rf_driver.c revision 1.97 1 /* $NetBSD: rf_driver.c,v 1.97 2004/03/20 04:22:05 oster Exp $ */
2 /*-
3 * Copyright (c) 1999 The NetBSD Foundation, Inc.
4 * All rights reserved.
5 *
6 * This code is derived from software contributed to The NetBSD Foundation
7 * by Greg Oster
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the NetBSD
20 * Foundation, Inc. and its contributors.
21 * 4. Neither the name of The NetBSD Foundation nor the names of its
22 * contributors may be used to endorse or promote products derived
23 * from this software without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
26 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
29 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38 /*
39 * Copyright (c) 1995 Carnegie-Mellon University.
40 * All rights reserved.
41 *
42 * Author: Mark Holland, Khalil Amiri, Claudson Bornstein, William V. Courtright II,
43 * Robby Findler, Daniel Stodolsky, Rachad Youssef, Jim Zelenka
44 *
45 * Permission to use, copy, modify and distribute this software and
46 * its documentation is hereby granted, provided that both the copyright
47 * notice and this permission notice appear in all copies of the
48 * software, derivative works or modified versions, and any portions
49 * thereof, and that both notices appear in supporting documentation.
50 *
51 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
52 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
53 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
54 *
55 * Carnegie Mellon requests users of this software to return to
56 *
57 * Software Distribution Coordinator or Software.Distribution (at) CS.CMU.EDU
58 * School of Computer Science
59 * Carnegie Mellon University
60 * Pittsburgh PA 15213-3890
61 *
62 * any improvements or extensions that they make and grant Carnegie the
63 * rights to redistribute these changes.
64 */
65
66 /******************************************************************************
67 *
68 * rf_driver.c -- main setup, teardown, and access routines for the RAID driver
69 *
70 * all routines are prefixed with rf_ (raidframe), to avoid conficts.
71 *
72 ******************************************************************************/
73
74
75 #include <sys/cdefs.h>
76 __KERNEL_RCSID(0, "$NetBSD: rf_driver.c,v 1.97 2004/03/20 04:22:05 oster Exp $");
77
78 #include "opt_raid_diagnostic.h"
79
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/ioctl.h>
83 #include <sys/fcntl.h>
84 #include <sys/vnode.h>
85
86
87 #include "rf_archs.h"
88 #include "rf_threadstuff.h"
89
90 #include <sys/errno.h>
91
92 #include "rf_raid.h"
93 #include "rf_dag.h"
94 #include "rf_aselect.h"
95 #include "rf_diskqueue.h"
96 #include "rf_parityscan.h"
97 #include "rf_alloclist.h"
98 #include "rf_dagutils.h"
99 #include "rf_utils.h"
100 #include "rf_etimer.h"
101 #include "rf_acctrace.h"
102 #include "rf_general.h"
103 #include "rf_desc.h"
104 #include "rf_states.h"
105 #include "rf_decluster.h"
106 #include "rf_map.h"
107 #include "rf_revent.h"
108 #include "rf_callback.h"
109 #include "rf_engine.h"
110 #include "rf_mcpair.h"
111 #include "rf_nwayxor.h"
112 #include "rf_copyback.h"
113 #include "rf_driver.h"
114 #include "rf_options.h"
115 #include "rf_shutdown.h"
116 #include "rf_kintf.h"
117
118 #include <sys/buf.h>
119
120 #ifndef RF_ACCESS_DEBUG
121 #define RF_ACCESS_DEBUG 0
122 #endif
123
124 /* rad == RF_RaidAccessDesc_t */
125 RF_DECLARE_MUTEX(rf_rad_lock)
126 #define RF_MAX_FREE_RAD 128
127 #define RF_MIN_FREE_RAD 32
128
129 /* debug variables */
130 char rf_panicbuf[2048]; /* a buffer to hold an error msg when we panic */
131
132 /* main configuration routines */
133 static int raidframe_booted = 0;
134
135 static void rf_ConfigureDebug(RF_Config_t * cfgPtr);
136 static void set_debug_option(char *name, long val);
137 static void rf_UnconfigureArray(void);
138 static void rf_ShutdownRDFreeList(void *);
139 static int rf_ConfigureRDFreeList(RF_ShutdownList_t **);
140
141 RF_DECLARE_MUTEX(rf_printf_mutex) /* debug only: avoids interleaved
142 * printfs by different stripes */
143
