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 65 struct g_class g_disk_class = { 66 .name = "DISK", 67 .init = g_disk_init, 68 .fini = g_disk_fini, 69 }; 70 71 static void 72 g_disk_init(struct g_class *mp __unused) 73 { 74 75 mtx_init(&g_disk_done_mtx, "g_disk_done", MTX_DEF, 0); 76 } 77 78 static void 79 g_disk_fini(struct g_class *mp __unused) 80 { 81 82 mtx_destroy(&g_disk_done_mtx); 83 } 84 85 DECLARE_GEOM_CLASS(g_disk_class, g_disk); 86 87 static void __inline 88 g_disk_lock_giant(struct disk *dp) 89 { 90 if (dp->d_flags & DISKFLAG_NOGIANT) 91 return; 92 mtx_lock(&Giant); 93 } 94 95 static void __inline 96 g_disk_unlock_giant(struct disk *dp) 97 { 98 if (dp->d_flags & DISKFLAG_NOGIANT) 99 return; 100 mtx_unlock(&Giant); 101 } 102 103 static int 104 g_disk_access(struct g_provider *pp, int r, int w, int e) 105 { 106 struct disk *dp; 107 int error; 108 109 g_trace(G_T_ACCESS, "g_disk_access(%s, %d, %d, %d)", 110 pp->name, r, w, e); 111 g_topology_assert(); 112 r += pp->acr; 113 w += pp->acw; 114 e += pp->ace; 115 dp = pp->geom->softc; 116 if (dp == NULL) 117 return (ENXIO); 118 error = 0; 119 if ((pp->acr + pp->acw + pp->ace) == 0 && (r + w + e) > 0) { 120 if (dp->d_open != NULL) { 121 g_disk_lock_giant(dp); 122 error = dp->d_open(dp); 123 if (error != 0) 124 printf("Opened disk %s -> %d\n", 125 pp->name, error); 126 g_disk_unlock_giant(dp); 127 } 128 pp->mediasize = dp->d_mediasize; 129 pp->sectorsize = dp->d_sectorsize; 130 dp->d_flags |= DISKFLAG_OPEN; 131 if (dp->d_maxsize == 0) { 132 printf("WARNING: Disk drive %s%d has no d_maxsize\n", 133 dp->d_name, dp->d_unit); 134 dp->d_maxsize = DFLTPHYS; 135 } 136 } else if ((pp->acr + pp->acw + pp->ace) > 0 && (r + w + e) == 0) { 137 if (dp->d_close != NULL) { 138 g_disk_lock_giant(dp); 139 error = dp->d_close(dp); 140 if (error != 0) 141 printf("Closed disk %s -> %d\n", 142 pp->name, error); 143 g_disk_unlock_giant(dp); 144 } 145 dp->d_flags &= ~DISKFLAG_OPEN; 146 } 147 return (error); 148 } 149 150 static void 151 g_disk_kerneldump(struct bio *bp, struct disk *dp) 152 { 153 int error; 154 struct g_kerneldump *gkd; 155 struct dumperinfo di; 156 struct g_geom *gp; 157 158 gkd = (struct g_kerneldump*)bp->bio_data; 159 gp = bp->bio_to->geom; 160 g_trace(G_T_TOPOLOGY, "g_disk_kernedump(%s, %jd, %jd)", 161 gp->name, (intmax_t)gkd->offset, (intmax_t)gkd->length); 162 di.dumper = dp->d_dump; 163 di.priv = dp; 164 di.blocksize = dp->d_sectorsize; 165 di.mediaoffset = gkd->offset; 166 di.mediasize = gkd->length; 167 error = set_dumper(&di); 168 g_io_deliver(bp, error); 169 } 170 171 static void 172 g_disk_done(struct bio *bp) 173 { 174 struct bio *bp2; 175 struct disk *dp; 176 177 /* See "notes" for why we need a mutex here */ 178 /* XXX: will witness accept a mix of Giant/unGiant drivers here ? */ 179 mtx_lock(&g_disk_done_mtx); 180 bp->bio_completed = bp->bio_length - bp->bio_resid; 181 182 bp2 = bp->bio_parent; 183 dp = bp2->bio_to->geom->softc; 184 if (bp2->bio_error == 0) 185 bp2->bio_error = bp->bio_error; 186 bp2->bio_completed += bp->bio_completed; 187 g_destroy_bio(bp); 188 bp2->bio_inbed++; 189 if (bp2->bio_children == bp2->bio_inbed) { 190 bp2->bio_resid = bp2->bio_bcount - bp2->bio_completed; 191 devstat_end_transaction_bio(dp->d_devstat, bp2); 192 g_io_deliver(bp2, bp2->bio_error); 193 } 194 mtx_unlock(&g_disk_done_mtx); 195 } 196 197 static void 198 g_disk_start(struct bio *bp) 199 { 200 struct bio *bp2, *bp3; 201 struct disk *dp; 202 struct g_ioctl *gio; 203 int error; 204 off_t off; 205 206 dp = bp->bio_to->geom->softc; 207 if (dp == NULL) 208 g_io_deliver(bp, ENXIO); 209 error = EJUSTRETURN; 210 switch(bp->bio_cmd) { 211 case BIO_DELETE: 212 if (!(dp->d_flags & DISKFLAG_CANDELETE)) { 213 error = 0; 214 break; 215 } 216 /* fall-through */ 217 case BIO_READ: 218 case BIO_WRITE: 219 off = 0; 220 bp3 = NULL; 221 bp2 = g_clone_bio(bp); 222 if (bp2 == NULL) { 223 error = ENOMEM; 224 break; 225 } 226 devstat_start_transaction_bio(dp->d_devstat, bp); 227 do { 228 bp2->bio_offset += off; 229 bp2->bio_length -= off; 230 bp2->bio_data += off; 231 if (bp2->bio_length > dp->d_maxsize) { 232 /* 233 * XXX: If we have a stripesize we should really 234 * use it here. 235 */ 236 bp2->bio_length = dp->d_maxsize; 237 off += dp->d_maxsize; 238 /* 239 * To avoid a race, we need to grab the next bio 240 * before we schedule this one. See "notes". 241 */ 242 bp3 = g_clone_bio(bp); 243 if (bp3 == NULL) 244 bp->bio_error = ENOMEM; 245 } 246 bp2->bio_done = g_disk_done; 247 bp2->bio_blkno = bp2->bio_offset >> DEV_BSHIFT; 248 bp2->bio_pblkno = bp2->bio_offset / dp->d_sectorsize; 249 bp2->bio_bcount = bp2->bio_length; 250 bp2->bio_disk = dp; 251 g_disk_lock_giant(dp); 252 dp->d_strategy(bp2); 253 g_disk_unlock_giant(dp); 254 bp2 = bp3; 255 bp3 = NULL; 256 } while (bp2 != NULL); 257 break; 258 case BIO_GETATTR: 259 if (g_handleattr_int(bp, "GEOM::fwsectors", dp->d_fwsectors)) 260 break; 261 else if (g_handleattr_int(bp, "GEOM::fwheads", dp->d_fwheads)) 262 break; 263 else if (g_handleattr_off_t(bp, "GEOM::frontstuff", 0)) 264 break; 265 else if (!strcmp(bp->bio_attribute, "GEOM::kerneldump")) 266 g_disk_kerneldump(bp, dp); 267 else if ((g_debugflags & G_F_DISKIOCTL) && 268 (dp->d_ioctl != NULL) && 269 !strcmp(bp->bio_attribute, "GEOM::ioctl") && 270 bp->bio_length == sizeof *gio) { 271 gio = (struct g_ioctl *)bp->bio_data; 272 gio->dev = dp; 273 gio->func = (d_ioctl_t *)(dp->d_ioctl); 274 error = EDIRIOCTL; 275 } else 276 error = ENOIOCTL; 277 break; 278 default: 279 error = EOPNOTSUPP; 280 break; 281 } 282 if (error != EJUSTRETURN) 283 g_io_deliver(bp, error); 284 return; 285 } 286 287 static void 288 g_disk_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) 289 { 290 struct disk *dp; 291 292 dp = gp->softc; 293 if (indent == NULL) { 294 sbuf_printf(sb, " hd %u", dp->d_fwheads); 295 sbuf_printf(sb, " sc %u", dp->d_fwsectors); 296 return; 297 } 298 if (pp != NULL) { 299 sbuf_printf(sb, "%s<fwheads>%u</fwheads>\n", 300 indent, dp->d_fwheads); 301 sbuf_printf(sb, "%s<fwsectors>%u</fwsectors>\n", 302 indent, dp->d_fwsectors); 303 } 304 } 305 306 static void 307 g_disk_create(void *arg, int flag) 308 { 309 struct g_geom *gp; 310 struct g_provider *pp; 311 struct disk *dp; 312 313 if (flag == EV_CANCEL) 314 