geom_io.c revision 110523 
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 110523 2003-02-07 23:08:24Z 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
59static struct g_bioq g_bio_run_down;
60static struct g_bioq g_bio_run_up;
61static struct g_bioq g_bio_idle;
62
63static u_int pace;
64
65#include <machine/atomic.h>
66
67static void
68g_bioq_lock(struct g_bioq *bq)
69{
70
71	mtx_lock(&bq->bio_queue_lock);
72}
73
74static void
75g_bioq_unlock(struct g_bioq *bq)
76{
77
78	mtx_unlock(&bq->bio_queue_lock);
79}
80
81#if 0
82static void
83g_bioq_destroy(struct g_bioq *bq)
84{
85
86	mtx_destroy(&bq->bio_queue_lock);
87}
88#endif
89
90static void
91g_bioq_init(struct g_bioq *bq)
92{
93
94	TAILQ_INIT(&bq->bio_queue);
95	mtx_init(&bq->bio_queue_lock, "bio queue", NULL, MTX_DEF);
96}
97
98static struct bio *
99g_bioq_first(struct g_bioq *bq)
100{
101	struct bio *bp;
102
103	g_bioq_lock(bq);
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	g_bioq_unlock(bq);
110	return (bp);
111}
112
113static void
114g_bioq_enqueue_tail(struct bio *bp, struct g_bioq *rq)
115{
116
117	g_bioq_lock(rq);
118	TAILQ_INSERT_TAIL(&rq->bio_queue, bp, bio_queue);
119	rq->bio_queue_length++;
120	g_bioq_unlock(rq);
121}
122
123struct bio *
124g_new_bio(void)
125{
126	struct bio *bp;
127
128	bp = g_bioq_first(&g_bio_idle);
129	if (bp == NULL)
130		bp = g_malloc(sizeof *bp, M_NOWAIT | M_ZERO);
131	/* g_trace(G_T_BIO, "g_new_bio() = %p", bp); */
132	return (bp);
133}
134
135void
136g_destroy_bio(struct bio *bp)
137{
138
139	/* g_trace(G_T_BIO, "g_destroy_bio(%p)", bp); */
140	bzero(bp, sizeof *bp);
141	g_bioq_enqueue_tail(bp, &g_bio_idle);
142}
143
144struct bio *
145g_clone_bio(struct bio *bp)
146{
147	struct bio *bp2;
148
149	bp2 = g_new_bio();
150	if (bp2 != NULL) {
151		bp2->bio_parent = bp;
152		bp2->bio_cmd = bp->bio_cmd;
153		bp2->bio_length = bp->bio_length;
154		bp2->bio_offset = bp->bio_offset;
155		bp2->bio_data = bp->bio_data;
156		bp2->bio_attribute = bp->bio_attribute;
157		bp->bio_children++;
158	}
159	/* g_trace(G_T_BIO, "g_clone_bio(%p) = %p", bp, bp2); */
160	return(bp2);
161}
162
163void
164g_io_init()
165{
166
167	g_bioq_init(&g_bio_run_down);
168	g_bioq_init(&g_bio_run_up);
169	g_bioq_init(&g_bio_idle);
170}
171
172int
173g_io_setattr(const char *attr, struct g_consumer *cp, int len, void *ptr)
174{
175	struct bio *bp;
176	int error;
177
178	g_trace(G_T_BIO, "bio_setattr(%s)", attr);
179	bp = g_new_bio();
180	bp->bio_cmd = BIO_SETATTR;
181	bp->bio_done = NULL;
182	bp->bio_attribute = attr;
183	bp->bio_length = len;
184	bp->bio_data = ptr;
