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("$FreeBSD$");
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/malloc.h>
42 #include <sys/kernel.h>
46 #include <sys/mutex.h>
48 #include <sys/errno.h>
51 #include <sys/fcntl.h>
52 #include <sys/limits.h>
53 #include <geom/geom.h>
54 #include <geom/geom_int.h>
56 static d_open_t g_dev_open;
57 static d_close_t g_dev_close;
58 static d_strategy_t g_dev_strategy;
59 static d_ioctl_t g_dev_ioctl;
61 static struct cdevsw g_dev_cdevsw = {
62 .d_version = D_VERSION,
64 .d_close = g_dev_close,
67 .d_ioctl = g_dev_ioctl,
68 .d_strategy = g_dev_strategy,
70 .d_flags = D_DISK | D_TRACKCLOSE,
73 static g_taste_t g_dev_taste;
74 static g_orphan_t g_dev_orphan;
75 static g_init_t g_dev_init;
77 static struct g_class g_dev_class = {
81 .orphan = g_dev_orphan,
85 static struct unrhdr *unithdr; /* Locked by topology */
88 g_dev_init(struct g_class *mp)
91 unithdr = new_unrhdr(0, minor2unit(MAXMINOR), NULL);
100 LIST_FOREACH(gp, &g_dev_class.geom, geom) {
101 printf("%s%s", p, gp->name);
108 g_dev_getprovider(struct cdev *dev)
110 struct g_consumer *cp;
115 if (dev->si_devsw != &g_dev_cdevsw)
118 return (cp->provider);
122 static struct g_geom *
123 g_dev_taste(struct g_class *mp, struct g_provider *pp, int insist __unused)
126 struct g_consumer *cp;
132 g_trace(G_T_TOPOLOGY, "dev_taste(%s,%s)", mp->name, pp->name);
134 LIST_FOREACH(cp, &pp->consumers, consumers)
135 if (cp->geom->class == mp)
137 gp = g_new_geomf(mp, pp->name);
138 cp = g_new_consumer(gp);
139 error = g_attach(cp, pp);
141 ("g_dev_taste(%s) failed to g_attach, err=%d", pp->name, error));
142 unit = alloc_unr(unithdr);
143 dev = make_dev(&g_dev_cdevsw, unit2minor(unit),
144 UID_ROOT, GID_OPERATOR, 0640, gp->name);
145 if (pp->flags & G_PF_CANDELETE)
146 dev->si_flags |= SI_CANDELETE;
147 dev->si_iosize_max = MAXPHYS;
154 alias = g_malloc(MAXPATHLEN, M_WAITOK | M_ZERO);
155 error = (pp->geom->ioctl == NULL) ? ENODEV :
156 pp->geom->ioctl(pp, DIOCGPROVIDERALIAS, alias, 0, curthread);
157 if (!error && alias[0] != '\0')
158 make_dev_alias(dev, "%s", alias);
166 g_dev_open(struct cdev *dev, int flags, int fmt, struct thread *td)
169 struct g_consumer *cp;
174 if (gp == NULL || cp == NULL || gp->softc != dev)
175 return(ENXIO); /* g_dev_taste() not done yet */
177 g_trace(G_T_ACCESS, "g_dev_open(%s, %d, %d, %p)",
178 gp->name, flags, fmt, td);
180 r = flags & FREAD ? 1 : 0;
181 w = flags & FWRITE ? 1 : 0;
183 e = flags & O_EXCL ? 1 : 0;
189 * When running in very secure mode, do not allow
190 * opens for writing of any disks.
192 error = securelevel_ge(td->td_ucred, 2);
197 if (dev->si_devsw == NULL)
198 error = ENXIO; /* We were orphaned */
200 error = g_access(cp, r, w, e);
206 g_dev_close(struct cdev *dev, int flags, int fmt, struct thread *td)
209 struct g_consumer *cp;
210 int error, r, w, e, i;
214 if (gp == NULL || cp == NULL)
216 g_trace(G_T_ACCESS, "g_dev_close(%s, %d, %d, %p)",
217 gp->name, flags, fmt, td);
218 r = flags & FREAD ? -1 : 0;
219 w = flags & FWRITE ? -1 : 0;
221 e = flags & O_EXCL ? -1 : 0;
226 if (dev->si_devsw == NULL)
227 error = ENXIO; /* We were orphaned */
229 error = g_access(cp, r, w, e);
230 for (i = 0; i < 10 * hz;) {
231 if (cp->acr != 0 || cp->acw != 0)
233 if (cp->nstart == cp->nend)
235 pause("gdevwclose", hz / 10);
238 if (cp->acr == 0 && cp->acw == 0 && cp->nstart != cp->nend) {
239 printf("WARNING: Final close of geom_dev(%s) %s %s\n",
241 "still has outstanding I/O after 10 seconds.",
242 "Completing close anyway, panic may happen later.");
249 * XXX: Until we have unmessed the ioctl situation, there is a race against
250 * XXX: a concurrent orphanization. We cannot close it by holding topology
251 * XXX: since that would prevent us from doing our job, and stalling events
252 * XXX: will break (actually: stall) the BSD disklabel hacks.
