1 /* 2 * Copyright (c) 2000-2006 Silicon Graphics, Inc. 3 * All Rights Reserved. 4 * 5 * This program is free software; you can redistribute it and/or 6 * modify it under the terms of the GNU General Public License as 7 * published by the Free Software Foundation. 8 * 9 * This program is distributed in the hope that it would be useful, 10 * but WITHOUT ANY WARRANTY; without even the implied warranty of 11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 12 * GNU General Public License for more details. 13 * 14 * You should have received a copy of the GNU General Public License 15 * along with this program; if not, write the Free Software Foundation, 16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA 17 */ 18 19 #include "xfs.h" 20 #include "xfs_log.h" 21 #include "xfs_inum.h" 22 #include "xfs_trans.h" 23 #include "xfs_sb.h" 24 #include "xfs_ag.h" 25 #include "xfs_dir2.h" 26 #include "xfs_alloc.h" 27 #include "xfs_quota.h" 28 #include "xfs_mount.h" 29 #include "xfs_bmap_btree.h" 30 #include "xfs_alloc_btree.h" 31 #include "xfs_ialloc_btree.h" 32 #include "xfs_dinode.h" 33 #include "xfs_inode.h" 34 #include "xfs_btree.h" 35 #include "xfs_ialloc.h" 36 #include "xfs_bmap.h" 37 #include "xfs_rtalloc.h" 38 #include "xfs_error.h" 39 #include "xfs_itable.h" 40 #include "xfs_fsops.h" 41 #include "xfs_attr.h" 42 #include "xfs_buf_item.h" 43 #include "xfs_utils.h" 44 #include "xfs_vnodeops.h" 45 #include "xfs_log_priv.h" 46 #include "xfs_trans_priv.h" 47 #include "xfs_filestream.h" 48 #include "xfs_da_btree.h" 49 #include "xfs_extfree_item.h" 50 #include "xfs_mru_cache.h" 51 #include "xfs_inode_item.h" 52 #include "xfs_icache.h" 53 #include "xfs_trace.h" 54 55 #include <linux/namei.h> 56 #include <linux/init.h> 57 #include <linux/slab.h> 58 #include <linux/mount.h> 59 #include <linux/mempool.h> 60 #include <linux/writeback.h> 61 #include <linux/kthread.h> 62 #include <linux/freezer.h> 63 #include <linux/parser.h> 64 65 static const struct super_operations xfs_super_operations; 66 static kmem_zone_t *xfs_ioend_zone; 67 mempool_t *xfs_ioend_pool; 68 69 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */ 70 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */ 71 #define MNTOPT_LOGDEV "logdev" /* log device */ 72 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */ 73 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */ 74 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */ 75 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */ 76 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */ 77 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */ 78 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */ 79 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */ 80 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */ 81 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */ 82 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */ 83 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */ 84 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */ 85 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */ 86 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */ 87 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and 88 * unwritten extent conversion */ 89 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */ 90 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */ 91 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to 92 * XFS_MAXINUMBER_32 */ 93 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */ 94 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */ 95 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */ 96 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes 97 * in stat(). */ 98 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */ 99 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */ 100 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */ 101 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */ 102 #define MNTOPT_NOQUOTA "noquota" /* no quotas */ 103 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */ 104 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */ 105 