rf_driver.c revision 1.44 1 1.44 lukem /* $NetBSD: rf_driver.c,v 1.44 2001/11/13 07:11:13 lukem Exp $ */
2 1.9 oster /*-
3 1.9 oster * Copyright (c) 1999 The NetBSD Foundation, Inc.
4 1.9 oster * All rights reserved.
5 1.9 oster *
6 1.9 oster * This code is derived from software contributed to The NetBSD Foundation
7 1.9 oster * by Greg Oster
8 1.9 oster *
9 1.9 oster * Redistribution and use in source and binary forms, with or without
10 1.9 oster * modification, are permitted provided that the following conditions
11 1.9 oster * are met:
12 1.9 oster * 1. Redistributions of source code must retain the above copyright
13 1.9 oster * notice, this list of conditions and the following disclaimer.
14 1.9 oster * 2. Redistributions in binary form must reproduce the above copyright
15 1.9 oster * notice, this list of conditions and the following disclaimer in the
16 1.9 oster * documentation and/or other materials provided with the distribution.
17 1.9 oster * 3. All advertising materials mentioning features or use of this software
18 1.9 oster * must display the following acknowledgement:
19 1.9 oster * This product includes software developed by the NetBSD
20 1.9 oster * Foundation, Inc. and its contributors.
21 1.9 oster * 4. Neither the name of The NetBSD Foundation nor the names of its
22 1.9 oster * contributors may be used to endorse or promote products derived
23 1.9 oster * from this software without specific prior written permission.
24 1.9 oster *
25 1.9 oster * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
26 1.9 oster * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
27 1.9 oster * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28 1.9 oster * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
29 1.9 oster * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 1.9 oster * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 1.9 oster * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 1.9 oster * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 1.9 oster * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 1.9 oster * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 1.9 oster * POSSIBILITY OF SUCH DAMAGE.
36 1.9 oster */
37 1.9 oster
38 1.1 oster /*
39 1.1 oster * Copyright (c) 1995 Carnegie-Mellon University.
40 1.1 oster * All rights reserved.
41 1.1 oster *
42 1.1 oster * Author: Mark Holland, Khalil Amiri, Claudson Bornstein, William V. Courtright II,
43 1.1 oster * Robby Findler, Daniel Stodolsky, Rachad Youssef, Jim Zelenka
44 1.1 oster *
45 1.1 oster * Permission to use, copy, modify and distribute this software and
46 1.1 oster * its documentation is hereby granted, provided that both the copyright
47 1.1 oster * notice and this permission notice appear in all copies of the
48 1.1 oster * software, derivative works or modified versions, and any portions
49 1.1 oster * thereof, and that both notices appear in supporting documentation.
50 1.1 oster *
51 1.1 oster * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
52 1.1 oster * CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND
53 1.1 oster * FOR ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
54 1.1 oster *
55 1.1 oster * Carnegie Mellon requests users of this software to return to
56 1.1 oster *
57 1.1 oster * Software Distribution Coordinator or Software.Distribution (at) CS.CMU.EDU
58 1.1 oster * School of Computer Science
59 1.1 oster * Carnegie Mellon University
60 1.1 oster * Pittsburgh PA 15213-3890
61 1.1 oster *
62 1.1 oster * any improvements or extensions that they make and grant Carnegie the
63 1.1 oster * rights to redistribute these changes.
64 1.1 oster */
65 1.1 oster
66 1.1 oster /******************************************************************************
67 1.1 oster *
68 1.1 oster * rf_driver.c -- main setup, teardown, and access routines for the RAID driver
69 1.1 oster *
70 1.1 oster * all routines are prefixed with rf_ (raidframe), to avoid conficts.
71 1.1 oster *
72 1.1 oster ******************************************************************************/
73 1.1 oster
74 1.44 lukem
75 1.44 lukem #include <sys/cdefs.h>
76 1.44 lukem __KERNEL_RCSID(0, "$NetBSD: rf_driver.c,v 1.44 2001/11/13 07:11:13 lukem Exp $");
77 1.1 oster
78 1.1 oster #include <sys/types.h>
79 1.1 oster #include <sys/param.h>
80 1.1 oster #include <sys/systm.h>
81 1.1 oster #include <sys/ioctl.h>
82 1.1 oster #include <sys/fcntl.h>
83 1.1 oster #include <sys/vnode.h>
84 1.1 oster
85 1.1 oster
86 1.1 oster #include "rf_archs.h"
87 1.1 oster #include "rf_threadstuff.h"
88 1.1 oster
89 1.1 oster #include <sys/errno.h>
90 1.1 oster
91 1.1 oster #include "rf_raid.h"
92 1.1 oster #include "rf_dag.h"
93 1.1 oster #include "rf_aselect.h"
94 1.1 oster #include "rf_diskqueue.h"
95 1.1 oster #include "rf_parityscan.h"
96 1.1 oster #include "rf_alloclist.h"
97 1.1 oster #include "rf_dagutils.h"
98 1.1 oster #include "rf_utils.h"
99 1.1 oster #include "rf_etimer.h"
100 1.1 oster #include "rf_acctrace.h"
101 1.1 oster #include "rf_general.h"
102 1.1 oster #include "rf_desc.h"
103 1.1 oster #include "rf_states.h"
104 1.1 oster #include "rf_freelist.h"
105 1.1 oster #include "rf_decluster.h"
106 1.1 oster #include "rf_map.h"
107 1.1 oster #include "rf_revent.h"
108 1.1 oster #include "rf_callback.h"
109 1.1 oster #include "rf_engine.h"
110 1.1 oster #include "rf_memchunk.h"
111 1.1 oster #include "rf_mcpair.h"
112 1.1 oster #include "rf_nwayxor.h"
113 1.1 oster #include "rf_debugprint.h"
114 1.1 oster #include "rf_copyback.h"
115 1.1 oster #include "rf_driver.h"
