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/fcntl.h> 48 #include <sys/malloc.h> 49 #include <sys/sysctl.h> 50 #include <sys/devicestat.h> 51 #include <machine/md_var.h> 52 53 #include <sys/lock.h> 54 #include <sys/mutex.h> 55 #include <geom/geom.h> 56 #include <geom/geom_disk.h> 57 #include <geom/geom_int.h> 58 59 static struct mtx g_disk_done_mtx; 60 61 static g_access_t g_disk_access; 62 static g_init_t g_disk_init; 63 static g_fini_t g_disk_fini; 64 static g_start_t g_disk_start; 65 static g_ioctl_t g_disk_ioctl; 66 static g_dumpconf_t g_disk_dumpconf; 67 68 static struct g_class g_disk_class = { 69 .name = "DISK", 70 .version = G_VERSION, 71 .init = g_disk_init, 72 .fini = g_disk_fini, 73 .start = g_disk_start, 74 .access = g_disk_access, 75 .ioctl = g_disk_ioctl, 76 .dumpconf = g_disk_dumpconf, 77 }; 78 79 static void 80 g_disk_init(struct g_class *mp __unused) 81 { 82 83 mtx_init(&g_disk_done_mtx, "g_disk_done", NULL, MTX_DEF); 84 } 85 86 static void 87 g_disk_fini(struct g_class *mp __unused) 88 { 89 90 mtx_destroy(&g_disk_done_mtx); 91 } 92 93 DECLARE_GEOM_CLASS(g_disk_class, g_disk); 94 95 static void __inline 96 g_disk_lock_giant(struct disk *dp) 97 { 98 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 99 mtx_lock(&Giant); 100 } 101 102 static void __inline 103 g_disk_unlock_giant(struct disk *dp) 104 { 105 if (dp->d_flags & DISKFLAG_NEEDSGIANT) 106 mtx_unlock(&Giant); 107 } 108 109 static int 110 g_disk_access(struct g_provider *pp, int r, int w, int e) 111 { 112 struct disk *dp; 113 int error; 114 115 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 116 pp->name, r, w, e); 117 g_topology_assert(); 118 dp = pp->geom->softc; 119 if (dp == NULL || dp->d_destroyed) { 120 /* 121 * Allow decreasing access count even if disk is not 122 * avaliable anymore. 123 */ 124 if (r <= 0 && w <= 0 && e <= 0) 125 return (0); 126 return (ENXIO); 127 } 128 r += pp->acr; 129 w += pp->acw; 130 e += pp->ace; 131 error = 0; 132 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 133 if (dp->d_open != NULL) { 134 g_disk_lock_giant(dp); 135 error = dp->d_open(dp); 136 if (bootverbose && error != 0) 137 printf("Opened disk %s -> %d\n", 138 pp->name, error); 139 g_disk_unlock_giant(dp); 140 } 141 pp->mediasize = dp->d_mediasize; 142 pp->sectorsize = dp->d_sectorsize; 143 dp->d_flags |= DISKFLAG_OPEN; 144 if (dp->d_maxsize == 0) { 145 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 146 dp->d_name, dp->d_unit); 147 dp->d_maxsize = DFLTPHYS; 148 } 149 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 150 if (dp->d_close != NULL) { 151 g_disk_lock_giant(dp); 152 error = dp->d_close(dp); 153 if (error != 0) 154 printf("Closed disk %s -> %d\n", 155 pp->name, error); 156 g_disk_unlock_giant(dp); 157 } 158 dp->d_flags &= ~DISKFLAG_OPEN; 159 } 160 return (error); 161 } 162 163 static void 164 g_disk_kerneldump(struct bio *bp, struct disk *dp) 165 { 166 int error; 167 struct g_kerneldump *gkd; 168 struct dumperinfo di; 169 struct g_geom *gp; 170 171 gkd = (struct g_kerneldump*)bp->bio_data; 172 gp = bp->bio_to->geom; 173 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)", 174 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 175 if (dp->d_dump == NULL) { 176 g_io_deliver(bp, ENODEV); 177 return; 178 } 179 di.dumper = dp->d_dump; 180 di.priv = dp; 181 di.blocksize = dp->d_sectorsize; 182 di.maxiosize = dp->d_maxsize; 183 di.mediaoffset = gkd->offset; 184 if ((gkd->offset + gkd->length) > dp->d_mediasize) 185 gkd->length = dp->d_mediasize - gkd->offset; 186 di.mediasize = gkd->length; 187 error = set_dumper(&di); 188 g_io_deliver(bp, error); 189 } 190 191 static void 192 g_disk_done(struct bio *bp) 193 { 194 struct bio *bp2; 195 struct disk *dp; 196 197 /* See "notes" for why we need a mutex here */ 198 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 199 mtx_lock(&g_disk_done_mtx); 200 bp->bio_completed = bp->bio_length - bp->bio_resid; 