1 /*- 2 * Copyright (c) 2002 Poul-Henning Kamp 3 * Copyright (c) 2002 Networks Associates Technology, Inc. 4 * All rights reserved. 5 * 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. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 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 21 * permission. 22 * 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 33 * SUCH DAMAGE. 34 * 35 * $FreeBSD$ 36 */ 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/malloc.h> 41 #include <sys/kernel.h> 42 #include <sys/conf.h> 43 #include <sys/bio.h> 44 #include <sys/lock.h> 45 #include <sys/mutex.h> 46 #include <sys/errno.h> 47 #include <sys/time.h> 48 #include <sys/disk.h> 49 #include <sys/fcntl.h> 50 #include <geom/geom.h> 51 #include <geom/geom_int.h> 52 #include <machine/limits.h> 53 54 static d_open_t g_dev_open; 55 static d_close_t g_dev_close; 56 static d_strategy_t g_dev_strategy; 57 static d_ioctl_t g_dev_ioctl; 58 59 static struct cdevsw g_dev_cdevsw = { 60 .d_open = g_dev_open, 61 .d_close = g_dev_close, 62 .d_read = physread, 63 .d_write = physwrite, 64 .d_ioctl = g_dev_ioctl, 65 .d_strategy = g_dev_strategy, 66 .d_name = "g_dev", 67 .d_maj = GEOM_MAJOR, 68 .d_flags = D_DISK | D_TRACKCLOSE, 69 }; 70 71 static g_taste_t g_dev_taste; 72 static g_orphan_t g_dev_orphan; 73 74 static struct g_class g_dev_class = { 75 .name = "DEV", 76 .taste = g_dev_taste, 77 G_CLASS_INITIALIZER 78 }; 79 80 int 81 g_dev_print(void) 82 { 83 struct g_geom *gp; 84 85 if (LIST_EMPTY(&g_dev_class.geom)) 86 return (0); 87 printf("List of GEOM disk devices:\n "); 88 LIST_FOREACH(gp, &g_dev_class.geom, geom) 89 printf(" %s", gp->name); 90 printf("\n"); 91 return (1); 92 } 93 94 /* 95 * XXX: This is disgusting and wrong in every way imaginable: The only reason 96 * XXX: we have a clone function is because of the root-mount hack we currently 97 * XXX: employ. An improvment would be to unregister this cloner once we know 98 * XXX: we no longer need it. Ideally, root-fs would be mounted through DEVFS 99 * XXX: eliminating the need for this hack. 100 */ 101 static void 102 g_dev_clone(void *arg __unused, char *name, int namelen __unused, dev_t *dev) 103 { 104 struct g_geom *gp; 105 106 if (*dev != NODEV) 107 return; 108 109 g_waitidle(); 110 111 /* g_topology_lock(); */ 112 LIST_FOREACH(gp, &g_dev_class.geom, geom) { 113 if (strcmp(gp->name, name)) 114 continue; 115 *dev = gp->softc; 116 g_trace(G_T_TOPOLOGY, "g_dev_clone(%s) = %p", name, *dev); 117 return; 118 } 119 /* g_topology_unlock(); */ 120 return; 121 } 122 123 static void 124 g_dev_register_cloner(void *foo __unused) 125 { 126 static int once; 127 128 /* XXX: why would this happen more than once ?? */ 129 if (!once) { 130 EVENTHANDLER_REGISTER(dev_clone, g_dev_clone, 0, 1000); 131 once++; 132 } 133 } 134 135 SYSINIT(geomdev,SI_SUB_DRIVERS,SI_ORDER_MIDDLE,g_dev_register_cloner,NULL); 136 137 static struct g_geom * 138 g_dev_taste(struct g_class *mp, struct g_provider *pp, int insist __unused) 139 { 140 struct g_geom *gp; 141 struct g_consumer *cp; 142 static int unit = GEOM_MINOR_PROVIDERS; 143 int error; 144 dev_t dev; 145 146 g_trace(G_T_TOPOLOGY, "dev_taste(%s,%s)", mp->name, pp->name); 147 g_topology_assert(); 148 LIST_FOREACH(cp, &pp->consumers, consumers) 149 if (cp->geom->class == mp) 150 return (NULL); 151 gp = g_new_geomf(mp, pp->name); 152 gp->orphan = g_dev_orphan; 153 cp = g_new_consumer(gp); 154 error = g_attach(cp, pp); 155 KASSERT(error == 0, 156 ("g_dev_taste(%s) failed to g_attach, err=%d", pp->name, error)); 157 /* 158 * XXX: I'm not 100% sure we can call make_dev(9) without Giant 159 * yet. Once we can, we don't need to drop topology here either. 