geom_io.c revision 112370 
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 112370 2003-03-18 09:42:33Z 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_setattr(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_setattr(%s)", attr);
181	bp = g_new_bio();
182	bp->bio_cmd = BIO_SETATTR;
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, "gsetattr");
189	g_destroy_bio(bp);
190	return (error);
191}
192
193
194int
195g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
196{
197	struct bio *bp;
198	int error;
199
200	g_trace(G_T_BIO, "bio_getattr(%s)", attr);
201	bp = g_new_bio();
202	bp->bio_cmd = BIO_GETATTR;
203	bp->bio_done = NULL;
204	bp->bio_attribute = attr;
205	bp->bio_length = *len;
206	bp->bio_data = ptr;
207	g_io_request(bp, cp);
208	error = biowait(bp, "ggetattr");
209	*len = bp->bio_completed;
210	g_destroy_bio(bp);
211	return (error);
212}
213
214static int
215g_io_check(struct bio *bp)
216{
217	struct g_consumer *cp;
218	struct g_provider *pp;
219
220	cp = bp->bio_from;
221	pp = bp->bio_to;
222
223	/* Fail if access counters dont allow the operation */
224	switch(bp->bio_cmd) {
225	case BIO_READ:
226	case BIO_GETATTR:
227		if (cp->acr == 0)
228			return (EPERM);
229		break;
230	case BIO_WRITE:
231	case BIO_DELETE:
232	case BIO_SETATTR:
233		if (cp->acw == 0)
234			return (EPERM);
235		break;
236	default:
237		return (EPERM);
238	}
239	/* if provider is marked for error, don't disturb. */
240	if (pp->error)
241		return (pp->error);
242
243	switch(bp->bio_cmd) {
244	case BIO_READ:
245	case BIO_WRITE:
246	case BIO_DELETE:
247		/* Reject I/O not on sector boundary */
248		if (bp->bio_offset % pp->sectorsize)
249			return (EINVAL);
250		/* Reject I/O not integral sector long */
251		if (bp->bio_length % pp->sectorsize)
252			return (EINVAL);
253		/* Reject requests past the end of media. */
254		if (bp->bio_offset > pp->mediasize)
255			return (EIO);
256		break;
257	default:
258		break;
259	}
260	return (0);
261}
262
263void
264g_io_request(struct bio *bp, struct g_consumer *cp)
265{
266	struct g_provider *pp;
267
268	pp = cp->provider;
269	KASSERT(cp != NULL, ("NULL cp in g_io_request"));
270	KASSERT(bp != NULL, ("NULL bp in g_io_request"));
271	KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
272	KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
273
274	bp->bio_from = cp;
275	bp->bio_to = pp;
276	bp->bio_error = 0;
277	bp->bio_completed = 0;
278
279	if (g_collectstats) {
280		devstat_start_transaction_bio(cp->stat, bp);
281		devstat_start_transaction_bio(pp->stat, bp);
282	}
283	cp->nstart++;
284	pp->nstart++;
285
286	/* Pass it on down. */
287	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
288	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
289	g_bioq_enqueue_tail(bp, &g_bio_run_down);
290	wakeup(&g_wait_down);
291}
292
293void
294g_io_deliver(struct bio *bp, int error)
295{
296	struct g_consumer *cp;
297	struct g_provider *pp;
298
299	cp = bp->bio_from;
300	pp = bp->bio_to;
301	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
302	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
303	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
304	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
305
306	g_trace(G_T_BIO,
307"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
308	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
309	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
310
311	bp->bio_bcount = bp->bio_length;
312	if (g_collectstats) {
313		bp->bio_resid = bp->bio_bcount - bp->bio_completed;
314		devstat_end_transaction_bio(cp->stat, bp);
315		devstat_end_transaction_bio(pp->stat, bp);
