rf_driver.c revision 1.119 1 /* $NetBSD: rf_driver.c,v 1.119 2008/11/18 14:29:55 ad 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 *
18 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
19 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
20 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
21 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
22 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28 * POSSIBILITY OF SUCH DAMAGE.
29 */
30
31 /*
32 * Copyright (c) 1995 Carnegie-Mellon University.
33 * All rights reserved.
34 *
35 * Author: Mark Holland, Khalil Amiri, Claudson Bornstein, William V. Courtright II,
36 * Robby Findler, Daniel Stodolsky, Rachad Youssef, Jim Zelenka
37 *
38 * Permission to use, copy, modify and distribute this software and
39 * its documentation is hereby granted, provided that both the copyright
40 * notice and this permission notice appear in all copies of the
41 * software, derivative works or modified versions, and any portions
42 * thereof, and that both notices appear in supporting documentation.
43 *
44 * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
45 * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
46 * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
47 *
48 * Carnegie Mellon requests users of this software to return to
49 *
50 * Software Distribution Coordinator or Software.Distribution (at) CS.CMU.EDU
51 * School of Computer Science
52 * Carnegie Mellon University
53 * Pittsburgh PA 15213-3890
54 *
55 * any improvements or extensions that they make and grant Carnegie the
56 * rights to redistribute these changes.
57 */
58
59 /******************************************************************************
60 *
61 * rf_driver.c -- main setup, teardown, and access routines for the RAID driver
62 *
63 * all routines are prefixed with rf_ (raidframe), to avoid conficts.
64 *
65 ******************************************************************************/
66
67
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: rf_driver.c,v 1.119 2008/11/18 14:29:55 ad Exp $");
70
71 #ifdef _KERNEL_OPT
72 #include "opt_raid_diagnostic.h"
73 #endif
74
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/ioctl.h>
78 #include <sys/fcntl.h>
79 #include <sys/vnode.h>
80
81
82 #include "rf_archs.h"
83 #include "rf_threadstuff.h"
84
85 #include <sys/errno.h>
86
87 #include "rf_raid.h"
88 #include "rf_dag.h"
89 #include "rf_aselect.h"
90 #include "rf_diskqueue.h"
91 #include "rf_parityscan.h"
92 #include "rf_alloclist.h"
93 #include "rf_dagutils.h"
94 #include "rf_utils.h"
95 #include "rf_etimer.h"
96 #include "rf_acctrace.h"
97 #include "rf_general.h"
98 #include "rf_desc.h"
99 #include "rf_states.h"
100 #include "rf_decluster.h"
101 #include "rf_map.h"
102 #include "rf_revent.h"
103 #include "rf_callback.h"
104 #include "rf_engine.h"
105 #include "rf_mcpair.h"
106 #include "rf_nwayxor.h"
107 #include "rf_copyback.h"
108 #include "rf_driver.h"
109 #include "rf_options.h"
110 #include "rf_shutdown.h"
111 #include "rf_kintf.h"
112
113 #include <sys/buf.h>
114
115 #ifndef RF_ACCESS_DEBUG
116 #define RF_ACCESS_DEBUG 0
117 #endif
118
119 /* rad == RF_RaidAccessDesc_t */
120 RF_DECLARE_MUTEX(rf_rad_lock)
121 #define RF_MAX_FREE_RAD 128
122 #define RF_MIN_FREE_RAD 32
123
124 /* debug variables */
125 char rf_panicbuf[2048]; /* a buffer to hold an error msg when we panic */
126
127 /* main configuration routines */
128 static int raidframe_booted = 0;
129
130 static void rf_ConfigureDebug(RF_Config_t * cfgPtr);
131 static void set_debug_option(char *name, long val);
132 static void rf_UnconfigureArray(void);
