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/sysctl.h>
51 #include <sys/devicestat.h>
52 #include <machine/md_var.h>
55 #include <sys/mutex.h>
56 #include <geom/geom.h>
57 #include <geom/geom_disk.h>
58 #include <geom/geom_int.h>
60 #include <dev/led/led.h>
65 struct sysctl_ctx_list sysctl_ctx;
66 struct sysctl_oid *sysctl_tree;
71 static g_access_t g_disk_access;
72 static g_start_t g_disk_start;
73 static g_ioctl_t g_disk_ioctl;
74 static g_dumpconf_t g_disk_dumpconf;
75 static g_provgone_t g_disk_providergone;
77 static struct g_class g_disk_class = {
78 .name = G_DISK_CLASS_NAME,
80 .start = g_disk_start,
81 .access = g_disk_access,
82 .ioctl = g_disk_ioctl,
83 .providergone = g_disk_providergone,
84 .dumpconf = g_disk_dumpconf,
87 SYSCTL_DECL(_kern_geom);
88 static SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0,
91 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
94 g_disk_lock_giant(struct disk *dp)
97 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
102 g_disk_unlock_giant(struct disk *dp)
105 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
110 g_disk_access(struct g_provider *pp, int r, int w, int e)
113 struct g_disk_softc *sc;
116 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
120 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
122 * Allow decreasing access count even if disk is not
125 if (r <= 0 && w <= 0 && e <= 0)
133 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
134 if (dp->d_open != NULL) {
135 g_disk_lock_giant(dp);
136 error = dp->d_open(dp);
137 if (bootverbose && error != 0)
138 printf("Opened disk %s -> %d\n",
140 g_disk_unlock_giant(dp);
144 pp->mediasize = dp->d_mediasize;
145 pp->sectorsize = dp->d_sectorsize;
146 if (dp->d_maxsize == 0) {
147 printf("WARNING: Disk drive %s%d has no d_maxsize\n",
148 dp->d_name, dp->d_unit);
149 dp->d_maxsize = DFLTPHYS;
151 if (dp->d_delmaxsize == 0) {
152 if (bootverbose && dp->d_flags & DISKFLAG_CANDELETE) {
153 printf("WARNING: Disk drive %s%d has no "
154 "d_delmaxsize\n", dp->d_name, dp->d_unit);
156 dp->d_delmaxsize = dp->d_maxsize;
158 pp->stripeoffset = dp->d_stripeoffset;
159 pp->stripesize = dp->d_stripesize;
160 dp->d_flags |= DISKFLAG_OPEN;
161 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
162 if (dp->d_close != NULL) {
163 g_disk_lock_giant(dp);
164 error = dp->d_close(dp);
166 printf("Closed disk %s -> %d\n",
168 g_disk_unlock_giant(dp);
170 sc->state = G_STATE_ACTIVE;
172 led_set(sc->led, "0");
173 dp->d_flags &= ~DISKFLAG_OPEN;
179 g_disk_kerneldump(struct bio *bp, struct disk *dp)
181 struct g_kerneldump *gkd;
184 gkd = (struct g_kerneldump*)bp->bio_data;
185 gp = bp->bio_to->geom;
186 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
187 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
188 if (dp->d_dump == NULL) {
189 g_io_deliver(bp, ENODEV);
192 gkd->di.dumper = dp->d_dump;
194 gkd->di.blocksize = dp->d_sectorsize;
195 gkd->di.maxiosize = dp->d_maxsize;
196 gkd->di.mediaoffset = gkd->offset;
197 if ((gkd->offset + gkd->length) > dp->d_mediasize)
198 gkd->length = dp->d_mediasize - gkd->offset;
199 gkd->di.mediasize = gkd->length;
204 g_disk_setstate(struct bio *bp, struct g_disk_softc *sc)
208 memcpy(&sc->state, bp->bio_data, sizeof(sc->state));
209 if (sc->led[0] != 0) {
214 case G_STATE_REBUILD:
224 led_set(sc->led, cmd);
230 g_disk_done(struct bio *bp)
233 struct g_disk_softc *sc;
235 /* See "notes" for why we need a mutex here */
236 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
237 bp2 = bp->bio_parent;
238 sc = bp2->bio_to->private;
239 bp->bio_completed = bp->bio_length - bp->bio_resid;
240 mtx_lock(&sc->done_mtx);
241 if (bp2->bio_error == 0)
242 bp2->bio_error = bp->bio_error;
243 bp2->bio_completed += bp->bio_completed;
244 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) != 0)
245 devstat_end_transaction_bio(sc->dp->d_devstat, bp);
248 if (bp2->bio_children == bp2->bio_inbed) {
249 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
250 g_io_deliver(bp2, bp2->bio_error);
252 mtx_unlock(&sc->done_mtx);
256 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
259 struct g_disk_softc *sc;
265 if (dp->d_ioctl == NULL)
267 g_disk_lock_giant(dp);
268 error = dp->d_ioctl(dp, cmd, data, fflag, td);
269 g_disk_unlock_giant(dp);
274 g_disk_start(struct bio *bp)
276 struct bio *bp2, *bp3;
278 struct g_disk_softc *sc;
282 sc = bp->bio_to->private;
283 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
284 g_io_deliver(bp, ENXIO);
288 switch(bp->bio_cmd) {
290 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
299 bp2 = g_clone_bio(bp);
307 d_maxsize = (bp->bio_cmd == BIO_DELETE) ?
