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;
72 static g_access_t g_disk_access;
73 static g_start_t g_disk_start;
74 static g_ioctl_t g_disk_ioctl;
75 static g_dumpconf_t g_disk_dumpconf;
76 static g_provgone_t g_disk_providergone;
78 static struct g_class g_disk_class = {
79 .name = G_DISK_CLASS_NAME,
81 .start = g_disk_start,
82 .access = g_disk_access,
83 .ioctl = g_disk_ioctl,
84 .providergone = g_disk_providergone,
85 .dumpconf = g_disk_dumpconf,
88 SYSCTL_DECL(_kern_geom);
89 static SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0,
92 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
95 g_disk_access(struct g_provider *pp, int r, int w, int e)
98 struct g_disk_softc *sc;
101 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
105 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
107 * Allow decreasing access count even if disk is not
110 if (r <= 0 && w <= 0 && e <= 0)
118 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
119 if (dp->d_open != NULL) {
120 error = dp->d_open(dp);
121 if (bootverbose && error != 0)
122 printf("Opened disk %s -> %d\n",
127 pp->mediasize = dp->d_mediasize;
128 pp->sectorsize = dp->d_sectorsize;
129 if (dp->d_maxsize == 0) {
130 printf("WARNING: Disk drive %s%d has no d_maxsize\n",
131 dp->d_name, dp->d_unit);
132 dp->d_maxsize = DFLTPHYS;
134 if (dp->d_delmaxsize == 0) {
135 if (bootverbose && dp->d_flags & DISKFLAG_CANDELETE) {
136 printf("WARNING: Disk drive %s%d has no "
137 "d_delmaxsize\n", dp->d_name, dp->d_unit);
139 dp->d_delmaxsize = dp->d_maxsize;
141 pp->stripeoffset = dp->d_stripeoffset;
142 pp->stripesize = dp->d_stripesize;
143 dp->d_flags |= DISKFLAG_OPEN;
144 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
145 if (dp->d_close != NULL) {
146 error = dp->d_close(dp);
148 printf("Closed disk %s -> %d\n",
151 sc->state = G_STATE_ACTIVE;
153 led_set(sc->led, "0");
154 dp->d_flags &= ~DISKFLAG_OPEN;
160 g_disk_kerneldump(struct bio *bp, struct disk *dp)
162 struct g_kerneldump *gkd;
165 gkd = (struct g_kerneldump*)bp->bio_data;
166 gp = bp->bio_to->geom;
167 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
168 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
169 if (dp->d_dump == NULL) {
170 g_io_deliver(bp, ENODEV);
173 gkd->di.dumper = dp->d_dump;
175 gkd->di.blocksize = dp->d_sectorsize;
176 gkd->di.maxiosize = dp->d_maxsize;
177 gkd->di.mediaoffset = gkd->offset;
178 if ((gkd->offset + gkd->length) > dp->d_mediasize)
179 gkd->length = dp->d_mediasize - gkd->offset;
180 gkd->di.mediasize = gkd->length;
185 g_disk_setstate(struct bio *bp, struct g_disk_softc *sc)
189 memcpy(&sc->state, bp->bio_data, sizeof(sc->state));
190 if (sc->led[0] != 0) {
195 case G_STATE_REBUILD:
205 led_set(sc->led, cmd);
211 g_disk_done(struct bio *bp)
215 struct g_disk_softc *sc;
217 /* See "notes" for why we need a mutex here */
218 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
219 bp2 = bp->bio_parent;
220 sc = bp2->bio_to->private;
221 bp->bio_completed = bp->bio_length - bp->bio_resid;
223 mtx_lock(&sc->done_mtx);
224 if (bp2->bio_error == 0)
225 bp2->bio_error = bp->bio_error;
226 bp2->bio_completed += bp->bio_completed;
227 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) != 0)
228 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
230 if (bp2->bio_children == bp2->bio_inbed) {
231 mtx_unlock(&sc->done_mtx);
232 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
233 g_io_deliver(bp2, bp2->bio_error);
235 mtx_unlock(&sc->done_mtx);
240 g_disk_done_single(struct bio *bp)
243 struct g_disk_softc *sc;
245 bp->bio_completed = bp->bio_length - bp->bio_resid;
246 bp->bio_done = (void *)bp->bio_to;
247 bp->bio_to = LIST_FIRST(&bp->bio_disk->d_geom->provider);
248 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) != 0) {
250 sc = bp->bio_to->private;
251 mtx_lock(&sc->done_mtx);
252 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
253 mtx_unlock(&sc->done_mtx);
255 g_io_deliver(bp, bp->bio_error);
259 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
262 struct g_disk_softc *sc;
268 if (dp->d_ioctl == NULL)
270 error = dp->d_ioctl(dp, cmd, data, fflag, td);
275 g_disk_start(struct bio *bp)
277 struct bio *bp2, *bp3;
279 struct g_disk_softc *sc;
281 off_t d_maxsize, off;
283 sc = bp->bio_to->private;
284 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
285 g_io_deliver(bp, ENXIO);
289 switch(bp->bio_cmd) {
291 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
298 d_maxsize = (bp->bio_cmd == BIO_DELETE) ?
