cmnd is needed
687
 * @cmnd_size:  size in bytes of @cmnd (must be <= BLK_MAX_CDB)
688
 * @sense_bytes: size of sense data to copy. or 0 (if != 0 @cmnd is ignored)
689
 *
690
 * This function is used to save a scsi command information before re-execution
691 692 693 694
 * as part of the error recovery process.  If @sense_bytes is 0 the command
 * sent must be one that does not transfer any data.  If @sense_bytes != 0
 * @cmnd is ignored and this functions sets up a REQUEST_SENSE command
 * and cmnd buffers to read @sense_bytes into @scmd->sense_buffer.
695
 */
696 697
void scsi_eh_prep_cmnd(struct scsi_cmnd *scmd, struct scsi_eh_save *ses,
			unsigned char *cmnd, int cmnd_size, unsigned sense_bytes)
L
Linus Torvalds 已提交
698
{
已提交
699
	struct scsi_device *sdev = scmd->device;
L
Linus Torvalds 已提交
700

701 702 703 704 705 706 707
	/*
	 * We need saved copies of a number of fields - this is because
	 * error handling may need to overwrite these with different values
	 * to run different commands, and once error handling is complete,
	 * we will need to restore these values prior to running the actual
	 * command.
	 */
708
	ses->cmd_len = scmd->cmd_len;
709
	ses->cmnd = scmd->cmnd;
710
	ses->data_direction = scmd->sc_data_direction;
B
Boaz Harrosh 已提交
711
	ses->sdb = scmd->sdb;
712
	ses->next_rq = scmd->request->next_rq;
713
	ses->result = scmd->result;
714
	ses->underflow = scmd->underflow;
715
	ses->prot_op = scmd->prot_op;
716

717
	scmd->prot_op = SCSI_PROT_NORMAL;
718 719
	scmd->cmnd = ses->eh_cmnd;
	memset(scmd->cmnd, 0, BLK_MAX_CDB);
B
Boaz Harrosh 已提交
720
	memset(&scmd->sdb, 0, sizeof(scmd->sdb));
721
	scmd->request->next_rq = NULL;
B
Boaz Harrosh 已提交
722

723
	if (sense_bytes) {
B
Boaz Harrosh 已提交
724 725
		scmd->sdb.length = min_t(unsigned, SCSI_SENSE_BUFFERSIZE,
					 sense_bytes);
726
		sg_init_one(&ses->sense_sgl, scmd->sense_buffer,
B
Boaz Harrosh 已提交
727 728
			    scmd->sdb.length);
		scmd->sdb.table.sgl = &ses->sense_sgl;
729
		scmd->sc_data_direction = DMA_FROM_DEVICE;
B
Boaz Harrosh 已提交
730
		scmd->sdb.table.nents = 1;
731
		scmd->cmnd[0] = REQUEST_SENSE;
B
Boaz Harrosh 已提交
732
		scmd->cmnd[4] = scmd->sdb.length;
733
		scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]);
734 735
	} else {
		scmd->sc_data_direction = DMA_NONE;
736
		if (cmnd) {
737
			BUG_ON(cmnd_size > BLK_MAX_CDB);
738 739 740
			memcpy(scmd->cmnd, cmnd, cmnd_size);
			scmd->cmd_len = COMMAND_SIZE(scmd->cmnd[0]);
		}
741 742 743 744
	}

	scmd->underflow = 0;

745
	if (sdev->scsi_level <= SCSI_2 && sdev->scsi_level != SCSI_UNKNOWN)
L
Linus Torvalds 已提交
746
		scmd->cmnd[1] = (scmd->cmnd[1] & 0x1f) |
已提交
747
			(sdev->lun << 5 & 0xe0);
L
Linus Torvalds 已提交
748

749 750 751 752
	/*
	 * Zero the sense buffer.  The scsi spec mandates that any
	 * untransferred sense data should be interpreted as being zero.
	 */
753
	memset(scmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
754 755 756 757
}
EXPORT_SYMBOL(scsi_eh_prep_cmnd);

/**
758
 * scsi_eh_restore_cmnd  - Restore a scsi command info as part of error recovery
759
 * @scmd:       SCSI command structure to restore
760
 * @ses:        saved information from a coresponding call to scsi_eh_prep_cmnd
761
 *
762
 * Undo any damage done by above scsi_eh_prep_cmnd().
763
 */
764 765 766 767 768 769
void scsi_eh_restore_cmnd(struct scsi_cmnd* scmd, struct scsi_eh_save *ses)
{
	/*
	 * Restore original data
	 */
	scmd->cmd_len = ses->cmd_len;
770
	scmd->cmnd = ses->cmnd;
771
	scmd->sc_data_direction = ses->data_direction;
B
Boaz Harrosh 已提交
772
	scmd->sdb = ses->sdb;
773
	scmd->request->next_rq = ses->next_rq;
774
	scmd->result = ses->result;
775
	scmd->underflow = ses->underflow;
776
	scmd->prot_op = ses->prot_op;
777 778
}
EXPORT_SYMBOL(scsi_eh_restore_cmnd);
779

780
/**
781
 * scsi_send_eh_cmnd  - submit a scsi command as part of error recovery
782 783 784 785 786 787 788 789 790 791 792
 * @scmd:       SCSI command structure to hijack
 * @cmnd:       CDB to send
 * @cmnd_size:  size in bytes of @cmnd
 * @timeout:    timeout for this request
 * @sense_bytes: size of sense data to copy or 0
 *
 * This function is used to send a scsi command down to a target device
 * as part of the error recovery process. See also scsi_eh_prep_cmnd() above.
 *
 * Return value:
 *    SUCCESS or FAILED or NEEDS_RETRY
793
 */
794 795 796 797 798 799
static int scsi_send_eh_cmnd(struct scsi_cmnd *scmd, unsigned char *cmnd,
			     int cmnd_size, int timeout, unsigned sense_bytes)
{
	struct scsi_device *sdev = scmd->device;
	struct Scsi_Host *shost = sdev->host;
	DECLARE_COMPLETION_ONSTACK(done);
800
	unsigned long timeleft = timeout;
801
	struct scsi_eh_save ses;
802
	const unsigned long stall_for = msecs_to_jiffies(100);
803 804
	int rtn;

805
retry:
806
	scsi_eh_prep_cmnd(scmd, &ses, cmnd, cmnd_size, sense_bytes);
807
	shost->eh_action = &done;
L
Linus Torvalds 已提交
808 809

	scsi_log_send(scmd);
J
Jeff Garzik 已提交
810
	scmd->scsi_done = scsi_eh_done;
811 812 813 814 815 816 817 818 819 820 821 822 823 824
	rtn = shost->hostt->queuecommand(shost, scmd);
	if (rtn) {
		if (timeleft > stall_for) {
			scsi_eh_restore_cmnd(scmd, &ses);
			timeleft -= stall_for;
			msleep(jiffies_to_msecs(stall_for));
			goto retry;
		}
		/* signal not to enter either branch of the if () below */
		timeleft = 0;
		rtn = NEEDS_RETRY;
	} else {
		timeleft = wait_for_completion_timeout(&done, timeout);
	}
L
Linus Torvalds 已提交
825

已提交
826
	shost->eh_action = NULL;
L
Linus Torvalds 已提交
827

828
	scsi_log_completion(scmd, rtn);
L
Linus Torvalds 已提交
829

830 831
	SCSI_LOG_ERROR_RECOVERY(3,
		printk("%s: scmd: %p, timeleft: %ld\n",
832
			__func__, scmd, timeleft));
L
Linus Torvalds 已提交
833 834

	/*
835 836 837 838 839 840 841
	 * If there is time left scsi_eh_done got called, and we will examine
	 * the actual status codes to see whether the command actually did
	 * complete normally, else if we have a zero return and no time left,
	 * the command must still be pending, so abort it and return FAILED.
	 * If we never actually managed to issue the command, because
	 * ->queuecommand() kept returning non zero, use the rtn = FAILED
	 * value above (so don't execute either branch of the if)
L
Linus Torvalds 已提交
842
	 */
843
	if (timeleft) {
L
Linus Torvalds 已提交
844 845 846
		rtn = scsi_eh_completed_normally(scmd);
		SCSI_LOG_ERROR_RECOVERY(3,
			printk("%s: scsi_eh_completed_normally %x\n",
847
			       __func__, rtn));
848

