geom_disk.c revision 169287 
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_disk.c 169287 2007-05-05 17:47:20Z pjd $");
38
39#include "opt_geom.h"
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/kernel.h>
44#include <sys/sysctl.h>
45#include <sys/bio.h>
46#include <sys/conf.h>
47#include <sys/fcntl.h>
48#include <sys/malloc.h>
49#include <sys/sysctl.h>
50#include <sys/devicestat.h>
51#include <machine/md_var.h>
52
53#include <sys/lock.h>
54#include <sys/mutex.h>
55#include <geom/geom.h>
56#include <geom/geom_disk.h>
57#include <geom/geom_int.h>
58
59static struct mtx g_disk_done_mtx;
60
61static g_access_t g_disk_access;
62static g_init_t g_disk_init;
63static g_fini_t g_disk_fini;
64static g_start_t g_disk_start;
65static g_ioctl_t g_disk_ioctl;
66static g_dumpconf_t g_disk_dumpconf;
67
68static struct g_class g_disk_class = {
69	.name = "DISK",
70	.version = G_VERSION,
71	.init = g_disk_init,
72	.fini = g_disk_fini,
73	.start = g_disk_start,
74	.access = g_disk_access,
75	.ioctl = g_disk_ioctl,
76	.dumpconf = g_disk_dumpconf,
77};
78
79static void
80g_disk_init(struct g_class *mp __unused)
81{
82
83	mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF);
84}
85
86static void
87g_disk_fini(struct g_class *mp __unused)
88{
89
90	mtx_destroy(&g_disk_done_mtx);
91}
92
93DECLARE_GEOM_CLASS(g_disk_class, g_disk);
94
95static void __inline
96g_disk_lock_giant(struct disk *dp)
97{
98	if (dp->d_flags & DISKFLAG_NEEDSGIANT)
99		mtx_lock(&Giant);
100}
101
102static void __inline
103g_disk_unlock_giant(struct disk *dp)
104{
105	if (dp->d_flags & DISKFLAG_NEEDSGIANT)
106		mtx_unlock(&Giant);
107}
108
109static int
110g_disk_access(struct g_provider *pp, int r, int w, int e)
111{
112	struct disk *dp;
113	int error;
114
115	g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
116	    pp->name, r, w, e);
117	g_topology_assert();
118	dp = pp->geom->softc;
119	if (dp == NULL || dp->d_destroyed) {
120		/*
121		 * Allow decreasing access count even if disk is not
122		 * avaliable anymore.
123		 */
124		if (r <= 0 && w <= 0 && e <= 0)
125			return (0);
126		return (ENXIO);
127	}
128	r += pp->acr;
129	w += pp->acw;
130	e += pp->ace;
131	error = 0;
132	if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
133		if (dp->d_open != NULL) {
134			g_disk_lock_giant(dp);
135			error = dp->d_open(dp);
136			if (bootverbose && error != 0)
137				printf("Opened disk %s -> %d\n",
138				    pp->name, error);
139			g_disk_unlock_giant(dp);
140		}
141		pp->mediasize = dp->d_mediasize;
142		pp->sectorsize = dp->d_sectorsize;
143		dp->d_flags |= DISKFLAG_OPEN;
144		if (dp->d_maxsize == 0) {
145			printf("WARNING: Disk drive %s%d has no d_maxsize\n",
146			    dp->d_name, dp->d_unit);
147			dp->d_maxsize = DFLTPHYS;
148		}
149	} else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
150		if (dp->d_close != NULL) {
151			g_disk_lock_giant(dp);
152			error = dp->d_close(dp);
153			if (error != 0)
154				printf("Closed disk %s -> %d\n",
155				    pp->name, error);
156			g_disk_unlock_giant(dp);
157		}
158		dp->d_flags &= ~DISKFLAG_OPEN;
159	}
160	return (error);
161}
162
163static void
164g_disk_kerneldump(struct bio *bp, struct disk *dp)
165{
166	int error;
167	struct g_kerneldump *gkd;
168	struct dumperinfo di;
169	struct g_geom *gp;
170
171	gkd = (struct g_kerneldump*)bp->bio_data;
172	gp = bp->bio_to->geom;
173	g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
174		gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
175	if (dp->d_dump == NULL) {
176		g_io_deliver(bp, ENODEV);
177		return;
178	}
179	di.dumper = dp->d_dump;
180	di.priv = dp;
181	di.blocksize = dp->d_sectorsize;
182	di.mediaoffset = gkd->offset;
183	if ((gkd->offset + gkd->length) > dp->d_mediasize)
184		gkd->length = dp->d_mediasize - gkd->offset;
185	di.mediasize = gkd->length;
186	error = set_dumper(&di);
187	g_io_deliver(bp, error);
188}
189
190static void
191g_disk_done(struct bio *bp)
192{
193	struct bio *bp2;
194	struct disk *dp;
195
196	/* See "notes" for why we need a mutex here */
197	/* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
198	mtx_lock(&g_disk_done_mtx);