144 #define SIGNAL_QUIESCENT_COND(_raid_) wakeup(&((_raid_)->accesses_suspended))
145 #define WAIT_FOR_QUIESCENCE(_raid_) \
146 ltsleep(&((_raid_)->accesses_suspended), PRIBIO, \
147 "raidframe quiesce", 0, &((_raid_)->access_suspend_mutex))
148
149 static int configureCount = 0; /* number of active configurations */
150 static int isconfigged = 0; /* is basic raidframe (non per-array)
151 * stuff configged */
152 RF_DECLARE_LKMGR_STATIC_MUTEX(configureMutex) /* used to lock the configuration
153 * stuff */
154 static RF_ShutdownList_t *globalShutdown; /* non array-specific
155 * stuff */
156
157 static int rf_ConfigureRDFreeList(RF_ShutdownList_t ** listp);
158
159 /* called at system boot time */
160 int
161 rf_BootRaidframe()
162 {
163
164 if (raidframe_booted)
165 return (EBUSY);
166 raidframe_booted = 1;
167 lockinit(&configureMutex, PRIBIO, "RAIDframe lock", 0, 0);
168 configureCount = 0;
169 isconfigged = 0;
170 globalShutdown = NULL;
171 return (0);
172 }
173
174 /*
175 * Called whenever an array is shutdown
176 */
177 static void
178 rf_UnconfigureArray()
179 {
180
181 RF_LOCK_LKMGR_MUTEX(configureMutex);
182 if (--configureCount == 0) { /* if no active configurations, shut
183 * everything down */
184 isconfigged = 0;
185 rf_ShutdownList(&globalShutdown);
186
187 /*
188 * We must wait until now, because the AllocList module
189 * uses the DebugMem module.
190 */
191 #if RF_DEBUG_MEM
192 if (rf_memDebug)
193 rf_print_unfreed();
194 #endif
195 }
196 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
197 }
198
199 /*
200 * Called to shut down an array.
201 */
202 int
203 rf_Shutdown(RF_Raid_t *raidPtr)
204 {
205 if (!raidPtr->valid) {
206 RF_ERRORMSG("Attempt to shut down unconfigured RAIDframe driver. Aborting shutdown\n");
207 return (EINVAL);
208 }
209 /*
210 * wait for outstanding IOs to land
211 * As described in rf_raid.h, we use the rad_freelist lock
212 * to protect the per-array info about outstanding descs
213 * since we need to do freelist locking anyway, and this
214 * cuts down on the amount of serialization we've got going
215 * on.
216 */
217 RF_LOCK_MUTEX(rf_rad_lock);
218 if (raidPtr->waitShutdown) {
219 RF_UNLOCK_MUTEX(rf_rad_lock);
220 return (EBUSY);
221 }
222 raidPtr->waitShutdown = 1;
223 while (raidPtr->nAccOutstanding) {
224 RF_WAIT_COND(raidPtr->outstandingCond, rf_rad_lock);
225 }
226 RF_UNLOCK_MUTEX(rf_rad_lock);
227
228 /* Wait for any parity re-writes to stop... */
229 while (raidPtr->parity_rewrite_in_progress) {
230 printf("Waiting for parity re-write to exit...\n");
231 tsleep(&raidPtr->parity_rewrite_in_progress, PRIBIO,
232 "rfprwshutdown", 0);
233 }
234
235 raidPtr->valid = 0;
236
237 rf_update_component_labels(raidPtr, RF_FINAL_COMPONENT_UPDATE);
238
239 rf_UnconfigureVnodes(raidPtr);
240
241 rf_ShutdownList(&raidPtr->shutdownList);
242
243 rf_UnconfigureArray();
244
245 return (0);
246 }
247
248
249 #define DO_INIT_CONFIGURE(f) { \
250 rc = f (&globalShutdown); \
251 if (rc) { \
252 RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
253 rf_ShutdownList(&globalShutdown); \
254 configureCount--; \
255 RF_UNLOCK_LKMGR_MUTEX(configureMutex); \
256 return(rc); \
257 } \
258 }
259
260 #define DO_RAID_FAIL() { \
261 rf_UnconfigureVnodes(raidPtr); \
262 rf_ShutdownList(&raidPtr->shutdownList); \
263 rf_UnconfigureArray(); \
264 }
265
266 #define DO_RAID_INIT_CONFIGURE(f) { \
267 rc = f (&raidPtr->shutdownList, raidPtr, cfgPtr); \
268 if (rc) { \
269 RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
270 DO_RAID_FAIL(); \
271 return(rc); \
272 } \
273 }
274
275 #define DO_RAID_MUTEX(_m_) { \
276 rf_mutex_init((_m_)); \
277 }
278
279 int
280 rf_Configure(RF_Raid_t *raidPtr, RF_Config_t *cfgPtr, RF_AutoConfig_t *ac)
281 {
282 RF_RowCol_t col;
283 RF_IOBufHeader_t *tmpbuf;
284 int rc, i;
285
286 RF_LOCK_LKMGR_MUTEX(configureMutex);
287 configureCount++;
288 if (isconfigged == 0) {
289 rf_mutex_init(&rf_printf_mutex);
290
291 /* initialize globals */
292
293 DO_INIT_CONFIGURE(rf_ConfigureAllocList);
294
295 /*
296 * Yes, this does make debugging general to the whole
297 * system instead of being array specific. Bummer, drag.
