1 /* $NetBSD: udf_strat_direct.c,v 1.9 2009/02/10 17:49:02 reinoud Exp $ */ 2 3 /* 4 * Copyright (c) 2006, 2008 Reinoud Zandijk 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 * 27 */ 28 29 #include <sys/cdefs.h> 30 #ifndef lint 31 __KERNEL_RCSID(0, "$NetBSD: udf_strat_direct.c,v 1.9 2009/02/10 17:49:02 reinoud Exp $"); 32 #endif /* not lint */ 33 34 35 #if defined(_KERNEL_OPT) 36 #include "opt_compat_netbsd.h" 37 #endif 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/sysctl.h> 42 #include <sys/namei.h> 43 #include <sys/proc.h> 44 #include <sys/kernel.h> 45 #include <sys/vnode.h> 46 #include <miscfs/genfs/genfs_node.h> 47 #include <sys/mount.h> 48 #include <sys/buf.h> 49 #include <sys/file.h> 50 #include <sys/device.h> 51 #include <sys/disklabel.h> 52 #include <sys/ioctl.h> 53 #include <sys/malloc.h> 54 #include <sys/dirent.h> 55 #include <sys/stat.h> 56 #include <sys/conf.h> 57 #include <sys/kauth.h> 58 #include <sys/kthread.h> 59 #include <dev/clock_subr.h> 60 61 #include <fs/udf/ecma167-udf.h> 62 #include <fs/udf/udf_mount.h> 63 64 #include "udf.h" 65 #include "udf_subr.h" 66 #include "udf_bswap.h" 67 68 69 #define VTOI(vnode) ((struct udf_node *) vnode->v_data) 70 #define PRIV(ump) ((struct strat_private *) ump->strategy_private) 71 72 /* --------------------------------------------------------------------- */ 73 74 /* BUFQ's */ 75 #define UDF_SHED_MAX 3 76 77 #define UDF_SHED_READING 0 78 #define UDF_SHED_WRITING 1 79 #define UDF_SHED_SEQWRITING 2 80 81 82 struct strat_private { 83 struct pool desc_pool; /* node descriptors */ 84 }; 85 86 /* --------------------------------------------------------------------- */ 87 88 static void 89 udf_wr_nodedscr_callback(struct buf *buf) 90 { 91 struct udf_node *udf_node; 92 93 KASSERT(buf); 94 KASSERT(buf->b_data); 95 96 /* called when write action is done */ 97 DPRINTF(WRITE, ("udf_wr_nodedscr_callback(): node written out\n")); 98 99 udf_node = VTOI(buf->b_vp); 100 if (udf_node == NULL) { 101 putiobuf(buf); 102 printf("udf_wr_node_callback: NULL node?\n"); 103 return; 104 } 105 106 /* XXX right flags to mark dirty again on error? */ 107 if (buf->b_error) { 108 /* write error on `defect free' media??? how to solve? */ 109 /* XXX lookup UDF standard for unallocatable space */ 110 udf_node->i_flags |= IN_MODIFIED | IN_ACCESSED; 111 } 112 113 /* decrement outstanding_nodedscr */ 114 KASSERT(udf_node->outstanding_nodedscr >= 1); 115 udf_node->outstanding_nodedscr--; 116 if (udf_node->outstanding_nodedscr == 0) { 117 /* unlock the node */ 118 KASSERT(udf_node->i_flags & IN_CALLBACK_ULK); 119 UDF_UNLOCK_NODE(udf_node, IN_CALLBACK_ULK); 120 121 wakeup(&udf_node->outstanding_nodedscr); 122 } 123 /* unreference the vnode so it can be recycled */ 124 holdrele(udf_node->vnode); 125 126 putiobuf(buf); 127 } 128 129 /* --------------------------------------------------------------------- */ 130 131 static int 132 udf_getblank_nodedscr_direct(struct udf_strat_args *args) 133 { 134 union dscrptr **dscrptr = &args->dscr; 135 struct udf_mount *ump = args->ump; 136 struct strat_private *priv = PRIV(ump); 137 uint32_t lb_size; 138 139 lb_size = udf_rw32(ump->logical_vol->lb_size); 140 *dscrptr = pool_get(&priv->desc_pool, PR_WAITOK); 141 memset(*dscrptr, 0, lb_size); 142 143 return 0; 144 } 145 146 147 static void 148 udf_free_nodedscr_direct(struct udf_strat_args *args) 149 { 150 union dscrptr *dscr = args->dscr; 151 struct udf_mount *ump = args->ump; 152 struct strat_private *priv = PRIV(ump); 153 154 pool_put(&priv->desc_pool, dscr); 155 } 156 157 158 static int 159 udf_read_nodedscr_direct(struct udf_strat_args *args) 160 { 161 union dscrptr **dscrptr = &args->dscr; 162 union dscrptr *tmpdscr; 163 struct udf_mount *ump = args->ump; 164 struct long_ad *icb = args->icb; 165 struct strat_private *priv = PRIV(ump); 166 uint32_t lb_size; 167 uint32_t sector, dummy; 168 int error; 169 170 lb_size = udf_rw32(ump->logical_vol->lb_size); 171 172 error = udf_translate_vtop(ump, icb, §or, &dummy); 173 if (error) 174 return error; 175 176 /* try to read in fe/efe */ 177 error = udf_read_phys_dscr(ump, sector, M_UDFTEMP, &tmpdscr); 178 if (error) 179 return error; 180 181 *dscrptr = pool_get(&priv->desc_pool, PR_WAITOK); 182 memcpy(*dscrptr, tmpdscr, lb_size); 183 free(tmpdscr, M_UDFTEMP); 184 185 return 0; 186 } 187 188 189 static int 190 udf_write_nodedscr_direct(struct udf_strat_args *args) 191 { 192 struct udf_mount *ump = args->ump; 193 struct udf_node *udf_node = args->udf_node; 194 union dscrptr *dscr = args->dscr; 195 struct long_ad *icb = args->icb; 196 int waitfor = args->waitfor; 197 uint32_t logsector, sector, dummy; 198 int error, vpart; 199 200 /* 201 * we have to decide if we write it out sequential or at its fixed 202 * position by examining the partition its (to be) written on. 203 */ 204 vpart = udf_rw16(udf_node->loc.loc.part_num); 205 logsector = udf_rw32(icb->loc.lb_num); 206 KASSERT(ump->vtop_tp[vpart] != UDF_VTOP_TYPE_VIRT); 207 208 sector = 0; 209 error = udf_translate_vtop(ump, icb, §or, &dummy); 210 if (error) 211 goto out; 212 213 /* add reference to the vnode to prevent recycling */ 214 vhold(udf_node->vnode); 215 216 if (waitfor) { 217 DPRINTF(WRITE, ("udf_write_nodedscr: sync write\n")); 218 219 error = udf_write_phys_dscr_sync(ump, udf_node, UDF_C_NODE, 220 dscr, sector, logsector); 221 } else { 222 DPRINTF(WRITE, ("udf_write_nodedscr: no wait, async write\n")); 223 224 error = udf_write_phys_dscr_async(ump, udf_node, UDF_C_NODE, 225 dscr, sector, logsector, udf_wr_nodedscr_callback); 226 /* will be UNLOCKED in call back */ 227 return error; 228 } 229 230 holdrele(udf_node->vnode); 231 out: 232 udf_node->outstanding_nodedscr--; 233 if (udf_node->outstanding_nodedscr == 0) { 234 UDF_UNLOCK_NODE(udf_node, 0); 235 wakeup(&udf_node->outstanding_nodedscr); 236 } 237 238 return error; 239 } 240 241 /* --------------------------------------------------------------------- */ 242 243 static void 244 udf_queue_buf_direct(struct udf_strat_args *args) 245 { 246 struct udf_mount *ump = args->ump; 247 struct buf *buf = args->nestbuf; 248 struct buf *nestbuf; 249 struct desc_tag *tag; 250 struct long_ad *node_ad_cpy; 251 uint64_t *lmapping, *pmapping, *lmappos, blknr, run_start; 252 uint32_t our_sectornr, sectornr; 253 uint32_t