308 dp->d_delmaxsize : dp->d_maxsize;
309 bp2->bio_offset += off;
310 bp2->bio_length -= off;
311 if ((bp->bio_flags & BIO_UNMAPPED) == 0) {
312 bp2->bio_data += off;
314 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO)
316 ("unmapped bio not supported by disk %s",
318 bp2->bio_ma += off / PAGE_SIZE;
319 bp2->bio_ma_offset += off;
320 bp2->bio_ma_offset %= PAGE_SIZE;
321 bp2->bio_ma_n -= off / PAGE_SIZE;
323 if (bp2->bio_length > d_maxsize) {
325 * XXX: If we have a stripesize we should really
326 * use it here. Care should be taken in the delete
327 * case if this is done as deletes can be very
328 * sensitive to size given how they are processed.
330 bp2->bio_length = d_maxsize;
331 if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
332 bp2->bio_ma_n = howmany(
334 bp2->bio_length, PAGE_SIZE);
338 * To avoid a race, we need to grab the next bio
339 * before we schedule this one. See "notes".
341 bp3 = g_clone_bio(bp);
343 bp->bio_error = ENOMEM;
345 bp2->bio_done = g_disk_done;
346 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
347 bp2->bio_bcount = bp2->bio_length;
349 devstat_start_transaction_bio(dp->d_devstat, bp2);
350 g_disk_lock_giant(dp);
352 g_disk_unlock_giant(dp);
355 } while (bp2 != NULL);
358 /* Give the driver a chance to override */
359 if (dp->d_getattr != NULL) {
360 if (bp->bio_disk == NULL)
362 error = dp->d_getattr(bp);
367 if (g_handleattr_int(bp, "GEOM::candelete",
368 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
370 else if (g_handleattr_int(bp, "GEOM::fwsectors",
373 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
375 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
377 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
379 else if (g_handleattr(bp, "GEOM::hba_vendor",
380 &dp->d_hba_vendor, 2))
382 else if (g_handleattr(bp, "GEOM::hba_device",
383 &dp->d_hba_device, 2))
385 else if (g_handleattr(bp, "GEOM::hba_subvendor",
386 &dp->d_hba_subvendor, 2))
388 else if (g_handleattr(bp, "GEOM::hba_subdevice",
389 &dp->d_hba_subdevice, 2))
391 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
392 g_disk_kerneldump(bp, dp);
393 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
394 g_disk_setstate(bp, sc);
399 g_trace(G_T_BIO, "g_disk_flushcache(%s)",
401 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
405 bp2 = g_clone_bio(bp);
407 g_io_deliver(bp, ENOMEM);
410 bp2->bio_done = g_disk_done;
412 g_disk_lock_giant(dp);
414 g_disk_unlock_giant(dp);
420 if (error != EJUSTRETURN)
421 g_io_deliver(bp, error);
426 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
430 struct g_disk_softc *sc;
435 if (sc == NULL || (dp = sc->dp) == NULL)
437 if (indent == NULL) {
438 sbuf_printf(sb, " hd %u", dp->d_fwheads);
439 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
443 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
444 indent, dp->d_fwheads);
445 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
446 indent, dp->d_fwsectors);
447 if (dp->d_getattr != NULL) {
448 buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK);
451 bp->bio_attribute = "GEOM::ident";
452 bp->bio_length = DISK_IDENT_SIZE;
454 res = dp->d_getattr(bp);
455 sbuf_printf(sb, "%s<ident>%s</ident>\n", indent,
456 res == 0 ? buf: dp->d_ident);
457 bp->bio_attribute = "GEOM::lunid";
458 bp->bio_length = DISK_IDENT_SIZE;
460 if (dp->d_getattr(bp) == 0)
461 sbuf_printf(sb, "%s<lunid>%s</lunid>\n",
463 bp->bio_attribute = "GEOM::lunname";
464 bp->bio_length = DISK_IDENT_SIZE;
466 if (dp->d_getattr(bp) == 0)
467 sbuf_printf(sb, "%s<lunname>%s</lunname>\n",
472 sbuf_printf(sb, "%s<ident>%s</ident>\n", indent,
474 sbuf_printf(sb, "%s<descr>%s</descr>\n", indent, dp->d_descr);
479 g_disk_resize(void *ptr, int flag)
483 struct g_provider *pp;
485 if (flag == EV_CANCEL)
492 if (dp->d_destroyed || gp == NULL)
495 LIST_FOREACH(pp, &gp->provider, provider) {
496 if (pp->sectorsize != 0 &&
497 pp->sectorsize != dp->d_sectorsize)
498 g_wither_provider(pp, ENXIO);
500 g_resize_provider(pp, dp->d_mediasize);
505 g_disk_create(void *arg, int flag)
508 struct g_provider *pp;
510 struct g_disk_softc *sc;
513 if (flag == EV_CANCEL)
517 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
518 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF);
520 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
522 pp = g_new_providerf(gp, "%s", gp->name);
523 pp->mediasize = dp->d_mediasize;
524 pp->sectorsize = dp->d_sectorsize;
525 pp->stripeoffset = dp->d_stripeoffset;
526 pp->stripesize = dp->d_stripesize;
527 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0)
528 pp->flags |= G_PF_ACCEPT_UNMAPPED;
530 printf("GEOM: new disk %s\n", gp->name);
531 sysctl_ctx_init(&sc->sysctl_ctx);
532 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
533 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
534 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
535 CTLFLAG_RD, 0, tmpstr);
536 if (sc->sysctl_tree != NULL) {
537 snprintf(tmpstr, sizeof(tmpstr),
538 "kern.geom.disk.%s.led", gp->name);
539 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led));
540 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
541 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
542 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led),
547 g_error_provider(pp, 0);
551 * We get this callback after all of the consumers have gone away, and just
552 * before the provider is freed. If the disk driver provided a d_gone
553 * callback, let them know that it is okay to free resources -- they won't
554 * be getting any more accesses from GEOM.
