geom_io.c revision 113032 
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 * $FreeBSD: head/sys/geom/geom_io.c 113032 2003-04-03 19:19:36Z phk $
36 */
37
38
39#include <sys/param.h>
40#ifndef _KERNEL
41#include <stdio.h>
42#include <string.h>
43#include <stdlib.h>
44#include <signal.h>
45#include <err.h>
46#include <sched.h>
47#else
48#include <sys/systm.h>
49#include <sys/kernel.h>
50#include <sys/malloc.h>
51#include <sys/bio.h>
52#endif
53
54#include <sys/errno.h>
55#include <geom/geom.h>
56#include <geom/geom_int.h>
57#include <sys/devicestat.h>
58
59static struct g_bioq g_bio_run_down;
60static struct g_bioq g_bio_run_up;
61static struct g_bioq g_bio_run_task;
62static struct g_bioq g_bio_idle;
63
64static u_int pace;
65
66#include <machine/atomic.h>
67
68static void
69g_bioq_lock(struct g_bioq *bq)
70{
71
72	mtx_lock(&bq->bio_queue_lock);
73}
74
75static void
76g_bioq_unlock(struct g_bioq *bq)
77{
78
79	mtx_unlock(&bq->bio_queue_lock);
80}
81
82#if 0
83static void
84g_bioq_destroy(struct g_bioq *bq)
85{
86
87	mtx_destroy(&bq->bio_queue_lock);
88}
89#endif
90
91static void
92g_bioq_init(struct g_bioq *bq)
93{
94
95	TAILQ_INIT(&bq->bio_queue);
96	mtx_init(&bq->bio_queue_lock, "bio queue", NULL, MTX_DEF);
97}
98
99static struct bio *
100g_bioq_first(struct g_bioq *bq)
101{
102	struct bio *bp;
103
104	bp = TAILQ_FIRST(&bq->bio_queue);
105	if (bp != NULL) {
106		TAILQ_REMOVE(&bq->bio_queue, bp, bio_queue);
107		bq->bio_queue_length--;
108	}
109	return (bp);
110}
111
112static void
113g_bioq_enqueue_tail(struct bio *bp, struct g_bioq *rq)
114{
115
116	g_bioq_lock(rq);
117	TAILQ_INSERT_TAIL(&rq->bio_queue, bp, bio_queue);
118	rq->bio_queue_length++;
119	g_bioq_unlock(rq);
120}
121
122struct bio *
123g_new_bio(void)
124{
125	struct bio *bp;
126
127	g_bioq_lock(&g_bio_idle);
128	bp = g_bioq_first(&g_bio_idle);
129	g_bioq_unlock(&g_bio_idle);
130	if (bp == NULL)
131		bp = g_malloc(sizeof *bp, M_NOWAIT | M_ZERO);
132	/* g_trace(G_T_BIO, "g_new_bio() = %p", bp); */
133	return (bp);
134}
135
136void
137g_destroy_bio(struct bio *bp)
138{
139
140	/* g_trace(G_T_BIO, "g_destroy_bio(%p)", bp); */
141	bzero(bp, sizeof *bp);
142	g_bioq_enqueue_tail(bp, &g_bio_idle);
143}
144
145struct bio *
146g_clone_bio(struct bio *bp)
147{
148	struct bio *bp2;
149
150	bp2 = g_new_bio();
151	if (bp2 != NULL) {
152		bp2->bio_parent = bp;
153		bp2->bio_cmd = bp->bio_cmd;
154		bp2->bio_length = bp->bio_length;
155		bp2->bio_offset = bp->bio_offset;
156		bp2->bio_data = bp->bio_data;
157		bp2->bio_attribute = bp->bio_attribute;
158		bp->bio_children++;
159	}
160	/* g_trace(G_T_BIO, "g_clone_bio(%p) = %p", bp, bp2); */
161	return(bp2);
162}
163
164void
165g_io_init()
166{
167
168	g_bioq_init(&g_bio_run_down);
169	g_bioq_init(&g_bio_run_up);
170	g_bioq_init(&g_bio_run_task);
171	g_bioq_init(&g_bio_idle);
172}
173
174int
175g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
176{
177	struct bio *bp;
178	int error;
179
180	g_trace(G_T_BIO, "bio_getattr(%s)", attr);
181	bp = g_new_bio();
182	bp->bio_cmd = BIO_GETATTR;
183	bp->bio_done = NULL;
184	bp->bio_attribute = attr;
185	bp->bio_length = *len;
186	bp->bio_data = ptr;
187	g_io_request(bp, cp);
