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>
64 struct sysctl_ctx_list sysctl_ctx;
65 struct sysctl_oid *sysctl_tree;
70 static struct mtx g_disk_done_mtx;
72 static g_access_t g_disk_access;
73 static g_init_t g_disk_init;
74 static g_fini_t g_disk_fini;
75 static g_start_t g_disk_start;
76 static g_ioctl_t g_disk_ioctl;
77 static g_dumpconf_t g_disk_dumpconf;
78 static g_provgone_t g_disk_providergone;
80 static struct g_class g_disk_class = {
85 .start = g_disk_start,
86 .access = g_disk_access,
87 .ioctl = g_disk_ioctl,
88 .providergone = g_disk_providergone,
89 .dumpconf = g_disk_dumpconf,
92 SYSCTL_DECL(_kern_geom);
93 SYSCTL_NODE(_kern_geom, OID_AUTO, disk, CTLFLAG_RW, 0, "GEOM_DISK stuff");
96 g_disk_init(struct g_class *mp __unused)
99 mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF);
103 g_disk_fini(struct g_class *mp __unused)
106 mtx_destroy(&g_disk_done_mtx);
109 DECLARE_GEOM_CLASS(g_disk_class, g_disk);
112 g_disk_lock_giant(struct disk *dp)
114 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
119 g_disk_unlock_giant(struct disk *dp)
121 if (dp->d_flags & DISKFLAG_NEEDSGIANT)
126 g_disk_access(struct g_provider *pp, int r, int w, int e)
129 struct g_disk_softc *sc;
132 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)",
135 sc = pp->geom->softc;
136 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
138 * Allow decreasing access count even if disk is not
141 if (r <= 0 && w <= 0 && e <= 0)
149 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) {
150 if (dp->d_open != NULL) {
151 g_disk_lock_giant(dp);
152 error = dp->d_open(dp);
153 if (bootverbose && error != 0)
154 printf("Opened disk %s -> %d\n",
156 g_disk_unlock_giant(dp);
158 pp->mediasize = dp->d_mediasize;
159 pp->sectorsize = dp->d_sectorsize;
160 if (dp->d_flags & DISKFLAG_CANDELETE)
161 pp->flags |= G_PF_CANDELETE;
163 pp->flags &= ~G_PF_CANDELETE;
164 pp->stripeoffset = dp->d_stripeoffset;
165 pp->stripesize = dp->d_stripesize;
166 dp->d_flags |= DISKFLAG_OPEN;
167 if (dp->d_maxsize == 0) {
168 printf("WARNING: Disk drive %s%d has no d_maxsize\n",
169 dp->d_name, dp->d_unit);
170 dp->d_maxsize = DFLTPHYS;
172 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) {
173 if (dp->d_close != NULL) {
174 g_disk_lock_giant(dp);
175 error = dp->d_close(dp);
177 printf("Closed disk %s -> %d\n",
179 g_disk_unlock_giant(dp);
181 sc->state = G_STATE_ACTIVE;
183 led_set(sc->led, "0");
184 dp->d_flags &= ~DISKFLAG_OPEN;
190 g_disk_kerneldump(struct bio *bp, struct disk *dp)
192 struct g_kerneldump *gkd;
195 gkd = (struct g_kerneldump*)bp->bio_data;
196 gp = bp->bio_to->geom;
197 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)",
198 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length);
199 if (dp->d_dump == NULL) {
200 g_io_deliver(bp, ENODEV);
203 gkd->di.dumper = dp->d_dump;
205 gkd->di.blocksize = dp->d_sectorsize;
206 gkd->di.maxiosize = dp->d_maxsize;
207 gkd->di.mediaoffset = gkd->offset;
208 if ((gkd->offset + gkd->length) > dp->d_mediasize)
209 gkd->length = dp->d_mediasize - gkd->offset;
210 gkd->di.mediasize = gkd->length;
215 g_disk_setstate(struct bio *bp, struct g_disk_softc *sc)
219 memcpy(&sc->state, bp->bio_data, sizeof(sc->state));
220 if (sc->led[0] != 0) {
225 case G_STATE_REBUILD:
235 led_set(sc->led, cmd);
241 g_disk_done(struct bio *bp)
245 struct g_disk_softc *sc;
247 /* See "notes" for why we need a mutex here */
248 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */
249 mtx_lock(&g_disk_done_mtx);
250 bp->bio_completed = bp->bio_length - bp->bio_resid;
252 bp2 = bp->bio_parent;
253 if (bp2->bio_error == 0)
254 bp2->bio_error = bp->bio_error;
255 bp2->bio_completed += bp->bio_completed;
256 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) &&
257 (sc = bp2->bio_to->geom->softc) &&
259 devstat_end_transaction_bio(dp->d_devstat, bp);
263 if (bp2->bio_children == bp2->bio_inbed) {
264 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed;
265 g_io_deliver(bp2, bp2->bio_error);
267 mtx_unlock(&g_disk_done_mtx);
271 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td)
275 struct g_disk_softc *sc;
282 if (dp->d_ioctl == NULL)
284 g_disk_lock_giant(dp);
285 error = dp->d_ioctl(dp, cmd, data, fflag, td);
286 g_disk_unlock_giant(dp);
291 g_disk_start(struct bio *bp)
293 struct bio *bp2, *bp3;
295 struct g_disk_softc *sc;
299 sc = bp->bio_to->geom->softc;
300 if (sc == NULL || (dp = sc->dp) == NULL || dp->d_destroyed) {
301 g_io_deliver(bp, ENXIO);
305 switch(bp->bio_cmd) {
307 if (!(dp->d_flags & DISKFLAG_CANDELETE)) {
316 bp2 = g_clone_bio(bp);
322 bp2->bio_offset += off;
323 bp2->bio_length -= off;
324 bp2->bio_data += off;
325 if (bp2->bio_length > dp->d_maxsize) {
327 * XXX: If we have a stripesize we should really
330 bp2->bio_length = dp->d_maxsize;
331 off += dp->d_maxsize;
333 * To avoid a race, we need to grab the next bio
334 * before we schedule this one. See "notes".