298 */
299 rf_ConfigureDebug(cfgPtr);
300 DO_INIT_CONFIGURE(rf_ConfigureDebugMem);
301 #if RF_ACC_TRACE > 0
302 DO_INIT_CONFIGURE(rf_ConfigureAccessTrace);
303 #endif
304 DO_INIT_CONFIGURE(rf_ConfigureMapModule);
305 DO_INIT_CONFIGURE(rf_ConfigureReconEvent);
306 DO_INIT_CONFIGURE(rf_ConfigureCallback);
307 DO_INIT_CONFIGURE(rf_ConfigureRDFreeList);
308 DO_INIT_CONFIGURE(rf_ConfigureNWayXor);
309 DO_INIT_CONFIGURE(rf_ConfigureStripeLockFreeList);
310 DO_INIT_CONFIGURE(rf_ConfigureMCPair);
311 DO_INIT_CONFIGURE(rf_ConfigureDAGs);
312 DO_INIT_CONFIGURE(rf_ConfigureDAGFuncs);
313 DO_INIT_CONFIGURE(rf_ConfigureReconstruction);
314 DO_INIT_CONFIGURE(rf_ConfigureCopyback);
315 DO_INIT_CONFIGURE(rf_ConfigureDiskQueueSystem);
316 isconfigged = 1;
317 }
318 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
319
320 DO_RAID_MUTEX(&raidPtr->mutex);
321 /* set up the cleanup list. Do this after ConfigureDebug so that
322 * value of memDebug will be set */
323
324 rf_MakeAllocList(raidPtr->cleanupList);
325 if (raidPtr->cleanupList == NULL) {
326 DO_RAID_FAIL();
327 return (ENOMEM);
328 }
329 rf_ShutdownCreate(&raidPtr->shutdownList,
330 (void (*) (void *)) rf_FreeAllocList,
331 raidPtr->cleanupList);
332
333 raidPtr->numCol = cfgPtr->numCol;
334 raidPtr->numSpare = cfgPtr->numSpare;
335
336 raidPtr->status = rf_rs_optimal;
337 raidPtr->reconControl = NULL;
338
339 TAILQ_INIT(&(raidPtr->iodone));
340 simple_lock_init(&(raidPtr->iodone_lock));
341
342 DO_RAID_INIT_CONFIGURE(rf_ConfigureEngine);
343 DO_RAID_INIT_CONFIGURE(rf_ConfigureStripeLocks);
344
345 raidPtr->outstandingCond = 0;
346
347 raidPtr->nAccOutstanding = 0;
348 raidPtr->waitShutdown = 0;
349
350 DO_RAID_MUTEX(&raidPtr->access_suspend_mutex);
351
352 raidPtr->waitForReconCond = 0;
353
354 if (ac!=NULL) {
355 /* We have an AutoConfig structure.. Don't do the
356 normal disk configuration... call the auto config
357 stuff */
358 rf_AutoConfigureDisks(raidPtr, cfgPtr, ac);
359 } else {
360 DO_RAID_INIT_CONFIGURE(rf_ConfigureDisks);
361 DO_RAID_INIT_CONFIGURE(rf_ConfigureSpareDisks);
362 }
363 /* do this after ConfigureDisks & ConfigureSpareDisks to be sure dev
364 * no. is set */
365 DO_RAID_INIT_CONFIGURE(rf_ConfigureDiskQueues);
366
367 DO_RAID_INIT_CONFIGURE(rf_ConfigureLayout);
368
369 DO_RAID_INIT_CONFIGURE(rf_ConfigurePSStatus);
370
371 #if RF_INCLUDE_CHAINDECLUSTER > 0
372 for (col = 0; col < raidPtr->numCol; col++) {
373 /*
374 * XXX better distribution
375 */
376 raidPtr->hist_diskreq[col] = 0;
377 }
378 #endif
379 raidPtr->numNewFailures = 0;
380 raidPtr->copyback_in_progress = 0;
381 raidPtr->parity_rewrite_in_progress = 0;
382 raidPtr->adding_hot_spare = 0;
383 raidPtr->recon_in_progress = 0;
