rf_driver.c revision 1.96 1 /* $NetBSD: rf_driver.c,v 1.96 2004/03/13 02:31:12 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 1.96 2004/03/13 02:31:12 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 int rc;
284
285 RF_LOCK_LKMGR_MUTEX(configureMutex);
286 configureCount++;
287 if (isconfigged == 0) {
288 rf_mutex_init(&rf_printf_mutex);
289
290 /* initialize globals */
291
292 DO_INIT_CONFIGURE(rf_ConfigureAllocList);
293
294 /*
295 * Yes, this does make debugging general to the whole
296 * system instead of being array specific. Bummer, drag.
297 */
298 rf_ConfigureDebug(cfgPtr);
299 DO_INIT_CONFIGURE(rf_ConfigureDebugMem);
300 #if RF_ACC_TRACE > 0
301 DO_INIT_CONFIGURE(rf_ConfigureAccessTrace);
302 #endif
303 DO_INIT_CONFIGURE(rf_ConfigureMapModule);
304 DO_INIT_CONFIGURE(rf_ConfigureReconEvent);
305 DO_INIT_CONFIGURE(rf_ConfigureCallback);
306 DO_INIT_CONFIGURE(rf_ConfigureRDFreeList);
307 DO_INIT_CONFIGURE(rf_ConfigureNWayXor);
308 DO_INIT_CONFIGURE(rf_ConfigureStripeLockFreeList);
309 DO_INIT_CONFIGURE(rf_ConfigureMCPair);
310 DO_INIT_CONFIGURE(rf_ConfigureDAGs);
311 DO_INIT_CONFIGURE(rf_ConfigureDAGFuncs);
312 DO_INIT_CONFIGURE(rf_ConfigureReconstruction);
313 DO_INIT_CONFIGURE(rf_ConfigureCopyback);
314 DO_INIT_CONFIGURE(rf_ConfigureDiskQueueSystem);
315 isconfigged = 1;
316 }
317 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
318
319 DO_RAID_MUTEX(&raidPtr->mutex);
320 /* set up the cleanup list. Do this after ConfigureDebug so that
321 * value of memDebug will be set */
322
323 rf_MakeAllocList(raidPtr->cleanupList);
324 if (raidPtr->cleanupList == NULL) {
325 DO_RAID_FAIL();
326 return (ENOMEM);
327 }
328 rf_ShutdownCreate(&raidPtr->shutdownList,
329 (void (*) (void *)) rf_FreeAllocList,
330 raidPtr->cleanupList);
331
332 raidPtr->numCol = cfgPtr->numCol;
333 raidPtr->numSpare = cfgPtr->numSpare;
334
335 raidPtr->status = rf_rs_optimal;
336 raidPtr->reconControl = NULL;
337
338 TAILQ_INIT(&(raidPtr->iodone));
339 simple_lock_init(&(raidPtr->iodone_lock));
340
341 DO_RAID_INIT_CONFIGURE(rf_ConfigureEngine);
342 DO_RAID_INIT_CONFIGURE(rf_ConfigureStripeLocks);
343
344 raidPtr->outstandingCond = 0;
345
346 raidPtr->nAccOutstanding = 0;
347 raidPtr->waitShutdown = 0;
348
349 DO_RAID_MUTEX(&raidPtr->access_suspend_mutex);
350
351 raidPtr->waitForReconCond = 0;
352
353 if (ac!=NULL) {
354 /* We have an AutoConfig structure.. Don't do the
355 normal disk configuration... call the auto config
356 stuff */
357 rf_AutoConfigureDisks(raidPtr, cfgPtr, ac);
358 } else {
359 DO_RAID_INIT_CONFIGURE(rf_ConfigureDisks);
