zfcp_erp.c 70.8 KB
Newer Older
S
Swen Schillig 已提交
1
/*
C
Christof Schmitt 已提交
2
 * zfcp device driver
L
Linus Torvalds 已提交
3
 *
C
Christof Schmitt 已提交
4
 * Error Recovery Procedures (ERP).
S
Swen Schillig 已提交
5
 *
C
Christof Schmitt 已提交
6
 * Copyright IBM Corporation 2002, 2008
L
Linus Torvalds 已提交
7 8 9 10
 */

#include "zfcp_ext.h"

11
static int zfcp_erp_adapter_reopen_internal(struct zfcp_adapter *, int, u8,
12
					    void *);
13
static int zfcp_erp_port_forced_reopen_internal(struct zfcp_port *, int, u8,
14 15 16
						void *);
static int zfcp_erp_port_reopen_internal(struct zfcp_port *, int, u8, void *);
static int zfcp_erp_unit_reopen_internal(struct zfcp_unit *, int, u8, void *);
L
Linus Torvalds 已提交
17

18
static int zfcp_erp_port_reopen_all_internal(struct zfcp_adapter *, int, u8,
19 20 21
					     void *);
static int zfcp_erp_unit_reopen_all_internal(struct zfcp_port *, int, u8,
					     void *);
L
Linus Torvalds 已提交
22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42

static void zfcp_erp_adapter_block(struct zfcp_adapter *, int);
static void zfcp_erp_adapter_unblock(struct zfcp_adapter *);
static void zfcp_erp_port_block(struct zfcp_port *, int);
static void zfcp_erp_port_unblock(struct zfcp_port *);
static void zfcp_erp_unit_block(struct zfcp_unit *, int);
static void zfcp_erp_unit_unblock(struct zfcp_unit *);

static int zfcp_erp_thread(void *);

static int zfcp_erp_strategy(struct zfcp_erp_action *);

static int zfcp_erp_strategy_do_action(struct zfcp_erp_action *);
static int zfcp_erp_strategy_memwait(struct zfcp_erp_action *);
static int zfcp_erp_strategy_check_target(struct zfcp_erp_action *, int);
static int zfcp_erp_strategy_check_unit(struct zfcp_unit *, int);
static int zfcp_erp_strategy_check_port(struct zfcp_port *, int);
static int zfcp_erp_strategy_check_adapter(struct zfcp_adapter *, int);
static int zfcp_erp_strategy_statechange(int, u32, struct zfcp_adapter *,
					 struct zfcp_port *,
					 struct zfcp_unit *, int);
43
static int zfcp_erp_strategy_statechange_detected(atomic_t *, u32);
L
Linus Torvalds 已提交
44 45 46 47 48 49 50 51 52 53 54 55 56
static int zfcp_erp_strategy_followup_actions(int, struct zfcp_adapter *,
					      struct zfcp_port *,
					      struct zfcp_unit *, int);
static int zfcp_erp_strategy_check_queues(struct zfcp_adapter *);
static int zfcp_erp_strategy_check_action(struct zfcp_erp_action *, int);

static int zfcp_erp_adapter_strategy(struct zfcp_erp_action *);
static int zfcp_erp_adapter_strategy_generic(struct zfcp_erp_action *, int);
static int zfcp_erp_adapter_strategy_close(struct zfcp_erp_action *);
static int zfcp_erp_adapter_strategy_open(struct zfcp_erp_action *);
static int zfcp_erp_adapter_strategy_open_qdio(struct zfcp_erp_action *);
static int zfcp_erp_adapter_strategy_open_fsf(struct zfcp_erp_action *);
static int zfcp_erp_adapter_strategy_open_fsf_xconfig(struct zfcp_erp_action *);
57
static int zfcp_erp_adapter_strategy_open_fsf_xport(struct zfcp_erp_action *);
L
Linus Torvalds 已提交
58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79
static int zfcp_erp_adapter_strategy_open_fsf_statusread(
	struct zfcp_erp_action *);

static int zfcp_erp_port_forced_strategy(struct zfcp_erp_action *);
static int zfcp_erp_port_forced_strategy_close(struct zfcp_erp_action *);

static int zfcp_erp_port_strategy(struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_clearstati(struct zfcp_port *);
static int zfcp_erp_port_strategy_close(struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_open(struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_open_nameserver(struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_open_nameserver_wakeup(
	struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_open_common(struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_open_common_lookup(struct zfcp_erp_action *);
static int zfcp_erp_port_strategy_open_port(struct zfcp_erp_action *);

static int zfcp_erp_unit_strategy(struct zfcp_erp_action *);
static int zfcp_erp_unit_strategy_clearstati(struct zfcp_unit *);
static int zfcp_erp_unit_strategy_close(struct zfcp_erp_action *);
static int zfcp_erp_unit_strategy_open(struct zfcp_erp_action *);

80
static void zfcp_erp_action_dismiss_adapter(struct zfcp_adapter *);
81 82 83
static void zfcp_erp_action_dismiss_port(struct zfcp_port *);
static void zfcp_erp_action_dismiss_unit(struct zfcp_unit *);
static void zfcp_erp_action_dismiss(struct zfcp_erp_action *);
L
Linus Torvalds 已提交
84 85

static int zfcp_erp_action_enqueue(int, struct zfcp_adapter *,
86
				   struct zfcp_port *, struct zfcp_unit *,
87
				   u8 id, void *ref);
L
Linus Torvalds 已提交
88 89 90 91 92 93 94 95
static int zfcp_erp_action_dequeue(struct zfcp_erp_action *);
static void zfcp_erp_action_cleanup(int, struct zfcp_adapter *,
				    struct zfcp_port *, struct zfcp_unit *,
				    int);

static void zfcp_erp_action_ready(struct zfcp_erp_action *);
static int  zfcp_erp_action_exists(struct zfcp_erp_action *);

96 97
static void zfcp_erp_action_to_ready(struct zfcp_erp_action *);
static void zfcp_erp_action_to_running(struct zfcp_erp_action *);
L
Linus Torvalds 已提交
98 99 100

static void zfcp_erp_memwait_handler(unsigned long);

101
/**
102
 * zfcp_close_fsf - stop FSF operations for an adapter
103
 *
104 105 106
 * Dismiss and cleanup all pending fsf_reqs (this wakes up all initiators of
 * requests waiting for completion; especially this returns SCSI commands
 * with error state).
L
Linus Torvalds 已提交
107
 */
108
static void zfcp_close_fsf(struct zfcp_adapter *adapter)
L
Linus Torvalds 已提交
109
{
110
	/* close queues to ensure that buffers are not accessed by adapter */
S
Swen Schillig 已提交
111
	zfcp_qdio_close(adapter);
112 113 114 115
	zfcp_fsf_req_dismiss_all(adapter);
	/* reset FSF request sequence number */
	adapter->fsf_req_seq_no = 0;
	/* all ports and units are closed */
116
	zfcp_erp_modify_adapter_status(adapter, 24, NULL,
117 118
				       ZFCP_STATUS_COMMON_OPEN, ZFCP_CLEAR);
}
L
Linus Torvalds 已提交
119

120 121 122 123 124 125 126 127 128 129 130 131
/**
 * zfcp_fsf_request_timeout_handler - called if a request timed out
 * @data: pointer to adapter for handler function
 *
 * This function needs to be called if requests (ELS, Generic Service,
 * or SCSI commands) exceed a certain time limit. The assumption is
 * that after the time limit the adapter get stuck. So we trigger a reopen of
 * the adapter.
 */
static void zfcp_fsf_request_timeout_handler(unsigned long data)
{
	struct zfcp_adapter *adapter = (struct zfcp_adapter *) data;
132 133
	zfcp_erp_adapter_reopen(adapter, ZFCP_STATUS_COMMON_ERP_FAILED, 62,
				NULL);
L
Linus Torvalds 已提交
134 135
}

136 137 138 139
void zfcp_fsf_start_timer(struct zfcp_fsf_req *fsf_req, unsigned long timeout)
{
	fsf_req->timer.function = zfcp_fsf_request_timeout_handler;
	fsf_req->timer.data = (unsigned long) fsf_req->adapter;
140
	fsf_req->timer.expires = jiffies + timeout;
141 142 143
	add_timer(&fsf_req->timer);
}

L
Linus Torvalds 已提交
144
/*
S
Swen Schillig 已提交
145
 * function:
L
Linus Torvalds 已提交
146 147 148 149 150
 *
 * purpose:	called if an adapter failed,
 *		initiates adapter recovery which is done
 *		asynchronously
 *
A
Andreas Mohr 已提交
151
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
152 153
 *		<0	- failed to initiate action
 */
154
static int zfcp_erp_adapter_reopen_internal(struct zfcp_adapter *adapter,
155
					    int clear_mask, u8 id, void *ref)
L
Linus Torvalds 已提交
156 157 158 159 160 161 162
{
	int retval;

	zfcp_erp_adapter_block(adapter, clear_mask);

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &adapter->status)) {
		/* ensure propagation of failed status to new devices */
163
		zfcp_erp_adapter_failed(adapter, 13, NULL);
L
Linus Torvalds 已提交
164 165 166 167
		retval = -EIO;
		goto out;
	}
	retval = zfcp_erp_action_enqueue(ZFCP_ERP_ACTION_REOPEN_ADAPTER,
168
					 adapter, NULL, NULL, id, ref);
L
Linus Torvalds 已提交
169 170 171 172 173 174

 out:
	return retval;
}

/*
S
Swen Schillig 已提交
175
 * function:
L
Linus Torvalds 已提交
176 177 178 179
 *
 * purpose:	Wrappper for zfcp_erp_adapter_reopen_internal
 *              used to ensure the correct locking
 *
A
Andreas Mohr 已提交
180
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
181 182
 *		<0	- failed to initiate action
 */
183
int zfcp_erp_adapter_reopen(struct zfcp_adapter *adapter, int clear_mask,
184
			    u8 id, void *ref)
L
Linus Torvalds 已提交
185 186 187 188 189 190
{
	int retval;
	unsigned long flags;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);
191
	retval = zfcp_erp_adapter_reopen_internal(adapter, clear_mask, id, ref);
L
Linus Torvalds 已提交
192 193 194 195 196 197
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);

	return retval;
}

198
int zfcp_erp_adapter_shutdown(struct zfcp_adapter *adapter, int clear_mask,
199
			      u8 id, void *ref)
L
Linus Torvalds 已提交
200 201 202 203 204 205
{
	int retval;

	retval = zfcp_erp_adapter_reopen(adapter,
					 ZFCP_STATUS_COMMON_RUNNING |
					 ZFCP_STATUS_COMMON_ERP_FAILED |
206
					 clear_mask, id, ref);
L
Linus Torvalds 已提交
207 208 209 210

	return retval;
}

211
int zfcp_erp_port_shutdown(struct zfcp_port *port, int clear_mask, u8 id,
212
			   void *ref)
L
Linus Torvalds 已提交
213 214 215 216 217 218
{
	int retval;

	retval = zfcp_erp_port_reopen(port,
				      ZFCP_STATUS_COMMON_RUNNING |
				      ZFCP_STATUS_COMMON_ERP_FAILED |
219
				      clear_mask, id, ref);
L
Linus Torvalds 已提交
220 221 222 223

	return retval;
}

224
int zfcp_erp_unit_shutdown(struct zfcp_unit *unit, int clear_mask, u8 id,
225
			   void *ref)
L
Linus Torvalds 已提交
226 227 228 229 230 231
{
	int retval;

	retval = zfcp_erp_unit_reopen(unit,
				      ZFCP_STATUS_COMMON_RUNNING |
				      ZFCP_STATUS_COMMON_ERP_FAILED |
232
				      clear_mask, id, ref);
L
Linus Torvalds 已提交
233 234 235 236 237

	return retval;
}

/*
S
Swen Schillig 已提交
238
 * function:
L
Linus Torvalds 已提交
239 240 241 242 243
 *
 * purpose:	called if a port failed to be opened normally
 *		initiates Forced Reopen recovery which is done
 *		asynchronously
 *
A
Andreas Mohr 已提交
244
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
245 246
 *		<0	- failed to initiate action
 */
247
static int zfcp_erp_port_forced_reopen_internal(struct zfcp_port *port,
248 249
						int clear_mask, u8 id,
						void *ref)
L
Linus Torvalds 已提交
250 251 252 253 254 255 256 257 258 259 260
{
	int retval;

	zfcp_erp_port_block(port, clear_mask);

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &port->status)) {
		retval = -EIO;
		goto out;
	}

	retval = zfcp_erp_action_enqueue(ZFCP_ERP_ACTION_REOPEN_PORT_FORCED,
261
					 port->adapter, port, NULL, id, ref);
L
Linus Torvalds 已提交
262 263 264 265 266 267

 out:
	return retval;
}

/*
S
Swen Schillig 已提交
268
 * function:
L
Linus Torvalds 已提交
269 270 271 272
 *
 * purpose:	Wrappper for zfcp_erp_port_forced_reopen_internal
 *              used to ensure the correct locking
 *
A
Andreas Mohr 已提交
273
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
274 275
 *		<0	- failed to initiate action
 */
276
int zfcp_erp_port_forced_reopen(struct zfcp_port *port, int clear_mask, u8 id,
277
				void *ref)
L
Linus Torvalds 已提交
278 279 280 281 282 283 284 285
{
	int retval;
	unsigned long flags;
	struct zfcp_adapter *adapter;

	adapter = port->adapter;
	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);
286 287
	retval = zfcp_erp_port_forced_reopen_internal(port, clear_mask, id,
						      ref);
L
Linus Torvalds 已提交
288 289 290 291 292 293 294
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);

	return retval;
}

/*
S
Swen Schillig 已提交
295
 * function:
L
Linus Torvalds 已提交
296 297 298 299 300
 *
 * purpose:	called if a port is to be opened
 *		initiates Reopen recovery which is done
 *		asynchronously
 *
A
Andreas Mohr 已提交
301
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
302 303
 *		<0	- failed to initiate action
 */
304
static int zfcp_erp_port_reopen_internal(struct zfcp_port *port, int clear_mask,
305
					 u8 id, void *ref)
L
Linus Torvalds 已提交
306 307 308 309 310 311 312
{
	int retval;

	zfcp_erp_port_block(port, clear_mask);

