2 * Copyright (c) 2002 Poul-Henning Kamp
3 * Copyright (c) 2002 Networks Associates Technology, Inc.
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.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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
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
36 #include <sys/cdefs.h>
37 __FBSDID("$FreeBSD$");
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/kernel.h>
44 #include <sys/sysctl.h>
46 #include <sys/ctype.h>
47 #include <sys/fcntl.h>
48 #include <sys/malloc.h>
50 #include <sys/devicestat.h>
51 #include <machine/md_var.h>
54 #include <sys/mutex.h>
55 #include <geom/geom.h>
56 #include <geom/geom_disk.h>
57 #include <geom/geom_int.h>
59 #include <dev/led/led.h>
64 struct sysctl_ctx_list sysctl_ctx;
65 struct sysctl_oid *sysctl_tree;
70 static g_access_t g_disk_access;
71 static g_start_t g_disk_start;
72 static g_ioctl_t g_disk_ioctl;
73 static g_dumpconf_t g_disk_dumpconf;
74 static g_provgone_t g_disk_providergone;
76 static struct g_class g_disk_class = {
79 .start = g_disk_start,
80 .access = g_disk_access,
81 .ioctl = g_disk_ioctl,
82 .providergone = g_disk_providergone,
83 .dumpconf = g_disk_dumpconf,
86 SYSCTL_DECL(_kern_geom);
87 static SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0,
90 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
93 g_disk_lock_giant(struct disk *dp)
96 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
101 g_disk_unlock_giant(struct disk *dp)
104 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
109 g_disk_access(struct g_provider *pp, int r, int w, int e)
112 struct g_disk_softc *sc;
115 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
119 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
121 * Allow decreasing access count even if disk is not
124 if (r <= 0 && w <= 0 && e <= 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",
139 g_disk_unlock_giant(dp);
143 pp->mediasize = dp->d_mediasize;
144 pp->sectorsize = dp->d_sectorsize;
145 if (dp->d_flags & DISKFLAG_CANDELETE)
146 pp->flags |= G_PF_CANDELETE;
148 pp->flags &= ~G_PF_CANDELETE;
149 pp->stripeoffset = dp->d_stripeoffset;
150 pp->stripesize = dp->d_stripesize;
151 dp->d_flags |= DISKFLAG_OPEN;
152 if (dp->d_maxsize == 0) {
153 printf("WARNING: Disk drive %s%d has no d_maxsize\n",
154 dp->d_name, dp->d_unit);
155 dp->d_maxsize = DFLTPHYS;
157 if (dp->d_delmaxsize == 0) {
158 if (bootverbose && dp->d_flags & DISKFLAG_CANDELETE) {
159 printf("WARNING: Disk drive %s%d has no "
160 "d_delmaxsize\n", dp->d_name, dp->d_unit);
162 dp->d_delmaxsize = dp->d_maxsize;
164 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
165 if (dp->d_close != NULL) {
166 g_disk_lock_giant(dp);
167 error = dp->d_close(dp);
169 printf("Closed disk %s -> %d\n",
171 g_disk_unlock_giant(dp);
173 sc->state = G_STATE_ACTIVE;
175 led_set(sc->led, "0");
176 dp->d_flags &= ~DISKFLAG_OPEN;
182 g_disk_kerneldump(struct bio *bp, struct disk *dp)
184 struct g_kerneldump *gkd;
187 gkd = (struct g_kerneldump*)bp->bio_data;
