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_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_kerneldump(%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)
234 struct g_disk_softc *sc;
236 /* See "notes" for why we need a mutex here */
237 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
238 bp2 = bp->bio_parent;
239 sc = bp2->bio_to->private;
240 bp->bio_completed = bp->bio_length - bp->bio_resid;
242 mtx_lock(&sc->done_mtx);
243 if (bp2->bio_error == 0)
244 bp2->bio_error = bp->bio_error;
245 bp2->bio_completed += bp->bio_completed;
246 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0)
247 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
249 if (bp2->bio_children == bp2->bio_inbed) {
250 mtx_unlock(&sc->done_mtx);
251 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
252 g_io_deliver(bp2, bp2->bio_error);
254 mtx_unlock(&sc->done_mtx);
259 g_disk_done_single(struct bio *bp)
262 struct g_disk_softc *sc;
264 bp->bio_completed = bp->bio_length - bp->bio_resid;
265 bp->bio_done = (void *)bp->bio_to;
266 bp->bio_to = LIST_FIRST(&bp->bio_disk->d_geom->provider);
267 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE|BIO_FLUSH)) != 0) {
269 sc = bp->bio_to->private;
270 mtx_lock(&sc->done_mtx);
271 devstat_end_transaction_bio_bt(sc->dp->d_devstat, bp, &now);
272 mtx_unlock(&sc->done_mtx);
274 g_io_deliver(bp, bp->bio_error);
278 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
281 struct g_disk_softc *sc;
287 if (dp->d_ioctl == NULL)
289 g_disk_lock_giant(dp);
290 error = dp->d_ioctl(dp, cmd, data, fflag, td);
291 g_disk_unlock_giant(dp);
296 g_disk_start(struct bio *bp)
298 struct bio *bp2, *bp3;
300 struct g_disk_softc *sc;
302 off_t d_maxsize, off;
304 sc = bp->bio_to->private;
305 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
306 g_io_deliver(bp, ENXIO);
310 switch(bp->bio_cmd) {
312 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
319 d_maxsize = (bp->bio_cmd == BIO_DELETE) ?
320 dp->d_delmaxsize : dp->d_maxsize;
321 if (bp->bio_length <= d_maxsize) {
323 bp->bio_to = (void *)bp->bio_done;
324 bp->bio_done = g_disk_done_single;
325 bp->bio_pblkno = bp->bio_offset / dp->d_sectorsize;
326 bp->bio_bcount = bp->bio_length;
327 mtx_lock(&sc->start_mtx);
328 devstat_start_transaction_bio(dp->d_devstat, bp);
329 mtx_unlock(&sc->start_mtx);
330 g_disk_lock_giant(dp);
332 g_disk_unlock_giant(dp);
337 bp2 = g_clone_bio(bp);
343 bp2->bio_offset += off;
344 bp2->bio_length -= off;
345 if ((bp->bio_flags & BIO_UNMAPPED) == 0) {
346 bp2->bio_data += off;
348 KASSERT((dp->d_flags & DISKFLAG_UNMAPPED_BIO)
350 ("unmapped bio not supported by disk %s",
352 bp2->bio_ma += off / PAGE_SIZE;
353 bp2->bio_ma_offset += off;
354 bp2->bio_ma_offset %= PAGE_SIZE;
355 bp2->bio_ma_n -= off / PAGE_SIZE;
357 if (bp2->bio_length > d_maxsize) {
359 * XXX: If we have a stripesize we should really
360 * use it here. Care should be taken in the delete
361 * case if this is done as deletes can be very
362 * sensitive to size given how they are processed.
364 bp2->bio_length = d_maxsize;
365 if ((bp->bio_flags & BIO_UNMAPPED) != 0) {
366 bp2->bio_ma_n = howmany(
368 bp2->bio_length, PAGE_SIZE);
372 * To avoid a race, we need to grab the next bio
373 * before we schedule this one. See "notes".
