1 /*- 2 * Copyright (c) 2004 Max Khon 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include <sys/param.h> 31 #include <sys/bio.h> 32 #include <sys/endian.h> 33 #include <sys/errno.h> 34 #include <sys/kernel.h> 35 #include <sys/lock.h> 36 #include <sys/mutex.h> 37 #include <sys/malloc.h> 38 #include <sys/systm.h> 39 #include <sys/sysctl.h> 40 41 #include <geom/geom.h> 42 #include <net/zlib.h> 43 44 FEATURE(geom_uzip, "GEOM uzip read-only compressed disks support"); 45 46 #undef GEOM_UZIP_DEBUG 47 #ifdef GEOM_UZIP_DEBUG 48 #define DPRINTF(a) printf a 49 #else 50 #define DPRINTF(a) 51 #endif 52 53 static MALLOC_DEFINE(M_GEOM_UZIP, "geom_uzip", "GEOM UZIP data structures"); 54 55 #define UZIP_CLASS_NAME "UZIP" 56 57 /* 58 * Maximum allowed valid block size (to prevent foot-shooting) 59 */ 60 #define MAX_BLKSZ (MAXPHYS - MAXPHYS / 1000 - 12) 61 62 /* 63 * Integer values (block size, number of blocks, offsets) 64 * are stored in big-endian (network) order on disk and struct cloop_header 65 * and in native order in struct g_uzip_softc 66 */ 67 68 #define CLOOP_MAGIC_LEN 128 69 static char CLOOP_MAGIC_START[] = "#!/bin/sh\n"; 70 71 struct cloop_header { 72 char magic[CLOOP_MAGIC_LEN]; /* cloop magic */ 73 uint32_t blksz; /* block size */ 74 uint32_t nblocks; /* number of blocks */ 75 }; 76 77 struct g_uzip_softc { 78 uint32_t blksz; /* block size */ 79 uint32_t nblocks; /* number of blocks */ 80 uint64_t *offsets; 81 82 struct mtx last_mtx; 83 uint32_t last_blk; /* last blk no */ 84 char *last_buf; /* last blk data */ 85 int req_total; /* total requests */ 86 int req_cached; /* cached requests */ 87 }; 88 89 static void 90 g_uzip_softc_free(struct g_uzip_softc *sc, struct g_geom *gp) 91 { 92 if (gp != NULL) { 93 printf("%s: %d requests, %d cached\n", 94 gp->name, sc->req_total, sc->req_cached); 95 } 96 if (sc->offsets != NULL) 97 free(sc->offsets, M_GEOM_UZIP); 98 mtx_destroy(&sc->last_mtx); 99 free(sc->last_buf, M_GEOM_UZIP); 100 free(sc, M_GEOM_UZIP); 101 } 102 103 static void * 104 z_alloc(void *nil, u_int type, u_int size) 105 { 106 void *ptr; 107 108 ptr = malloc(type * size, M_GEOM_UZIP, M_NOWAIT); 109 return ptr; 110 } 111 112 static void 113 z_free(void *nil, void *ptr) 114 { 115 free(ptr, M_GEOM_UZIP); 116 } 117 118 static void 119 g_uzip_done(struct bio *bp) 120 { 121 int err; 122 struct bio *bp2; 123 z_stream zs; 124 struct g_provider *pp, *pp2; 125 struct g_consumer *cp; 126 struct g_geom *gp; 127 struct g_uzip_softc *sc; 128 off_t iolen, pos, upos; 129 uint32_t start_blk, i; 130 size_t bsize; 131 132 bp2 = bp->bio_parent; 133 pp = bp2->bio_to; 134 gp = pp->geom; 135 cp = LIST_FIRST(&gp->consumer); 136 pp2 = cp->provider; 137 sc = gp->softc; 138 DPRINTF(("%s: done\n", gp->name)); 139 140 bp2->bio_error = bp->bio_error; 141 if (bp2->bio_error != 0) 142 goto done; 143 144 /* 145 * Uncompress data. 146 */ 147 zs.zalloc = z_alloc; 148 