133 static void rf_ShutdownRDFreeList(void *);
134 static int rf_ConfigureRDFreeList(RF_ShutdownList_t **);
135
136 RF_DECLARE_MUTEX(rf_printf_mutex) /* debug only: avoids interleaved
137 * printfs by different stripes */
138
139 #define SIGNAL_QUIESCENT_COND(_raid_) wakeup(&((_raid_)->accesses_suspended))
140 #define WAIT_FOR_QUIESCENCE(_raid_) \
141 ltsleep(&((_raid_)->accesses_suspended), PRIBIO, \
142 "raidframe quiesce", 0, &((_raid_)->access_suspend_mutex))
143
144 static int configureCount = 0; /* number of active configurations */
145 static int isconfigged = 0; /* is basic raidframe (non per-array)
146 * stuff configured */
147 RF_DECLARE_LKMGR_STATIC_MUTEX(configureMutex) /* used to lock the configuration
148 * stuff */
149 static RF_ShutdownList_t *globalShutdown; /* non array-specific
150 * stuff */
151
152 static int rf_ConfigureRDFreeList(RF_ShutdownList_t ** listp);
153 static int rf_AllocEmergBuffers(RF_Raid_t *);
154 static void rf_FreeEmergBuffers(RF_Raid_t *);
155
156 /* called at system boot time */
157 int
158 rf_BootRaidframe()
159 {
160
161 if (raidframe_booted)
162 return (EBUSY);
163 raidframe_booted = 1;
164 mutex_init(&configureMutex, MUTEX_DEFAULT, IPL_NONE);
165 configureCount = 0;
166 isconfigged = 0;
167 globalShutdown = NULL;
168 return (0);
169 }
170
171 /*
172 * Called whenever an array is shutdown
173 */
174 static void
175 rf_UnconfigureArray()
176 {
177
178 RF_LOCK_LKMGR_MUTEX(configureMutex);
179 if (--configureCount == 0) { /* if no active configurations, shut
180 * everything down */
181 isconfigged = 0;
182 rf_ShutdownList(&globalShutdown);
183
184 /*
185 * We must wait until now, because the AllocList module
186 * uses the DebugMem module.
187 */
188 #if RF_DEBUG_MEM
189 if (rf_memDebug)
190 rf_print_unfreed();
191 #endif
192 }
193 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
194 }
195
196 /*
197 * Called to shut down an array.
198 */
199 int
200 rf_Shutdown(RF_Raid_t *raidPtr)
201 {
202
203 if (!raidPtr->valid) {
204 RF_ERRORMSG("Attempt to shut down unconfigured RAIDframe driver. Aborting shutdown\n");
205 return (EINVAL);
206 }
207 /*
208 * wait for outstanding IOs to land
209 * As described in rf_raid.h, we use the rad_freelist lock
210 * to protect the per-array info about outstanding descs
211 * since we need to do freelist locking anyway, and this
212 * cuts down on the amount of serialization we've got going
213 * on.
214 */
215 RF_LOCK_MUTEX(rf_rad_lock);
216 if (raidPtr->waitShutdown) {
217 RF_UNLOCK_MUTEX(rf_rad_lock);
218 return (EBUSY);
219 }
220 raidPtr->waitShutdown = 1;
221 while (raidPtr->nAccOutstanding) {
222 RF_WAIT_COND(raidPtr->outstandingCond, rf_rad_lock);
223 }
224 RF_UNLOCK_MUTEX(rf_rad_lock);
225
226 /* Wait for any parity re-writes to stop... */
227 while (raidPtr->parity_rewrite_in_progress) {
228 printf("Waiting for parity re-write to exit...\n");
229 tsleep(&raidPtr->parity_rewrite_in_progress, PRIBIO,
230 "rfprwshutdown", 0);
231 }
232
233 raidPtr->valid = 0;
234
235 rf_update_component_labels(raidPtr, RF_FINAL_COMPONENT_UPDATE);
236
237 rf_UnconfigureVnodes(raidPtr);
238
239 rf_FreeEmergBuffers(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_FreeEmergBuffers(raidPtr); \
263 rf_ShutdownList(&raidPtr->shutdownList); \
264 rf_UnconfigureArray(); \
265 }
266
267 #define DO_RAID_INIT_CONFIGURE(f) { \
268 rc = f (&raidPtr->shutdownList, raidPtr, cfgPtr); \
269 if (rc) { \
270 RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
271 DO_RAID_FAIL(); \
272 return(rc); \
273 } \
274 }
275
276 #define DO_RAID_MUTEX(_m_) { \
277 rf_mutex_init((_m_)); \
278 }
279
280 int