return; 315 g_topology_assert(); 316 dp = arg; 317 gp = g_new_geomf(&g_disk_class, "%s%d", dp->d_name, dp->d_unit); 318 gp->start = g_disk_start; 319 gp->access = g_disk_access; 320 gp->softc = dp; 321 gp->dumpconf = g_disk_dumpconf; 322 pp = g_new_providerf(gp, "%s", gp->name); 323 pp->mediasize = dp->d_mediasize; 324 pp->sectorsize = dp->d_sectorsize; 325 if (dp->d_flags & DISKFLAG_CANDELETE) 326 pp->flags |= G_PF_CANDELETE; 327 pp->stripeoffset = dp->d_stripeoffset; 328 pp->stripesize = dp->d_stripesize; 329 if (bootverbose) 330 printf("GEOM: new disk %s\n", gp->name); 331 dp->d_geom = gp; 332 g_error_provider(pp, 0); 333 } 334 335 336 337 void 338 disk_create(int unit, struct disk *dp, int flags, void *unused __unused, void * unused2 __unused) 339 { 340 341 dp->d_unit = unit; 342 dp->d_flags = flags; 343 KASSERT(dp->d_strategy != NULL, ("disk_create need d_strategy")); 344 KASSERT(dp->d_name != NULL, ("disk_create need d_name")); 345 KASSERT(*dp->d_name != 0, ("disk_create need d_name")); 346 KASSERT(strlen(dp->d_name) < SPECNAMELEN - 4, ("disk name too long")); 347 dp->d_devstat = devstat_new_entry(dp->d_name, dp->d_unit, 348 dp->d_sectorsize, DEVSTAT_ALL_SUPPORTED, 349 DEVSTAT_TYPE_DIRECT, DEVSTAT_PRIORITY_MAX); 350 dp->d_geom = NULL; 351 g_post_event(g_disk_create, dp, M_WAITOK, dp, NULL); 352 } 353 354 /* 355 * XXX: There is a race if disk_destroy() is called while the g_disk_create() 356 * XXX: event is running. I belive the current result is that disk_destroy() 357 * XXX: actually doesn't do anything. Considering that the driver owns the 358 * XXX: struct disk and is likely to free it in a few moments, this can 359 * XXX: hardly be said to be optimal. To what extent we can sleep in 360 * XXX: disk_create() and disk_destroy() is currently undefined (but generally 361 * XXX: undesirable) so any solution seems to involve an intrusive decision. 362 */ 363 364 static void 365 disk_destroy_event(void *ptr, int flag) 366 { 367 368 g_topology_assert(); 369 g_wither_geom(ptr, ENXIO); 370 } 371 372 void 373 disk_destroy(struct disk *dp) 374 { 375 struct g_geom *gp; 376 377 g_cancel_event(dp); 378 gp = dp->d_geom; 379 if (gp == NULL) 380 return; 381 gp->softc = NULL; 382 devstat_remove_entry(dp->d_devstat); 383 g_post_event(disk_destroy_event, gp, M_WAITOK, NULL, NULL); 384 } 385 386 static void 387 g_kern_disks(void *p, int flag __unused) 388 { 389 struct sbuf *sb; 390 struct g_geom *gp; 391 char *sp; 392 393 sb = p; 394 sp = ""; 395 g_topology_assert(); 396 LIST_FOREACH(gp, &g_disk_class.geom, geom) { 397 sbuf_printf(sb, "%s%s", sp, gp->name); 398 sp = " "; 399 } 400 sbuf_finish(sb); 401 } 402 403 static int 404 sysctl_disks(SYSCTL_HANDLER_ARGS) 405 { 406 int error; 407 struct sbuf *sb; 408 409 sb = sbuf_new(NULL, NULL, 0, SBUF_AUTOEXTEND); 410 sbuf_clear(sb); 411 g_waitfor_event(g_kern_disks, sb, M_WAITOK, NULL); 412 error = SYSCTL_OUT(req, sbuf_data(sb), sbuf_len(sb) + 1); 413 sbuf_delete(sb); 414 return error; 415 } 416 417 SYSCTL_PROC(_kern, OID_AUTO, disks, CTLTYPE_STRING | CTLFLAG_RD | CTLFLAG_NOLOCK, 0, 0, 418 sysctl_disks, "A", "names of available disks"); 419 420