lb_size, buf_offset, rbuflen, bpos; 254 uint16_t vpart_num; 255 uint8_t *fidblk; 256 off_t rblk; 257 int sector_size = ump->discinfo.sector_size; 258 int blks = sector_size / DEV_BSIZE; 259 int len, buf_len, sector, sectors, run_length; 260 int what, class, queue; 261 262 KASSERT(ump); 263 KASSERT(buf); 264 KASSERT(buf->b_iodone == nestiobuf_iodone); 265 266 what = buf->b_udf_c_type; 267 queue = UDF_SHED_READING; 268 if ((buf->b_flags & B_READ) == 0) { 269 /* writing */ 270 queue = UDF_SHED_SEQWRITING; 271 if (what == UDF_C_ABSOLUTE) 272 queue = UDF_SHED_WRITING; 273 if (what == UDF_C_DSCR) 274 queue = UDF_SHED_WRITING; 275 if (what == UDF_C_NODE) 276 queue = UDF_SHED_WRITING; 277 } 278 279 /* use disc sheduler */ 280 class = ump->discinfo.mmc_class; 281 KASSERT((class == MMC_CLASS_UNKN) || (class == MMC_CLASS_DISC) || 282 (ump->discinfo.mmc_cur & MMC_CAP_HW_DEFECTFREE) || 283 (ump->vfs_mountp->mnt_flag & MNT_RDONLY)); 284 285 if (queue == UDF_SHED_READING) { 286 DPRINTF(SHEDULE, ("\nudf_issue_buf READ %p : sector %d type %d," 287 "b_resid %d, b_bcount %d, b_bufsize %d\n", 288 buf, (uint32_t) buf->b_blkno / blks, buf->b_udf_c_type, 289 buf->b_resid, buf->b_bcount, buf->b_bufsize)); 290 VOP_STRATEGY(ump->devvp, buf); 291 return; 292 } 293 294 /* (sectorsize == lb_size) for UDF */ 295 lb_size = udf_rw32(ump->logical_vol->lb_size); 296 blknr = buf->b_blkno; 297 our_sectornr = blknr / blks; 298 299 if (queue == UDF_SHED_WRITING) { 300 DPRINTF(SHEDULE, ("\nudf_issue_buf WRITE %p : sector %d " 301 "type %d, b_resid %d, b_bcount %d, b_bufsize %d\n", 302 buf, (uint32_t) buf->b_blkno / blks, buf->b_udf_c_type, 303 buf->b_resid, buf->b_bcount, buf->b_bufsize)); 304 KASSERT(buf->b_udf_c_type == UDF_C_DSCR || 305 buf->b_udf_c_type == UDF_C_ABSOLUTE || 306 buf->b_udf_c_type == UDF_C_NODE); 307 udf_fixup_node_internals(ump, buf->b_data, buf->b_udf_c_type); 308 VOP_STRATEGY(ump->devvp, buf); 309 return; 310 } 311 312 /* UDF_SHED_SEQWRITING */ 313 KASSERT(queue == UDF_SHED_SEQWRITING); 314 DPRINTF(SHEDULE, ("\nudf_issue_buf SEQWRITE %p : sector XXXX " 315 "type %d, b_resid %d, b_bcount %d, b_bufsize %d\n", 316 buf, buf->b_udf_c_type, buf->b_resid, buf->b_bcount, 317 buf->b_bufsize)); 318 319 /* 320 * Buffers should not have been allocated to disc addresses yet on 321 * this queue. Note that a buffer can get multiple extents allocated. 322 * 323 * lmapping contains lb_num relative to base partition. 324 */ 325 lmapping = ump->la_lmapping; 326 node_ad_cpy = ump->la_node_ad_cpy; 327 328 /* logically allocate buf and map it in the file */ 329 udf_late_allocate_buf(ump, buf, lmapping, node_ad_cpy, &vpart_num); 330 331 /* if we have FIDs, fixup using the new allocation table */ 332 if (buf->b_udf_c_type == UDF_C_FIDS) { 333 buf_len = buf->b_bcount; 334 bpos = 0; 335 lmappos = lmapping; 336 while (buf_len) { 337 sectornr = *lmappos++; 338 len = MIN(buf_len, sector_size); 339 fidblk = (uint8_t *) buf->b_data + bpos; 340 udf_fixup_fid_block(fidblk, sector_size, 341 0, len, sectornr); 342 bpos += len; 343 buf_len -= len; 344 } 345 } 346 if (buf->b_udf_c_type == UDF_C_METADATA_SBM) { 347 if (buf->b_lblkno == 0) { 348 /* update the tag location inside */ 349 tag = (struct desc_tag *) buf->b_data; 350 tag->tag_loc = udf_rw32(*lmapping); 351 udf_validate_tag_and_crc_sums(buf->b_data); 352 } 353 } 354 udf_fixup_node_internals(ump, buf->b_data, buf->b_udf_c_type); 355 356 /* 357 * Translate new mappings in lmapping to pmappings and try to 358 * conglomerate extents to reduce the number of writes. 