281 rf_Configure(RF_Raid_t *raidPtr, RF_Config_t *cfgPtr, RF_AutoConfig_t *ac)
282 {
283 RF_RowCol_t col;
284 int rc;
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 DO_INIT_CONFIGURE(rf_ConfigurePSStatus);
317 isconfigged = 1;
318 }
319 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
320
321 DO_RAID_MUTEX(&raidPtr->mutex);
322 /* set up the cleanup list. Do this after ConfigureDebug so that
323 * value of memDebug will be set */
324
325 rf_MakeAllocList(raidPtr->cleanupList);
326 if (raidPtr->cleanupList == NULL) {
327 DO_RAID_FAIL();
328 return (ENOMEM);
329 }
330 rf_ShutdownCreate(&raidPtr->shutdownList,
331 (void (*) (void *)) rf_FreeAllocList,
332 raidPtr->cleanupList);
333
334 raidPtr->numCol = cfgPtr->numCol;
335 raidPtr->numSpare = cfgPtr->numSpare;
336
337 raidPtr->status = rf_rs_optimal;
338 raidPtr->reconControl = NULL;
339
340 TAILQ_INIT(&(raidPtr->iodone));
341 simple_lock_init(&(raidPtr->iodone_lock));
342
343 DO_RAID_INIT_CONFIGURE(rf_ConfigureEngine);
344 DO_RAID_INIT_CONFIGURE(rf_ConfigureStripeLocks);
345
346 raidPtr->outstandingCond = 0;
347
348 raidPtr->nAccOutstanding = 0;
349 raidPtr->waitShutdown = 0;
350
351 DO_RAID_MUTEX(&raidPtr->access_suspend_mutex);
352
353 raidPtr->waitForReconCond = 0;
354
355 if (ac!=NULL) {
356 /* We have an AutoConfig structure.. Don't do the
357 normal disk configuration... call the auto config
358 stuff */
359 rf_AutoConfigureDisks(raidPtr, cfgPtr, ac);
360 } else {
361 DO_RAID_INIT_CONFIGURE(rf_ConfigureDisks);
362 DO_RAID_INIT_CONFIGURE(rf_ConfigureSpareDisks);
363 }
364 /* do this after ConfigureDisks & ConfigureSpareDisks to be sure dev
365 * no. is set */
366 DO_RAID_INIT_CONFIGURE(rf_ConfigureDiskQueues);
367
368 DO_RAID_INIT_CONFIGURE(rf_ConfigureLayout);
369
370 /* Initialize per-RAID PSS bits */
371 rf_InitPSStatus(raidPtr);
372
373 #if RF_INCLUDE_CHAINDECLUSTER > 0
374 for (col = 0; col < raidPtr->numCol; col++) {
375 /*
376 * XXX better distribution
377 */
378 raidPtr->hist_diskreq[col] = 0;
379 }
380 #endif
381 raidPtr->numNewFailures = 0;
382 raidPtr->copyback_in_progress = 0;
383 raidPtr->parity_rewrite_in_progress = 0;
384 raidPtr->adding_hot_spare = 0;
385 raidPtr->recon_in_progress = 0;
386 raidPtr->maxOutstanding = cfgPtr->maxOutstandingDiskReqs;
387
388 /* autoconfigure and root_partition will actually get filled in
389 after the config is done */
390 raidPtr->autoconfigure = 0;
391 raidPtr->root_partition = 0;
392 raidPtr->last_unit = raidPtr->raidid;
393 raidPtr->config_order = 0;
394
395 if (rf_keepAccTotals) {
396 raidPtr->keep_acc_totals = 1;
397 }
398
399 /* Allocate a bunch of buffers to be used in low-memory conditions */
400 raidPtr->iobuf = NULL;
401
402 rc = rf_AllocEmergBuffers(raidPtr);
403 if (rc) {
404 printf("raid%d: Unable to allocate emergency buffers.\n",
405 raidPtr->raidid);
406 DO_RAID_FAIL();
407 return(rc);
408 }
409
410 raidPtr->valid = 1;
411
412 printf("raid%d: %s\n", raidPtr->raidid,
413 raidPtr->Layout.map->configName);
414 printf("raid%d: Components:", raidPtr->raidid);
415
416 for (col = 0; col < raidPtr->numCol; col++) {
417 printf(" %s", raidPtr->Disks[col].devname);
418 if (RF_DEAD_DISK(raidPtr->Disks[col].status)) {
419 printf("[**FAILED**]");
420 }
421 }
422 printf("\n");
423 printf("raid%d: Total Sectors: %" PRIu64 " (%" PRIu64 " MB)\n",
424 raidPtr->raidid,
425 raidPtr->totalSectors,
426 (raidPtr->totalSectors / 1024 *
427 (1 << raidPtr->logBytesPerSector) / 1024));
428
429 return (0);
430 }
431
432
433 /*
434
435 Routines to allocate and free the "emergency buffers" for a given
436 RAID set. These emergency buffers will be used when the kernel runs
437 out of kernel memory.