316	}
317	cp->nend++;
318	pp->nend++;
319
320	if (error == ENOMEM) {
321		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
322		g_io_request(bp, cp);
323		pace++;
324		return;
325	}
326	bp->bio_error = error;
327	g_bioq_enqueue_tail(bp, &g_bio_run_up);
328	wakeup(&g_wait_up);
329}
330
331void
332g_io_schedule_down(struct thread *tp __unused)
333{
334	struct bio *bp;
335	off_t excess;
336	int error;
337	struct mtx mymutex;
338
339	bzero(&mymutex, sizeof mymutex);
340	mtx_init(&mymutex, "g_xdown", MTX_DEF, 0);
341
342	for(;;) {
343		g_bioq_lock(&g_bio_run_down);
344		bp = g_bioq_first(&g_bio_run_down);
345		if (bp == NULL) {
346			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
347			    PRIBIO | PDROP, "g_down", hz/10);
348			continue;
349		}
350		g_bioq_unlock(&g_bio_run_down);
351		error = g_io_check(bp);
352		if (error) {
353			g_io_deliver(bp, error);
354			continue;
355		}
356		switch (bp->bio_cmd) {
357		case BIO_READ:
358		case BIO_WRITE:
359		case BIO_DELETE:
360			/* Truncate requests to the end of providers media. */
361			excess = bp->bio_offset + bp->bio_length;
362			if (excess > bp->bio_to->mediasize) {
363				excess -= bp->bio_to->mediasize;
364				bp->bio_length -= excess;
365			}
366			/* Deliver zero length transfers right here. */
367			if (bp->bio_length == 0) {
368				g_io_deliver(bp, 0);
369				continue;
370			}
371			break;
372		default:
373			break;
374		}
375		mtx_lock(&mymutex);
376		bp->bio_to->geom->start(bp);
377		mtx_unlock(&mymutex);
378		if (pace) {
379			pace--;
380			break;
381		}
382	}
383}
384
385void
386bio_taskqueue(struct bio *bp, bio_task_t *func, void *arg)
387{
388	bp->bio_task = func;
389	bp->bio_task_arg = arg;
390	/*
391	 * The taskqueue is actually just a second queue off the "up"
392	 * queue, so we use the same lock.
393	 */
394	g_bioq_lock(&g_bio_run_up);
395	TAILQ_INSERT_TAIL(&g_bio_run_task.bio_queue, bp, bio_queue);
396	g_bio_run_task.bio_queue_length++;
397	wakeup(&g_wait_up);
398	g_bioq_unlock(&g_bio_run_up);
399}
400
401
402void
403g_io_schedule_up(struct thread *tp __unused)
404{
405	struct bio *bp;
406	struct mtx mymutex;
407
408	bzero(&mymutex, sizeof mymutex);
409	mtx_init(&mymutex, "g_xup", MTX_DEF, 0);
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			mtx_lock(&mymutex);
416			bp->bio_task(bp, bp->bio_task_arg);
417			mtx_unlock(&mymutex);
418			continue;
419		}
420		bp = g_bioq_first(&g_bio_run_up);
421		if (bp != NULL) {
422			g_bioq_unlock(&g_bio_run_up);
423			mtx_lock(&mymutex);
424			biodone(bp);
425			mtx_unlock(&mymutex);
426			continue;
427		}
428		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
429		    PRIBIO | PDROP, "g_up", hz/10);
430	}
431}
432
433void *
434g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
435{
436	struct bio *bp;
437	void *ptr;
438	int errorc;
439
440	bp = g_new_bio();
441	bp->bio_cmd = BIO_READ;
442	bp->bio_done = NULL;
443	bp->bio_offset = offset;
444	bp->bio_length = length;
445	ptr = g_malloc(length, M_WAITOK);
446	bp->bio_data = ptr;
447	g_io_request(bp, cp);
448	errorc = biowait(bp, "gread");
449	if (error != NULL)
450		*error = errorc;
451	g_destroy_bio(bp);
452	if (errorc) {
453		g_free(ptr);
454		ptr = NULL;
455	}
456	return (ptr);
457}
458
459int
460g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
461{
462	struct bio *bp;
463	int error;
464
465	bp = g_new_bio();
466	bp->bio_cmd = BIO_WRITE;
467	bp->bio_done = NULL;
468	bp->bio_offset = offset;
469	bp->bio_length = length;
470	bp->bio_data = ptr;
471	g_io_request(bp, cp);
472	error = biowait(bp, "gwrite");
473	g_destroy_bio(bp);
474	return (error);
475}
476