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */ 106 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */ 107 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */ 108 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */ 109 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */ 110 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */ 111 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */ 112 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */ 113 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */ 114 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */ 115 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */ 116 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */ 117 118 /* 119 * Table driven mount option parser. 120 * 121 * Currently only used for remount, but it will be used for mount 122 * in the future, too. 123 */ 124 enum { 125 Opt_barrier, 126 Opt_nobarrier, 127 Opt_inode64, 128 Opt_inode32, 129 Opt_err 130 }; 131 132 static const match_table_t tokens = { 133 {Opt_barrier, "barrier"}, 134 {Opt_nobarrier, "nobarrier"}, 135 {Opt_inode64, "inode64"}, 136 {Opt_inode32, "inode32"}, 137 {Opt_err, NULL} 138 }; 139 140 141 STATIC unsigned long 142 suffix_strtoul(char *s, char **endp, unsigned int base) 143 { 144 int last, shift_left_factor = 0; 145 char *value = s; 146 147 last = strlen(value) - 1; 148 if (value[last] == 'K' || value[last] == 'k') { 149 shift_left_factor = 10; 150 value[last] = '\0'; 151 } 152 if (value[last] == 'M' || value[last] == 'm') { 153 shift_left_factor = 20; 154 value[last] = '\0'; 155 } 156 if (value[last] == 'G' || value[last] == 'g') { 157 shift_left_factor = 30; 158 value[last] = '\0'; 159 } 160 161 return simple_strtoul((const char *)s, endp, base) << shift_left_factor; 162 } 163 164 /* 165 * This function fills in xfs_mount_t fields based on mount args. 166 * Note: the superblock has _not_ yet been read in. 167 * 168 * Note that this function leaks the various device name allocations on 169 * failure. The caller takes care of them. 170 */ 171 STATIC int 172 xfs_parseargs( 173 struct xfs_mount *mp, 174 char *options) 175 { 176 struct super_block *sb = mp->m_super; 177 char *this_char, *value, *eov; 178 int dsunit = 0; 179 int dswidth = 0; 180 int iosize = 0; 181 __uint8_t iosizelog = 0; 182 183 /* 184 * set up the mount name first so all the errors will refer to the 185 * correct device. 186 */ 187 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL); 188 if (!mp->m_fsname) 189 return ENOMEM; 190 mp->m_fsname_len = strlen(mp->m_fsname) + 1; 191 192 /* 193 * Copy binary VFS mount flags we are interested in. 194 */ 195 if (sb->s_flags & MS_RDONLY) 196 mp->m_flags |= XFS_MOUNT_RDONLY; 197 if (sb->s_flags & MS_DIRSYNC) 198 mp->m_flags |= XFS_MOUNT_DIRSYNC; 199 if (sb->s_flags & MS_SYNCHRONOUS) 200 mp->m_flags |= XFS_MOUNT_WSYNC; 201 202 /* 203 * Set some default flags that could be cleared by the mount option 204 * parsing. 205 */ 206 mp->m_flags |= XFS_MOUNT_BARRIER; 207 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE; 208 #if !XFS_BIG_INUMS 209 mp->m_flags |= XFS_MOUNT_SMALL_INUMS; 210 #endif 211 212 /* 213 * These can be overridden by the mount option parsing. 214 */ 215 mp->m_logbufs = -1; 216 mp->m_logbsize = -1; 217 218 if (!options) 219 goto done; 220 221 while ((this_char = strsep(&options, ",")) != NULL) { 222 if (!*this_char) 223 continue; 224 if ((value = strchr(this_char, '=')) != NULL) 225 *value++ = 0; 226 227 if (!strcmp(this_char, MNTOPT_LOGBUFS)) { 228 if (!value || !*value) { 229 xfs_warn(mp, "%s option requires an argument", 230 this_char); 231 return EINVAL; 232 } 233 mp->m_logbufs = simple_strtoul(value, &eov, 10); 234 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) { 235 if (!value || !*value) { 236 xfs_warn(mp, "%s option requires an argument", 237 this_char); 238 return EINVAL; 239 } 240 mp->m_logbsize = suffix_strtoul(value, &eov, 10); 241 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) { 242 if (!value || !