188 gp = bp->bio_to->geom;
189 g_trace(G_T_TOPOLOGY, "g_disk_kerneldump(%s, %jd, %jd)",
190 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
191 if (dp->d_dump == NULL) {
192 g_io_deliver(bp, ENODEV);
195 gkd->di.dumper = dp->d_dump;
197 gkd->di.blocksize = dp->d_sectorsize;
198 gkd->di.maxiosize = dp->d_maxsize;
199 gkd->di.mediaoffset = gkd->offset;
200 if ((gkd->offset + gkd->length) > dp->d_mediasize)
201 gkd->length = dp->d_mediasize - gkd->offset;
202 gkd->di.mediasize = gkd->length;
207 g_disk_setstate(struct bio *bp, struct g_disk_softc *sc)
211 memcpy(&sc->state, bp->bio_data, sizeof(sc->state));
212 if (sc->led[0] != 0) {
217 case G_STATE_REBUILD:
227 led_set(sc->led, cmd);
233 g_disk_done(struct bio *bp)
237 struct g_disk_softc *sc;
239 /* See "notes" for why we need a mutex here */
240 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
241 bp2 = bp->bio_parent;
242 sc = bp2->bio_to->private;
243 bp->bio_completed = bp->bio_length - bp->bio_resid;
245 mtx_lock(&sc->done_mtx);
246 if (bp2->bio_error == 0)
247 bp2->bio_error = bp->bio_error;
248 bp2->bio_completed += bp->bio_completed;
249 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) != 0)
250 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
252 if (bp2->bio_children == bp2->bio_inbed) {
253 mtx_unlock(&sc->done_mtx);
254 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
255 g_io_deliver(bp2, bp2->bio_error);
257 mtx_unlock(&sc->done_mtx);
262 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
265 struct g_disk_softc *sc;
271 if (dp->d_ioctl == NULL)
273 g_disk_lock_giant(dp);
274 error = dp->d_ioctl(dp, cmd, data, fflag, td);
275 g_disk_unlock_giant(dp);
280 g_disk_start(struct bio *bp)
282 struct bio *bp2, *bp3;
284 struct g_disk_softc *sc;
288 sc = bp->bio_to->private;
289 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
290 g_io_deliver(bp, ENXIO);
294 switch(bp->bio_cmd) {
296 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
305 bp2 = g_clone_bio(bp);
313 d_maxsize = (bp->bio_cmd == BIO_DELETE &&
314 (dp->d_flags & DISKFLAG_LACKS_DELMAX) == 0) ?
315 dp->d_delmaxsize : dp->d_maxsize;
317 bp2->bio_offset += off;
318 bp2->bio_length -= off;
319 if ((bp->bio_flags & BIO_UNMAPPED) == 0) {
320 bp2->bio_data += off;
322 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO)
324 ("unmapped bio not supported by disk %s",
326 bp2->bio_ma += off / PAGE_SIZE;
327 bp2->bio_ma_offset += off;
328 bp2->bio_ma_offset %= PAGE_SIZE;
329 bp2->bio_ma_n -= off / PAGE_SIZE;
331 if (bp2->bio_length > d_maxsize) {
333 * XXX: If we have a stripesize we should really
334 * use it here. Care should be taken in the delete
335 * case if this is done as deletes can be very
336 * sensitive to size given how they are processed.
338 bp2->bio_length = d_maxsize;
339 if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
340 bp2->bio_ma_n = howmany(
342 bp2->bio_length, PAGE_SIZE);
346 * To avoid a race, we need to grab the next bio
347 * before we schedule this one. See "notes".