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &port->status)) {
		/* ensure propagation of failed status to new devices */
313
		zfcp_erp_port_failed(port, 14, NULL);
L
Linus Torvalds 已提交
314 315 316 317 318
		retval = -EIO;
		goto out;
	}

	retval = zfcp_erp_action_enqueue(ZFCP_ERP_ACTION_REOPEN_PORT,
319
					 port->adapter, port, NULL, id, ref);
L
Linus Torvalds 已提交
320 321 322 323 324 325 326 327 328 329 330 331 332 333 334

 out:
	return retval;
}

/**
 * zfcp_erp_port_reopen - initiate reopen of a remote port
 * @port: port to be reopened
 * @clear_mask: specifies flags in port status to be cleared
 * Return: 0 on success, < 0 on error
 *
 * This is a wrappper function for zfcp_erp_port_reopen_internal. It ensures
 * correct locking. An error recovery task is initiated to do the reopen.
 * To wait for the completion of the reopen zfcp_erp_wait should be used.
 */
335 336
int zfcp_erp_port_reopen(struct zfcp_port *port, int clear_mask, u8 id,
			 void *ref)
L
Linus Torvalds 已提交
337 338 339 340 341 342 343
{
	int retval;
	unsigned long flags;
	struct zfcp_adapter *adapter = port->adapter;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);
344
	retval = zfcp_erp_port_reopen_internal(port, clear_mask, id, ref);
L
Linus Torvalds 已提交
345 346 347 348 349 350 351
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);

	return retval;
}

/*
S
Swen Schillig 已提交
352
 * function:
L
Linus Torvalds 已提交
353 354 355 356 357
 *
 * purpose:	called if a unit is to be opened
 *		initiates Reopen recovery which is done
 *		asynchronously
 *
A
Andreas Mohr 已提交
358
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
359 360
 *		<0	- failed to initiate action
 */
361
static int zfcp_erp_unit_reopen_internal(struct zfcp_unit *unit, int clear_mask,
362
					 u8 id, void *ref)
L
Linus Torvalds 已提交
363 364 365 366 367 368 369 370 371 372 373 374
{
	int retval;
	struct zfcp_adapter *adapter = unit->port->adapter;

	zfcp_erp_unit_block(unit, clear_mask);

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &unit->status)) {
		retval = -EIO;
		goto out;
	}

	retval = zfcp_erp_action_enqueue(ZFCP_ERP_ACTION_REOPEN_UNIT,
375
					 adapter, unit->port, unit, id, ref);
L
Linus Torvalds 已提交
376 377 378 379 380 381 382 383 384 385 386 387 388 389
 out:
	return retval;
}

/**
 * zfcp_erp_unit_reopen - initiate reopen of a unit
 * @unit: unit to be reopened
 * @clear_mask: specifies flags in unit status to be cleared
 * Return: 0 on success, < 0 on error
 *
 * This is a wrappper for zfcp_erp_unit_reopen_internal. It ensures correct
 * locking. An error recovery task is initiated to do the reopen.
 * To wait for the completion of the reopen zfcp_erp_wait should be used.
 */
390 391
int zfcp_erp_unit_reopen(struct zfcp_unit *unit, int clear_mask, u8 id,
			 void *ref)
L
Linus Torvalds 已提交
392 393 394 395 396 397 398 399 400 401 402
{
	int retval;
	unsigned long flags;
	struct zfcp_adapter *adapter;
	struct zfcp_port *port;

	port = unit->port;
	adapter = port->adapter;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);
403
	retval = zfcp_erp_unit_reopen_internal(unit, clear_mask, id, ref);
L
Linus Torvalds 已提交
404 405 406 407 408 409
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);

	return retval;
}

410 411
/**
 * zfcp_erp_adapter_block - mark adapter as blocked, block scsi requests
L
Linus Torvalds 已提交
412
 */
413
static void zfcp_erp_adapter_block(struct zfcp_adapter *adapter, int clear_mask)
L
Linus Torvalds 已提交
414
{
415
	zfcp_erp_modify_adapter_status(adapter, 15, NULL,
L
Linus Torvalds 已提交
416 417 418 419
				       ZFCP_STATUS_COMMON_UNBLOCKED |
				       clear_mask, ZFCP_CLEAR);
}

420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443
/* FIXME: isn't really atomic */
/*
 * returns the mask which has not been set so far, i.e.
 * 0 if no bit has been changed, !0 if some bit has been changed
 */
static int atomic_test_and_set_mask(unsigned long mask, atomic_t *v)
{
	int changed_bits = (atomic_read(v) /*XOR*/^ mask) & mask;
	atomic_set_mask(mask, v);
	return changed_bits;
}

/* FIXME: isn't really atomic */
/*
 * returns the mask which has not been cleared so far, i.e.
 * 0 if no bit has been changed, !0 if some bit has been changed
 */
static int atomic_test_and_clear_mask(unsigned long mask, atomic_t *v)
{
	int changed_bits = atomic_read(v) & mask;
	atomic_clear_mask(mask, v);
	return changed_bits;
}

444 445
/**
 * zfcp_erp_adapter_unblock - mark adapter as unblocked, allow scsi requests
L
Linus Torvalds 已提交
446
 */
447
static void zfcp_erp_adapter_unblock(struct zfcp_adapter *adapter)
L
Linus Torvalds 已提交
448
{
449 450
	if (atomic_test_and_set_mask(ZFCP_STATUS_COMMON_UNBLOCKED,
				     &adapter->status))
451
		zfcp_rec_dbf_event_adapter(16, NULL, adapter);
L
Linus Torvalds 已提交
452 453 454
}

/*
S
Swen Schillig 已提交
455
 * function:
L
Linus Torvalds 已提交
456 457 458 459 460 461 462 463 464 465
 *
 * purpose:	disable I/O,
 *		return any open requests and clean them up,
 *		aim: no pending and incoming I/O
 *
 * returns:
 */
static void
zfcp_erp_port_block(struct zfcp_port *port, int clear_mask)
{
466
	zfcp_erp_modify_port_status(port, 17, NULL,
L
Linus Torvalds 已提交
467 468 469 470 471
				    ZFCP_STATUS_COMMON_UNBLOCKED | clear_mask,
				    ZFCP_CLEAR);
}

/*
S
Swen Schillig 已提交
472
 * function:
L
Linus Torvalds 已提交
473 474 475 476 477 478 479 480
 *
 * purpose:	enable I/O
 *
 * returns:
 */
static void
zfcp_erp_port_unblock(struct zfcp_port *port)
{
481 482
	if (atomic_test_and_set_mask(ZFCP_STATUS_COMMON_UNBLOCKED,
				     &port->status))
483
		zfcp_rec_dbf_event_port(18, NULL, port);
L
Linus Torvalds 已提交
484 485 486
}

/*
S
Swen Schillig 已提交
487
 * function:
L
Linus Torvalds 已提交
488 489 490 491 492 493 494 495 496 497
 *
 * purpose:	disable I/O,
 *		return any open requests and clean them up,
 *		aim: no pending and incoming I/O
 *
 * returns:
 */
static void
zfcp_erp_unit_block(struct zfcp_unit *unit, int clear_mask)
{
498
	zfcp_erp_modify_unit_status(unit, 19, NULL,
L
Linus Torvalds 已提交
499 500 501 502 503
				    ZFCP_STATUS_COMMON_UNBLOCKED | clear_mask,
				    ZFCP_CLEAR);
}

/*
S
Swen Schillig 已提交
504
 * function:
L
Linus Torvalds 已提交
505 506 507 508 509 510 511 512
 *
 * purpose:	enable I/O
 *
 * returns:
 */
static void
zfcp_erp_unit_unblock(struct zfcp_unit *unit)
{
513 514
	if (atomic_test_and_set_mask(ZFCP_STATUS_COMMON_UNBLOCKED,
				     &unit->status))
515
		zfcp_rec_dbf_event_unit(20, NULL, unit);
L
Linus Torvalds 已提交
516 517 518 519 520 521 522 523 524
}

static void
zfcp_erp_action_ready(struct zfcp_erp_action *erp_action)
{
	struct zfcp_adapter *adapter = erp_action->adapter;

	zfcp_erp_action_to_ready(erp_action);
	up(&adapter->erp_ready_sem);
525
	zfcp_rec_dbf_event_thread(2, adapter);
L
Linus Torvalds 已提交
526 527 528
}

/*
S
Swen Schillig 已提交
529
 * function:
L
Linus Torvalds 已提交
530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
 *
 * purpose:
 *
 * returns:	<0			erp_action not found in any list
 *		ZFCP_ERP_ACTION_READY	erp_action is in ready list
 *		ZFCP_ERP_ACTION_RUNNING	erp_action is in running list
 *
 * locks:	erp_lock must be held
 */
static int
zfcp_erp_action_exists(struct zfcp_erp_action *erp_action)
{
	int retval = -EINVAL;
	struct list_head *entry;
	struct zfcp_erp_action *entry_erp_action;
	struct zfcp_adapter *adapter = erp_action->adapter;

	/* search in running list */
	list_for_each(entry, &adapter->erp_running_head) {
		entry_erp_action =
		    list_entry(entry, struct zfcp_erp_action, list);
		if (entry_erp_action == erp_action) {
			retval = ZFCP_ERP_ACTION_RUNNING;
			goto out;
		}
	}
	/* search in ready list */
	list_for_each(entry, &adapter->erp_ready_head) {
		entry_erp_action =
		    list_entry(entry, struct zfcp_erp_action, list);
		if (entry_erp_action == erp_action) {
			retval = ZFCP_ERP_ACTION_READY;
			goto out;
		}
	}

 out:
	return retval;
}

/*
 * purpose:	checks current status of action (timed out, dismissed, ...)
 *		and does appropriate preparations (dismiss fsf request, ...)
 *
 * locks:	called under erp_lock (disabled interrupts)
 */
576
static void
L
Linus Torvalds 已提交
577 578 579 580 581
zfcp_erp_strategy_check_fsfreq(struct zfcp_erp_action *erp_action)
{
	struct zfcp_adapter *adapter = erp_action->adapter;

	if (erp_action->fsf_req) {
582 583
		/* take lock to ensure that request is not deleted meanwhile */
		spin_lock(&adapter->req_list_lock);
584 585
		if (zfcp_reqlist_find_safe(adapter, erp_action->fsf_req) &&
		    erp_action->fsf_req->erp_action == erp_action) {
L
Linus Torvalds 已提交
586
			/* fsf_req still exists */
587
			/* dismiss fsf_req of timed out/dismissed erp_action */
L
Linus Torvalds 已提交
588 589
			if (erp_action->status & (ZFCP_STATUS_ERP_DISMISSED |
						  ZFCP_STATUS_ERP_TIMEDOUT)) {
590 591
				erp_action->fsf_req->status |=
					ZFCP_STATUS_FSFREQ_DISMISSED;
592
				zfcp_rec_dbf_event_action(142, erp_action);
L
Linus Torvalds 已提交
593
			}
C
Christof Schmitt 已提交
594
			if (erp_action->status & ZFCP_STATUS_ERP_TIMEDOUT)
595
				zfcp_rec_dbf_event_action(143, erp_action);
L
Linus Torvalds 已提交
596 597 598 599 600
			/*
			 * If fsf_req is neither dismissed nor completed
			 * then keep it running asynchronously and don't mess
			 * with the association of erp_action and fsf_req.
			 */
601 602
			if (erp_action->fsf_req->status &
					(ZFCP_STATUS_FSFREQ_COMPLETED |
L
Linus Torvalds 已提交
603 604 605 606 607 608 609 610 611 612 613 614
					       ZFCP_STATUS_FSFREQ_DISMISSED)) {
				/* forget about association between fsf_req
				   and erp_action */
				erp_action->fsf_req = NULL;
			}
		} else {
			/*
			 * even if this fsf_req has gone, forget about
			 * association between erp_action and fsf_req
			 */
			erp_action->fsf_req = NULL;
		}
615
		spin_unlock(&adapter->req_list_lock);
616
	}
L
Linus Torvalds 已提交
617 618
}

619 620
/**
 * zfcp_erp_async_handler_nolock - complete erp_action
L
Linus Torvalds 已提交
621
 *
622 623
 * Used for normal completion, time-out, dismissal and failure after
 * low memory condition.
L
Linus Torvalds 已提交
624
 */
625 626
static void zfcp_erp_async_handler_nolock(struct zfcp_erp_action *erp_action,
					  unsigned long set_mask)
L
Linus Torvalds 已提交
627 628 629 630 631 632 633 634 635
{
	if (zfcp_erp_action_exists(erp_action) == ZFCP_ERP_ACTION_RUNNING) {
		erp_action->status |= set_mask;
		zfcp_erp_action_ready(erp_action);
	} else {
		/* action is ready or gone - nothing to do */
	}
}