*value) { 243 xfs_warn(mp, "%s option requires an argument", 244 this_char); 245 return EINVAL; 246 } 247 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL); 248 if (!mp->m_logname) 249 return ENOMEM; 250 } else if (!strcmp(this_char, MNTOPT_MTPT)) { 251 xfs_warn(mp, "%s option not allowed on this system", 252 this_char); 253 return EINVAL; 254 } else if (!strcmp(this_char, MNTOPT_RTDEV)) { 255 if (!value || !*value) { 256 xfs_warn(mp, "%s option requires an argument", 257 this_char); 258 return EINVAL; 259 } 260 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL); 261 if (!mp->m_rtname) 262 return ENOMEM; 263 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) { 264 if (!value || !*value) { 265 xfs_warn(mp, "%s option requires an argument", 266 this_char); 267 return EINVAL; 268 } 269 iosize = simple_strtoul(value, &eov, 10); 270 iosizelog = ffs(iosize) - 1; 271 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) { 272 if (!value || !*value) { 273 xfs_warn(mp, "%s option requires an argument", 274 this_char); 275 return EINVAL; 276 } 277 iosize = suffix_strtoul(value, &eov, 10); 278 iosizelog = ffs(iosize) - 1; 279 } else if (!strcmp(this_char, MNTOPT_GRPID) || 280 !strcmp(this_char, MNTOPT_BSDGROUPS)) { 281 mp->m_flags |= XFS_MOUNT_GRPID; 282 } else if (!strcmp(this_char, MNTOPT_NOGRPID) || 283 !strcmp(this_char, MNTOPT_SYSVGROUPS)) { 284 mp->m_flags &= ~XFS_MOUNT_GRPID; 285 } else if (!strcmp(this_char, MNTOPT_WSYNC)) { 286 mp->m_flags |= XFS_MOUNT_WSYNC; 287 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) { 288 mp->m_flags |= XFS_MOUNT_NORECOVERY; 289 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) { 290 mp->m_flags |= XFS_MOUNT_NOALIGN; 291 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) { 292 mp->m_flags |= XFS_MOUNT_SWALLOC; 293 } else if (!strcmp(this_char, MNTOPT_SUNIT)) { 294 if (!value || !*value) { 295 xfs_warn(mp, "%s option requires an argument", 296 this_char); 297 return EINVAL; 298 } 299 dsunit = simple_strtoul(value, &eov, 10); 300 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) { 301 if (!value || !*value) { 302 xfs_warn(mp, "%s option requires an argument", 303 this_char); 304 return EINVAL; 305 } 306 dswidth = simple_strtoul(value, &eov, 10); 307 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) { 308 mp->m_flags |= XFS_MOUNT_SMALL_INUMS; 309 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) { 310 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; 311 #if !XFS_BIG_INUMS 312 xfs_warn(mp, "%s option not allowed on this system", 313 this_char); 314 return EINVAL; 315 #endif 316 } else if (!strcmp(this_char, MNTOPT_NOUUID)) { 317 mp->m_flags |= XFS_MOUNT_NOUUID; 318 } else if (!strcmp(this_char, MNTOPT_BARRIER)) { 319 mp->m_flags |= XFS_MOUNT_BARRIER; 320 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) { 321 mp->m_flags &= ~XFS_MOUNT_BARRIER; 322 } else if (!strcmp(this_char, MNTOPT_IKEEP)) { 323 mp->m_flags |= XFS_MOUNT_IKEEP; 324 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) { 325 mp->m_flags &= ~XFS_MOUNT_IKEEP; 326 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) { 327 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE; 328 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) { 329 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE; 330 } else if (!strcmp(this_char, MNTOPT_ATTR2)) { 331 mp->m_flags |= XFS_MOUNT_ATTR2; 332 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) { 333 mp->m_flags &= ~XFS_MOUNT_ATTR2; 334 mp->m_flags |= XFS_MOUNT_NOATTR2; 335 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) { 336 mp->m_flags |= XFS_MOUNT_FILESTREAMS; 337 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) { 338 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT; 339 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD; 340 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE; 341 } else if (!strcmp(this_char, MNTOPT_QUOTA) || 342 !strcmp(this_char, MNTOPT_UQUOTA) || 343 !strcmp(this_char, MNTOPT_USRQUOTA)) { 344 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE | 345 XFS_UQUOTA_ENFD); 346 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) || 347 !strcmp(this_char, MNTOPT_UQUOTANOENF)) { 348 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE); 349 