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 36 #include <sys/cdefs.h> 37 __FBSDID("$FreeBSD$"); 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/malloc.h> 42 #include <sys/kernel.h> 43 #include <sys/conf.h> 44 #include <sys/bio.h> 45 #include <sys/lock.h> 46 #include <sys/mutex.h> 47 #include <sys/proc.h> 48 #include <sys/errno.h> 49 #include <sys/time.h> 50 #include <sys/disk.h> 51 #include <sys/fcntl.h> 52 #include <sys/limits.h> 53 #include <geom/geom.h> 54 #include <geom/geom_int.h> 55 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; 60 61 static struct cdevsw g_dev_cdevsw = { 62 .d_version = D_VERSION, 63 .d_open = g_dev_open, 64 .d_close = g_dev_close, 65 .d_read = physread, 66 .d_write = physwrite, 67 .d_ioctl = g_dev_ioctl, 68 .d_strategy = g_dev_strategy, 69 .d_name = "g_dev", 70 .d_flags = D_DISK | D_TRACKCLOSE, 71 }; 72 73 static g_taste_t g_dev_taste; 74 static g_orphan_t g_dev_orphan; 75 static g_init_t g_dev_init; 76 77 static struct g_class g_dev_class = { 78 .name = "DEV", 79 .version = G_VERSION, 80 .taste = g_dev_taste, 81 .orphan = g_dev_orphan, 82 .init = g_dev_init, 83 }; 84 85 static struct unrhdr *unithdr; /* Locked by topology */ 86 87 static void 88 g_dev_init(struct g_class *mp) 89 { 90 91 /* XXX: should have a #define MAX_UNIT_MINOR */ 92 unithdr = new_unrhdr(0, 0xffffff); 93 } 94 95 void 96 g_dev_print(void) 97 { 98 struct g_geom *gp; 99 char const *p = ""; 100 101 LIST_FOREACH(gp, &g_dev_class.geom, geom) { 102 printf("%s%s", p, gp->name); 103 p = " "; 104 } 105 printf("\n"); 106 } 107 108 struct g_provider * 109 g_dev_getprovider(struct cdev *dev) 110 { 111 struct g_consumer *cp; 112 113 g_topology_assert(); 114 if (dev == NULL) 115 return (NULL); 116 if (dev->si_devsw != &g_dev_cdevsw) 117 cp = NULL; 118 else 119 cp = dev->si_drv2; 120 return (cp->provider); 121 } 122 123 124 static struct g_geom * 125 g_dev_taste(struct g_class *mp, struct g_provider *pp, int insist __unused) 126 { 127 struct g_geom *gp; 128 struct g_consumer *cp; 129 int error; 130 struct cdev *dev; 131 u_int unit; 132 133 g_trace(G_T_TOPOLOGY, "dev_taste(%s,%s)", mp->name, pp->name); 134 g_topology_assert(); 135 LIST_FOREACH(cp, &pp->consumers, consumers) 136 if (cp->geom->class == mp) 137 return (NULL); 138 gp = g_new_geomf(mp, pp->name); 139 cp = g_new_consumer(gp); 140 error = g_attach(cp, pp); 141 KASSERT(error == 0, 142 ("g_dev_taste(%s) failed to g_attach, err=%d", pp->name, error)); 143 /* 144 * XXX: I'm not 100% sure we can call make_dev(9) without Giant 145 * yet. Once we can, we don't need to drop topology here either. 146 */ 147 unit = alloc_unr(unithdr); 148 g_topology_unlock(); 149 mtx_lock(&Giant); 150 dev = make_dev(&g_dev_cdevsw, unit2minor(unit), 151 UID_ROOT, GID_OPERATOR, 0640, gp->name); 152 if (pp->flags & G_PF_CANDELETE) 153 dev->si_flags |= SI_CANDELETE; 154 mtx_unlock(&Giant); 155 g_topology_lock(); 156 dev->si_iosize_max = MAXPHYS; 157 gp->softc = dev; 158 dev->si_drv1 = gp; 159 dev->si_drv2 = cp; 160 return (gp); 161 } 162 163 static int 164 g_dev_open(struct cdev *dev, int flags, int fmt, struct thread *td) 165 { 166 struct g_geom *gp; 167 struct g_consumer *cp; 168 int error, r, w, e; 169 170 gp = dev->si_drv1; 171 cp = dev->si_drv2; 172 if (gp == NULL || cp == NULL || gp->softc != dev) 173 return(ENXIO); /* g_dev_taste() not done yet */ 174 175 g_trace(G_T_ACCESS, "g_dev_open(%s, %d, %d, %p)", 176 gp->name, flags, fmt, td); 177 178 r = flags & FREAD ? 