geom_io.c revision 110541 
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 110541 2003-02-08 13:03:57Z phk $
36 */
37
38
39#include <sys/param.h>
40#include <sys/stdint.h>
41#ifndef _KERNEL
42#include <stdio.h>
43#include <string.h>
44#include <stdlib.h>
45#include <signal.h>
46#include <err.h>
47#include <sched.h>
48#else
49#include <sys/systm.h>
50#include <sys/kernel.h>
51#include <sys/malloc.h>
52#include <sys/bio.h>
53#endif
54
55#include <sys/errno.h>
56#include <geom/geom.h>
57#include <geom/geom_int.h>
58#include <geom/geom_stats.h>
59
60static struct g_bioq g_bio_run_down;
61static struct g_bioq g_bio_run_up;
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	g_bioq_lock(bq);
105	bp = TAILQ_FIRST(&bq->bio_queue);
106	if (bp != NULL) {
107		TAILQ_REMOVE(&bq->bio_queue, bp, bio_queue);
108		bq->bio_queue_length--;
109	}
110	g_bioq_unlock(bq);
111	return (bp);
112}
113
114static void
115g_bioq_enqueue_tail(struct bio *bp, struct g_bioq *rq)
116{
117
118	g_bioq_lock(rq);
119	TAILQ_INSERT_TAIL(&rq->bio_queue, bp, bio_queue);
120	rq->bio_queue_length++;
121	g_bioq_unlock(rq);
122}
123
124struct bio *
125g_new_bio(void)
126{
127	struct bio *bp;
128
129	bp = g_bioq_first(&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_idle);
171}
172
173int
174g_io_setattr(const char *attr, struct g_consumer *cp, int len, void *ptr)
175{
176	struct bio *bp;
177	int error;
178
179	g_trace(G_T_BIO, "bio_setattr(%s)", attr);
180	bp = g_new_bio();
181	bp->bio_cmd = BIO_SETATTR;
182	bp->bio_done = NULL;
183	bp->bio_attribute = attr;
184	bp->bio_length = len;
185	bp->bio_data = ptr;
186	g_io_request(bp, cp);
187	error = biowait(bp, "gsetattr");
188	g_destroy_bio(bp);
189	return (error);
190}
191
192
193int
194g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
195{
196	struct bio *bp;
197	int error;
198
199	g_trace(G_T_BIO, "bio_getattr(%s)", attr);
200	bp = g_new_bio();
201	bp->bio_cmd = BIO_GETATTR;
202	bp->bio_done = NULL;
203	bp->bio_attribute = attr;
204	bp->bio_length = *len;
205	bp->bio_data = ptr;
206	g_io_request(bp, cp);
207	error = biowait(bp, "ggetattr");
208	*len = bp->bio_completed;
209	g_destroy_bio(bp);
210	return (error);
211}
212
213static int
214g_io_check(struct bio *bp)
215{
216	struct g_consumer *cp;
217	struct g_provider *pp;
218
219	cp = bp->bio_from;
220	pp = bp->bio_to;
221
222	/* Fail if access counters dont allow the operation */
223	switch(bp->bio_cmd) {
224	case BIO_READ:
225	case BIO_GETATTR:
226		if (cp->acr == 0)
227			return (EPERM);
228		break;
229	case BIO_WRITE:
230	case BIO_DELETE:
231	case BIO_SETATTR:
232		if (cp->acw == 0)
233			return (EPERM);
234		break;
235	default:
236		return (EPERM);
237	}
238	/* if provider is marked for error, don't disturb. */
239	if (pp->error)
240		return (pp->error);
241
242	switch(bp->bio_cmd) {
243	case BIO_READ:
244	case BIO_WRITE:
245	case BIO_DELETE:
246		/* Reject I/O not on sector boundary */
247		if (bp->bio_offset % pp->sectorsize)
248			return (EINVAL);
249		/* Reject I/O not integral sector long */
250		if (bp->bio_length % pp->sectorsize)
251			return (EINVAL);
252		/* Reject requests past the end of media. */
253		if (bp->bio_offset > pp->mediasize)
254			return (EIO);
255		break;
256	default:
257		break;
258	}
259	return (0);
260}
261
262void
263g_io_request(struct bio *bp, struct g_consumer *cp)
264{
265	struct g_provider *pp;
266	struct bintime bt;
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		/* Collect statistics */
281		binuptime(&bp->bio_t0);
282		if (cp->stat->nop == cp->stat->nend) {
283			/* Consumer is idle */
284			bt = bp->bio_t0;
285			bintime_sub(&bt, &cp->stat->wentidle);
286			bintime_add(&cp->stat->it, &bt);
287			if (pp->stat->nop == pp->stat->nend) {
288				/*
289				 * NB: Provider can only be idle if the
290				 * consumer is but we cannot trust them
291				 * to have gone idle at the same time.
