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 "opt_geom.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/kernel.h> 44 #include <sys/sysctl.h> 45 #include <sys/bio.h> 46 #include <sys/conf.h> 47 #include <sys/ctype.h> 48 #include <sys/fcntl.h> 49 #include <sys/malloc.h> 50 #include <sys/sysctl.h> 51 #include <sys/devicestat.h> 52 #include <machine/md_var.h> 53 54 #include <sys/lock.h> 55 #include <sys/mutex.h> 56 #include <geom/geom.h> 57 #include <geom/geom_disk.h> 58 #include <geom/geom_int.h> 59 60 static struct mtx g_disk_done_mtx; 61 62 static g_access_t g_disk_access; 63 static g_init_t g_disk_init; 64 static g_fini_t g_disk_fini; 65 static g_start_t g_disk_start; 66 static g_ioctl_t g_disk_ioctl; 67 static g_dumpconf_t g_disk_dumpconf; 68 69 static struct g_class g_disk_class = { 70 .name = "DISK", 71 .version = G_VERSION, 72 .init = g_disk_init, 73 .fini = g_disk_fini, 74 .start = g_disk_start, 75 .access = g_disk_access, 76 .ioctl = g_disk_ioctl, 77 .dumpconf = g_disk_dumpconf, 78 }; 79 80 static void 81 g_disk_init(struct g_class *mp __unused) 82 { 83 84 mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF); 85 } 86 87 static void 88 g_disk_fini(struct g_class *mp __unused) 89 { 90 91 mtx_destroy(&g_disk_done_mtx); 92 } 93 94 DECLARE_GEOM_CLASS(g_disk_class, g_disk); 95 96 static void __inline 97 g_disk_lock_giant(struct disk *dp) 98 { 99 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 100 mtx_lock(&Giant); 101 } 102 103 static void __inline 104 g_disk_unlock_giant(struct disk *dp) 105 { 106 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 107 mtx_unlock(&Giant); 108 } 109 110 static int 111 g_disk_access(struct g_provider *pp, int r, int w, int e) 112 { 113 struct disk *dp; 114 int error; 115 116 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 117 pp->name, r, w, e); 118 g_topology_assert(); 119 dp = pp->geom->softc; 120 if (dp == NULL || dp->d_destroyed) { 121 /* 122 * Allow decreasing access count even if disk is not 123 * avaliable anymore. 124 */ 125 if (r <= 0 && w <= 0 && e <= 0) 126 return (0); 127 return (ENXIO); 128 } 129 r += pp->acr; 130 w += pp->acw; 131 e += pp->ace; 132 error = 0; 133 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 134 if (dp->d_open != NULL) { 135 g_disk_lock_giant(dp); 136 error = dp->d_open(dp); 137 if (bootverbose && error != 0) 138 printf("Opened disk %s -> %d\n", 139 pp->name, error); 140 g_disk_unlock_giant(dp); 141 } 142 pp->mediasize = dp->d_mediasize; 143 pp->sectorsize = dp->d_sectorsize; 144 dp->d_flags |= DISKFLAG_OPEN; 145 if (dp->d_maxsize == 0) { 146 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 147 dp->d_name, dp->d_unit); 148 dp->d_maxsize = DFLTPHYS; 149 } 150 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 151 if (dp->d_close != NULL) { 152 g_disk_lock_giant(dp); 153 error = dp->d_close(dp); 154 if (error != 0) 155 printf("Closed disk %s -> %d\n", 156 pp->name, error); 157 g_disk_unlock_giant(dp); 158 } 159 dp->d_flags &= ~DISKFLAG_OPEN; 160 } 161 return (error); 162 } 163 164 static void 165 g_disk_kerneldump(struct bio *bp, struct disk *dp) 166 { 167 int error; 168 struct g_kerneldump *gkd; 169 struct dumperinfo di; 170 struct g_geom *gp; 171 172 gkd = (struct g_kerneldump*)bp->bio_data; 173 gp = bp->bio_to->geom; 174 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)", 175 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 176 if (dp->d_dump == NULL) { 177 g_io_deliver(bp, ENODEV); 178 return; 179 } 180 di.dumper = dp->d_dump; 181 di.priv = dp; 182 di.blocksize = dp->d_sectorsize; 183 di.maxiosize = dp->d_maxsize; 184 di.mediaoffset = gkd->offset; 185 if ((gkd->offset + gkd->length) > dp->d_mediasize) 186 gkd->length = dp->d_mediasize - gkd->offset; 187 di.mediasize = gkd->length; 188 error = set_dumper(&di); 189 g_io_deliver(bp, error); 190 } 191 192 static void 193 g_disk_done(struct bio *bp) 194 { 195 struct bio *bp2; 196 struct disk *dp; 197 198 /* See "notes" for