mp->m_qflags &= ~XFS_UQUOTA_ENFD; 350 } else if (!strcmp(this_char, MNTOPT_PQUOTA) || 351 !strcmp(this_char, MNTOPT_PRJQUOTA)) { 352 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE | 353 XFS_OQUOTA_ENFD); 354 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) { 355 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE); 356 mp->m_qflags &= ~XFS_OQUOTA_ENFD; 357 } else if (!strcmp(this_char, MNTOPT_GQUOTA) || 358 !strcmp(this_char, MNTOPT_GRPQUOTA)) { 359 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE | 360 XFS_OQUOTA_ENFD); 361 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) { 362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE); 363 mp->m_qflags &= ~XFS_OQUOTA_ENFD; 364 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) { 365 xfs_warn(mp, 366 "delaylog is the default now, option is deprecated."); 367 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) { 368 xfs_warn(mp, 369 "nodelaylog support has been removed, option is deprecated."); 370 } else if (!strcmp(this_char, MNTOPT_DISCARD)) { 371 mp->m_flags |= XFS_MOUNT_DISCARD; 372 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) { 373 mp->m_flags &= ~XFS_MOUNT_DISCARD; 374 } else if (!strcmp(this_char, "ihashsize")) { 375 xfs_warn(mp, 376 "ihashsize no longer used, option is deprecated."); 377 } else if (!strcmp(this_char, "osyncisdsync")) { 378 xfs_warn(mp, 379 "osyncisdsync has no effect, option is deprecated."); 380 } else if (!strcmp(this_char, "osyncisosync")) { 381 xfs_warn(mp, 382 "osyncisosync has no effect, option is deprecated."); 383 } else if (!strcmp(this_char, "irixsgid")) { 384 xfs_warn(mp, 385 "irixsgid is now a sysctl(2) variable, option is deprecated."); 386 } else { 387 xfs_warn(mp, "unknown mount option [%s].", this_char); 388 return EINVAL; 389 } 390 } 391 392 /* 393 * no recovery flag requires a read-only mount 394 */ 395 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) && 396 !(mp->m_flags & XFS_MOUNT_RDONLY)) { 397 xfs_warn(mp, "no-recovery mounts must be read-only."); 398 return EINVAL; 399 } 400 401 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) { 402 xfs_warn(mp, 403 "sunit and swidth options incompatible with the noalign option"); 404 return EINVAL; 405 } 406 407 #ifndef CONFIG_XFS_QUOTA 408 if (XFS_IS_QUOTA_RUNNING(mp)) { 409 xfs_warn(mp, "quota support not available in this kernel."); 410 return EINVAL; 411 } 412 #endif 413 414 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) && 415 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) { 416 xfs_warn(mp, "cannot mount with both project and group quota"); 417 return EINVAL; 418 } 419 420 if ((dsunit && !dswidth) || (!dsunit && dswidth)) { 421 xfs_warn(mp, "sunit and swidth must be specified together"); 422 return EINVAL; 423 } 424 425 if (dsunit && (dswidth % dsunit != 0)) { 426 xfs_warn(mp, 427 "stripe width (%d) must be a multiple of the stripe unit (%d)", 428 dswidth, dsunit); 429 return EINVAL; 430 } 431 432 done: 433 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) { 434 /* 435 * At this point the superblock has not been read 436 * in, therefore we do not know the block size. 437 * Before the mount call ends we will convert 438 * these to FSBs. 439 */ 440 if (dsunit) { 441 mp->m_dalign = dsunit; 442 mp->m_flags |= XFS_MOUNT_RETERR; 443 } 444 445 if (dswidth) 446 mp->m_swidth = dswidth; 447 } 448 449 if (mp->m_logbufs != -1 && 450 mp->m_logbufs != 0 && 451 (mp->m_logbufs < XLOG_MIN_ICLOGS || 452 mp->m_logbufs > XLOG_MAX_ICLOGS)) { 453 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]", 454 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS); 455 return XFS_ERROR(EINVAL); 456 } 457 if (mp->m_logbsize != -1 && 458 mp->m_logbsize != 0 && 459 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE || 460 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE || 461 !is_power_of_2(mp->m_logbsize))) { 462 xfs_warn(mp, 463 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]", 464 mp->m_logbsize); 465 return XFS_ERROR(EINVAL); 466 } 467 468 if (iosizelog) { 469 if (iosizelog > XFS_MAX_IO_LOG || 470 iosizelog < XFS_MIN_IO_LOG) { 471 