375 bp3 = g_clone_bio(bp);
377 bp->bio_error = ENOMEM;
379 bp2->bio_done = g_disk_done;
380 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
381 bp2->bio_bcount = bp2->bio_length;
383 mtx_lock(&sc->start_mtx);
384 devstat_start_transaction_bio(dp->d_devstat, bp2);
385 mtx_unlock(&sc->start_mtx);
386 g_disk_lock_giant(dp);
388 g_disk_unlock_giant(dp);
391 } while (bp2 != NULL);
394 /* Give the driver a chance to override */
395 if (dp->d_getattr != NULL) {
396 if (bp->bio_disk == NULL)
398 error = dp->d_getattr(bp);
403 if (g_handleattr_int(bp, "GEOM::candelete",
404 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
406 else if (g_handleattr_int(bp, "GEOM::fwsectors",
409 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
411 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
413 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
415 else if (g_handleattr(bp, "GEOM::hba_vendor",
416 &dp->d_hba_vendor, 2))
418 else if (g_handleattr(bp, "GEOM::hba_device",
419 &dp->d_hba_device, 2))
421 else if (g_handleattr(bp, "GEOM::hba_subvendor",
422 &dp->d_hba_subvendor, 2))
424 else if (g_handleattr(bp, "GEOM::hba_subdevice",
425 &dp->d_hba_subdevice, 2))
427 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
428 g_disk_kerneldump(bp, dp);
429 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
430 g_disk_setstate(bp, sc);
435 g_trace(G_T_BIO, "g_disk_flushcache(%s)",
437 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
442 bp->bio_to = (void *)bp->bio_done;
443 bp->bio_done = g_disk_done_single;
444 mtx_lock(&sc->start_mtx);
445 devstat_start_transaction_bio(dp->d_devstat, bp);
446 mtx_unlock(&sc->start_mtx);
447 g_disk_lock_giant(dp);
449 g_disk_unlock_giant(dp);
455 if (error != EJUSTRETURN)
456 g_io_deliver(bp, error);
461 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
465 struct g_disk_softc *sc;
470 if (sc == NULL || (dp = sc->dp) == NULL)
472 if (indent == NULL) {
473 sbuf_printf(sb, " hd %u", dp->d_fwheads);
474 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
478 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
479 indent, dp->d_fwheads);
480 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
481 indent, dp->d_fwsectors);
482 if (dp->d_getattr != NULL) {
483 buf = g_malloc(DISK_IDENT_SIZE, M_WAITOK);
486 bp->bio_attribute = "GEOM::ident";
487 bp->bio_length = DISK_IDENT_SIZE;
489 res = dp->d_getattr(bp);
490 sbuf_printf(sb, "%s<ident>", indent);
491 g_conf_printf_escaped(sb, "%s",
492 res == 0 ? buf: dp->d_ident);
493 sbuf_printf(sb, "</ident>\n");
494 bp->bio_attribute = "GEOM::lunid";
495 bp->bio_length = DISK_IDENT_SIZE;
497 if (dp->d_getattr(bp) == 0) {
498 sbuf_printf(sb, "%s<lunid>", indent);
499 g_conf_printf_escaped(sb, "%s", buf);
500 sbuf_printf(sb, "</lunid>\n");
502 bp->bio_attribute = "GEOM::lunname";
503 bp->bio_length = DISK_IDENT_SIZE;
505 if (dp->d_getattr(bp) == 0) {
506 sbuf_printf(sb, "%s<lunname>", indent);
507 g_conf_printf_escaped(sb, "%s", buf);
508 sbuf_printf(sb, "</lunname>\n");
513 sbuf_printf(sb, "%s<ident>", indent);
514 g_conf_printf_escaped(sb, "%s", dp->d_ident);
515 sbuf_printf(sb, "</ident>\n");
517 sbuf_printf(sb, "%s<descr>", indent);
518 g_conf_printf_escaped(sb, "%s", dp->d_descr);
519 sbuf_printf(sb, "</descr>\n");
524 g_disk_resize(void *ptr, int flag)
528 struct g_provider *pp;
530 if (flag == EV_CANCEL)
537 if (dp->d_destroyed || gp == NULL)
540 LIST_FOREACH(pp, &gp->provider, provider) {
541 if (pp->sectorsize != 0 &&
542 pp->sectorsize != dp->d_sectorsize)
543 g_wither_provider(pp, ENXIO);
545 g_resize_provider(pp, dp->d_mediasize);
550 g_disk_create(void *arg, int flag)
553 struct g_provider *pp;
555 struct g_disk_softc *sc;
558 if (flag == EV_CANCEL)
562 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
563 mtx_init(&sc->start_mtx, "g_disk_start", NULL, MTX_DEF);
564 mtx_init(&sc->done_mtx, "g_disk_done", NULL, MTX_DEF);
566 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
568 pp = g_new_providerf(gp, "%s", gp->name);
569 devstat_remove_entry(pp->stat);
571 dp->d_devstat->id = pp;
572 pp->mediasize = dp->d_mediasize;
573 pp->sectorsize = dp->d_sectorsize;
574 pp->stripeoffset = dp->d_stripeoffset;
575 pp->stripesize = dp->d_stripesize;
576 if ((dp->d_flags & DISKFLAG_UNMAPPED_BIO) != 0)
577 pp->flags |= G_PF_ACCEPT_UNMAPPED;
578 if ((dp->d_flags & DISKFLAG_DIRECT_COMPLETION) != 0)
579 pp->flags |= G_PF_DIRECT_SEND;
580 pp->flags |= G_PF_DIRECT_RECEIVE;
582 printf("GEOM: new disk %s\n", gp->name);
583 sysctl_ctx_init(&sc->sysctl_ctx);
584 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
585 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
586 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
587 CTLFLAG_RD, 0, tmpstr);
588 if (sc->sysctl_tree != NULL) {
589 snprintf(tmpstr, sizeof(tmpstr),
590 "kern.geom.disk.%s.led", gp->name);
591 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led));
592 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
593 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
594 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led),
599 g_error_provider(pp, 0);
603 * We get this callback after all of the consumers have gone away, and just
604 * before the provider is freed. If the disk driver provided a d_gone
605 * callback, let them know that it is okay to free resources -- they won't
606 * be getting any more accesses from GEOM.
