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;
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_access(struct g_provider *pp, int r, int w, int e)
97 struct g_disk_softc *sc;
100 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
104 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
106 * Allow decreasing access count even if disk is not
109 if (r <= 0 && w <= 0 && e <= 0)
117 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
118 if (dp->d_open != NULL) {
119 error = dp->d_open(dp);
120 if (bootverbose && error != 0)
121 printf("Opened disk %s -> %d\n",
126 pp->mediasize = dp->d_mediasize;
127 pp->sectorsize = dp->d_sectorsize;
128 if (dp->d_maxsize == 0) {
129 printf("WARNING: Disk drive %s%d has no d_maxsize\n",
130 dp->d_name, dp->d_unit);
131 dp->d_maxsize = DFLTPHYS;
133 if (dp->d_delmaxsize == 0) {
134 if (bootverbose && dp->d_flags & DISKFLAG_CANDELETE) {
135 printf("WARNING: Disk drive %s%d has no "
136 "d_delmaxsize\n", dp->d_name, dp->d_unit);
138 dp->d_delmaxsize = dp->d_maxsize;
140 pp->stripeoffset = dp->d_stripeoffset;
141 pp->stripesize = dp->d_stripesize;
142 dp->d_flags |= DISKFLAG_OPEN;
143 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
144 if (dp->d_close != NULL) {
145 error = dp->d_close(dp);
147 printf("Closed disk %s -> %d\n",
150 sc->state = G_STATE_ACTIVE;
152 led_set(sc->led, "0");
153 dp->d_flags &= ~DISKFLAG_OPEN;
159 g_disk_kerneldump(struct bio *bp, struct disk *dp)
161 struct g_kerneldump *gkd;
164 gkd = (struct g_kerneldump*)bp->bio_data;
165 gp = bp->bio_to->geom;
166 g_trace(G_T_TOPOLOGY, "g_disk_kerneldump(%s, %jd, %jd)",
167 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
168 if (dp->d_dump == NULL) {
169 g_io_deliver(bp, ENODEV);
172 gkd->di.dumper = dp->d_dump;
174 gkd->di.blocksize = dp->d_sectorsize;
175 gkd->di.maxiosize = dp->d_maxsize;
176 gkd->di.mediaoffset = gkd->offset;
177 if ((gkd->offset + gkd->length) > dp->d_mediasize)
178 gkd->length = dp->d_mediasize - gkd->offset;
179 gkd->di.mediasize = gkd->length;
184 g_disk_setstate(struct bio *bp, struct g_disk_softc *sc)
188 memcpy(&sc->state, bp->bio_data, sizeof(sc->state));
189 if (sc->led[0] != 0) {
194 case G_STATE_REBUILD:
204 led_set(sc->led, cmd);
210 g_disk_done(struct bio *bp)
214 struct g_disk_softc *sc;
216 /* See "notes" for why we need a mutex here */
217 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
218 bp2 = bp->bio_parent;
219 sc = bp2->bio_to->private;
220 bp->bio_completed = bp->bio_length - bp->bio_resid;
222 mtx_lock(&sc->done_mtx);
223 if (bp2->bio_error == 0)
224 bp2->bio_error = bp->bio_error;
225 bp2->bio_completed += bp->bio_completed;
226 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0)
227 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
229 if (bp2->bio_children == bp2->bio_inbed) {
230 mtx_unlock(&sc->done_mtx);
231 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
232 g_io_deliver(bp2, bp2->bio_error);
234 mtx_unlock(&sc->done_mtx);
239 g_disk_done_single(struct bio *bp)
242 struct g_disk_softc *sc;
244 bp->bio_completed = bp->bio_length - bp->bio_resid;
245 bp->bio_done = (void *)bp->bio_to;
246 bp->bio_to = LIST_FIRST(&bp->bio_disk->d_geom->provider);
247 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0) {
249 sc = bp->bio_to->private;
250 mtx_lock(&sc->done_mtx);
251 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
252 mtx_unlock(&sc->done_mtx);
254 g_io_deliver(bp, bp->bio_error);
258 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
261 struct g_disk_softc *sc;
267 if (dp->d_ioctl == NULL)
269 error = dp->d_ioctl(dp, cmd, data, fflag, td);
274 g_disk_start(struct bio *bp)
276 struct bio *bp2, *bp3;
278 struct g_disk_softc *sc;
280 off_t d_maxsize, off;
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)) {
297 d_maxsize = (bp->bio_cmd == BIO_DELETE) ?
