geom_io.c revision 114526 
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 114526 2003-05-02 12:36:12Z phk $
36 */
37
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		/* Reject I/O not on sector boundary */
213		if (bp->bio_offset % pp->sectorsize)
214			return (EINVAL);
215		/* Reject I/O not integral sector long */
216		if (bp->bio_length % pp->sectorsize)
217			return (EINVAL);
218		/* Reject requests past the end of media. */
219		if (bp->bio_offset > pp->mediasize)
220			return (EIO);
221		break;
222	default:
223		break;
224	}
225	return (0);
226}
227
228void
229g_io_request(struct bio *bp, struct g_consumer *cp)
230{
231	struct g_provider *pp;
232
233	pp = cp->provider;
234	KASSERT(cp != NULL, ("NULL cp in g_io_request"));
235	KASSERT(bp != NULL, ("NULL bp in g_io_request"));
236	KASSERT(bp->bio_data != NULL, ("NULL bp->data in g_io_request"));
237	KASSERT(pp != NULL, ("consumer not attached in g_io_request"));
238
239	bp->bio_from = cp;
240	bp->bio_to = pp;
241	bp->bio_error = 0;
242	bp->bio_completed = 0;
243
244	if (g_collectstats) {
245		devstat_start_transaction_bio(cp->stat, bp);
246		devstat_start_transaction_bio(pp->stat, bp);
247	}
248	cp->nstart++;
249	pp->nstart++;
250
251	/* Pass it on down. */
252	g_trace(G_T_BIO, "bio_request(%p) from %p(%s) to %p(%s) cmd %d",
253	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd);
254	g_bioq_enqueue_tail(bp, &g_bio_run_down);
255	wakeup(&g_wait_down);
256}
257
258void
259g_io_deliver(struct bio *bp, int error)
260{
261	struct g_consumer *cp;
262	struct g_provider *pp;
263
264	cp = bp->bio_from;
265	pp = bp->bio_to;
266	KASSERT(bp != NULL, ("NULL bp in g_io_deliver"));
267	KASSERT(cp != NULL, ("NULL bio_from in g_io_deliver"));
268	KASSERT(cp->geom != NULL, ("NULL bio_from->geom in g_io_deliver"));
269	KASSERT(pp != NULL, ("NULL bio_to in g_io_deliver"));
270
271	g_trace(G_T_BIO,
272"g_io_deliver(%p) from %p(%s) to %p(%s) cmd %d error %d off %jd len %jd",
273	    bp, cp, cp->geom->name, pp, pp->name, bp->bio_cmd, error,
274	    (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length);
275
276	bp->bio_bcount = bp->bio_length;
277	if (g_collectstats) {
278		bp->bio_resid = bp->bio_bcount - bp->bio_completed;
279		devstat_end_transaction_bio(cp->stat, bp);
280		devstat_end_transaction_bio(pp->stat, bp);
281	}
282	cp->nend++;
283	pp->nend++;
284
285	if (error == ENOMEM) {
286		printf("ENOMEM %p on %p(%s)\n", bp, pp, pp->name);
287		g_io_request(bp, cp);
288		pace++;
289		return;
290	}
291	bp->bio_error = error;
292	g_bioq_enqueue_tail(bp, &g_bio_run_up);
293	wakeup(&g_wait_up);
294}
295
296void
297g_io_schedule_down(struct thread *tp __unused)
298{
299	struct bio *bp;
300	off_t excess;
301	int error;
302	struct mtx mymutex;
303
304	bzero(&mymutex, sizeof mymutex);
305	mtx_init(&mymutex, "g_xdown", MTX_DEF, 0);
306
307	for(;;) {
308		g_bioq_lock(&g_bio_run_down);
309		bp = g_bioq_first(&g_bio_run_down);
310		if (bp == NULL) {
311			msleep(&g_wait_down, &g_bio_run_down.bio_queue_lock,
312			    PRIBIO | PDROP, "g_down", hz/10);
313			continue;
314		}
315		g_bioq_unlock(&g_bio_run_down);
316		if (pace > 0) {
317			msleep(&error, NULL, PRIBIO, "g_down", hz/10);
318			pace--;
319		}
320		error = g_io_check(bp);
321		if (error) {
322			g_io_deliver(bp, error);
323			continue;
324		}
325		switch (bp->bio_cmd) {
326		case BIO_READ:
327		case BIO_WRITE:
328		case BIO_DELETE:
329			/* Truncate requests to the end of providers media. */
330			excess = bp->bio_offset + bp->bio_length;
331			if (excess > bp->bio_to->mediasize) {
332				excess -= bp->bio_to->mediasize;
333				bp->bio_length -= excess;
334			}
335			/* Deliver zero length transfers right here. */
336			if (bp->bio_length == 0) {
337				g_io_deliver(bp, 0);
338				continue;
339			}
340			break;
341		default:
342			break;
343		}
344		mtx_lock(&mymutex);
345		bp->bio_to->geom->start(bp);
346		mtx_unlock(&mymutex);
347	}
348}
349
350void
351g_io_schedule_up(struct thread *tp __unused)
352{
353	struct bio *bp;
354	struct mtx mymutex;
355
356	bzero(&mymutex, sizeof mymutex);
357	mtx_init(&mymutex, "g_xup", MTX_DEF, 0);
358	for(;;) {
359		g_bioq_lock(&g_bio_run_up);
360		bp = g_bioq_first(&g_bio_run_up);
361		if (bp != NULL) {
362			g_bioq_unlock(&g_bio_run_up);
363			mtx_lock(&mymutex);
364			biodone(bp);
365			mtx_unlock(&mymutex);
366			continue;
367		}
368		msleep(&g_wait_up, &g_bio_run_up.bio_queue_lock,
369		    PRIBIO | PDROP, "g_up", hz/10);
370	}
371}
372
373void *
374g_read_data(struct g_consumer *cp, off_t offset, off_t length, int *error)
375{
376	struct bio *bp;
377	void *ptr;
378	int errorc;
379
380	bp = g_new_bio();
381	bp->bio_cmd = BIO_READ;
382	bp->bio_done = NULL;
383	bp->bio_offset = offset;
384	bp->bio_length = length;
385	ptr = g_malloc(length, M_WAITOK);
386	bp->bio_data = ptr;
387	g_io_request(bp, cp);
388	errorc = biowait(bp, "gread");
389	if (error != NULL)
390		*error = errorc;
391	g_destroy_bio(bp);
392	if (errorc) {
393		g_free(ptr);
394		ptr = NULL;
395	}
396	return (ptr);
397}
398
399int
400g_write_data(struct g_consumer *cp, off_t offset, void *ptr, off_t length)
401{
402	struct bio *bp;
403	int error;
404
405	bp = g_new_bio();
406	bp->bio_cmd = BIO_WRITE;
407	bp->bio_done = NULL;
408	bp->bio_offset = offset;
409	bp->bio_length = length;
410	bp->bio_data = ptr;
411	g_io_request(bp, cp);
412	error = biowait(bp, "gwrite");
413	g_destroy_bio(bp);
414	return (error);
415}
416