FreeBSD/Linux Kernel Cross Reference
sys/geom/geom_io.c
1 /*-
2 * Copyright (c) 2002 Poul-Henning Kamp
3 * Copyright (c) 2002 Networks Associates Technology, Inc.
4 * All rights reserved.
5 *
6 * This software was developed for the FreeBSD Project by Poul-Henning Kamp
7 * and NAI Labs, the Security Research Division of Network Associates, Inc.
8 * under DARPA/SPAWAR contract N66001-01-C-8035 ("CBOSS"), as part of the
9 * DARPA CHATS research program.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. The names of the authors may not be used to endorse or promote
20 * products derived from this software without specific prior written
21 * permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 */
35
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD: releng/5.3/sys/geom/geom_io.c 136754 2004-10-21 17:49:45Z phk $");
38
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/malloc.h>
43 #include <sys/bio.h>
44
45 #include <sys/errno.h>
46 #include <geom/geom.h>
47 #include <geom/geom_int.h>
48 #include <sys/devicestat.h>
49
50 #include <vm/uma.h>
51
52 static struct g_bioq g_bio_run_down;
53 static struct g_bioq g_bio_run_up;
54 static struct g_bioq g_bio_run_task;
55
56 static u_int pace;
57 static uma_zone_t biozone;
58
59 #include <machine/atomic.h>
60
61 static void
62 g_bioq_lock(struct g_bioq *bq)
63 {
64
65 mtx_lock(&bq->bio_queue_lock);
66 }
67
68 static void
69 g_bioq_unlock(struct g_bioq *bq)
70 {
71
72 mtx_unlock(&bq->bio_queue_lock);
73 }
74
75 #if 0
76 static void
77 g_bioq_destroy(struct g_bioq *bq)
78 {
79
80 mtx_destroy(&bq->bio_queue_lock);
81 }
82 #endif
83
84 static void
85 g_bioq_init(struct g_bioq *bq)
86 {
87
88 TAILQ_INIT(&bq->bio_queue);
89 mtx_init(&bq->bio_queue_lock, "bio queue", NULL, MTX_DEF);
90 }
91
92 static struct bio *
93 g_bioq_first(struct g_bioq *bq)
94 {
95 struct bio *bp;
96
97 bp = TAILQ_FIRST(&bq->bio_queue);
98 if (bp != NULL) {
99 KASSERT((bp->bio_flags & BIO_ONQUEUE),
100 ("Bio not on queue bp=%p target %p", bp, bq));
101 bp->bio_flags &= ~BIO_ONQUEUE;
102 TAILQ_REMOVE(&bq->bio_queue, bp, bio_queue);
103 bq->bio_queue_length--;
104 }
105 return (bp);
106 }
107
108 struct bio *
109 g_new_bio(void)
110 {
111 struct bio *bp;
112
113 bp = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
114 return (bp);
115 }
116
117 struct bio *
118 g_alloc_bio(void)
119 {
120 struct bio *bp;
121
122 bp = uma_zalloc(biozone, M_WAITOK | M_ZERO);
123 return (bp);
124 }
125
126 void
127 g_destroy_bio(struct bio *bp)
128 {
129
130 uma_zfree(biozone, bp);
131 }
132
133 struct bio *
134 g_clone_bio(struct bio *bp)
135 {
136 struct bio *bp2;
137
138 bp2 = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
139 if (bp2 != NULL) {
140 bp2->bio_parent = bp;
141 bp2->bio_cmd = bp->bio_cmd;
142 bp2->bio_length = bp->bio_length;
143 bp2->bio_offset = bp->bio_offset;
144 bp2->bio_data = bp->bio_data;
145 bp2->bio_attribute = bp->bio_attribute;
146 bp->bio_children++;
147 }
148 return(bp2);
149 }
150
151 void
152 g_io_init()
153 {
154
155 g_bioq_init(&g_bio_run_down);
156 g_bioq_init(&g_bio_run_up);
157 g_bioq_init(&g_bio_run_task);
158 biozone = uma_zcreate("g_bio", sizeof (struct bio),
159 NULL, NULL,
160 NULL, NULL,
161 0, 0);
162 }
163
164 int
165 g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
166 {
167 struct bio *bp;
168 int error;
169
170 g_trace(G_T_BIO, "bio_getattr(%s)", attr);
