Expand PMF_FN_* macros.
[netbsd-mini2440.git] / sys / dev / raidframe / rf_driver.c
blobedca28fbfdf361baa3fffc9c7a1cbce5e536b9a0
1 /* $NetBSD: rf_driver.c,v 1.121 2009/03/15 17:17:23 cegger Exp $ */
2 /*-
3 * Copyright (c) 1999 The NetBSD Foundation, Inc.
4 * All rights reserved.
6 * This code is derived from software contributed to The NetBSD Foundation
7 * by Greg Oster
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.
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.
32 * Copyright (c) 1995 Carnegie-Mellon University.
33 * All rights reserved.
35 * Author: Mark Holland, Khalil Amiri, Claudson Bornstein, William V. Courtright II,
36 * Robby Findler, Daniel Stodolsky, Rachad Youssef, Jim Zelenka
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.
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.
48 * Carnegie Mellon requests users of this software to return to
50 * Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU
51 * School of Computer Science
52 * Carnegie Mellon University
53 * Pittsburgh PA 15213-3890
55 * any improvements or extensions that they make and grant Carnegie the
56 * rights to redistribute these changes.
59 /******************************************************************************
61 * rf_driver.c -- main setup, teardown, and access routines for the RAID driver
63 * all routines are prefixed with rf_ (raidframe), to avoid conficts.
65 ******************************************************************************/
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: rf_driver.c,v 1.121 2009/03/15 17:17:23 cegger Exp $");
71 #ifdef _KERNEL_OPT
72 #include "opt_raid_diagnostic.h"
73 #endif
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>
82 #include "rf_archs.h"
83 #include "rf_threadstuff.h"
85 #include <sys/errno.h>
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 #include "rf_paritymap.h"
114 #include <sys/buf.h>
116 #ifndef RF_ACCESS_DEBUG
117 #define RF_ACCESS_DEBUG 0
118 #endif
120 /* rad == RF_RaidAccessDesc_t */
121 RF_DECLARE_MUTEX(rf_rad_lock)
122 #define RF_MAX_FREE_RAD 128
123 #define RF_MIN_FREE_RAD 32
125 /* debug variables */
126 char rf_panicbuf[2048]; /* a buffer to hold an error msg when we panic */
128 /* main configuration routines */
129 static int raidframe_booted = 0;
131 static void rf_ConfigureDebug(RF_Config_t * cfgPtr);
132 static void set_debug_option(char *name, long val);
133 static void rf_UnconfigureArray(void);
134 static void rf_ShutdownRDFreeList(void *);
135 static int rf_ConfigureRDFreeList(RF_ShutdownList_t **);
137 RF_DECLARE_MUTEX(rf_printf_mutex) /* debug only: avoids interleaved
138 * printfs by different stripes */
140 #define SIGNAL_QUIESCENT_COND(_raid_) wakeup(&((_raid_)->accesses_suspended))
141 #define WAIT_FOR_QUIESCENCE(_raid_) \
142 ltsleep(&((_raid_)->accesses_suspended), PRIBIO, \
143 "raidframe quiesce", 0, &((_raid_)->access_suspend_mutex))
145 static int configureCount = 0; /* number of active configurations */
146 static int isconfigged = 0; /* is basic raidframe (non per-array)
147 * stuff configured */
148 RF_DECLARE_LKMGR_STATIC_MUTEX(configureMutex) /* used to lock the configuration
149 * stuff */
150 static RF_ShutdownList_t *globalShutdown; /* non array-specific
151 * stuff */
153 static int rf_ConfigureRDFreeList(RF_ShutdownList_t ** listp);
154 static int rf_AllocEmergBuffers(RF_Raid_t *);
155 static void rf_FreeEmergBuffers(RF_Raid_t *);
157 /* called at system boot time */
159 rf_BootRaidframe(void)
162 if (raidframe_booted)
163 return (EBUSY);
164 raidframe_booted = 1;
165 mutex_init(&configureMutex, MUTEX_DEFAULT, IPL_NONE);
166 configureCount = 0;
167 isconfigged = 0;
168 globalShutdown = NULL;
169 return (0);
173 * Called whenever an array is shutdown
175 static void
176 rf_UnconfigureArray(void)
179 RF_LOCK_LKMGR_MUTEX(configureMutex);
180 if (--configureCount == 0) { /* if no active configurations, shut
181 * everything down */
182 isconfigged = 0;
183 rf_ShutdownList(&globalShutdown);
186 * We must wait until now, because the AllocList module
187 * uses the DebugMem module.