160 */ 161 g_topology_unlock(); 162 mtx_lock(&Giant); 163 dev = make_dev(&g_dev_cdevsw, unit2minor(unit++), 164 UID_ROOT, GID_OPERATOR, 0640, gp->name); 165 if (pp->flags & G_PF_CANDELETE) 166 dev->si_flags |= SI_CANDELETE; 167 mtx_unlock(&Giant); 168 g_topology_lock(); 169 dev->si_iosize_max = MAXPHYS; 170 dev->si_stripesize = pp->stripesize; 171 dev->si_stripeoffset = pp->stripeoffset; 172 gp->softc = dev; 173 dev->si_drv1 = gp; 174 dev->si_drv2 = cp; 175 return (gp); 176 } 177 178 static int 179 g_dev_open(dev_t dev, int flags, int fmt, struct thread *td) 180 { 181 struct g_geom *gp; 182 struct g_consumer *cp; 183 int error, r, w, e; 184 185 gp = dev->si_drv1; 186 cp = dev->si_drv2; 187 if (gp == NULL || cp == NULL) 188 return(ENXIO); 189 g_trace(G_T_ACCESS, "g_dev_open(%s, %d, %d, %p)", 190 gp->name, flags, fmt, td); 191 DROP_GIANT(); 192 g_topology_lock(); 193 r = flags & FREAD ? 1 : 0; 194 w = flags & FWRITE ? 1 : 0; 195 #ifdef notyet 196 e = flags & O_EXCL ? 1 : 0; 197 #else 198 e = 0; 199 #endif 200 error = g_access_rel(cp, r, w, e); 201 g_topology_unlock(); 202 PICKUP_GIANT(); 203 g_waitidle(); 204 dev->si_bsize_phys = cp->provider->sectorsize; 205 return(error); 206 } 207 208 static int 209 g_dev_close(dev_t dev, int flags, int fmt, struct thread *td) 210 { 211 struct g_geom *gp; 212 struct g_consumer *cp; 213 int error, r, w, e; 214 215 gp = dev->si_drv1; 216 cp = dev->si_drv2; 217 if (gp == NULL || cp == NULL) 218 return(ENXIO); 219 g_trace(G_T_ACCESS, "g_dev_close(%s, %d, %d, %p)", 220 gp->name, flags, fmt, td); 221 DROP_GIANT(); 222 g_topology_lock(); 223 r = flags & FREAD ? -1 : 0; 224 w = flags & FWRITE ? -1 : 0; 225 #ifdef notyet 226 e = flags & O_EXCL ? -1 : 0; 227 #else 228 e = 0; 229 #endif 230 error = g_access_rel(cp, r, w, e); 231 g_topology_unlock(); 232 PICKUP_GIANT(); 233 g_waitidle(); 234 return (error); 235 } 236 237 static int 238 g_dev_ioctl(dev_t dev, u_long cmd, caddr_t data, int fflag, struct thread *td) 239 { 240 struct g_geom *gp, *gp2; 241 struct g_consumer *cp; 242 struct g_provider *pp2; 243 struct g_kerneldump kd; 244 int i, error; 245 u_int u; 246 struct g_ioctl *gio; 247 248 gp = dev->si_drv1; 249 cp = dev->si_drv2; 250 pp2 = cp->provider; 251 gp2 = pp2->geom; 252 gio = NULL; 253 254 error = 0; 255 DROP_GIANT(); 256 257 gio = NULL; 258 i = IOCPARM_LEN(cmd); 259 switch (cmd) { 260 case DIOCGSECTORSIZE: 261 *(u_int *)data = cp->provider->sectorsize; 262 if (*(u_int *)data == 0) 263 error = ENOENT; 264 break; 265 case DIOCGMEDIASIZE: 266 *(off_t *)data = cp->provider->mediasize; 267 if (*(off_t *)data == 0) 268 error = ENOENT; 269 break; 270 case DIOCGFWSECTORS: 271 error = g_io_getattr("GEOM::fwsectors", cp, &i, data); 272 if (error == 0 && *(u_int *)data == 0) 273 error = ENOENT; 274 break; 275 case DIOCGFWHEADS: 276 error = g_io_getattr("GEOM::fwheads", cp, &i, data); 277 if (error == 0 && *(u_int *)data == 0) 278 error = ENOENT; 279 break; 280 case DIOCGFRONTSTUFF: 281 error = g_io_getattr("GEOM::frontstuff", cp, &i, data); 282 break; 283 case DIOCSKERNELDUMP: 284 u = *((u_int *)data); 285 if (!u) { 286 set_dumper(NULL); 287 error = 0; 288 break; 289 } 290 kd.offset = 0; 291 kd.length = OFF_MAX; 292 i = sizeof kd; 293 error = g_io_getattr("GEOM::kerneldump", cp, &i, &kd); 294 if (!error) 295 dev->si_flags |= SI_DUMPDEV; 296 break; 297 298 default: 299 gio = g_malloc(sizeof *gio, M_WAITOK | M_ZERO); 300 gio->cmd = cmd; 301 gio->data = data; 302 gio->fflag = fflag; 303 gio->td = td; 304 i = sizeof *gio; 305 /* 306 * We always issue ioctls as getattr since the direction of data 307 * movement in ioctl is no indication of the ioctl being a "set" 308 * or "get" type ioctl or if such simplistic terms even