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>
49 #include <sys/sysctl.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>
63 struct sysctl_ctx_list sysctl_ctx;
64 struct sysctl_oid *sysctl_tree;
69 static struct mtx g_disk_done_mtx;
71 static g_access_t g_disk_access;
72 static g_init_t g_disk_init;
73 static g_fini_t g_disk_fini;
74 static g_start_t g_disk_start;
75 static g_ioctl_t g_disk_ioctl;
76 static g_dumpconf_t g_disk_dumpconf;
78 static struct g_class g_disk_class = {
83 .start = g_disk_start,
84 .access = g_disk_access,
85 .ioctl = g_disk_ioctl,
86 .dumpconf = g_disk_dumpconf,
89 SYSCTL_DECL(_kern_geom);
90 SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0, "GEOM_DISK stuff");
93 g_disk_init(struct g_class *mp __unused)
96 mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF);
100 g_disk_fini(struct g_class *mp __unused)
103 mtx_destroy(&g_disk_done_mtx);
106 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
109 g_disk_lock_giant(struct disk *dp)
111 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
116 g_disk_unlock_giant(struct disk *dp)
118 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
123 g_disk_access(struct g_provider *pp, int r, int w, int e)
126 struct g_disk_softc *sc;
129 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
132 sc = pp->geom->softc;
133 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
135 * Allow decreasing access count even if disk is not
138 if (r <= 0 && w <= 0 && e <= 0)
146 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
147 if (dp->d_open != NULL) {
148 g_disk_lock_giant(dp);
149 error = dp->d_open(dp);
150 if (bootverbose && error != 0)
151 printf("Opened disk %s -> %d\n",
153 g_disk_unlock_giant(dp);
155 pp->mediasize = dp->d_mediasize;
156 pp->sectorsize = dp->d_sectorsize;
157 if (dp->d_flags & DISKFLAG_CANDELETE)
158 pp->flags |= G_PF_CANDELETE;
160 pp->flags &= ~G_PF_CANDELETE;
161 pp->stripeoffset = dp->d_stripeoffset;
162 pp->stripesize = dp->d_stripesize;
163 dp->d_flags |= DISKFLAG_OPEN;
164 if (dp->d_maxsize == 0) {
165 printf("WARNING: Disk drive %s%d has no d_maxsize\n",
166 dp->d_name, dp->d_unit);
167 dp->d_maxsize = DFLTPHYS;
169 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
170 if (dp->d_close != NULL) {
171 g_disk_lock_giant(dp);
172 error = dp->d_close(dp);
174 printf("Closed disk %s -> %d\n",
176 g_disk_unlock_giant(dp);
178 sc->state = G_STATE_ACTIVE;
180 led_set(sc->led, "0");
181 dp->d_flags &= ~DISKFLAG_OPEN;
187 g_disk_kerneldump(struct bio *bp, struct disk *dp)
189 struct g_kerneldump *gkd;
192 gkd = (struct g_kerneldump*)bp->bio_data;
193 gp = bp->bio_to->geom;
194 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
195 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
196 if (dp->d_dump == NULL) {
197 g_io_deliver(bp, ENODEV);
200 gkd->di.dumper = dp->d_dump;
202 gkd->di.blocksize = dp->d_sectorsize;
203 gkd->di.maxiosize = dp->d_maxsize;
204 gkd->di.mediaoffset = gkd->offset;
205 if ((gkd->offset + gkd->length) > dp->d_mediasize)
206 gkd->length = dp->d_mediasize - gkd->offset;
207 gkd->di.mediasize = gkd->length;
212 g_disk_setstate(struct bio *bp, struct g_disk_softc *sc)
216 memcpy(&sc->state, bp->bio_data, sizeof(sc->state));
217 if (sc->led[0] != 0) {
222 case G_STATE_REBUILD:
232 led_set(sc->led, cmd);
238 g_disk_done(struct bio *bp)
242 struct g_disk_softc *sc;
244 /* See "notes" for why we need a mutex here */
245 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
246 mtx_lock(&g_disk_done_mtx);
247 bp->bio_completed = bp->bio_length - bp->bio_resid;
249 bp2 = bp->bio_parent;
250 if (bp2->bio_error == 0)
251 bp2->bio_error = bp->bio_error;
252 bp2->bio_completed += bp->bio_completed;
253 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) &&
254 (sc = bp2->bio_to->geom->softc) &&
256 devstat_end_transaction_bio(dp->d_devstat, bp);
260 if (bp2->bio_children == bp2->bio_inbed) {
261 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
262 g_io_deliver(bp2, bp2->bio_error);
264 mtx_unlock(&g_disk_done_mtx);
268 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
272 struct g_disk_softc *sc;
279 if (dp->d_ioctl == NULL)
281 g_disk_lock_giant(dp);
282 error = dp->d_ioctl(dp, cmd, data, fflag, td);
283 g_disk_unlock_giant(dp);
288 g_disk_start(struct bio *bp)
290 struct bio *bp2, *bp3;
292 struct g_disk_softc *sc;
296 sc = bp->bio_to->geom->softc;
297 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
298 g_io_deliver(bp, ENXIO);
302 switch(bp->bio_cmd) {
304 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
313 bp2 = g_clone_bio(bp);
319 bp2->bio_offset += off;
320 bp2->bio_length -= off;
321 bp2->bio_data += off;
322 if (bp2->bio_length > dp->d_maxsize) {
324 * XXX: If we have a stripesize we should really
327 bp2->bio_length = dp->d_maxsize;
328 off += dp->d_maxsize;
330 * To avoid a race, we need to grab the next bio
331 * before we schedule this one. See "notes".