189 #if RF_DEBUG_MEM
190 if (rf_memDebug)
191 rf_print_unfreed();
192 #endif
194 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
198 * Called to shut down an array.
201 rf_Shutdown(RF_Raid_t *raidPtr)
204 if (!raidPtr->valid) {
205 RF_ERRORMSG("Attempt to shut down unconfigured RAIDframe driver. Aborting shutdown\n");
206 return (EINVAL);
209 * wait for outstanding IOs to land
210 * As described in rf_raid.h, we use the rad_freelist lock
211 * to protect the per-array info about outstanding descs
212 * since we need to do freelist locking anyway, and this
213 * cuts down on the amount of serialization we've got going
214 * on.
216 RF_LOCK_MUTEX(rf_rad_lock);
217 if (raidPtr->waitShutdown) {
218 RF_UNLOCK_MUTEX(rf_rad_lock);
219 return (EBUSY);
221 raidPtr->waitShutdown = 1;
222 while (raidPtr->nAccOutstanding) {
223 RF_WAIT_COND(raidPtr->outstandingCond, rf_rad_lock);
225 RF_UNLOCK_MUTEX(rf_rad_lock);
227 /* Wait for any parity re-writes to stop... */
228 while (raidPtr->parity_rewrite_in_progress) {
229 printf("raid%d: Waiting for parity re-write to exit...\n",
230 raidPtr->raidid);
231 tsleep(&raidPtr->parity_rewrite_in_progress, PRIBIO,
232 "rfprwshutdown", 0);
235 /* Wait for any reconstruction to stop... */
236 while (raidPtr->reconInProgress) {
237 printf("raid%d: Waiting for reconstruction to stop...\n",
238 raidPtr->raidid);
239 tsleep(&raidPtr->waitForReconCond, PRIBIO,
240 "rfreshutdown",0);
243 raidPtr->valid = 0;
245 if (raidPtr->parity_map != NULL)
246 rf_paritymap_detach(raidPtr);
248 rf_update_component_labels(raidPtr, RF_FINAL_COMPONENT_UPDATE);
250 rf_UnconfigureVnodes(raidPtr);
252 rf_FreeEmergBuffers(raidPtr);
254 rf_ShutdownList(&raidPtr->shutdownList);
256 rf_UnconfigureArray();
258 return (0);
262 #define DO_INIT_CONFIGURE(f) { \
263 rc = f (&globalShutdown); \
264 if (rc) { \
265 RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
266 rf_ShutdownList(&globalShutdown); \
267 configureCount--; \
268 RF_UNLOCK_LKMGR_MUTEX(configureMutex); \
269 return(rc); \
273 #define DO_RAID_FAIL() { \
274 rf_UnconfigureVnodes(raidPtr); \
275 rf_FreeEmergBuffers(raidPtr); \
276 rf_ShutdownList(&raidPtr->shutdownList); \
277 rf_UnconfigureArray(); \
280 #define DO_RAID_INIT_CONFIGURE(f) { \
281 rc = f (&raidPtr->shutdownList, raidPtr, cfgPtr); \
282 if (rc) { \
283 RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
284 DO_RAID_FAIL(); \
285 return(rc); \
289 #define DO_RAID_MUTEX(_m_) { \
290 rf_mutex_init((_m_)); \
294 rf_Configure(RF_Raid_t *raidPtr, RF_Config_t *cfgPtr, RF_AutoConfig_t *ac)
296 RF_RowCol_t col;
297 int rc;
299 RF_LOCK_LKMGR_MUTEX(configureMutex);
300 configureCount++;
301 if (isconfigged == 0) {
302 rf_mutex_init(&rf_printf_mutex);
304 /* initialize globals */
306 DO_INIT_CONFIGURE(rf_ConfigureAllocList);
309 * Yes, this does make debugging general to the whole
310 * system instead of being array specific. Bummer, drag.