201 202 bp2 = bp->bio_parent; 203 if (bp2->bio_error == 0) 204 bp2->bio_error = bp->bio_error; 205 bp2->bio_completed += bp->bio_completed; 206 if ((bp->bio_cmd & (BIO_READ|BIO_WRITE|BIO_DELETE)) && 207 (dp = bp2->bio_to->geom->softc)) { 208 devstat_end_transaction_bio(dp->d_devstat, bp); 209 } 210 g_destroy_bio(bp); 211 bp2->bio_inbed++; 212 if (bp2->bio_children == bp2->bio_inbed) { 213 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 214 g_io_deliver(bp2, bp2->bio_error); 215 } 216 mtx_unlock(&g_disk_done_mtx); 217 } 218 219 static int 220 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, int fflag, struct thread *td) 221 { 222 struct g_geom *gp; 223 struct disk *dp; 224 int error; 225 226 gp = pp->geom; 227 dp = gp->softc; 228 229 if (dp->d_ioctl == NULL) 230 return (ENOIOCTL); 231 g_disk_lock_giant(dp); 232 error = dp->d_ioctl(dp, cmd, data, fflag, td); 233 g_disk_unlock_giant(dp); 234 return(error); 235 } 236 237 static void 238 g_disk_start(struct bio *bp) 239 { 240 struct bio *bp2, *bp3; 241 struct disk *dp; 242 int error; 243 off_t off; 244 245 dp = bp->bio_to->geom->softc; 246 if (dp == NULL || dp->d_destroyed) { 247 g_io_deliver(bp, ENXIO); 248 return; 249 } 250 error = EJUSTRETURN; 251 switch(bp->bio_cmd) { 252 case BIO_DELETE: 253 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 254 error = 0; 255 break; 256 } 257 /* fall-through */ 258 case BIO_READ: 259 case BIO_WRITE: 260 off = 0; 261 bp3 = NULL; 262 bp2 = g_clone_bio(bp); 263 if (bp2 == NULL) { 264 error = ENOMEM; 265 break; 266 } 267 do { 268 bp2->bio_offset += off; 269 bp2->bio_length -= off; 270 bp2->bio_data += off; 271 if (bp2->bio_length > dp->d_maxsize) { 272 /* 273 * XXX: If we have a stripesize we should really 274 * use it here. 275 */ 276 bp2->bio_length = dp->d_maxsize; 277 off += dp->d_maxsize; 278 /* 279 * To avoid a race, we need to grab the next bio 280 * before we schedule this one. See "notes". 281 */ 282 bp3 = g_clone_bio(bp); 283 if (bp3 == NULL) 284 bp->bio_error = ENOMEM; 285 } 286 bp2->bio_done = g_disk_done; 287 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 288 bp2->bio_bcount = bp2->bio_length; 289 bp2->bio_disk = dp; 290 devstat_start_transaction_bio(dp->d_devstat, bp2); 291 g_disk_lock_giant(dp); 292 dp->d_strategy(bp2); 293 g_disk_unlock_giant(dp); 294 bp2 = bp3; 295 bp3 = NULL; 296 } while (bp2 != NULL); 297 break; 298 case BIO_GETATTR: 299 if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors)) 300 break; 301 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 302 break; 303 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 304 break; 305 else if (g_handleattr_str(bp, "GEOM::ident", dp->d_ident)) 306 break; 307 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 308 g_disk_kerneldump(bp, dp); 309 else 310 error = ENOIOCTL; 311 break; 312 case BIO_FLUSH: 313 g_trace(G_T_TOPOLOGY, "g_disk_flushcache(%s)", 314 bp->bio_to->name); 315 if (!(dp->d_flags & DISKFLAG_CANFLUSHCACHE)) { 316 g_io_deliver(bp, ENODEV); 317 return; 318 } 319 bp2 = g_clone_bio(bp); 320 if (bp2 == NULL) { 321 g_io_deliver(bp, ENOMEM); 322 return; 323 } 324 bp2->bio_done = g_disk_done; 325 bp2->bio_disk = dp; 326 g_disk_lock_giant(dp); 327 dp->d_strategy(bp2); 328 g_disk_unlock_giant(dp); 329 break; 330 default: 331 error = EOPNOTSUPP; 332 break; 333 } 334 if (error != EJUSTRETURN) 335 g_io_deliver(bp, error); 336 return; 337 } 338 339 static void 340 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 341 { 342 struct disk *dp; 343 344 dp = gp->softc; 345 if (dp == NULL) 346 return; 347 if (indent == NULL) { 348 sbuf_printf(sb, " hd %u", dp->d_fwheads); 349 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 350 return; 351 } 352 if (pp != NULL) { 353 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 354 indent, dp->d_fwheads); 355 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 356 indent, dp->d_fwsectors); 357 } 358 } 359 360 static void 361 