geom_io.c revision 136399 
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 136399 2004-10-11 21:22:59Z ups $");
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	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
258	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
259
260	bp->bio_from = cp;
261	bp->bio_to = pp;
262	bp->bio_error = 0;
263	bp->bio_completed = 0;
264
265	KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
266	    ("Bio already on queue bp=%p", bp));
267	bp->bio_flags |= BIO_ONQUEUE;
268
269	binuptime(&bp->bio_t0);
270	if (g_collectstats & 4)
271		g_bioq_lock(&g_bio_run_down);
272	if (g_collectstats & 1)
273		devstat_start_transaction(pp->stat, &bp->bio_t0);
274	if (g_collectstats & 2)
275		devstat_start_transaction(cp->stat, &bp->bio_t0);
276
277	if (!(g_collectstats & 4))
278		g_bioq_lock(&g_bio_run_down);
279	pp->nstart++;
280	cp->nstart++;
281	TAILQ_INSERT_TAIL(&g_bio_run_down.bio_queue, bp, bio_queue);
282	g_bio_run_down.bio_queue_length++;
283	g_bioq_unlock(&g_bio_run_down);
284
285	/* Pass it on down. */
286	wakeup(&g_wait_down);
287}
288
289void
290g_io_deliver(struct bio *bp, int error)
291{
292	struct g_consumer *cp;
293	struct g_provider *pp;
294
295	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
296	pp = bp->bio_to;
297	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
298	cp = bp->bio_from;
299	if (cp == NULL) {
300		bp->bio_error = error;
301		bp->bio_done(bp);
302		return;
303	}
304	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
305	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
306	KASSERT(bp->bio_completed >= 0, ("bio_completed can't be less than 0"));
307	KASSERT(bp->bio_completed <= bp->bio_length,
308	    ("bio_completed can't be greater than bio_length"));
309
310	g_trace(G_T_BIO,
311"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
312	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
313	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
314
315	KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
316	    ("Bio already on queue bp=%p", bp));
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		bp->bio_flags |= BIO_ONQUEUE;
338		g_bio_run_up.bio_queue_length++;
339		g_bioq_unlock(&g_bio_run_up);
340		wakeup(&g_wait_up);
341		return;
342	}
343	g_bioq_unlock(&g_bio_run_up);
344
345	if (bootverbose)
346		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
347	bp->bio_children = 0;
348	bp->bio_inbed = 0;
349	g_io_request(bp, cp);
350	pace++;
351	return;
352}
353
354void
355g_io_schedule_down(struct thread *tp __unused)
356{
357	struct bio *bp;
358	off_t excess;
359	int error;
360#ifdef WITNESS
361	struct mtx mymutex;
362
363	bzero(&mymutex, sizeof mymutex);
364	mtx_init(&mymutex, "g_xdown", NULL, MTX_DEF);
365#endif
366
367	for(;;) {
368		g_bioq_lock(&g_bio_run_down);
369		bp = g_bioq_first(&g_bio_run_down);
370		if (bp == NULL) {
371			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
372			    PRIBIO | PDROP, "-", hz/10);
373			continue;
374		}
375		g_bioq_unlock(&g_bio_run_down);
376		if (pace > 0) {
377			msleep(&error, NULL, PRIBIO, "g_down", hz/10);
378			pace--;
379		}
380		error = g_io_check(bp);
381		if (error) {
382			g_io_deliver(bp, error);
383			continue;
384		}
385		switch (bp->bio_cmd) {
386		case BIO_READ:
387		case BIO_WRITE:
388		case BIO_DELETE:
389			/* Truncate requests to the end of providers media. */
390			excess = bp->bio_offset + bp->bio_length;
391			if (excess > bp->bio_to->mediasize) {
392				excess -= bp->bio_to->mediasize;
393				bp->bio_length -= excess;
394			}
395			/* Deliver zero length transfers right here. */
396			if (bp->bio_length == 0) {
397				g_io_deliver(bp, 0);
398				continue;
399			}
400			break;
401		default:
402			break;
403		}
404#ifdef WITNESS
405		mtx_lock(&mymutex);
406#endif
407		bp->bio_to->geom->start(bp);
408#ifdef WITNESS
409		mtx_unlock(&mymutex);
410#endif
411	}
412}
413
414void
415bio_taskqueue(struct bio *bp, bio_task_t *func, void *arg)
416{
417	bp->bio_task = func;
418	bp->bio_task_arg = arg;
419	/*
420	 * The taskqueue is actually just a second queue off the "up"
421	 * queue, so we use the same lock.