why we need a mutex here */ 199 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 200 mtx_lock(&g_disk_done_mtx); 201 bp->bio_completed = bp->bio_length - bp->bio_resid; 202 203 bp2 = bp->bio_parent; 204 if (bp2->bio_error == 0) 205 bp2->bio_error = bp->bio_error; 206 bp2->bio_completed += bp->bio_completed; 207 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) && 208 (dp = bp2->bio_to->geom->softc)) { 209 devstat_end_transaction_bio(dp->d_devstat, bp); 210 } 211 g_destroy_bio(bp); 212 bp2->bio_inbed++; 213 if (bp2->bio_children == bp2->bio_inbed) { 214 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 215 g_io_deliver(bp2, bp2->bio_error); 216 } 217 mtx_unlock(&g_disk_done_mtx); 218 } 219 220 static int 221 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) 222 { 223 struct g_geom *gp; 224 struct disk *dp; 225 int error; 226 227 gp = pp->geom; 228 dp = gp->softc; 229 230 if (dp->d_ioctl == NULL) 231 return (ENOIOCTL); 232 g_disk_lock_giant(dp); 233 error = dp->d_ioctl(dp, cmd, data, fflag, td); 234 g_disk_unlock_giant(dp); 235 return(error); 236 } 237 238 static void 239 g_disk_start(struct bio *bp) 240 { 241 struct bio *bp2, *bp3; 242 struct disk *dp; 243 int error; 244 off_t off; 245 246 dp = bp->bio_to->geom->softc; 247 if (dp == NULL || dp->d_destroyed) { 248 g_io_deliver(bp, ENXIO); 249 return; 250 } 251 error = EJUSTRETURN; 252 switch(bp->bio_cmd) { 253 case BIO_DELETE: 254 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 255 error = 0; 256 break; 257 } 258 /* fall-through */ 259 case BIO_READ: 260 case BIO_WRITE: 261 off = 0; 262 bp3 = NULL; 263 bp2 = g_clone_bio(bp); 264 if (bp2 == NULL) { 265 error = ENOMEM; 266 break; 267 } 268 do { 269 bp2->bio_offset += off; 270 bp2->bio_length -= off; 271 bp2->bio_data += off; 272 if (bp2->bio_length > dp->d_maxsize) { 273 /* 274 * XXX: If we have a stripesize we should really 275 * use it here. 276 */ 277 bp2->bio_length = dp->d_maxsize; 278 off += dp->d_maxsize; 279 /* 280 * To avoid a race, we need to grab the next bio 281 * before we schedule this one. See "notes". 282 */ 283 bp3 = g_clone_bio(bp); 284 if (bp3 == NULL) 285 bp->bio_error = ENOMEM; 286 } 287 bp2->bio_done = g_disk_done; 288 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 289 bp2->bio_bcount = bp2->bio_length; 290 bp2->bio_disk = dp; 291 devstat_start_transaction_bio(dp->d_devstat, bp2); 292 g_disk_lock_giant(dp); 293 dp->d_strategy(bp2); 294 g_disk_unlock_giant(dp); 295 bp2 = bp3; 296 bp3 = NULL; 297 } while (bp2 != NULL); 298 break; 299 case BIO_GETATTR: 300 if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors)) 301 break; 302 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 303 break; 304 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 305 break; 306 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) 307 break; 308 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 309 g_disk_kerneldump(bp, dp); 310 else 311 error = ENOIOCTL; 312 break; 313 case BIO_FLUSH: 314 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)", 315 bp->bio_to->name); 316 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { 317 g_io_deliver(bp, ENODEV); 318 return; 319 } 320 bp2 = g_clone_bio(bp); 321 if (bp2 == NULL) { 322 g_io_deliver(bp, ENOMEM); 323 return; 324 } 325 bp2->bio_done = g_disk_done; 326 bp2->bio_disk = dp; 327 g_disk_lock_giant(dp); 328 dp->d_strategy(bp2); 329 g_disk_unlock_giant(dp); 330 break; 331 default: 332 error = EOPNOTSUPP; 333 break; 334 } 335 if (error != EJUSTRETURN) 336 g_io_deliver(bp, error); 337 return; 338 } 339 340 static void 341 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 342 { 343 struct disk *dp; 344 345 dp = gp->softc; 346 if (dp == NULL) 347 return; 348 if (indent == NULL) { 349 sbuf_printf(sb, " hd %u", dp->d_fwheads); 350 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 351 return; 352 } 353 if (pp != NULL) { 354 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 355 indent, dp->d_fwheads); 