636 637
/**
 * zfcp_erp_async_handler - wrapper for erp_async_handler_nolock w/ locking
L
Linus Torvalds 已提交
638
 */
639 640
void zfcp_erp_async_handler(struct zfcp_erp_action *erp_action,
			    unsigned long set_mask)
L
Linus Torvalds 已提交
641 642 643 644 645
{
	struct zfcp_adapter *adapter = erp_action->adapter;
	unsigned long flags;

	write_lock_irqsave(&adapter->erp_lock, flags);
646
	zfcp_erp_async_handler_nolock(erp_action, set_mask);
L
Linus Torvalds 已提交
647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667
	write_unlock_irqrestore(&adapter->erp_lock, flags);
}

/*
 * purpose:	is called for erp_action which was slept waiting for
 *		memory becoming avaliable,
 *		will trigger that this action will be continued
 */
static void
zfcp_erp_memwait_handler(unsigned long data)
{
	struct zfcp_erp_action *erp_action = (struct zfcp_erp_action *) data;

	zfcp_erp_async_handler(erp_action, 0);
}

/*
 * purpose:	is called if an asynchronous erp step timed out,
 *		action gets an appropriate flag and will be processed
 *		accordingly
 */
668
static void zfcp_erp_timeout_handler(unsigned long data)
L
Linus Torvalds 已提交
669 670 671 672 673 674
{
	struct zfcp_erp_action *erp_action = (struct zfcp_erp_action *) data;

	zfcp_erp_async_handler(erp_action, ZFCP_STATUS_ERP_TIMEDOUT);
}

675 676
/**
 * zfcp_erp_action_dismiss - dismiss an erp_action
L
Linus Torvalds 已提交
677
 *
678
 * adapter->erp_lock must be held
S
Swen Schillig 已提交
679
 *
680 681
 * Dismissal of an erp_action is usually required if an erp_action of
 * higher priority is generated.
L
Linus Torvalds 已提交
682
 */
683
static void zfcp_erp_action_dismiss(struct zfcp_erp_action *erp_action)
L
Linus Torvalds 已提交
684
{
685 686 687
	erp_action->status |= ZFCP_STATUS_ERP_DISMISSED;
	if (zfcp_erp_action_exists(erp_action) == ZFCP_ERP_ACTION_RUNNING)
		zfcp_erp_action_ready(erp_action);
L
Linus Torvalds 已提交
688 689 690 691 692 693 694 695 696 697
}

int
zfcp_erp_thread_setup(struct zfcp_adapter *adapter)
{
	int retval = 0;

	atomic_clear_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_UP, &adapter->status);

	retval = kernel_thread(zfcp_erp_thread, adapter, SIGCHLD);
C
Christof Schmitt 已提交
698 699 700 701
	if (retval < 0)
		dev_err(&adapter->ccw_device->dev,
			"Creation of ERP thread failed.\n");
	else {
L
Linus Torvalds 已提交
702 703 704 705 706 707 708 709 710
		wait_event(adapter->erp_thread_wqh,
			   atomic_test_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_UP,
					    &adapter->status));
	}

	return (retval < 0);
}

/*
S
Swen Schillig 已提交
711
 * function:
L
Linus Torvalds 已提交
712
 *
S
Swen Schillig 已提交
713
 * purpose:
L
Linus Torvalds 已提交
714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730
 *
 * returns:
 *
 * context:	process (i.e. proc-fs or rmmod/insmod)
 *
 * note:	The caller of this routine ensures that the specified
 *		adapter has been shut down and that this operation
 *		has been completed. Thus, there are no pending erp_actions
 *		which would need to be handled here.
 */
int
zfcp_erp_thread_kill(struct zfcp_adapter *adapter)
{
	int retval = 0;

	atomic_set_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_KILL, &adapter->status);
	up(&adapter->erp_ready_sem);
731
	zfcp_rec_dbf_event_thread_lock(2, adapter);
L
Linus Torvalds 已提交
732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767

	wait_event(adapter->erp_thread_wqh,
		   !atomic_test_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_UP,
				     &adapter->status));

	atomic_clear_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_KILL,
			  &adapter->status);

	return retval;
}

/*
 * purpose:	is run as a kernel thread,
 *		goes through list of error recovery actions of associated adapter
 *		and delegates single action to execution
 *
 * returns:	0
 */
static int
zfcp_erp_thread(void *data)
{
	struct zfcp_adapter *adapter = (struct zfcp_adapter *) data;
	struct list_head *next;
	struct zfcp_erp_action *erp_action;
	unsigned long flags;

	daemonize("zfcperp%s", zfcp_get_busid_by_adapter(adapter));
	/* Block all signals */
	siginitsetinv(&current->blocked, 0);
	atomic_set_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_UP, &adapter->status);
	wake_up(&adapter->erp_thread_wqh);

	while (!atomic_test_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_KILL,
				 &adapter->status)) {

		write_lock_irqsave(&adapter->erp_lock, flags);
768
		next = adapter->erp_ready_head.next;
L
Linus Torvalds 已提交
769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785
		write_unlock_irqrestore(&adapter->erp_lock, flags);

		if (next != &adapter->erp_ready_head) {
			erp_action =
			    list_entry(next, struct zfcp_erp_action, list);
			/*
			 * process action (incl. [re]moving it
			 * from 'ready' queue)
			 */
			zfcp_erp_strategy(erp_action);
		}

		/*
		 * sleep as long as there is nothing to do, i.e.
		 * no action in 'ready' queue to be processed and
		 * thread is not to be killed
		 */
786
		zfcp_rec_dbf_event_thread_lock(4, adapter);
L
Linus Torvalds 已提交
787
		down_interruptible(&adapter->erp_ready_sem);
788
		zfcp_rec_dbf_event_thread_lock(5, adapter);
L
Linus Torvalds 已提交
789 790 791 792 793 794 795 796 797
	}

	atomic_clear_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_UP, &adapter->status);
	wake_up(&adapter->erp_thread_wqh);

	return 0;
}

/*
S
Swen Schillig 已提交
798
 * function:
L
Linus Torvalds 已提交
799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
 *
 * purpose:	drives single error recovery action and schedules higher and
 *		subordinate actions, if necessary
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_SUCCEEDED	- action finished successfully (deqd)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully (deqd)
 *		ZFCP_ERP_EXIT		- action finished (dequeued), offline
 *		ZFCP_ERP_DISMISSED	- action canceled (dequeued)
 */
static int
zfcp_erp_strategy(struct zfcp_erp_action *erp_action)
{
	int retval = 0;
	struct zfcp_adapter *adapter = erp_action->adapter;
	struct zfcp_port *port = erp_action->port;
	struct zfcp_unit *unit = erp_action->unit;
	int action = erp_action->action;
	u32 status = erp_action->status;
	unsigned long flags;

	/* serialise dismissing, timing out, moving, enqueueing */
	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);

	/* dequeue dismissed action and leave, if required */
	retval = zfcp_erp_strategy_check_action(erp_action, retval);
	if (retval == ZFCP_ERP_DISMISSED) {
		goto unlock;
	}

	/*
	 * move action to 'running' queue before processing it
	 * (to avoid a race condition regarding moving the
	 * action to the 'running' queue and back)
	 */
	zfcp_erp_action_to_running(erp_action);

	/*
	 * try to process action as far as possible,
	 * no lock to allow for blocking operations (kmalloc, qdio, ...),
	 * afterwards the lock is required again for the following reasons:
	 * - dequeueing of finished action and enqueueing of
	 *   follow-up actions must be atomic so that any other
	 *   reopen-routine does not believe there is nothing to do
	 *   and that it is safe to enqueue something else,
	 * - we want to force any control thread which is dismissing
	 *   actions to finish this before we decide about
	 *   necessary steps to be taken here further
	 */
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);
	retval = zfcp_erp_strategy_do_action(erp_action);
	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);

	/*
	 * check for dismissed status again to avoid follow-up actions,
857 858
	 * failing of targets and so on for dismissed actions,
	 * we go through down() here because there has been an up()
L
Linus Torvalds 已提交
859
	 */
860 861
	if (erp_action->status & ZFCP_STATUS_ERP_DISMISSED)
		retval = ZFCP_ERP_CONTINUES;
L
Linus Torvalds 已提交
862 863 864 865 866 867 868 869 870 871 872 873 874 875

	switch (retval) {
	case ZFCP_ERP_NOMEM:
		/* no memory to continue immediately, let it sleep */
		if (!(erp_action->status & ZFCP_STATUS_ERP_LOWMEM)) {
			++adapter->erp_low_mem_count;
			erp_action->status |= ZFCP_STATUS_ERP_LOWMEM;
		}
		/* This condition is true if there is no memory available
		   for any erp_action on this adapter. This implies that there
		   are no elements in the memory pool(s) left for erp_actions.
		   This might happen if an erp_action that used a memory pool
		   element was timed out.
		 */
C
Christof Schmitt 已提交
876
		if (adapter->erp_total_count == adapter->erp_low_mem_count)
877
			zfcp_erp_adapter_reopen_internal(adapter, 0, 66, NULL);
C
Christof Schmitt 已提交
878 879
		else
			retval = zfcp_erp_strategy_memwait(erp_action);
L
Linus Torvalds 已提交
880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
		goto unlock;
	case ZFCP_ERP_CONTINUES:
		/* leave since this action runs asynchronously */
		if (erp_action->status & ZFCP_STATUS_ERP_LOWMEM) {
			--adapter->erp_low_mem_count;
			erp_action->status &= ~ZFCP_STATUS_ERP_LOWMEM;
		}
		goto unlock;
	}
	/* ok, finished action (whatever its result is) */

	/* check for unrecoverable targets */
	retval = zfcp_erp_strategy_check_target(erp_action, retval);

	/* action must be dequeued (here to allow for further ones) */
	zfcp_erp_action_dequeue(erp_action);

	/*
	 * put this target through the erp mill again if someone has
S
Swen Schillig 已提交
899
	 * requested to change the status of a target being online
L
Linus Torvalds 已提交
900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
	 * to offline or the other way around
	 * (old retval is preserved if nothing has to be done here)
	 */
	retval = zfcp_erp_strategy_statechange(action, status, adapter,
					       port, unit, retval);

	/*
	 * leave if target is in permanent error state or if
	 * action is repeated in order to process state change
	 */
	if (retval == ZFCP_ERP_EXIT) {
		goto unlock;
	}

	/* trigger follow up actions */
	zfcp_erp_strategy_followup_actions(action, adapter, port, unit, retval);

 unlock:
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);
S
Swen Schillig 已提交
920

L
Linus Torvalds 已提交
921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939
	if (retval != ZFCP_ERP_CONTINUES)
		zfcp_erp_action_cleanup(action, adapter, port, unit, retval);

	/*
	 * a few tasks remain when the erp queues are empty
	 * (don't do that if the last action evaluated was dismissed
	 * since this clearly indicates that there is more to come) :
	 * - close the name server port if it is open yet
	 *   (enqueues another [probably] final action)
	 * - otherwise, wake up whoever wants to be woken when we are
	 *   done with erp
	 */
	if (retval != ZFCP_ERP_DISMISSED)
		zfcp_erp_strategy_check_queues(adapter);

	return retval;
}

/*
S
Swen Schillig 已提交
940
 * function:
L
Linus Torvalds 已提交
941
 *
S
Swen Schillig 已提交
942
 * purpose:
L
Linus Torvalds 已提交
943 944 945 946 947 948 949 950 951 952 953 954
 *
 * returns:	ZFCP_ERP_DISMISSED	- if action has been dismissed
 *		retval			- otherwise
 */
static int
zfcp_erp_strategy_check_action(struct zfcp_erp_action *erp_action, int retval)
{
	zfcp_erp_strategy_check_fsfreq(erp_action);

	if (erp_action->status & ZFCP_STATUS_ERP_DISMISSED) {
		zfcp_erp_action_dequeue(erp_action);
		retval = ZFCP_ERP_DISMISSED;
955
	}
L
Linus Torvalds 已提交
956 957 958 959 960 961 962 963 964 965 966 967 968 969

	return retval;
}

static int
zfcp_erp_strategy_do_action(struct zfcp_erp_action *erp_action)
{
	int retval = ZFCP_ERP_FAILED;

	/*
	 * try to execute/continue action as far as possible,
	 * note: no lock in subsequent strategy routines
	 * (this allows these routine to call schedule, e.g.
	 * kmalloc with such flags or qdio_initialize & friends)
970
	 * Note: in case of timeout, the separate strategies will fail
L
Linus Torvalds 已提交
971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996
	 * anyhow. No need for a special action. Even worse, a nameserver
	 * failure would not wake up waiting ports without the call.
	 */
	switch (erp_action->action) {

	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		retval = zfcp_erp_adapter_strategy(erp_action);
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
		retval = zfcp_erp_port_forced_strategy(erp_action);
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT:
		retval = zfcp_erp_port_strategy(erp_action);
		break;

	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		retval = zfcp_erp_unit_strategy(erp_action);
		break;
	}

	return retval;
}

/*
S
Swen Schillig 已提交
997
 * function:
L
Linus Torvalds 已提交
998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017
 *
 * purpose:	triggers retry of this action after a certain amount of time
 *		by means of timer provided by erp_action
 *
 * returns:	ZFCP_ERP_CONTINUES - erp_action sleeps in erp running queue
 */
static int
zfcp_erp_strategy_memwait(struct zfcp_erp_action *erp_action)
{
	int retval = ZFCP_ERP_CONTINUES;

	init_timer(&erp_action->timer);
	erp_action->timer.function = zfcp_erp_memwait_handler;
	erp_action->timer.data = (unsigned long) erp_action;
	erp_action->timer.expires = jiffies + ZFCP_ERP_MEMWAIT_TIMEOUT;
	add_timer(&erp_action->timer);

	return retval;
}

S
Swen Schillig 已提交
1018
/*
L
Linus Torvalds 已提交
1019 1020 1021 1022 1023 1024
 * function:    zfcp_erp_adapter_failed
 *
 * purpose:     sets the adapter and all underlying devices to ERP_FAILED
 *
 */
void
1025
zfcp_erp_adapter_failed(struct zfcp_adapter *adapter, u8 id, void *ref)
L
Linus Torvalds 已提交
1026
{
1027
	zfcp_erp_modify_adapter_status(adapter, id, ref,
L
Linus Torvalds 已提交
1028
				       ZFCP_STATUS_COMMON_ERP_FAILED, ZFCP_SET);
C
Christof Schmitt 已提交
1029
	dev_err(&adapter->ccw_device->dev, "Adapter ERP failed.\n");
L
Linus Torvalds 已提交
1030 1031
}

