geom_io.c revision 134379 
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 134379 2004-08-27 14:43:11Z 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		TAILQ_REMOVE(&bq->bio_queue, bp, bio_queue);
100		bq->bio_queue_length--;
101	}
102	return (bp);
103}
104
105static void
106g_bioq_enqueue_tail(struct bio *bp, struct g_bioq *rq)
107{
108
109	g_bioq_lock(rq);
110	TAILQ_INSERT_TAIL(&rq->bio_queue, bp, bio_queue);
111	rq->bio_queue_length++;
112	g_bioq_unlock(rq);
113}
114
115struct bio *
116g_new_bio(void)
117{
118	struct bio *bp;
119
120	bp = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
121	return (bp);
122}
123
124struct bio *
125g_alloc_bio(void)
126{
127	struct bio *bp;
128
129	bp = uma_zalloc(biozone, M_WAITOK | M_ZERO);
130	return (bp);
131}
132
133void
134g_destroy_bio(struct bio *bp)
135{
136
137	uma_zfree(biozone, bp);
138}
139
140struct bio *
141g_clone_bio(struct bio *bp)
142{
143	struct bio *bp2;
144
145	bp2 = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
146	if (bp2 != NULL) {
147		bp2->bio_parent = bp;
148		bp2->bio_cmd = bp->bio_cmd;
149		bp2->bio_length = bp->bio_length;
150		bp2->bio_offset = bp->bio_offset;
151		bp2->bio_data = bp->bio_data;
152		bp2->bio_attribute = bp->bio_attribute;
153		bp->bio_children++;
154	}
155	return(bp2);
156}
157
158void
159g_io_init()
160{
161
162	g_bioq_init(&g_bio_run_down);
163	g_bioq_init(&g_bio_run_up);
164	g_bioq_init(&g_bio_run_task);
165	biozone = uma_zcreate("g_bio", sizeof (struct bio),
166	    NULL, NULL,
167	    NULL, NULL,
168	    0, 0);
169}
170
171int
172g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
173{
174	struct bio *bp;
175	int error;
176
177	g_trace(G_T_BIO, "bio_getattr(%s)", attr);
178	bp = g_alloc_bio();
179	bp->bio_cmd = BIO_GETATTR;
180	bp->bio_done = NULL;
181	bp->bio_attribute = attr;
182	bp->bio_length = *len;
183	bp->bio_data = ptr;
184	g_io_request(bp, cp);
185	error = biowait(bp, "ggetattr");
186	*len = bp->bio_completed;
187	g_destroy_bio(bp);
188	return (error);
189}
190
191static int
192g_io_check(struct bio *bp)
193{
194	struct g_consumer *cp;
195	struct g_provider *pp;
196
197	cp = bp->bio_from;
198	pp = bp->bio_to;
199
200	/* Fail if access counters dont allow the operation */
201	switch(bp->bio_cmd) {
202	case BIO_READ:
203	case BIO_GETATTR:
204		if (cp->acr == 0)
205			return (EPERM);
206		break;
207	case BIO_WRITE:
208	case BIO_DELETE:
209		if (cp->acw == 0)
210			return (EPERM);
211		break;
212	default:
213		return (EPERM);
214	}
215	/* if provider is marked for error, don't disturb. */
216	if (pp->error)
217		return (pp->error);
218
219	switch(bp->bio_cmd) {
220	case BIO_READ:
221	case BIO_WRITE:
222	case BIO_DELETE:
223		/* Zero sectorsize is a probably lack of media */
224		if (pp->sectorsize == 0)
225			return (ENXIO);
226		/* Reject I/O not on sector boundary */
227		if (bp->bio_offset % pp->sectorsize)
228			return (EINVAL);
229		/* Reject I/O not integral sector long */
230		if (bp->bio_length % pp->sectorsize)
231			return (EINVAL);
232		/* Reject requests before or past the end of media. */
233		if (bp->bio_offset < 0)
234			return (EIO);
235		if (bp->bio_offset > pp->mediasize)
236			return (EIO);
237		break;
238	default:
239		break;
240	}
241	return (0);
242}
243
244void
245g_io_request(struct bio *bp, struct g_consumer *cp)
246{
247	struct g_provider *pp;
248
249	KASSERT(cp != NULL, ("NULL cp in g_io_request"));
250	KASSERT(bp != NULL, ("NULL bp in g_io_request"));
251	KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
252	pp = cp->provider;
253	KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
254
255	bp->bio_from = cp;