171 bp = g_alloc_bio();
172 bp->bio_cmd = BIO_GETATTR;
173 bp->bio_done = NULL;
174 bp->bio_attribute = attr;
175 bp->bio_length = *len;
176 bp->bio_data = ptr;
177 g_io_request(bp, cp);
178 error = biowait(bp, "ggetattr");
179 *len = bp->bio_completed;
180 g_destroy_bio(bp);
181 return (error);
182 }
183
184 static int
185 g_io_check(struct bio *bp)
186 {
187 struct g_consumer *cp;
188 struct g_provider *pp;
189
190 cp = bp->bio_from;
191 pp = bp->bio_to;
192
193 /* Fail if access counters dont allow the operation */
194 switch(bp->bio_cmd) {
195 case BIO_READ:
196 case BIO_GETATTR:
197 if (cp->acr == 0)
198 return (EPERM);
199 break;
200 case BIO_WRITE:
201 case BIO_DELETE:
202 if (cp->acw == 0)
203 return (EPERM);
204 break;
205 default:
206 return (EPERM);
207 }
208 /* if provider is marked for error, don't disturb. */
209 if (pp->error)
210 return (pp->error);
211
212 switch(bp->bio_cmd) {
213 case BIO_READ:
214 case BIO_WRITE:
215 case BIO_DELETE:
216 /* Zero sectorsize is a probably lack of media */
217 if (pp->sectorsize == 0)
218 return (ENXIO);
219 /* Reject I/O not on sector boundary */
220 if (bp->bio_offset % pp->sectorsize)
221 return (EINVAL);
222 /* Reject I/O not integral sector long */
223 if (bp->bio_length % pp->sectorsize)
224 return (EINVAL);
225 /* Reject requests before or past the end of media. */
226 if (bp->bio_offset < 0)
227 return (EIO);
228 if (bp->bio_offset > pp->mediasize)
229 return (EIO);
230 break;
231 default:
232 break;
233 }
234 return (0);
235 }
236
237 void
238 g_io_request(struct bio *bp, struct g_consumer *cp)
239 {
240 struct g_provider *pp;
241
242 KASSERT(cp != NULL, ("NULL cp in g_io_request"));
243 KASSERT(bp != NULL, ("NULL bp in g_io_request"));
244 KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
245 pp = cp->provider;
246 KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
247
248 #ifdef DIAGNOSTIC
249 if (bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) {
250 KASSERT(bp->bio_offset % cp->provider->sectorsize == 0,
251 ("wrong offset %jd for sectorsize %u",
252 bp->bio_offset, cp->provider->sectorsize));
253 KASSERT(bp->bio_length % cp->provider->sectorsize == 0,
254 ("wrong length %jd for sectorsize %u",
255 bp->bio_length, cp->provider->sectorsize));
256 }
257 #endif /* DIAGNOSTIC */
258
259 g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
260 bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
261
262 bp->bio_from = cp;
263 bp->bio_to = pp;
264 bp->bio_error = 0;
265 bp->bio_completed = 0;
266
267 KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
268 ("Bio already on queue bp=%p", bp));
269 bp->bio_flags |= BIO_ONQUEUE;
270
271 binuptime(&bp->bio_t0);
272 if (g_collectstats & 4)
273 g_bioq_lock(&g_bio_run_down);
274 if (g_collectstats & 1)
275 devstat_start_transaction(pp->stat, &bp->bio_t0);
276 if (g_collectstats & 2)
277 devstat_start_transaction(cp->stat, &bp->bio_t0);
278
279 if (!(g_collectstats & 4))
280 g_bioq_lock(&g_bio_run_down);
281 pp->nstart++;
282 cp->nstart++;
283 TAILQ_INSERT_TAIL(&g_bio_run_down.bio_queue, bp, bio_queue);
284 g_bio_run_down.bio_queue_length++;
285 g_bioq_unlock(&g_bio_run_down);
286
287 /* Pass it on down. */
288 wakeup(&g_wait_down);
289 }
290
291 void
292 g_io_deliver(struct bio *bp, int error)
293 {
294 struct g_consumer *cp;
295 struct g_provider *pp;
296