359 * 360 * pmapping to contain lb_nums as used for disc adressing. 361 */ 362 pmapping = ump->la_pmapping; 363 sectors = (buf->b_bcount + sector_size -1) / sector_size; 364 udf_translate_vtop_list(ump, sectors, vpart_num, lmapping, pmapping); 365 366 for (sector = 0; sector < sectors; sector++) { 367 buf_offset = sector * sector_size; 368 DPRINTF(WRITE, ("\tprocessing rel sector %d\n", sector)); 369 370 DPRINTF(WRITE, ("\tissue write sector %"PRIu64"\n", 371 pmapping[sector])); 372 373 run_start = pmapping[sector]; 374 run_length = 1; 375 while (sector < sectors-1) { 376 if (pmapping[sector+1] != pmapping[sector]+1) 377 break; 378 run_length++; 379 sector++; 380 } 381 382 /* nest an iobuf for the extent */ 383 rbuflen = run_length * sector_size; 384 rblk = run_start * (sector_size/DEV_BSIZE); 385 386 nestbuf = getiobuf(NULL, true); 387 nestiobuf_setup(buf, nestbuf, buf_offset, rbuflen); 388 /* nestbuf is B_ASYNC */ 389 390 /* identify this nestbuf */ 391 nestbuf->b_lblkno = sector; 392 assert(nestbuf->b_vp == buf->b_vp); 393 394 /* CD shedules on raw blkno */ 395 nestbuf->b_blkno = rblk; 396 nestbuf->b_proc = NULL; 397 nestbuf->b_rawblkno = rblk; 398 nestbuf->b_udf_c_type = UDF_C_PROCESSED; 399 400 VOP_STRATEGY(ump->devvp, nestbuf); 401 } 402 } 403 404 405 static void 406 udf_discstrat_init_direct(struct udf_strat_args *args) 407 { 408 struct udf_mount *ump = args->ump; 409 struct strat_private *priv = PRIV(ump); 410 uint32_t lb_size; 411 412 KASSERT(priv == NULL); 413 ump->strategy_private = malloc(sizeof(struct strat_private), 414 M_UDFTEMP, M_WAITOK); 415 priv = ump->strategy_private; 416 memset(priv, 0 , sizeof(struct strat_private)); 417 418 /* 419 * Initialise pool for descriptors associated with nodes. This is done 420 * in lb_size units though currently lb_size is dictated to be 421 * sector_size. 422 */ 423 memset(&priv->desc_pool, 0, sizeof(struct pool)); 424 425 lb_size = udf_rw32(ump->logical_vol->lb_size); 426 pool_init(&priv->desc_pool, lb_size, 0, 0, 0, "udf_desc_pool", NULL, 427 IPL_NONE); 428 } 429 430 431 static void 432 udf_discstrat_finish_direct(struct udf_strat_args *args) 433 { 434 struct udf_mount *ump = args->ump; 435 struct strat_private *priv = PRIV(ump); 436 437 /* destroy our pool */ 438 pool_destroy(&priv->desc_pool); 439 440 /* free our private space */ 441 free(ump->strategy_private, M_UDFTEMP); 442 ump->strategy_private = NULL; 443 } 444 445 /* --------------------------------------------------------------------- */ 446 447 struct udf_strategy udf_strat_direct = 448 { 449 udf_getblank_nodedscr_direct, 450 udf_free_nodedscr_direct, 451 udf_read_nodedscr_direct, 452 udf_write_nodedscr_direct, 453 udf_queue_buf_direct, 454 udf_discstrat_init_direct, 455 udf_discstrat_finish_direct 456 }; 457 458