384 raidPtr->maxOutstanding = cfgPtr->maxOutstandingDiskReqs;
385
386 /* autoconfigure and root_partition will actually get filled in
387 after the config is done */
388 raidPtr->autoconfigure = 0;
389 raidPtr->root_partition = 0;
390 raidPtr->last_unit = raidPtr->raidid;
391 raidPtr->config_order = 0;
392
393 if (rf_keepAccTotals) {
394 raidPtr->keep_acc_totals = 1;
395 }
396
397 /* Allocate a bunch of buffers to be used in low-memory conditions */
398 raidPtr->iobuf = NULL;
399 /* XXX next line needs tuning... */
400 raidPtr->numEmergencyBuffers = 10 * raidPtr->numCol;
401 #if DEBUG
402 printf("raid%d: allocating %d buffers of %d bytes.\n",
403 raidPtr->raidid,
404 raidPtr->numEmergencyBuffers,
405 (int)(raidPtr->Layout.sectorsPerStripeUnit <<
406 raidPtr->logBytesPerSector));
407 #endif
408 for (i = 0; i < raidPtr->numEmergencyBuffers; i++) {
409 tmpbuf = malloc( raidPtr->Layout.sectorsPerStripeUnit <<
410 raidPtr->logBytesPerSector,
411 M_RAIDFRAME, M_NOWAIT);
412 if (tmpbuf) {
413 tmpbuf->next = raidPtr->iobuf;
414 raidPtr->iobuf = tmpbuf;
415 raidPtr->iobuf_count++;
416 } else {
417 printf("raid%d: failed to allocate emergency buffer!\n",
418 raidPtr->raidid);
419 }
420 }
421
422 raidPtr->valid = 1;
423
424 printf("raid%d: %s\n", raidPtr->raidid,
425 raidPtr->Layout.map->configName);
426 printf("raid%d: Components:", raidPtr->raidid);
427
428 for (col = 0; col < raidPtr->numCol; col++) {
429 printf(" %s", raidPtr->Disks[col].devname);
430 if (RF_DEAD_DISK(raidPtr->Disks[col].status)) {
431 printf("[**FAILED**]");
432 }
433 }
434 printf("\n");
435 printf("raid%d: Total Sectors: %lu (%lu MB)\n",
436 raidPtr->raidid,
437 (unsigned long) raidPtr->totalSectors,
438 (unsigned long) (raidPtr->totalSectors / 1024 *
439 (1 << raidPtr->logBytesPerSector) / 1024));
440
441 return (0);
442 }
443
444 static void
445 rf_ShutdownRDFreeList(void *ignored)
446 {
447 pool_destroy(&rf_pools.rad);
448 }
449
450 static int
451 rf_ConfigureRDFreeList(RF_ShutdownList_t **listp)
452 {
453
454 rf_pool_init(&rf_pools.rad, sizeof(RF_RaidAccessDesc_t),
455 "rf_rad_pl", RF_MIN_FREE_RAD, RF_MAX_FREE_RAD);
456 rf_ShutdownCreate(listp, rf_ShutdownRDFreeList, NULL);
457 simple_lock_init(&rf_rad_lock);
458 return (0);
459 }
460
461 RF_RaidAccessDesc_t *
462 rf_AllocRaidAccDesc(RF_Raid_t *raidPtr, RF_IoType_t type,
463 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
464 caddr_t bufPtr, void *bp, RF_RaidAccessFlags_t flags,
465 RF_AccessState_t *states)
466 {
467 RF_RaidAccessDesc_t *desc;
468
469 desc = pool_get(&rf_pools.rad, PR_WAITOK);
470
471 RF_LOCK_MUTEX(rf_rad_lock);
472 if (raidPtr->waitShutdown) {
473 /*
474 * Actually, we're shutting the array down. Free the desc
475 * and return NULL.