apply 309 */ 310 error = g_io_getattr("GEOM::ioctl", cp, &i, gio); 311 break; 312 } 313 314 PICKUP_GIANT(); 315 if (error == EDIRIOCTL) { 316 KASSERT(gio != NULL, ("NULL gio but EDIRIOCTL")); 317 KASSERT(gio->func != NULL, ("NULL function but EDIRIOCTL")); 318 error = (gio->func)(gio->dev, cmd, data, fflag, td); 319 } 320 g_waitidle(); 321 if (gio != NULL && (error == EOPNOTSUPP || error == ENOIOCTL)) { 322 if (g_debugflags & G_T_TOPOLOGY) { 323 i = IOCGROUP(cmd); 324 printf("IOCTL(0x%lx) \"%s\"", cmd, gp->name); 325 if (i > ' ' && i <= '~') 326 printf(" '%c'", (int)IOCGROUP(cmd)); 327 else 328 printf(" 0x%lx", IOCGROUP(cmd)); 329 printf("/%ld ", cmd & 0xff); 330 if (cmd & IOC_IN) 331 printf("I"); 332 if (cmd & IOC_OUT) 333 printf("O"); 334 printf("(%ld) = ENOIOCTL\n", IOCPARM_LEN(cmd)); 335 } 336 error = ENOTTY; 337 } 338 if (gio != NULL) 339 g_free(gio); 340 return (error); 341 } 342 343 static void 344 g_dev_done(struct bio *bp2) 345 { 346 struct bio *bp; 347 348 bp = bp2->bio_parent; 349 bp->bio_error = bp2->bio_error; 350 if (bp->bio_error != 0) { 351 g_trace(G_T_BIO, "g_dev_done(%p) had error %d", 352 bp2, bp->bio_error); 353 bp->bio_flags |= BIO_ERROR; 354 } else { 355 g_trace(G_T_BIO, "g_dev_done(%p/%p) resid %ld completed %jd", 356 bp2, bp, bp->bio_resid, (intmax_t)bp2->bio_completed); 357 } 358 bp->bio_resid = bp->bio_bcount - bp2->bio_completed; 359 g_destroy_bio(bp2); 360 mtx_lock(&Giant); 361 biodone(bp); 362 mtx_unlock(&Giant); 363 } 364 365 static void 366 g_dev_strategy(struct bio *bp) 367 { 368 struct g_geom *gp; 369 struct g_consumer *cp; 370 struct bio *bp2; 371 dev_t dev; 372 373 KASSERT(bp->bio_cmd == BIO_READ || 374 bp->bio_cmd == BIO_WRITE || 375 bp->bio_cmd == BIO_DELETE, 376 ("Wrong bio_cmd bio=%p cmd=%d", bp, bp->bio_cmd)); 377 dev = bp->bio_dev; 378 gp = dev->si_drv1; 379 cp = dev->si_drv2; 380 bp2 = g_clone_bio(bp); 381 KASSERT(bp2 != NULL, ("XXX: ENOMEM in a bad place")); 382 bp2->bio_offset = (off_t)bp->bio_blkno << DEV_BSHIFT; 383 KASSERT(bp2->bio_offset >= 0, 384 ("Negative bio_offset (%jd) on bio %p", 385 (intmax_t)bp2->bio_offset, bp)); 386 bp2->bio_length = (off_t)bp->bio_bcount; 387 bp2->bio_done = g_dev_done; 388 g_trace(G_T_BIO, 389 "g_dev_strategy(%p/%p) offset %jd length %jd data %p cmd %d", 390 bp, bp2, (intmax_t)bp->bio_offset, (intmax_t)bp2->bio_length, 391 bp2->bio_data, bp2->bio_cmd); 392 g_io_request(bp2, cp); 393 } 394 395 /* 396 * g_dev_orphan() 397 * 398 * Called from below when the provider orphaned us. 399 * - Clear any dump settings. 400 * - Destroy the dev_t to prevent any more request from coming in. The 401 * provider is already marked with an error, so anything which comes in 402 * in the interrim will be returned immediately. 403 * - Wait for any outstanding I/O to finish. 404 * - Set our access counts to zero, whatever they were. 405 * - Detach and self-destruct. 406 */ 407 408 static void 409 g_dev_orphan(struct g_consumer *cp) 410 { 411 struct g_geom *gp; 412 dev_t dev; 413 414 g_topology_assert(); 415 gp = cp->geom; 416 dev = gp->softc; 417 g_trace(G_T_TOPOLOGY, "g_dev_orphan(%p(%s))", cp, gp->name); 418 419 /* Reset any dump-area set on this device */ 420 if (dev->si_flags & SI_DUMPDEV) 421 set_dumper(NULL); 422 423 /* Destroy the dev_t so we get no more requests */ 424 destroy_dev(dev); 425 426 /* Wait for the cows to come home */ 427 while (cp->nstart != cp->nend) 428 msleep(&dev, NULL, PRIBIO, "gdevorphan", 1); 429 430 if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0) 431 g_access_rel(cp, -cp->acr, -cp->acw, -cp->ace); 432 433 g_detach(cp); 434 g_destroy_consumer(cp); 435 g_destroy_geom(gp); 436 } 437 438 DECLARE_GEOM_CLASS(g_dev_class, g_dev); 439