geom_io.c revision 135876 
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: head/sys/geom/geom_io.c 135876 2004-09-28 11:56:37Z 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
52static struct g_bioq g_bio_run_down;
53static struct g_bioq g_bio_run_up;
54static struct g_bioq g_bio_run_task;
55
56static u_int pace;
57static uma_zone_t	biozone;
58
59#include <machine/atomic.h>
60
61static void
62g_bioq_lock(struct g_bioq *bq)
63{
64
65	mtx_lock(&bq->bio_queue_lock);
66}
67
68static void
69g_bioq_unlock(struct g_bioq *bq)
70{
71
72	mtx_unlock(&bq->bio_queue_lock);
73}
74
75#if 0
76static void
77g_bioq_destroy(struct g_bioq *bq)
78{
79
80	mtx_destroy(&bq->bio_queue_lock);
81}
82#endif
83
84static void
85g_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
92static struct bio *
93g_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
108struct bio *
109g_new_bio(void)
110{
111	struct bio *bp;
112
113	bp = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
114	return (bp);
115}
116
117struct bio *
118g_alloc_bio(void)
119{
120	struct bio *bp;
121
122	bp = uma_zalloc(biozone, M_WAITOK | M_ZERO);
123	return (bp);
124}
125
126void
127g_destroy_bio(struct bio *bp)
128{
129
130	uma_zfree(biozone, bp);
131}
132
133struct bio *
134g_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
151void
152g_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
164int
165g_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
184static int
185g_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
237void
238g_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	if (bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) {
249		KASSERT(bp->bio_offset % cp->provider->sectorsize == 0,
250		    ("wrong offset %jd for sectorsize %u",
251		    bp->bio_offset, cp->provider->sectorsize));
252		KASSERT(bp->bio_length % cp->provider->sectorsize == 0,
253		    ("wrong length %jd for sectorsize %u",
254		    bp->bio_length, cp->provider->sectorsize));
255	}
256
257	bp->bio_from = cp;
258	bp->bio_to = pp;
259	bp->bio_error = 0;
260	bp->bio_completed = 0;
261
262	KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
263	    ("Bio already on queue bp=%p", bp));
264	bp->bio_flags |= BIO_ONQUEUE;
265
266	binuptime(&bp->bio_t0);
267	if (g_collectstats & 4)
268		g_bioq_lock(&g_bio_run_down);
269	if (g_collectstats & 1)
270		devstat_start_transaction(pp->stat, &bp->bio_t0);
271	if (g_collectstats & 2)
272		devstat_start_transaction(cp->stat, &bp->bio_t0);
273
274	if (!(g_collectstats & 4))
275		g_bioq_lock(&g_bio_run_down);
276	pp->nstart++;
277	cp->nstart++;
278	TAILQ_INSERT_TAIL(&g_bio_run_down.bio_queue, bp, bio_queue);
279	g_bio_run_down.bio_queue_length++;
280	g_bioq_unlock(&g_bio_run_down);
281
282	/* Pass it on down. */
283	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
284	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
285	wakeup(&g_wait_down);
286}
287
288void
289g_io_deliver(struct bio *bp, int error)
290{
291	struct g_consumer *cp;
292	struct g_provider *pp;
293
294	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
295	pp = bp->bio_to;
296	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
297	cp = bp->bio_from;
298	if (cp == NULL) {
299		bp->bio_error = error;
300		bp->bio_done(bp);
301		return;
302	}
303	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
304	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
305	KASSERT(bp->bio_completed >= 0, ("bio_completed can't be less than 0"));
306	KASSERT(bp->bio_completed <= bp->bio_length,
307	    ("bio_completed can't be greater than bio_length"));
308
309	g_trace(G_T_BIO,
310"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
311	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
312	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
313
314	KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
315	    ("Bio already on queue bp=%p", bp));
316	bp->bio_flags |= BIO_ONQUEUE;
317
318	/*
319	 * XXX: next two doesn't belong here
320	 */
321	bp->bio_bcount = bp->bio_length;
322	bp->bio_resid = bp->bio_bcount - bp->bio_completed;
323
324	if (g_collectstats & 4)
325		g_bioq_lock(&g_bio_run_up);
326	if (g_collectstats & 1)
327		devstat_end_transaction_bio(pp->stat, bp);
328	if (g_collectstats & 2)
329		devstat_end_transaction_bio(cp->stat, bp);
330	if (!(g_collectstats & 4))
331		g_bioq_lock(&g_bio_run_up);
332	cp->nend++;
333	pp->nend++;
334	if (error != ENOMEM) {
335		bp->bio_error = error;
336		TAILQ_INSERT_TAIL(&g_bio_run_up.bio_queue, bp, bio_queue);
337		g_bio_run_up.bio_queue_length++;
338		g_bioq_unlock(&g_bio_run_up);
339		wakeup(&g_wait_up);
340		return;
341	}
342	g_bioq_unlock(&g_bio_run_up);
343
344	if (bootverbose)
345		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
346	bp->bio_children = 0;
347	bp->bio_inbed = 0;
348	g_io_request(bp, cp);
349	pace++;
350	return;
351}
352
353void
354g_io_schedule_down(struct thread *tp __unused)
355{
356	struct bio *bp;
357	off_t excess;
358	int error;
359#ifdef WITNESS
360	struct mtx mymutex;
361
362	bzero(&mymutex, sizeof mymutex);
363	mtx_init(&mymutex, "g_xdown", NULL, MTX_DEF);
364#endif
365
366	for(;;) {
367		g_bioq_lock(&g_bio_run_down);
368		bp = g_bioq_first(&g_bio_run_down);
369		if (bp == NULL) {
370			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
371			    PRIBIO | PDROP, "-", hz/10);
372			continue;
373		}
374		g_bioq_unlock(&g_bio_run_down);
375		if (pace > 0) {
376			msleep(&error, NULL, PRIBIO, "g_down", hz/10);
377			pace--;
378		}
379		error = g_io_check(bp);
380		if (error) {
381			g_io_deliver(bp, error);
382			continue;
383		}
384		switch (bp->bio_cmd) {
385		case BIO_READ:
386		case BIO_WRITE:
387		case BIO_DELETE:
388			/* Truncate requests to the end of providers media. */
389			excess = bp->bio_offset + bp->bio_length;
390			if (excess > bp->bio_to->mediasize) {
391				excess -= bp->bio_to->mediasize;
392				bp->bio_length -= excess;
393			}
394			/* Deliver zero length transfers right here. */
395			if (bp->bio_length == 0) {
396				g_io_deliver(bp, 0);
397				continue;
398			}
399			break;
400		default:
401			break;
402		}
403#ifdef WITNESS
404		mtx_lock(&mymutex);
405#endif
406		bp->bio_to->geom->start(bp);
407#ifdef WITNESS
408		mtx_unlock(&mymutex);
409#endif
410	}
411}
412
413void
414bio_taskqueue(struct bio *bp, bio_task_t *func, void *arg)
415{
416	bp->bio_task = func;
417	bp->bio_task_arg = arg;
418	/*
419	 * The taskqueue is actually just a second queue off the "up"
420	 * queue, so we use the same lock.