360 DO_RAID_INIT_CONFIGURE(rf_ConfigureSpareDisks);
361 }
362 /* do this after ConfigureDisks & ConfigureSpareDisks to be sure dev
363 * no. is set */
364 DO_RAID_INIT_CONFIGURE(rf_ConfigureDiskQueues);
365
366 DO_RAID_INIT_CONFIGURE(rf_ConfigureLayout);
367
368 DO_RAID_INIT_CONFIGURE(rf_ConfigurePSStatus);
369
370 #if RF_INCLUDE_CHAINDECLUSTER > 0
371 for (col = 0; col < raidPtr->numCol; col++) {
372 /*
373 * XXX better distribution
374 */
375 raidPtr->hist_diskreq[col] = 0;
376 }
377 #endif
378 raidPtr->numNewFailures = 0;
379 raidPtr->copyback_in_progress = 0;
380 raidPtr->parity_rewrite_in_progress = 0;
381 raidPtr->adding_hot_spare = 0;
382 raidPtr->recon_in_progress = 0;
383 raidPtr->maxOutstanding = cfgPtr->maxOutstandingDiskReqs;
384
385 /* autoconfigure and root_partition will actually get filled in
386 after the config is done */
387 raidPtr->autoconfigure = 0;
388 raidPtr->root_partition = 0;
389 raidPtr->last_unit = raidPtr->raidid;
390 raidPtr->config_order = 0;
391
392 if (rf_keepAccTotals) {
393 raidPtr->keep_acc_totals = 1;
394 }
395
396 raidPtr->valid = 1;
397
398 printf("raid%d: %s\n", raidPtr->raidid,
399 raidPtr->Layout.map->configName);
400 printf("raid%d: Components:", raidPtr->raidid);
401
402 for (col = 0; col < raidPtr->numCol; col++) {
403 printf(" %s", raidPtr->Disks[col].devname);
404 if (RF_DEAD_DISK(raidPtr->Disks[col].status)) {
405 printf("[**FAILED**]");
406 }
407 }
408 printf("\n");
409 printf("raid%d: Total Sectors: %lu (%lu MB)\n",
410 raidPtr->raidid,
411 (unsigned long) raidPtr->totalSectors,
412 (unsigned long) (raidPtr->totalSectors / 1024 *
413 (1 << raidPtr->logBytesPerSector) / 1024));
414
415 return (0);
416 }
417
418 static void
419 rf_ShutdownRDFreeList(void *ignored)
420 {
421 pool_destroy(&rf_pools.rad);
422 }
423
424 static int
425 rf_ConfigureRDFreeList(RF_ShutdownList_t **listp)
426 {
427
428 rf_pool_init(&rf_pools.rad, sizeof(RF_RaidAccessDesc_t),
429 "rf_rad_pl", RF_MIN_FREE_RAD, RF_MAX_FREE_RAD);
430 rf_ShutdownCreate(listp, rf_ShutdownRDFreeList, NULL);
431 simple_lock_init(&rf_rad_lock);
432 return (0);
433 }
434
435 RF_RaidAccessDesc_t *
436 rf_AllocRaidAccDesc(RF_Raid_t *raidPtr, RF_IoType_t type,
437 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
438 caddr_t bufPtr, void *bp, RF_RaidAccessFlags_t flags,
439 RF_AccessState_t *states)
440 {
441 RF_RaidAccessDesc_t *desc;
442
443 desc = pool_get(&rf_pools.rad, PR_WAITOK);
444
445 RF_LOCK_MUTEX(rf_rad_lock);
446 if (raidPtr->waitShutdown) {
447 /*
448 * Actually, we're shutting the array down. Free the desc
449 * and return NULL.