438
439 */
440
441 static int
442 rf_AllocEmergBuffers(RF_Raid_t *raidPtr)
443 {
444 void *tmpbuf;
445 RF_VoidPointerListElem_t *vple;
446 int i;
447
448 /* XXX next line needs tuning... */
449 raidPtr->numEmergencyBuffers = 10 * raidPtr->numCol;
450 #if DEBUG
451 printf("raid%d: allocating %d buffers of %d bytes.\n",
452 raidPtr->raidid,
453 raidPtr->numEmergencyBuffers,
454 (int)(raidPtr->Layout.sectorsPerStripeUnit <<
455 raidPtr->logBytesPerSector));
456 #endif
457 for (i = 0; i < raidPtr->numEmergencyBuffers; i++) {
458 tmpbuf = malloc( raidPtr->Layout.sectorsPerStripeUnit <<
459 raidPtr->logBytesPerSector,
460 M_RAIDFRAME, M_WAITOK);
461 if (tmpbuf) {
462 vple = rf_AllocVPListElem();
463 vple->p= tmpbuf;
464 vple->next = raidPtr->iobuf;
465 raidPtr->iobuf = vple;
466 raidPtr->iobuf_count++;
467 } else {
468 printf("raid%d: failed to allocate emergency buffer!\n",
469 raidPtr->raidid);
470 return 1;
471 }
472 }
473
474 /* XXX next line needs tuning too... */
475 raidPtr->numEmergencyStripeBuffers = 10;
476 for (i = 0; i < raidPtr->numEmergencyStripeBuffers; i++) {
477 tmpbuf = malloc( raidPtr->numCol * (raidPtr->Layout.sectorsPerStripeUnit <<
478 raidPtr->logBytesPerSector),
479 M_RAIDFRAME, M_WAITOK);
480 if (tmpbuf) {
481 vple = rf_AllocVPListElem();
482 vple->p= tmpbuf;
483 vple->next = raidPtr->stripebuf;
484 raidPtr->stripebuf = vple;
485 raidPtr->stripebuf_count++;
486 } else {
487 printf("raid%d: failed to allocate emergency stripe buffer!\n",
488 raidPtr->raidid);
489 return 1;
490 }
491 }
492
493 return (0);
494 }
495
496 static void
497 rf_FreeEmergBuffers(RF_Raid_t *raidPtr)
498 {
499 RF_VoidPointerListElem_t *tmp;
500
501 /* Free the emergency IO buffers */
502 while (raidPtr->iobuf != NULL) {
503 tmp = raidPtr->iobuf;
504 raidPtr->iobuf = raidPtr->iobuf->next;
505 free(tmp->p, M_RAIDFRAME);
506 rf_FreeVPListElem(tmp);
507 }
508
509 /* Free the emergency stripe buffers */
510 while (raidPtr->stripebuf != NULL) {
511 tmp = raidPtr->stripebuf;
512 raidPtr->stripebuf = raidPtr->stripebuf->next;
513 free(tmp->p, M_RAIDFRAME);
514 rf_FreeVPListElem(tmp);
515 }
516 }
517
518
519 static void
520 rf_ShutdownRDFreeList(void *ignored)
521 {
522 pool_destroy(&rf_pools.rad);
523 }
524
525 static int
526 rf_ConfigureRDFreeList(RF_ShutdownList_t **listp)
527 {
528
529 rf_pool_init(&rf_pools.rad, sizeof(RF_RaidAccessDesc_t),
530 "rf_rad_pl", RF_MIN_FREE_RAD, RF_MAX_FREE_RAD);
531 rf_ShutdownCreate(listp, rf_ShutdownRDFreeList, NULL);
532 simple_lock_init(&rf_rad_lock);
533 return (0);
534 }
535
536 RF_RaidAccessDesc_t *
537 rf_AllocRaidAccDesc(RF_Raid_t *raidPtr, RF_IoType_t type,
538 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
539 void *bufPtr, void *bp, RF_RaidAccessFlags_t flags,
540 const RF_AccessState_t *states)
541 {
542 RF_RaidAccessDesc_t *desc;
543
544 desc = pool_get(&rf_pools.rad, PR_WAITOK);
545
546 RF_LOCK_MUTEX(rf_rad_lock);
547 if (raidPtr->waitShutdown) {
548 /*
549 * Actually, we're shutting the array down. Free the desc
550 * and return NULL.