292				 */
293				bt = bp->bio_t0;
294				bintime_sub(&bt, &pp->stat->wentidle);
295				bintime_add(&pp->stat->it, &bt);
296			}
297		}
298	}
299	cp->stat->nop++;
300	pp->stat->nop++;
301
302	/* Pass it on down. */
303	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
304	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
305	g_bioq_enqueue_tail(bp, &g_bio_run_down);
306	wakeup(&g_wait_down);
307}
308
309void
310g_io_deliver(struct bio *bp, int error)
311{
312	struct g_consumer *cp;
313	struct g_provider *pp;
314	struct bintime t1;
315	int idx;
316
317	cp = bp->bio_from;
318	pp = bp->bio_to;
319	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
320	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
321	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
322	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
323
324	g_trace(G_T_BIO,
325"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
326	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
327	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
328
329	switch (bp->bio_cmd) {
330	case BIO_READ:    idx =  G_STAT_IDX_READ;    break;
331	case BIO_WRITE:   idx =  G_STAT_IDX_WRITE;   break;
332	case BIO_DELETE:  idx =  G_STAT_IDX_DELETE;  break;
333	case BIO_GETATTR: idx =  -1; break;
334	case BIO_SETATTR: idx =  -1; break;
335	default:
336		panic("unknown bio_cmd in g_io_deliver");
337		break;
338	}
339
340	/* Collect statistics */
341	if (g_collectstats) {
342		binuptime(&t1);
343		pp->stat->wentidle = t1;
344		cp->stat->wentidle = t1;
345
346		if (idx >= 0) {
347			bintime_sub(&t1, &bp->bio_t0);
348			bintime_add(&cp->stat->ops[idx].dt, &t1);
349			bintime_add(&pp->stat->ops[idx].dt, &t1);
350			pp->stat->ops[idx].nbyte += bp->bio_completed;
351			cp->stat->ops[idx].nbyte += bp->bio_completed;
352			pp->stat->ops[idx].nop++;
353			cp->stat->ops[idx].nop++;
354			if (error == ENOMEM) {
355				cp->stat->ops[idx].nmem++;
356				pp->stat->ops[idx].nmem++;
357			} else if (error != 0) {
358				cp->stat->ops[idx].nerr++;
359				pp->stat->ops[idx].nerr++;
360			}
361		}
362	}
363
364	pp->stat->nend++; /* In reverse order of g_io_request() */
365	cp->stat->nend++;
366
367	if (error == ENOMEM) {
368		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
369		g_io_request(bp, cp);
370		pace++;
371		return;
372	}
373	bp->bio_error = error;
374	g_bioq_enqueue_tail(bp, &g_bio_run_up);
375	wakeup(&g_wait_up);
376}
377
378void
379g_io_schedule_down(struct thread *tp __unused)
380{
381	struct bio *bp;
382	off_t excess;
383	int error;
384
385	for(;;) {
386		bp = g_bioq_first(&g_bio_run_down);
387		if (bp == NULL)
388			break;
389		error = g_io_check(bp);
390		if (error) {
391			g_io_deliver(bp, error);
392			continue;
393		}
394		/* Truncate requests to the end of providers media. */
395		excess = bp->bio_offset + bp->bio_length;
396		if (excess > bp->bio_to->mediasize) {
397			excess -= bp->bio_to->mediasize;
398			bp->bio_length -= excess;
399		}
400		/* Deliver zero length transfers right here. */
401		if (bp->bio_length == 0) {
402			g_io_deliver(bp, 0);
403			continue;
404		}
405		bp->bio_to->geom->start(bp);
406		if (pace) {
407			pace--;
408			break;
409		}
410	}
411}
412
413void
414g_io_schedule_up(struct thread *tp __unused)
415{
416	struct bio *bp;
417	struct g_consumer *cp;
418
419	for(;;) {
420		bp = g_bioq_first(&g_bio_run_up);
421		if (bp == NULL)
422			break;
423
424		cp = bp->bio_from;
425		biodone(bp);
426	}
427}
428
429void *
430g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
431{
432	struct bio *bp;
433	void *ptr;
434	int errorc;
435
436	bp = g_new_bio();
437	bp->bio_cmd = BIO_READ;
438	bp->bio_done = NULL;
439	bp->bio_offset = offset;
440	bp->bio_length = length;
441	ptr = g_malloc(length, 0);
442	bp->bio_data = ptr;
443	g_io_request(bp, cp);
444	errorc = biowait(bp, "gread");
445	if (error != NULL)
446		*error = errorc;
447	g_destroy_bio(bp);
448	if (errorc) {
449		g_free(ptr);
450		ptr = NULL;
451	}
452	return (ptr);
453}
454
455int
456g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
457{
458	struct bio *bp;
459	int error;
460
461	bp = g_new_bio();
462	bp->bio_cmd = BIO_WRITE;
463	bp->bio_done = NULL;
464	bp->bio_offset = offset;
465	bp->bio_length = length;
466	bp->bio_data = ptr;
467	g_io_request(bp, cp);
468	error = biowait(bp, "gwrite");
469	g_destroy_bio(bp);
470	return (error);
471}
472