S
Swen Schillig 已提交
1032
/*
L
Linus Torvalds 已提交
1033 1034 1035 1036 1037 1038
 * function:    zfcp_erp_port_failed
 *
 * purpose:     sets the port and all underlying devices to ERP_FAILED
 *
 */
void
1039
zfcp_erp_port_failed(struct zfcp_port *port, u8 id, void *ref)
L
Linus Torvalds 已提交
1040
{
1041
	zfcp_erp_modify_port_status(port, id, ref,
L
Linus Torvalds 已提交
1042 1043 1044
				    ZFCP_STATUS_COMMON_ERP_FAILED, ZFCP_SET);

	if (atomic_test_mask(ZFCP_STATUS_PORT_WKA, &port->status))
C
Christof Schmitt 已提交
1045 1046 1047
		dev_err(&port->adapter->ccw_device->dev,
			"Port ERP failed for WKA port d_id=0x%06x.\n",
			port->d_id);
L
Linus Torvalds 已提交
1048
	else
C
Christof Schmitt 已提交
1049 1050 1051
		dev_err(&port->adapter->ccw_device->dev,
			"Port ERP failed for port wwpn=0x%016Lx.\n",
			port->wwpn);
L
Linus Torvalds 已提交
1052 1053
}

S
Swen Schillig 已提交
1054
/*
L
Linus Torvalds 已提交
1055 1056 1057 1058 1059 1060
 * function:    zfcp_erp_unit_failed
 *
 * purpose:     sets the unit to ERP_FAILED
 *
 */
void
1061
zfcp_erp_unit_failed(struct zfcp_unit *unit, u8 id, void *ref)
L
Linus Torvalds 已提交
1062
{
1063
	zfcp_erp_modify_unit_status(unit, id, ref,
L
Linus Torvalds 已提交
1064 1065
				    ZFCP_STATUS_COMMON_ERP_FAILED, ZFCP_SET);

C
Christof Schmitt 已提交
1066 1067 1068
	dev_err(&unit->port->adapter->ccw_device->dev,
		"Unit ERP failed for unit 0x%016Lx on port 0x%016Lx.\n",
		unit->fcp_lun, unit->port->wwpn);
L
Linus Torvalds 已提交
1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
}

/*
 * function:	zfcp_erp_strategy_check_target
 *
 * purpose:	increments the erp action count on the device currently in
 *              recovery if the action failed or resets the count in case of
 *              success. If a maximum count is exceeded the device is marked
 *              as ERP_FAILED.
 *		The 'blocked' state of a target which has been recovered
 *              successfully is reset.
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (not considered)
S
Swen Schillig 已提交
1082
 *		ZFCP_ERP_SUCCEEDED	- action finished successfully
L
Linus Torvalds 已提交
1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
 *		ZFCP_ERP_EXIT		- action failed and will not continue
 */
static int
zfcp_erp_strategy_check_target(struct zfcp_erp_action *erp_action, int result)
{
	struct zfcp_adapter *adapter = erp_action->adapter;
	struct zfcp_port *port = erp_action->port;
	struct zfcp_unit *unit = erp_action->unit;

	switch (erp_action->action) {

	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		result = zfcp_erp_strategy_check_unit(unit, result);
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
	case ZFCP_ERP_ACTION_REOPEN_PORT:
		result = zfcp_erp_strategy_check_port(port, result);
		break;

	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		result = zfcp_erp_strategy_check_adapter(adapter, result);
		break;
	}

	return result;
}

static int
zfcp_erp_strategy_statechange(int action,
			      u32 status,
			      struct zfcp_adapter *adapter,
			      struct zfcp_port *port,
			      struct zfcp_unit *unit, int retval)
{
	switch (action) {

	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		if (zfcp_erp_strategy_statechange_detected(&adapter->status,
							   status)) {
1123 1124
			zfcp_erp_adapter_reopen_internal(adapter,
						ZFCP_STATUS_COMMON_ERP_FAILED,
1125
						67, NULL);
L
Linus Torvalds 已提交
1126 1127 1128 1129 1130 1131 1132 1133
			retval = ZFCP_ERP_EXIT;
		}
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
	case ZFCP_ERP_ACTION_REOPEN_PORT:
		if (zfcp_erp_strategy_statechange_detected(&port->status,
							   status)) {
1134 1135
			zfcp_erp_port_reopen_internal(port,
						ZFCP_STATUS_COMMON_ERP_FAILED,
1136
						68, NULL);
L
Linus Torvalds 已提交
1137 1138 1139 1140 1141 1142 1143
			retval = ZFCP_ERP_EXIT;
		}
		break;

	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		if (zfcp_erp_strategy_statechange_detected(&unit->status,
							   status)) {
1144 1145
			zfcp_erp_unit_reopen_internal(unit,
						ZFCP_STATUS_COMMON_ERP_FAILED,
1146
						69, NULL);
L
Linus Torvalds 已提交
1147 1148 1149 1150 1151 1152 1153 1154
			retval = ZFCP_ERP_EXIT;
		}
		break;
	}

	return retval;
}

1155
static int
L
Linus Torvalds 已提交
1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177
zfcp_erp_strategy_statechange_detected(atomic_t * target_status, u32 erp_status)
{
	return
	    /* take it online */
	    (atomic_test_mask(ZFCP_STATUS_COMMON_RUNNING, target_status) &&
	     (ZFCP_STATUS_ERP_CLOSE_ONLY & erp_status)) ||
	    /* take it offline */
	    (!atomic_test_mask(ZFCP_STATUS_COMMON_RUNNING, target_status) &&
	     !(ZFCP_STATUS_ERP_CLOSE_ONLY & erp_status));
}

static int
zfcp_erp_strategy_check_unit(struct zfcp_unit *unit, int result)
{
	switch (result) {
	case ZFCP_ERP_SUCCEEDED :
		atomic_set(&unit->erp_counter, 0);
		zfcp_erp_unit_unblock(unit);
		break;
	case ZFCP_ERP_FAILED :
		atomic_inc(&unit->erp_counter);
		if (atomic_read(&unit->erp_counter) > ZFCP_MAX_ERPS)
1178
			zfcp_erp_unit_failed(unit, 21, NULL);
L
Linus Torvalds 已提交
1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
		break;
	case ZFCP_ERP_EXIT :
		/* nothing */
		break;
	}

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &unit->status)) {
		zfcp_erp_unit_block(unit, 0); /* for ZFCP_ERP_SUCCEEDED */
		result = ZFCP_ERP_EXIT;
	}

	return result;
}

static int
zfcp_erp_strategy_check_port(struct zfcp_port *port, int result)
{
	switch (result) {
	case ZFCP_ERP_SUCCEEDED :
		atomic_set(&port->erp_counter, 0);
		zfcp_erp_port_unblock(port);
		break;
	case ZFCP_ERP_FAILED :
		atomic_inc(&port->erp_counter);
		if (atomic_read(&port->erp_counter) > ZFCP_MAX_ERPS)
1204
			zfcp_erp_port_failed(port, 22, NULL);
L
Linus Torvalds 已提交
1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
		break;
	case ZFCP_ERP_EXIT :
		/* nothing */
		break;
	}

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &port->status)) {
		zfcp_erp_port_block(port, 0); /* for ZFCP_ERP_SUCCEEDED */
		result = ZFCP_ERP_EXIT;
	}

	return result;
}

static int
zfcp_erp_strategy_check_adapter(struct zfcp_adapter *adapter, int result)
{
	switch (result) {
	case ZFCP_ERP_SUCCEEDED :
		atomic_set(&adapter->erp_counter, 0);
		zfcp_erp_adapter_unblock(adapter);
		break;
	case ZFCP_ERP_FAILED :
		atomic_inc(&adapter->erp_counter);
		if (atomic_read(&adapter->erp_counter) > ZFCP_MAX_ERPS)
1230
			zfcp_erp_adapter_failed(adapter, 23, NULL);
L
Linus Torvalds 已提交
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244
		break;
	case ZFCP_ERP_EXIT :
		/* nothing */
		break;
	}

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_FAILED, &adapter->status)) {
		zfcp_erp_adapter_block(adapter, 0); /* for ZFCP_ERP_SUCCEEDED */
		result = ZFCP_ERP_EXIT;
	}

	return result;
}

1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279
struct zfcp_erp_add_work {
	struct zfcp_unit  *unit;
	struct work_struct work;
};

/**
 * zfcp_erp_scsi_scan
 * @data: pointer to a struct zfcp_erp_add_work
 *
 * Registers a logical unit with the SCSI stack.
 */
static void zfcp_erp_scsi_scan(struct work_struct *work)
{
	struct zfcp_erp_add_work *p =
		container_of(work, struct zfcp_erp_add_work, work);
	struct zfcp_unit *unit = p->unit;
	struct fc_rport *rport = unit->port->rport;
	scsi_scan_target(&rport->dev, 0, rport->scsi_target_id,
			 unit->scsi_lun, 0);
	atomic_clear_mask(ZFCP_STATUS_UNIT_SCSI_WORK_PENDING, &unit->status);
	zfcp_unit_put(unit);
	kfree(p);
}

/**
 * zfcp_erp_schedule_work
 * @unit: pointer to unit which should be registered with SCSI stack
 *
 * Schedules work which registers a unit with the SCSI stack
 */
static void
zfcp_erp_schedule_work(struct zfcp_unit *unit)
{
	struct zfcp_erp_add_work *p;

1280
	p = kzalloc(sizeof(*p), GFP_KERNEL);
1281
	if (!p) {
C
Christof Schmitt 已提交
1282 1283 1284 1285
		dev_err(&unit->port->adapter->ccw_device->dev,
			"Out of resources. Could not register unit 0x%016Lx "
			"on port 0x%016Lx with SCSI stack.\n",
			unit->fcp_lun, unit->port->wwpn);
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
		return;
	}

	zfcp_unit_get(unit);
	atomic_set_mask(ZFCP_STATUS_UNIT_SCSI_WORK_PENDING, &unit->status);
	INIT_WORK(&p->work, zfcp_erp_scsi_scan);
	p->unit = unit;
	schedule_work(&p->work);
}

L
Linus Torvalds 已提交
1296
/*
S
Swen Schillig 已提交
1297
 * function:
L
Linus Torvalds 已提交
1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314
 *
 * purpose:	remaining things in good cases,
 *		escalation in bad cases
 *
 * returns:
 */
static int
zfcp_erp_strategy_followup_actions(int action,
				   struct zfcp_adapter *adapter,
				   struct zfcp_port *port,
				   struct zfcp_unit *unit, int status)
{
	/* initiate follow-up actions depending on success of finished action */
	switch (action) {

	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		if (status == ZFCP_ERP_SUCCEEDED)
1315
			zfcp_erp_port_reopen_all_internal(adapter, 0, 70, NULL);
L
Linus Torvalds 已提交
1316
		else
1317
			zfcp_erp_adapter_reopen_internal(adapter, 0, 71, NULL);
L
Linus Torvalds 已提交
1318 1319 1320 1321
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
		if (status == ZFCP_ERP_SUCCEEDED)
1322
			zfcp_erp_port_reopen_internal(port, 0, 72, NULL);
L
Linus Torvalds 已提交
1323
		else
1324
			zfcp_erp_adapter_reopen_internal(adapter, 0, 73, NULL);
L
Linus Torvalds 已提交
1325 1326 1327 1328
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT:
		if (status == ZFCP_ERP_SUCCEEDED)
1329
			zfcp_erp_unit_reopen_all_internal(port, 0, 74, NULL);
L
Linus Torvalds 已提交
1330
		else
1331
			zfcp_erp_port_forced_reopen_internal(port, 0, 75, NULL);
L
Linus Torvalds 已提交
1332 1333 1334
		break;

	case ZFCP_ERP_ACTION_REOPEN_UNIT:
1335 1336
		/* Nothing to do if status == ZFCP_ERP_SUCCEEDED */
		if (status != ZFCP_ERP_SUCCEEDED)
1337
			zfcp_erp_port_reopen_internal(unit->port, 0, 76, NULL);
L
Linus Torvalds 已提交
1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
		break;
	}

	return 0;
}

static int
zfcp_erp_strategy_check_queues(struct zfcp_adapter *adapter)
{
	unsigned long flags;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
	read_lock(&adapter->erp_lock);
	if (list_empty(&adapter->erp_ready_head) &&
	    list_empty(&adapter->erp_running_head)) {
			atomic_clear_mask(ZFCP_STATUS_ADAPTER_ERP_PENDING,
					  &adapter->status);
			wake_up(&adapter->erp_done_wqh);
1356
	}
L
Linus Torvalds 已提交
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379
	read_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);

	return 0;
}

/**
 * zfcp_erp_wait - wait for completion of error recovery on an adapter
 * @adapter: adapter for which to wait for completion of its error recovery
 * Return: 0
 */
int
zfcp_erp_wait(struct zfcp_adapter *adapter)
{
	int retval = 0;

	wait_event(adapter->erp_done_wqh,
		   !atomic_test_mask(ZFCP_STATUS_ADAPTER_ERP_PENDING,
				     &adapter->status));

	return retval;
}

1380
void zfcp_erp_modify_adapter_status(struct zfcp_adapter *adapter, u8 id,
1381
				    void *ref, u32 mask, int set_or_clear)
L
Linus Torvalds 已提交
1382 1383
{
	struct zfcp_port *port;
1384
	u32 changed, common_mask = mask & ZFCP_COMMON_FLAGS;
L
Linus Torvalds 已提交
1385 1386

	if (set_or_clear == ZFCP_SET) {
1387
		changed = atomic_test_and_set_mask(mask, &adapter->status);
L
Linus Torvalds 已提交
1388
	} else {
1389
		changed = atomic_test_and_clear_mask(mask, &adapter->status);
L
Linus Torvalds 已提交
1390 1391 1392
		if (mask & ZFCP_STATUS_COMMON_ERP_FAILED)
			atomic_set(&adapter->erp_counter, 0);
	}
1393 1394
	if (changed)
		zfcp_rec_dbf_event_adapter(id, ref, adapter);
L
Linus Torvalds 已提交
1395 1396 1397 1398