609 g_disk_providergone(struct g_provider *pp)
612 struct g_disk_softc *sc;
614 sc = (struct g_disk_softc *)pp->private;
616 if (dp != NULL && dp->d_gone != NULL)
618 if (sc->sysctl_tree != NULL) {
619 sysctl_ctx_free(&sc->sysctl_ctx);
620 sc->sysctl_tree = NULL;
622 if (sc->led[0] != 0) {
623 led_set(sc->led, "0");
627 pp->geom->softc = NULL;
628 mtx_destroy(&sc->done_mtx);
629 mtx_destroy(&sc->start_mtx);
634 g_disk_destroy(void *ptr, int flag)
638 struct g_disk_softc *sc;
648 g_wither_geom(gp, ENXIO);
654 * We only allow printable characters in disk ident,
655 * the rest is converted to 'x<HH>'.
658 g_disk_ident_adjust(char *ident, size_t size)
660 char *p, tmp[4], newid[DISK_IDENT_SIZE];
663 for (p = ident; *p != '\0'; p++) {
668 snprintf(tmp, sizeof(tmp), "x%02hhx",
669 *(unsigned char *)p);
671 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
675 strlcpy(ident, newid, size);
682 return (g_malloc(sizeof(struct disk), M_WAITOK | M_ZERO));
686 disk_create(struct disk *dp, int version)
689 if (version != DISK_VERSION) {
690 printf("WARNING: Attempt to add disk %s%d %s",
691 dp->d_name, dp->d_unit,
692 " using incompatible ABI version of disk(9)\n");
693 printf("WARNING: Ignoring disk %s%d\n",
694 dp->d_name, dp->d_unit);
697 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
698 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
699 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
700 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
701 if (dp->d_devstat == NULL)
702 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
703 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
704 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
706 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
707 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
711 disk_destroy(struct disk *dp)
716 if (dp->d_devstat != NULL)
717 devstat_remove_entry(dp->d_devstat);
718 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
722 disk_gone(struct disk *dp)
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_wither_provider(pp, ENXIO);
739 disk_attr_changed(struct disk *dp, const char *attr, int flag)
742 struct g_provider *pp;
746 LIST_FOREACH(pp, &gp->provider, provider)
747 (void)g_attr_changed(pp, attr, flag);
751 disk_media_changed(struct disk *dp, int flag)
754 struct g_provider *pp;
758 pp = LIST_FIRST(&gp->provider);
760 KASSERT(LIST_NEXT(pp, provider) == NULL,
761 ("geom %p has more than one provider", gp));
762 g_media_changed(pp, flag);
768 disk_media_gone(struct disk *dp, int flag)
771 struct g_provider *pp;
775 pp = LIST_FIRST(&gp->provider);
777 KASSERT(LIST_NEXT(pp, provider) == NULL,
778 ("geom %p has more than one provider", gp));
779 g_media_gone(pp, flag);
785 disk_resize(struct disk *dp, int flag)
788 if (dp->d_destroyed || dp->d_geom == NULL)
791 return (g_post_event(g_disk_resize, dp, flag, NULL));
795 g_kern_disks(void *p, int flag __unused)
804 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
805 sbuf_printf(sb, "%s%s", sp, gp->name);
812 sysctl_disks(SYSCTL_HANDLER_ARGS)
817 sb = sbuf_new_auto();
818 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
819 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
824 SYSCTL_PROC(_kern, OID_AUTO, disks,
825 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
826 sysctl_disks, "A", "names of available disks");