450 */
451
452 RF_UNLOCK_MUTEX(rf_rad_lock);
453 pool_put(&rf_pools.rad, desc);
454 return (NULL);
455 }
456 raidPtr->nAccOutstanding++;
457
458 RF_UNLOCK_MUTEX(rf_rad_lock);
459
460 desc->raidPtr = (void *) raidPtr;
461 desc->type = type;
462 desc->raidAddress = raidAddress;
463 desc->numBlocks = numBlocks;
464 desc->bufPtr = bufPtr;
465 desc->bp = bp;
466 desc->flags = flags;
467 desc->states = states;
468 desc->state = 0;
469
470 desc->status = 0;
471 #if RF_ACC_TRACE > 0
472 memset((char *) &desc->tracerec, 0, sizeof(RF_AccTraceEntry_t));
473 #endif
474 desc->callbackFunc = NULL;
475 desc->callbackArg = NULL;
476 desc->next = NULL;
477 desc->cleanupList = NULL;
478 rf_MakeAllocList(desc->cleanupList);
479 return (desc);
480 }
481
482 void
483 rf_FreeRaidAccDesc(RF_RaidAccessDesc_t *desc)
484 {
485 RF_Raid_t *raidPtr = desc->raidPtr;
486 RF_DagList_t *dagList, *temp;
487
488 RF_ASSERT(desc);
489
490 /* Cleanup the dagList(s) */
491 dagList = desc->dagList;
492 while(dagList != NULL) {
493 temp = dagList;
494 dagList = dagList->next;
495 rf_FreeDAGList(temp);
496 }
497
498 rf_FreeAllocList(desc->cleanupList);
499 pool_put(&rf_pools.rad, desc);
500 RF_LOCK_MUTEX(rf_rad_lock);
501 raidPtr->nAccOutstanding--;
502 if (raidPtr->waitShutdown) {
503 RF_SIGNAL_COND(raidPtr->outstandingCond);
504 }
505 RF_UNLOCK_MUTEX(rf_rad_lock);
506 }
507 /*********************************************************************
508 * Main routine for performing an access.
509 * Accesses are retried until a DAG can not be selected. This occurs
510 * when either the DAG library is incomplete or there are too many
511 * failures in a parity group.
512 *
513 * type should be read or write async_flag should be RF_TRUE or
514 * RF_FALSE bp_in is a buf pointer. void * to facilitate ignoring it
515 * outside the kernel
516 ********************************************************************/
517 int
518 rf_DoAccess(RF_Raid_t * raidPtr, RF_IoType_t type, int async_flag,
519 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
520 caddr_t bufPtr, void *bp_in, RF_RaidAccessFlags_t flags)
521 {
522 RF_RaidAccessDesc_t *desc;
523 caddr_t lbufPtr = bufPtr;
524 struct buf *bp = (struct buf *) bp_in;
525
526 raidAddress += rf_raidSectorOffset;
527
528 #if RF_ACCESS_DEBUG
529 if (rf_accessDebug) {
530
531 printf("logBytes is: %d %d %d\n", raidPtr->raidid,
532 raidPtr->logBytesPerSector,
533 (int) rf_RaidAddressToByte(raidPtr, numBlocks));
534 printf("raid%d: %s raidAddr %d (stripeid %d-%d) numBlocks %d (%d bytes) buf 0x%lx\n", raidPtr->raidid,
535 (type == RF_IO_TYPE_READ) ? "READ" : "WRITE", (int) raidAddress,
536 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress),
537 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress + numBlocks - 1),
538 (int) numBlocks,