L
Linus Torvalds 已提交
849 850 851 852 853
		switch (rtn) {
		case SUCCESS:
		case NEEDS_RETRY:
		case FAILED:
			break;
854 855 856
		case ADD_TO_MLQUEUE:
			rtn = NEEDS_RETRY;
			break;
L
Linus Torvalds 已提交
857 858 859 860
		default:
			rtn = FAILED;
			break;
		}
861
	} else if (!rtn) {
862
		scsi_abort_eh_cmnd(scmd);
863
		rtn = FAILED;
L
Linus Torvalds 已提交
864 865
	}

866
	scsi_eh_restore_cmnd(scmd, &ses);
867

868 869 870 871 872
	if (scmd->request->cmd_type != REQ_TYPE_BLOCK_PC) {
		struct scsi_driver *sdrv = scsi_cmd_to_driver(scmd);
		if (sdrv->eh_action)
			rtn = sdrv->eh_action(scmd, cmnd, cmnd_size, rtn);
	}
873

L
Linus Torvalds 已提交
874 875 876 877 878 879 880 881 882 883 884
	return rtn;
}

/**
 * scsi_request_sense - Request sense data from a particular target.
 * @scmd:	SCSI cmd for request sense.
 *
 * Notes:
 *    Some hosts automatically obtain this information, others require
 *    that we obtain it on our own. This function will *not* return until
 *    the command either times out, or it completes.
885
 */
L
Linus Torvalds 已提交
886 887
static int scsi_request_sense(struct scsi_cmnd *scmd)
{
888
	return scsi_send_eh_cmnd(scmd, NULL, 0, scmd->device->eh_timeout, ~0);
L
Linus Torvalds 已提交
889 890 891 892 893 894 895 896 897 898
}

/**
 * scsi_eh_finish_cmd - Handle a cmd that eh is finished with.
 * @scmd:	Original SCSI cmd that eh has finished.
 * @done_q:	Queue for processed commands.
 *
 * Notes:
 *    We don't want to use the normal command completion while we are are
 *    still handling errors - it may cause other commands to be queued,
899
 *    and that would disturb what we are doing.  Thus we really want to
L
Linus Torvalds 已提交
900 901
 *    keep a list of pending commands for final completion, and once we
 *    are ready to leave error handling we handle completion for real.
902
 */
903
void scsi_eh_finish_cmd(struct scsi_cmnd *scmd, struct list_head *done_q)
L
Linus Torvalds 已提交
904 905
{
	scmd->device->host->host_failed--;
906
	scmd->eh_eflags = 0;
L
Linus Torvalds 已提交
907 908
	list_move_tail(&scmd->eh_entry, done_q);
}
909
EXPORT_SYMBOL(scsi_eh_finish_cmd);
L
Linus Torvalds 已提交
910 911 912 913

/**
 * scsi_eh_get_sense - Get device sense data.
 * @work_q:	Queue of commands to process.
914
 * @done_q:	Queue of processed commands.
L
Linus Torvalds 已提交
915 916 917
 *
 * Description:
 *    See if we need to request sense information.  if so, then get it
918
 *    now, so we have a better idea of what to do.
L
Linus Torvalds 已提交
919 920 921
 *
 * Notes:
 *    This has the unfortunate side effect that if a shost adapter does
922
 *    not automatically request sense information, we end up shutting
L
Linus Torvalds 已提交
923 924 925 926 927 928 929
 *    it down before we request it.
 *
 *    All drivers should request sense information internally these days,
 *    so for now all I have to say is tough noogies if you end up in here.
 *
 *    XXX: Long term this code should go away, but that needs an audit of
 *         all LLDDs first.
930
 */
931 932
int scsi_eh_get_sense(struct list_head *work_q,
		      struct list_head *done_q)
L
Linus Torvalds 已提交
933
{
934
	struct scsi_cmnd *scmd, *next;
L
Linus Torvalds 已提交
935 936
	int rtn;

937
	list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
938
		if ((scmd->eh_eflags & SCSI_EH_CANCEL_CMD) ||
L
Linus Torvalds 已提交
939 940 941
		    SCSI_SENSE_VALID(scmd))
			continue;

942 943 944
		SCSI_LOG_ERROR_RECOVERY(2, scmd_printk(KERN_INFO, scmd,
						  "%s: requesting sense\n",
						  current->comm));
L
Linus Torvalds 已提交
945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973
		rtn = scsi_request_sense(scmd);
		if (rtn != SUCCESS)
			continue;

		SCSI_LOG_ERROR_RECOVERY(3, printk("sense requested for %p"
						  " result %x\n", scmd,
						  scmd->result));
		SCSI_LOG_ERROR_RECOVERY(3, scsi_print_sense("bh", scmd));

		rtn = scsi_decide_disposition(scmd);

		/*
		 * if the result was normal, then just pass it along to the
		 * upper level.
		 */
		if (rtn == SUCCESS)
			/* we don't want this command reissued, just
			 * finished with the sense data, so set
			 * retries to the max allowed to ensure it
			 * won't get reissued */
			scmd->retries = scmd->allowed;
		else if (rtn != NEEDS_RETRY)
			continue;

		scsi_eh_finish_cmd(scmd, done_q);
	}

	return list_empty(work_q);
}
974
EXPORT_SYMBOL_GPL(scsi_eh_get_sense);
L
Linus Torvalds 已提交
975 976 977

/**
 * scsi_eh_tur - Send TUR to device.
978
 * @scmd:	&scsi_cmnd to send TUR
L
Linus Torvalds 已提交
979 980 981
 *
 * Return value:
 *    0 - Device is ready. 1 - Device NOT ready.
982
 */
L
Linus Torvalds 已提交
983 984 985 986 987 988
static int scsi_eh_tur(struct scsi_cmnd *scmd)
{
	static unsigned char tur_command[6] = {TEST_UNIT_READY, 0, 0, 0, 0, 0};
	int retry_cnt = 1, rtn;

retry_tur:
989 990
	rtn = scsi_send_eh_cmnd(scmd, tur_command, 6,
				scmd->device->eh_timeout, 0);
L
Linus Torvalds 已提交
991 992

	SCSI_LOG_ERROR_RECOVERY(3, printk("%s: scmd %p rtn %x\n",
993
		__func__, scmd, rtn));
994 995 996

	switch (rtn) {
	case NEEDS_RETRY:
L
Linus Torvalds 已提交
997 998
		if (retry_cnt--)
			goto retry_tur;
999 1000
		/*FALLTHRU*/
	case SUCCESS:
1001
		return 0;
1002 1003
	default:
		return 1;
1004
	}
L
Linus Torvalds 已提交
1005 1006
}

1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
/**
 * scsi_eh_test_devices - check if devices are responding from error recovery.
 * @cmd_list:	scsi commands in error recovery.
 * @work_q:     queue for commands which still need more error recovery
 * @done_q:     queue for commands which are finished
 * @try_stu:    boolean on if a STU command should be tried in addition to TUR.
 *
 * Decription:
 *    Tests if devices are in a working state.  Commands to devices now in
 *    a working state are sent to the done_q while commands to devices which
 *    are still failing to respond are returned to the work_q for more
 *    processing.
 **/
static int scsi_eh_test_devices(struct list_head *cmd_list,
				struct list_head *work_q,
				struct list_head *done_q, int try_stu)
{
	struct scsi_cmnd *scmd, *next;
	struct scsi_device *sdev;
	int finish_cmds;

	while (!list_empty(cmd_list)) {
		scmd = list_entry(cmd_list->next, struct scsi_cmnd, eh_entry);
		sdev = scmd->device;

		finish_cmds = !scsi_device_online(scmd->device) ||
			(try_stu && !scsi_eh_try_stu(scmd) &&
			 !scsi_eh_tur(scmd)) ||
			!scsi_eh_tur(scmd);

		list_for_each_entry_safe(scmd, next, cmd_list, eh_entry)
			if (scmd->device == sdev) {
				if (finish_cmds)
					scsi_eh_finish_cmd(scmd, done_q);
				else
					list_move_tail(&scmd->eh_entry, work_q);
			}
	}
	return list_empty(work_q);
}