185	g_io_request(bp, cp);
186	error = biowait(bp, "gsetattr");
187	g_destroy_bio(bp);
188	return (error);
189}
190
191
192int
193g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
194{
195	struct bio *bp;
196	int error;
197
198	g_trace(G_T_BIO, "bio_getattr(%s)", attr);
199	bp = g_new_bio();
200	bp->bio_cmd = BIO_GETATTR;
201	bp->bio_done = NULL;
202	bp->bio_attribute = attr;
203	bp->bio_length = *len;
204	bp->bio_data = ptr;
205	g_io_request(bp, cp);
206	error = biowait(bp, "ggetattr");
207	*len = bp->bio_completed;
208	g_destroy_bio(bp);
209	return (error);
210}
211
212static int
213g_io_check(struct bio *bp)
214{
215	struct g_consumer *cp;
216	struct g_provider *pp;
217
218	cp = bp->bio_from;
219	pp = bp->bio_to;
220
221	/* Fail if access counters dont allow the operation */
222	switch(bp->bio_cmd) {
223	case BIO_READ:
224	case BIO_GETATTR:
225		if (cp->acr == 0)
226			return (EPERM);
227		break;
228	case BIO_WRITE:
229	case BIO_DELETE:
230	case BIO_SETATTR:
231		if (cp->acw == 0)
232			return (EPERM);
233		break;
234	default:
235		return (EPERM);
236	}
237	/* if provider is marked for error, don't disturb. */
238	if (pp->error)
239		return (pp->error);
240
241	switch(bp->bio_cmd) {
242	case BIO_READ:
243	case BIO_WRITE:
244	case BIO_DELETE:
245		/* Reject I/O not on sector boundary */
246		if (bp->bio_offset % pp->sectorsize)
247			return (EINVAL);
248		/* Reject I/O not integral sector long */
249		if (bp->bio_length % pp->sectorsize)
250			return (EINVAL);
251		/* Reject requests past the end of media. */
252		if (bp->bio_offset > pp->mediasize)
253			return (EIO);
254		break;
255	default:
256		break;
257	}
258	return (0);
259}
260
261void
262g_io_request(struct bio *bp, struct g_consumer *cp)
263{
264	struct g_provider *pp;
265	struct bintime bt;
266
267	pp = cp->provider;
268	KASSERT(cp != NULL, ("NULL cp in g_io_request"));
269	KASSERT(bp != NULL, ("NULL bp in g_io_request"));
270	KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
271	KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
272
273	bp->bio_from = cp;
274	bp->bio_to = pp;
275	bp->bio_error = 0;
276	bp->bio_completed = 0;
277
278	if (g_collectstats) {
279		/* Collect statistics */
280		binuptime(&bp->bio_t0);
281		if (cp->stat.nop == cp->stat.nend) {
282			/* Consumer is idle */
283			bt = bp->bio_t0;
284			bintime_sub(&bt, &cp->stat.wentidle);
285			bintime_add(&cp->stat.it, &bt);
286			if (pp->stat.nop == pp->stat.nend) {
287				/*
288				 * NB: Provider can only be idle if the
289				 * consumer is but we cannot trust them
290				 * to have gone idle at the same time.