116 1.1 oster #include "rf_options.h"
117 1.1 oster #include "rf_shutdown.h"
118 1.24 oster #include "rf_kintf.h"
119 1.1 oster
120 1.1 oster #include <sys/buf.h>
121 1.1 oster
122 1.1 oster /* rad == RF_RaidAccessDesc_t */
123 1.1 oster static RF_FreeList_t *rf_rad_freelist;
124 1.1 oster #define RF_MAX_FREE_RAD 128
125 1.1 oster #define RF_RAD_INC 16
126 1.1 oster #define RF_RAD_INITIAL 32
127 1.1 oster
128 1.1 oster /* debug variables */
129 1.6 oster char rf_panicbuf[2048]; /* a buffer to hold an error msg when we panic */
130 1.1 oster
131 1.1 oster /* main configuration routines */
132 1.1 oster static int raidframe_booted = 0;
133 1.1 oster
134 1.6 oster static void rf_ConfigureDebug(RF_Config_t * cfgPtr);
135 1.1 oster static void set_debug_option(char *name, long val);
136 1.1 oster static void rf_UnconfigureArray(void);
137 1.1 oster static int init_rad(RF_RaidAccessDesc_t *);
138 1.1 oster static void clean_rad(RF_RaidAccessDesc_t *);
139 1.1 oster static void rf_ShutdownRDFreeList(void *);
140 1.1 oster static int rf_ConfigureRDFreeList(RF_ShutdownList_t **);
141 1.1 oster
142 1.6 oster RF_DECLARE_MUTEX(rf_printf_mutex) /* debug only: avoids interleaved
143 1.6 oster * printfs by different stripes */
144 1.1 oster
145 1.1 oster #define SIGNAL_QUIESCENT_COND(_raid_) wakeup(&((_raid_)->accesses_suspended))
146 1.1 oster #define WAIT_FOR_QUIESCENCE(_raid_) \
147 1.38 oster ltsleep(&((_raid_)->accesses_suspended), PRIBIO, \
148 1.38 oster "raidframe quiesce", 0, &((_raid_)->access_suspend_mutex))
149 1.1 oster
150 1.20 oster #define IO_BUF_ERR(bp, err) { \
151 1.1 oster bp->b_flags |= B_ERROR; \
152 1.1 oster bp->b_resid = bp->b_bcount; \
153 1.1 oster bp->b_error = err; \
154 1.1 oster biodone(bp); \
155 1.1 oster }
156 1.1 oster
157 1.9 oster static int configureCount = 0; /* number of active configurations */
158 1.9 oster static int isconfigged = 0; /* is basic raidframe (non per-array)
159 1.9 oster * stuff configged */
160 1.6 oster RF_DECLARE_STATIC_MUTEX(configureMutex) /* used to lock the configuration
161 1.6 oster * stuff */
162 1.9 oster static RF_ShutdownList_t *globalShutdown; /* non array-specific
163 1.9 oster * stuff */
164 1.1 oster
165 1.9 oster static int rf_ConfigureRDFreeList(RF_ShutdownList_t ** listp);
166 1.1 oster
167 1.1 oster /* called at system boot time */
168 1.7 oster int
169 1.7 oster rf_BootRaidframe()
170 1.1 oster {
171 1.6 oster int rc;
172 1.1 oster
173 1.6 oster if (raidframe_booted)
174 1.6 oster return (EBUSY);
175 1.6 oster raidframe_booted = 1;
176 1.1 oster
177 1.6 oster rc = rf_mutex_init(&configureMutex);
178 1.6 oster if (rc) {
179 1.6 oster RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n", __FILE__,
180 1.6 oster __LINE__, rc);
181 1.6 oster RF_PANIC();
182 1.6 oster }
183 1.6 oster configureCount = 0;
184 1.6 oster isconfigged = 0;
185 1.6 oster globalShutdown = NULL;
186 1.6 oster return (0);
187 1.1 oster }
188 1.1 oster /*
189 1.1 oster * This function is really just for debugging user-level stuff: it
190 1.1 oster * frees up all memory, other RAIDframe resources which might otherwise
191 1.1 oster * be kept around. This is used with systems like "sentinel" to detect
192 1.1 oster * memory leaks.
193 1.1 oster */
194 1.6 oster int
195 1.6 oster rf_UnbootRaidframe()
196 1.1 oster {
197 1.6 oster int rc;
198 1.1 oster
199 1.1 oster RF_LOCK_MUTEX(configureMutex);
200 1.1 oster if (configureCount) {
201 1.1 oster RF_UNLOCK_MUTEX(configureMutex);
202 1.6 oster return (EBUSY);
203 1.1 oster }
204 1.1 oster raidframe_booted = 0;
205 1.1 oster RF_UNLOCK_MUTEX(configureMutex);
206 1.1 oster rc = rf_mutex_destroy(&configureMutex);
207 1.1 oster if (rc) {
208 1.1 oster RF_ERRORMSG3("Unable to destroy mutex file %s line %d rc=%d\n", __FILE__,
209 1.6 oster __LINE__, rc);
210 1.1 oster RF_PANIC();
211 1.1 oster }
212 1.6 oster return (0);
213 1.1 oster }
214 1.1 oster /*
215 1.1 oster * Called whenever an array is shutdown
216 1.1 oster */
217 1.6 oster static void
218 1.6 oster rf_UnconfigureArray()
219 1.1 oster {
220 1.6 oster int rc;
221 1.1 oster
222 1.6 oster RF_LOCK_MUTEX(configureMutex);
223 1.6 oster if (--configureCount == 0) { /* if no active configurations, shut
224 1.6 oster * everything down */
225 1.6 oster isconfigged = 0;
226 1.6 oster
227 1.6 oster rc = rf_ShutdownList(&globalShutdown);
228 1.6 oster if (rc) {
229 1.6 oster RF_ERRORMSG1("RAIDFRAME: unable to do global shutdown, rc=%d\n", rc);
230 1.6 oster }
231 1.6 oster
232 1.6 oster /*
233 1.6 oster * We must wait until now, because the AllocList module
234 1.6 oster * uses the DebugMem module.
235 1.6 oster */
236 1.6 oster if (rf_memDebug)
237 1.6 oster rf_print_unfreed();
238 1.6 oster }
239 1.6 oster RF_UNLOCK_MUTEX(configureMutex);
240 1.9 oster }
241 1.9 oster
242 1.1 oster /*
243 1.1 oster * Called to shut down an array.
244 1.1 oster */
245 1.6 oster int
246 1.6 oster rf_Shutdown(raidPtr)
247 1.6 oster RF_Raid_t *raidPtr;
248 1.1 oster {
249 1.1 oster
250 1.6 oster if (!raidPtr->valid) {
251 1.6 oster RF_ERRORMSG("Attempt to shut down unconfigured RAIDframe driver. Aborting shutdown\n");
252 1.6 oster return (EINVAL);
253 1.6 oster }
254 1.6 oster /*
255 1.6 oster * wait for outstanding IOs to land
256 1.6 oster * As described in rf_raid.h, we use the rad_freelist lock
257 1.6 oster * to protect the per-array info about outstanding descs
258 1.6 oster * since we need to do freelist locking anyway, and this
259 1.6 oster * cuts down on the amount of serialization we've got going
260 1.6 oster * on.