L
Linus Torvalds 已提交
1049
/**
1050 1051 1052
 * scsi_eh_abort_cmds - abort pending commands.
 * @work_q:	&list_head for pending commands.
 * @done_q:	&list_head for processed commands.
L
Linus Torvalds 已提交
1053 1054 1055
 *
 * Decription:
 *    Try and see whether or not it makes sense to try and abort the
1056 1057
 *    running command.  This only works out to be the case if we have one
 *    command that has timed out.  If the command simply failed, it makes
L
Linus Torvalds 已提交
1058 1059
 *    no sense to try and abort the command, since as far as the shost
 *    adapter is concerned, it isn't running.
1060
 */
L
Linus Torvalds 已提交
1061 1062 1063
static int scsi_eh_abort_cmds(struct list_head *work_q,
			      struct list_head *done_q)
{
1064
	struct scsi_cmnd *scmd, *next;
1065
	LIST_HEAD(check_list);
L
Linus Torvalds 已提交
1066 1067
	int rtn;

1068
	list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
1069
		if (!(scmd->eh_eflags & SCSI_EH_CANCEL_CMD))
L
Linus Torvalds 已提交
1070 1071 1072 1073
			continue;
		SCSI_LOG_ERROR_RECOVERY(3, printk("%s: aborting cmd:"
						  "0x%p\n", current->comm,
						  scmd));
1074
		rtn = scsi_try_to_abort_cmd(scmd->device->host->hostt, scmd);
1075
		if (rtn == SUCCESS || rtn == FAST_IO_FAIL) {
1076
			scmd->eh_eflags &= ~SCSI_EH_CANCEL_CMD;
1077
			if (rtn == FAST_IO_FAIL)
L
Linus Torvalds 已提交
1078
				scsi_eh_finish_cmd(scmd, done_q);
1079 1080
			else
				list_move_tail(&scmd->eh_entry, &check_list);
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086 1087 1088
		} else
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: aborting"
							  " cmd failed:"
							  "0x%p\n",
							  current->comm,
							  scmd));
	}

1089
	return scsi_eh_test_devices(&check_list, work_q, done_q, 0);
L
Linus Torvalds 已提交
1090 1091 1092 1093
}

/**
 * scsi_eh_try_stu - Send START_UNIT to device.
1094
 * @scmd:	&scsi_cmnd to send START_UNIT
L
Linus Torvalds 已提交
1095 1096 1097
 *
 * Return value:
 *    0 - Device is ready. 1 - Device NOT ready.
1098
 */
L
Linus Torvalds 已提交
1099 1100 1101 1102
static int scsi_eh_try_stu(struct scsi_cmnd *scmd)
{
	static unsigned char stu_command[6] = {START_STOP, 0, 0, 0, 1, 0};

1103
	if (scmd->device->allow_restart) {
1104 1105 1106
		int i, rtn = NEEDS_RETRY;

		for (i = 0; rtn == NEEDS_RETRY && i < 2; i++)
1107
			rtn = scsi_send_eh_cmnd(scmd, stu_command, 6, scmd->device->request_queue->rq_timeout, 0);
L
Linus Torvalds 已提交
1108

1109 1110 1111
		if (rtn == SUCCESS)
			return 0;
	}
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116 1117

	return 1;
}

 /**
 * scsi_eh_stu - send START_UNIT if needed
1118 1119 1120
 * @shost:	&scsi host being recovered.
 * @work_q:     &list_head for pending commands.
 * @done_q:	&list_head for processed commands.
L
Linus Torvalds 已提交
1121 1122 1123
 *
 * Notes:
 *    If commands are failing due to not ready, initializing command required,
1124
 *	try revalidating the device, which will end up sending a start unit.
1125
 */
L
Linus Torvalds 已提交
1126 1127 1128 1129
static int scsi_eh_stu(struct Scsi_Host *shost,
			      struct list_head *work_q,
			      struct list_head *done_q)
{
1130
	struct scsi_cmnd *scmd, *stu_scmd, *next;
L
Linus Torvalds 已提交
1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
	struct scsi_device *sdev;

	shost_for_each_device(sdev, shost) {
		stu_scmd = NULL;
		list_for_each_entry(scmd, work_q, eh_entry)
			if (scmd->device == sdev && SCSI_SENSE_VALID(scmd) &&
			    scsi_check_sense(scmd) == FAILED ) {
				stu_scmd = scmd;
				break;
			}

		if (!stu_scmd)
			continue;

		SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending START_UNIT to sdev:"
						  " 0x%p\n", current->comm, sdev));

		if (!scsi_eh_try_stu(stu_scmd)) {
			if (!scsi_device_online(sdev) ||
			    !scsi_eh_tur(stu_scmd)) {
1151 1152
				list_for_each_entry_safe(scmd, next,
							  work_q, eh_entry) {
L
Linus Torvalds 已提交
1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170
					if (scmd->device == sdev)
						scsi_eh_finish_cmd(scmd, done_q);
				}
			}
		} else {
			SCSI_LOG_ERROR_RECOVERY(3,
						printk("%s: START_UNIT failed to sdev:"
						       " 0x%p\n", current->comm, sdev));
		}
	}

	return list_empty(work_q);
}


/**
 * scsi_eh_bus_device_reset - send bdr if needed
 * @shost:	scsi host being recovered.
1171 1172
 * @work_q:     &list_head for pending commands.
 * @done_q:	&list_head for processed commands.
L
Linus Torvalds 已提交
1173 1174
 *
 * Notes:
1175
 *    Try a bus device reset.  Still, look to see whether we have multiple
L
Linus Torvalds 已提交
1176 1177
 *    devices that are jammed or not - if we have multiple devices, it
 *    makes no sense to try bus_device_reset - we really would need to try
1178
 *    a bus_reset instead.
1179
 */
L
Linus Torvalds 已提交
1180 1181 1182 1183
static int scsi_eh_bus_device_reset(struct Scsi_Host *shost,
				    struct list_head *work_q,
				    struct list_head *done_q)
{
1184
	struct scsi_cmnd *scmd, *bdr_scmd, *next;
L
Linus Torvalds 已提交
1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
	struct scsi_device *sdev;
	int rtn;

	shost_for_each_device(sdev, shost) {
		bdr_scmd = NULL;
		list_for_each_entry(scmd, work_q, eh_entry)
			if (scmd->device == sdev) {
				bdr_scmd = scmd;
				break;
			}

		if (!bdr_scmd)
			continue;

		SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending BDR sdev:"
						  " 0x%p\n", current->comm,
						  sdev));
		rtn = scsi_try_bus_device_reset(bdr_scmd);
1203
		if (rtn == SUCCESS || rtn == FAST_IO_FAIL) {
L
Linus Torvalds 已提交
1204
			if (!scsi_device_online(sdev) ||
1205
			    rtn == FAST_IO_FAIL ||
L
Linus Torvalds 已提交
1206
			    !scsi_eh_tur(bdr_scmd)) {
1207 1208
				list_for_each_entry_safe(scmd, next,
							 work_q, eh_entry) {
L
Linus Torvalds 已提交
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
					if (scmd->device == sdev)
						scsi_eh_finish_cmd(scmd,
								   done_q);
				}
			}
		} else {
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: BDR"
							  " failed sdev:"
							  "0x%p\n",
							  current->comm,
							   sdev));
		}
	}

	return list_empty(work_q);
}

1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
/**
 * scsi_eh_target_reset - send target reset if needed
 * @shost:	scsi host being recovered.
 * @work_q:     &list_head for pending commands.
 * @done_q:	&list_head for processed commands.
 *
 * Notes:
 *    Try a target reset.
 */
static int scsi_eh_target_reset(struct Scsi_Host *shost,
				struct list_head *work_q,
				struct list_head *done_q)
{
1239
	LIST_HEAD(tmp_list);
1240
	LIST_HEAD(check_list);
1241