298 dp->d_delmaxsize : dp->d_maxsize;
299 if (bp->bio_length <= d_maxsize) {
301 bp->bio_to = (void *)bp->bio_done;
302 bp->bio_done = g_disk_done_single;
303 bp->bio_pblkno = bp->bio_offset / dp->d_sectorsize;
304 bp->bio_bcount = bp->bio_length;
305 mtx_lock(&sc->start_mtx);
306 devstat_start_transaction_bio(dp->d_devstat, bp);
307 mtx_unlock(&sc->start_mtx);
313 bp2 = g_clone_bio(bp);
319 bp2->bio_offset += off;
320 bp2->bio_length -= off;
321 if ((bp->bio_flags & BIO_UNMAPPED) == 0) {
322 bp2->bio_data += off;
324 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO)
326 ("unmapped bio not supported by disk %s",
328 bp2->bio_ma += off / PAGE_SIZE;
329 bp2->bio_ma_offset += off;
330 bp2->bio_ma_offset %= PAGE_SIZE;
331 bp2->bio_ma_n -= off / PAGE_SIZE;
333 if (bp2->bio_length > d_maxsize) {
335 * XXX: If we have a stripesize we should really
336 * use it here. Care should be taken in the delete
337 * case if this is done as deletes can be very
338 * sensitive to size given how they are processed.
340 bp2->bio_length = d_maxsize;
341 if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
342 bp2->bio_ma_n = howmany(
344 bp2->bio_length, PAGE_SIZE);
348 * To avoid a race, we need to grab the next bio
349 * before we schedule this one. See "notes".
351 bp3 = g_clone_bio(bp);
353 bp->bio_error = ENOMEM;
355 bp2->bio_done = g_disk_done;
356 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
357 bp2->bio_bcount = bp2->bio_length;
359 mtx_lock(&sc->start_mtx);
360 devstat_start_transaction_bio(dp->d_devstat, bp2);
361 mtx_unlock(&sc->start_mtx);
365 } while (bp2 != NULL);
368 /* Give the driver a chance to override */
369 if (dp->d_getattr != NULL) {
370 if (bp->bio_disk == NULL)
372 error = dp->d_getattr(bp);
377 if (g_handleattr_int(bp, "GEOM::candelete",
378 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
380 else if (g_handleattr_int(bp, "GEOM::fwsectors",
383 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
385 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
387 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
389 else if (g_handleattr_uint16_t(bp, "GEOM::hba_vendor",
392 else if (g_handleattr_uint16_t(bp, "GEOM::hba_device",
395 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subvendor",
396 dp->d_hba_subvendor))
398 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subdevice",
399 dp->d_hba_subdevice))
401 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
402 g_disk_kerneldump(bp, dp);
403 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
404 g_disk_setstate(bp, sc);
405 else if (g_handleattr_uint16_t(bp, "GEOM::rotation_rate",
406 dp->d_rotation_rate))
412 g_trace(G_T_BIO, "g_disk_flushcache(%s)",
414 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
419 bp->bio_to = (void *)bp->bio_done;
420 bp->bio_done = g_disk_done_single;
421 mtx_lock(&sc->start_mtx);
422 devstat_start_transaction_bio(dp->d_devstat, bp);
423 mtx_unlock(&sc->start_mtx);
430 if (error != EJUSTRETURN)
431 g_io_deliver(bp, error);
436 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
440 struct g_disk_softc *sc;
445 if (sc == NULL || (dp = sc->dp) == NULL)
447 if (indent == NULL) {
448 sbuf_printf(sb, " hd %u", dp->d_fwheads);
449 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