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]", 472 iosizelog, XFS_MIN_IO_LOG, 473 XFS_MAX_IO_LOG); 474 return XFS_ERROR(EINVAL); 475 } 476 477 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE; 478 mp->m_readio_log = iosizelog; 479 mp->m_writeio_log = iosizelog; 480 } 481 482 return 0; 483 } 484 485 struct proc_xfs_info { 486 int flag; 487 char *str; 488 }; 489 490 STATIC int 491 xfs_showargs( 492 struct xfs_mount *mp, 493 struct seq_file *m) 494 { 495 static struct proc_xfs_info xfs_info_set[] = { 496 /* the few simple ones we can get from the mount struct */ 497 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP }, 498 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC }, 499 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN }, 500 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC }, 501 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID }, 502 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY }, 503 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 }, 504 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM }, 505 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID }, 506 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD }, 507 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE }, 508 { 0, NULL } 509 }; 510 static struct proc_xfs_info xfs_info_unset[] = { 511 /* the few simple ones we can get from the mount struct */ 512 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO }, 513 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER }, 514 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE }, 515 { 0, NULL } 516 }; 517 struct proc_xfs_info *xfs_infop; 518 519 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) { 520 if (mp->m_flags & xfs_infop->flag) 521 seq_puts(m, xfs_infop->str); 522 } 523 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) { 524 if (!(mp->m_flags & xfs_infop->flag)) 525 seq_puts(m, xfs_infop->str); 526 } 527 528 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) 529 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk", 530 (int)(1 << mp->m_writeio_log) >> 10); 531 532 if (mp->m_logbufs > 0) 533 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs); 534 if (mp->m_logbsize > 0) 535 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10); 536 537 if (mp->m_logname) 538 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname); 539 if (mp->m_rtname) 540 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname); 541 542 if (mp->m_dalign > 0) 543 seq_printf(m, "," MNTOPT_SUNIT "=%d", 544 (int)XFS_FSB_TO_BB(mp, mp->m_dalign)); 545 if (mp->m_swidth > 0) 546 seq_printf(m, "," MNTOPT_SWIDTH "=%d", 547 (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); 548 549 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD)) 550 seq_puts(m, "," MNTOPT_USRQUOTA); 551 else if (mp->m_qflags & XFS_UQUOTA_ACCT) 552 seq_puts(m, "," MNTOPT_UQUOTANOENF); 553 554 /* Either project or group quotas can be active, not both */ 555 556 if (mp->m_qflags & XFS_PQUOTA_ACCT) { 557 if (mp->m_qflags & XFS_OQUOTA_ENFD) 558 seq_puts(m, "," MNTOPT_PRJQUOTA); 559 else 560 seq_puts(m, "," MNTOPT_PQUOTANOENF); 561 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) { 562 if (mp->m_qflags & XFS_OQUOTA_ENFD) 563 seq_puts(m, "," MNTOPT_GRPQUOTA); 564 else 565 seq_puts(m, "," MNTOPT_GQUOTANOENF); 566 } 567 568 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT)) 569 seq_puts(m, "," MNTOPT_NOQUOTA); 570 571 return 0; 572 } 573 __uint64_t 574 xfs_max_file_offset( 575 unsigned int blockshift) 576 { 577 unsigned int pagefactor = 1; 578 unsigned int bitshift = BITS_PER_LONG - 1; 579 580 /* Figure out maximum filesize, on Linux this can depend on 581 * the filesystem blocksize (on 32 bit platforms). 582 * __block_write_begin does this in an [unsigned] long... 583 * page->index << (PAGE_CACHE_SHIFT - bbits) 584 * So, for page sized blocks (4K on 32 bit platforms), 585 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is 586 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) 587 * but for smaller blocksizes it is less (bbits = log2 bsize). 