356 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 357 indent, dp->d_fwsectors); 358 } 359 } 360 361 static void 362 g_disk_create(void *arg, int flag) 363 { 364 struct g_geom *gp; 365 struct g_provider *pp; 366 struct disk *dp; 367 368 if (flag == EV_CANCEL) 369 return; 370 g_topology_assert(); 371 dp = arg; 372 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 373 gp->softc = dp; 374 pp = g_new_providerf(gp, "%s", gp->name); 375 pp->mediasize = dp->d_mediasize; 376 pp->sectorsize = dp->d_sectorsize; 377 if (dp->d_flags & DISKFLAG_CANDELETE) 378 pp->flags |= G_PF_CANDELETE; 379 pp->stripeoffset = dp->d_stripeoffset; 380 pp->stripesize = dp->d_stripesize; 381 if (bootverbose) 382 printf("GEOM: new disk %s\n", gp->name); 383 dp->d_geom = gp; 384 g_error_provider(pp, 0); 385 } 386 387 static void 388 g_disk_destroy(void *ptr, int flag) 389 { 390 struct disk *dp; 391 struct g_geom *gp; 392 393 g_topology_assert(); 394 dp = ptr; 395 gp = dp->d_geom; 396 if (gp != NULL) { 397 gp->softc = NULL; 398 g_wither_geom(gp, ENXIO); 399 } 400 g_free(dp); 401 } 402 403 /* 404 * We only allow printable characters in disk ident, 405 * the rest is converted to 'x<HH>'. 406 */ 407 static void 408 g_disk_ident_adjust(char *ident, size_t size) 409 { 410 char *p, tmp[4], newid[DISK_IDENT_SIZE]; 411 412 newid[0] = '\0'; 413 for (p = ident; *p != '\0'; p++) { 414 if (isprint(*p)) { 415 tmp[0] = *p; 416 tmp[1] = '\0'; 417 } else { 418 snprintf(tmp, sizeof(tmp), "x%02hhx", 419 *(unsigned char *)p); 420 } 421 if (strlcat(newid, tmp, sizeof(newid)) >= sizeof(newid)) 422 break; 423 } 424 bzero(ident, size); 425 strlcpy(ident, newid, size); 426 } 427 428 struct disk * 429 disk_alloc() 430 { 431 struct disk *dp; 432 433 dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO); 434 return (dp); 435 } 436 437 void 438 disk_create(struct disk *dp, int version) 439 { 440 if (version != DISK_VERSION_00 && version != DISK_VERSION_01) { 441 printf("WARNING: Attempt to add disk %s%d %s", 442 dp->d_name, dp->d_unit, 443 " using incompatible ABI version of disk(9)\n"); 444 printf("WARNING: Ignoring disk %s%d\n", 445 dp->d_name, dp->d_unit); 446 return; 447 } 448 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 449 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 450 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 451 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 452 if (dp->d_devstat == NULL) 453 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 454 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 455 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 456 dp->d_geom = NULL; 457 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); 458 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 459 } 460 461 void 462 disk_destroy(struct disk *dp) 463 { 464 465 g_cancel_event(dp); 466 dp->d_destroyed = 1; 467 if (dp->d_devstat != NULL) 468 devstat_remove_entry(dp->d_devstat); 469 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 470 } 471 472 void 473 disk_gone(struct disk *dp) 474 { 475 struct g_geom *gp; 476 struct g_provider *pp; 477 478 gp = dp->d_geom; 479 if (gp != NULL) 480 LIST_FOREACH(pp, &gp->provider, provider) 481 g_wither_provider(pp, ENXIO); 482 } 483 484 static void 485 g_kern_disks(void *p, int flag __unused) 486 { 487 struct sbuf *sb; 488 struct g_geom *gp; 489 char *sp; 490 491 sb = p; 492 sp = ""; 493 g_topology_assert(); 494 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 495 sbuf_printf(sb, "%s%s", sp, gp->name); 496 sp = " "; 497 } 498 sbuf_finish(sb); 499 } 500 501 static int 502 sysctl_disks(SYSCTL_HANDLER_ARGS) 503 { 504 int error; 505 struct sbuf *sb; 506 507 sb = sbuf_new_auto(); 508 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 509 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 510 sbuf_delete(sb); 511 return error; 512 } 513 514 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0, 515 sysctl_disks, "A", "names of available disks"); 516 517