188	error = biowait(bp, "ggetattr");
189	*len = bp->bio_completed;
190	g_destroy_bio(bp);
191	return (error);
192}
193
194static int
195g_io_check(struct bio *bp)
196{
197	struct g_consumer *cp;
198	struct g_provider *pp;
199
200	cp = bp->bio_from;
201	pp = bp->bio_to;
202
203	/* Fail if access counters dont allow the operation */
204	switch(bp->bio_cmd) {
205	case BIO_READ:
206	case BIO_GETATTR:
207		if (cp->acr == 0)
208			return (EPERM);
209		break;
210	case BIO_WRITE:
211	case BIO_DELETE:
212		if (cp->acw == 0)
213			return (EPERM);
214		break;
215	default:
216		return (EPERM);
217	}
218	/* if provider is marked for error, don't disturb. */
219	if (pp->error)
220		return (pp->error);
221
222	switch(bp->bio_cmd) {
223	case BIO_READ:
224	case BIO_WRITE:
225	case BIO_DELETE:
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 past the end of media. */
233		if (bp->bio_offset > pp->mediasize)
234			return (EIO);
235		break;
236	default:
237		break;
238	}
239	return (0);
240}
241
242void
243g_io_request(struct bio *bp, struct g_consumer *cp)
244{
245	struct g_provider *pp;
246
247	pp = cp->provider;
248	KASSERT(cp != NULL, ("NULL cp in g_io_request"));
249	KASSERT(bp != NULL, ("NULL bp in g_io_request"));
250	KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
251	KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
252
253	bp->bio_from = cp;
254	bp->bio_to = pp;
255	bp->bio_error = 0;
256	bp->bio_completed = 0;
257
258	if (g_collectstats) {
259		devstat_start_transaction_bio(cp->stat, bp);
260		devstat_start_transaction_bio(pp->stat, bp);
261	}
262	cp->nstart++;
263	pp->nstart++;
264
265	/* Pass it on down. */
266	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
267	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
268	g_bioq_enqueue_tail(bp, &g_bio_run_down);
269	wakeup(&g_wait_down);
270}
271
272void
273g_io_deliver(struct bio *bp, int error)
274{
275	struct g_consumer *cp;
276	struct g_provider *pp;
277
278	cp = bp->bio_from;
279	pp = bp->bio_to;
280	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
281	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
282	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
283	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
284
285	g_trace(G_T_BIO,
286"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
287	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
288	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
289
290	bp->bio_bcount = bp->bio_length;
291	if (g_collectstats) {
292		bp->bio_resid = bp->bio_bcount - bp->bio_completed;
293		devstat_end_transaction_bio(cp->stat, bp);
294		devstat_end_transaction_bio(pp->stat, bp);
295	}
296	cp->nend++;
297	pp->nend++;
298
299	if (error == ENOMEM) {
300		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
301		g_io_request(bp, cp);
302		pace++;
303		return;
304	}
305	bp->bio_error = error;
306	g_bioq_enqueue_tail(bp, &g_bio_run_up);
307	wakeup(&g_wait_up);
308}
309
310void
311g_io_schedule_down(struct thread *tp __unused)
312{
313	struct bio *bp;
314	off_t excess;
315	int error;
316	struct mtx mymutex;
317
318	bzero(&mymutex, sizeof mymutex);
319	mtx_init(&mymutex, "g_xdown", MTX_DEF, 0);
320
321	for(;;) {
322		g_bioq_lock(&g_bio_run_down);