1242 1243 1244 1245 1246 1247 1248 1249 1250
	list_splice_init(work_q, &tmp_list);

	while (!list_empty(&tmp_list)) {
		struct scsi_cmnd *next, *scmd;
		int rtn;
		unsigned int id;

		scmd = list_entry(tmp_list.next, struct scsi_cmnd, eh_entry);
		id = scmd_id(scmd);
1251 1252 1253 1254

		SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending target reset "
						  "to target %d\n",
						  current->comm, id));
1255 1256
		rtn = scsi_try_target_reset(scmd);
		if (rtn != SUCCESS && rtn != FAST_IO_FAIL)
1257 1258 1259 1260
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Target reset"
							  " failed target: "
							  "%d\n",
							  current->comm, id));
1261 1262 1263 1264
		list_for_each_entry_safe(scmd, next, &tmp_list, eh_entry) {
			if (scmd_id(scmd) != id)
				continue;

1265 1266 1267
			if (rtn == SUCCESS)
				list_move_tail(&scmd->eh_entry, &check_list);
			else if (rtn == FAST_IO_FAIL)
1268 1269 1270 1271 1272 1273
				scsi_eh_finish_cmd(scmd, done_q);
			else
				/* push back on work queue for further processing */
				list_move(&scmd->eh_entry, work_q);
		}
	}
1274

1275
	return scsi_eh_test_devices(&check_list, work_q, done_q, 0);
1276 1277
}

L
Linus Torvalds 已提交
1278
/**
1279
 * scsi_eh_bus_reset - send a bus reset
1280 1281 1282
 * @shost:	&scsi host being recovered.
 * @work_q:     &list_head for pending commands.
 * @done_q:	&list_head for processed commands.
1283
 */
L
Linus Torvalds 已提交
1284 1285 1286 1287
static int scsi_eh_bus_reset(struct Scsi_Host *shost,
			     struct list_head *work_q,
			     struct list_head *done_q)
{
1288
	struct scsi_cmnd *scmd, *chan_scmd, *next;
1289
	LIST_HEAD(check_list);
L
Linus Torvalds 已提交
1290 1291 1292 1293 1294 1295 1296
	unsigned int channel;
	int rtn;

	/*
	 * we really want to loop over the various channels, and do this on
	 * a channel by channel basis.  we should also check to see if any
	 * of the failed commands are on soft_reset devices, and if so, skip
1297
	 * the reset.
L
Linus Torvalds 已提交
1298 1299 1300 1301 1302
	 */

	for (channel = 0; channel <= shost->max_channel; channel++) {
		chan_scmd = NULL;
		list_for_each_entry(scmd, work_q, eh_entry) {
1303
			if (channel == scmd_channel(scmd)) {
L
Linus Torvalds 已提交
1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
				chan_scmd = scmd;
				break;
				/*
				 * FIXME add back in some support for
				 * soft_reset devices.
				 */
			}
		}

		if (!chan_scmd)
			continue;
		SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending BRST chan:"
						  " %d\n", current->comm,
						  channel));
		rtn = scsi_try_bus_reset(chan_scmd);
1319
		if (rtn == SUCCESS || rtn == FAST_IO_FAIL) {
1320
			list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
1321 1322
				if (channel == scmd_channel(scmd)) {
					if (rtn == FAST_IO_FAIL)
L
Linus Torvalds 已提交
1323 1324
						scsi_eh_finish_cmd(scmd,
								   done_q);
1325 1326 1327 1328
					else
						list_move_tail(&scmd->eh_entry,
							       &check_list);
				}
L
Linus Torvalds 已提交
1329 1330 1331 1332 1333 1334 1335 1336
			}
		} else {
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: BRST"
							  " failed chan: %d\n",
							  current->comm,
							  channel));
		}
	}
1337
	return scsi_eh_test_devices(&check_list, work_q, done_q, 0);
L
Linus Torvalds 已提交
1338 1339 1340
}

/**
1341
 * scsi_eh_host_reset - send a host reset
L
Linus Torvalds 已提交
1342 1343
 * @work_q:	list_head for processed commands.
 * @done_q:	list_head for processed commands.
1344
 */
L
Linus Torvalds 已提交
1345 1346 1347
static int scsi_eh_host_reset(struct list_head *work_q,
			      struct list_head *done_q)
{
1348
	struct scsi_cmnd *scmd, *next;
1349
	LIST_HEAD(check_list);
L
Linus Torvalds 已提交
1350 1351 1352 1353 1354 1355 1356 1357 1358 1359
	int rtn;

	if (!list_empty(work_q)) {
		scmd = list_entry(work_q->next,
				  struct scsi_cmnd, eh_entry);

		SCSI_LOG_ERROR_RECOVERY(3, printk("%s: Sending HRST\n"
						  , current->comm));

		rtn = scsi_try_host_reset(scmd);
1360 1361 1362
		if (rtn == SUCCESS) {
			list_splice_init(work_q, &check_list);
		} else if (rtn == FAST_IO_FAIL) {
1363
			list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
L
Linus Torvalds 已提交
1364 1365 1366 1367 1368 1369 1370 1371
					scsi_eh_finish_cmd(scmd, done_q);
			}
		} else {
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: HRST"
							  " failed\n",
							  current->comm));
		}
	}
1372
	return scsi_eh_test_devices(&check_list, work_q, done_q, 1);
L
Linus Torvalds 已提交
1373 1374 1375 1376 1377 1378
}

/**
 * scsi_eh_offline_sdevs - offline scsi devices that fail to recover
 * @work_q:	list_head for processed commands.
 * @done_q:	list_head for processed commands.
1379
 */
L
Linus Torvalds 已提交
1380 1381 1382
static void scsi_eh_offline_sdevs(struct list_head *work_q,
				  struct list_head *done_q)
{
1383
	struct scsi_cmnd *scmd, *next;
L
Linus Torvalds 已提交
1384

1385
	list_for_each_entry_safe(scmd, next, work_q, eh_entry) {
1386 1387
		sdev_printk(KERN_INFO, scmd->device, "Device offlined - "
			    "not ready after error recovery\n");
L
Linus Torvalds 已提交
1388
		scsi_device_set_state(scmd->device, SDEV_OFFLINE);
1389
		if (scmd->eh_eflags & SCSI_EH_CANCEL_CMD) {
L
Linus Torvalds 已提交
1390 1391 1392 1393 1394 1395 1396 1397 1398
			/*
			 * FIXME: Handle lost cmds.
			 */
		}
		scsi_eh_finish_cmd(scmd, done_q);
	}
	return;
}

1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
/**
 * scsi_noretry_cmd - determinte if command should be failed fast
 * @scmd:	SCSI cmd to examine.
 */
int scsi_noretry_cmd(struct scsi_cmnd *scmd)
{
	switch (host_byte(scmd->result)) {
	case DID_OK:
		break;
	case DID_BUS_BUSY:
1409
		return (scmd->request->cmd_flags & REQ_FAILFAST_TRANSPORT);
1410
	case DID_PARITY:
1411
		return (scmd->request->cmd_flags & REQ_FAILFAST_DEV);
1412 1413 1414 1415 1416 1417
	case DID_ERROR:
		if (msg_byte(scmd->result) == COMMAND_COMPLETE &&
		    status_byte(scmd->result) == RESERVATION_CONFLICT)
			return 0;
		/* fall through */
	case DID_SOFT_ERROR:
1418
		return (scmd->request->cmd_flags & REQ_FAILFAST_DRIVER);
1419 1420 1421 1422 1423 1424 1425 1426
	}

	switch (status_byte(scmd->result)) {
	case CHECK_CONDITION:
		/*
		 * assume caller has checked sense and determinted
		 * the check condition was retryable.
		 */
1427 1428 1429
		if (scmd->request->cmd_flags & REQ_FAILFAST_DEV ||
		    scmd->request->cmd_type == REQ_TYPE_BLOCK_PC)
			return 1;
1430 1431 1432 1433 1434
	}

	return 0;
}

L
Linus Torvalds 已提交
1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447
/**
 * scsi_decide_disposition - Disposition a cmd on return from LLD.
 * @scmd:	SCSI cmd to examine.
 *
 * Notes:
 *    This is *only* called when we are examining the status after sending
 *    out the actual data command.  any commands that are queued for error
 *    recovery (e.g. test_unit_ready) do *not* come through here.
 *
 *    When this routine returns failed, it means the error handler thread
 *    is woken.  In cases where the error code indicates an error that
 *    doesn't require the error handler read (i.e. we don't need to
 *    abort/reset), this function should return SUCCESS.
1448
 */
L
Linus Torvalds 已提交
1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459
int scsi_decide_disposition(struct scsi_cmnd *scmd)
{
	int rtn;

