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_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
242 struct g_disk_softc *sc;
248 if (dp->d_ioctl == NULL)
250 error = dp->d_ioctl(dp, cmd, data, fflag, td);
255 g_disk_start(struct bio *bp)
257 struct bio *bp2, *bp3;
259 struct g_disk_softc *sc;
263 sc = bp->bio_to->private;
264 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
265 g_io_deliver(bp, ENXIO);
269 switch(bp->bio_cmd) {
271 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
280 bp2 = g_clone_bio(bp);
288 d_maxsize = (bp->bio_cmd == BIO_DELETE) ?
289 dp->d_delmaxsize : dp->d_maxsize;
290 bp2->bio_offset += off;
291 bp2->bio_length -= off;
292 if ((bp->bio_flags & BIO_UNMAPPED) == 0) {
293 bp2->bio_data += off;
295 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO)
297 ("unmapped bio not supported by disk %s",
299 bp2->bio_ma += off / PAGE_SIZE;
300 bp2->bio_ma_offset += off;
301 bp2->bio_ma_offset %= PAGE_SIZE;
302 bp2->bio_ma_n -= off / PAGE_SIZE;
304 if (bp2->bio_length > d_maxsize) {
306 * XXX: If we have a stripesize we should really
307 * use it here. Care should be taken in the delete
308 * case if this is done as deletes can be very
309 * sensitive to size given how they are processed.
311 bp2->bio_length = d_maxsize;
312 if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
313 bp2->bio_ma_n = howmany(
315 bp2->bio_length, PAGE_SIZE);
319 * To avoid a race, we need to grab the next bio
320 * before we schedule this one. See "notes".
322 bp3 = g_clone_bio(bp);
324 bp->bio_error = ENOMEM;
326 bp2->bio_done = g_disk_done;
327 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
328 bp2->bio_bcount = bp2->bio_length;
330 mtx_lock(&sc->start_mtx);
331 devstat_start_transaction_bio(dp->d_devstat, bp2);
332 mtx_unlock(&sc->start_mtx);
336 } while (bp2 != NULL);
339 /* Give the driver a chance to override */
340 if (dp->d_getattr != NULL) {
341 if (bp->bio_disk == NULL)
343 error = dp->d_getattr(bp);
348 if (g_handleattr_int(bp, "GEOM::candelete",
349 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
351 else if (g_handleattr_int(bp, "GEOM::fwsectors",
354 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
356 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
358 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
360 else if (g_handleattr_uint16_t(bp, "GEOM::hba_vendor",
363 else if (g_handleattr_uint16_t(bp, "GEOM::hba_device",
366 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subvendor",
367 dp->d_hba_subvendor))
369 else if (g_handleattr_uint16_t(bp, "GEOM::hba_subdevice",
370 dp->d_hba_subdevice))
372 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
373 g_disk_kerneldump(bp, dp);
374 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
375 g_disk_setstate(bp, sc);
376 else if (g_handleattr_uint16_t(bp, "GEOM::rotation_rate",
377 dp->d_rotation_rate))
383 g_trace(G_T_BIO, "g_disk_flushcache(%s)",
385 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
389 bp2 = g_clone_bio(bp);
391 g_io_deliver(bp, ENOMEM);
394 bp2->bio_done = g_disk_done;
396 mtx_lock(&sc->start_mtx);
397 devstat_start_transaction_bio(dp->d_devstat, bp2);
398 mtx_unlock(&sc->start_mtx);
405 if (error != EJUSTRETURN)
406 g_io_deliver(bp, error);
411 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
415 struct g_disk_softc *sc;
420 if (sc == NULL || (dp = sc->dp) == NULL)
422 if (indent == NULL) {
423 sbuf_printf(sb, " hd %u", dp->d_fwheads);
424 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
428 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
429 indent, dp->d_fwheads);
430 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
431 indent, dp->d_fwsectors);
432 if (dp->d_getattr != NULL) {
433 buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK);
436 bp->bio_attribute = "GEOM::ident";
437 bp->bio_length = DISK_IDENT_SIZE;
439 res = dp->d_getattr(bp);
440 sbuf_printf(sb, "%s<ident>", indent);
441 g_conf_printf_escaped(sb, "%s",
442 res == 0 ? buf: dp->d_ident);
443 sbuf_printf(sb, "</ident>\n");
444 bp->bio_attribute = "GEOM::lunid";
445 bp->bio_length = DISK_IDENT_SIZE;
447 if (dp->d_getattr(bp) == 0) {
448 sbuf_printf(sb, "%s<lunid>", indent);
449 g_conf_printf_escaped(sb, "%s", buf);
450 sbuf_printf(sb, "</lunid>\n");
452 bp->bio_attribute = "GEOM::lunname";
453 bp->bio_length = DISK_IDENT_SIZE;
455 if (dp->d_getattr(bp) == 0) {
456 sbuf_printf(sb, "%s<lunname>", indent);
457 g_conf_printf_escaped(sb, "%s", buf);
458 sbuf_printf(sb, "</lunname>\n");
463 sbuf_printf(sb, "%s<ident>", indent);
464 g_conf_printf_escaped(sb, "%s", dp->d_ident);
465 sbuf_printf(sb, "</ident>\n");
467 sbuf_printf(sb, "%s<descr>", indent);
468 g_conf_printf_escaped(sb, "%s", dp->d_descr);
469 sbuf_printf(sb, "</descr>\n");
474 g_disk_resize(void *ptr, int flag)