261 1.6 oster */
262 1.6 oster RF_FREELIST_DO_LOCK(rf_rad_freelist);
263 1.6 oster if (raidPtr->waitShutdown) {
264 1.6 oster RF_FREELIST_DO_UNLOCK(rf_rad_freelist);
265 1.6 oster return (EBUSY);
266 1.6 oster }
267 1.6 oster raidPtr->waitShutdown = 1;
268 1.6 oster while (raidPtr->nAccOutstanding) {
269 1.6 oster RF_WAIT_COND(raidPtr->outstandingCond, RF_FREELIST_MUTEX_OF(rf_rad_freelist));
270 1.6 oster }
271 1.6 oster RF_FREELIST_DO_UNLOCK(rf_rad_freelist);
272 1.35 oster
273 1.35 oster /* Wait for any parity re-writes to stop... */
274 1.35 oster while (raidPtr->parity_rewrite_in_progress) {
275 1.35 oster printf("Waiting for parity re-write to exit...\n");
276 1.35 oster tsleep(&raidPtr->parity_rewrite_in_progress, PRIBIO,
277 1.35 oster "rfprwshutdown", 0);
278 1.35 oster }
279 1.6 oster
280 1.6 oster raidPtr->valid = 0;
281 1.6 oster
282 1.37 oster rf_update_component_labels(raidPtr, RF_FINAL_COMPONENT_UPDATE);
283 1.6 oster
284 1.7 oster rf_UnconfigureVnodes(raidPtr);
285 1.7 oster
286 1.7 oster rf_ShutdownList(&raidPtr->shutdownList);
287 1.7 oster
288 1.7 oster rf_UnconfigureArray();
289 1.7 oster
290 1.7 oster return (0);
291 1.7 oster }
292 1.1 oster
293 1.6 oster
294 1.1 oster #define DO_INIT_CONFIGURE(f) { \
295 1.1 oster rc = f (&globalShutdown); \
296 1.1 oster if (rc) { \
297 1.1 oster RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
298 1.1 oster rf_ShutdownList(&globalShutdown); \
299 1.1 oster configureCount--; \
300 1.1 oster RF_UNLOCK_MUTEX(configureMutex); \
301 1.1 oster return(rc); \
302 1.1 oster } \
303 1.1 oster }
304 1.1 oster
305 1.1 oster #define DO_RAID_FAIL() { \
306 1.12 oster rf_UnconfigureVnodes(raidPtr); \
307 1.1 oster rf_ShutdownList(&raidPtr->shutdownList); \
308 1.1 oster rf_UnconfigureArray(); \
309 1.1 oster }
310 1.1 oster
311 1.1 oster #define DO_RAID_INIT_CONFIGURE(f) { \
312 1.1 oster rc = f (&raidPtr->shutdownList, raidPtr, cfgPtr); \
313 1.1 oster if (rc) { \
314 1.1 oster RF_ERRORMSG2("RAIDFRAME: failed %s with %d\n", RF_STRING(f), rc); \
315 1.1 oster DO_RAID_FAIL(); \
316 1.1 oster return(rc); \
317 1.1 oster } \
318 1.1 oster }
319 1.1 oster
320 1.1 oster #define DO_RAID_MUTEX(_m_) { \
321 1.1 oster rc = rf_create_managed_mutex(&raidPtr->shutdownList, (_m_)); \
322 1.1 oster if (rc) { \
323 1.1 oster RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n", \
324 1.1 oster __FILE__, __LINE__, rc); \
325 1.1 oster DO_RAID_FAIL(); \
326 1.1 oster return(rc); \
327 1.1 oster } \
328 1.1 oster }
329 1.1 oster
330 1.1 oster #define DO_RAID_COND(_c_) { \
331 1.1 oster rc = rf_create_managed_cond(&raidPtr->shutdownList, (_c_)); \
332 1.1 oster if (rc) { \
333 1.1 oster RF_ERRORMSG3("Unable to init cond file %s line %d rc=%d\n", \
334 1.1 oster __FILE__, __LINE__, rc); \
335 1.1 oster DO_RAID_FAIL(); \
336 1.1 oster return(rc); \
337 1.1 oster } \
338 1.1 oster }
339 1.1 oster
340 1.6 oster int
341 1.28 oster rf_Configure(raidPtr, cfgPtr, ac)
342 1.6 oster RF_Raid_t *raidPtr;
343 1.6 oster RF_Config_t *cfgPtr;
344 1.28 oster RF_AutoConfig_t *ac;
345 1.6 oster {
346 1.6 oster RF_RowCol_t row, col;
347 1.6 oster int i, rc;
348 1.6 oster
349 1.28 oster /* XXX This check can probably be removed now, since
350 1.28 oster RAIDFRAME_CONFIGURRE now checks to make sure that the
351 1.28 oster RAID set is not already valid
352 1.28 oster */
353 1.6 oster if (raidPtr->valid) {
354 1.6 oster RF_ERRORMSG("RAIDframe configuration not shut down. Aborting configure.\n");
355 1.6 oster return (EINVAL);
356 1.6 oster }
357 1.6 oster RF_LOCK_MUTEX(configureMutex);
358 1.6 oster configureCount++;
359 1.6 oster if (isconfigged == 0) {
360 1.6 oster rc = rf_create_managed_mutex(&globalShutdown, &rf_printf_mutex);
361 1.6 oster if (rc) {
362 1.6 oster RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n", __FILE__,
363 1.6 oster __LINE__, rc);
364 1.6 oster rf_ShutdownList(&globalShutdown);
365 1.6 oster return (rc);
366 1.6 oster }
367 1.6 oster /* initialize globals */
368 1.28 oster printf("RAIDFRAME: protectedSectors is %ld\n",
369 1.28 oster rf_protectedSectors);
370 1.6 oster
371 1.6 oster rf_clear_debug_print_buffer();
372 1.6 oster
373 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureAllocList);
374 1.28 oster
375 1.6 oster /*
376 1.28 oster * Yes, this does make debugging general to the whole
377 1.28 oster * system instead of being array specific. Bummer, drag.