312 rf_ConfigureDebug(cfgPtr);
313 DO_INIT_CONFIGURE(rf_ConfigureDebugMem);
314 #if RF_ACC_TRACE > 0
315 DO_INIT_CONFIGURE(rf_ConfigureAccessTrace);
316 #endif
317 DO_INIT_CONFIGURE(rf_ConfigureMapModule);
318 DO_INIT_CONFIGURE(rf_ConfigureReconEvent);
319 DO_INIT_CONFIGURE(rf_ConfigureCallback);
320 DO_INIT_CONFIGURE(rf_ConfigureRDFreeList);
321 DO_INIT_CONFIGURE(rf_ConfigureNWayXor);
322 DO_INIT_CONFIGURE(rf_ConfigureStripeLockFreeList);
323 DO_INIT_CONFIGURE(rf_ConfigureMCPair);
324 DO_INIT_CONFIGURE(rf_ConfigureDAGs);
325 DO_INIT_CONFIGURE(rf_ConfigureDAGFuncs);
326 DO_INIT_CONFIGURE(rf_ConfigureReconstruction);
327 DO_INIT_CONFIGURE(rf_ConfigureCopyback);
328 DO_INIT_CONFIGURE(rf_ConfigureDiskQueueSystem);
329 DO_INIT_CONFIGURE(rf_ConfigurePSStatus);
330 isconfigged = 1;
332 RF_UNLOCK_LKMGR_MUTEX(configureMutex);
334 DO_RAID_MUTEX(&raidPtr->mutex);
335 /* set up the cleanup list. Do this after ConfigureDebug so that
336 * value of memDebug will be set */
338 rf_MakeAllocList(raidPtr->cleanupList);
339 if (raidPtr->cleanupList == NULL) {
340 DO_RAID_FAIL();
341 return (ENOMEM);
343 rf_ShutdownCreate(&raidPtr->shutdownList,
344 (void (*) (void *)) rf_FreeAllocList,
345 raidPtr->cleanupList);
347 raidPtr->numCol = cfgPtr->numCol;
348 raidPtr->numSpare = cfgPtr->numSpare;
350 raidPtr->status = rf_rs_optimal;
351 raidPtr->reconControl = NULL;
353 TAILQ_INIT(&(raidPtr->iodone));
354 simple_lock_init(&(raidPtr->iodone_lock));
356 DO_RAID_INIT_CONFIGURE(rf_ConfigureEngine);
357 DO_RAID_INIT_CONFIGURE(rf_ConfigureStripeLocks);
359 raidPtr->outstandingCond = 0;
361 raidPtr->nAccOutstanding = 0;
362 raidPtr->waitShutdown = 0;
364 DO_RAID_MUTEX(&raidPtr->access_suspend_mutex);
366 raidPtr->waitForReconCond = 0;
368 if (ac!=NULL) {
369 /* We have an AutoConfig structure.. Don't do the
370 normal disk configuration... call the auto config
371 stuff */
372 rf_AutoConfigureDisks(raidPtr, cfgPtr, ac);
373 } else {
374 DO_RAID_INIT_CONFIGURE(rf_ConfigureDisks);
375 DO_RAID_INIT_CONFIGURE(rf_ConfigureSpareDisks);
377 /* do this after ConfigureDisks & ConfigureSpareDisks to be sure dev
378 * no. is set */
379 DO_RAID_INIT_CONFIGURE(rf_ConfigureDiskQueues);
381 DO_RAID_INIT_CONFIGURE(rf_ConfigureLayout);
383 /* Initialize per-RAID PSS bits */
384 rf_InitPSStatus(raidPtr);
386 #if RF_INCLUDE_CHAINDECLUSTER > 0
387 for (col = 0; col < raidPtr->numCol; col++) {
389 * XXX better distribution
391 raidPtr->hist_diskreq[col] = 0;
393 #endif
394 raidPtr->numNewFailures = 0;
395 raidPtr->copyback_in_progress = 0;
396 raidPtr->parity_rewrite_in_progress = 0;
397 raidPtr->adding_hot_spare = 0;
398 raidPtr->recon_in_progress = 0;
399 raidPtr->maxOutstanding = cfgPtr->maxOutstandingDiskReqs;
401 /* autoconfigure and root_partition will actually get filled in
402 after the config is done */
403 raidPtr->autoconfigure = 0;
404 raidPtr->root_partition = 0;
405 raidPtr->last_unit = raidPtr->raidid;
406 raidPtr->config_order = 0;
408 if (rf_keepAccTotals) {
409 raidPtr->keep_acc_totals = 1;
412 /* Allocate a bunch of buffers to be used in low-memory conditions */
413 raidPtr->iobuf = NULL;
415 rc = rf_AllocEmergBuffers(raidPtr);
416 if (rc) {
417 printf("raid%d: Unable to allocate emergency buffers.\n",
418 raidPtr->raidid);
419 DO_RAID_FAIL();
420 return(rc);
423 /* Set up parity map stuff, if applicable. */
424 #ifndef RF_NO_PARITY_MAP
425 rf_paritymap_attach(raidPtr, cfgPtr->force);
426 #endif
428 raidPtr->valid = 1;
430 printf("raid%d: %s\n", raidPtr->raidid,
431 raidPtr->Layout.map->configName);
432 printf("raid%d: Components:", raidPtr->raidid);
434 for (col = 0; col < raidPtr->numCol; col++) {
435 printf(" %s", raidPtr->Disks[col].devname);
436 if (RF_DEAD_DISK(raidPtr->Disks[col].status)) {
437 printf("[**FAILED**]");
440 printf("\n");
441 printf("raid%d: Total Sectors: %" PRIu64 " (%" PRIu64 " MB)\n",
442 raidPtr->raidid,
443 raidPtr->totalSectors,
444 (raidPtr->totalSectors / 1024 *
445 (1 << raidPtr->logBytesPerSector) / 1024));
447 return (0);
453 Routines to allocate and free the "emergency buffers" for a given
454 RAID set. These emergency buffers will be used when the kernel runs
455 out of kernel memory.