478 struct g_provider *pp;
480 if (flag == EV_CANCEL)
487 if (dp->d_destroyed || gp == NULL)
490 LIST_FOREACH(pp, &gp->provider, provider) {
491 if (pp->sectorsize != 0 &&
492 pp->sectorsize != dp->d_sectorsize)
493 g_wither_provider(pp, ENXIO);
495 g_resize_provider(pp, dp->d_mediasize);
500 g_disk_create(void *arg, int flag)
503 struct g_provider *pp;
505 struct g_disk_softc *sc;
508 if (flag == EV_CANCEL)
512 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
513 mtx_init(&sc->start_mtx, "g_disk_start", NULL, MTX_DEF);
514 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF);
516 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
518 pp = g_new_providerf(gp, "%s", gp->name);
519 devstat_remove_entry(pp->stat);
521 dp->d_devstat->id = pp;
522 pp->mediasize = dp->d_mediasize;
523 pp->sectorsize = dp->d_sectorsize;
524 pp->stripeoffset = dp->d_stripeoffset;
525 pp->stripesize = dp->d_stripesize;
526 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0)
527 pp->flags |= G_PF_ACCEPT_UNMAPPED;
528 if ((dp->d_flags & DISKFLAG_DIRECT_COMPLETION) != 0)
529 pp->flags |= G_PF_DIRECT_SEND;
530 pp->flags |= G_PF_DIRECT_RECEIVE;
532 printf("GEOM: new disk %s\n", gp->name);
533 sysctl_ctx_init(&sc->sysctl_ctx);
534 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
535 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
536 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
537 CTLFLAG_RD, 0, tmpstr);
538 if (sc->sysctl_tree != NULL) {
539 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
540 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
541 CTLFLAG_RWTUN, sc->led, sizeof(sc->led),
546 g_error_provider(pp, 0);
550 * We get this callback after all of the consumers have gone away, and just
551 * before the provider is freed. If the disk driver provided a d_gone
552 * callback, let them know that it is okay to free resources -- they won't
553 * be getting any more accesses from GEOM.
556 g_disk_providergone(struct g_provider *pp)
559 struct g_disk_softc *sc;
561 sc = (struct g_disk_softc *)pp->private;
563 if (dp != NULL && dp->d_gone != NULL)
565 if (sc->sysctl_tree != NULL) {
566 sysctl_ctx_free(&sc->sysctl_ctx);
567 sc->sysctl_tree = NULL;
569 if (sc->led[0] != 0) {
570 led_set(sc->led, "0");
574 pp->geom->softc = NULL;
575 mtx_destroy(&sc->done_mtx);
576 mtx_destroy(&sc->start_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 if (dp->d_flags & DISKFLAG_RESERVED) {
645 printf("WARNING: Attempt to add non-MPSAFE disk %s%d\n",
646 dp->d_name, dp->d_unit);
647 printf("WARNING: Ignoring disk %s%d\n",
648 dp->d_name, dp->d_unit);
651 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
652 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
653 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
654 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
655 if (dp->d_devstat == NULL)
656 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
657 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
658 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
660 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
661 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
665 disk_destroy(struct disk *dp)
670 if (dp->d_devstat != NULL)
671 devstat_remove_entry(dp->d_devstat);
672 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
676 disk_gone(struct disk *dp)
679 struct g_provider *pp;
683 pp = LIST_FIRST(&gp->provider);
685 KASSERT(LIST_NEXT(pp, provider) == NULL,
686 ("geom %p has more than one provider", gp));
687 g_wither_provider(pp, ENXIO);
693 disk_attr_changed(struct disk *dp, const char *attr, int flag)
696 struct g_provider *pp;
700 LIST_FOREACH(pp, &gp->provider, provider)
701 (void)g_attr_changed(pp, attr, flag);
705 disk_media_changed(struct disk *dp, int flag)
708 struct g_provider *pp;
712 pp = LIST_FIRST(&gp->provider);
714 KASSERT(LIST_NEXT(pp, provider) == NULL,
715 ("geom %p has more than one provider", gp));
716 g_media_changed(pp, flag);
722 disk_media_gone(struct disk *dp, int flag)
725 struct g_provider *pp;
729 pp = LIST_FIRST(&gp->provider);
731 KASSERT(LIST_NEXT(pp, provider) == NULL,
732 ("geom %p has more than one provider", gp));
733 g_media_gone(pp, flag);
739 disk_resize(struct disk *dp, int flag)
742 if (dp->d_destroyed || dp->d_geom == NULL)
745 return (g_post_event(g_disk_resize, dp, flag, NULL));
749 g_kern_disks(void *p, int flag __unused)
758 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
759 sbuf_printf(sb, "%s%s", sp, gp->name);
766 sysctl_disks(SYSCTL_HANDLER_ARGS)
771 sb = sbuf_new_auto();
772 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
773 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
778 SYSCTL_PROC(_kern, OID_AUTO, disks,
779 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
780 sysctl_disks, "A", "names of available disks");