256	bp->bio_to = pp;
257	bp->bio_error = 0;
258	bp->bio_completed = 0;
259
260	if (g_collectstats & 1)
261		devstat_start_transaction_bio(pp->stat, bp);
262	pp->nstart++;
263	if (g_collectstats & 2)
264		devstat_start_transaction_bio(cp->stat, bp);
265	cp->nstart++;
266
267	/* Pass it on down. */
268	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
269	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
270	g_bioq_enqueue_tail(bp, &g_bio_run_down);
271	wakeup(&g_wait_down);
272}
273
274void
275g_io_deliver(struct bio *bp, int error)
276{
277	struct g_consumer *cp;
278	struct g_provider *pp;
279
280	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
281	pp = bp->bio_to;
282	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
283	cp = bp->bio_from;
284	if (cp == NULL) {
285		bp->bio_error = error;
286		bp->bio_done(bp);
287		return;
288	}
289	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
290	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
291	KASSERT(bp->bio_completed >= 0, ("bio_completed can't be less than 0"));
292	KASSERT(bp->bio_completed <= bp->bio_length,
293	    ("bio_completed can't be greater than bio_length"));
294
295	g_trace(G_T_BIO,
296"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
297	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
298	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
299
300	bp->bio_bcount = bp->bio_length;
301	bp->bio_resid = bp->bio_bcount - bp->bio_completed;
302	if (g_collectstats & 1)
303		devstat_end_transaction_bio(pp->stat, bp);
304	if (g_collectstats & 2)
305		devstat_end_transaction_bio(cp->stat, bp);
306	cp->nend++;
307	pp->nend++;
308
309	if (error == ENOMEM) {
310		if (bootverbose)
311			printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
312		bp->bio_children = 0;
313		bp->bio_inbed = 0;
314		g_io_request(bp, cp);
315		pace++;
316		return;
317	}
318	bp->bio_error = error;
319	g_bioq_enqueue_tail(bp, &g_bio_run_up);
320	wakeup(&g_wait_up);
321}
322
323void
324g_io_schedule_down(struct thread *tp __unused)
325{
326	struct bio *bp;
327	off_t excess;
328	int error;
329#ifdef WITNESS
330	struct mtx mymutex;
331
332	bzero(&mymutex, sizeof mymutex);
333	mtx_init(&mymutex, "g_xdown", NULL, MTX_DEF);
334#endif
335
336	for(;;) {
337		g_bioq_lock(&g_bio_run_down);
338		bp = g_bioq_first(&g_bio_run_down);
339		if (bp == NULL) {
340			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
341			    PRIBIO | PDROP, "-", hz/10);
342			continue;
343		}
344		g_bioq_unlock(&g_bio_run_down);
345		if (pace > 0) {
346			msleep(&error, NULL, PRIBIO, "g_down", hz/10);
347			pace--;
348		}
349		error = g_io_check(bp);
350		if (error) {
351			g_io_deliver(bp, error);
352			continue;
353		}
354		switch (bp->bio_cmd) {
355		case BIO_READ:
356		case BIO_WRITE:
357		case BIO_DELETE:
358			/* Truncate requests to the end of providers media. */
359			excess = bp->bio_offset + bp->bio_length;
360			if (excess > bp->bio_to->mediasize) {
361				excess -= bp->bio_to->mediasize;
362				bp->bio_length -= excess;
363			}
364			/* Deliver zero length transfers right here. */
365			if (bp->bio_length == 0) {
366				g_io_deliver(bp, 0);
367				continue;
368			}
369			break;
370		default:
371			break;
372		}
373#ifdef WITNESS
374		mtx_lock(&mymutex);
375#endif
376		bp->bio_to->geom->start(bp);
377#ifdef WITNESS
378		mtx_unlock(&mymutex);
379#endif
380	}
381}
382
383void
384bio_taskqueue(struct bio *bp, bio_task_t *func, void *arg)
385{
386	bp->bio_task = func;
387	bp->bio_task_arg = arg;
388	/*
389	 * The taskqueue is actually just a second queue off the "up"
390	 * queue, so we use the same lock.