	/*
	 * if the device is offline, then we clearly just pass the result back
	 * up to the top level.
	 */
	if (!scsi_device_online(scmd->device)) {
		SCSI_LOG_ERROR_RECOVERY(5, printk("%s: device offline - report"
						  " as SUCCESS\n",
1460
						  __func__));
L
Linus Torvalds 已提交
1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492
		return SUCCESS;
	}

	/*
	 * first check the host byte, to see if there is anything in there
	 * that would indicate what we need to do.
	 */
	switch (host_byte(scmd->result)) {
	case DID_PASSTHROUGH:
		/*
		 * no matter what, pass this through to the upper layer.
		 * nuke this special code so that it looks like we are saying
		 * did_ok.
		 */
		scmd->result &= 0xff00ffff;
		return SUCCESS;
	case DID_OK:
		/*
		 * looks good.  drop through, and check the next byte.
		 */
		break;
	case DID_NO_CONNECT:
	case DID_BAD_TARGET:
	case DID_ABORT:
		/*
		 * note - this means that we just report the status back
		 * to the top level driver, not that we actually think
		 * that it indicates SUCCESS.
		 */
		return SUCCESS;
		/*
		 * when the low level driver returns did_soft_error,
1493
		 * it is responsible for keeping an internal retry counter
L
Linus Torvalds 已提交
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
		 * in order to avoid endless loops (db)
		 *
		 * actually this is a bug in this function here.  we should
		 * be mindful of the maximum number of retries specified
		 * and not get stuck in a loop.
		 */
	case DID_SOFT_ERROR:
		goto maybe_retry;
	case DID_IMM_RETRY:
		return NEEDS_RETRY;

已提交
1505 1506
	case DID_REQUEUE:
		return ADD_TO_MLQUEUE;
1507 1508 1509 1510 1511
	case DID_TRANSPORT_DISRUPTED:
		/*
		 * LLD/transport was disrupted during processing of the IO.
		 * The transport class is now blocked/blocking,
		 * and the transport will decide what to do with the IO
1512 1513
		 * based on its timers and recovery capablilities if
		 * there are enough retries.
1514
		 */
1515
		goto maybe_retry;
1516 1517 1518 1519 1520 1521
	case DID_TRANSPORT_FAILFAST:
		/*
		 * The transport decided to failfast the IO (most likely
		 * the fast io fail tmo fired), so send IO directly upwards.
		 */
		return SUCCESS;
L
Linus Torvalds 已提交
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562
	case DID_ERROR:
		if (msg_byte(scmd->result) == COMMAND_COMPLETE &&
		    status_byte(scmd->result) == RESERVATION_CONFLICT)
			/*
			 * execute reservation conflict processing code
			 * lower down
			 */
			break;
		/* fallthrough */
	case DID_BUS_BUSY:
	case DID_PARITY:
		goto maybe_retry;
	case DID_TIME_OUT:
		/*
		 * when we scan the bus, we get timeout messages for
		 * these commands if there is no device available.
		 * other hosts report did_no_connect for the same thing.
		 */
		if ((scmd->cmnd[0] == TEST_UNIT_READY ||
		     scmd->cmnd[0] == INQUIRY)) {
			return SUCCESS;
		} else {
			return FAILED;
		}
	case DID_RESET:
		return SUCCESS;
	default:
		return FAILED;
	}

	/*
	 * next, check the message byte.
	 */
	if (msg_byte(scmd->result) != COMMAND_COMPLETE)
		return FAILED;

	/*
	 * check the status byte to see if this indicates anything special.
	 */
	switch (status_byte(scmd->result)) {
	case QUEUE_FULL:
1563
		scsi_handle_queue_full(scmd->device);
L
Linus Torvalds 已提交
1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576
		/*
		 * the case of trying to send too many commands to a
		 * tagged queueing device.
		 */
	case BUSY:
		/*
		 * device can't talk to us at the moment.  Should only
		 * occur (SAM-3) when the task queue is empty, so will cause
		 * the empty queue handling to trigger a stall in the
		 * device.
		 */
		return ADD_TO_MLQUEUE;
	case GOOD:
V
Vasu Dev 已提交
1577
		scsi_handle_queue_ramp_up(scmd->device);
L
Linus Torvalds 已提交
1578 1579
	case COMMAND_TERMINATED:
		return SUCCESS;
1580 1581
	case TASK_ABORTED:
		goto maybe_retry;
L
Linus Torvalds 已提交
1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600
	case CHECK_CONDITION:
		rtn = scsi_check_sense(scmd);
		if (rtn == NEEDS_RETRY)
			goto maybe_retry;
		/* if rtn == FAILED, we have no sense information;
		 * returning FAILED will wake the error handler thread
		 * to collect the sense and redo the decide
		 * disposition */
		return rtn;
	case CONDITION_GOOD:
	case INTERMEDIATE_GOOD:
	case INTERMEDIATE_C_GOOD:
	case ACA_ACTIVE:
		/*
		 * who knows?  FIXME(eric)
		 */
		return SUCCESS;

	case RESERVATION_CONFLICT:
1601 1602
		sdev_printk(KERN_INFO, scmd->device,
			    "reservation conflict\n");
1603
		set_host_byte(scmd, DID_NEXUS_FAILURE);
L
Linus Torvalds 已提交
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
		return SUCCESS; /* causes immediate i/o error */
	default:
		return FAILED;
	}
	return FAILED;

      maybe_retry:

	/* we requeue for retry because the error was retryable, and
	 * the request was not marked fast fail.  Note that above,
	 * even if the request is marked fast fail, we still requeue
	 * for queue congestion conditions (QUEUE_FULL or BUSY) */
1616
	if ((++scmd->retries) <= scmd->allowed
1617
	    && !scsi_noretry_cmd(scmd)) {
L
Linus Torvalds 已提交
1618 1619 1620 1621 1622 1623 1624 1625 1626
		return NEEDS_RETRY;
	} else {
		/*
		 * no more retries - report this one back to upper level.
		 */
		return SUCCESS;
	}
}

F
FUJITA Tomonori 已提交
1627 1628 1629 1630 1631
static void eh_lock_door_done(struct request *req, int uptodate)
{
	__blk_put_request(req->q, req);
}

L
Linus Torvalds 已提交
1632 1633 1634 1635 1636
/**
 * scsi_eh_lock_door - Prevent medium removal for the specified device
 * @sdev:	SCSI device to prevent medium removal
 *
 * Locking:
1637
 * 	We must be called from process context.
L
Linus Torvalds 已提交
1638 1639 1640 1641
 *
 * Notes:
 * 	We queue up an asynchronous "ALLOW MEDIUM REMOVAL" request on the
 * 	head of the devices request queue, and continue.
1642
 */
L
Linus Torvalds 已提交
1643 1644
static void scsi_eh_lock_door(struct scsi_device *sdev)
{
F
FUJITA Tomonori 已提交
1645
	struct request *req;
L
Linus Torvalds 已提交
1646

1647 1648 1649 1650
	/*
	 * blk_get_request with GFP_KERNEL (__GFP_WAIT) sleeps until a
	 * request becomes available
	 */
F
FUJITA Tomonori 已提交
1651
	req = blk_get_request(sdev->request_queue, READ, GFP_KERNEL);
L
Linus Torvalds 已提交
1652