476 */
477
478 RF_UNLOCK_MUTEX(rf_rad_lock);
479 pool_put(&rf_pools.rad, desc);
480 return (NULL);
481 }
482 raidPtr->nAccOutstanding++;
483
484 RF_UNLOCK_MUTEX(rf_rad_lock);
485
486 desc->raidPtr = (void *) raidPtr;
487 desc->type = type;
488 desc->raidAddress = raidAddress;
489 desc->numBlocks = numBlocks;
490 desc->bufPtr = bufPtr;
491 desc->bp = bp;
492 desc->flags = flags;
493 desc->states = states;
494 desc->state = 0;
495
496 desc->status = 0;
497 #if RF_ACC_TRACE > 0
498 memset((char *) &desc->tracerec, 0, sizeof(RF_AccTraceEntry_t));
499 #endif
500 desc->callbackFunc = NULL;
501 desc->callbackArg = NULL;
502 desc->next = NULL;
503 desc->cleanupList = NULL;
504 rf_MakeAllocList(desc->cleanupList);
505 return (desc);
506 }
507
508 void
509 rf_FreeRaidAccDesc(RF_RaidAccessDesc_t *desc)
510 {
511 RF_Raid_t *raidPtr = desc->raidPtr;
512 RF_DagList_t *dagList, *temp;
513
514 RF_ASSERT(desc);
515
516 /* Cleanup the dagList(s) */
517 dagList = desc->dagList;
518 while(dagList != NULL) {
519 temp = dagList;
520 dagList = dagList->next;
521 rf_FreeDAGList(temp);
522 }
523
524 rf_FreeAllocList(desc->cleanupList);
525 pool_put(&rf_pools.rad, desc);
526 RF_LOCK_MUTEX(rf_rad_lock);
527 raidPtr->nAccOutstanding--;
528 if (raidPtr->waitShutdown) {
529 RF_SIGNAL_COND(raidPtr->outstandingCond);
530 }
531 RF_UNLOCK_MUTEX(rf_rad_lock);
532 }
533 /*********************************************************************
534 * Main routine for performing an access.
535 * Accesses are retried until a DAG can not be selected. This occurs
536 * when either the DAG library is incomplete or there are too many
537 * failures in a parity group.
538 *
539 * type should be read or write async_flag should be RF_TRUE or
540 * RF_FALSE bp_in is a buf pointer. void * to facilitate ignoring it
541 * outside the kernel
542 ********************************************************************/
543 int
544 rf_DoAccess(RF_Raid_t * raidPtr, RF_IoType_t type, int async_flag,
545 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
546 caddr_t bufPtr, void *bp_in, RF_RaidAccessFlags_t flags)
547 {
548 RF_RaidAccessDesc_t *desc;
549 caddr_t lbufPtr = bufPtr;
550 struct buf *bp = (struct buf *) bp_in;
551
552 raidAddress += rf_raidSectorOffset;
553
554 #if RF_ACCESS_DEBUG
555 if (rf_accessDebug) {
556
557 printf("logBytes is: %d %d %d\n", raidPtr->raidid,
558 raidPtr->logBytesPerSector,
559 (int) rf_RaidAddressToByte(raidPtr, numBlocks));
560 printf("raid%d: %s raidAddr %d (stripeid %d-%d) numBlocks %d (%d bytes) buf 0x%lx\n", raidPtr->raidid,
561 (type == RF_IO_TYPE_READ) ? "READ" : "WRITE", (int) raidAddress,
562 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress),
563 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress + numBlocks - 1),
564 (int) numBlocks,