378 1.28 oster */
379 1.6 oster rf_ConfigureDebug(cfgPtr);
380 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureDebugMem);
381 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureAccessTrace);
382 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureMapModule);
383 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureReconEvent);
384 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureCallback);
385 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureMemChunk);
386 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureRDFreeList);
387 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureNWayXor);
388 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureStripeLockFreeList);
389 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureMCPair);
390 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureDAGs);
391 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureDAGFuncs);
392 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureDebugPrint);
393 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureReconstruction);
394 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureCopyback);
395 1.6 oster DO_INIT_CONFIGURE(rf_ConfigureDiskQueueSystem);
396 1.6 oster isconfigged = 1;
397 1.6 oster }
398 1.6 oster RF_UNLOCK_MUTEX(configureMutex);
399 1.6 oster
400 1.6 oster DO_RAID_MUTEX(&raidPtr->mutex);
401 1.6 oster /* set up the cleanup list. Do this after ConfigureDebug so that
402 1.6 oster * value of memDebug will be set */
403 1.6 oster
404 1.6 oster rf_MakeAllocList(raidPtr->cleanupList);
405 1.6 oster if (raidPtr->cleanupList == NULL) {
406 1.6 oster DO_RAID_FAIL();
407 1.6 oster return (ENOMEM);
408 1.6 oster }
409 1.6 oster rc = rf_ShutdownCreate(&raidPtr->shutdownList,
410 1.6 oster (void (*) (void *)) rf_FreeAllocList,
411 1.6 oster raidPtr->cleanupList);
412 1.6 oster if (rc) {
413 1.6 oster RF_ERRORMSG3("Unable to add to shutdown list file %s line %d rc=%d\n",
414 1.6 oster __FILE__, __LINE__, rc);
415 1.6 oster DO_RAID_FAIL();
416 1.6 oster return (rc);
417 1.6 oster }
418 1.6 oster raidPtr->numRow = cfgPtr->numRow;
419 1.6 oster raidPtr->numCol = cfgPtr->numCol;
420 1.6 oster raidPtr->numSpare = cfgPtr->numSpare;
421 1.6 oster
422 1.6 oster /* XXX we don't even pretend to support more than one row in the
423 1.6 oster * kernel... */
424 1.6 oster if (raidPtr->numRow != 1) {
425 1.6 oster RF_ERRORMSG("Only one row supported in kernel.\n");
426 1.6 oster DO_RAID_FAIL();
427 1.6 oster return (EINVAL);
428 1.6 oster }
429 1.6 oster RF_CallocAndAdd(raidPtr->status, raidPtr->numRow, sizeof(RF_RowStatus_t),
430 1.6 oster (RF_RowStatus_t *), raidPtr->cleanupList);
431 1.6 oster if (raidPtr->status == NULL) {
432 1.6 oster DO_RAID_FAIL();
433 1.6 oster return (ENOMEM);
434 1.6 oster }
435 1.6 oster RF_CallocAndAdd(raidPtr->reconControl, raidPtr->numRow,
436 1.6 oster sizeof(RF_ReconCtrl_t *), (RF_ReconCtrl_t **), raidPtr->cleanupList);
437 1.6 oster if (raidPtr->reconControl == NULL) {
438 1.6 oster DO_RAID_FAIL();
439 1.6 oster return (ENOMEM);
440 1.6 oster }
441 1.6 oster for (i = 0; i < raidPtr->numRow; i++) {
442 1.6 oster raidPtr->status[i] = rf_rs_optimal;
443 1.6 oster raidPtr->reconControl[i] = NULL;
444 1.6 oster }
445 1.6 oster
446 1.6 oster DO_RAID_INIT_CONFIGURE(rf_ConfigureEngine);
447 1.6 oster DO_RAID_INIT_CONFIGURE(rf_ConfigureStripeLocks);
448 1.6 oster
449 1.6 oster DO_RAID_COND(&raidPtr->outstandingCond);
450 1.6 oster
451 1.6 oster raidPtr->nAccOutstanding = 0;
452 1.6 oster raidPtr->waitShutdown = 0;
453 1.6 oster
454 1.6 oster DO_RAID_MUTEX(&raidPtr->access_suspend_mutex);
455 1.6 oster DO_RAID_COND(&raidPtr->quiescent_cond);
456 1.6 oster
457 1.6 oster DO_RAID_COND(&raidPtr->waitForReconCond);
458 1.6 oster
459 1.6 oster DO_RAID_MUTEX(&raidPtr->recon_done_proc_mutex);
460 1.28 oster
461 1.28 oster if (ac!=NULL) {
462 1.28 oster /* We have an AutoConfig structure.. Don't do the
463 1.28 oster normal disk configuration... call the auto config
464 1.28 oster stuff */
465 1.28 oster rf_AutoConfigureDisks(raidPtr, cfgPtr, ac);
466 1.28 oster } else {
467 1.28 oster DO_RAID_INIT_CONFIGURE(rf_ConfigureDisks);
468 1.28 oster DO_RAID_INIT_CONFIGURE(rf_ConfigureSpareDisks);
469 1.28 oster }
470 1.6 oster /* do this after ConfigureDisks & ConfigureSpareDisks to be sure dev
471 1.6 oster * no. is set */
472 1.6 oster DO_RAID_INIT_CONFIGURE(rf_ConfigureDiskQueues);
473 1.6 oster
474 1.6 oster DO_RAID_INIT_CONFIGURE(rf_ConfigureLayout);
475 1.6 oster
476 1.6 oster DO_RAID_INIT_CONFIGURE(rf_ConfigurePSStatus);
477 1.6 oster
478 1.6 oster for (row = 0; row < raidPtr->numRow; row++) {
479 1.6 oster for (col = 0; col < raidPtr->numCol; col++) {
480 1.6 oster /*
481 1.6 oster * XXX better distribution
482 1.6 oster */
483 1.6 oster raidPtr->hist_diskreq[row][col] = 0;
484 1.6 oster }
485 1.6 oster }
486 1.28 oster
487 1.30 oster raidPtr->numNewFailures = 0;
488 1.28 oster raidPtr->copyback_in_progress = 0;
489 1.28 oster raidPtr->parity_rewrite_in_progress = 0;
490 1.28 oster raidPtr->recon_in_progress = 0;
491 1.29 oster raidPtr->maxOutstanding = cfgPtr->maxOutstandingDiskReqs;
492 1.29 oster
493 1.29 oster /* autoconfigure and root_partition will actually get filled in
494 1.29 oster after the config is done */