459 static int
460 rf_AllocEmergBuffers(RF_Raid_t *raidPtr)
462 void *tmpbuf;
463 RF_VoidPointerListElem_t *vple;
464 int i;
466 /* XXX next line needs tuning... */
467 raidPtr->numEmergencyBuffers = 10 * raidPtr->numCol;
468 #if DEBUG
469 printf("raid%d: allocating %d buffers of %d bytes.\n",
470 raidPtr->raidid,
471 raidPtr->numEmergencyBuffers,
472 (int)(raidPtr->Layout.sectorsPerStripeUnit <<
473 raidPtr->logBytesPerSector));
474 #endif
475 for (i = 0; i < raidPtr->numEmergencyBuffers; i++) {
476 tmpbuf = malloc( raidPtr->Layout.sectorsPerStripeUnit <<
477 raidPtr->logBytesPerSector,
478 M_RAIDFRAME, M_WAITOK);
479 if (tmpbuf) {
480 vple = rf_AllocVPListElem();
481 vple->p= tmpbuf;
482 vple->next = raidPtr->iobuf;
483 raidPtr->iobuf = vple;
484 raidPtr->iobuf_count++;
485 } else {
486 printf("raid%d: failed to allocate emergency buffer!\n",
487 raidPtr->raidid);
488 return 1;
492 /* XXX next line needs tuning too... */
493 raidPtr->numEmergencyStripeBuffers = 10;
494 for (i = 0; i < raidPtr->numEmergencyStripeBuffers; i++) {
495 tmpbuf = malloc( raidPtr->numCol * (raidPtr->Layout.sectorsPerStripeUnit <<
496 raidPtr->logBytesPerSector),
497 M_RAIDFRAME, M_WAITOK);
498 if (tmpbuf) {
499 vple = rf_AllocVPListElem();
500 vple->p= tmpbuf;
501 vple->next = raidPtr->stripebuf;
502 raidPtr->stripebuf = vple;
503 raidPtr->stripebuf_count++;
504 } else {
505 printf("raid%d: failed to allocate emergency stripe buffer!\n",
506 raidPtr->raidid);
507 return 1;
511 return (0);
514 static void
515 rf_FreeEmergBuffers(RF_Raid_t *raidPtr)
517 RF_VoidPointerListElem_t *tmp;
519 /* Free the emergency IO buffers */
520 while (raidPtr->iobuf != NULL) {
521 tmp = raidPtr->iobuf;
522 raidPtr->iobuf = raidPtr->iobuf->next;
523 free(tmp->p, M_RAIDFRAME);
524 rf_FreeVPListElem(tmp);
527 /* Free the emergency stripe buffers */
528 while (raidPtr->stripebuf != NULL) {
529 tmp = raidPtr->stripebuf;
530 raidPtr->stripebuf = raidPtr->stripebuf->next;
531 free(tmp->p, M_RAIDFRAME);
532 rf_FreeVPListElem(tmp);
537 static void
538 rf_ShutdownRDFreeList(void *ignored)
540 pool_destroy(&rf_pools.rad);
543 static int
544 rf_ConfigureRDFreeList(RF_ShutdownList_t **listp)
547 rf_pool_init(&rf_pools.rad, sizeof(RF_RaidAccessDesc_t),
548 "rf_rad_pl", RF_MIN_FREE_RAD, RF_MAX_FREE_RAD);
549 rf_ShutdownCreate(listp, rf_ShutdownRDFreeList, NULL);
550 simple_lock_init(&rf_rad_lock);
551 return (0);
554 RF_RaidAccessDesc_t *
555 rf_AllocRaidAccDesc(RF_Raid_t *raidPtr, RF_IoType_t type,
556 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
557 void *bufPtr, void *bp, RF_RaidAccessFlags_t flags,
558 const RF_AccessState_t *states)
560 RF_RaidAccessDesc_t *desc;
562 desc = pool_get(&rf_pools.rad, PR_WAITOK);
564 RF_LOCK_MUTEX(rf_rad_lock);
565 if (raidPtr->waitShutdown) {
567 * Actually, we're shutting the array down. Free the desc
568 * and return NULL.