551 */
552
553 RF_UNLOCK_MUTEX(rf_rad_lock);
554 pool_put(&rf_pools.rad, desc);
555 return (NULL);
556 }
557 raidPtr->nAccOutstanding++;
558
559 RF_UNLOCK_MUTEX(rf_rad_lock);
560
561 desc->raidPtr = (void *) raidPtr;
562 desc->type = type;
563 desc->raidAddress = raidAddress;
564 desc->numBlocks = numBlocks;
565 desc->bufPtr = bufPtr;
566 desc->bp = bp;
567 desc->flags = flags;
568 desc->states = states;
569 desc->state = 0;
570 desc->dagList = NULL;
571
572 desc->status = 0;
573 desc->numRetries = 0;
574 #if RF_ACC_TRACE > 0
575 memset((char *) &desc->tracerec, 0, sizeof(RF_AccTraceEntry_t));
576 #endif
577 desc->callbackFunc = NULL;
578 desc->callbackArg = NULL;
579 desc->next = NULL;
580 desc->iobufs = NULL;
581 desc->stripebufs = NULL;
582
583 return (desc);
584 }
585
586 void
587 rf_FreeRaidAccDesc(RF_RaidAccessDesc_t *desc)
588 {
589 RF_Raid_t *raidPtr = desc->raidPtr;
590 RF_DagList_t *dagList, *temp;
591 RF_VoidPointerListElem_t *tmp;
592
593 RF_ASSERT(desc);
594
595 /* Cleanup the dagList(s) */
596 dagList = desc->dagList;
597 while(dagList != NULL) {
598 temp = dagList;
599 dagList = dagList->next;
600 rf_FreeDAGList(temp);
601 }
602
603 while (desc->iobufs) {
604 tmp = desc->iobufs;
605 desc->iobufs = desc->iobufs->next;
606 rf_FreeIOBuffer(raidPtr, tmp);
607 }
608
609 while (desc->stripebufs) {
610 tmp = desc->stripebufs;
611 desc->stripebufs = desc->stripebufs->next;
612 rf_FreeStripeBuffer(raidPtr, tmp);
613 }
614
615 pool_put(&rf_pools.rad, desc);
616 RF_LOCK_MUTEX(rf_rad_lock);
617 raidPtr->nAccOutstanding--;
618 if (raidPtr->waitShutdown) {
619 RF_SIGNAL_COND(raidPtr->outstandingCond);
620 }
621 RF_UNLOCK_MUTEX(rf_rad_lock);
622 }
623 /*********************************************************************
624 * Main routine for performing an access.
625 * Accesses are retried until a DAG can not be selected. This occurs
626 * when either the DAG library is incomplete or there are too many
627 * failures in a parity group.