zs.zfree = z_free; 149 err = inflateInit(&zs); 150 if (err != Z_OK) { 151 bp2->bio_error = EIO; 152 goto done; 153 } 154 start_blk = bp2->bio_offset / sc->blksz; 155 bsize = pp2->sectorsize; 156 iolen = bp->bio_completed; 157 pos = sc->offsets[start_blk] % bsize; 158 upos = 0; 159 DPRINTF(("%s: done: start_blk %d, pos %jd, upos %jd, iolen %jd " 160 "(%jd, %d, %zd)\n", 161 gp->name, start_blk, (intmax_t)pos, (intmax_t)upos, 162 (intmax_t)iolen, (intmax_t)bp2->bio_offset, sc->blksz, bsize)); 163 for (i = start_blk; upos < bp2->bio_length; i++) { 164 off_t len, ulen, uoff; 165 166 uoff = i == start_blk ? bp2->bio_offset % sc->blksz : 0; 167 ulen = MIN(sc->blksz - uoff, bp2->bio_length - upos); 168 len = sc->offsets[i + 1] - sc->offsets[i]; 169 170 if (len == 0) { 171 /* All zero block: no cache update */ 172 bzero(bp2->bio_data + upos, ulen); 173 upos += ulen; 174 bp2->bio_completed += ulen; 175 continue; 176 } 177 if (len > iolen) { 178 DPRINTF(("%s: done: early termination: len (%jd) > " 179 "iolen (%jd)\n", 180 gp->name, (intmax_t)len, (intmax_t)iolen)); 181 break; 182 } 183 zs.next_in = bp->bio_data + pos; 184 zs.avail_in = len; 185 zs.next_out = sc->last_buf; 186 zs.avail_out = sc->blksz; 187 mtx_lock(&sc->last_mtx); 188 err = inflate(&zs, Z_FINISH); 189 if (err != Z_STREAM_END) { 190 sc->last_blk = -1; 191 mtx_unlock(&sc->last_mtx); 192 DPRINTF(("%s: done: inflate failed (%jd + %jd -> %jd + %jd + %jd)\n", 193 gp->name, (intmax_t)pos, (intmax_t)len, 194 (intmax_t)uoff, (intmax_t)upos, (intmax_t)ulen)); 195 inflateEnd(&zs); 196 bp2->bio_error = EIO; 197 goto done; 198 } 199 sc->last_blk = i; 200 DPRINTF(("%s: done: inflated %jd + %jd -> %jd + %jd + %jd\n", 201 gp->name, (intmax_t)pos, (intmax_t)len, (intmax_t)uoff, 202 (intmax_t)upos, (intmax_t)ulen)); 203 memcpy(bp2->bio_data + upos, sc->last_buf + uoff, ulen); 204 mtx_unlock(&sc->last_mtx); 205 206 pos += len; 207 iolen -= len; 208 upos += ulen; 209 bp2->bio_completed += ulen; 210 err = inflateReset(&zs); 211 if (err != Z_OK) { 212 inflateEnd(&zs); 213 bp2->bio_error = EIO; 214 goto done; 215 } 216 } 217 err = inflateEnd(&zs); 218 if (err != Z_OK) { 219 bp2->bio_error = EIO; 220 goto done; 221 } 222 223 done: 224 /* 225 * Finish processing the request. 226 */ 227 DPRINTF(("%s: done: (%d, %jd, %ld)\n", 228 gp->name, bp2->bio_error, (intmax_t)bp2->bio_completed, 229 bp2->bio_resid)); 230 free(bp->bio_data, M_GEOM_UZIP); 231 g_destroy_bio(bp); 232 g_io_deliver(bp2, bp2->bio_error); 233 } 234 235 static void 236 g_uzip_start(struct bio *bp) 237 { 238 struct bio *bp2; 239 struct g_provider *pp, *pp2; 240 struct g_geom *gp; 241 struct g_consumer *cp; 242 struct g_uzip_softc *sc; 243 uint32_t start_blk, end_blk; 244 size_t bsize; 245 246 pp = bp->bio_to; 247 gp = pp->geom; 248 DPRINTF(("%s: start (%d)\n", gp->name, bp->bio_cmd)); 249 250 if (bp->bio_cmd != BIO_READ) { 251 g_io_deliver(bp, EOPNOTSUPP); 