571 RF_UNLOCK_MUTEX(rf_rad_lock);
572 pool_put(&rf_pools.rad, desc);
573 return (NULL);
575 raidPtr->nAccOutstanding++;
577 RF_UNLOCK_MUTEX(rf_rad_lock);
579 desc->raidPtr = (void *) raidPtr;
580 desc->type = type;
581 desc->raidAddress = raidAddress;
582 desc->numBlocks = numBlocks;
583 desc->bufPtr = bufPtr;
584 desc->bp = bp;
585 desc->flags = flags;
586 desc->states = states;
587 desc->state = 0;
588 desc->dagList = NULL;
590 desc->status = 0;
591 desc->numRetries = 0;
592 #if RF_ACC_TRACE > 0
593 memset((char *) &desc->tracerec, 0, sizeof(RF_AccTraceEntry_t));
594 #endif
595 desc->callbackFunc = NULL;
596 desc->callbackArg = NULL;
597 desc->next = NULL;
598 desc->iobufs = NULL;
599 desc->stripebufs = NULL;
601 return (desc);
604 void
605 rf_FreeRaidAccDesc(RF_RaidAccessDesc_t *desc)
607 RF_Raid_t *raidPtr = desc->raidPtr;
608 RF_DagList_t *dagList, *temp;
609 RF_VoidPointerListElem_t *tmp;
611 RF_ASSERT(desc);
613 /* Cleanup the dagList(s) */
614 dagList = desc->dagList;
615 while(dagList != NULL) {
616 temp = dagList;
617 dagList = dagList->next;
618 rf_FreeDAGList(temp);
621 while (desc->iobufs) {
622 tmp = desc->iobufs;
623 desc->iobufs = desc->iobufs->next;
624 rf_FreeIOBuffer(raidPtr, tmp);
627 while (desc->stripebufs) {
628 tmp = desc->stripebufs;
629 desc->stripebufs = desc->stripebufs->next;
630 rf_FreeStripeBuffer(raidPtr, tmp);
633 pool_put(&rf_pools.rad, desc);
634 RF_LOCK_MUTEX(rf_rad_lock);
635 raidPtr->nAccOutstanding--;
636 if (raidPtr->waitShutdown) {
637 RF_SIGNAL_COND(raidPtr->outstandingCond);
639 RF_UNLOCK_MUTEX(rf_rad_lock);
641 /*********************************************************************
642 * Main routine for performing an access.
643 * Accesses are retried until a DAG can not be selected. This occurs
644 * when either the DAG library is incomplete or there are too many
645 * failures in a parity group.