1 : 0; 179 w = flags & FWRITE ? 1 : 0; 180 #ifdef notyet 181 e = flags & O_EXCL ? 1 : 0; 182 #else 183 e = 0; 184 #endif 185 if (w) { 186 /* 187 * When running in very secure mode, do not allow 188 * opens for writing of any disks. 189 */ 190 error = securelevel_ge(td->td_ucred, 2); 191 if (error) 192 return (error); 193 } 194 g_topology_lock(); 195 if (dev->si_devsw == NULL) 196 error = ENXIO; /* We were orphaned */ 197 else 198 error = g_access(cp, r, w, e); 199 g_topology_unlock(); 200 return(error); 201 } 202 203 static int 204 g_dev_close(struct cdev *dev, int flags, int fmt, struct thread *td) 205 { 206 struct g_geom *gp; 207 struct g_consumer *cp; 208 int error, r, w, e, i; 209 210 gp = dev->si_drv1; 211 cp = dev->si_drv2; 212 if (gp == NULL || cp == NULL) 213 return(ENXIO); 214 g_trace(G_T_ACCESS, "g_dev_close(%s, %d, %d, %p)", 215 gp->name, flags, fmt, td); 216 r = flags & FREAD ? -1 : 0; 217 w = flags & FWRITE ? -1 : 0; 218 #ifdef notyet 219 e = flags & O_EXCL ? -1 : 0; 220 #else 221 e = 0; 222 #endif 223 g_topology_lock(); 224 if (dev->si_devsw == NULL) 225 error = ENXIO; /* We were orphaned */ 226 else 227 error = g_access(cp, r, w, e); 228 for (i = 0; i < 10 * hz;) { 229 if (cp->acr != 0 || cp->acw != 0) 230 break; 231 if (cp->nstart == cp->nend) 232 break; 233 tsleep(&i, PRIBIO, "gdevwclose", hz / 10); 234 i += hz / 10; 235 } 236 if (cp->acr == 0 && cp->acw == 0 && cp->nstart != cp->nend) { 237 printf("WARNING: Final close of geom_dev(%s) %s %s\n", 238 gp->name, 239 "still has outstanding I/O after 10 seconds.", 240 "Completing close anyway, panic may happen later."); 241 } 242 g_topology_unlock(); 243 return (error); 244 } 245 246 /* 247 * XXX: Until we have unmessed the ioctl situation, there is a race against 248 * XXX: a concurrent orphanization. We cannot close it by holding topology 249 * XXX: since that would prevent us from doing our job, and stalling events 250 * XXX: will break (actually: stall) the BSD disklabel hacks. 251 */ 252 static int 253 g_dev_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag, struct thread *td) 254 { 255 struct g_geom *gp; 256 struct g_consumer *cp; 257 struct g_kerneldump kd; 258 int i, error; 259 u_int u; 260 261 gp = dev->si_drv1; 262 cp = dev->si_drv2; 263 264 error = 0; 265 KASSERT(cp->acr || cp->acw, 266 ("Consumer with zero access count in g_dev_ioctl")); 267 268 i = IOCPARM_LEN(cmd); 269 switch (cmd) { 270 case DIOCGSECTORSIZE: 271 *(u_int *)data = cp->provider->sectorsize; 272 if (*(u_int *)data == 0) 273 error = ENOENT; 274 break; 275 case DIOCGMEDIASIZE: 276 *(off_t *)data = cp->provider->mediasize; 277 if (*(off_t *)data == 0) 278 error = ENOENT; 279 break; 280 case DIOCGFWSECTORS: 281 error = g_io_getattr("GEOM::fwsectors", cp, &i, data); 282 if (error == 0 && *(u_int *)data == 0) 283 error = ENOENT; 284 break; 285 case DIOCGFWHEADS: 286 error = g_io_getattr("GEOM::fwheads", cp, &i, data); 287 if (error == 0 && *(u_int *)data == 0) 288 error = ENOENT; 289 break; 290 case DIOCGFRONTSTUFF: 291 error = g_io_getattr("GEOM::frontstuff", cp, &i, data); 292 break; 293 case DIOCSKERNELDUMP: 294 u = *((u_int *)data); 295 if (!u) { 296 set_dumper(NULL); 297 error = 0; 298 break; 299 } 300 kd.offset = 0; 301 kd.length = OFF_MAX; 302 i = sizeof kd; 303 error = g_io_getattr("GEOM::kerneldump", cp, &i, &kd); 304 if (!error) 305 dev->si_flags |= SI_DUMPDEV; 306 