F
FUJITA Tomonori 已提交
1653 1654 1655 1656 1657 1658
	req->cmd[0] = ALLOW_MEDIUM_REMOVAL;
	req->cmd[1] = 0;
	req->cmd[2] = 0;
	req->cmd[3] = 0;
	req->cmd[4] = SCSI_REMOVAL_PREVENT;
	req->cmd[5] = 0;
L
Linus Torvalds 已提交
1659

F
FUJITA Tomonori 已提交
1660 1661 1662 1663 1664 1665 1666 1667 1668
	req->cmd_len = COMMAND_SIZE(req->cmd[0]);

	req->cmd_type = REQ_TYPE_BLOCK_PC;
	req->cmd_flags |= REQ_QUIET;
	req->timeout = 10 * HZ;
	req->retries = 5;

	blk_execute_rq_nowait(req->q, NULL, req, 1, eh_lock_door_done);
}
L
Linus Torvalds 已提交
1669 1670 1671 1672 1673 1674 1675 1676

/**
 * scsi_restart_operations - restart io operations to the specified host.
 * @shost:	Host we are restarting.
 *
 * Notes:
 *    When we entered the error handler, we blocked all further i/o to
 *    this device.  we need to 'reverse' this process.
1677
 */
L
Linus Torvalds 已提交
1678 1679 1680
static void scsi_restart_operations(struct Scsi_Host *shost)
{
	struct scsi_device *sdev;
1681
	unsigned long flags;
L
Linus Torvalds 已提交
1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698

	/*
	 * If the door was locked, we need to insert a door lock request
	 * onto the head of the SCSI request queue for the device.  There
	 * is no point trying to lock the door of an off-line device.
	 */
	shost_for_each_device(sdev, shost) {
		if (scsi_device_online(sdev) && sdev->locked)
			scsi_eh_lock_door(sdev);
	}

	/*
	 * next free up anything directly waiting upon the host.  this
	 * will be requests for character device operations, and also for
	 * ioctls to queued block devices.
	 */
	SCSI_LOG_ERROR_RECOVERY(3, printk("%s: waking up host to restart\n",
1699
					  __func__));
L
Linus Torvalds 已提交
1700

1701 1702 1703 1704 1705
	spin_lock_irqsave(shost->host_lock, flags);
	if (scsi_host_set_state(shost, SHOST_RUNNING))
		if (scsi_host_set_state(shost, SHOST_CANCEL))
			BUG_ON(scsi_host_set_state(shost, SHOST_DEL));
	spin_unlock_irqrestore(shost->host_lock, flags);
L
Linus Torvalds 已提交
1706 1707 1708 1709 1710 1711 1712 1713 1714 1715

	wake_up(&shost->host_wait);

	/*
	 * finally we need to re-initiate requests that may be pending.  we will
	 * have had everything blocked while error handling is taking place, and
	 * now that error recovery is done, we will need to ensure that these
	 * requests are started.
	 */
	scsi_run_host_queues(shost);
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729

	/*
	 * if eh is active and host_eh_scheduled is pending we need to re-run
	 * recovery.  we do this check after scsi_run_host_queues() to allow
	 * everything pent up since the last eh run a chance to make forward
	 * progress before we sync again.  Either we'll immediately re-run
	 * recovery or scsi_device_unbusy() will wake us again when these
	 * pending commands complete.
	 */
	spin_lock_irqsave(shost->host_lock, flags);
	if (shost->host_eh_scheduled)
		if (scsi_host_set_state(shost, SHOST_RECOVERY))
			WARN_ON(scsi_host_set_state(shost, SHOST_CANCEL_RECOVERY));
	spin_unlock_irqrestore(shost->host_lock, flags);
L
Linus Torvalds 已提交
1730 1731 1732 1733 1734
}

/**
 * scsi_eh_ready_devs - check device ready state and recover if not.
 * @shost: 	host to be recovered.
1735 1736
 * @work_q:     &list_head for pending commands.
 * @done_q:	&list_head for processed commands.
1737
 */
1738 1739 1740
void scsi_eh_ready_devs(struct Scsi_Host *shost,
			struct list_head *work_q,
			struct list_head *done_q)
L
Linus Torvalds 已提交
1741 1742 1743
{
	if (!scsi_eh_stu(shost, work_q, done_q))
		if (!scsi_eh_bus_device_reset(shost, work_q, done_q))
1744 1745 1746 1747 1748
			if (!scsi_eh_target_reset(shost, work_q, done_q))
				if (!scsi_eh_bus_reset(shost, work_q, done_q))
					if (!scsi_eh_host_reset(work_q, done_q))
						scsi_eh_offline_sdevs(work_q,
								      done_q);
L
Linus Torvalds 已提交
1749
}
1750
EXPORT_SYMBOL_GPL(scsi_eh_ready_devs);
L
Linus Torvalds 已提交
1751 1752 1753 1754

/**
 * scsi_eh_flush_done_q - finish processed commands or retry them.
 * @done_q:	list_head of processed commands.
1755
 */
1756
void scsi_eh_flush_done_q(struct list_head *done_q)
L
Linus Torvalds 已提交
1757
{
1758
	struct scsi_cmnd *scmd, *next;
L
Linus Torvalds 已提交
1759

1760 1761
	list_for_each_entry_safe(scmd, next, done_q, eh_entry) {
		list_del_init(&scmd->eh_entry);
L
Linus Torvalds 已提交
1762
		if (scsi_device_online(scmd->device) &&
1763
		    !scsi_noretry_cmd(scmd) &&
1764
		    (++scmd->retries <= scmd->allowed)) {
L
Linus Torvalds 已提交
1765 1766 1767 1768 1769 1770
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: flush"
							  " retry cmd: %p\n",
							  current->comm,
							  scmd));
				scsi_queue_insert(scmd, SCSI_MLQUEUE_EH_RETRY);
		} else {
1771 1772 1773 1774 1775
			/*
			 * If just we got sense for the device (called
			 * scsi_eh_get_sense), scmd->result is already
			 * set, do not set DRIVER_TIMEOUT.
			 */
L
Linus Torvalds 已提交
1776 1777 1778 1779 1780 1781 1782 1783 1784
			if (!scmd->result)
				scmd->result |= (DRIVER_TIMEOUT << 24);
			SCSI_LOG_ERROR_RECOVERY(3, printk("%s: flush finish"
							" cmd: %p\n",
							current->comm, scmd));
			scsi_finish_command(scmd);
		}
	}
}
1785
EXPORT_SYMBOL(scsi_eh_flush_done_q);
L
Linus Torvalds 已提交
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808

/**
 * scsi_unjam_host - Attempt to fix a host which has a cmd that failed.
 * @shost:	Host to unjam.
 *
 * Notes:
 *    When we come in here, we *know* that all commands on the bus have
 *    either completed, failed or timed out.  we also know that no further
 *    commands are being sent to the host, so things are relatively quiet
 *    and we have freedom to fiddle with things as we wish.
 *
 *    This is only the *default* implementation.  it is possible for
 *    individual drivers to supply their own version of this function, and
 *    if the maintainer wishes to do this, it is strongly suggested that
 *    this function be taken as a template and modified.  this function
 *    was designed to correctly handle problems for about 95% of the
 *    different cases out there, and it should always provide at least a
 *    reasonable amount of error recovery.
 *
 *    Any command marked 'failed' or 'timeout' must eventually have
 *    scsi_finish_cmd() called for it.  we do all of the retry stuff
 *    here, so when we restart the host after we return it should have an
 *    empty queue.
1809
 */
L
Linus Torvalds 已提交
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829
static void scsi_unjam_host(struct Scsi_Host *shost)
{
	unsigned long flags;
	LIST_HEAD(eh_work_q);
	LIST_HEAD(eh_done_q);

	spin_lock_irqsave(shost->host_lock, flags);
	list_splice_init(&shost->eh_cmd_q, &eh_work_q);
	spin_unlock_irqrestore(shost->host_lock, flags);