647 * type should be read or write async_flag should be RF_TRUE or
648 * RF_FALSE bp_in is a buf pointer. void *to facilitate ignoring it
649 * outside the kernel
650 ********************************************************************/
652 rf_DoAccess(RF_Raid_t * raidPtr, RF_IoType_t type, int async_flag,
653 RF_RaidAddr_t raidAddress, RF_SectorCount_t numBlocks,
654 void *bufPtr, struct buf *bp, RF_RaidAccessFlags_t flags)
656 RF_RaidAccessDesc_t *desc;
657 void *lbufPtr = bufPtr;
659 raidAddress += rf_raidSectorOffset;
661 #if RF_ACCESS_DEBUG
662 if (rf_accessDebug) {
664 printf("logBytes is: %d %d %d\n", raidPtr->raidid,
665 raidPtr->logBytesPerSector,
666 (int) rf_RaidAddressToByte(raidPtr, numBlocks));
667 printf("raid%d: %s raidAddr %d (stripeid %d-%d) numBlocks %d (%d bytes) buf 0x%lx\n", raidPtr->raidid,
668 (type == RF_IO_TYPE_READ) ? "READ" : "WRITE", (int) raidAddress,
669 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress),
670 (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress + numBlocks - 1),
671 (int) numBlocks,
672 (int) rf_RaidAddressToByte(raidPtr, numBlocks),
673 (long) bufPtr);
675 #endif
677 desc = rf_AllocRaidAccDesc(raidPtr, type, raidAddress,
678 numBlocks, lbufPtr, bp, flags, raidPtr->Layout.map->states);
680 if (desc == NULL) {
681 return (ENOMEM);
683 #if RF_ACC_TRACE > 0
684 RF_ETIMER_START(desc->tracerec.tot_timer);
685 #endif
686 desc->async_flag = async_flag;
688 if (raidPtr->parity_map != NULL &&
689 type == RF_IO_TYPE_WRITE)
690 rf_paritymap_begin(raidPtr->parity_map, raidAddress,
691 numBlocks);
693 rf_ContinueRaidAccess(desc);
695 return (0);
697 #if 0
698 /* force the array into reconfigured mode without doing reconstruction */
700 rf_SetReconfiguredMode(RF_Raid_t *raidPtr, int col)
702 if (!(raidPtr->Layout.map->flags & RF_DISTRIBUTE_SPARE)) {
703 printf("Can't set reconfigured mode in dedicated-spare array\n");
704 RF_PANIC();
706 RF_LOCK_MUTEX(raidPtr->mutex);
707 raidPtr->numFailures++;
708 raidPtr->Disks[col].status = rf_ds_dist_spared;
709 raidPtr->status = rf_rs_reconfigured;
710 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
711 /* install spare table only if declustering + distributed sparing
712 * architecture. */
713 if (raidPtr->Layout.map->flags & RF_BD_DECLUSTERED)
714 rf_InstallSpareTable(raidPtr, col);
715 RF_UNLOCK_MUTEX(raidPtr->mutex);
716 return (0);
718 #endif
721 rf_FailDisk(RF_Raid_t *raidPtr, int fcol, int initRecon)
724 /* need to suspend IO's here -- if there are DAGs in flight
725 and we pull the rug out from under ci_vp, Bad Things
726 can happen. */
728 rf_SuspendNewRequestsAndWait(raidPtr);
730 RF_LOCK_MUTEX(raidPtr->mutex);
731 if (raidPtr->Disks[fcol].status != rf_ds_failed) {
732 /* must be failing something that is valid, or else it's
733 already marked as failed (in which case we don't
734 want to mark it failed again!) */
735 raidPtr->numFailures++;
736 raidPtr->Disks[fcol].status = rf_ds_failed;
737 raidPtr->status = rf_rs_degraded;
739 RF_UNLOCK_MUTEX(raidPtr->mutex);
741 rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
743 /* Close the component, so that it's not "locked" if someone
744 else want's to use it! */
746 rf_close_component(raidPtr, raidPtr->raid_cinfo[fcol].ci_vp,
747 raidPtr->Disks[fcol].auto_configured);
749 RF_LOCK_MUTEX(raidPtr->mutex);
750 raidPtr->raid_cinfo[fcol].ci_vp = NULL;
752 /* Need to mark the component as not being auto_configured
753 (in case it was previously). */
755 raidPtr->Disks[fcol].auto_configured = 0;
756 RF_UNLOCK_MUTEX(raidPtr->mutex);
757 /* now we can allow IO to continue -- we'll be suspending it
758 again in rf_ReconstructFailedDisk() if we have to.. */
760 rf_ResumeNewRequests(raidPtr);
762 if (initRecon)
763 rf_ReconstructFailedDisk(raidPtr, fcol);
764 return (0);
766 /* releases a thread that is waiting for the array to become quiesced.