336 bp3 = g_clone_bio(bp);
338 bp->bio_error = ENOMEM;
340 bp2->bio_done = g_disk_done;
341 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize;
342 bp2->bio_bcount = bp2->bio_length;
344 devstat_start_transaction_bio(dp->d_devstat, bp2);
345 g_disk_lock_giant(dp);
347 g_disk_unlock_giant(dp);
350 } while (bp2 != NULL);
353 /* Give the driver a chance to override */
354 if (dp->d_getattr != NULL) {
355 if (bp->bio_disk == NULL)
357 error = dp->d_getattr(bp);
362 if (g_handleattr_int(bp, "GEOM::candelete",
363 (dp->d_flags & DISKFLAG_CANDELETE) != 0))
365 else if (g_handleattr_int(bp, "GEOM::fwsectors",
368 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads))
370 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0))
372 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident))
374 else if (g_handleattr(bp, "GEOM::hba_vendor",
375 &dp->d_hba_vendor, 2))
377 else if (g_handleattr(bp, "GEOM::hba_device",
378 &dp->d_hba_device, 2))
380 else if (g_handleattr(bp, "GEOM::hba_subvendor",
381 &dp->d_hba_subvendor, 2))
383 else if (g_handleattr(bp, "GEOM::hba_subdevice",
384 &dp->d_hba_subdevice, 2))
386 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump"))
387 g_disk_kerneldump(bp, dp);
388 else if (!strcmp(bp->bio_attribute, "GEOM::setstate"))
389 g_disk_setstate(bp, sc);
394 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)",
396 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) {
397 g_io_deliver(bp, ENODEV);
400 bp2 = g_clone_bio(bp);
402 g_io_deliver(bp, ENOMEM);
405 bp2->bio_done = g_disk_done;
407 g_disk_lock_giant(dp);
409 g_disk_unlock_giant(dp);
415 if (error != EJUSTRETURN)
416 g_io_deliver(bp, error);
421 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp)
424 struct g_disk_softc *sc;
427 if (sc == NULL || (dp = sc->dp) == NULL)
429 if (indent == NULL) {
430 sbuf_printf(sb, " hd %u", dp->d_fwheads);
431 sbuf_printf(sb, " sc %u", dp->d_fwsectors);
435 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n",
436 indent, dp->d_fwheads);
437 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n",
438 indent, dp->d_fwsectors);
439 sbuf_printf(sb, "%s<ident>%s</ident>\n", indent, dp->d_ident);
440 sbuf_printf(sb, "%s<descr>%s</descr>\n", indent, dp->d_descr);
445 g_disk_create(void *arg, int flag)
448 struct g_provider *pp;
450 struct g_disk_softc *sc;
453 if (flag == EV_CANCEL)
457 sc = g_malloc(sizeof(*sc), M_WAITOK | M_ZERO);
459 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit);
461 pp = g_new_providerf(gp, "%s", gp->name);
462 pp->mediasize = dp->d_mediasize;
463 pp->sectorsize = dp->d_sectorsize;
464 if (dp->d_flags & DISKFLAG_CANDELETE)
465 pp->flags |= G_PF_CANDELETE;
466 pp->stripeoffset = dp->d_stripeoffset;
467 pp->stripesize = dp->d_stripesize;
469 printf("GEOM: new disk %s\n", gp->name);
470 sysctl_ctx_init(&sc->sysctl_ctx);
471 snprintf(tmpstr, sizeof(tmpstr), "GEOM disk %s", gp->name);
472 sc->sysctl_tree = SYSCTL_ADD_NODE(&sc->sysctl_ctx,
473 SYSCTL_STATIC_CHILDREN(_kern_geom_disk), OID_AUTO, gp->name,
474 CTLFLAG_RD, 0, tmpstr);
475 if (sc->sysctl_tree != NULL) {
476 snprintf(tmpstr, sizeof(tmpstr),
477 "kern.geom.disk.%s.led", gp->name);
478 TUNABLE_STR_FETCH(tmpstr, sc->led, sizeof(sc->led));
479 SYSCTL_ADD_STRING(&sc->sysctl_ctx,
480 SYSCTL_CHILDREN(sc->sysctl_tree), OID_AUTO, "led",
481 CTLFLAG_RW | CTLFLAG_TUN, sc->led, sizeof(sc->led),
486 g_error_provider(pp, 0);
490 * We get this callback after all of the consumers have gone away, and just
491 * before the provider is freed. If the disk driver provided a d_gone
492 * callback, let them know that it is okay to free resources -- they won't
493 * be getting any more accesses from GEOM.