	/* Deal with all underlying devices, only pass common_mask */
	if (common_mask)
		list_for_each_entry(port, &adapter->port_list_head, list)
1399 1400
			zfcp_erp_modify_port_status(port, id, ref, common_mask,
						    set_or_clear);
L
Linus Torvalds 已提交
1401 1402 1403 1404 1405 1406 1407 1408
}

/*
 * function:	zfcp_erp_modify_port_status
 *
 * purpose:	sets the port and all underlying devices to ERP_FAILED
 *
 */
1409
void zfcp_erp_modify_port_status(struct zfcp_port *port, u8 id, void *ref,
1410
				 u32 mask, int set_or_clear)
L
Linus Torvalds 已提交
1411 1412
{
	struct zfcp_unit *unit;
1413
	u32 changed, common_mask = mask & ZFCP_COMMON_FLAGS;
L
Linus Torvalds 已提交
1414 1415

	if (set_or_clear == ZFCP_SET) {
1416
		changed = atomic_test_and_set_mask(mask, &port->status);
L
Linus Torvalds 已提交
1417
	} else {
1418
		changed = atomic_test_and_clear_mask(mask, &port->status);
L
Linus Torvalds 已提交
1419 1420 1421
		if (mask & ZFCP_STATUS_COMMON_ERP_FAILED)
			atomic_set(&port->erp_counter, 0);
	}
1422 1423
	if (changed)
		zfcp_rec_dbf_event_port(id, ref, port);
L
Linus Torvalds 已提交
1424 1425 1426 1427

	/* Modify status of all underlying devices, only pass common mask */
	if (common_mask)
		list_for_each_entry(unit, &port->unit_list_head, list)
1428 1429
			zfcp_erp_modify_unit_status(unit, id, ref, common_mask,
						    set_or_clear);
L
Linus Torvalds 已提交
1430 1431 1432 1433 1434 1435 1436 1437
}

/*
 * function:	zfcp_erp_modify_unit_status
 *
 * purpose:	sets the unit to ERP_FAILED
 *
 */
1438
void zfcp_erp_modify_unit_status(struct zfcp_unit *unit, u8 id, void *ref,
1439
				 u32 mask, int set_or_clear)
L
Linus Torvalds 已提交
1440
{
1441 1442
	u32 changed;

L
Linus Torvalds 已提交
1443
	if (set_or_clear == ZFCP_SET) {
1444
		changed = atomic_test_and_set_mask(mask, &unit->status);
L
Linus Torvalds 已提交
1445
	} else {
1446
		changed = atomic_test_and_clear_mask(mask, &unit->status);
L
Linus Torvalds 已提交
1447 1448 1449 1450
		if (mask & ZFCP_STATUS_COMMON_ERP_FAILED) {
			atomic_set(&unit->erp_counter, 0);
		}
	}
1451 1452
	if (changed)
		zfcp_rec_dbf_event_unit(id, ref, unit);
L
Linus Torvalds 已提交
1453 1454 1455
}

/*
S
Swen Schillig 已提交
1456
 * function:
L
Linus Torvalds 已提交
1457 1458 1459 1460
 *
 * purpose:	Wrappper for zfcp_erp_port_reopen_all_internal
 *              used to ensure the correct locking
 *
A
Andreas Mohr 已提交
1461
 * returns:	0	- initiated action successfully
L
Linus Torvalds 已提交
1462 1463
 *		<0	- failed to initiate action
 */
1464
int zfcp_erp_port_reopen_all(struct zfcp_adapter *adapter, int clear_mask,
1465
			     u8 id, void *ref)
L
Linus Torvalds 已提交
1466 1467 1468 1469 1470 1471
{
	int retval;
	unsigned long flags;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
	write_lock(&adapter->erp_lock);
1472 1473
	retval = zfcp_erp_port_reopen_all_internal(adapter, clear_mask, id,
						   ref);
L
Linus Torvalds 已提交
1474 1475 1476 1477 1478 1479
	write_unlock(&adapter->erp_lock);
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);

	return retval;
}

1480
static int zfcp_erp_port_reopen_all_internal(struct zfcp_adapter *adapter,
1481
					     int clear_mask, u8 id, void *ref)
L
Linus Torvalds 已提交
1482 1483 1484 1485 1486 1487
{
	int retval = 0;
	struct zfcp_port *port;

	list_for_each_entry(port, &adapter->port_list_head, list)
		if (!atomic_test_mask(ZFCP_STATUS_PORT_WKA, &port->status))
1488 1489
			zfcp_erp_port_reopen_internal(port, clear_mask, id,
						      ref);
L
Linus Torvalds 已提交
1490 1491 1492 1493 1494

	return retval;
}

/*
S
Swen Schillig 已提交
1495
 * function:
L
Linus Torvalds 已提交
1496
 *
S
Swen Schillig 已提交
1497
 * purpose:
L
Linus Torvalds 已提交
1498 1499 1500
 *
 * returns:	FIXME
 */
1501
static int zfcp_erp_unit_reopen_all_internal(struct zfcp_port *port,
1502
					     int clear_mask, u8 id, void *ref)
L
Linus Torvalds 已提交
1503 1504 1505 1506 1507
{
	int retval = 0;
	struct zfcp_unit *unit;

	list_for_each_entry(unit, &port->unit_list_head, list)
1508
		zfcp_erp_unit_reopen_internal(unit, clear_mask, id, ref);
L
Linus Torvalds 已提交
1509 1510 1511 1512 1513

	return retval;
}

/*
S
Swen Schillig 已提交
1514
 * function:
L
Linus Torvalds 已提交
1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
 *
 * purpose:	this routine executes the 'Reopen Adapter' action
 *		(the entire action is processed synchronously, since
 *		there are no actions which might be run concurrently
 *		per definition)
 *
 * returns:	ZFCP_ERP_SUCCEEDED	- action finished successfully
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_adapter_strategy(struct zfcp_erp_action *erp_action)
{
	int retval;

	retval = zfcp_erp_adapter_strategy_close(erp_action);
	if (erp_action->status & ZFCP_STATUS_ERP_CLOSE_ONLY)
		retval = ZFCP_ERP_EXIT;
	else
		retval = zfcp_erp_adapter_strategy_open(erp_action);

C
Christof Schmitt 已提交
1535
	if (retval == ZFCP_ERP_FAILED)
1536
		ssleep(ZFCP_TYPE2_RECOVERY_TIME);
L
Linus Torvalds 已提交
1537 1538 1539 1540 1541

	return retval;
}

/*
S
Swen Schillig 已提交
1542
 * function:
L
Linus Torvalds 已提交
1543
 *
S
Swen Schillig 已提交
1544
 * purpose:
L
Linus Torvalds 已提交
1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563
 *
 * returns:	ZFCP_ERP_SUCCEEDED      - action finished successfully
 *              ZFCP_ERP_FAILED         - action finished unsuccessfully
 */
static int
zfcp_erp_adapter_strategy_close(struct zfcp_erp_action *erp_action)
{
	int retval;

	atomic_set_mask(ZFCP_STATUS_COMMON_CLOSING,
			&erp_action->adapter->status);
	retval = zfcp_erp_adapter_strategy_generic(erp_action, 1);
	atomic_clear_mask(ZFCP_STATUS_COMMON_CLOSING,
			  &erp_action->adapter->status);

	return retval;
}

/*
S
Swen Schillig 已提交
1564
 * function:
L
Linus Torvalds 已提交
1565
 *
S
Swen Schillig 已提交
1566
 * purpose:
L
Linus Torvalds 已提交
1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590
 *
 * returns:	ZFCP_ERP_SUCCEEDED      - action finished successfully
 *              ZFCP_ERP_FAILED         - action finished unsuccessfully
 */
static int
zfcp_erp_adapter_strategy_open(struct zfcp_erp_action *erp_action)
{
	int retval;

	atomic_set_mask(ZFCP_STATUS_COMMON_OPENING,
			&erp_action->adapter->status);
	retval = zfcp_erp_adapter_strategy_generic(erp_action, 0);
	atomic_clear_mask(ZFCP_STATUS_COMMON_OPENING,
			  &erp_action->adapter->status);

	return retval;
}

/*
 * function:    zfcp_register_adapter
 *
 * purpose:	allocate the irq associated with this devno and register
 *		the FSF adapter with the SCSI stack
 *
S
Swen Schillig 已提交
1591
 * returns:
L
Linus Torvalds 已提交
1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
 */
static int
zfcp_erp_adapter_strategy_generic(struct zfcp_erp_action *erp_action, int close)
{
	int retval = ZFCP_ERP_SUCCEEDED;

	if (close)
		goto close_only;

	retval = zfcp_erp_adapter_strategy_open_qdio(erp_action);
	if (retval != ZFCP_ERP_SUCCEEDED)
		goto failed_qdio;

	retval = zfcp_erp_adapter_strategy_open_fsf(erp_action);
	if (retval != ZFCP_ERP_SUCCEEDED)
		goto failed_openfcp;

	atomic_set_mask(ZFCP_STATUS_COMMON_OPEN, &erp_action->adapter->status);
	goto out;

 close_only:
	atomic_clear_mask(ZFCP_STATUS_COMMON_OPEN,
			  &erp_action->adapter->status);

 failed_openfcp:
1617
	zfcp_close_fsf(erp_action->adapter);
L
Linus Torvalds 已提交
1618
 failed_qdio:
1619 1620 1621 1622
	atomic_clear_mask(ZFCP_STATUS_ADAPTER_XCONFIG_OK |
			  ZFCP_STATUS_ADAPTER_LINK_UNPLUGGED |
			  ZFCP_STATUS_ADAPTER_XPORT_OK,
			  &erp_action->adapter->status);
L
Linus Torvalds 已提交
1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634
 out:
	return retval;
}

/*
 * function:    zfcp_qdio_init
 *
 * purpose:	setup QDIO operation for specified adapter
 *
 * returns:	0 - successful setup
 *		!0 - failed setup
 */
1635
static int
L
Linus Torvalds 已提交
1636 1637 1638 1639
zfcp_erp_adapter_strategy_open_qdio(struct zfcp_erp_action *erp_action)
{
	struct zfcp_adapter *adapter = erp_action->adapter;

S
Swen Schillig 已提交
1640 1641
	if (zfcp_qdio_open(adapter))
		return ZFCP_ERP_FAILED;
L
Linus Torvalds 已提交
1642 1643 1644 1645 1646 1647

	/* initialize waitqueue used to wait for free SBALs in requests queue */
	init_waitqueue_head(&adapter->request_wq);

	/* ok, we did it - skip all cleanups for different failures */
	atomic_set_mask(ZFCP_STATUS_ADAPTER_QDIOUP, &adapter->status);
S
Swen Schillig 已提交
1648
	return ZFCP_ERP_SUCCEEDED;
L
Linus Torvalds 已提交
1649 1650 1651 1652 1653 1654
}


static int
zfcp_erp_adapter_strategy_open_fsf(struct zfcp_erp_action *erp_action)
{
1655
	int retval;
1656

1657 1658 1659 1660 1661 1662
	retval = zfcp_erp_adapter_strategy_open_fsf_xconfig(erp_action);
	if (retval == ZFCP_ERP_FAILED)
		return ZFCP_ERP_FAILED;

	retval = zfcp_erp_adapter_strategy_open_fsf_xport(erp_action);
	if (retval == ZFCP_ERP_FAILED)
1663
		return ZFCP_ERP_FAILED;
L
Linus Torvalds 已提交
1664

1665
	return zfcp_erp_adapter_strategy_open_fsf_statusread(erp_action);
L
Linus Torvalds 已提交
1666 1667 1668 1669 1670 1671 1672
}

static int
zfcp_erp_adapter_strategy_open_fsf_xconfig(struct zfcp_erp_action *erp_action)
{
	int retval = ZFCP_ERP_SUCCEEDED;
	int retries;
1673
	int sleep = ZFCP_EXCHANGE_CONFIG_DATA_FIRST_SLEEP;
L
Linus Torvalds 已提交
1674 1675 1676 1677
	struct zfcp_adapter *adapter = erp_action->adapter;

	atomic_clear_mask(ZFCP_STATUS_ADAPTER_XCONFIG_OK, &adapter->status);