628 *
629 * type should be read or write async_flag should be RF_TRUE or
630 * RF_FALSE bp_in is a buf pointer. void *to facilitate ignoring it
631 * outside the kernel
632 ********************************************************************/
633 int
634 rf_DoAccess(RF_Raid_t * raidPtr, RF_IoType_t type, int async_flag,
635 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
636 void *bufPtr, struct buf *bp, RF_RaidAccessFlags_t flags)
637 {
638 RF_RaidAccessDesc_t *desc;
639 void *lbufPtr = bufPtr;
640
641 raidAddress += rf_raidSectorOffset;
642
643 #if RF_ACCESS_DEBUG
644 if (rf_accessDebug) {
645
646 printf("logBytes is: %d %d %d\n", raidPtr->raidid,
647 raidPtr->logBytesPerSector,
648 (int) rf_RaidAddressToByte(raidPtr, numBlocks));
649 printf("raid%d: %s raidAddr %d (stripeid %d-%d) numBlocks %d (%d bytes) buf 0x%lx\n", raidPtr->raidid,
650 (type == RF_IO_TYPE_READ) ? "READ" : "WRITE", (int) raidAddress,
651 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress),
652 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress + numBlocks - 1),
653 (int) numBlocks,
654 (int) rf_RaidAddressToByte(raidPtr, numBlocks),
655 (long) bufPtr);
656 }
657 #endif
658
659 desc = rf_AllocRaidAccDesc(raidPtr, type, raidAddress,
660 numBlocks, lbufPtr, bp, flags, raidPtr->Layout.map->states);
661
662 if (desc == NULL) {
663 return (ENOMEM);
664 }
665 #if RF_ACC_TRACE > 0
666 RF_ETIMER_START(desc->tracerec.tot_timer);
667 #endif
668 desc->async_flag = async_flag;
669
670 rf_ContinueRaidAccess(desc);
671
672 return (0);
673 }
674 #if 0
675 /* force the array into reconfigured mode without doing reconstruction */
676 int
677 rf_SetReconfiguredMode(RF_Raid_t *raidPtr, int col)
678 {
679 if (!(raidPtr->Layout.map->flags & RF_DISTRIBUTE_SPARE)) {
680 printf("Can't set reconfigured mode in dedicated-spare array\n");
681 RF_PANIC();
682 }
683 RF_LOCK_MUTEX(raidPtr->mutex);
684 raidPtr->numFailures++;
685 raidPtr->Disks[col].status = rf_ds_dist_spared;
686 raidPtr->status = rf_rs_reconfigured;
687 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
688 /* install spare table only if declustering + distributed sparing
689 * architecture. */
690 if (raidPtr->Layout.map->flags & RF_BD_DECLUSTERED)
691 rf_InstallSpareTable(raidPtr, col);
692 RF_UNLOCK_MUTEX(raidPtr->mutex);
693 return (0);
694 }
695 #endif
696
697 int
698 rf_FailDisk(RF_Raid_t *raidPtr, int fcol, int initRecon)
699 {
700
701 /* need to suspend IO's here -- if there are DAGs in flight
702 and we pull the rug out from under ci_vp, Bad Things
703 can happen. */
704
705 rf_SuspendNewRequestsAndWait(raidPtr);
706
707 RF_LOCK_MUTEX(raidPtr->mutex);
708 if (raidPtr->Disks[fcol].status != rf_ds_failed) {
709 /* must be failing something that is valid, or else it's
710 already marked as failed (in which case we don't
711 want to mark it failed again!) */
712 raidPtr->numFailures++;
713 raidPtr->Disks[fcol].status = rf_ds_failed;
714 raidPtr->status = rf_rs_degraded;
715 }
716 RF_UNLOCK_MUTEX(raidPtr->mutex);
717
718 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
719
720 /* Close the component, so that it's not "locked" if someone
721 else want's to use it! */
722
723 rf_close_component(raidPtr, raidPtr->raid_cinfo[fcol].ci_vp,
724 raidPtr->Disks[fcol].auto_configured);
725
726 RF_LOCK_MUTEX(raidPtr->mutex);
727 raidPtr->raid_cinfo[fcol].ci_vp = NULL;
728
729 /* Need to mark the component as not being auto_configured
730 (in case it was previously). */
731
732 raidPtr->Disks[fcol].auto_configured = 0;
733 RF_UNLOCK_MUTEX(raidPtr->mutex);
734 /* now we can allow IO to continue -- we'll be suspending it
735 again in rf_ReconstructFailedDisk() if we have to.. */
736
737 rf_ResumeNewRequests(raidPtr);
738
739 if (initRecon)
740 rf_ReconstructFailedDisk(raidPtr, fcol);
741 return (0);
742 }
743 /* releases a thread that is waiting for the array to become quiesced.