496 g_disk_providergone(struct g_provider *pp)
499 struct g_disk_softc *sc;
501 sc = (struct g_disk_softc *)pp->geom->softc;
504 * If the softc is already NULL, then we've probably been through
505 * g_disk_destroy already; there is nothing for us to do anyway.
513 * FreeBSD 9 started with VERSION_01 of the struct disk structure.
514 * However, g_gone was added in the middle of the branch. To
515 * cope with version being missing from struct disk, we set a flag
516 * in g_disk_create for VERSION_01 and avoid touching the d_gone
517 * field for old consumers.
519 if (!(dp->d_flags & DISKFLAG_LACKS_GONE) && dp->d_gone != NULL)
524 g_disk_destroy(void *ptr, int flag)
528 struct g_disk_softc *sc;
535 if (sc->sysctl_tree != NULL) {
536 sysctl_ctx_free(&sc->sysctl_ctx);
537 sc->sysctl_tree = NULL;
539 if (sc->led[0] != 0) {
540 led_set(sc->led, "0");
545 g_wither_geom(gp, ENXIO);
551 * We only allow printable characters in disk ident,
552 * the rest is converted to 'x<HH>'.
555 g_disk_ident_adjust(char *ident, size_t size)
557 char *p, tmp[4], newid[DISK_IDENT_SIZE];
560 for (p = ident; *p != '\0'; p++) {
565 snprintf(tmp, sizeof(tmp), "x%02hhx",
566 *(unsigned char *)p);
568 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid))
572 strlcpy(ident, newid, size);
580 dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO);
585 disk_create(struct disk *dp, int version)
587 if (version != DISK_VERSION_02 && version != DISK_VERSION_01) {
588 printf("WARNING: Attempt to add disk %s%d %s",
589 dp->d_name, dp->d_unit,
590 " using incompatible ABI version of disk(9)\n");
591 printf("WARNING: Ignoring disk %s%d\n",
592 dp->d_name, dp->d_unit);
595 if (version == DISK_VERSION_01)
596 dp->d_flags |= DISKFLAG_LACKS_GONE;
597 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy"));
598 KASSERT(dp->d_name != NULL, ("disk_create need d_name"));
599 KASSERT(*dp->d_name != 0, ("disk_create need d_name"));
600 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long"));
601 if (dp->d_devstat == NULL)
602 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit,
603 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED,
604 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX);
606 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident));
607 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL);
611 disk_destroy(struct disk *dp)
616 if (dp->d_devstat != NULL)
617 devstat_remove_entry(dp->d_devstat);
618 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL);
622 disk_gone(struct disk *dp)
625 struct g_provider *pp;
629 LIST_FOREACH(pp, &gp->provider, provider)
630 g_wither_provider(pp, ENXIO);
634 disk_attr_changed(struct disk *dp, const char *attr, int flag)
637 struct g_provider *pp;
641 LIST_FOREACH(pp, &gp->provider, provider)
642 (void)g_attr_changed(pp, attr, flag);
646 g_kern_disks(void *p, int flag __unused)
655 LIST_FOREACH(gp, &g_disk_class.geom, geom) {
656 sbuf_printf(sb, "%s%s", sp, gp->name);
663 sysctl_disks(SYSCTL_HANDLER_ARGS)
668 sb = sbuf_new_auto();
669 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL);
670 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1);
675 SYSCTL_PROC(_kern, OID_AUTO, disks,
676 CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
677 sysctl_disks, "A", "names of available disks");