453 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
454 indent, dp->d_fwheads);
455 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
456 indent, dp->d_fwsectors);
457 if (dp->d_getattr != NULL) {
458 buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK);
461 bp->bio_attribute = "GEOM::ident";
462 bp->bio_length = DISK_IDENT_SIZE;
464 res = dp->d_getattr(bp);
465 sbuf_printf(sb, "%s<ident>", indent);
466 g_conf_printf_escaped(sb, "%s",
467 res == 0 ? buf: dp->d_ident);
468 sbuf_printf(sb, "</ident>\n");
469 bp->bio_attribute = "GEOM::lunid";
470 bp->bio_length = DISK_IDENT_SIZE;
472 if (dp->d_getattr(bp) == 0) {
473 sbuf_printf(sb, "%s<lunid>", indent);
474 g_conf_printf_escaped(sb, "%s", buf);
475 sbuf_printf(sb, "</lunid>\n");
477 bp->bio_attribute = "GEOM::lunname";
478 bp->bio_length = DISK_IDENT_SIZE;
480 if (dp->d_getattr(bp) == 0) {
481 sbuf_printf(sb, "%s<lunname>", indent);
482 g_conf_printf_escaped(sb, "%s", buf);
483 sbuf_printf(sb, "</lunname>\n");
488 sbuf_printf(sb, "%s<ident>", indent);
489 g_conf_printf_escaped(sb, "%s", dp->d_ident);
490 sbuf_printf(sb, "</ident>\n");
492 sbuf_printf(sb, "%s<descr>", indent);
493 g_conf_printf_escaped(sb, "%s", dp->d_descr);
494 sbuf_printf(sb, "</descr>\n");
499 g_disk_resize(void *ptr, int flag)
503 struct g_provider *pp;
505 if (flag == EV_CANCEL)
512 if (dp->d_destroyed || gp == NULL)
515 LIST_FOREACH(pp, &gp->provider, provider) {
516 if (pp->sectorsize != 0 &&
517 pp->sectorsize != dp->d_sectorsize)
518 g_wither_provider(pp, ENXIO);
520 g_resize_provider(pp, dp->d_mediasize);
525 g_disk_create(void *arg, int flag)
528 struct g_provider *pp;
530 struct g_disk_softc *sc;
533 if (flag == EV_CANCEL)
537 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
538 mtx_init(&sc->start_mtx, "g_disk_start", NULL, MTX_DEF);
539 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF);
541 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
543 pp = g_new_providerf(gp, "%s", gp->name);
544 devstat_remove_entry(pp->stat);
546 dp->d_devstat->id = pp;
547 pp->mediasize = dp->d_mediasize;
548 pp->sectorsize = dp->d_sectorsize;
549 pp->stripeoffset = dp->d_stripeoffset;
550 pp->stripesize = dp->d_stripesize;
551 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0)
552 pp->flags |= G_PF_ACCEPT_UNMAPPED;
553 if ((dp->d_flags & DISKFLAG_DIRECT_COMPLETION) != 0)
554 pp->flags |= G_PF_DIRECT_SEND;
555 pp->flags |= G_PF_DIRECT_RECEIVE;
557 printf("GEOM: new disk %s\n", gp->name);
558 sysctl_ctx_init(&sc->sysctl_ctx);
559 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
560 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
561 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
562 CTLFLAG_RD, 0, tmpstr);
563 if (sc->sysctl_tree != NULL) {
564 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
565 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
566 CTLFLAG_RWTUN, sc->led, sizeof(sc->led),
571 g_error_provider(pp, 0);
575 * We get this callback after all of the consumers have gone away, and just
576 * before the provider is freed. If the disk driver provided a d_gone
577 * callback, let them know that it is okay to free resources -- they won't
578 * be getting any more accesses from GEOM.