297 KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
298 pp = bp->bio_to;
299 KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
300 cp = bp->bio_from;
301 if (cp == NULL) {
302 bp->bio_error = error;
303 bp->bio_done(bp);
304 return;
305 }
306 KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
307 KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
308 KASSERT(bp->bio_completed >= 0, ("bio_completed can't be less than 0"));
309 KASSERT(bp->bio_completed <= bp->bio_length,
310 ("bio_completed can't be greater than bio_length"));
311
312 g_trace(G_T_BIO,
313 "g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
314 bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
315 (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
316
317 KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
318 ("Bio already on queue bp=%p", bp));
319
320 /*
321 * XXX: next two doesn't belong here
322 */
323 bp->bio_bcount = bp->bio_length;
324 bp->bio_resid = bp->bio_bcount - bp->bio_completed;
325
326 if (g_collectstats & 4)
327 g_bioq_lock(&g_bio_run_up);
328 if (g_collectstats & 1)
329 devstat_end_transaction_bio(pp->stat, bp);
330 if (g_collectstats & 2)
331 devstat_end_transaction_bio(cp->stat, bp);
332 if (!(g_collectstats & 4))
333 g_bioq_lock(&g_bio_run_up);
334 cp->nend++;
335 pp->nend++;
336 if (error != ENOMEM) {
337 bp->bio_error = error;
338 TAILQ_INSERT_TAIL(&g_bio_run_up.bio_queue, bp, bio_queue);
339 bp->bio_flags |= BIO_ONQUEUE;
340 g_bio_run_up.bio_queue_length++;
341 g_bioq_unlock(&g_bio_run_up);
342 wakeup(&g_wait_up);
343 return;
344 }
345 g_bioq_unlock(&g_bio_run_up);
346
347 if (bootverbose)
348 printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
349 bp->bio_children = 0;
350 bp->bio_inbed = 0;
351 g_io_request(bp, cp);
352 pace++;
353 return;
354 }
355
356 void
357 g_io_schedule_down(struct thread *tp __unused)
358 {
359 struct bio *bp;
360 off_t excess;
361 int error;
362 #ifdef WITNESS
363 struct mtx mymutex;
364
365 bzero(&mymutex, sizeof mymutex);
366 mtx_init(&mymutex, "g_xdown", NULL, MTX_DEF);
367 #endif
368
369 for(;;) {
370 g_bioq_lock(&g_bio_run_down);
371 bp = g_bioq_first(&g_bio_run_down);
372 if (bp == NULL) {
373 msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
374 PRIBIO | PDROP, "-", hz/10);
375 continue;
376 }
377 g_bioq_unlock(&g_bio_run_down);
378 if (pace > 0) {
379 msleep(&error, NULL, PRIBIO, "g_down", hz/10);
380 pace--;
381 }
382 error = g_io_check(bp);
383 if (error) {
384 g_io_deliver(bp, error);
385 continue;
386 }
387 switch (bp->bio_cmd) {
388 case BIO_READ:
389 case BIO_WRITE:
390 case BIO_DELETE:
391 /* Truncate requests to the end of providers media. */
392 excess = bp->bio_offset + bp->bio_length;
393 if (excess > bp->bio_to->mediasize) {
394 excess -= bp->bio_to->mediasize;
395 bp->bio_length -= excess;
396 }
397 /* Deliver zero length transfers right here. */
398 if (bp->bio_length == 0) {
399 g_io_deliver(bp, 0);
400 continue;
401 }
402 break;
403 default:
404 break;
405 }
406 #ifdef WITNESS
407 mtx_lock(&mymutex);
408 #endif
409 bp->bio_to->geom->start(bp);
410 #ifdef WITNESS
411 mtx_unlock(&mymutex);
412 #endif
413 }
414 }
415
416 void
417 bio_taskqueue(struct bio *bp, bio_task_t *func, void *arg)
418 {
419 bp->bio_task = func;
420 bp->bio_task_arg = arg;
421 /*
422 * The taskqueue is actually just a second queue off the "up"
423 * queue, so we use the same lock.
