geom_io.c revision 121323 
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 121323 2003-10-22 06:32:20Z 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;
54
55static u_int pace;
56static uma_zone_t	biozone;
57
58#include <machine/atomic.h>
59
60static void
61g_bioq_lock(struct g_bioq *bq)
62{
63
64	mtx_lock(&bq->bio_queue_lock);
65}
66
67static void
68g_bioq_unlock(struct g_bioq *bq)
69{
70
71	mtx_unlock(&bq->bio_queue_lock);
72}
73
74#if 0
75static void
76g_bioq_destroy(struct g_bioq *bq)
77{
78
79	mtx_destroy(&bq->bio_queue_lock);
80}
81#endif
82
83static void
84g_bioq_init(struct g_bioq *bq)
85{
86
87	TAILQ_INIT(&bq->bio_queue);
88	mtx_init(&bq->bio_queue_lock, "bio queue", NULL, MTX_DEF);
89}
90
91static struct bio *
92g_bioq_first(struct g_bioq *bq)
93{
94	struct bio *bp;
95
96	bp = TAILQ_FIRST(&bq->bio_queue);
97	if (bp != NULL) {
98		TAILQ_REMOVE(&bq->bio_queue, bp, bio_queue);
99		bq->bio_queue_length--;
100	}
101	return (bp);
102}
103
104static void
105g_bioq_enqueue_tail(struct bio *bp, struct g_bioq *rq)
106{
107
108	g_bioq_lock(rq);
109	TAILQ_INSERT_TAIL(&rq->bio_queue, bp, bio_queue);
110	rq->bio_queue_length++;
111	g_bioq_unlock(rq);
112}
113
114struct bio *
115g_new_bio(void)
116{
117	struct bio *bp;
118
119	bp = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
120	return (bp);
121}
122
123void
124g_destroy_bio(struct bio *bp)
125{
126
127	uma_zfree(biozone, bp);
128}
129
130struct bio *
131g_clone_bio(struct bio *bp)
132{
133	struct bio *bp2;
134
135	bp2 = uma_zalloc(biozone, M_NOWAIT | M_ZERO);
136	if (bp2 != NULL) {
137		bp2->bio_parent = bp;
138		bp2->bio_cmd = bp->bio_cmd;
139		bp2->bio_length = bp->bio_length;
140		bp2->bio_offset = bp->bio_offset;
141		bp2->bio_data = bp->bio_data;
142		bp2->bio_attribute = bp->bio_attribute;
143		bp->bio_children++;
144	}
145	return(bp2);
146}
147
148void
149g_io_init()
150{
151
152	g_bioq_init(&g_bio_run_down);
153	g_bioq_init(&g_bio_run_up);
154	biozone = uma_zcreate("g_bio", sizeof (struct bio),
155	    NULL, NULL,
156	    NULL, NULL,
157	    0, 0);
158}
159
160int
161g_io_getattr(const char *attr, struct g_consumer *cp, int *len, void *ptr)
162{
163	struct bio *bp;
164	int error;
165
166	g_trace(G_T_BIO, "bio_getattr(%s)", attr);
167	bp = g_new_bio();
168	bp->bio_cmd = BIO_GETATTR;
169	bp->bio_done = NULL;
170	bp->bio_attribute = attr;
171	bp->bio_length = *len;
172	bp->bio_data = ptr;
173	g_io_request(bp, cp);
174	error = biowait(bp, "ggetattr");
175	*len = bp->bio_completed;
176	g_destroy_bio(bp);
177	return (error);
178}
179
180static int
181g_io_check(struct bio *bp)
182{
183	struct g_consumer *cp;
184	struct g_provider *pp;
185
186	cp = bp->bio_from;
187	pp = bp->bio_to;
188
189	/* Fail if access counters dont allow the operation */
190	switch(bp->bio_cmd) {
191	case BIO_READ:
192	case BIO_GETATTR:
193		if (cp->acr == 0)
194			return (EPERM);
195		break;
196	case BIO_WRITE:
197	case BIO_DELETE:
198		if (cp->acw == 0)
199			return (EPERM);
200		break;
201	default:
202		return (EPERM);
203	}
204	/* if provider is marked for error, don't disturb. */
205	if (pp->error)
206		return (pp->error);
207
208	switch(bp->bio_cmd) {
209	case BIO_READ:
210	case BIO_WRITE:
211	case BIO_DELETE:
212		/* Zero sectorsize is a probably lack of media */
213		if (pp->sectorsize == 0)
214			return (ENXIO);
215		/* Reject I/O not on sector boundary */
216		if (bp->bio_offset % pp->sectorsize)
217			return (EINVAL);
218		/* Reject I/O not integral sector long */
219		if (bp->bio_length % pp->sectorsize)
220			return (EINVAL);
221		/* Reject requests before or past the end of media. */
222		if (bp->bio_offset < 0)
223			return (EIO);
224		if (bp->bio_offset > pp->mediasize)
225			return (EIO);
226		break;
227	default:
228		break;
229	}
230	return (0);
231}
232
233void
234g_io_request(struct bio *bp, struct g_consumer *cp)
235{
236	struct g_provider *pp;
237
238	KASSERT(cp != NULL, ("NULL cp in g_io_request"));
239	KASSERT(bp != NULL, ("NULL bp in g_io_request"));
240	KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
241	pp = cp->provider;
242	KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
243
244	bp->bio_from = cp;
245	bp->bio_to = pp;
246	bp->bio_error = 0;
247	bp->bio_completed = 0;
248
249	if (g_collectstats) {
250		devstat_start_transaction_bio(cp->stat, bp);
251		devstat_start_transaction_bio(pp->stat, bp);
252	}
253	cp->nstart++;
254	pp->nstart++;
255
256	/* Pass it on down. */