	SCSI_LOG_ERROR_RECOVERY(1, scsi_eh_prt_fail_stats(shost, &eh_work_q));

	if (!scsi_eh_get_sense(&eh_work_q, &eh_done_q))
		if (!scsi_eh_abort_cmds(&eh_work_q, &eh_done_q))
			scsi_eh_ready_devs(shost, &eh_work_q, &eh_done_q);

	scsi_eh_flush_done_q(&eh_done_q);
}

/**
1830
 * scsi_error_handler - SCSI error handler thread
L
Linus Torvalds 已提交
1831 1832 1833
 * @data:	Host for which we are running.
 *
 * Notes:
1834 1835
 *    This is the main error handling loop.  This is run as a kernel thread
 *    for every SCSI host and handles all error handling activity.
1836
 */
L
Linus Torvalds 已提交
1837 1838
int scsi_error_handler(void *data)
{
1839
	struct Scsi_Host *shost = data;
L
Linus Torvalds 已提交
1840 1841

	/*
1842 1843 1844
	 * We use TASK_INTERRUPTIBLE so that the thread is not
	 * counted against the load average as a running process.
	 * We never actually get interrupted because kthread_run
1845
	 * disables signal delivery for the created thread.
L
Linus Torvalds 已提交
1846
	 */
1847
	while (!kthread_should_stop()) {
1848
		set_current_state(TASK_INTERRUPTIBLE);
1849
		if ((shost->host_failed == 0 && shost->host_eh_scheduled == 0) ||
1850
		    shost->host_failed != shost->host_busy) {
1851 1852 1853
			SCSI_LOG_ERROR_RECOVERY(1,
				printk("Error handler scsi_eh_%d sleeping\n",
					shost->host_no));
1854 1855 1856
			schedule();
			continue;
		}
L
Linus Torvalds 已提交
1857

1858
		__set_current_state(TASK_RUNNING);
1859 1860 1861
		SCSI_LOG_ERROR_RECOVERY(1,
			printk("Error handler scsi_eh_%d waking up\n",
				shost->host_no));
L
Linus Torvalds 已提交
1862 1863 1864 1865 1866 1867

		/*
		 * We have a host that is failing for some reason.  Figure out
		 * what we need to do to get it up and online again (if we can).
		 * If we fail, we end up taking the thing offline.
		 */
1868
		if (!shost->eh_noresume && scsi_autopm_get_host(shost) != 0) {
1869 1870 1871 1872 1873 1874 1875
			SCSI_LOG_ERROR_RECOVERY(1,
				printk(KERN_ERR "Error handler scsi_eh_%d "
						"unable to autoresume\n",
						shost->host_no));
			continue;
		}

1876 1877
		if (shost->transportt->eh_strategy_handler)
			shost->transportt->eh_strategy_handler(shost);
L
Linus Torvalds 已提交
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
		else
			scsi_unjam_host(shost);

		/*
		 * Note - if the above fails completely, the action is to take
		 * individual devices offline and flush the queue of any
		 * outstanding requests that may have been pending.  When we
		 * restart, we restart any I/O to any other devices on the bus
		 * which are still online.
		 */
		scsi_restart_operations(shost);
1889 1890
		if (!shost->eh_noresume)
			scsi_autopm_put_host(shost);
L
Linus Torvalds 已提交
1891
	}
1892 1893
	__set_current_state(TASK_RUNNING);

1894 1895
	SCSI_LOG_ERROR_RECOVERY(1,
		printk("Error handler scsi_eh_%d exiting\n", shost->host_no));
1896
	shost->ehandler = NULL;
L
Linus Torvalds 已提交
1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925
	return 0;
}

/*
 * Function:    scsi_report_bus_reset()
 *
 * Purpose:     Utility function used by low-level drivers to report that
 *		they have observed a bus reset on the bus being handled.
 *
 * Arguments:   shost       - Host in question
 *		channel     - channel on which reset was observed.
 *
 * Returns:     Nothing
 *
 * Lock status: Host lock must be held.
 *
 * Notes:       This only needs to be called if the reset is one which
 *		originates from an unknown location.  Resets originated
 *		by the mid-level itself don't need to call this, but there
 *		should be no harm.
 *
 *		The main purpose of this is to make sure that a CHECK_CONDITION
 *		is properly treated.
 */
void scsi_report_bus_reset(struct Scsi_Host *shost, int channel)
{
	struct scsi_device *sdev;

	__shost_for_each_device(sdev, shost) {
1926 1927
		if (channel == sdev_channel(sdev))
			__scsi_report_device_reset(sdev, NULL);
L
Linus Torvalds 已提交
1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	}
}
EXPORT_SYMBOL(scsi_report_bus_reset);

/*
 * Function:    scsi_report_device_reset()
 *
 * Purpose:     Utility function used by low-level drivers to report that
 *		they have observed a device reset on the device being handled.
 *
 * Arguments:   shost       - Host in question
 *		channel     - channel on which reset was observed
 *		target	    - target on which reset was observed
 *
 * Returns:     Nothing
 *
 * Lock status: Host lock must be held
 *
 * Notes:       This only needs to be called if the reset is one which
 *		originates from an unknown location.  Resets originated
 *		by the mid-level itself don't need to call this, but there
 *		should be no harm.
 *
 *		The main purpose of this is to make sure that a CHECK_CONDITION
 *		is properly treated.
 */
void scsi_report_device_reset(struct Scsi_Host *shost, int channel, int target)
{
	struct scsi_device *sdev;

	__shost_for_each_device(sdev, shost) {
1959
		if (channel == sdev_channel(sdev) &&
1960 1961
		    target == sdev_id(sdev))
			__scsi_report_device_reset(sdev, NULL);
L
Linus Torvalds 已提交
1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
	}
}
EXPORT_SYMBOL(scsi_report_device_reset);

static void
scsi_reset_provider_done_command(struct scsi_cmnd *scmd)
{
}

/*
 * Function:	scsi_reset_provider
 *
 * Purpose:	Send requested reset to a bus or device at any phase.
 *
 * Arguments:	device	- device to send reset to
 *		flag - reset type (see scsi.h)
 *
 * Returns:	SUCCESS/FAILURE.
 *
 * Notes:	This is used by the SCSI Generic driver to provide
 *		Bus/Device reset capability.
 */
int
scsi_reset_provider(struct scsi_device *dev, int flag)
{
1987
	struct scsi_cmnd *scmd;
1988
	struct Scsi_Host *shost = dev->host;
L
Linus Torvalds 已提交
1989
	struct request req;
1990
	unsigned long flags;
L
Linus Torvalds 已提交
1991 1992
	int rtn;

1993 1994 1995 1996
	if (scsi_autopm_get_host(shost) < 0)
		return FAILED;

	scmd = scsi_get_command(dev, GFP_KERNEL);
1997
	blk_rq_init(NULL, &req);
L
Linus Torvalds 已提交
1998
	scmd->request = &req;
1999

2000 2001
	scmd->cmnd = req.cmd;

L
Linus Torvalds 已提交
2002
	scmd->scsi_done		= scsi_reset_provider_done_command;
B
Boaz Harrosh 已提交
2003
	memset(&scmd->sdb, 0, sizeof(scmd->sdb));
L
Linus Torvalds 已提交
2004 2005 2006 2007 2008

	scmd->cmd_len			= 0;

	scmd->sc_data_direction		= DMA_BIDIRECTIONAL;

2009 2010 2011 2012
	spin_lock_irqsave(shost->host_lock, flags);
	shost->tmf_in_progress = 1;
	spin_unlock_irqrestore(shost->host_lock, flags);

L
Linus Torvalds 已提交
2013 2014 2015 2016 2017 2018
	switch (flag) {
	case SCSI_TRY_RESET_DEVICE:
		rtn = scsi_try_bus_device_reset(scmd);
		if (rtn == SUCCESS)
			break;
		/* FALLTHROUGH */
2019 2020 2021 2022 2023
	case SCSI_TRY_RESET_TARGET:
		rtn = scsi_try_target_reset(scmd);
		if (rtn == SUCCESS)
			break;
		/* FALLTHROUGH */
L
Linus Torvalds 已提交
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
	case SCSI_TRY_RESET_BUS:
		rtn = scsi_try_bus_reset(scmd);
		if (rtn == SUCCESS)
			break;
		/* FALLTHROUGH */
	case SCSI_TRY_RESET_HOST:
		rtn = scsi_try_host_reset(scmd);
		break;
	default:
		rtn = FAILED;
	}