199	bp->bio_completed = bp->bio_length - bp->bio_resid;
200
201	bp2 = bp->bio_parent;
202	if (bp2->bio_error == 0)
203		bp2->bio_error = bp->bio_error;
204	bp2->bio_completed += bp->bio_completed;
205	if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) &&
206	    (dp = bp2->bio_to->geom->softc)) {
207		devstat_end_transaction_bio(dp->d_devstat, bp);
208	}
209	g_destroy_bio(bp);
210	bp2->bio_inbed++;
211	if (bp2->bio_children == bp2->bio_inbed) {
212		bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
213		g_io_deliver(bp2, bp2->bio_error);
214	}
215	mtx_unlock(&g_disk_done_mtx);
216}
217
218static int
219g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
220{
221	struct g_geom *gp;
222	struct disk *dp;
223	int error;
224
225	gp = pp->geom;
226	dp = gp->softc;
227
228	if (dp->d_ioctl == NULL)
229		return (ENOIOCTL);
230	g_disk_lock_giant(dp);
231	error = dp->d_ioctl(dp, cmd, data, fflag, td);
232	g_disk_unlock_giant(dp);
233	return(error);
234}
235
236static void
237g_disk_start(struct bio *bp)
238{
239	struct bio *bp2, *bp3;
240	struct disk *dp;
241	int error;
242	off_t off;
243
244	dp = bp->bio_to->geom->softc;
245	if (dp == NULL || dp->d_destroyed) {
246		g_io_deliver(bp, ENXIO);
247		return;
248	}
249	error = EJUSTRETURN;
250	switch(bp->bio_cmd) {
251	case BIO_DELETE:
252		if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
253			error = 0;
254			break;
255		}
256		/* fall-through */
257	case BIO_READ:
258	case BIO_WRITE:
259		off = 0;
260		bp3 = NULL;
261		bp2 = g_clone_bio(bp);
262		if (bp2 == NULL) {
263			error = ENOMEM;
264			break;
265		}
266		do {
267			bp2->bio_offset += off;
268			bp2->bio_length -= off;
269			bp2->bio_data += off;
270			if (bp2->bio_length > dp->d_maxsize) {
271				/*
272				 * XXX: If we have a stripesize we should really
273				 * use it here.
274				 */
275				bp2->bio_length = dp->d_maxsize;
276				off += dp->d_maxsize;
277				/*
278				 * To avoid a race, we need to grab the next bio
279				 * before we schedule this one.  See "notes".
280				 */
281				bp3 = g_clone_bio(bp);
282				if (bp3 == NULL)
283					bp->bio_error = ENOMEM;
284			}
285			bp2->bio_done = g_disk_done;
286			bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
287			bp2->bio_bcount = bp2->bio_length;
288			bp2->bio_disk = dp;
289			devstat_start_transaction_bio(dp->d_devstat, bp2);
290			g_disk_lock_giant(dp);
291			dp->d_strategy(bp2);
292			g_disk_unlock_giant(dp);
293			bp2 = bp3;
294			bp3 = NULL;
295		} while (bp2 != NULL);
296		break;
297	case BIO_GETATTR:
298		if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors))
299			break;
300		else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
301			break;
302		else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
303			break;
304		else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
305			break;
306		else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
307			g_disk_kerneldump(bp, dp);
308		else
309			error = ENOIOCTL;
310		break;
311	case BIO_FLUSH:
312		g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)",
313		    bp->bio_to->name);
314		if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
315			g_io_deliver(bp, ENODEV);
316			return;
317		}
318		bp2 = g_clone_bio(bp);
319		if (bp2 == NULL) {
320			g_io_deliver(bp, ENOMEM);
321			return;
322		}
323		bp2->bio_done = g_disk_done;
324		bp2->bio_disk = dp;
325		g_disk_lock_giant(dp);
326		dp->d_strategy(bp2);
327		g_disk_unlock_giant(dp);
328		break;
329	default:
330		error = EOPNOTSUPP;
331		break;
332	}
333	if (error != EJUSTRETURN)
334		g_io_deliver(bp, error);
335	return;
336}
337
338static void
339g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
340{
341	struct disk *dp;
342
343	dp = gp->softc;
344	if (dp == NULL)
345		return;
346	if (indent == NULL) {
347		sbuf_printf(sb, " hd %u", dp->d_fwheads);
348		sbuf_printf(sb, " sc %u", dp->d_fwsectors);
349		return;
350	}
351	if (pp != NULL) {
352		sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
353		    indent, dp->d_fwheads);
354		sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
355		    indent, dp->d_fwsectors);
356	}
357}
358
359static void
360g_disk_create(void *arg, int flag)