424 */
425 g_bioq_lock(&g_bio_run_up);
426 KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
427 ("Bio already on queue bp=%p target taskq", bp));
428 bp->bio_flags |= BIO_ONQUEUE;
429 TAILQ_INSERT_TAIL(&g_bio_run_task.bio_queue, bp, bio_queue);
430 g_bio_run_task.bio_queue_length++;
431 wakeup(&g_wait_up);
432 g_bioq_unlock(&g_bio_run_up);
433 }
434
435
436 void
437 g_io_schedule_up(struct thread *tp __unused)
438 {
439 struct bio *bp;
440 #ifdef WITNESS
441 struct mtx mymutex;
442
443 bzero(&mymutex, sizeof mymutex);
444 mtx_init(&mymutex, "g_xup", NULL, MTX_DEF);
445 #endif
446 for(;;) {
447 g_bioq_lock(&g_bio_run_up);
448 bp = g_bioq_first(&g_bio_run_task);
449 if (bp != NULL) {
450 g_bioq_unlock(&g_bio_run_up);
451 #ifdef WITNESS
452 mtx_lock(&mymutex);
453 #endif
454 bp->bio_task(bp->bio_task_arg);
455 #ifdef WITNESS
456 mtx_unlock(&mymutex);
457 #endif
458 continue;
459 }
460 bp = g_bioq_first(&g_bio_run_up);
461 if (bp != NULL) {
462 g_bioq_unlock(&g_bio_run_up);
463 #ifdef WITNESS
464 mtx_lock(&mymutex);
465 #endif
466 biodone(bp);
467 #ifdef WITNESS
468 mtx_unlock(&mymutex);
469 #endif
470 continue;
471 }
472 msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
473 PRIBIO | PDROP, "-", hz/10);
474 }
475 }
476
477 void *
478 g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
479 {
480 struct bio *bp;
481 void *ptr;
482 int errorc;
483
484 KASSERT(length > 0 && length >= cp->provider->sectorsize &&
485 length <= MAXPHYS, ("g_read_data(): invalid length %jd",
486 (intmax_t)length));
487
488 bp = g_alloc_bio();
489 bp->bio_cmd = BIO_READ;
490 bp->bio_done = NULL;
491 bp->bio_offset = offset;
492 bp->bio_length = length;
493 ptr = g_malloc(length, M_WAITOK);
494 bp->bio_data = ptr;
495 g_io_request(bp, cp);
496 errorc = biowait(bp, "gread");
497 if (error != NULL)
498 *error = errorc;
499 g_destroy_bio(bp);
500 if (errorc) {
501 g_free(ptr);
502 ptr = NULL;
503 }
504 return (ptr);
505 }
506
507 int
508 g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
509 {
510 struct bio *bp;
511 int error;
512
513 KASSERT(length > 0 && length >= cp->provider->sectorsize &&
514 length <= MAXPHYS, ("g_write_data(): invalid length %jd",
515 (intmax_t)length));
516
517 bp = g_alloc_bio();
518 bp->bio_cmd = BIO_WRITE;
519 bp->bio_done = NULL;
520 bp->bio_offset = offset;
521 bp->bio_length = length;
522 bp->bio_data = ptr;
523 g_io_request(bp, cp);
524 error = biowait(bp, "gwrite");
525 g_destroy_bio(bp);
526 return (error);
527 }
528
529 void
530 g_print_bio(struct bio *bp)
531 {
532 const char *pname, *cmd = NULL;
533
534 if (bp->bio_to != NULL)
535 pname = bp->bio_to->name;
536 else
537 pname = "[unknown]";
538
539 switch (bp->bio_cmd) {
540 case BIO_GETATTR:
541 cmd = "GETATTR";
542 printf("%s[%s(attr=%s)]", pname, cmd, bp->bio_attribute);
543 return;
544 case BIO_READ:
545 cmd = "READ";
546 case BIO_WRITE:
547 if (cmd == NULL)
548 cmd = "WRITE";
549 case BIO_DELETE:
550 if (cmd == NULL)
551 cmd = "DELETE";
552 printf("%s[%s(offset=%jd, length=%jd)]", pname, cmd,
553 (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
554 return;
555 default:
556 cmd = "UNKNOWN";
557 printf("%s[%s()]", pname, cmd);
558 return;
559 }
560 /* NOTREACHED */
561 }
Cache object: f85de7a58ed7060dc1352272d2cdb85c
|