252 return; 253 } 254 255 cp = LIST_FIRST(&gp->consumer); 256 pp2 = cp->provider; 257 sc = gp->softc; 258 259 start_blk = bp->bio_offset / sc->blksz; 260 end_blk = (bp->bio_offset + bp->bio_length + sc->blksz - 1) / sc->blksz; 261 KASSERT(start_blk < sc->nblocks, 262 ("start_blk out of range")); 263 KASSERT(end_blk <= sc->nblocks, 264 ("end_blk out of range")); 265 266 sc->req_total++; 267 if (start_blk + 1 == end_blk) { 268 mtx_lock(&sc->last_mtx); 269 if (start_blk == sc->last_blk) { 270 off_t uoff; 271 272 uoff = bp->bio_offset % sc->blksz; 273 KASSERT(bp->bio_length <= sc->blksz - uoff, 274 ("cached data error")); 275 memcpy(bp->bio_data, sc->last_buf + uoff, 276 bp->bio_length); 277 sc->req_cached++; 278 mtx_unlock(&sc->last_mtx); 279 280 DPRINTF(("%s: start: cached 0 + %jd, %jd + 0 + %jd\n", 281 gp->name, (intmax_t)bp->bio_length, (intmax_t)uoff, 282 (intmax_t)bp->bio_length)); 283 bp->bio_completed = bp->bio_length; 284 g_io_deliver(bp, 0); 285 return; 286 } 287 mtx_unlock(&sc->last_mtx); 288 } 289 290 bp2 = g_clone_bio(bp); 291 if (bp2 == NULL) { 292 g_io_deliver(bp, ENOMEM); 293 return; 294 } 295 bp2->bio_done = g_uzip_done; 296 DPRINTF(("%s: start (%d..%d), %s: %d + %jd, %s: %d + %jd\n", 297 gp->name, start_blk, end_blk, 298 pp->name, pp->sectorsize, (intmax_t)pp->mediasize, 299 pp2->name, pp2->sectorsize, (intmax_t)pp2->mediasize)); 300 bsize = pp2->sectorsize; 301 bp2->bio_offset = sc->offsets[start_blk] - sc->offsets[start_blk] % bsize; 302 while (1) { 303 bp2->bio_length = sc->offsets[end_blk] - bp2->bio_offset; 304 bp2->bio_length = (bp2->bio_length + bsize - 1) / bsize * bsize; 305 if (bp2->bio_length < MAXPHYS) 306 break; 307 308 end_blk--; 309 DPRINTF(("%s: bio_length (%jd) > MAXPHYS: lowering end_blk " 310 "to %u\n", gp->name, (intmax_t)bp2->bio_length, end_blk)); 311 } 312 DPRINTF(("%s: start %jd + %jd -> %ju + %ju -> %jd + %jd\n", 313 gp->name, 314 (intmax_t)bp->bio_offset, (intmax_t)bp->bio_length, 315 (uintmax_t)sc->offsets[start_blk], 316 (uintmax_t)sc->offsets[end_blk] - sc->offsets[start_blk], 317 (intmax_t)bp2->bio_offset, (intmax_t)bp2->bio_length)); 318 bp2->bio_data = malloc(bp2->bio_length, M_GEOM_UZIP, M_NOWAIT); 319 if (bp2->bio_data == NULL) { 320 g_destroy_bio(bp2); 321 g_io_deliver(bp, ENOMEM); 322 return; 323 } 324 325 g_io_request(bp2, cp); 326 DPRINTF(("%s: start ok\n", gp->name)); 327 } 328 329 static void 330 g_uzip_orphan(struct g_consumer *cp) 331 { 332 struct g_geom *gp; 333 334 g_trace(G_T_TOPOLOGY, "g_uzip_orphan(%p/%s)", cp, cp->provider->name); 335 g_topology_assert(); 336 337 gp = cp->geom; 338 g_uzip_softc_free(gp->softc, gp); 339 gp->softc = NULL; 340 g_wither_geom(gp, ENXIO); 341 } 342 343 static int 344 g_uzip_access(struct g_provider *pp, int dr, int dw, int de) 345 { 346 struct g_geom *gp; 347 struct g_consumer *cp; 348 349 gp = pp->geom; 350 cp = LIST_FIRST(&gp->consumer); 351 KASSERT (cp != NULL, ("g_uzip_access but no consumer")); 352 353 if (cp->acw + dw > 0) 354 return EROFS; 355 356 return (g_access(cp, dr, dw, de)); 357 } 358 359 static void 360 g_uzip_spoiled(struct g_consumer *cp) 361 { 362 struct g_geom *gp; 363 364 gp = cp->geom; 365 g_trace(G_T_TOPOLOGY, "g_uzip_spoiled(%p/%s)", cp, gp->name); 366 g_topology_assert(); 367 368 g_uzip_softc_free(gp->softc, gp); 369 gp->softc = NULL; 370 g_wither_geom(gp, ENXIO); 371 } 372 373 static struct g_geom * 374 g_uzip_taste(struct g_class *mp, struct g_provider *pp, int flags) 375 { 376 int error; 377 uint32_t i, total_offsets, offsets_read, blk; 378 void *buf; 379 struct cloop_header *header; 380 struct g_consumer *cp; 381 struct g_geom *gp; 382 struct g_provider *pp2; 383 struct g_uzip_softc *sc; 384 385 g_trace(G_T_TOPOLOGY, "g_uzip_taste(%s,%s)", mp->name, pp->name); 386 g_topology_assert(); 387 388 /* Skip providers that are already open for writing. */ 389 if (pp->acw > 0) 390 return (NULL); 391 392 buf = NULL; 393 394 /* 395 * Create geom instance. 396 */ 397 gp = g_new_geomf(mp, "%s.uzip", pp->name); 398 cp = g_new_consumer(gp); 399 error = g_attach(cp, pp); 400 if (error == 0) 401 error = g_access(cp, 1, 0, 0); 402 if (error) { 403 g_detach(cp); 404 g_destroy_consumer(cp); 405 g_destroy_geom(gp); 406 return (NULL); 407 } 408 g_topology_unlock(); 409 410 /* 411 * Read cloop header, look for CLOOP magic, perform 412 * other validity checks. 413 */ 414 DPRINTF(("%s: media sectorsize %u, mediasize %jd\n", 415 gp->name, pp->sectorsize, (intmax_t)pp->mediasize)); 416 buf = g_read_data(cp, 0, pp->sectorsize, NULL); 417 if (buf == NULL) 418 goto err; 419 header = (struct cloop_header *) buf; 420 if (strncmp(header->magic, CLOOP_MAGIC_START, 421 sizeof(CLOOP_MAGIC_START) - 1) != 0) { 422 DPRINTF(("%s: no CLOOP magic\n", gp->name)); 423 goto err; 424 } 425 if (header->magic[0x0b] != 'V' || header->magic[0x0c] < '2') { 426 DPRINTF(("%s: image version too old\n", gp->name)); 427 goto err; 428 } 429 430 /* 431 * Initialize softc and read offsets. 432 */ 433 sc = malloc(sizeof(*sc), M_GEOM_UZIP, M_WAITOK | M_ZERO); 434 gp->softc = sc; 435 sc->blksz = ntohl(header->blksz); 436 sc->nblocks = ntohl(header->nblocks); 437 if (sc->blksz % 512 != 0) { 438 printf("%s: block size (%u) should be multiple of 512.\n", 439 gp->name, sc->blksz); 440 goto err; 441 } 442 if (sc->blksz > MAX_BLKSZ) { 443 printf("%s: block size (%u) should not be larger than %d.\n", 444 gp->name, sc->blksz, MAX_BLKSZ); 445 } 446 total_offsets = sc->nblocks + 1; 447 if (sizeof(struct cloop_header) + 448 total_offsets * sizeof(uint64_t) > pp->mediasize) { 449 printf("%s: media too small for %u blocks\n", 450 gp->name, sc->nblocks); 451 goto err; 452 } 453 sc->offsets = malloc( 454 total_offsets * sizeof(uint64_t), M_GEOM_UZIP, M_WAITOK); 455 offsets_read = MIN(total_offsets, 456 (pp->sectorsize - sizeof(*header)) / sizeof(uint64_t)); 