break; 307 308 default: 309 if (cp->provider->geom->ioctl != NULL) { 310 error = cp->provider->geom->ioctl(cp->provider, cmd, data, fflag, td); 311 } else { 312 error = ENOIOCTL; 313 } 314 } 315 316 return (error); 317 } 318 319 static void 320 g_dev_done(struct bio *bp2) 321 { 322 struct bio *bp; 323 324 bp = bp2->bio_parent; 325 bp->bio_error = bp2->bio_error; 326 if (bp->bio_error != 0) { 327 g_trace(G_T_BIO, "g_dev_done(%p) had error %d", 328 bp2, bp->bio_error); 329 bp->bio_flags |= BIO_ERROR; 330 } else { 331 g_trace(G_T_BIO, "g_dev_done(%p/%p) resid %ld completed %jd", 332 bp2, bp, bp->bio_resid, (intmax_t)bp2->bio_completed); 333 } 334 bp->bio_resid = bp->bio_length - bp2->bio_completed; 335 bp->bio_completed = bp2->bio_completed; 336 g_destroy_bio(bp2); 337 biodone(bp); 338 } 339 340 static void 341 g_dev_strategy(struct bio *bp) 342 { 343 struct g_consumer *cp; 344 struct bio *bp2; 345 struct cdev *dev; 346 347 KASSERT(bp->bio_cmd == BIO_READ || 348 bp->bio_cmd == BIO_WRITE || 349 bp->bio_cmd == BIO_DELETE, 350 ("Wrong bio_cmd bio=%p cmd=%d", bp, bp->bio_cmd)); 351 dev = bp->bio_dev; 352 cp = dev->si_drv2; 353 KASSERT(cp->acr || cp->acw, 354 ("Consumer with zero access count in g_dev_strategy")); 355 356 if ((bp->bio_offset % cp->provider->sectorsize) != 0 || 357 (bp->bio_bcount % cp->provider->sectorsize) != 0) { 358 biofinish(bp, NULL, EINVAL); 359 return; 360 } 361 362 for (;;) { 363 /* 364 * XXX: This is not an ideal solution, but I belive it to 365 * XXX: deadlock safe, all things considered. 366 */ 367 bp2 = g_clone_bio(bp); 368 if (bp2 != NULL) 369 break; 370 tsleep(&bp, PRIBIO, "gdstrat", hz / 10); 371 } 372 KASSERT(bp2 != NULL, ("XXX: ENOMEM in a bad place")); 373 bp2->bio_done = g_dev_done; 374 g_trace(G_T_BIO, 375 "g_dev_strategy(%p/%p) offset %jd length %jd data %p cmd %d", 376 bp, bp2, (intmax_t)bp->bio_offset, (intmax_t)bp2->bio_length, 377 bp2->bio_data, bp2->bio_cmd); 378 g_io_request(bp2, cp); 379 KASSERT(cp->acr || cp->acw, 380 ("g_dev_strategy raced with g_dev_close and lost")); 381 382 } 383 384 /* 385 * g_dev_orphan() 386 * 387 * Called from below when the provider orphaned us. 388 * - Clear any dump settings. 389 * - Destroy the struct cdev *to prevent any more request from coming in. The 390 * provider is already marked with an error, so anything which comes in 391 * in the interrim will be returned immediately. 392 * - Wait for any outstanding I/O to finish. 393 * - Set our access counts to zero, whatever they were. 394 * - Detach and self-destruct. 395 */ 396 397 static void 398 g_dev_orphan(struct g_consumer *cp) 399 { 400 struct g_geom *gp; 401 struct cdev *dev; 402 u_int unit; 403 404 g_topology_assert(); 405 gp = cp->geom; 406 dev = gp->softc; 407 g_trace(G_T_TOPOLOGY, "g_dev_orphan(%p(%s))", cp, gp->name); 408 409 /* Reset any dump-area set on this device */ 410 if (dev->si_flags & SI_DUMPDEV) 411 set_dumper(NULL); 412 413 /* Destroy the struct cdev *so we get no more requests */ 414 unit = dev2unit(dev); 415 destroy_dev(dev); 416 free_unr(unithdr, unit); 417 418 /* Wait for the cows to come home */ 419 while (cp->nstart != cp->nend) 420 msleep(&dev, NULL, PRIBIO, "gdevorphan", hz / 10); 421 422 if (cp->acr > 0 || cp->acw > 0 || cp->ace > 0) 423 g_access(cp, -cp->acr, -cp->acw, -cp->ace); 424 425 g_detach(cp); 426 g_destroy_consumer(cp); 427 g_destroy_geom(gp); 428 } 429 430 DECLARE_GEOM_CLASS(g_dev_class, g_dev); 431