361{
362	struct g_geom *gp;
363	struct g_provider *pp;
364	struct disk *dp;
365
366	if (flag == EV_CANCEL)
367		return;
368	g_topology_assert();
369	dp = arg;
370	gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
371	gp->softc = dp;
372	pp = g_new_providerf(gp, "%s", gp->name);
373	pp->mediasize = dp->d_mediasize;
374	pp->sectorsize = dp->d_sectorsize;
375	if (dp->d_flags & DISKFLAG_CANDELETE)
376		pp->flags |= G_PF_CANDELETE;
377	pp->stripeoffset = dp->d_stripeoffset;
378	pp->stripesize = dp->d_stripesize;
379	if (bootverbose)
380		printf("GEOM: new disk %s\n", gp->name);
381	dp->d_geom = gp;
382	g_error_provider(pp, 0);
383}
384
385static void
386g_disk_destroy(void *ptr, int flag)
387{
388	struct disk *dp;
389	struct g_geom *gp;
390
391	g_topology_assert();
392	dp = ptr;
393	gp = dp->d_geom;
394	if (gp != NULL) {
395		gp->softc = NULL;
396		g_wither_geom(gp, ENXIO);
397	}
398	g_free(dp);
399}
400
401/*
402 * We only allow [a-zA-Z0-9-_@#%.] characters, the rest is converted to 'x<HH>'.
403 */
404static void
405g_disk_ident_adjust(char *ident, size_t size)
406{
407	char newid[DISK_IDENT_SIZE], tmp[4];
408	size_t len;
409	char *p;
410
411	bzero(newid, sizeof(newid));
412	len = 0;
413	for (p = ident; *p != '\0' && len < sizeof(newid) - 1; p++) {
414		switch (*p) {
415		default:
416			if ((*p < 'a' || *p > 'z') &&
417			    (*p < 'A' || *p > 'Z') &&
418			    (*p < '0' || *p > '9')) {
419				snprintf(tmp, sizeof(tmp), "x%02hhx", *p);
420				strlcat(newid, tmp, sizeof(newid));
421				len += 3;
422				break;
423			}
424			/* FALLTHROUGH */
425		case '-':
426		case '_':
427		case '@':
428		case '#':
429		case '%':
430		case '.':
431			newid[len++] = *p;
432			break;
433		}
434	}
435	bzero(ident, size);
436	strlcpy(ident, newid, size);
437}
438
439struct disk *
440disk_alloc()
441{
442	struct disk *dp;
443
444	dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO);
445	return (dp);
446}
447
448void
449disk_create(struct disk *dp, int version)
450{
451	if (version != DISK_VERSION_00 && version != DISK_VERSION_01) {
452		printf("WARNING: Attempt to add disk %s%d %s",
453		    dp->d_name, dp->d_unit,
454		    " using incompatible ABI version of disk(9)\n");
455		printf("WARNING: Ignoring disk %s%d\n",
456		    dp->d_name, dp->d_unit);
457		return;
458	}
459	KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
460	KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
461	KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
462	KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
463	if (dp->d_devstat == NULL)
464		dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
465		    dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
466		    DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
467	dp->d_geom = NULL;
468	g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
469	g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
470}
471
472void
473disk_destroy(struct disk *dp)
474{
475
476	g_cancel_event(dp);
477	dp->d_destroyed = 1;
478	if (dp->d_devstat != NULL)
479		devstat_remove_entry(dp->d_devstat);
480	g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
481}
482
483void
484disk_gone(struct disk *dp)
485{
486	struct g_geom *gp;
487	struct g_provider *pp;
488
489	gp = dp->d_geom;
490	if (gp != NULL)
491		LIST_FOREACH(pp, &gp->provider, provider)
492			g_wither_provider(pp, ENXIO);
493}
494
495static void
496g_kern_disks(void *p, int flag __unused)
497{
498	struct sbuf *sb;
499	struct g_geom *gp;
500	char *sp;
501
502	sb = p;
503	sp = "";
504	g_topology_assert();
505	LIST_FOREACH(gp, &g_disk_class.geom, geom) {
506		sbuf_printf(sb, "%s%s", sp, gp->name);
507		sp = " ";
508	}
509	sbuf_finish(sb);
510}
511
512static int
513sysctl_disks(SYSCTL_HANDLER_ARGS)
514{
515	int error;
516	struct sbuf *sb;
517
518	sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND);
519	g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
520	error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
521	sbuf_delete(sb);
522	return error;
523}
524
525SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0,
526    sysctl_disks, "A", "names of available disks");
527
528