2036 2037 2038 2039 2040 2041 2042 2043 2044 2045
	spin_lock_irqsave(shost->host_lock, flags);
	shost->tmf_in_progress = 0;
	spin_unlock_irqrestore(shost->host_lock, flags);

	/*
	 * be sure to wake up anyone who was sleeping or had their queue
	 * suspended while we performed the TMF.
	 */
	SCSI_LOG_ERROR_RECOVERY(3,
		printk("%s: waking up host to restart after TMF\n",
2046
		__func__));
2047 2048 2049 2050 2051

	wake_up(&shost->host_wait);

	scsi_run_host_queues(shost);

L
Linus Torvalds 已提交
2052
	scsi_next_command(scmd);
2053
	scsi_autopm_put_host(shost);
L
Linus Torvalds 已提交
2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
	return rtn;
}
EXPORT_SYMBOL(scsi_reset_provider);

/**
 * scsi_normalize_sense - normalize main elements from either fixed or
 *			descriptor sense data format into a common format.
 *
 * @sense_buffer:	byte array containing sense data returned by device
 * @sb_len:		number of valid bytes in sense_buffer
 * @sshdr:		pointer to instance of structure that common
 *			elements are written to.
 *
 * Notes:
 *	The "main elements" from sense data are: response_code, sense_key,
 *	asc, ascq and additional_length (only for descriptor format).
 *
 *	Typically this function can be called after a device has
 *	responded to a SCSI command with the CHECK_CONDITION status.
 *
 * Return value:
 *	1 if valid sense data information found, else 0;
2076
 */
L
Linus Torvalds 已提交
2077 2078 2079
int scsi_normalize_sense(const u8 *sense_buffer, int sb_len,
                         struct scsi_sense_hdr *sshdr)
{
2080
	if (!sense_buffer || !sb_len)
L
Linus Torvalds 已提交
2081 2082 2083 2084 2085
		return 0;

	memset(sshdr, 0, sizeof(struct scsi_sense_hdr));

	sshdr->response_code = (sense_buffer[0] & 0x7f);
2086 2087 2088 2089

	if (!scsi_sense_valid(sshdr))
		return 0;

L
Linus Torvalds 已提交
2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102
	if (sshdr->response_code >= 0x72) {
		/*
		 * descriptor format
		 */
		if (sb_len > 1)
			sshdr->sense_key = (sense_buffer[1] & 0xf);
		if (sb_len > 2)
			sshdr->asc = sense_buffer[2];
		if (sb_len > 3)
			sshdr->ascq = sense_buffer[3];
		if (sb_len > 7)
			sshdr->additional_length = sense_buffer[7];
	} else {
2103
		/*
L
Linus Torvalds 已提交
2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
		 * fixed format
		 */
		if (sb_len > 2)
			sshdr->sense_key = (sense_buffer[2] & 0xf);
		if (sb_len > 7) {
			sb_len = (sb_len < (sense_buffer[7] + 8)) ?
					 sb_len : (sense_buffer[7] + 8);
			if (sb_len > 12)
				sshdr->asc = sense_buffer[12];
			if (sb_len > 13)
				sshdr->ascq = sense_buffer[13];
		}
	}

	return 1;
}
EXPORT_SYMBOL(scsi_normalize_sense);

int scsi_command_normalize_sense(struct scsi_cmnd *cmd,
				 struct scsi_sense_hdr *sshdr)
{
	return scsi_normalize_sense(cmd->sense_buffer,
2126
			SCSI_SENSE_BUFFERSIZE, sshdr);
L
Linus Torvalds 已提交
2127 2128 2129 2130
}
EXPORT_SYMBOL(scsi_command_normalize_sense);

/**
2131
 * scsi_sense_desc_find - search for a given descriptor type in	descriptor sense data format.
L
Linus Torvalds 已提交
2132 2133 2134 2135 2136 2137 2138 2139 2140 2141
 * @sense_buffer:	byte array of descriptor format sense data
 * @sb_len:		number of valid bytes in sense_buffer
 * @desc_type:		value of descriptor type to find
 *			(e.g. 0 -> information)
 *
 * Notes:
 *	only valid when sense data is in descriptor format
 *
 * Return value:
 *	pointer to start of (first) descriptor if found else NULL
2142
 */
L
Linus Torvalds 已提交
2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169
const u8 * scsi_sense_desc_find(const u8 * sense_buffer, int sb_len,
				int desc_type)
{
	int add_sen_len, add_len, desc_len, k;
	const u8 * descp;

	if ((sb_len < 8) || (0 == (add_sen_len = sense_buffer[7])))
		return NULL;
	if ((sense_buffer[0] < 0x72) || (sense_buffer[0] > 0x73))
		return NULL;
	add_sen_len = (add_sen_len < (sb_len - 8)) ?
			add_sen_len : (sb_len - 8);
	descp = &sense_buffer[8];
	for (desc_len = 0, k = 0; k < add_sen_len; k += desc_len) {
		descp += desc_len;
		add_len = (k < (add_sen_len - 1)) ? descp[1]: -1;
		desc_len = add_len + 2;
		if (descp[0] == desc_type)
			return descp;
		if (add_len < 0) // short descriptor ??
			break;
	}
	return NULL;
}
EXPORT_SYMBOL(scsi_sense_desc_find);

/**
2170
 * scsi_get_sense_info_fld - get information field from sense data (either fixed or descriptor format)
L
Linus Torvalds 已提交
2171 2172 2173 2174 2175 2176 2177
 * @sense_buffer:	byte array of sense data
 * @sb_len:		number of valid bytes in sense_buffer
 * @info_out:		pointer to 64 integer where 8 or 4 byte information
 *			field will be placed if found.
 *
 * Return value:
 *	1 if information field found, 0 if not found.
2178
 */
L
Linus Torvalds 已提交
2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217
int scsi_get_sense_info_fld(const u8 * sense_buffer, int sb_len,
			    u64 * info_out)
{
	int j;
	const u8 * ucp;
	u64 ull;

	if (sb_len < 7)
		return 0;
	switch (sense_buffer[0] & 0x7f) {
	case 0x70:
	case 0x71:
		if (sense_buffer[0] & 0x80) {
			*info_out = (sense_buffer[3] << 24) +
				    (sense_buffer[4] << 16) +
				    (sense_buffer[5] << 8) + sense_buffer[6];
			return 1;
		} else
			return 0;
	case 0x72:
	case 0x73:
		ucp = scsi_sense_desc_find(sense_buffer, sb_len,
					   0 /* info desc */);
		if (ucp && (0xa == ucp[1])) {
			ull = 0;
			for (j = 0; j < 8; ++j) {
				if (j > 0)
					ull <<= 8;
				ull |= ucp[4 + j];
			}
			*info_out = ull;
			return 1;
		} else
			return 0;
	default:
		return 0;
	}
}
EXPORT_SYMBOL(scsi_get_sense_info_fld);
2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245

/**
 * scsi_build_sense_buffer - build sense data in a buffer
 * @desc:	Sense format (non zero == descriptor format,
 * 		0 == fixed format)
 * @buf:	Where to build sense data
 * @key:	Sense key
 * @asc:	Additional sense code
 * @ascq:	Additional sense code qualifier
 *
 **/
void scsi_build_sense_buffer(int desc, u8 *buf, u8 key, u8 asc, u8 ascq)
{
	if (desc) {
		buf[0] = 0x72;	/* descriptor, current */
		buf[1] = key;
		buf[2] = asc;
		buf[3] = ascq;
		buf[7] = 0;
	} else {
		buf[0] = 0x70;	/* fixed, current */
		buf[2] = key;
		buf[7] = 0xa;
		buf[12] = asc;
		buf[13] = ascq;
	}
}
EXPORT_SYMBOL(scsi_build_sense_buffer);