744 * access_suspend_mutex should be locked upon calling this
745 */
746 void
747 rf_SignalQuiescenceLock(RF_Raid_t *raidPtr)
748 {
749 #if RF_DEBUG_QUIESCE
750 if (rf_quiesceDebug) {
751 printf("raid%d: Signalling quiescence lock\n",
752 raidPtr->raidid);
753 }
754 #endif
755 raidPtr->access_suspend_release = 1;
756
757 if (raidPtr->waiting_for_quiescence) {
758 SIGNAL_QUIESCENT_COND(raidPtr);
759 }
760 }
761 /* suspends all new requests to the array. No effect on accesses that are in flight. */
762 int
763 rf_SuspendNewRequestsAndWait(RF_Raid_t *raidPtr)
764 {
765 #if RF_DEBUG_QUIESCE
766 if (rf_quiesceDebug)
767 printf("raid%d: Suspending new reqs\n", raidPtr->raidid);
768 #endif
769 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
770 raidPtr->accesses_suspended++;
771 raidPtr->waiting_for_quiescence = (raidPtr->accs_in_flight == 0) ? 0 : 1;
772
773 if (raidPtr->waiting_for_quiescence) {
774 raidPtr->access_suspend_release = 0;
775 while (!raidPtr->access_suspend_release) {
776 #if RF_DEBUG_QUIESCE
777 printf("raid%d: Suspending: Waiting for Quiescence\n",
778 raidPtr->raidid);
779 #endif
780 WAIT_FOR_QUIESCENCE(raidPtr);
781 raidPtr->waiting_for_quiescence = 0;
782 }
783 }
784 #if RF_DEBUG_QUIESCE
785 printf("raid%d: Quiescence reached..\n", raidPtr->raidid);
786 #endif
787
788 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
789 return (raidPtr->waiting_for_quiescence);
790 }
791 /* wake up everyone waiting for quiescence to be released */
792 void
793 rf_ResumeNewRequests(RF_Raid_t *raidPtr)
794 {
795 RF_CallbackDesc_t *t, *cb;
796
797 #if RF_DEBUG_QUIESCE
798 if (rf_quiesceDebug)
799 printf("raid%d: Resuming new requests\n", raidPtr->raidid);
800 #endif
801
802 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
803 raidPtr->accesses_suspended--;
804 if (raidPtr->accesses_suspended == 0)
805 cb = raidPtr->quiesce_wait_list;
806 else
807 cb = NULL;
808 raidPtr->quiesce_wait_list = NULL;
809 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
810
811 while (cb) {
812 t = cb;
813 cb = cb->next;
814 (t->callbackFunc) (t->callbackArg);
815 rf_FreeCallbackDesc(t);
816 }
817 }
818 /*****************************************************************************************
819 *
820 * debug routines
821 *
822 ****************************************************************************************/
823
824 static void
825 set_debug_option(char *name, long val)
826 {
827 RF_DebugName_t *p;
828
829 for (p = rf_debugNames; p->name; p++) {
830 if (!strcmp(p->name, name)) {
831 *(p->ptr) = val;
832 printf("[Set debug variable %s to %ld]\n", name, val);
833 return;
834 }
835 }
836 RF_ERRORMSG1("Unknown debug string \"%s\"\n", name);
837 }
838
839
840 /* would like to use sscanf here, but apparently not available in kernel */
841 /*ARGSUSED*/
842 static void
843 rf_ConfigureDebug(RF_Config_t *cfgPtr)
844 {
845 char *val_p, *name_p, *white_p;
846 long val;
847 int i;
848
849 rf_ResetDebugOptions();
850 for (i = 0; cfgPtr->debugVars[i][0] && i < RF_MAXDBGV; i++) {
851 name_p = rf_find_non_white(&cfgPtr->debugVars[i][0]);
852 white_p = rf_find_white(name_p); /* skip to start of 2nd
853 * word */
854 val_p = rf_find_non_white(white_p);
855 if (*val_p == '0' && *(val_p + 1) == 'x')
856 val = rf_htoi(val_p + 2);
857 else
858 val = rf_atoi(val_p);
859 *white_p = '\0';
860 set_debug_option(name_p, val);
861 }
862 }
863
864 void
865 rf_print_panic_message(int line, const char *file)
866 {
867 snprintf(rf_panicbuf, sizeof(rf_panicbuf),
868 "raidframe error at line %d file %s", line, file);
869 }
870
871 #ifdef RAID_DIAGNOSTIC
872 void
873 rf_print_assert_panic_message(int line, const char *file, const char *condition)
874 {
875 snprintf(rf_panicbuf, sizeof(rf_panicbuf),
876 "raidframe error at line %d file %s (failed asserting %s)\n",
877 line, file, condition);
878 }
879 #endif
880
881 void
882 rf_print_unable_to_init_mutex(const char *file, int line, int rc)
883 {
884 RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n",
885 file, line, rc);
886 }
887
888 void
889 rf_print_unable_to_add_shutdown(const char *file, int line, int rc)
890 {
891 RF_ERRORMSG3("Unable to add to shutdown list file %s line %d rc=%d\n",
892 file, line, rc);
893 }
894