539 (int) rf_RaidAddressToByte(raidPtr, numBlocks),
540 (long) bufPtr);
541 }
542 #endif
543 if (raidAddress + numBlocks > raidPtr->totalSectors) {
544
545 printf("DoAccess: raid addr %lu too large to access %lu sectors. Max legal addr is %lu\n",
546 (u_long) raidAddress, (u_long) numBlocks, (u_long) raidPtr->totalSectors);
547
548
549 bp->b_flags |= B_ERROR;
550 bp->b_resid = bp->b_bcount;
551 bp->b_error = ENOSPC;
552 biodone(bp);
553 return (ENOSPC);
554 }
555 desc = rf_AllocRaidAccDesc(raidPtr, type, raidAddress,
556 numBlocks, lbufPtr, bp, flags, raidPtr->Layout.map->states);
557
558 if (desc == NULL) {
559 return (ENOMEM);
560 }
561 #if RF_ACC_TRACE > 0
562 RF_ETIMER_START(desc->tracerec.tot_timer);
563 #endif
564 desc->async_flag = async_flag;
565
566 rf_ContinueRaidAccess(desc);
567
568 return (0);
569 }
570 #if 0
571 /* force the array into reconfigured mode without doing reconstruction */
572 int
573 rf_SetReconfiguredMode(RF_Raid_t *raidPtr, int col)
574 {
575 if (!(raidPtr->Layout.map->flags & RF_DISTRIBUTE_SPARE)) {
576 printf("Can't set reconfigured mode in dedicated-spare array\n");
577 RF_PANIC();
578 }
579 RF_LOCK_MUTEX(raidPtr->mutex);
580 raidPtr->numFailures++;
581 raidPtr->Disks[col].status = rf_ds_dist_spared;
582 raidPtr->status = rf_rs_reconfigured;
583 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
584 /* install spare table only if declustering + distributed sparing
585 * architecture. */
586 if (raidPtr->Layout.map->flags & RF_BD_DECLUSTERED)
587 rf_InstallSpareTable(raidPtr, col);
588 RF_UNLOCK_MUTEX(raidPtr->mutex);
589 return (0);
590 }
591 #endif
592
593 int
594 rf_FailDisk(RF_Raid_t *raidPtr, int fcol, int initRecon)
595 {
596 RF_LOCK_MUTEX(raidPtr->mutex);
597 if (raidPtr->Disks[fcol].status != rf_ds_failed) {
598 /* must be failing something that is valid, or else it's
599 already marked as failed (in which case we don't
600 want to mark it failed again!) */
601 raidPtr->numFailures++;
602 raidPtr->Disks[fcol].status = rf_ds_failed;
603 raidPtr->status = rf_rs_degraded;
604 }
605 RF_UNLOCK_MUTEX(raidPtr->mutex);
606
607 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
608
609 /* Close the component, so that it's not "locked" if someone
610 else want's to use it! */
611
612 rf_close_component(raidPtr, raidPtr->raid_cinfo[fcol].ci_vp,
613 raidPtr->Disks[fcol].auto_configured);
614
615 RF_LOCK_MUTEX(raidPtr->mutex);
616 raidPtr->raid_cinfo[fcol].ci_vp = NULL;
617
618 /* Need to mark the component as not being auto_configured
619 (in case it was previously). */
620
621 raidPtr->Disks[fcol].auto_configured = 0;
622 RF_UNLOCK_MUTEX(raidPtr->mutex);
623
624 if (initRecon)
625 rf_ReconstructFailedDisk(raidPtr, fcol);
626 return (0);
627 }
628 /* releases a thread that is waiting for the array to become quiesced.