422	 */
423	g_bioq_lock(&g_bio_run_up);
424	KASSERT(!(bp->bio_flags & BIO_ONQUEUE),
425	    ("Bio already on queue bp=%p target taskq", bp));
426	bp->bio_flags |= BIO_ONQUEUE;
427	TAILQ_INSERT_TAIL(&g_bio_run_task.bio_queue, bp, bio_queue);
428	g_bio_run_task.bio_queue_length++;
429	wakeup(&g_wait_up);
430	g_bioq_unlock(&g_bio_run_up);
431}
432
433
434void
435g_io_schedule_up(struct thread *tp __unused)
436{
437	struct bio *bp;
438#ifdef WITNESS
439	struct mtx mymutex;
440
441	bzero(&mymutex, sizeof mymutex);
442	mtx_init(&mymutex, "g_xup", NULL, MTX_DEF);
443#endif
444	for(;;) {
445		g_bioq_lock(&g_bio_run_up);
446		bp = g_bioq_first(&g_bio_run_task);
447		if (bp != NULL) {
448			g_bioq_unlock(&g_bio_run_up);
449#ifdef WITNESS
450			mtx_lock(&mymutex);
451#endif
452			bp->bio_task(bp->bio_task_arg);
453#ifdef WITNESS
454			mtx_unlock(&mymutex);
455#endif
456			continue;
457		}
458		bp = g_bioq_first(&g_bio_run_up);
459		if (bp != NULL) {
460			g_bioq_unlock(&g_bio_run_up);
461#ifdef WITNESS
462			mtx_lock(&mymutex);
463#endif
464			biodone(bp);
465#ifdef WITNESS
466			mtx_unlock(&mymutex);
467#endif
468			continue;
469		}
470		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
471		    PRIBIO | PDROP, "-", hz/10);
472	}
473}
474
475void *
476g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
477{
478	struct bio *bp;
479	void *ptr;
480	int errorc;
481
482	KASSERT(length > 0 && length >= cp->provider->sectorsize &&
483	    length <= MAXPHYS, ("g_read_data(): invalid length %jd",
484	    (intmax_t)length));
485
486	bp = g_alloc_bio();
487	bp->bio_cmd = BIO_READ;
488	bp->bio_done = NULL;
489	bp->bio_offset = offset;
490	bp->bio_length = length;
491	ptr = g_malloc(length, M_WAITOK);
492	bp->bio_data = ptr;
493	g_io_request(bp, cp);
494	errorc = biowait(bp, "gread");
495	if (error != NULL)
496		*error = errorc;
497	g_destroy_bio(bp);
498	if (errorc) {
499		g_free(ptr);
500		ptr = NULL;
501	}
502	return (ptr);
503}
504
505int
506g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
507{
508	struct bio *bp;
509	int error;
510
511	KASSERT(length > 0 && length >= cp->provider->sectorsize &&
512	    length <= MAXPHYS, ("g_write_data(): invalid length %jd",
513	    (intmax_t)length));
514
515	bp = g_alloc_bio();
516	bp->bio_cmd = BIO_WRITE;
517	bp->bio_done = NULL;
518	bp->bio_offset = offset;
519	bp->bio_length = length;
520	bp->bio_data = ptr;
521	g_io_request(bp, cp);
522	error = biowait(bp, "gwrite");
523	g_destroy_bio(bp);
524	return (error);
525}
526
527void
528g_print_bio(struct bio *bp)
529{
530	const char *pname, *cmd = NULL;
531
532	if (bp->bio_to != NULL)
533		pname = bp->bio_to->name;
534	else
535		pname = "[unknown]";
536
537	switch (bp->bio_cmd) {
538	case BIO_GETATTR:
539		cmd = "GETATTR";
540		printf("%s[%s(attr=%s)]", pname, cmd, bp->bio_attribute);
541		return;
542	case BIO_READ:
543		cmd = "READ";
544	case BIO_WRITE:
545		if (cmd == NULL)
546			cmd = "WRITE";
547	case BIO_DELETE:
548		if (cmd == NULL)
549			cmd = "DELETE";
550		printf("%s[%s(offset=%jd, length=%jd)]", pname, cmd,
551		    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
552		return;
553	default:
554		cmd = "UNKNOWN";
555		printf("%s[%s()]", pname, cmd);
556		return;
557	}
558	/* NOTREACHED */
559}
560