457 for (i = 0; i < offsets_read; i++) 458 sc->offsets[i] = be64toh(((uint64_t *) (header + 1))[i]); 459 DPRINTF(("%s: %u offsets in the first sector\n", 460 gp->name, offsets_read)); 461 for (blk = 1; offsets_read < total_offsets; blk++) { 462 uint32_t nread; 463 464 free(buf, M_GEOM); 465 buf = g_read_data( 466 cp, blk * pp->sectorsize, pp->sectorsize, NULL); 467 if (buf == NULL) 468 goto err; 469 nread = MIN(total_offsets - offsets_read, 470 pp->sectorsize / sizeof(uint64_t)); 471 DPRINTF(("%s: %u offsets read from sector %d\n", 472 gp->name, nread, blk)); 473 for (i = 0; i < nread; i++) { 474 sc->offsets[offsets_read + i] = 475 be64toh(((uint64_t *) buf)[i]); 476 } 477 offsets_read += nread; 478 } 479 DPRINTF(("%s: done reading offsets\n", gp->name)); 480 mtx_init(&sc->last_mtx, "geom_uzip cache", NULL, MTX_DEF); 481 sc->last_blk = -1; 482 sc->last_buf = malloc(sc->blksz, M_GEOM_UZIP, M_WAITOK); 483 sc->req_total = 0; 484 sc->req_cached = 0; 485 486 g_topology_lock(); 487 pp2 = g_new_providerf(gp, "%s", gp->name); 488 pp2->sectorsize = 512; 489 pp2->mediasize = (off_t)sc->nblocks * sc->blksz; 490 pp2->stripesize = pp->stripesize; 491 pp2->stripeoffset = pp->stripeoffset; 492 g_error_provider(pp2, 0); 493 g_access(cp, -1, 0, 0); 494 495 DPRINTF(("%s: taste ok (%d, %jd), (%d, %d), %x\n", 496 gp->name, 497 pp2->sectorsize, (intmax_t)pp2->mediasize, 498 pp2->stripeoffset, pp2->stripesize, pp2->flags)); 499 printf("%s: %u x %u blocks\n", 500 gp->name, sc->nblocks, sc->blksz); 501 return (gp); 502 503 err: 504 g_topology_lock(); 505 g_access(cp, -1, 0, 0); 506 if (buf != NULL) 507 free(buf, M_GEOM); 508 if (gp->softc != NULL) { 509 g_uzip_softc_free(gp->softc, NULL); 510 gp->softc = NULL; 511 } 512 g_detach(cp); 513 g_destroy_consumer(cp); 514 g_destroy_geom(gp); 515 return (NULL); 516 } 517 518 static int 519 g_uzip_destroy_geom(struct gctl_req *req, struct g_class *mp, struct g_geom *gp) 520 { 521 struct g_provider *pp; 522 523 g_trace(G_T_TOPOLOGY, "g_uzip_destroy_geom(%s, %s)", mp->name, gp->name); 524 g_topology_assert(); 525 526 if (gp->softc == NULL) { 527 printf("%s(%s): gp->softc == NULL\n", __func__, gp->name); 528 return (ENXIO); 529 } 530 531 KASSERT(gp != NULL, ("NULL geom")); 532 pp = LIST_FIRST(&gp->provider); 533 KASSERT(pp != NULL, ("NULL provider")); 534 if (pp->acr > 0 || pp->acw > 0 || pp->ace > 0) 535 return (EBUSY); 536 537 g_uzip_softc_free(gp->softc, gp); 538 gp->softc = NULL; 539 g_wither_geom(gp, ENXIO); 540 return (0); 541 } 542 543 static struct g_class g_uzip_class = { 544 .name = UZIP_CLASS_NAME, 545 .version = G_VERSION, 546 .taste = g_uzip_taste, 547 .destroy_geom = g_uzip_destroy_geom, 548 549 .start = g_uzip_start, 550 .orphan = g_uzip_orphan, 551 .access = g_uzip_access, 552 .spoiled = g_uzip_spoiled, 553 }; 554 555 DECLARE_GEOM_CLASS(g_uzip_class, g_uzip); 556 MODULE_DEPEND(g_uzip, zlib, 1, 1, 1); 557