257	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
258	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
259	g_bioq_enqueue_tail(bp, &g_bio_run_down);
260	wakeup(&g_wait_down);
261}
262
263void
264g_io_deliver(struct bio *bp, int error)
265{
266	struct g_consumer *cp;
267	struct g_provider *pp;
268
269	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
270	pp = bp->bio_to;
271	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
272	cp = bp->bio_from;
273	if (cp == NULL) {
274		bp->bio_error = error;
275		bp->bio_done(bp);
276		return;
277	}
278	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
279	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
280
281	g_trace(G_T_BIO,
282"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
283	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
284	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
285
286	bp->bio_bcount = bp->bio_length;
287	if (g_collectstats) {
288		bp->bio_resid = bp->bio_bcount - bp->bio_completed;
289		devstat_end_transaction_bio(cp->stat, bp);
290		devstat_end_transaction_bio(pp->stat, bp);
291	}
292	cp->nend++;
293	pp->nend++;
294
295	if (error == ENOMEM) {
296		if (bootverbose)
297			printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
298		g_io_request(bp, cp);
299		pace++;
300		return;
301	}
302	bp->bio_error = error;
303	g_bioq_enqueue_tail(bp, &g_bio_run_up);
304	wakeup(&g_wait_up);
305}
306
307void
308g_io_schedule_down(struct thread *tp __unused)
309{
310	struct bio *bp;
311	off_t excess;
312	int error;
313	struct mtx mymutex;
314
315	bzero(&mymutex, sizeof mymutex);
316	mtx_init(&mymutex, "g_xdown", MTX_DEF, 0);
317
318	for(;;) {
319		g_bioq_lock(&g_bio_run_down);
320		bp = g_bioq_first(&g_bio_run_down);
321		if (bp == NULL) {
322			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
323			    PRIBIO | PDROP, "-", hz/10);
324			continue;
325		}
326		g_bioq_unlock(&g_bio_run_down);
327		if (pace > 0) {
328			msleep(&error, NULL, PRIBIO, "g_down", hz/10);
329			pace--;
330		}
331		error = g_io_check(bp);
332		if (error) {
333			g_io_deliver(bp, error);
334			continue;
335		}
336		switch (bp->bio_cmd) {
337		case BIO_READ:
338		case BIO_WRITE:
339		case BIO_DELETE:
340			/* Truncate requests to the end of providers media. */
341			excess = bp->bio_offset + bp->bio_length;
342			if (excess > bp->bio_to->mediasize) {
343				excess -= bp->bio_to->mediasize;
344				bp->bio_length -= excess;
345			}
346			/* Deliver zero length transfers right here. */
347			if (bp->bio_length == 0) {
348				g_io_deliver(bp, 0);
349				continue;
350			}
351			break;
352		default:
353			break;
354		}
355		mtx_lock(&mymutex);
356		bp->bio_to->geom->start(bp);
357		mtx_unlock(&mymutex);
358	}
359}
360
361void
362g_io_schedule_up(struct thread *tp __unused)
363{
364	struct bio *bp;
365	struct mtx mymutex;
366
367	bzero(&mymutex, sizeof mymutex);
368	mtx_init(&mymutex, "g_xup", MTX_DEF, 0);
369	for(;;) {
370		g_bioq_lock(&g_bio_run_up);
371		bp = g_bioq_first(&g_bio_run_up);
372		if (bp != NULL) {
373			g_bioq_unlock(&g_bio_run_up);
374			mtx_lock(&mymutex);
375			biodone(bp);
376			mtx_unlock(&mymutex);
377			continue;
378		}
379		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
380		    PRIBIO | PDROP, "-", hz/10);
381	}
382}
383
384void *
385g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
386{
387	struct bio *bp;
388	void *ptr;
389	int errorc;
390
391	KASSERT(length >= 512 && length <= DFLTPHYS,
392		("g_read_data(): invalid length %jd", (intmax_t)length));
393
394	bp = g_new_bio();
395	bp->bio_cmd = BIO_READ;
396	bp->bio_done = NULL;
397	bp->bio_offset = offset;
398	bp->bio_length = length;
399	ptr = g_malloc(length, M_WAITOK);
400	bp->bio_data = ptr;
401	g_io_request(bp, cp);
402	errorc = biowait(bp, "gread");
403	if (error != NULL)
404		*error = errorc;
405	g_destroy_bio(bp);
406	if (errorc) {
407		g_free(ptr);
408		ptr = NULL;
409	}
410	return (ptr);
411}
412
413int
414g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
415{
416	struct bio *bp;
417	int error;
418
419	KASSERT(length >= 512 && length <= DFLTPHYS,
420		("g_write_data(): invalid length %jd", (intmax_t)length));
421
422	bp = g_new_bio();
423	bp->bio_cmd = BIO_WRITE;
424	bp->bio_done = NULL;
425	bp->bio_offset = offset;
426	bp->bio_length = length;
427	bp->bio_data = ptr;
428	g_io_request(bp, cp);
429	error = biowait(bp, "gwrite");
430	g_destroy_bio(bp);
431	return (error);
432}
433