581 g_disk_providergone(struct g_provider *pp)
584 struct g_disk_softc *sc;
586 sc = (struct g_disk_softc *)pp->private;
588 if (dp != NULL && dp->d_gone != NULL)
590 if (sc->sysctl_tree != NULL) {
591 sysctl_ctx_free(&sc->sysctl_ctx);
592 sc->sysctl_tree = NULL;
594 if (sc->led[0] != 0) {
595 led_set(sc->led, "0");
599 pp->geom->softc = NULL;
600 mtx_destroy(&sc->done_mtx);
601 mtx_destroy(&sc->start_mtx);
606 g_disk_destroy(void *ptr, int flag)
610 struct g_disk_softc *sc;
620 g_wither_geom(gp, ENXIO);
626 * We only allow printable characters in disk ident,
627 * the rest is converted to 'x<HH>'.
630 g_disk_ident_adjust(char *ident, size_t size)
632 char *p, tmp[4], newid[DISK_IDENT_SIZE];
635 for (p = ident; *p != '\0'; p++) {
640 snprintf(tmp, sizeof(tmp), "x%02hhx",
641 *(unsigned char *)p);
643 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
647 strlcpy(ident, newid, size);
654 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO));
658 disk_create(struct disk *dp, int version)
661 if (version != DISK_VERSION) {
662 printf("WARNING: Attempt to add disk %s%d %s",
663 dp->d_name, dp->d_unit,
664 " using incompatible ABI version of disk(9)\n");
665 printf("WARNING: Ignoring disk %s%d\n",
666 dp->d_name, dp->d_unit);
669 if (dp->d_flags & DISKFLAG_RESERVED) {
670 printf("WARNING: Attempt to add non-MPSAFE disk %s%d\n",
671 dp->d_name, dp->d_unit);
672 printf("WARNING: Ignoring disk %s%d\n",
673 dp->d_name, dp->d_unit);
676 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
677 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
678 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
679 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
680 if (dp->d_devstat == NULL)
681 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
682 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
683 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
685 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
686 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
690 disk_destroy(struct disk *dp)
695 if (dp->d_devstat != NULL)
696 devstat_remove_entry(dp->d_devstat);
697 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
701 disk_gone(struct disk *dp)
704 struct g_provider *pp;
708 pp = LIST_FIRST(&gp->provider);
710 KASSERT(LIST_NEXT(pp, provider) == NULL,
711 ("geom %p has more than one provider", gp));
712 g_wither_provider(pp, ENXIO);
718 disk_attr_changed(struct disk *dp, const char *attr, int flag)
721 struct g_provider *pp;
725 LIST_FOREACH(pp, &gp->provider, provider)
726 (void)g_attr_changed(pp, attr, flag);
730 disk_media_changed(struct disk *dp, int flag)
733 struct g_provider *pp;
737 pp = LIST_FIRST(&gp->provider);
739 KASSERT(LIST_NEXT(pp, provider) == NULL,
740 ("geom %p has more than one provider", gp));
741 g_media_changed(pp, flag);
747 disk_media_gone(struct disk *dp, int flag)
750 struct g_provider *pp;
754 pp = LIST_FIRST(&gp->provider);
756 KASSERT(LIST_NEXT(pp, provider) == NULL,
757 ("geom %p has more than one provider", gp));
758 g_media_gone(pp, flag);
764 disk_resize(struct disk *dp, int flag)
767 if (dp->d_destroyed || dp->d_geom == NULL)
770 return (g_post_event(g_disk_resize, dp, flag, NULL));
774 g_kern_disks(void *p, int flag __unused)
783 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
784 sbuf_printf(sb, "%s%s", sp, gp->name);
791 sysctl_disks(SYSCTL_HANDLER_ARGS)
796 sb = sbuf_new_auto();
797 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
798 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
803 SYSCTL_PROC(_kern, OID_AUTO, disks,
804 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
805 sysctl_disks, "A", "names of available disks");