495 1.29 oster raidPtr->autoconfigure = 0;
496 1.29 oster raidPtr->root_partition = 0;
497 1.29 oster raidPtr->last_unit = raidPtr->raidid;
498 1.29 oster raidPtr->config_order = 0;
499 1.6 oster
500 1.6 oster if (rf_keepAccTotals) {
501 1.6 oster raidPtr->keep_acc_totals = 1;
502 1.6 oster }
503 1.6 oster rf_StartUserStats(raidPtr);
504 1.1 oster
505 1.6 oster raidPtr->valid = 1;
506 1.6 oster return (0);
507 1.1 oster }
508 1.1 oster
509 1.6 oster static int
510 1.6 oster init_rad(desc)
511 1.6 oster RF_RaidAccessDesc_t *desc;
512 1.1 oster {
513 1.6 oster int rc;
514 1.1 oster
515 1.1 oster rc = rf_mutex_init(&desc->mutex);
516 1.1 oster if (rc) {
517 1.1 oster RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n", __FILE__,
518 1.6 oster __LINE__, rc);
519 1.6 oster return (rc);
520 1.1 oster }
521 1.1 oster rc = rf_cond_init(&desc->cond);
522 1.1 oster if (rc) {
523 1.1 oster RF_ERRORMSG3("Unable to init cond file %s line %d rc=%d\n", __FILE__,
524 1.6 oster __LINE__, rc);
525 1.1 oster rf_mutex_destroy(&desc->mutex);
526 1.6 oster return (rc);
527 1.1 oster }
528 1.6 oster return (0);
529 1.1 oster }
530 1.1 oster
531 1.6 oster static void
532 1.6 oster clean_rad(desc)
533 1.6 oster RF_RaidAccessDesc_t *desc;
534 1.1 oster {
535 1.1 oster rf_mutex_destroy(&desc->mutex);
536 1.1 oster rf_cond_destroy(&desc->cond);
537 1.1 oster }
538 1.1 oster
539 1.6 oster static void
540 1.6 oster rf_ShutdownRDFreeList(ignored)
541 1.6 oster void *ignored;
542 1.1 oster {
543 1.6 oster RF_FREELIST_DESTROY_CLEAN(rf_rad_freelist, next, (RF_RaidAccessDesc_t *), clean_rad);
544 1.1 oster }
545 1.1 oster
546 1.6 oster static int
547 1.6 oster rf_ConfigureRDFreeList(listp)
548 1.6 oster RF_ShutdownList_t **listp;
549 1.1 oster {
550 1.6 oster int rc;
551 1.1 oster
552 1.1 oster RF_FREELIST_CREATE(rf_rad_freelist, RF_MAX_FREE_RAD,
553 1.6 oster RF_RAD_INC, sizeof(RF_RaidAccessDesc_t));
554 1.1 oster if (rf_rad_freelist == NULL) {
555 1.6 oster return (ENOMEM);
556 1.1 oster }
557 1.1 oster rc = rf_ShutdownCreate(listp, rf_ShutdownRDFreeList, NULL);
558 1.1 oster if (rc) {
559 1.1 oster RF_ERRORMSG3("Unable to add to shutdown list file %s line %d rc=%d\n", __FILE__,
560 1.6 oster __LINE__, rc);
561 1.1 oster rf_ShutdownRDFreeList(NULL);
562 1.6 oster return (rc);
563 1.1 oster }
564 1.6 oster RF_FREELIST_PRIME_INIT(rf_rad_freelist, RF_RAD_INITIAL, next,
565 1.6 oster (RF_RaidAccessDesc_t *), init_rad);
566 1.6 oster return (0);
567 1.6 oster }
568 1.6 oster
569 1.6 oster RF_RaidAccessDesc_t *
570 1.6 oster rf_AllocRaidAccDesc(
571 1.6 oster RF_Raid_t * raidPtr,
572 1.6 oster RF_IoType_t type,
573 1.6 oster RF_RaidAddr_t raidAddress,
574 1.6 oster RF_SectorCount_t numBlocks,
575 1.6 oster caddr_t bufPtr,
576 1.6 oster void *bp,
577 1.6 oster RF_RaidAccessFlags_t flags,
578 1.6 oster RF_AccessState_t * states)
579 1.6 oster {
580 1.6 oster RF_RaidAccessDesc_t *desc;
581 1.6 oster
582 1.6 oster RF_FREELIST_GET_INIT_NOUNLOCK(rf_rad_freelist, desc, next, (RF_RaidAccessDesc_t *), init_rad);
583 1.6 oster if (raidPtr->waitShutdown) {
584 1.6 oster /*
585 1.6 oster * Actually, we're shutting the array down. Free the desc
586 1.6 oster * and return NULL.
587 1.6 oster */
588 1.6 oster RF_FREELIST_DO_UNLOCK(rf_rad_freelist);
589 1.6 oster RF_FREELIST_FREE_CLEAN(rf_rad_freelist, desc, next, clean_rad);
590 1.6 oster return (NULL);
591 1.6 oster }
592 1.6 oster raidPtr->nAccOutstanding++;
593 1.6 oster RF_FREELIST_DO_UNLOCK(rf_rad_freelist);
594 1.6 oster
595 1.6 oster desc->raidPtr = (void *) raidPtr;
596 1.6 oster desc->type = type;
597 1.6 oster desc->raidAddress = raidAddress;
598 1.6 oster desc->numBlocks = numBlocks;
599 1.6 oster desc->bufPtr = bufPtr;
600 1.6 oster desc->bp = bp;
601 1.41 oster desc->paramDAG = NULL;
602 1.41 oster desc->paramASM = NULL;
603 1.6 oster desc->flags = flags;
604 1.6 oster desc->states = states;
605 1.6 oster desc->state = 0;
606 1.6 oster
607 1.6 oster desc->status = 0;
608 1.40 thorpej memset((char *) &desc->tracerec, 0, sizeof(RF_AccTraceEntry_t));
609 1.41 oster desc->callbackFunc = NULL;
610 1.41 oster desc->callbackArg = NULL;
611 1.6 oster desc->next = NULL;
612 1.6 oster desc->head = desc;
613 1.6 oster desc->numPending = 0;
614 1.6 oster desc->cleanupList = NULL;
615 1.6 oster rf_MakeAllocList(desc->cleanupList);
616 1.6 oster return (desc);
617 1.6 oster }
618 1.6 oster
619 1.6 oster void
620 1.6 oster rf_FreeRaidAccDesc(RF_RaidAccessDesc_t * desc)
621 1.6 oster {
622 1.6 oster RF_Raid_t *raidPtr = desc->raidPtr;
623 1.6 oster
624 1.6 oster RF_ASSERT(desc);
625 1.6 oster
626 1.6 oster rf_FreeAllocList(desc->cleanupList);
627 1.6 oster RF_FREELIST_FREE_CLEAN_NOUNLOCK(rf_rad_freelist, desc, next, clean_rad);
628 1.6 oster raidPtr->nAccOutstanding--;
629 1.6 oster if (raidPtr->waitShutdown) {
630 1.6 oster RF_SIGNAL_COND(raidPtr->outstandingCond);
631 1.6 oster }
632 1.6 oster RF_FREELIST_DO_UNLOCK(rf_rad_freelist);
633 1.1 oster }
634 1.1 oster /*********************************************************************
635 1.1 oster * Main routine for performing an access.
636 1.1 oster * Accesses are retried until a DAG can not be selected. This occurs
637 1.1 oster * when either the DAG library is incomplete or there are too many
638 1.1 oster * failures in a parity group.