421	 */
422	g_bioq_lock(&g_bio_run_up);
423	KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
424	    ("Bio already on queue bp=%p target taskq", bp));
425	bp->bio_flags |= BIO_ONQUEUE;
426	TAILQ_INSERT_TAIL(&g_bio_run_task.bio_queue, bp, bio_queue);
427	g_bio_run_task.bio_queue_length++;
428	wakeup(&g_wait_up);
429	g_bioq_unlock(&g_bio_run_up);
430}
431
432
433void
434g_io_schedule_up(struct thread *tp __unused)
435{
436	struct bio *bp;
437#ifdef WITNESS
438	struct mtx mymutex;
439
440	bzero(&mymutex, sizeof mymutex);
441	mtx_init(&mymutex, "g_xup", NULL, MTX_DEF);
442#endif
443	for(;;) {
444		g_bioq_lock(&g_bio_run_up);
445		bp = g_bioq_first(&g_bio_run_task);
446		if (bp != NULL) {
447			g_bioq_unlock(&g_bio_run_up);
448#ifdef WITNESS
449			mtx_lock(&mymutex);
450#endif
451			bp->bio_task(bp->bio_task_arg);
452#ifdef WITNESS
453			mtx_unlock(&mymutex);
454#endif
455			continue;
456		}
457		bp = g_bioq_first(&g_bio_run_up);
458		if (bp != NULL) {
459			g_bioq_unlock(&g_bio_run_up);
460#ifdef WITNESS
461			mtx_lock(&mymutex);
462#endif
463			biodone(bp);
464#ifdef WITNESS
465			mtx_unlock(&mymutex);
466#endif
467			continue;
468		}
469		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
470		    PRIBIO | PDROP, "-", hz/10);
471	}
472}
473
474void *
475g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
476{
477	struct bio *bp;
478	void *ptr;
479	int errorc;
480
481	KASSERT(length > 0 && length >= cp->provider->sectorsize &&
482	    length <= MAXPHYS, ("g_read_data(): invalid length %jd",
483	    (intmax_t)length));
484
485	bp = g_alloc_bio();
486	bp->bio_cmd = BIO_READ;
487	bp->bio_done = NULL;
488	bp->bio_offset = offset;
489	bp->bio_length = length;
490	ptr = g_malloc(length, M_WAITOK);
491	bp->bio_data = ptr;
492	g_io_request(bp, cp);
493	errorc = biowait(bp, "gread");
494	if (error != NULL)
495		*error = errorc;
496	g_destroy_bio(bp);
497	if (errorc) {
498		g_free(ptr);
499		ptr = NULL;
500	}
501	return (ptr);
502}
503
504int
505g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
506{
507	struct bio *bp;
508	int error;
509
510	KASSERT(length > 0 && length >= cp->provider->sectorsize &&
511	    length <= MAXPHYS, ("g_write_data(): invalid length %jd",
512	    (intmax_t)length));
513
514	bp = g_alloc_bio();
515	bp->bio_cmd = BIO_WRITE;
516	bp->bio_done = NULL;
517	bp->bio_offset = offset;
518	bp->bio_length = length;
519	bp->bio_data = ptr;
520	g_io_request(bp, cp);
521	error = biowait(bp, "gwrite");
522	g_destroy_bio(bp);
523	return (error);
524}
525
526void
527g_print_bio(struct bio *bp)
528{
529	const char *pname, *cmd = NULL;
530
531	if (bp->bio_to != NULL)
532		pname = bp->bio_to->name;
533	else
534		pname = "[unknown]";
535
536	switch (bp->bio_cmd) {
537	case BIO_GETATTR:
538		cmd = "GETATTR";
539		printf("%s[%s(attr=%s)]", pname, cmd, bp->bio_attribute);
540		return;
541	case BIO_READ:
542		cmd = "READ";
543	case BIO_WRITE:
544		if (cmd == NULL)
545			cmd = "WRITE";
546	case BIO_DELETE:
547		if (cmd == NULL)
548			cmd = "DELETE";
549		printf("%s[%s(offset=%jd, length=%jd)]", pname, cmd,
550		    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
551		return;
552	default:
553		cmd = "UNKNOWN";
554		printf("%s[%s()]", pname, cmd);
555		return;
556	}
557	/* NOTREACHED */
558}
559