299 dp->d_delmaxsize : dp->d_maxsize;
300 if (bp->bio_length <= d_maxsize) {
302 bp->bio_to = (void *)bp->bio_done;
303 bp->bio_done = g_disk_done_single;
304 bp->bio_pblkno = bp->bio_offset / dp->d_sectorsize;
305 bp->bio_bcount = bp->bio_length;
306 mtx_lock(&sc->start_mtx);
307 devstat_start_transaction_bio(dp->d_devstat, bp);
308 mtx_unlock(&sc->start_mtx);
314 bp2 = g_clone_bio(bp);
320 bp2->bio_offset += off;
321 bp2->bio_length -= off;
322 if ((bp->bio_flags & BIO_UNMAPPED) == 0) {
323 bp2->bio_data += off;
325 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO)
327 ("unmapped bio not supported by disk %s",
329 bp2->bio_ma += off / PAGE_SIZE;
330 bp2->bio_ma_offset += off;
331 bp2->bio_ma_offset %= PAGE_SIZE;
332 bp2->bio_ma_n -= off / PAGE_SIZE;
334 if (bp2->bio_length > d_maxsize) {
336 * XXX: If we have a stripesize we should really
337 * use it here. Care should be taken in the delete
338 * case if this is done as deletes can be very
339 * sensitive to size given how they are processed.
341 bp2->bio_length = d_maxsize;
342 if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
343 bp2->bio_ma_n = howmany(
345 bp2->bio_length, PAGE_SIZE);
349 * To avoid a race, we need to grab the next bio
350 * before we schedule this one. See "notes".
352 bp3 = g_clone_bio(bp);
354 bp->bio_error = ENOMEM;
356 bp2->bio_done = g_disk_done;
357 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
358 bp2->bio_bcount = bp2->bio_length;
360 mtx_lock(&sc->start_mtx);
361 devstat_start_transaction_bio(dp->d_devstat, bp2);
362 mtx_unlock(&sc->start_mtx);
366 } while (bp2 != NULL);
369 /* Give the driver a chance to override */
370 if (dp->d_getattr != NULL) {
371 if (bp->bio_disk == NULL)
373 error = dp->d_getattr(bp);
378 if (g_handleattr_int(bp, "GEOM::candelete",
379 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
381 else if (g_handleattr_int(bp, "GEOM::fwsectors",
384 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
386 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
388 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
390 else if (g_handleattr_uint16_t(bp, "GEOM::hba_vendor",
393 else if (g_handleattr_uint16_t(bp, "GEOM::hba_device",
396 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subvendor",
397 dp->d_hba_subvendor))
399 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subdevice",
400 dp->d_hba_subdevice))
402 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
403 g_disk_kerneldump(bp, dp);
404 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
405 g_disk_setstate(bp, sc);
406 else if (g_handleattr_uint16_t(bp, "GEOM::rotation_rate",
407 dp->d_rotation_rate))
413 g_trace(G_T_BIO, "g_disk_flushcache(%s)",
415 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
420 bp->bio_to = (void *)bp->bio_done;
421 bp->bio_done = g_disk_done_single;
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_resize(void *ptr, int flag)
501 struct g_provider *pp;
503 if (flag == EV_CANCEL)
510 if (dp->d_destroyed || gp == NULL)
513 LIST_FOREACH(pp, &gp->provider, provider) {
514 if (pp->sectorsize != 0 &&
515 pp->sectorsize != dp->d_sectorsize)
516 g_wither_provider(pp, ENXIO);
518 g_resize_provider(pp, dp->d_mediasize);
523 g_disk_create(void *arg, int flag)
526 struct g_provider *pp;
528 struct g_disk_softc *sc;
531 if (flag == EV_CANCEL)
535 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
536 mtx_init(&sc->start_mtx, "g_disk_start", NULL, MTX_DEF);
537 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF);
539 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
541 pp = g_new_providerf(gp, "%s", gp->name);
542 devstat_remove_entry(pp->stat);
544 dp->d_devstat->id = pp;
545 pp->mediasize = dp->d_mediasize;
546 pp->sectorsize = dp->d_sectorsize;
547 pp->stripeoffset = dp->d_stripeoffset;
548 pp->stripesize = dp->d_stripesize;
549 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0)
550 pp->flags |= G_PF_ACCEPT_UNMAPPED;
551 if ((dp->d_flags & DISKFLAG_DIRECT_COMPLETION) != 0)
552 pp->flags |= G_PF_DIRECT_SEND;
553 pp->flags |= G_PF_DIRECT_RECEIVE;
555 printf("GEOM: new disk %s\n", gp->name);
556 sysctl_ctx_init(&sc->sysctl_ctx);
557 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
558 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
559 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
560 CTLFLAG_RD, 0, tmpstr);
561 if (sc->sysctl_tree != NULL) {
562 snprintf(tmpstr, sizeof(tmpstr),
563 "kern.geom.disk.%s.led", gp->name);
564 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led));
565 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
566 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
567 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led),
572 g_error_provider(pp, 0);
576 * We get this callback after all of the consumers have gone away, and just
577 * before the provider is freed. If the disk driver provided a d_gone
578 * callback, let them know that it is okay to free resources -- they won't
579 * be getting any more accesses from GEOM.