1678
	for (retries = ZFCP_EXCHANGE_CONFIG_DATA_RETRIES; retries; retries--) {
L
Linus Torvalds 已提交
1679 1680
		atomic_clear_mask(ZFCP_STATUS_ADAPTER_HOST_CON_INIT,
				  &adapter->status);
1681
		write_lock_irq(&adapter->erp_lock);
L
Linus Torvalds 已提交
1682
		zfcp_erp_action_to_running(erp_action);
1683
		write_unlock_irq(&adapter->erp_lock);
L
Linus Torvalds 已提交
1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
		if (zfcp_fsf_exchange_config_data(erp_action)) {
			retval = ZFCP_ERP_FAILED;
			break;
		}

		/*
		 * Why this works:
		 * Both the normal completion handler as well as the timeout
		 * handler will do an 'up' when the 'exchange config data'
		 * request completes or times out. Thus, the signal to go on
		 * won't be lost utilizing this semaphore.
		 * Furthermore, this 'adapter_reopen' action is
		 * guaranteed to be the only action being there (highest action
		 * which prevents other actions from being created).
		 * Resulting from that, the wake signal recognized here
		 * _must_ be the one belonging to the 'exchange config
		 * data' request.
		 */
1702
		zfcp_rec_dbf_event_thread_lock(6, adapter);
L
Linus Torvalds 已提交
1703
		down(&adapter->erp_ready_sem);
1704
		zfcp_rec_dbf_event_thread_lock(7, adapter);
C
Christof Schmitt 已提交
1705
		if (erp_action->status & ZFCP_STATUS_ERP_TIMEDOUT)
L
Linus Torvalds 已提交
1706
			break;
1707 1708 1709 1710 1711 1712

		if (!atomic_test_mask(ZFCP_STATUS_ADAPTER_HOST_CON_INIT,
				     &adapter->status))
			break;

		/* sleep a little bit before retry */
1713
		ssleep(sleep);
1714 1715
		sleep *= 2;
	}
L
Linus Torvalds 已提交
1716

1717 1718 1719
	atomic_clear_mask(ZFCP_STATUS_ADAPTER_HOST_CON_INIT,
			  &adapter->status);

L
Linus Torvalds 已提交
1720
	if (!atomic_test_mask(ZFCP_STATUS_ADAPTER_XCONFIG_OK,
C
Christof Schmitt 已提交
1721
			      &adapter->status))
L
Linus Torvalds 已提交
1722 1723 1724 1725 1726
		retval = ZFCP_ERP_FAILED;

	return retval;
}

1727 1728 1729
static int
zfcp_erp_adapter_strategy_open_fsf_xport(struct zfcp_erp_action *erp_action)
{
1730
	int ret;
1731
	struct zfcp_adapter *adapter;
1732

1733
	adapter = erp_action->adapter;
1734 1735
	atomic_clear_mask(ZFCP_STATUS_ADAPTER_XPORT_OK, &adapter->status);

1736
	write_lock_irq(&adapter->erp_lock);
1737
	zfcp_erp_action_to_running(erp_action);
1738
	write_unlock_irq(&adapter->erp_lock);
1739

1740
	ret = zfcp_fsf_exchange_port_data(erp_action);
1741 1742 1743 1744 1745 1746 1747
	if (ret == -EOPNOTSUPP) {
		return ZFCP_ERP_SUCCEEDED;
	} else if (ret) {
		return ZFCP_ERP_FAILED;
	}

	ret = ZFCP_ERP_SUCCEEDED;
1748
	zfcp_rec_dbf_event_thread_lock(8, adapter);
1749
	down(&adapter->erp_ready_sem);
1750
	zfcp_rec_dbf_event_thread_lock(9, adapter);
C
Christof Schmitt 已提交
1751
	if (erp_action->status & ZFCP_STATUS_ERP_TIMEDOUT)
1752
		ret = ZFCP_ERP_FAILED;
1753

1754
	return ret;
1755 1756
}

L
Linus Torvalds 已提交
1757 1758 1759 1760 1761 1762
static int
zfcp_erp_adapter_strategy_open_fsf_statusread(struct zfcp_erp_action
					      *erp_action)
{
	struct zfcp_adapter *adapter = erp_action->adapter;

1763 1764
	atomic_set(&adapter->stat_miss, 16);
	return zfcp_status_read_refill(adapter);
L
Linus Torvalds 已提交
1765 1766 1767
}

/*
S
Swen Schillig 已提交
1768
 * function:
L
Linus Torvalds 已提交
1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816
 *
 * purpose:	this routine executes the 'Reopen Physical Port' action
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_SUCCEEDED	- action finished successfully
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_port_forced_strategy(struct zfcp_erp_action *erp_action)
{
	int retval = ZFCP_ERP_FAILED;
	struct zfcp_port *port = erp_action->port;

	switch (erp_action->step) {

		/*
		 * FIXME:
		 * the ULP spec. begs for waiting for oustanding commands
		 */
	case ZFCP_ERP_STEP_UNINITIALIZED:
		zfcp_erp_port_strategy_clearstati(port);
		/*
		 * it would be sufficient to test only the normal open flag
		 * since the phys. open flag cannot be set if the normal
		 * open flag is unset - however, this is for readabilty ...
		 */
		if (atomic_test_mask((ZFCP_STATUS_PORT_PHYS_OPEN |
				      ZFCP_STATUS_COMMON_OPEN),
			             &port->status)) {
			retval =
			    zfcp_erp_port_forced_strategy_close(erp_action);
		} else
			retval = ZFCP_ERP_FAILED;
		break;

	case ZFCP_ERP_STEP_PHYS_PORT_CLOSING:
		if (atomic_test_mask(ZFCP_STATUS_PORT_PHYS_OPEN,
				     &port->status)) {
			retval = ZFCP_ERP_FAILED;
		} else
			retval = ZFCP_ERP_SUCCEEDED;
		break;
	}

	return retval;
}

/*
S
Swen Schillig 已提交
1817
 * function:
L
Linus Torvalds 已提交
1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886
 *
 * purpose:	this routine executes the 'Reopen Port' action
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_SUCCEEDED	- action finished successfully
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_port_strategy(struct zfcp_erp_action *erp_action)
{
	int retval = ZFCP_ERP_FAILED;
	struct zfcp_port *port = erp_action->port;

	switch (erp_action->step) {

		/*
		 * FIXME:
		 * the ULP spec. begs for waiting for oustanding commands
		 */
	case ZFCP_ERP_STEP_UNINITIALIZED:
		zfcp_erp_port_strategy_clearstati(port);
		if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &port->status)) {
			retval = zfcp_erp_port_strategy_close(erp_action);
			goto out;
		}		/* else it's already closed, open it */
		break;

	case ZFCP_ERP_STEP_PORT_CLOSING:
		if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &port->status)) {
			retval = ZFCP_ERP_FAILED;
			goto out;
		}		/* else it's closed now, open it */
		break;
	}
	if (erp_action->status & ZFCP_STATUS_ERP_CLOSE_ONLY)
		retval = ZFCP_ERP_EXIT;
	else
		retval = zfcp_erp_port_strategy_open(erp_action);

 out:
	return retval;
}

static int
zfcp_erp_port_strategy_open(struct zfcp_erp_action *erp_action)
{
	int retval;

	if (atomic_test_mask(ZFCP_STATUS_PORT_WKA,
			     &erp_action->port->status))
		retval = zfcp_erp_port_strategy_open_nameserver(erp_action);
	else
		retval = zfcp_erp_port_strategy_open_common(erp_action);

	return retval;
}

static int
zfcp_erp_port_strategy_open_common(struct zfcp_erp_action *erp_action)
{
	int retval = 0;
	struct zfcp_adapter *adapter = erp_action->adapter;
	struct zfcp_port *port = erp_action->port;

	switch (erp_action->step) {

	case ZFCP_ERP_STEP_UNINITIALIZED:
	case ZFCP_ERP_STEP_PHYS_PORT_CLOSING:
	case ZFCP_ERP_STEP_PORT_CLOSING:
1887
		if (fc_host_port_type(adapter->scsi_host) == FC_PORTTYPE_PTP) {
已提交
1888
			if (port->wwpn != adapter->peer_wwpn) {
C
Christof Schmitt 已提交
1889 1890 1891 1892
				dev_err(&adapter->ccw_device->dev,
					"Failed to open port 0x%016Lx, "
					"Peer WWPN 0x%016Lx does not match.\n",
					port->wwpn, adapter->peer_wwpn);
1893
				zfcp_erp_port_failed(port, 25, NULL);
已提交
1894 1895 1896 1897 1898 1899 1900 1901
				retval = ZFCP_ERP_FAILED;
				break;
			}
			port->d_id = adapter->peer_d_id;
			atomic_set_mask(ZFCP_STATUS_PORT_DID_DID, &port->status);
			retval = zfcp_erp_port_strategy_open_port(erp_action);
			break;
		}
L
Linus Torvalds 已提交
1902 1903 1904
		if (!(adapter->nameserver_port)) {
			retval = zfcp_nameserver_enqueue(adapter);
			if (retval != 0) {
C
Christof Schmitt 已提交
1905 1906
				dev_err(&adapter->ccw_device->dev,
					"Nameserver port unavailable.\n");
L
Linus Torvalds 已提交
1907 1908 1909 1910 1911 1912 1913 1914 1915
				retval = ZFCP_ERP_FAILED;
				break;
			}
		}
		if (!atomic_test_mask(ZFCP_STATUS_COMMON_UNBLOCKED,
				      &adapter->nameserver_port->status)) {
			/* nameserver port may live again */
			atomic_set_mask(ZFCP_STATUS_COMMON_RUNNING,
					&adapter->nameserver_port->status);
1916
			if (zfcp_erp_port_reopen(adapter->nameserver_port, 0,
1917
						 77, erp_action) >= 0) {
L
Linus Torvalds 已提交
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
				erp_action->step =
					ZFCP_ERP_STEP_NAMESERVER_OPEN;
				retval = ZFCP_ERP_CONTINUES;
			} else
				retval = ZFCP_ERP_FAILED;
			break;
		}
		/* else nameserver port is already open, fall through */
	case ZFCP_ERP_STEP_NAMESERVER_OPEN:
		if (!atomic_test_mask(ZFCP_STATUS_COMMON_OPEN,
C
Christof Schmitt 已提交
1928
				      &adapter->nameserver_port->status))
L
Linus Torvalds 已提交
1929
			retval = ZFCP_ERP_FAILED;
C
Christof Schmitt 已提交
1930
		else
L
Linus Torvalds 已提交
1931 1932 1933 1934 1935 1936 1937 1938
			retval = zfcp_erp_port_strategy_open_common_lookup
				(erp_action);
		break;

	case ZFCP_ERP_STEP_NAMESERVER_LOOKUP:
		if (!atomic_test_mask(ZFCP_STATUS_PORT_DID_DID, &port->status)) {
			if (atomic_test_mask
			    (ZFCP_STATUS_PORT_INVALID_WWPN, &port->status)) {
1939
				zfcp_erp_port_failed(port, 26, NULL);
L
Linus Torvalds 已提交
1940
				retval = ZFCP_ERP_EXIT;
C
Christof Schmitt 已提交
1941
			} else
L
Linus Torvalds 已提交
1942
				retval = ZFCP_ERP_FAILED;
C
Christof Schmitt 已提交
1943
		} else
L
Linus Torvalds 已提交
1944 1945 1946 1947 1948 1949 1950
			retval = zfcp_erp_port_strategy_open_port(erp_action);
		break;

	case ZFCP_ERP_STEP_PORT_OPENING:
		/* D_ID might have changed during open */
		if (atomic_test_mask((ZFCP_STATUS_COMMON_OPEN |
				      ZFCP_STATUS_PORT_DID_DID),
C
Christof Schmitt 已提交
1951
				     &port->status))
L
Linus Torvalds 已提交
1952
			retval = ZFCP_ERP_SUCCEEDED;
C
Christof Schmitt 已提交
1953
		else
L
Linus Torvalds 已提交
1954 1955 1956 1957
			retval = ZFCP_ERP_FAILED;
		break;

	default:
C
Christof Schmitt 已提交
1958
		/* unknown erp step */
L
Linus Torvalds 已提交
1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979
		retval = ZFCP_ERP_FAILED;
	}

	return retval;
}

static int
zfcp_erp_port_strategy_open_nameserver(struct zfcp_erp_action *erp_action)
{
	int retval;
	struct zfcp_port *port = erp_action->port;

	switch (erp_action->step) {

	case ZFCP_ERP_STEP_UNINITIALIZED:
	case ZFCP_ERP_STEP_PHYS_PORT_CLOSING:
	case ZFCP_ERP_STEP_PORT_CLOSING:
		retval = zfcp_erp_port_strategy_open_port(erp_action);
		break;

	case ZFCP_ERP_STEP_PORT_OPENING:
C
Christof Schmitt 已提交
1980
		if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &port->status))
L
Linus Torvalds 已提交
1981
			retval = ZFCP_ERP_SUCCEEDED;
C
Christof Schmitt 已提交
1982
		else
L
Linus Torvalds 已提交
1983 1984 1985 1986 1987 1988
			retval = ZFCP_ERP_FAILED;
		/* this is needed anyway (dont care for retval of wakeup) */
		zfcp_erp_port_strategy_open_nameserver_wakeup(erp_action);
		break;

	default:
C
Christof Schmitt 已提交
1989
		/* unknown erp step */
L
Linus Torvalds 已提交
1990 1991 1992 1993 1994 1995 1996
		retval = ZFCP_ERP_FAILED;
	}