323		bp = g_bioq_first(&g_bio_run_down);
324		if (bp == NULL) {
325			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
326			    PRIBIO | PDROP, "g_down", hz/10);
327			continue;
328		}
329		g_bioq_unlock(&g_bio_run_down);
330		if (pace > 0) {
331			msleep(&error, NULL, PRIBIO, "g_down", hz/10);
332			pace--;
333		}
334		error = g_io_check(bp);
335		if (error) {
336			g_io_deliver(bp, error);
337			continue;
338		}
339		switch (bp->bio_cmd) {
340		case BIO_READ:
341		case BIO_WRITE:
342		case BIO_DELETE:
343			/* Truncate requests to the end of providers media. */
344			excess = bp->bio_offset + bp->bio_length;
345			if (excess > bp->bio_to->mediasize) {
346				excess -= bp->bio_to->mediasize;
347				bp->bio_length -= excess;
348			}
349			/* Deliver zero length transfers right here. */
350			if (bp->bio_length == 0) {
351				g_io_deliver(bp, 0);
352				continue;
353			}
354			break;
355		default:
356			break;
357		}
358		mtx_lock(&mymutex);
359		bp->bio_to->geom->start(bp);
360		mtx_unlock(&mymutex);
361	}
362}
363
364void
365bio_taskqueue(struct bio *bp, bio_task_t *func, void *arg)
366{
367	bp->bio_task = func;
368	bp->bio_task_arg = arg;
369	/*
370	 * The taskqueue is actually just a second queue off the "up"
371	 * queue, so we use the same lock.
372	 */
373	g_bioq_lock(&g_bio_run_up);
374	TAILQ_INSERT_TAIL(&g_bio_run_task.bio_queue, bp, bio_queue);
375	g_bio_run_task.bio_queue_length++;
376	wakeup(&g_wait_up);
377	g_bioq_unlock(&g_bio_run_up);
378}
379
380
381void
382g_io_schedule_up(struct thread *tp __unused)
383{
384	struct bio *bp;
385	struct mtx mymutex;
386
387	bzero(&mymutex, sizeof mymutex);
388	mtx_init(&mymutex, "g_xup", MTX_DEF, 0);
389	for(;;) {
390		g_bioq_lock(&g_bio_run_up);
391		bp = g_bioq_first(&g_bio_run_task);
392		if (bp != NULL) {
393			g_bioq_unlock(&g_bio_run_up);
394			mtx_lock(&mymutex);
395			bp->bio_task(bp, bp->bio_task_arg);
396			mtx_unlock(&mymutex);
397			continue;
398		}
399		bp = g_bioq_first(&g_bio_run_up);
400		if (bp != NULL) {
401			g_bioq_unlock(&g_bio_run_up);
402			mtx_lock(&mymutex);
403			biodone(bp);
404			mtx_unlock(&mymutex);
405			continue;
406		}
407		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
408		    PRIBIO | PDROP, "g_up", hz/10);
409	}
410}
411
412void *
413g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
414{
415	struct bio *bp;
416	void *ptr;
417	int errorc;
418
419	bp = g_new_bio();
420	bp->bio_cmd = BIO_READ;
421	bp->bio_done = NULL;
422	bp->bio_offset = offset;
423	bp->bio_length = length;
424	ptr = g_malloc(length, M_WAITOK);
425	bp->bio_data = ptr;
426	g_io_request(bp, cp);
427	errorc = biowait(bp, "gread");
428	if (error != NULL)
429		*error = errorc;
430	g_destroy_bio(bp);
431	if (errorc) {
432		g_free(ptr);
433		ptr = NULL;
434	}
435	return (ptr);
436}
437
438int
439g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
440{
441	struct bio *bp;
442	int error;
443
444	bp = g_new_bio();
445	bp->bio_cmd = BIO_WRITE;
446	bp->bio_done = NULL;
447	bp->bio_offset = offset;
448	bp->bio_length = length;
449	bp->bio_data = ptr;
450	g_io_request(bp, cp);
451	error = biowait(bp, "gwrite");
452	g_destroy_bio(bp);
453	return (error);
454}
455