767 * access_suspend_mutex should be locked upon calling this
769 void
770 rf_SignalQuiescenceLock(RF_Raid_t *raidPtr)
772 #if RF_DEBUG_QUIESCE
773 if (rf_quiesceDebug) {
774 printf("raid%d: Signalling quiescence lock\n",
775 raidPtr->raidid);
777 #endif
778 raidPtr->access_suspend_release = 1;
780 if (raidPtr->waiting_for_quiescence) {
781 SIGNAL_QUIESCENT_COND(raidPtr);
784 /* suspends all new requests to the array. No effect on accesses that are in flight. */
786 rf_SuspendNewRequestsAndWait(RF_Raid_t *raidPtr)
788 #if RF_DEBUG_QUIESCE
789 if (rf_quiesceDebug)
790 printf("raid%d: Suspending new reqs\n", raidPtr->raidid);
791 #endif
792 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
793 raidPtr->accesses_suspended++;
794 raidPtr->waiting_for_quiescence = (raidPtr->accs_in_flight == 0) ? 0 : 1;
796 if (raidPtr->waiting_for_quiescence) {
797 raidPtr->access_suspend_release = 0;
798 while (!raidPtr->access_suspend_release) {
799 #if RF_DEBUG_QUIESCE
800 printf("raid%d: Suspending: Waiting for Quiescence\n",
801 raidPtr->raidid);
802 #endif
803 WAIT_FOR_QUIESCENCE(raidPtr);
804 raidPtr->waiting_for_quiescence = 0;
807 #if RF_DEBUG_QUIESCE
808 printf("raid%d: Quiescence reached..\n", raidPtr->raidid);
809 #endif
811 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
812 return (raidPtr->waiting_for_quiescence);
814 /* wake up everyone waiting for quiescence to be released */
815 void
816 rf_ResumeNewRequests(RF_Raid_t *raidPtr)
818 RF_CallbackDesc_t *t, *cb;
820 #if RF_DEBUG_QUIESCE
821 if (rf_quiesceDebug)
822 printf("raid%d: Resuming new requests\n", raidPtr->raidid);
823 #endif
825 RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
826 raidPtr->accesses_suspended--;
827 if (raidPtr->accesses_suspended == 0)
828 cb = raidPtr->quiesce_wait_list;
829 else
830 cb = NULL;
831 raidPtr->quiesce_wait_list = NULL;
832 RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
834 while (cb) {
835 t = cb;
836 cb = cb->next;
837 (t->callbackFunc) (t->callbackArg);
838 rf_FreeCallbackDesc(t);
841 /*****************************************************************************************
843 * debug routines
845 ****************************************************************************************/
847 static void
848 set_debug_option(char *name, long val)
850 RF_DebugName_t *p;
852 for (p = rf_debugNames; p->name; p++) {
853 if (!strcmp(p->name, name)) {
854 *(p->ptr) = val;
855 printf("[Set debug variable %s to %ld]\n", name, val);
856 return;
859 RF_ERRORMSG1("Unknown debug string \"%s\"\n", name);
863 /* would like to use sscanf here, but apparently not available in kernel */
864 /*ARGSUSED*/
865 static void
866 rf_ConfigureDebug(RF_Config_t *cfgPtr)
868 char *val_p, *name_p, *white_p;
869 long val;
870 int i;
872 rf_ResetDebugOptions();
873 for (i = 0; cfgPtr->debugVars[i][0] && i < RF_MAXDBGV; i++) {
874 name_p = rf_find_non_white(&cfgPtr->debugVars[i][0]);
875 white_p = rf_find_white(name_p); /* skip to start of 2nd
876 * word */
877 val_p = rf_find_non_white(white_p);
878 if (*val_p == '0' && *(val_p + 1) == 'x')
879 val = rf_htoi(val_p + 2);
880 else
881 val = rf_atoi(val_p);
882 *white_p = '\0';
883 set_debug_option(name_p, val);
887 void
888 rf_print_panic_message(int line, const char *file)
890 snprintf(rf_panicbuf, sizeof(rf_panicbuf),
891 "raidframe error at line %d file %s", line, file);
894 #ifdef RAID_DIAGNOSTIC
895 void
896 rf_print_assert_panic_message(int line, const char *file, const char *condition)
898 snprintf(rf_panicbuf, sizeof(rf_panicbuf),
899 "raidframe error at line %d file %s (failed asserting %s)\n",
900 line, file, condition);
902 #endif
904 void
905 rf_print_unable_to_init_mutex(const char *file, int line, int rc)
907 RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n",
908 file, line, rc);
911 void
912 rf_print_unable_to_add_shutdown(const char *file, int line, int rc)
914 RF_ERRORMSG3("Unable to add to shutdown list file %s line %d rc=%d\n",
915 file, line, rc);