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 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 (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.mediaoffset = gkd->offset; 183 di.mediasize = gkd->length; 184 error = set_dumper(&di); 185 g_io_deliver(bp, error); 186 } 187 188 static void 189 g_disk_done(struct bio *bp) 190 { 191 struct bio *bp2; 192 struct disk *dp; 193 194 /* See "notes" for why we need a mutex here */ 195 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 196 mtx_lock(&g_disk_done_mtx); 197 bp->bio_completed = bp->bio_length - bp->bio_resid; 198 199 bp2 = bp->bio_parent; 200 if (bp2->bio_error == 0) 201 bp2->bio_error = bp->bio_error; 202 bp2->bio_completed += bp->bio_completed; 203 g_destroy_bio(bp); 204 bp2->bio_inbed++; 205 if (bp2->bio_children == bp2->bio_inbed) { 206 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 207 if ((dp = bp2->bio_to->geom->softc)) 208 devstat_end_transaction_bio(dp->d_devstat, bp2); 209 g_io_deliver(bp2, bp2->bio_error); 210 } 211 mtx_unlock(&g_disk_done_mtx); 212 } 213 214 static int 215 g_disk_ioctl(struct g_provider *pp, u_long cmd, void * data, struct thread *td) 216 { 217 struct g_geom *gp; 218 struct disk *dp; 219 int error; 220 221 gp = pp->geom; 222 dp = gp->softc; 223 224 if (dp->d_ioctl == NULL) 225 return (ENOIOCTL); 226 g_disk_lock_giant(dp); 227 error = dp->d_ioctl(dp, cmd, data, 0, td); 228 g_disk_unlock_giant(dp); 229 return(error); 230 } 231 232 static void 233 g_disk_start(struct bio *bp) 234 { 235 struct bio *bp2, *bp3; 236 struct disk *dp; 237 int error; 238 off_t off; 239 240 dp = bp->bio_to->geom->softc; 241 if (dp == NULL || dp->d_destroyed) 242 g_io_deliver(bp, ENXIO); 243 error = EJUSTRETURN; 244 switch(bp->bio_cmd) { 245 case BIO_DELETE: 246 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 247 error = 0; 248 break; 249 } 250 /* fall-through */ 251 case BIO_READ: 252 case BIO_WRITE: 253 off = 0; 254 bp3 = NULL; 255 bp2 = g_clone_bio(bp); 256 if (bp2 == NULL) { 257 error = ENOMEM; 258 break; 259 } 260 devstat_start_transaction_bio(dp->d_devstat, bp); 261 do { 262 bp2->bio_offset += off; 263 bp2->bio_length -= off; 264 bp2->bio_data += off; 265 if (bp2->bio_length > dp->d_maxsize) { 266 /* 267 * XXX: If we have a stripesize we should really 268 * use it here. 269 */ 270 bp2->bio_length = dp->d_maxsize; 271 off += dp->d_maxsize; 272 /* 273 * To avoid a race, we need to grab the next bio 274 * before we schedule this one. See "notes". 275 */ 276 bp3 = g_clone_bio(bp); 277 if (bp3 == NULL) 278 bp->bio_error = ENOMEM; 279 } 280 bp2->bio_done = g_disk_done; 281 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 282 bp2->bio_bcount = bp2->bio_length; 283 bp2->bio_disk = dp; 284 g_disk_lock_giant(dp); 285 dp->d_strategy(bp2); 286 g_disk_unlock_giant(dp); 287 bp2 = bp3; 288 bp3 = NULL; 289 } while (bp2 != NULL); 290 break; 291 case BIO_GETATTR: 292 if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors)) 293 break; 294 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 295 break; 296 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 297 break; 298 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 299 g_disk_kerneldump(bp, dp); 300 else 301 error = ENOIOCTL; 302 break; 303 default: 304 error = EOPNOTSUPP; 305 break; 306 } 307 if (error != EJUSTRETURN) 308 g_io_deliver(bp, error); 309 return; 310 } 311 312 static void 313 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 314 { 315 struct disk *dp; 316 317 dp = gp->softc; 318 if (dp == NULL) 319 return; 320 if (indent == NULL) { 321 sbuf_printf(sb, " hd %u", dp->d_fwheads); 322 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 323 return; 324 } 325 if (pp != NULL) { 326 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 327 indent, dp->d_fwheads); 328 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 329 indent, dp->d_fwsectors); 330 } 331 } 332 333 static void 334 g_disk_create(void *arg, int flag) 335 { 336 struct g_geom *gp; 337 struct g_provider *pp; 338 struct disk *dp; 339 340 if (flag == EV_CANCEL) 341 return; 342 g_topology_assert(); 343 dp = arg; 344 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 345 gp->softc = dp; 346 pp = g_new_providerf(gp, "%s", gp->name); 347 pp->mediasize = dp->d_mediasize; 348 pp->sectorsize = dp->d_sectorsize; 349 if (dp->d_flags & DISKFLAG_CANDELETE) 350 pp->flags |= G_PF_CANDELETE; 351 pp->stripeoffset = dp->d_stripeoffset; 352 pp->stripesize = dp->d_stripesize; 353 if (bootverbose) 354 printf("GEOM: new disk %s\n", gp->name); 355 dp->d_geom = gp; 356 g_error_provider(pp, 0); 357 } 358 359 static void 360 g_disk_destroy(void *ptr, int flag) 361 { 362 struct disk *dp; 363 struct g_geom *gp; 364 365 g_topology_assert(); 366 dp = ptr; 367 gp = dp->d_geom; 368 gp->softc = NULL; 369 g_wither_geom(gp, ENXIO); 370 g_free(dp); 371 } 372 373 struct disk * 374 disk_alloc() 375 { 376 struct disk *dp; 377 378 dp = g_malloc(sizeof *dp, M_WAITOK | M_ZERO); 379 return (dp); 380 } 381 382 void 383 disk_create(struct disk *dp, int version) 384 { 385 if (version != DISK_VERSION_00) { 386 printf("WARNING: Attempt to add disk %s%d %s", 387 dp->d_name, dp->d_unit, 388 " using incompatible ABI version of disk(9)\n"); 389 printf("WARNING: Ignoring disk %s%d\n", 390 dp->d_name, dp->d_unit); 391 return; 392 } 393 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 394 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 395 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 396 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 397 if (dp->d_devstat == NULL) 398 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 399 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 400 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 401 dp->d_geom = NULL; 402 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 403 } 404 405 void 406 disk_destroy(struct disk *dp) 407 { 408 409 g_cancel_event(dp); 410 dp->d_destroyed = 1; 411 if (dp->d_devstat != NULL) 412 devstat_remove_entry(dp->d_devstat); 413 g_post_event(g_disk_destroy, dp, M_WAITOK, NULL); 414 } 415 416 static void 417 g_kern_disks(void *p, int flag __unused) 418 { 419 struct sbuf *sb; 420 struct g_geom *gp; 421 char *sp; 422 423 sb = p; 424 sp = ""; 425 g_topology_assert(); 426 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 427 sbuf_printf(sb, "%s%s", sp, gp->name); 428 sp = " "; 429 } 430 sbuf_finish(sb); 431 } 432 433 static int 434 sysctl_disks(SYSCTL_HANDLER_ARGS) 435 { 436 int error; 437 struct sbuf *sb; 438 439 sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND); 440 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 441 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 442 sbuf_delete(sb); 443 return error; 444 } 445 446 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0, 447 sysctl_disks, "A", "names of available disks"); 448 449