291				 */
292				bt = bp->bio_t0;
293				bintime_sub(&bt, &pp->stat.wentidle);
294				bintime_add(&pp->stat.it, &bt);
295			}
296		}
297	}
298	cp->stat.nop++;
299	pp->stat.nop++;
300
301	/* Pass it on down. */
302	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
303	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
304	g_bioq_enqueue_tail(bp, &g_bio_run_down);
305	wakeup(&g_wait_down);
306}
307
308void
309g_io_deliver(struct bio *bp, int error)
310{
311	struct g_consumer *cp;
312	struct g_provider *pp;
313	struct bintime t1;
314	int idx;
315
316	cp = bp->bio_from;
317	pp = bp->bio_to;
318	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
319	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
320	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
321	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
322
323	g_trace(G_T_BIO,
324"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
325	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
326	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
327
328	switch (bp->bio_cmd) {
329	case BIO_READ:    idx =  G_STAT_IDX_READ;    break;
330	case BIO_WRITE:   idx =  G_STAT_IDX_WRITE;   break;
331	case BIO_DELETE:  idx =  G_STAT_IDX_DELETE;  break;
332	case BIO_GETATTR: idx =  -1; break;
333	case BIO_SETATTR: idx =  -1; break;
334	default:
335		panic("unknown bio_cmd in g_io_deliver");
336		break;
337	}
338
339	/* Collect statistics */
340	if (g_collectstats) {
341		binuptime(&t1);
342		pp->stat.wentidle = t1;
343		cp->stat.wentidle = t1;
344
345		if (idx >= 0) {
346			bintime_sub(&t1, &bp->bio_t0);
347			bintime_add(&cp->stat.ops[idx].dt, &t1);
348			bintime_add(&pp->stat.ops[idx].dt, &t1);
349			pp->stat.ops[idx].nbyte += bp->bio_completed;
350			cp->stat.ops[idx].nbyte += bp->bio_completed;
351			pp->stat.ops[idx].nop++;
352			cp->stat.ops[idx].nop++;
353			if (error == ENOMEM) {
354				cp->stat.ops[idx].nmem++;
355				pp->stat.ops[idx].nmem++;
356			} else if (error != 0) {
357				cp->stat.ops[idx].nerr++;
358				pp->stat.ops[idx].nerr++;
359			}
360		}
361	}
362
363	pp->stat.nend++; /* In reverse order of g_io_request() */
364	cp->stat.nend++;
365
366	if (error == ENOMEM) {
367		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
368		g_io_request(bp, cp);
369		pace++;
370		return;
371	}
372	bp->bio_error = error;
373	g_bioq_enqueue_tail(bp, &g_bio_run_up);
374	wakeup(&g_wait_up);
375}
376
377void
378g_io_schedule_down(struct thread *tp __unused)
379{
380	struct bio *bp;
381	off_t excess;
382	int error;
383
384	for(;;) {
385		bp = g_bioq_first(&g_bio_run_down);
386		if (bp == NULL)
387			break;
388		error = g_io_check(bp);
389		if (error) {
390			g_io_deliver(bp, error);
391			continue;
392		}
393		/* Truncate requests to the end of providers media. */
394		excess = bp->bio_offset + bp->bio_length;
395		if (excess > bp->bio_to->mediasize) {
396			excess -= bp->bio_to->mediasize;
397			bp->bio_length -= excess;
398		}
399		/* Deliver zero length transfers right here. */
400		if (bp->bio_length == 0) {
401			g_io_deliver(bp, 0);
402			continue;
403		}
404		bp->bio_to->geom->start(bp);
405		if (pace) {
406			pace--;
407			break;
408		}
409	}
410}
411
412void
413g_io_schedule_up(struct thread *tp __unused)
414{
415	struct bio *bp;
416	struct g_consumer *cp;
417
418	for(;;) {
419		bp = g_bioq_first(&g_bio_run_up);
420		if (bp == NULL)
421			break;
422
423		cp = bp->bio_from;
424		biodone(bp);
425	}
426}
427
428void *
429g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
430{
431	struct bio *bp;
432	void *ptr;
433	int errorc;
434
435	bp = g_new_bio();
436	bp->bio_cmd = BIO_READ;
437	bp->bio_done = NULL;
438	bp->bio_offset = offset;
439	bp->bio_length = length;
440	ptr = g_malloc(length, 0);
441	bp->bio_data = ptr;
442	g_io_request(bp, cp);
443	errorc = biowait(bp, "gread");
444	if (error != NULL)
445		*error = errorc;
446	g_destroy_bio(bp);
447	if (errorc) {
448		g_free(ptr);
449		ptr = NULL;
450	}
451	return (ptr);
452}
453
454int
455g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
456{
457	struct bio *bp;
458	int error;
459
460	bp = g_new_bio();
461	bp->bio_cmd = BIO_WRITE;
462	bp->bio_done = NULL;
463	bp->bio_offset = offset;
464	bp->bio_length = length;
465	bp->bio_data = ptr;
466	g_io_request(bp, cp);
467	error = biowait(bp, "gwrite");
468	g_destroy_bio(bp);
469	return (error);
470}
471