349 bp3 = g_clone_bio(bp);
351 bp->bio_error = ENOMEM;
353 bp2->bio_done = g_disk_done;
354 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
355 bp2->bio_bcount = bp2->bio_length;
357 devstat_start_transaction_bio(dp->d_devstat, bp2);
358 g_disk_lock_giant(dp);
360 g_disk_unlock_giant(dp);
363 } while (bp2 != NULL);
366 /* Give the driver a chance to override */
367 if (dp->d_getattr != NULL) {
368 if (bp->bio_disk == NULL)
370 error = dp->d_getattr(bp);
375 if (g_handleattr_int(bp, "GEOM::candelete",
376 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
378 else if (g_handleattr_int(bp, "GEOM::fwsectors",
381 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
383 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
385 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
387 else if (g_handleattr(bp, "GEOM::hba_vendor",
388 &dp->d_hba_vendor, 2))
390 else if (g_handleattr(bp, "GEOM::hba_device",
391 &dp->d_hba_device, 2))
393 else if (g_handleattr(bp, "GEOM::hba_subvendor",
394 &dp->d_hba_subvendor, 2))
396 else if (g_handleattr(bp, "GEOM::hba_subdevice",
397 &dp->d_hba_subdevice, 2))
399 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
400 g_disk_kerneldump(bp, dp);
401 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
402 g_disk_setstate(bp, sc);
407 g_trace(G_T_BIO, "g_disk_flushcache(%s)",
409 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
413 bp2 = g_clone_bio(bp);
415 g_io_deliver(bp, ENOMEM);
418 bp2->bio_done = g_disk_done;
420 g_disk_lock_giant(dp);
422 g_disk_unlock_giant(dp);
428 if (error != EJUSTRETURN)
429 g_io_deliver(bp, error);
434 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
438 struct g_disk_softc *sc;
443 if (sc == NULL || (dp = sc->dp) == NULL)
445 if (indent == NULL) {
446 sbuf_printf(sb, " hd %u", dp->d_fwheads);
447 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
451 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
452 indent, dp->d_fwheads);
453 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
454 indent, dp->d_fwsectors);
455 if (dp->d_getattr != NULL) {
456 buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK);
459 bp->bio_attribute = "GEOM::ident";
460 bp->bio_length = DISK_IDENT_SIZE;
462 res = dp->d_getattr(bp);
463 sbuf_printf(sb, "%s<ident>", indent);
464 g_conf_printf_escaped(sb, "%s",
465 res == 0 ? buf: dp->d_ident);
466 sbuf_printf(sb, "</ident>\n");
467 bp->bio_attribute = "GEOM::lunid";
468 bp->bio_length = DISK_IDENT_SIZE;
470 if (dp->d_getattr(bp) == 0) {
471 sbuf_printf(sb, "%s<lunid>", indent);
472 g_conf_printf_escaped(sb, "%s", buf);
473 sbuf_printf(sb, "</lunid>\n");
475 bp->bio_attribute = "GEOM::lunname";
476 bp->bio_length = DISK_IDENT_SIZE;
478 if (dp->d_getattr(bp) == 0) {
479 sbuf_printf(sb, "%s<lunname>", indent);
480 g_conf_printf_escaped(sb, "%s", buf);
481 sbuf_printf(sb, "</lunname>\n");
486 sbuf_printf(sb, "%s<ident>", indent);
487 g_conf_printf_escaped(sb, "%s", dp->d_ident);
488 sbuf_printf(sb, "</ident>\n");
490 sbuf_printf(sb, "%s<descr>", indent);
491 g_conf_printf_escaped(sb, "%s", dp->d_descr);
492 sbuf_printf(sb, "</descr>\n");
497 g_disk_create(void *arg, int flag)
500 struct g_provider *pp;
502 struct g_disk_softc *sc;
505 if (flag == EV_CANCEL)
509 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
510 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF);
512 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
514 pp = g_new_providerf(gp, "%s", gp->name);
515 pp->mediasize = dp->d_mediasize;
516 pp->sectorsize = dp->d_sectorsize;
517 if (dp->d_flags & DISKFLAG_CANDELETE)
518 pp->flags |= G_PF_CANDELETE;
519 pp->stripeoffset = dp->d_stripeoffset;
520 pp->stripesize = dp->d_stripesize;
521 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0)
522 pp->flags |= G_PF_ACCEPT_UNMAPPED;
524 printf("GEOM: new disk %s\n", gp->name);
525 sysctl_ctx_init(&sc->sysctl_ctx);
526 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
527 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
528 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
529 CTLFLAG_RD, 0, tmpstr);
530 if (sc->sysctl_tree != NULL) {
531 snprintf(tmpstr, sizeof(tmpstr),
532 "kern.geom.disk.%s.led", gp->name);
533 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led));
534 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
535 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
536 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led),
541 g_error_provider(pp, 0);
545 * We get this callback after all of the consumers have gone away, and just
546 * before the provider is freed. If the disk driver provided a d_gone
547 * callback, let them know that it is okay to free resources -- they won't
548 * be getting any more accesses from GEOM.