565 (int) rf_RaidAddressToByte(raidPtr, numBlocks),
566 (long) bufPtr);
567 }
568 #endif
569 if (raidAddress + numBlocks > raidPtr->totalSectors) {
570
571 printf("DoAccess: raid addr %lu too large to access %lu sectors. Max legal addr is %lu\n",
572 (u_long) raidAddress, (u_long) numBlocks, (u_long) raidPtr->totalSectors);
573
574
575 bp->b_flags |= B_ERROR;
576 bp->b_resid = bp->b_bcount;
577 bp->b_error = ENOSPC;
578 biodone(bp);
579 return (ENOSPC);
580 }
581 desc = rf_AllocRaidAccDesc(raidPtr, type, raidAddress,
582 numBlocks, lbufPtr, bp, flags, raidPtr->Layout.map->states);
583
584 if (desc == NULL) {
585 return (ENOMEM);
586 }
587 #if RF_ACC_TRACE > 0
588 RF_ETIMER_START(desc->tracerec.tot_timer);
589 #endif
590 desc->async_flag = async_flag;
591
592 rf_ContinueRaidAccess(desc);
593
594 return (0);
595 }
596 #if 0
597 /* force the array into reconfigured mode without doing reconstruction */
598 int
599 rf_SetReconfiguredMode(RF_Raid_t *raidPtr, int col)
600 {
601 if (!(raidPtr->Layout.map->flags & RF_DISTRIBUTE_SPARE)) {
602 printf("Can't set reconfigured mode in dedicated-spare array\n");
603 RF_PANIC();
604 }
605 RF_LOCK_MUTEX(raidPtr->mutex);
606 raidPtr->numFailures++;
607 raidPtr->Disks[col].status = rf_ds_dist_spared;
608 raidPtr->status = rf_rs_reconfigured;
609 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
610 /* install spare table only if declustering + distributed sparing
611 * architecture. */
612 if (raidPtr->Layout.map->flags & RF_BD_DECLUSTERED)
613 rf_InstallSpareTable(raidPtr, col);
614 RF_UNLOCK_MUTEX(raidPtr->mutex);
615 return (0);
616 }
617 #endif
618
619 int
620 rf_FailDisk(RF_Raid_t *raidPtr, int fcol, int initRecon)
621 {
622 RF_LOCK_MUTEX(raidPtr->mutex);
623 if (raidPtr->Disks[fcol].status != rf_ds_failed) {
624 /* must be failing something that is valid, or else it's
625 already marked as failed (in which case we don't
626 want to mark it failed again!) */
627 raidPtr->numFailures++;
628 raidPtr->Disks[fcol].status = rf_ds_failed;
629 raidPtr->status = rf_rs_degraded;
630 }
631 RF_UNLOCK_MUTEX(raidPtr->mutex);
632
633 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
634
635 /* Close the component, so that it's not "locked" if someone
636 else want's to use it! */
637
638 rf_close_component(raidPtr, raidPtr->raid_cinfo[fcol].ci_vp,
639 raidPtr->Disks[fcol].auto_configured);
640
641 RF_LOCK_MUTEX(raidPtr->mutex);
642 raidPtr->raid_cinfo[fcol].ci_vp = NULL;
643
644 /* Need to mark the component as not being auto_configured
645 (in case it was previously). */
646
647 raidPtr->Disks[fcol].auto_configured = 0;
648 RF_UNLOCK_MUTEX(raidPtr->mutex);
649
650 if (initRecon)
651 rf_ReconstructFailedDisk(raidPtr, fcol);
652 return (0);
653 }
654 /* releases a thread that is waiting for the array to become quiesced.