582 g_disk_providergone(struct g_provider *pp)
585 struct g_disk_softc *sc;
587 sc = (struct g_disk_softc *)pp->private;
589 if (dp != NULL && dp->d_gone != NULL)
591 if (sc->sysctl_tree != NULL) {
592 sysctl_ctx_free(&sc->sysctl_ctx);
593 sc->sysctl_tree = NULL;
595 if (sc->led[0] != 0) {
596 led_set(sc->led, "0");
600 pp->geom->softc = NULL;
601 mtx_destroy(&sc->done_mtx);
602 mtx_destroy(&sc->start_mtx);
607 g_disk_destroy(void *ptr, int flag)
611 struct g_disk_softc *sc;
621 g_wither_geom(gp, ENXIO);
627 * We only allow printable characters in disk ident,
628 * the rest is converted to 'x<HH>'.
631 g_disk_ident_adjust(char *ident, size_t size)
633 char *p, tmp[4], newid[DISK_IDENT_SIZE];
636 for (p = ident; *p != '\0'; p++) {
641 snprintf(tmp, sizeof(tmp), "x%02hhx",
642 *(unsigned char *)p);
644 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
648 strlcpy(ident, newid, size);
655 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO));
659 disk_create(struct disk *dp, int version)
662 if (version != DISK_VERSION) {
663 printf("WARNING: Attempt to add disk %s%d %s",
664 dp->d_name, dp->d_unit,
665 " using incompatible ABI version of disk(9)\n");
666 printf("WARNING: Ignoring disk %s%d\n",
667 dp->d_name, dp->d_unit);
670 if (dp->d_flags & DISKFLAG_RESERVED) {
671 printf("WARNING: Attempt to add non-MPSAFE disk %s%d\n",
672 dp->d_name, dp->d_unit);
673 printf("WARNING: Ignoring disk %s%d\n",
674 dp->d_name, dp->d_unit);
677 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
678 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
679 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
680 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
681 if (dp->d_devstat == NULL)
682 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
683 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
684 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
686 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
687 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
691 disk_destroy(struct disk *dp)
696 if (dp->d_devstat != NULL)
697 devstat_remove_entry(dp->d_devstat);
698 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
702 disk_gone(struct disk *dp)
705 struct g_provider *pp;
709 pp = LIST_FIRST(&gp->provider);
711 KASSERT(LIST_NEXT(pp, provider) == NULL,
712 ("geom %p has more than one provider", gp));
713 g_wither_provider(pp, ENXIO);
719 disk_attr_changed(struct disk *dp, const char *attr, int flag)
722 struct g_provider *pp;
726 LIST_FOREACH(pp, &gp->provider, provider)
727 (void)g_attr_changed(pp, attr, flag);
731 disk_media_changed(struct disk *dp, int flag)
734 struct g_provider *pp;
738 pp = LIST_FIRST(&gp->provider);
740 KASSERT(LIST_NEXT(pp, provider) == NULL,
741 ("geom %p has more than one provider", gp));
742 g_media_changed(pp, flag);
748 disk_media_gone(struct disk *dp, int flag)
751 struct g_provider *pp;
755 pp = LIST_FIRST(&gp->provider);
757 KASSERT(LIST_NEXT(pp, provider) == NULL,
758 ("geom %p has more than one provider", gp));
759 g_media_gone(pp, flag);
765 disk_resize(struct disk *dp, int flag)
768 if (dp->d_destroyed || dp->d_geom == NULL)
771 return (g_post_event(g_disk_resize, dp, flag, NULL));
775 g_kern_disks(void *p, int flag __unused)
784 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
785 sbuf_printf(sb, "%s%s", sp, gp->name);
792 sysctl_disks(SYSCTL_HANDLER_ARGS)
797 sb = sbuf_new_auto();
798 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
799 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
804 SYSCTL_PROC(_kern, OID_AUTO, disks,
805 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
806 sysctl_disks, "A", "names of available disks");