551 g_disk_providergone(struct g_provider *pp)
554 struct g_disk_softc *sc;
556 sc = (struct g_disk_softc *)pp->private;
560 * FreeBSD 9 started with VERSION_01 of the struct disk structure.
561 * However, g_gone was added in the middle of the branch. To
562 * cope with version being missing from struct disk, we set a flag
563 * in g_disk_create for VERSION_01 and avoid touching the d_gone
564 * field for old consumers.
566 if (dp != NULL && (dp->d_flags & DISKFLAG_LACKS_GONE) == 0 &&
569 if (sc->sysctl_tree != NULL) {
570 sysctl_ctx_free(&sc->sysctl_ctx);
571 sc->sysctl_tree = NULL;
573 if (sc->led[0] != 0) {
574 led_set(sc->led, "0");
578 pp->geom->softc = NULL;
579 mtx_destroy(&sc->done_mtx);
584 g_disk_destroy(void *ptr, int flag)
588 struct g_disk_softc *sc;
598 g_wither_geom(gp, ENXIO);
604 * We only allow printable characters in disk ident,
605 * the rest is converted to 'x<HH>'.
608 g_disk_ident_adjust(char *ident, size_t size)
610 char *p, tmp[4], newid[DISK_IDENT_SIZE];
613 for (p = ident; *p != '\0'; p++) {
618 snprintf(tmp, sizeof(tmp), "x%02hhx",
619 *(unsigned char *)p);
621 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
625 strlcpy(ident, newid, size);
632 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO));
636 disk_create(struct disk *dp, int version)
639 if (version != DISK_VERSION_03 && version != DISK_VERSION_02 &&
640 version != DISK_VERSION_01) {
641 printf("WARNING: Attempt to add disk %s%d %s",
642 dp->d_name, dp->d_unit,
643 " using incompatible ABI version of disk(9)\n");
644 printf("WARNING: Ignoring disk %s%d\n",
645 dp->d_name, dp->d_unit);
648 if (version == DISK_VERSION_01)
649 dp->d_flags |= DISKFLAG_LACKS_GONE;
650 if (version < DISK_VERSION_03)
651 dp->d_flags |= DISKFLAG_LACKS_DELMAX;
652 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
653 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
654 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
655 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
656 if (dp->d_devstat == NULL)
657 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
658 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
659 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
661 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
662 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
666 disk_destroy(struct disk *dp)
671 if (dp->d_devstat != NULL)
672 devstat_remove_entry(dp->d_devstat);
673 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
677 disk_gone(struct disk *dp)
680 struct g_provider *pp;
684 pp = LIST_FIRST(&gp->provider);
686 KASSERT(LIST_NEXT(pp, provider) == NULL,
687 ("geom %p has more than one provider", gp));
688 g_wither_provider(pp, ENXIO);
694 disk_attr_changed(struct disk *dp, const char *attr, int flag)
697 struct g_provider *pp;
701 LIST_FOREACH(pp, &gp->provider, provider)
702 (void)g_attr_changed(pp, attr, flag);
706 disk_media_changed(struct disk *dp, int flag)
709 struct g_provider *pp;
713 pp = LIST_FIRST(&gp->provider);
715 KASSERT(LIST_NEXT(pp, provider) == NULL,
716 ("geom %p has more than one provider", gp));
717 g_media_changed(pp, flag);
723 disk_media_gone(struct disk *dp, int flag)
726 struct g_provider *pp;
730 pp = LIST_FIRST(&gp->provider);
732 KASSERT(LIST_NEXT(pp, provider) == NULL,
733 ("geom %p has more than one provider", gp));
734 g_media_gone(pp, flag);
740 g_kern_disks(void *p, int flag __unused)
749 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
750 sbuf_printf(sb, "%s%s", sp, gp->name);
757 sysctl_disks(SYSCTL_HANDLER_ARGS)
762 sb = sbuf_new_auto();
763 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
764 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
769 SYSCTL_PROC(_kern, OID_AUTO, disks,
770 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
771 sysctl_disks, "A", "names of available disks");