639 1.1 oster ********************************************************************/
640 1.6 oster int
641 1.6 oster rf_DoAccess(
642 1.6 oster RF_Raid_t * raidPtr,
643 1.6 oster RF_IoType_t type,
644 1.6 oster int async_flag,
645 1.6 oster RF_RaidAddr_t raidAddress,
646 1.6 oster RF_SectorCount_t numBlocks,
647 1.6 oster caddr_t bufPtr,
648 1.6 oster void *bp_in,
649 1.41 oster RF_RaidAccessFlags_t flags)
650 1.1 oster /*
651 1.1 oster type should be read or write
652 1.1 oster async_flag should be RF_TRUE or RF_FALSE
653 1.1 oster bp_in is a buf pointer. void * to facilitate ignoring it outside the kernel
654 1.1 oster */
655 1.1 oster {
656 1.6 oster RF_RaidAccessDesc_t *desc;
657 1.6 oster caddr_t lbufPtr = bufPtr;
658 1.6 oster struct buf *bp = (struct buf *) bp_in;
659 1.6 oster
660 1.6 oster raidAddress += rf_raidSectorOffset;
661 1.6 oster
662 1.6 oster if (!raidPtr->valid) {
663 1.6 oster RF_ERRORMSG("RAIDframe driver not successfully configured. Rejecting access.\n");
664 1.20 oster IO_BUF_ERR(bp, EINVAL);
665 1.6 oster return (EINVAL);
666 1.6 oster }
667 1.6 oster
668 1.6 oster if (rf_accessDebug) {
669 1.1 oster
670 1.6 oster printf("logBytes is: %d %d %d\n", raidPtr->raidid,
671 1.6 oster raidPtr->logBytesPerSector,
672 1.6 oster (int) rf_RaidAddressToByte(raidPtr, numBlocks));
673 1.22 oster printf("raid%d: %s raidAddr %d (stripeid %d-%d) numBlocks %d (%d bytes) buf 0x%lx\n", raidPtr->raidid,
674 1.6 oster (type == RF_IO_TYPE_READ) ? "READ" : "WRITE", (int) raidAddress,
675 1.6 oster (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress),
676 1.6 oster (int) rf_RaidAddressToStripeID(&raidPtr->Layout, raidAddress + numBlocks - 1),
677 1.6 oster (int) numBlocks,
678 1.6 oster (int) rf_RaidAddressToByte(raidPtr, numBlocks),
679 1.6 oster (long) bufPtr);
680 1.6 oster }
681 1.6 oster if (raidAddress + numBlocks > raidPtr->totalSectors) {
682 1.1 oster
683 1.6 oster printf("DoAccess: raid addr %lu too large to access %lu sectors. Max legal addr is %lu\n",
684 1.6 oster (u_long) raidAddress, (u_long) numBlocks, (u_long) raidPtr->totalSectors);
685 1.1 oster
686 1.20 oster IO_BUF_ERR(bp, ENOSPC);
687 1.16 oster return (ENOSPC);
688 1.6 oster }
689 1.6 oster desc = rf_AllocRaidAccDesc(raidPtr, type, raidAddress,
690 1.41 oster numBlocks, lbufPtr, bp, flags, raidPtr->Layout.map->states);
691 1.1 oster
692 1.6 oster if (desc == NULL) {
693 1.6 oster return (ENOMEM);
694 1.6 oster }
695 1.6 oster RF_ETIMER_START(desc->tracerec.tot_timer);
696 1.1 oster
697 1.6 oster desc->async_flag = async_flag;
698 1.3 explorer
699 1.6 oster rf_ContinueRaidAccess(desc);
700 1.1 oster
701 1.6 oster return (0);
702 1.1 oster }
703 1.1 oster /* force the array into reconfigured mode without doing reconstruction */
704 1.6 oster int
705 1.6 oster rf_SetReconfiguredMode(raidPtr, row, col)
706 1.6 oster RF_Raid_t *raidPtr;
707 1.6 oster int row;
708 1.6 oster int col;
709 1.6 oster {
710 1.6 oster if (!(raidPtr->Layout.map->flags & RF_DISTRIBUTE_SPARE)) {
711 1.6 oster printf("Can't set reconfigured mode in dedicated-spare array\n");
712 1.6 oster RF_PANIC();
713 1.6 oster }
714 1.6 oster RF_LOCK_MUTEX(raidPtr->mutex);
715 1.6 oster raidPtr->numFailures++;
716 1.6 oster raidPtr->Disks[row][col].status = rf_ds_dist_spared;
717 1.6 oster raidPtr->status[row] = rf_rs_reconfigured;
718 1.37 oster rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
719 1.6 oster /* install spare table only if declustering + distributed sparing
720 1.6 oster * architecture. */
721 1.6 oster if (raidPtr->Layout.map->flags & RF_BD_DECLUSTERED)
722 1.6 oster rf_InstallSpareTable(raidPtr, row, col);
723 1.6 oster RF_UNLOCK_MUTEX(raidPtr->mutex);
724 1.6 oster return (0);
725 1.1 oster }
726 1.1 oster
727 1.6 oster int
728 1.6 oster rf_FailDisk(
729 1.6 oster RF_Raid_t * raidPtr,
730 1.6 oster int frow,
731 1.6 oster int fcol,
732 1.6 oster int initRecon)
733 1.6 oster {
734 1.22 oster printf("raid%d: Failing disk r%d c%d\n", raidPtr->raidid, frow, fcol);
735 1.6 oster RF_LOCK_MUTEX(raidPtr->mutex);
736 1.6 oster raidPtr->numFailures++;
737 1.6 oster raidPtr->Disks[frow][fcol].status = rf_ds_failed;
738 1.6 oster raidPtr->status[frow] = rf_rs_degraded;
739 1.37 oster rf_update_component_labels(raidPtr, RF_NORMAL_COMPONENT_UPDATE);
740 1.6 oster RF_UNLOCK_MUTEX(raidPtr->mutex);
741 1.6 oster if (initRecon)
742 1.6 oster rf_ReconstructFailedDisk(raidPtr, frow, fcol);
743 1.6 oster return (0);
744 1.1 oster }
745 1.1 oster /* releases a thread that is waiting for the array to become quiesced.