g_disk_create(void *arg, int flag) 362 { 363 struct g_geom *gp; 364 struct g_provider *pp; 365 struct disk *dp; 366 367 if (flag == EV_CANCEL) 368 return; 369 g_topology_assert(); 370 dp = arg; 371 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 372 gp->softc = dp; 373 pp = g_new_providerf(gp, "%s", gp->name); 374 pp->mediasize = dp->d_mediasize; 375 pp->sectorsize = dp->d_sectorsize; 376 if (dp->d_flags & DISKFLAG_CANDELETE) 377 pp->flags |= G_PF_CANDELETE; 378 pp->stripeoffset = dp->d_stripeoffset; 379 pp->stripesize = dp->d_stripesize; 380 if (bootverbose) 381 printf("GEOM: new disk %s\n", gp->name); 382 dp->d_geom = gp; 383 g_error_provider(pp, 0); 384 root_mount_rel(dp->d_roothold); 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 [a-zA-Z0-9-_@#%.:] characters, the rest is converted to 'x<HH>'. 405 */ 406 static void 407 g_disk_ident_adjust(char *ident, size_t size) 408 { 409 char newid[DISK_IDENT_SIZE], tmp[4]; 410 size_t len; 411 char *p; 412 413 bzero(newid, sizeof(newid)); 414 len = 0; 415 for (p = ident; *p != '\0' && len < sizeof(newid) - 1; p++) { 416 switch (*p) { 417 default: 418 if ((*p < 'a' || *p > 'z') && 419 (*p < 'A' || *p > 'Z') && 420 (*p < '0' || *p > '9')) { 421 snprintf(tmp, sizeof(tmp), "x%02hhx", *p); 422 strlcat(newid, tmp, sizeof(newid)); 423 len += 3; 424 break; 425 } 426 /* FALLTHROUGH */ 427 case '-': 428 case '_': 429 case '@': 430 case '#': 431 case '%': 432 case '.': 433 case ':': 434 newid[len++] = *p; 435 break; 436 } 437 } 438 bzero(ident, size); 439 strlcpy(ident, newid, size); 440 } 441 442 struct disk * 443 disk_alloc() 444 { 445 struct disk *dp; 446 447 dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO); 448 return (dp); 449 } 450 451 void 452 disk_create(struct disk *dp, int version) 453 { 454 if (version != DISK_VERSION_00 && version != DISK_VERSION_01) { 455 printf("WARNING: Attempt to add disk %s%d %s", 456 dp->d_name, dp->d_unit, 457 " using incompatible ABI version of disk(9)\n"); 458 printf("WARNING: Ignoring disk %s%d\n", 459 dp->d_name, dp->d_unit); 460 return; 461 } 462 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 463 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 464 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 465 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 466 if (dp->d_devstat == NULL) 467 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 468 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 469 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 470 dp->d_geom = NULL; 471 dp->d_roothold = root_mount_hold(dp->d_name, M_WAITOK); 472 g_disk_ident_adjust(dp->d_ident, sizeof(dp->d_ident)); 473 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 474 } 475 476 void 477 disk_destroy(struct disk *dp) 478 { 479 480 g_cancel_event(dp); 481 dp->d_destroyed = 1; 482 if (dp->d_devstat != NULL) 483 devstat_remove_entry(dp->d_devstat); 484 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 485 } 486 487 void 488 disk_gone(struct disk *dp) 489 { 490 struct g_geom *gp; 491 struct g_provider *pp; 492 493 gp = dp->d_geom; 494 if (gp != NULL) 495 LIST_FOREACH(pp, &gp->provider, provider) 496 g_wither_provider(pp, ENXIO); 497 } 498 499 static void 500 g_kern_disks(void *p, int flag __unused) 501 { 502 struct sbuf *sb; 503 struct g_geom *gp; 504 char *sp; 505 506 sb = p; 507 sp = ""; 508 g_topology_assert(); 509 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 510 sbuf_printf(sb, "%s%s", sp, gp->name); 511 sp = " "; 512 } 513 sbuf_finish(sb); 514 } 515 516 static int 517 sysctl_disks(SYSCTL_HANDLER_ARGS) 518 { 519 int error; 520 struct sbuf *sb; 521 522 sb = sbuf_new_auto(); 523 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 524 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 525 sbuf_delete(sb); 526 return error; 527 } 528 529 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0, 530 sysctl_disks, "A", "names of available disks"); 531 532