255 g_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag, struct thread *td)
258 struct g_consumer *cp;
259 struct g_kerneldump kd;
260 off_t offset, length;
268 KASSERT(cp->acr || cp->acw,
269 ("Consumer with zero access count in g_dev_ioctl"));
271 i = IOCPARM_LEN(cmd);
273 case DIOCGSECTORSIZE:
274 *(u_int *)data = cp->provider->sectorsize;
275 if (*(u_int *)data == 0)
279 *(off_t *)data = cp->provider->mediasize;
280 if (*(off_t *)data == 0)
284 error = g_io_getattr("GEOM::fwsectors", cp, &i, data);
285 if (error == 0 && *(u_int *)data == 0)
289 error = g_io_getattr("GEOM::fwheads", cp, &i, data);
290 if (error == 0 && *(u_int *)data == 0)
293 case DIOCGFRONTSTUFF:
294 error = g_io_getattr("GEOM::frontstuff", cp, &i, data);
296 case DIOCSKERNELDUMP:
297 u = *((u_int *)data);
306 error = g_io_getattr("GEOM::kerneldump", cp, &i, &kd);
308 dev->si_flags |= SI_DUMPDEV;
311 error = g_io_flush(cp);
314 offset = ((off_t *)data)[0];
315 length = ((off_t *)data)[1];
316 if ((offset % cp->provider->sectorsize) != 0 ||
317 (length % cp->provider->sectorsize) != 0 ||
318 length <= 0 || length > MAXPHYS) {
319 printf("%s: offset=%jd length=%jd\n", __func__, offset,
324 error = g_delete_data(cp, offset, length);
327 error = g_io_getattr("GEOM::ident", cp, &i, data);
331 if (cp->provider->geom->ioctl != NULL) {
332 error = cp->provider->geom->ioctl(cp->provider, cmd, data, fflag, td);
342 g_dev_done(struct bio *bp2)
346 bp = bp2->bio_parent;
347 bp->bio_error = bp2->bio_error;
348 if (bp->bio_error != 0) {
349 g_trace(G_T_BIO, "g_dev_done(%p) had error %d",
351 bp->bio_flags |= BIO_ERROR;
353 g_trace(G_T_BIO, "g_dev_done(%p/%p) resid %ld completed %jd",
354 bp2, bp, bp->bio_resid, (intmax_t)bp2->bio_completed);
356 bp->bio_resid = bp->bio_length - bp2->bio_completed;
357 bp->bio_completed = bp2->bio_completed;
363 g_dev_strategy(struct bio *bp)
365 struct g_consumer *cp;
369 KASSERT(bp->bio_cmd == BIO_READ ||
370 bp->bio_cmd == BIO_WRITE ||
371 bp->bio_cmd == BIO_DELETE,
372 ("Wrong bio_cmd bio=%p cmd=%d", bp, bp->bio_cmd));
375 KASSERT(cp->acr || cp->acw,
376 ("Consumer with zero access count in g_dev_strategy"));
378 if ((bp->bio_offset % cp->provider->sectorsize) != 0 ||
379 (bp->bio_bcount % cp->provider->sectorsize) != 0) {
380 bp->bio_resid = bp->bio_bcount;
381 biofinish(bp, NULL, EINVAL);
387 * XXX: This is not an ideal solution, but I belive it to
388 * XXX: deadlock safe, all things considered.
390 bp2 = g_clone_bio(bp);
393 pause("gdstrat", hz / 10);
395 KASSERT(bp2 != NULL, ("XXX: ENOMEM in a bad place"));
396 bp2->bio_done = g_dev_done;
398 "g_dev_strategy(%p/%p) offset %jd length %jd data %p cmd %d",
399 bp, bp2, (intmax_t)bp->bio_offset, (intmax_t)bp2->bio_length,
400 bp2->bio_data, bp2->bio_cmd);
401 g_io_request(bp2, cp);
402 KASSERT(cp->acr || cp->acw,
403 ("g_dev_strategy raced with g_dev_close and lost"));
410 * Called from below when the provider orphaned us.
411 * - Clear any dump settings.
412 * - Destroy the struct cdev *to prevent any more request from coming in. The
413 * provider is already marked with an error, so anything which comes in
414 * in the interrim will be returned immediately.
415 * - Wait for any outstanding I/O to finish.
416 * - Set our access counts to zero, whatever they were.
417 * - Detach and self-destruct.
421 g_dev_orphan(struct g_consumer *cp)
430 g_trace(G_T_TOPOLOGY, "g_dev_orphan(%p(%s))", cp, gp->name);
432 /* Reset any dump-area set on this device */
433 if (dev->si_flags & SI_DUMPDEV)
436 /* Destroy the struct cdev *so we get no more requests */
437 unit = dev2unit(dev);
439 free_unr(unithdr, unit);
441 /* Wait for the cows to come home */
442 while (cp->nstart != cp->nend)
443 pause("gdevorphan", hz / 10);
445 if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0)
446 g_access(cp, -cp->acr, -cp->acw, -cp->ace);
449 g_destroy_consumer(cp);
453 DECLARE_GEOM_CLASS(g_dev_class, g_dev);