557 g_disk_providergone(struct g_provider *pp)
560 struct g_disk_softc *sc;
562 sc = (struct g_disk_softc *)pp->private;
564 if (dp != NULL && dp->d_gone != NULL)
566 if (sc->sysctl_tree != NULL) {
567 sysctl_ctx_free(&sc->sysctl_ctx);
568 sc->sysctl_tree = NULL;
570 if (sc->led[0] != 0) {
571 led_set(sc->led, "0");
575 pp->geom->softc = NULL;
576 mtx_destroy(&sc->done_mtx);
581 g_disk_destroy(void *ptr, int flag)
585 struct g_disk_softc *sc;
595 g_wither_geom(gp, ENXIO);
601 * We only allow printable characters in disk ident,
602 * the rest is converted to 'x<HH>'.
605 g_disk_ident_adjust(char *ident, size_t size)
607 char *p, tmp[4], newid[DISK_IDENT_SIZE];
610 for (p = ident; *p != '\0'; p++) {
615 snprintf(tmp, sizeof(tmp), "x%02hhx",
616 *(unsigned char *)p);
618 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
622 strlcpy(ident, newid, size);
629 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO));
633 disk_create(struct disk *dp, int version)
636 if (version != DISK_VERSION) {
637 printf("WARNING: Attempt to add disk %s%d %s",
638 dp->d_name, dp->d_unit,
639 " using incompatible ABI version of disk(9)\n");
640 printf("WARNING: Ignoring disk %s%d\n",
641 dp->d_name, dp->d_unit);
644 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
645 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
646 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
647 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
648 if (dp->d_devstat == NULL)
649 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
650 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
651 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
653 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
654 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
658 disk_destroy(struct disk *dp)
663 if (dp->d_devstat != NULL)
664 devstat_remove_entry(dp->d_devstat);
665 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
669 disk_gone(struct disk *dp)
672 struct g_provider *pp;
676 pp = LIST_FIRST(&gp->provider);
678 KASSERT(LIST_NEXT(pp, provider) == NULL,
679 ("geom %p has more than one provider", gp));
680 g_wither_provider(pp, ENXIO);
686 disk_attr_changed(struct disk *dp, const char *attr, int flag)
689 struct g_provider *pp;
693 LIST_FOREACH(pp, &gp->provider, provider)
694 (void)g_attr_changed(pp, attr, flag);
698 disk_media_changed(struct disk *dp, int flag)
701 struct g_provider *pp;
705 pp = LIST_FIRST(&gp->provider);
707 KASSERT(LIST_NEXT(pp, provider) == NULL,
708 ("geom %p has more than one provider", gp));
709 g_media_changed(pp, flag);
715 disk_media_gone(struct disk *dp, int flag)
718 struct g_provider *pp;
722 pp = LIST_FIRST(&gp->provider);
724 KASSERT(LIST_NEXT(pp, provider) == NULL,
725 ("geom %p has more than one provider", gp));
726 g_media_gone(pp, flag);
732 disk_resize(struct disk *dp, int flag)
735 if (dp->d_destroyed || dp->d_geom == NULL)
738 return (g_post_event(g_disk_resize, dp, flag, NULL));
742 g_kern_disks(void *p, int flag __unused)
751 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
752 sbuf_printf(sb, "%s%s", sp, gp->name);
759 sysctl_disks(SYSCTL_HANDLER_ARGS)
764 sb = sbuf_new_auto();
765 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
766 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
771 SYSCTL_PROC(_kern, OID_AUTO, disks,
772 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
773 sysctl_disks, "A", "names of available disks");