391	 */
392	g_bioq_lock(&g_bio_run_up);
393	TAILQ_INSERT_TAIL(&g_bio_run_task.bio_queue, bp, bio_queue);
394	g_bio_run_task.bio_queue_length++;
395	wakeup(&g_wait_up);
396	g_bioq_unlock(&g_bio_run_up);
397}
398
399
400void
401g_io_schedule_up(struct thread *tp __unused)
402{
403	struct bio *bp;
404#ifdef WITNESS
405	struct mtx mymutex;
406
407	bzero(&mymutex, sizeof mymutex);
408	mtx_init(&mymutex, "g_xup", NULL, MTX_DEF);
409#endif
410	for(;;) {
411		g_bioq_lock(&g_bio_run_up);
412		bp = g_bioq_first(&g_bio_run_task);
413		if (bp != NULL) {
414			g_bioq_unlock(&g_bio_run_up);
415#ifdef WITNESS
416			mtx_lock(&mymutex);
417#endif
418			bp->bio_task(bp->bio_task_arg);
419#ifdef WITNESS
420			mtx_unlock(&mymutex);
421#endif
422			continue;
423		}
424		bp = g_bioq_first(&g_bio_run_up);
425		if (bp != NULL) {
426			g_bioq_unlock(&g_bio_run_up);
427#ifdef WITNESS
428			mtx_lock(&mymutex);
429#endif
430			biodone(bp);
431#ifdef WITNESS
432			mtx_unlock(&mymutex);
433#endif
434			continue;
435		}
436		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
437		    PRIBIO | PDROP, "-", hz/10);
438	}
439}
440
441void *
442g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
443{
444	struct bio *bp;
445	void *ptr;
446	int errorc;
447
448	KASSERT(length >= 512 && length <= DFLTPHYS,
449		("g_read_data(): invalid length %jd", (intmax_t)length));
450
451	bp = g_alloc_bio();
452	bp->bio_cmd = BIO_READ;
453	bp->bio_done = NULL;
454	bp->bio_offset = offset;
455	bp->bio_length = length;
456	ptr = g_malloc(length, M_WAITOK);
457	bp->bio_data = ptr;
458	g_io_request(bp, cp);
459	errorc = biowait(bp, "gread");
460	if (error != NULL)
461		*error = errorc;
462	g_destroy_bio(bp);
463	if (errorc) {
464		g_free(ptr);
465		ptr = NULL;
466	}
467	return (ptr);
468}
469
470int
471g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
472{
473	struct bio *bp;
474	int error;
475
476	KASSERT(length >= 512 && length <= DFLTPHYS,
477		("g_write_data(): invalid length %jd", (intmax_t)length));
478
479	bp = g_alloc_bio();
480	bp->bio_cmd = BIO_WRITE;
481	bp->bio_done = NULL;
482	bp->bio_offset = offset;
483	bp->bio_length = length;
484	bp->bio_data = ptr;
485	g_io_request(bp, cp);
486	error = biowait(bp, "gwrite");
487	g_destroy_bio(bp);
488	return (error);
489}
490
491void
492g_print_bio(struct bio *bp)
493{
494	const char *pname, *cmd = NULL;
495
496	if (bp->bio_to != NULL)
497		pname = bp->bio_to->name;
498	else
499		pname = "[unknown]";
500
501	switch (bp->bio_cmd) {
502	case BIO_GETATTR:
503		cmd = "GETATTR";
504		printf("%s[%s(attr=%s)]", pname, cmd, bp->bio_attribute);
505		return;
506	case BIO_READ:
507		cmd = "READ";
508	case BIO_WRITE:
509		if (cmd == NULL)
510			cmd = "WRITE";
511	case BIO_DELETE:
512		if (cmd == NULL)
513			cmd = "DELETE";
514		printf("%s[%s(offset=%jd, length=%jd)]", pname, cmd,
515		    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
516		return;
517	default:
518		cmd = "UNKNOWN";
519		printf("%s[%s()]", pname, cmd);
520		return;
521	}
522	/* NOTREACHED */
523}
524