655 * access_suspend_mutex should be locked upon calling this
656 */
657 void
658 rf_SignalQuiescenceLock(RF_Raid_t *raidPtr)
659 {
660 #if RF_DEBUG_QUIESCE
661 if (rf_quiesceDebug) {
662 printf("raid%d: Signalling quiescence lock\n",
663 raidPtr->raidid);
664 }
665 #endif
666 raidPtr->access_suspend_release = 1;
667
668 if (raidPtr->waiting_for_quiescence) {
669 SIGNAL_QUIESCENT_COND(raidPtr);
670 }
671 }
672 /* suspends all new requests to the array. No effect on accesses that are in flight. */
673 int
674 rf_SuspendNewRequestsAndWait(RF_Raid_t *raidPtr)
675 {
676 #if RF_DEBUG_QUIESCE
677 if (rf_quiesceDebug)
678 printf("raid%d: Suspending new reqs\n", raidPtr->raidid);
679 #endif
680 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
681 raidPtr->accesses_suspended++;
682 raidPtr->waiting_for_quiescence = (raidPtr->accs_in_flight == 0) ? 0 : 1;
683
684 if (raidPtr->waiting_for_quiescence) {
685 raidPtr->access_suspend_release = 0;
686 while (!raidPtr->access_suspend_release) {
687 #if RF_DEBUG_QUIESCE
688 printf("raid%d: Suspending: Waiting for Quiescence\n",
689 raidPtr->raidid);
690 #endif
691 WAIT_FOR_QUIESCENCE(raidPtr);
692 raidPtr->waiting_for_quiescence = 0;
693 }
694 }
695 #if RF_DEBUG_QUIESCE
696 printf("raid%d: Quiescence reached..\n", raidPtr->raidid);
697 #endif
698
699 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
700 return (raidPtr->waiting_for_quiescence);
701 }
702 /* wake up everyone waiting for quiescence to be released */
703 void
704 rf_ResumeNewRequests(RF_Raid_t *raidPtr)
705 {
706 RF_CallbackDesc_t *t, *cb;
707
708 #if RF_DEBUG_QUIESCE
709 if (rf_quiesceDebug)
710 printf("Resuming new reqs\n");
711 #endif
712
713 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
714 raidPtr->accesses_suspended--;
715 if (raidPtr->accesses_suspended == 0)
716 cb = raidPtr->quiesce_wait_list;
717 else
718 cb = NULL;
719 raidPtr->quiesce_wait_list = NULL;
720 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
721
722 while (cb) {
723 t = cb;
724 cb = cb->next;
725 (t->callbackFunc) (t->callbackArg);
726 rf_FreeCallbackDesc(t);
727 }
728 }
729 /*****************************************************************************************
730 *
731 * debug routines
732 *
733 ****************************************************************************************/
734
735 static void
736 set_debug_option(char *name, long val)
737 {
738 RF_DebugName_t *p;
739
740 for (p = rf_debugNames; p->name; p++) {
741 if (!strcmp(p->name, name)) {
742 *(p->ptr) = val;
743 printf("[Set debug variable %s to %ld]\n", name, val);
744 return;
745 }
746 }
747 RF_ERRORMSG1("Unknown debug string \"%s\"\n", name);
748 }
749
750
751 /* would like to use sscanf here, but apparently not available in kernel */
752 /*ARGSUSED*/
753 static void
754 rf_ConfigureDebug(RF_Config_t *cfgPtr)
755 {
756 char *val_p, *name_p, *white_p;
757 long val;
758 int i;
759
760 rf_ResetDebugOptions();
761 for (i = 0; cfgPtr->debugVars[i][0] && i < RF_MAXDBGV; i++) {
762 name_p = rf_find_non_white(&cfgPtr->debugVars[i][0]);
763 white_p = rf_find_white(name_p); /* skip to start of 2nd
764 * word */
765 val_p = rf_find_non_white(white_p);
766 if (*val_p == '0' && *(val_p + 1) == 'x')
767 val = rf_htoi(val_p + 2);
768 else
769 val = rf_atoi(val_p);
770 *white_p = '\0';
771 set_debug_option(name_p, val);
772 }
773 }
774
775 void
776 rf_print_panic_message(int line, char *file)
777 {
778 sprintf(rf_panicbuf,"raidframe error at line %d file %s",
779 line, file);
780 }
781
782 #ifdef RAID_DIAGNOSTIC
783 void
784 rf_print_assert_panic_message(int line, char *file, char *condition)
785 {
786 sprintf(rf_panicbuf,
787 "raidframe error at line %d file %s (failed asserting %s)\n",
788 line, file, condition);
789 }
790 #endif
791
792 void
793 rf_print_unable_to_init_mutex(char *file, int line, int rc)
794 {
795 RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n",
796 file, line, rc);
797 }
798
799 void
800 rf_print_unable_to_add_shutdown(char *file, int line, int rc)
801 {
802 RF_ERRORMSG3("Unable to add to shutdown list file %s line %d rc=%d\n",
803 file, line, rc);
804 }
805