746 1.1 oster * access_suspend_mutex should be locked upon calling this
747 1.1 oster */
748 1.6 oster void
749 1.6 oster rf_SignalQuiescenceLock(raidPtr, reconDesc)
750 1.6 oster RF_Raid_t *raidPtr;
751 1.6 oster RF_RaidReconDesc_t *reconDesc;
752 1.6 oster {
753 1.6 oster if (rf_quiesceDebug) {
754 1.22 oster printf("raid%d: Signalling quiescence lock\n",
755 1.22 oster raidPtr->raidid);
756 1.6 oster }
757 1.6 oster raidPtr->access_suspend_release = 1;
758 1.6 oster
759 1.6 oster if (raidPtr->waiting_for_quiescence) {
760 1.6 oster SIGNAL_QUIESCENT_COND(raidPtr);
761 1.6 oster }
762 1.1 oster }
763 1.1 oster /* suspends all new requests to the array. No effect on accesses that are in flight. */
764 1.6 oster int
765 1.6 oster rf_SuspendNewRequestsAndWait(raidPtr)
766 1.6 oster RF_Raid_t *raidPtr;
767 1.6 oster {
768 1.6 oster if (rf_quiesceDebug)
769 1.6 oster printf("Suspending new reqs\n");
770 1.6 oster
771 1.6 oster RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
772 1.6 oster raidPtr->accesses_suspended++;
773 1.6 oster raidPtr->waiting_for_quiescence = (raidPtr->accs_in_flight == 0) ? 0 : 1;
774 1.6 oster
775 1.6 oster if (raidPtr->waiting_for_quiescence) {
776 1.6 oster raidPtr->access_suspend_release = 0;
777 1.6 oster while (!raidPtr->access_suspend_release) {
778 1.36 oster printf("Suspending: Waiting for Quiescence\n");
779 1.6 oster WAIT_FOR_QUIESCENCE(raidPtr);
780 1.6 oster raidPtr->waiting_for_quiescence = 0;
781 1.6 oster }
782 1.6 oster }
783 1.36 oster printf("Quiescence reached..\n");
784 1.1 oster
785 1.6 oster RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
786 1.6 oster return (raidPtr->waiting_for_quiescence);
787 1.1 oster }
788 1.1 oster /* wake up everyone waiting for quiescence to be released */
789 1.6 oster void
790 1.6 oster rf_ResumeNewRequests(raidPtr)
791 1.6 oster RF_Raid_t *raidPtr;
792 1.6 oster {
793 1.6 oster RF_CallbackDesc_t *t, *cb;
794 1.6 oster
795 1.6 oster if (rf_quiesceDebug)
796 1.6 oster printf("Resuming new reqs\n");
797 1.6 oster
798 1.6 oster RF_LOCK_MUTEX(raidPtr->access_suspend_mutex);
799 1.6 oster raidPtr->accesses_suspended--;
800 1.6 oster if (raidPtr->accesses_suspended == 0)
801 1.6 oster cb = raidPtr->quiesce_wait_list;
802 1.6 oster else
803 1.6 oster cb = NULL;
804 1.6 oster raidPtr->quiesce_wait_list = NULL;
805 1.6 oster RF_UNLOCK_MUTEX(raidPtr->access_suspend_mutex);
806 1.6 oster
807 1.6 oster while (cb) {
808 1.6 oster t = cb;
809 1.6 oster cb = cb->next;
810 1.6 oster (t->callbackFunc) (t->callbackArg);
811 1.6 oster rf_FreeCallbackDesc(t);
812 1.6 oster }
813 1.1 oster }
814 1.1 oster /*****************************************************************************************
815 1.1 oster *
816 1.1 oster * debug routines
817 1.1 oster *
818 1.1 oster ****************************************************************************************/
819 1.1 oster
820 1.6 oster static void
821 1.6 oster set_debug_option(name, val)
822 1.6 oster char *name;
823 1.6 oster long val;
824 1.6 oster {
825 1.6 oster RF_DebugName_t *p;
826 1.6 oster
827 1.6 oster for (p = rf_debugNames; p->name; p++) {
828 1.6 oster if (!strcmp(p->name, name)) {
829 1.6 oster *(p->ptr) = val;
830 1.6 oster printf("[Set debug variable %s to %ld]\n", name, val);
831 1.6 oster return;
832 1.6 oster }
833 1.6 oster }
834 1.6 oster RF_ERRORMSG1("Unknown debug string \"%s\"\n", name);
835 1.1 oster }
836 1.1 oster
837 1.1 oster
838 1.1 oster /* would like to use sscanf here, but apparently not available in kernel */
839 1.1 oster /*ARGSUSED*/
840 1.6 oster static void
841 1.6 oster rf_ConfigureDebug(cfgPtr)
842 1.6 oster RF_Config_t *cfgPtr;
843 1.6 oster {
844 1.6 oster char *val_p, *name_p, *white_p;
845 1.6 oster long val;
846 1.6 oster int i;
847 1.6 oster
848 1.6 oster rf_ResetDebugOptions();
849 1.6 oster for (i = 0; cfgPtr->debugVars[i][0] && i < RF_MAXDBGV; i++) {
850 1.6 oster name_p = rf_find_non_white(&cfgPtr->debugVars[i][0]);
851 1.6 oster white_p = rf_find_white(name_p); /* skip to start of 2nd
852 1.6 oster * word */
853 1.6 oster val_p = rf_find_non_white(white_p);
854 1.6 oster if (*val_p == '0' && *(val_p + 1) == 'x')
855 1.6 oster val = rf_htoi(val_p + 2);
856 1.6 oster else
857 1.6 oster val = rf_atoi(val_p);
858 1.6 oster *white_p = '\0';
859 1.6 oster set_debug_option(name_p, val);
860 1.6 oster }
861 1.1 oster }
862 1.1 oster /* performance monitoring stuff */
863 1.1 oster
864 1.1 oster #define TIMEVAL_TO_US(t) (((long) t.tv_sec) * 1000000L + (long) t.tv_usec)
865 1.1 oster
866 1.4 oster #if !defined(_KERNEL) && !defined(SIMULATE)
867 1.1 oster
868 1.1 oster /*
869 1.1 oster * Throughput stats currently only used in user-level RAIDframe
870 1.1 oster */
871 1.1 oster
872 1.6 oster static int
873 1.6 oster rf_InitThroughputStats(
874 1.6 oster RF_ShutdownList_t ** listp,
875 1.6 oster RF_Raid_t * raidPtr,
876 1.6 oster RF_Config_t * cfgPtr)
877 1.1 oster {
878 1.6 oster int rc;
879 1.1 oster
880 1.6 oster /* these used by user-level raidframe only */
881 1.6 oster rc = rf_create_managed_mutex(listp, &raidPtr->throughputstats.mutex);
882 1.6 oster if (rc) {