	return retval;
}

/*
S
Swen Schillig 已提交
1997
 * function:
L
Linus Torvalds 已提交
1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
 *
 * purpose:	makes the erp thread continue with reopen (physical) port
 *		actions which have been paused until the name server port
 *		is opened (or failed)
 *
 * returns:	0	(a kind of void retval, its not used)
 */
static int
zfcp_erp_port_strategy_open_nameserver_wakeup(struct zfcp_erp_action
					      *ns_erp_action)
{
	int retval = 0;
	unsigned long flags;
	struct zfcp_adapter *adapter = ns_erp_action->adapter;
	struct zfcp_erp_action *erp_action, *tmp;

	read_lock_irqsave(&adapter->erp_lock, flags);
	list_for_each_entry_safe(erp_action, tmp, &adapter->erp_running_head,
				 list) {
		if (erp_action->step == ZFCP_ERP_STEP_NAMESERVER_OPEN) {
			if (atomic_test_mask(
				    ZFCP_STATUS_COMMON_ERP_FAILED,
				    &adapter->nameserver_port->status))
2021 2022
				zfcp_erp_port_failed(erp_action->port, 27,
						     NULL);
L
Linus Torvalds 已提交
2023 2024 2025 2026 2027 2028 2029 2030 2031
			zfcp_erp_action_ready(erp_action);
		}
	}
	read_unlock_irqrestore(&adapter->erp_lock, flags);

	return retval;
}

/*
S
Swen Schillig 已提交
2032
 * function:
L
Linus Torvalds 已提交
2033
 *
S
Swen Schillig 已提交
2034
 * purpose:
L
Linus Torvalds 已提交
2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_port_forced_strategy_close(struct zfcp_erp_action *erp_action)
{
	int retval;

	retval = zfcp_fsf_close_physical_port(erp_action);
	if (retval == -ENOMEM) {
		retval = ZFCP_ERP_NOMEM;
		goto out;
	}
	erp_action->step = ZFCP_ERP_STEP_PHYS_PORT_CLOSING;
	if (retval != 0) {
		/* could not send 'open', fail */
		retval = ZFCP_ERP_FAILED;
		goto out;
	}
	retval = ZFCP_ERP_CONTINUES;
 out:
	return retval;
}

static int
zfcp_erp_port_strategy_clearstati(struct zfcp_port *port)
{
	int retval = 0;

	atomic_clear_mask(ZFCP_STATUS_COMMON_OPENING |
			  ZFCP_STATUS_COMMON_CLOSING |
			  ZFCP_STATUS_COMMON_ACCESS_DENIED |
			  ZFCP_STATUS_PORT_DID_DID |
			  ZFCP_STATUS_PORT_PHYS_CLOSING |
			  ZFCP_STATUS_PORT_INVALID_WWPN,
			  &port->status);
	return retval;
}

/*
S
Swen Schillig 已提交
2076
 * function:
L
Linus Torvalds 已提交
2077
 *
S
Swen Schillig 已提交
2078
 * purpose:
L
Linus Torvalds 已提交
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_port_strategy_close(struct zfcp_erp_action *erp_action)
{
	int retval;

	retval = zfcp_fsf_close_port(erp_action);
	if (retval == -ENOMEM) {
		retval = ZFCP_ERP_NOMEM;
		goto out;
	}
	erp_action->step = ZFCP_ERP_STEP_PORT_CLOSING;
	if (retval != 0) {
		/* could not send 'close', fail */
		retval = ZFCP_ERP_FAILED;
		goto out;
	}
	retval = ZFCP_ERP_CONTINUES;
 out:
	return retval;
}

/*
S
Swen Schillig 已提交
2105
 * function:
L
Linus Torvalds 已提交
2106
 *
S
Swen Schillig 已提交
2107
 * purpose:
L
Linus Torvalds 已提交
2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_port_strategy_open_port(struct zfcp_erp_action *erp_action)
{
	int retval;

	retval = zfcp_fsf_open_port(erp_action);
	if (retval == -ENOMEM) {
		retval = ZFCP_ERP_NOMEM;
		goto out;
	}
	erp_action->step = ZFCP_ERP_STEP_PORT_OPENING;
	if (retval != 0) {
		/* could not send 'open', fail */
		retval = ZFCP_ERP_FAILED;
		goto out;
	}
	retval = ZFCP_ERP_CONTINUES;
 out:
	return retval;
}

/*
S
Swen Schillig 已提交
2134
 * function:
L
Linus Torvalds 已提交
2135
 *
S
Swen Schillig 已提交
2136
 * purpose:
L
Linus Torvalds 已提交
2137 2138 2139 2140 2141 2142 2143 2144 2145
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_port_strategy_open_common_lookup(struct zfcp_erp_action *erp_action)
{
	int retval;

2146
	retval = zfcp_fc_ns_gid_pn_request(erp_action);
L
Linus Torvalds 已提交
2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162
	if (retval == -ENOMEM) {
		retval = ZFCP_ERP_NOMEM;
		goto out;
	}
	erp_action->step = ZFCP_ERP_STEP_NAMESERVER_LOOKUP;
	if (retval != 0) {
		/* could not send nameserver request, fail */
		retval = ZFCP_ERP_FAILED;
		goto out;
	}
	retval = ZFCP_ERP_CONTINUES;
 out:
	return retval;
}

/*
S
Swen Schillig 已提交
2163
 * function:
L
Linus Torvalds 已提交
2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191
 *
 * purpose:	this routine executes the 'Reopen Unit' action
 *		currently no retries
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_SUCCEEDED	- action finished successfully
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_unit_strategy(struct zfcp_erp_action *erp_action)
{
	int retval = ZFCP_ERP_FAILED;
	struct zfcp_unit *unit = erp_action->unit;

	switch (erp_action->step) {

		/*
		 * FIXME:
		 * the ULP spec. begs for waiting for oustanding commands
		 */
	case ZFCP_ERP_STEP_UNINITIALIZED:
		zfcp_erp_unit_strategy_clearstati(unit);
		if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &unit->status)) {
			retval = zfcp_erp_unit_strategy_close(erp_action);
			break;
		}
		/* else it's already closed, fall through */
	case ZFCP_ERP_STEP_UNIT_CLOSING:
C
Christof Schmitt 已提交
2192
		if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &unit->status))
L
Linus Torvalds 已提交
2193
			retval = ZFCP_ERP_FAILED;
C
Christof Schmitt 已提交
2194
		else
L
Linus Torvalds 已提交
2195 2196
			if (erp_action->status & ZFCP_STATUS_ERP_CLOSE_ONLY)
				retval = ZFCP_ERP_EXIT;
C
Christof Schmitt 已提交
2197
			else
L
Linus Torvalds 已提交
2198 2199 2200 2201 2202
				retval =
				    zfcp_erp_unit_strategy_open(erp_action);
		break;

	case ZFCP_ERP_STEP_UNIT_OPENING:
C
Christof Schmitt 已提交
2203
		if (atomic_test_mask(ZFCP_STATUS_COMMON_OPEN, &unit->status))
L
Linus Torvalds 已提交
2204
			retval = ZFCP_ERP_SUCCEEDED;
C
Christof Schmitt 已提交
2205
		else
L
Linus Torvalds 已提交
2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228
			retval = ZFCP_ERP_FAILED;
		break;
	}

	return retval;
}

static int
zfcp_erp_unit_strategy_clearstati(struct zfcp_unit *unit)
{
	int retval = 0;

	atomic_clear_mask(ZFCP_STATUS_COMMON_OPENING |
			  ZFCP_STATUS_COMMON_CLOSING |
			  ZFCP_STATUS_COMMON_ACCESS_DENIED |
			  ZFCP_STATUS_UNIT_SHARED |
			  ZFCP_STATUS_UNIT_READONLY,
			  &unit->status);

	return retval;
}

/*
S
Swen Schillig 已提交
2229
 * function:
L
Linus Torvalds 已提交
2230
 *
S
Swen Schillig 已提交
2231
 * purpose:
L
Linus Torvalds 已提交
2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_unit_strategy_close(struct zfcp_erp_action *erp_action)
{
	int retval;

	retval = zfcp_fsf_close_unit(erp_action);
	if (retval == -ENOMEM) {
		retval = ZFCP_ERP_NOMEM;
		goto out;
	}
	erp_action->step = ZFCP_ERP_STEP_UNIT_CLOSING;
	if (retval != 0) {
		/* could not send 'close', fail */
		retval = ZFCP_ERP_FAILED;
		goto out;
	}
	retval = ZFCP_ERP_CONTINUES;

 out:
	return retval;
}

/*
S
Swen Schillig 已提交
2259
 * function:
L
Linus Torvalds 已提交
2260
 *
S
Swen Schillig 已提交
2261
 * purpose:
L
Linus Torvalds 已提交
2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
 *
 * returns:	ZFCP_ERP_CONTINUES	- action continues (asynchronously)
 *		ZFCP_ERP_FAILED		- action finished unsuccessfully
 */
static int
zfcp_erp_unit_strategy_open(struct zfcp_erp_action *erp_action)
{
	int retval;

	retval = zfcp_fsf_open_unit(erp_action);
	if (retval == -ENOMEM) {
		retval = ZFCP_ERP_NOMEM;
		goto out;
	}
	erp_action->step = ZFCP_ERP_STEP_UNIT_OPENING;
	if (retval != 0) {
		/* could not send 'open', fail */
		retval = ZFCP_ERP_FAILED;
		goto out;
	}
	retval = ZFCP_ERP_CONTINUES;
 out:
	return retval;
}

2287
void zfcp_erp_start_timer(struct zfcp_fsf_req *fsf_req)
L
Linus Torvalds 已提交
2288
{
2289 2290 2291 2292 2293
	BUG_ON(!fsf_req->erp_action);
	fsf_req->timer.function = zfcp_erp_timeout_handler;
	fsf_req->timer.data = (unsigned long) fsf_req->erp_action;
	fsf_req->timer.expires = jiffies + ZFCP_ERP_FSFREQ_TIMEOUT;
	add_timer(&fsf_req->timer);
L
Linus Torvalds 已提交
2294 2295 2296
}

/*
S
Swen Schillig 已提交
2297
 * function:
L
Linus Torvalds 已提交
2298 2299 2300 2301 2302
 *
 * purpose:	enqueue the specified error recovery action, if needed
 *
 * returns:
 */
2303 2304
static int zfcp_erp_action_enqueue(int want, struct zfcp_adapter *adapter,
				   struct zfcp_port *port,
2305
				   struct zfcp_unit *unit, u8 id, void *ref)
L
Linus Torvalds 已提交
2306
{
2307
	int retval = 1, need = want;
L
Linus Torvalds 已提交
2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327
	struct zfcp_erp_action *erp_action = NULL;
	u32 status = 0;

	/*
	 * We need some rules here which check whether we really need
	 * this action or whether we should just drop it.
	 * E.g. if there is a unfinished 'Reopen Port' request then we drop a
	 * 'Reopen Unit' request for an associated unit since we can't
	 * satisfy this request now. A 'Reopen Port' action will trigger
	 * 'Reopen Unit' actions when it completes.
	 * Thus, there are only actions in the queue which can immediately be
	 * executed. This makes the processing of the action queue more
	 * efficient.
	 */

	if (!atomic_test_mask(ZFCP_STATUS_ADAPTER_ERP_THREAD_UP,
			      &adapter->status))
		return -EIO;

	/* check whether we really need this */
2328
	switch (want) {
L
Linus Torvalds 已提交
2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340
	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		if (atomic_test_mask
		    (ZFCP_STATUS_COMMON_ERP_INUSE, &unit->status)) {
			goto out;
		}
		if (!atomic_test_mask
		    (ZFCP_STATUS_COMMON_RUNNING, &port->status) ||
		    atomic_test_mask
		    (ZFCP_STATUS_COMMON_ERP_FAILED, &port->status)) {
			goto out;
		}
		if (!atomic_test_mask
2341 2342
		    (ZFCP_STATUS_COMMON_UNBLOCKED, &port->status))
			need = ZFCP_ERP_ACTION_REOPEN_PORT;
L
Linus Torvalds 已提交
2343 2344 2345 2346 2347 2348 2349 2350 2351 2352
		/* fall through !!! */

	case ZFCP_ERP_ACTION_REOPEN_PORT:
		if (atomic_test_mask
		    (ZFCP_STATUS_COMMON_ERP_INUSE, &port->status)) {
			goto out;
		}
		/* fall through !!! */

	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
2353
		if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_INUSE,
C
Christof Schmitt 已提交
2354
				     &port->status))
L
Linus Torvalds 已提交
2355 2356 2357 2358 2359 2360 2361 2362
			goto out;
		if (!atomic_test_mask
		    (ZFCP_STATUS_COMMON_RUNNING, &adapter->status) ||
		    atomic_test_mask
		    (ZFCP_STATUS_COMMON_ERP_FAILED, &adapter->status)) {
			goto out;
		}
		if (!atomic_test_mask
2363 2364
		    (ZFCP_STATUS_COMMON_UNBLOCKED, &adapter->status))
			need = ZFCP_ERP_ACTION_REOPEN_ADAPTER;
L
Linus Torvalds 已提交
2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
		/* fall through !!! */

	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		if (atomic_test_mask
		    (ZFCP_STATUS_COMMON_ERP_INUSE, &adapter->status)) {
			goto out;
		}
		break;

	default:
C
Christof Schmitt 已提交
2375
		/* unknown erp action */
L
Linus Torvalds 已提交
2376 2377 2378 2379 2380 2381 2382
		goto out;
	}

	/* mark adapter to have some error recovery pending */
	atomic_set_mask(ZFCP_STATUS_ADAPTER_ERP_PENDING, &adapter->status);

	/* setup error recovery action */
2383
	switch (need) {
L
Linus Torvalds 已提交
2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419