629 * access_suspend_mutex should be locked upon calling this
630 */
631 void
632 rf_SignalQuiescenceLock(RF_Raid_t *raidPtr)
633 {
634 #if RF_DEBUG_QUIESCE
635 if (rf_quiesceDebug) {
636 printf("raid%d: Signalling quiescence lock\n",
637 raidPtr->raidid);
638 }
639 #endif
640 raidPtr->access_suspend_release = 1;
641
642 if (raidPtr->waiting_for_quiescence) {
643 SIGNAL_QUIESCENT_COND(raidPtr);
644 }
645 }
646 /* suspends all new requests to the array. No effect on accesses that are in flight. */
647 int
648 rf_SuspendNewRequestsAndWait(RF_Raid_t *raidPtr)
649 {
650 #if RF_DEBUG_QUIESCE
651 if (rf_quiesceDebug)
652 printf("raid%d: Suspending new reqs\n", raidPtr->raidid);
653 #endif
654 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
655 raidPtr->accesses_suspended++;
656 raidPtr->waiting_for_quiescence = (raidPtr->accs_in_flight == 0) ? 0 : 1;
657
658 if (raidPtr->waiting_for_quiescence) {
659 raidPtr->access_suspend_release = 0;
660 while (!raidPtr->access_suspend_release) {
661 #if RF_DEBUG_QUIESCE
662 printf("raid%d: Suspending: Waiting for Quiescence\n",
663 raidPtr->raidid);
664 #endif
665 WAIT_FOR_QUIESCENCE(raidPtr);
666 raidPtr->waiting_for_quiescence = 0;
667 }
668 }
669 #if RF_DEBUG_QUIESCE
670 printf("raid%d: Quiescence reached..\n", raidPtr->raidid);
671 #endif
672
673 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
674 return (raidPtr->waiting_for_quiescence);
675 }
676 /* wake up everyone waiting for quiescence to be released */
677 void
678 rf_ResumeNewRequests(RF_Raid_t *raidPtr)
679 {
680 RF_CallbackDesc_t *t, *cb;
681
682 #if RF_DEBUG_QUIESCE
683 if (rf_quiesceDebug)
684 printf("Resuming new reqs\n");
685 #endif
686
687 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
688 raidPtr->accesses_suspended--;
689 if (raidPtr->accesses_suspended == 0)
690 cb = raidPtr->quiesce_wait_list;
691 else
692 cb = NULL;
693 raidPtr->quiesce_wait_list = NULL;
694 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
695
696 while (cb) {
697 t = cb;
698 cb = cb->next;
699 (t->callbackFunc) (t->callbackArg);
700 rf_FreeCallbackDesc(t);
701 }
702 }
703 /*****************************************************************************************
704 *
705 * debug routines
706 *
707 ****************************************************************************************/
708
709 static void
710 set_debug_option(char *name, long val)
711 {
712 RF_DebugName_t *p;
713
714 for (p = rf_debugNames; p->name; p++) {
715 if (!strcmp(p->name, name)) {
716 *(p->ptr) = val;
717 printf("[Set debug variable %s to %ld]\n", name, val);
718 return;
719 }
720 }
721 RF_ERRORMSG1("Unknown debug string \"%s\"\n", name);
722 }
723
724
725 /* would like to use sscanf here, but apparently not available in kernel */
726 /*ARGSUSED*/
727 static void
728 rf_ConfigureDebug(RF_Config_t *cfgPtr)
729 {
730 char *val_p, *name_p, *white_p;
731 long val;
732 int i;
733
734 rf_ResetDebugOptions();
735 for (i = 0; cfgPtr->debugVars[i][0] && i < RF_MAXDBGV; i++) {
736 name_p = rf_find_non_white(&cfgPtr->debugVars[i][0]);
737 white_p = rf_find_white(name_p); /* skip to start of 2nd
738 * word */
739 val_p = rf_find_non_white(white_p);
740 if (*val_p == '0' && *(val_p + 1) == 'x')
741 val = rf_htoi(val_p + 2);
742 else
743 val = rf_atoi(val_p);
744 *white_p = '\0';
745 set_debug_option(name_p, val);
746 }
747 }
748
749 void
750 rf_print_panic_message(int line, char *file)
751 {
752 sprintf(rf_panicbuf,"raidframe error at line %d file %s",
753 line, file);
754 }
755
756 #ifdef RAID_DIAGNOSTIC
757 void
758 rf_print_assert_panic_message(int line, char *file, char *condition)
759 {
760 sprintf(rf_panicbuf,
761 "raidframe error at line %d file %s (failed asserting %s)\n",
762 line, file, condition);
763 }
764 #endif
765
766 void
767 rf_print_unable_to_init_mutex(char *file, int line, int rc)
768 {
769 RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n",
770 file, line, rc);
771 }
772
773 void
774 rf_print_unable_to_add_shutdown(char *file, int line, int rc)
775 {
776 RF_ERRORMSG3("Unable to add to shutdown list file %s line %d rc=%d\n",
777 file, line, rc);
778 }
779