883 1.6 oster RF_ERRORMSG3("Unable to init mutex file %s line %d rc=%d\n", __FILE__,
884 1.6 oster __LINE__, rc);
885 1.6 oster return (rc);
886 1.6 oster }
887 1.6 oster raidPtr->throughputstats.sum_io_us = 0;
888 1.6 oster raidPtr->throughputstats.num_ios = 0;
889 1.6 oster raidPtr->throughputstats.num_out_ios = 0;
890 1.6 oster return (0);
891 1.6 oster }
892 1.6 oster
893 1.6 oster void
894 1.6 oster rf_StartThroughputStats(RF_Raid_t * raidPtr)
895 1.6 oster {
896 1.6 oster RF_LOCK_MUTEX(raidPtr->throughputstats.mutex);
897 1.6 oster raidPtr->throughputstats.num_ios++;
898 1.6 oster raidPtr->throughputstats.num_out_ios++;
899 1.6 oster if (raidPtr->throughputstats.num_out_ios == 1)
900 1.6 oster RF_GETTIME(raidPtr->throughputstats.start);
901 1.6 oster RF_UNLOCK_MUTEX(raidPtr->throughputstats.mutex);
902 1.6 oster }
903 1.6 oster
904 1.6 oster static void
905 1.6 oster rf_StopThroughputStats(RF_Raid_t * raidPtr)
906 1.6 oster {
907 1.6 oster struct timeval diff;
908 1.6 oster
909 1.6 oster RF_LOCK_MUTEX(raidPtr->throughputstats.mutex);
910 1.6 oster raidPtr->throughputstats.num_out_ios--;
911 1.6 oster if (raidPtr->throughputstats.num_out_ios == 0) {
912 1.6 oster RF_GETTIME(raidPtr->throughputstats.stop);
913 1.6 oster RF_TIMEVAL_DIFF(&raidPtr->throughputstats.start, &raidPtr->throughputstats.stop, &diff);
914 1.6 oster raidPtr->throughputstats.sum_io_us += TIMEVAL_TO_US(diff);
915 1.6 oster }
916 1.6 oster RF_UNLOCK_MUTEX(raidPtr->throughputstats.mutex);
917 1.1 oster }
918 1.1 oster
919 1.6 oster static void
920 1.6 oster rf_PrintThroughputStats(RF_Raid_t * raidPtr)
921 1.1 oster {
922 1.6 oster RF_ASSERT(raidPtr->throughputstats.num_out_ios == 0);
923 1.6 oster if (raidPtr->throughputstats.sum_io_us != 0) {
924 1.6 oster printf("[Througphut: %8.2f IOs/second]\n", raidPtr->throughputstats.num_ios
925 1.6 oster / (raidPtr->throughputstats.sum_io_us / 1000000.0));
926 1.6 oster }
927 1.1 oster }
928 1.6 oster #endif /* !KERNEL && !SIMULATE */
929 1.1 oster
930 1.6 oster void
931 1.6 oster rf_StartUserStats(RF_Raid_t * raidPtr)
932 1.1 oster {
933 1.6 oster RF_GETTIME(raidPtr->userstats.start);
934 1.6 oster raidPtr->userstats.sum_io_us = 0;
935 1.6 oster raidPtr->userstats.num_ios = 0;
936 1.6 oster raidPtr->userstats.num_sect_moved = 0;
937 1.1 oster }
938 1.1 oster
939 1.6 oster void
940 1.6 oster rf_StopUserStats(RF_Raid_t * raidPtr)
941 1.1 oster {
942 1.6 oster RF_GETTIME(raidPtr->userstats.stop);
943 1.1 oster }
944 1.1 oster
945 1.6 oster void
946 1.6 oster rf_UpdateUserStats(raidPtr, rt, numsect)
947 1.6 oster RF_Raid_t *raidPtr;
948 1.6 oster int rt; /* resp time in us */
949 1.6 oster int numsect; /* number of sectors for this access */
950 1.1 oster {
951 1.6 oster raidPtr->userstats.sum_io_us += rt;
952 1.6 oster raidPtr->userstats.num_ios++;
953 1.6 oster raidPtr->userstats.num_sect_moved += numsect;
954 1.1 oster }
955 1.1 oster
956 1.6 oster void
957 1.6 oster rf_PrintUserStats(RF_Raid_t * raidPtr)
958 1.1 oster {
959 1.6 oster long elapsed_us, mbs, mbs_frac;
960 1.6 oster struct timeval diff;
961 1.1 oster
962 1.6 oster RF_TIMEVAL_DIFF(&raidPtr->userstats.start, &raidPtr->userstats.stop, &diff);
963 1.6 oster elapsed_us = TIMEVAL_TO_US(diff);
964 1.1 oster
965 1.6 oster /* 2000 sectors per megabyte, 10000000 microseconds per second */
966 1.6 oster if (elapsed_us)
967 1.6 oster mbs = (raidPtr->userstats.num_sect_moved / 2000) / (elapsed_us / 1000000);
968 1.6 oster else
969 1.6 oster mbs = 0;
970 1.1 oster
971 1.6 oster /* this computes only the first digit of the fractional mb/s moved */
972 1.6 oster if (elapsed_us) {
973 1.6 oster mbs_frac = ((raidPtr->userstats.num_sect_moved / 200) / (elapsed_us / 1000000))
974 1.6 oster - (mbs * 10);
975 1.6 oster } else {
976 1.6 oster mbs_frac = 0;
977 1.6 oster }
978 1.1 oster
979 1.6 oster printf("Number of I/Os: %ld\n", raidPtr->userstats.num_ios);
980 1.6 oster printf("Elapsed time (us): %ld\n", elapsed_us);
981 1.6 oster printf("User I/Os per second: %ld\n", RF_DB0_CHECK(raidPtr->userstats.num_ios, (elapsed_us / 1000000)));
982 1.6 oster printf("Average user response time: %ld us\n", RF_DB0_CHECK(raidPtr->userstats.sum_io_us, raidPtr->userstats.num_ios));
983 1.6 oster printf("Total sectors moved: %ld\n", raidPtr->userstats.num_sect_moved);
984 1.6 oster printf("Average access size (sect): %ld\n", RF_DB0_CHECK(raidPtr->userstats.num_sect_moved, raidPtr->userstats.num_ios));
985 1.6 oster printf("Achieved data rate: %ld.%ld MB/sec\n", mbs, mbs_frac);
986 1.39 oster }
987 1.39 oster
988 1.39 oster
989 1.39 oster void
990 1.39 oster rf_print_panic_message(line,file)
991 1.39 oster int line;
992 1.39 oster char *file;
993 1.39 oster {
994 1.39 oster sprintf(rf_panicbuf,"raidframe error at line %d file %s",
995 1.39 oster line, file);
996 1.39 oster }
997 1.39 oster
998 1.39 oster void
999 1.39 oster rf_print_assert_panic_message(line,file,condition)
1000 1.39 oster int line;
1001 1.39 oster char *file;
1002 1.39 oster char *condition;
1003 1.39 oster {
1004 1.39 oster sprintf(rf_panicbuf,
1005 1.39 oster "raidframe error at line %d file %s (failed asserting %s)\n",
1006 1.39 oster line, file, condition);
1007 1.1 oster }
1008