333 bp3 = g_clone_bio(bp);
335 bp->bio_error = ENOMEM;
337 bp2->bio_done = g_disk_done;
338 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
339 bp2->bio_bcount = bp2->bio_length;
341 devstat_start_transaction_bio(dp->d_devstat, bp2);
342 g_disk_lock_giant(dp);
344 g_disk_unlock_giant(dp);
347 } while (bp2 != NULL);
350 if (g_handleattr_int(bp, "GEOM::candelete",
351 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
353 else if (g_handleattr_int(bp, "GEOM::fwsectors",
356 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
358 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
360 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
362 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
363 g_disk_kerneldump(bp, dp);
364 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
365 g_disk_setstate(bp, sc);
370 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)",
372 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
373 g_io_deliver(bp, ENODEV);
376 bp2 = g_clone_bio(bp);
378 g_io_deliver(bp, ENOMEM);
381 bp2->bio_done = g_disk_done;
383 g_disk_lock_giant(dp);
385 g_disk_unlock_giant(dp);
391 if (error != EJUSTRETURN)
392 g_io_deliver(bp, error);
397 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
400 struct g_disk_softc *sc;
403 if (sc == NULL || (dp = sc->dp) == NULL)
405 if (indent == NULL) {
406 sbuf_printf(sb, " hd %u", dp->d_fwheads);
407 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
411 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
412 indent, dp->d_fwheads);
413 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
414 indent, dp->d_fwsectors);
419 g_disk_create(void *arg, int flag)
422 struct g_provider *pp;
424 struct g_disk_softc *sc;
427 if (flag == EV_CANCEL)
431 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
433 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
435 pp = g_new_providerf(gp, "%s", gp->name);
436 pp->mediasize = dp->d_mediasize;
437 pp->sectorsize = dp->d_sectorsize;
438 if (dp->d_flags & DISKFLAG_CANDELETE)
439 pp->flags |= G_PF_CANDELETE;
440 pp->stripeoffset = dp->d_stripeoffset;
441 pp->stripesize = dp->d_stripesize;
443 printf("GEOM: new disk %s\n", gp->name);
444 sysctl_ctx_init(&sc->sysctl_ctx);
445 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
446 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
447 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
448 CTLFLAG_RD, 0, tmpstr);
449 if (sc->sysctl_tree != NULL) {
450 snprintf(tmpstr, sizeof(tmpstr),
451 "kern.geom.disk.%s.led", gp->name);
452 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led));
453 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
454 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
455 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led),
460 g_error_provider(pp, 0);
464 g_disk_destroy(void *ptr, int flag)
468 struct g_disk_softc *sc;
475 if (sc->sysctl_tree != NULL) {
476 sysctl_ctx_free(&sc->sysctl_ctx);
477 sc->sysctl_tree = NULL;
479 if (sc->led[0] != 0) {
480 led_set(sc->led, "0");
485 g_wither_geom(gp, ENXIO);
491 * We only allow printable characters in disk ident,
492 * the rest is converted to 'x<HH>'.
495 g_disk_ident_adjust(char *ident, size_t size)
497 char *p, tmp[4], newid[DISK_IDENT_SIZE];
500 for (p = ident; *p != '\0'; p++) {
505 snprintf(tmp, sizeof(tmp), "x%02hhx",
506 *(unsigned char *)p);
508 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
512 strlcpy(ident, newid, size);
520 dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO);
525 disk_create(struct disk *dp, int version)
527 if (version != DISK_VERSION_00 && version != DISK_VERSION_01) {
528 printf("WARNING: Attempt to add disk %s%d %s",
529 dp->d_name, dp->d_unit,
530 " using incompatible ABI version of disk(9)\n");
531 printf("WARNING: Ignoring disk %s%d\n",
532 dp->d_name, dp->d_unit);
535 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
536 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
537 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
538 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
539 if (dp->d_devstat == NULL)
540 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
541 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
542 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
544 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
545 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
549 disk_destroy(struct disk *dp)
554 if (dp->d_devstat != NULL)
555 devstat_remove_entry(dp->d_devstat);
556 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
560 disk_gone(struct disk *dp)
563 struct g_provider *pp;
567 LIST_FOREACH(pp, &gp->provider, provider)
568 g_wither_provider(pp, ENXIO);
572 g_kern_disks(void *p, int flag __unused)
581 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
582 sbuf_printf(sb, "%s%s", sp, gp->name);
589 sysctl_disks(SYSCTL_HANDLER_ARGS)
594 sb = sbuf_new_auto();
595 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
596 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
601 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0,
602 sysctl_disks, "A", "names of available disks");