	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		zfcp_unit_get(unit);
		atomic_set_mask(ZFCP_STATUS_COMMON_ERP_INUSE, &unit->status);
		erp_action = &unit->erp_action;
		if (!atomic_test_mask
		    (ZFCP_STATUS_COMMON_RUNNING, &unit->status))
			status = ZFCP_STATUS_ERP_CLOSE_ONLY;
		break;

	case ZFCP_ERP_ACTION_REOPEN_PORT:
	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
		zfcp_port_get(port);
		zfcp_erp_action_dismiss_port(port);
		atomic_set_mask(ZFCP_STATUS_COMMON_ERP_INUSE, &port->status);
		erp_action = &port->erp_action;
		if (!atomic_test_mask
		    (ZFCP_STATUS_COMMON_RUNNING, &port->status))
			status = ZFCP_STATUS_ERP_CLOSE_ONLY;
		break;

	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		zfcp_adapter_get(adapter);
		zfcp_erp_action_dismiss_adapter(adapter);
		atomic_set_mask(ZFCP_STATUS_COMMON_ERP_INUSE, &adapter->status);
		erp_action = &adapter->erp_action;
		if (!atomic_test_mask
		    (ZFCP_STATUS_COMMON_RUNNING, &adapter->status))
			status = ZFCP_STATUS_ERP_CLOSE_ONLY;
		break;
	}

	memset(erp_action, 0, sizeof (struct zfcp_erp_action));
	erp_action->adapter = adapter;
	erp_action->port = port;
	erp_action->unit = unit;
2420
	erp_action->action = need;
L
Linus Torvalds 已提交
2421 2422 2423 2424 2425
	erp_action->status = status;

	++adapter->erp_total_count;

	/* finally put it into 'ready' queue and kick erp thread */
2426
	list_add_tail(&erp_action->list, &adapter->erp_ready_head);
L
Linus Torvalds 已提交
2427
	up(&adapter->erp_ready_sem);
2428
	zfcp_rec_dbf_event_thread(1, adapter);
L
Linus Torvalds 已提交
2429 2430
	retval = 0;
 out:
2431
	zfcp_rec_dbf_event_trigger(id, ref, want, need, erp_action,
2432
				   adapter, port, unit);
L
Linus Torvalds 已提交
2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448
	return retval;
}

static int
zfcp_erp_action_dequeue(struct zfcp_erp_action *erp_action)
{
	int retval = 0;
	struct zfcp_adapter *adapter = erp_action->adapter;

	--adapter->erp_total_count;
	if (erp_action->status & ZFCP_STATUS_ERP_LOWMEM) {
		--adapter->erp_low_mem_count;
		erp_action->status &= ~ZFCP_STATUS_ERP_LOWMEM;
	}

	list_del(&erp_action->list);
2449 2450
	zfcp_rec_dbf_event_action(144, erp_action);

L
Linus Torvalds 已提交
2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474
	switch (erp_action->action) {
	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		atomic_clear_mask(ZFCP_STATUS_COMMON_ERP_INUSE,
				  &erp_action->unit->status);
		break;
	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
	case ZFCP_ERP_ACTION_REOPEN_PORT:
		atomic_clear_mask(ZFCP_STATUS_COMMON_ERP_INUSE,
				  &erp_action->port->status);
		break;
	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
		atomic_clear_mask(ZFCP_STATUS_COMMON_ERP_INUSE,
				  &erp_action->adapter->status);
		break;
	default:
		/* bug */
		break;
	}
	return retval;
}

/**
 * zfcp_erp_action_cleanup
 *
2475
 * Register unit with scsi stack if appropriate and fix reference counts.
L
Linus Torvalds 已提交
2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487
 * Note: Temporary units are not registered with scsi stack.
 */
static void
zfcp_erp_action_cleanup(int action, struct zfcp_adapter *adapter,
			struct zfcp_port *port, struct zfcp_unit *unit,
			int result)
{
	switch (action) {
	case ZFCP_ERP_ACTION_REOPEN_UNIT:
		if ((result == ZFCP_ERP_SUCCEEDED)
		    && (!atomic_test_mask(ZFCP_STATUS_UNIT_TEMPORARY,
					  &unit->status))
2488
		    && !unit->device
2489 2490 2491
		    && port->rport) {
			atomic_set_mask(ZFCP_STATUS_UNIT_REGISTERED,
					&unit->status);
2492 2493 2494
			if (atomic_test_mask(ZFCP_STATUS_UNIT_SCSI_WORK_PENDING,
					     &unit->status) == 0)
				zfcp_erp_schedule_work(unit);
2495
		}
L
Linus Torvalds 已提交
2496 2497 2498 2499
		zfcp_unit_put(unit);
		break;
	case ZFCP_ERP_ACTION_REOPEN_PORT_FORCED:
	case ZFCP_ERP_ACTION_REOPEN_PORT:
2500 2501 2502 2503 2504 2505
		if (atomic_test_mask(ZFCP_STATUS_PORT_NO_WWPN,
				     &port->status)) {
			zfcp_port_put(port);
			break;
		}

2506 2507 2508 2509 2510 2511 2512 2513 2514 2515
		if ((result == ZFCP_ERP_SUCCEEDED)
		    && !port->rport) {
			struct fc_rport_identifiers ids;
			ids.node_name = port->wwnn;
			ids.port_name = port->wwpn;
			ids.port_id = port->d_id;
			ids.roles = FC_RPORT_ROLE_FCP_TARGET;
			port->rport =
				fc_remote_port_add(adapter->scsi_host, 0, &ids);
			if (!port->rport)
C
Christof Schmitt 已提交
2516 2517 2518
				dev_err(&adapter->ccw_device->dev,
					"Failed registration of rport "
					"0x%016Lx.\n", port->wwpn);
2519
			else {
2520
				scsi_target_unblock(&port->rport->dev);
2521 2522 2523 2524
				port->rport->maxframe_size = port->maxframe_size;
				port->rport->supported_classes =
					port->supported_classes;
			}
2525
		}
2526 2527 2528 2529
		if ((result != ZFCP_ERP_SUCCEEDED) && port->rport) {
			fc_remote_port_delete(port->rport);
			port->rport = NULL;
		}
L
Linus Torvalds 已提交
2530 2531 2532
		zfcp_port_put(port);
		break;
	case ZFCP_ERP_ACTION_REOPEN_ADAPTER:
2533 2534 2535 2536 2537 2538 2539 2540 2541
		if (result != ZFCP_ERP_SUCCEEDED) {
			list_for_each_entry(port, &adapter->port_list_head, list)
				if (port->rport &&
				    !atomic_test_mask(ZFCP_STATUS_PORT_WKA,
						      &port->status)) {
					fc_remote_port_delete(port->rport);
					port->rport = NULL;
				}
		}
L
Linus Torvalds 已提交
2542 2543 2544 2545 2546 2547 2548 2549
		zfcp_adapter_put(adapter);
		break;
	default:
		break;
	}
}


2550
static void zfcp_erp_action_dismiss_adapter(struct zfcp_adapter *adapter)
L
Linus Torvalds 已提交
2551 2552 2553 2554 2555 2556 2557 2558 2559 2560
{
	struct zfcp_port *port;

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_INUSE, &adapter->status))
		zfcp_erp_action_dismiss(&adapter->erp_action);
	else
		list_for_each_entry(port, &adapter->port_list_head, list)
		    zfcp_erp_action_dismiss_port(port);
}

2561
static void zfcp_erp_action_dismiss_port(struct zfcp_port *port)
L
Linus Torvalds 已提交
2562 2563 2564 2565 2566 2567 2568 2569 2570 2571
{
	struct zfcp_unit *unit;

	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_INUSE, &port->status))
		zfcp_erp_action_dismiss(&port->erp_action);
	else
		list_for_each_entry(unit, &port->unit_list_head, list)
		    zfcp_erp_action_dismiss_unit(unit);
}

2572
static void zfcp_erp_action_dismiss_unit(struct zfcp_unit *unit)
L
Linus Torvalds 已提交
2573 2574 2575 2576 2577
{
	if (atomic_test_mask(ZFCP_STATUS_COMMON_ERP_INUSE, &unit->status))
		zfcp_erp_action_dismiss(&unit->erp_action);
}

2578
static void zfcp_erp_action_to_running(struct zfcp_erp_action *erp_action)
L
Linus Torvalds 已提交
2579 2580
{
	list_move(&erp_action->list, &erp_action->adapter->erp_running_head);
2581
	zfcp_rec_dbf_event_action(145, erp_action);
L
Linus Torvalds 已提交
2582 2583
}

2584
static void zfcp_erp_action_to_ready(struct zfcp_erp_action *erp_action)
L
Linus Torvalds 已提交
2585 2586
{
	list_move(&erp_action->list, &erp_action->adapter->erp_ready_head);
2587
	zfcp_rec_dbf_event_action(146, erp_action);
L
Linus Torvalds 已提交
2588 2589
}

2590
void zfcp_erp_port_boxed(struct zfcp_port *port, u8 id, void *ref)
2591 2592 2593 2594
{
	unsigned long flags;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
2595 2596
	zfcp_erp_modify_port_status(port, id, ref,
				    ZFCP_STATUS_COMMON_ACCESS_BOXED, ZFCP_SET);
2597
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);
2598
	zfcp_erp_port_reopen(port, ZFCP_STATUS_COMMON_ERP_FAILED, id, ref);
2599 2600
}

2601
void zfcp_erp_unit_boxed(struct zfcp_unit *unit, u8 id, void *ref)
2602
{
2603 2604
	zfcp_erp_modify_unit_status(unit, id, ref,
				    ZFCP_STATUS_COMMON_ACCESS_BOXED, ZFCP_SET);
2605
	zfcp_erp_unit_reopen(unit, ZFCP_STATUS_COMMON_ERP_FAILED, id, ref);
2606 2607
}

2608
void zfcp_erp_port_access_denied(struct zfcp_port *port, u8 id, void *ref)
L
Linus Torvalds 已提交
2609 2610 2611 2612
{
	unsigned long flags;

	read_lock_irqsave(&zfcp_data.config_lock, flags);
2613 2614 2615
	zfcp_erp_modify_port_status(port, id, ref,
				    ZFCP_STATUS_COMMON_ERP_FAILED |
				    ZFCP_STATUS_COMMON_ACCESS_DENIED, ZFCP_SET);
L
Linus Torvalds 已提交
2616 2617 2618
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);
}

2619
void zfcp_erp_unit_access_denied(struct zfcp_unit *unit, u8 id, void *ref)
L
Linus Torvalds 已提交
2620
{
2621 2622 2623
	zfcp_erp_modify_unit_status(unit, id, ref,
				    ZFCP_STATUS_COMMON_ERP_FAILED |
				    ZFCP_STATUS_COMMON_ACCESS_DENIED, ZFCP_SET);
L
Linus Torvalds 已提交
2624 2625
}

2626
void zfcp_erp_adapter_access_changed(struct zfcp_adapter *adapter, u8 id,
2627
				     void *ref)
L
Linus Torvalds 已提交
2628 2629 2630 2631
{
	struct zfcp_port *port;
	unsigned long flags;

2632 2633 2634
	if (adapter->connection_features & FSF_FEATURE_NPIV_MODE)
		return;

L
Linus Torvalds 已提交
2635
	read_lock_irqsave(&zfcp_data.config_lock, flags);
已提交
2636
	if (adapter->nameserver_port)
2637
		zfcp_erp_port_access_changed(adapter->nameserver_port, id, ref);
L
Linus Torvalds 已提交
2638 2639
	list_for_each_entry(port, &adapter->port_list_head, list)
		if (port != adapter->nameserver_port)
2640
			zfcp_erp_port_access_changed(port, id, ref);
L
Linus Torvalds 已提交
2641 2642 2643
	read_unlock_irqrestore(&zfcp_data.config_lock, flags);
}

2644
void zfcp_erp_port_access_changed(struct zfcp_port *port, u8 id, void *ref)
L
Linus Torvalds 已提交
2645 2646 2647 2648
{
	struct zfcp_unit *unit;

	if (!atomic_test_mask(ZFCP_STATUS_COMMON_ACCESS_DENIED,
2649 2650
			      &port->status) &&
	    !atomic_test_mask(ZFCP_STATUS_COMMON_ACCESS_BOXED,
L
Linus Torvalds 已提交
2651 2652 2653
			      &port->status)) {
		if (!atomic_test_mask(ZFCP_STATUS_PORT_WKA, &port->status))
			list_for_each_entry(unit, &port->unit_list_head, list)
2654
				zfcp_erp_unit_access_changed(unit, id, ref);
L
Linus Torvalds 已提交
2655 2656 2657
		return;
	}

C
Christof Schmitt 已提交
2658
	zfcp_erp_port_reopen(port, ZFCP_STATUS_COMMON_ERP_FAILED, id, ref);
L
Linus Torvalds 已提交
2659 2660
}

2661
void zfcp_erp_unit_access_changed(struct zfcp_unit *unit, u8 id, void *ref)
L
Linus Torvalds 已提交
2662
{
2663 2664 2665 2666
	if (!atomic_test_mask(ZFCP_STATUS_COMMON_ACCESS_DENIED,
			      &unit->status) &&
	    !atomic_test_mask(ZFCP_STATUS_COMMON_ACCESS_BOXED,
			      &unit->status))
L
Linus Torvalds 已提交
2667 2668
		return;

C
Christof Schmitt 已提交
2669
	zfcp_erp_unit_reopen(unit, ZFCP_STATUS_COMMON_ERP_FAILED, id, ref);
L
Linus Torvalds 已提交
2670
}