588 * Note1: get_block_t takes a long (implicit cast from above) 589 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch 590 * can optionally convert the [unsigned] long from above into 591 * an [unsigned] long long. 592 */ 593 594 #if BITS_PER_LONG == 32 595 # if defined(CONFIG_LBDAF) 596 ASSERT(sizeof(sector_t) == 8); 597 pagefactor = PAGE_CACHE_SIZE; 598 bitshift = BITS_PER_LONG; 599 # else 600 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); 601 # endif 602 #endif 603 604 return (((__uint64_t)pagefactor) << bitshift) - 1; 605 } 606 607 xfs_agnumber_t 608 xfs_set_inode32(struct xfs_mount *mp) 609 { 610 xfs_agnumber_t index = 0; 611 xfs_agnumber_t maxagi = 0; 612 xfs_sb_t *sbp = &mp->m_sb; 613 xfs_agnumber_t max_metadata; 614 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0); 615 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino); 616 xfs_perag_t *pag; 617 618 /* Calculate how much should be reserved for inodes to meet 619 * the max inode percentage. 620 */ 621 if (mp->m_maxicount) { 622 __uint64_t icount; 623 624 icount = sbp->sb_dblocks * sbp->sb_imax_pct; 625 do_div(icount, 100); 626 icount += sbp->sb_agblocks - 1; 627 do_div(icount, sbp->sb_agblocks); 628 max_metadata = icount; 629 } else { 630 max_metadata = sbp->sb_agcount; 631 } 632 633 for (index = 0; index < sbp->sb_agcount; index++) { 634 ino = XFS_AGINO_TO_INO(mp, index, agino); 635 636 if (ino > XFS_MAXINUMBER_32) { 637 pag = xfs_perag_get(mp, index); 638 pag->pagi_inodeok = 0; 639 pag->pagf_metadata = 0; 640 xfs_perag_put(pag); 641 continue; 642 } 643 644 pag = xfs_perag_get(mp, index); 645 pag->pagi_inodeok = 1; 646 maxagi++; 647 if (index < max_metadata) 648 pag->pagf_metadata = 1; 649 xfs_perag_put(pag); 650 } 651 mp->m_flags |= (XFS_MOUNT_32BITINODES | 652 XFS_MOUNT_SMALL_INUMS); 653 654 return maxagi; 655 } 656 657 xfs_agnumber_t 658 xfs_set_inode64(struct xfs_mount *mp) 659 { 660 xfs_agnumber_t index = 0; 661 662 for (index = 0; index < mp->m_sb.sb_agcount; index++) { 663 struct xfs_perag *pag; 664 665 pag = xfs_perag_get(mp, index); 666 pag->pagi_inodeok = 1; 667 pag->pagf_metadata = 0; 668 xfs_perag_put(pag); 669 } 670 671 /* There is no need for lock protection on m_flags, 672 * the rw_semaphore of the VFS superblock is locked 673 * during mount/umount/remount operations, so this is 674 * enough to avoid concurency on the m_flags field 675 */ 676 mp->m_flags &= ~(XFS_MOUNT_32BITINODES | 677 XFS_MOUNT_SMALL_INUMS); 678 return index; 679 } 680 681 STATIC int 682 xfs_blkdev_get( 683 xfs_mount_t *mp, 684 const char *name, 685 struct block_device **bdevp) 686 { 687 int error = 0; 688 689 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL, 690 mp); 691 if (IS_ERR(*bdevp)) { 692 error = PTR_ERR(*bdevp); 693 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error); 694 } 695 696 return -error; 697 } 698 699 STATIC void 700 xfs_blkdev_put( 701 struct block_device *bdev) 702 { 703 if (bdev) 704 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 705 } 706 707 void 708 xfs_blkdev_issue_flush( 709 xfs_buftarg_t *buftarg) 710 { 711 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL); 712 } 713 714 STATIC void 715 xfs_close_devices( 716 struct xfs_mount *mp) 717 { 718 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { 719 struct block_device *logdev = mp->m_logdev_targp->bt_bdev; 720 xfs_free_buftarg(mp, mp->m_logdev_targp); 721 xfs_blkdev_put(logdev); 722 } 723 if (mp->m_rtdev_targp) { 724 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev; 725 xfs_free_buftarg(mp, mp->m_rtdev_targp); 726 xfs_blkdev_put(rtdev); 727 } 728 xfs_free_buftarg(mp, mp->m_ddev_targp); 729 } 730 731 /* 732 * The file system configurations are: 733 * (1) device (partition) with data and internal log 734 * (2) logical volume with data and log subvolumes. 735 * (3) logical volume with data, log, and realtime subvolumes. 736 * 737 * We only have to handle opening the log and realtime volumes here if 738 * they are present. The data subvolume has already been opened by 739 * get_sb_bdev() and is stored in sb->s_bdev. 740 */ 741 STATIC int 742 xfs_open_devices( 743 struct xfs_mount *mp) 744 { 745 struct block_device *ddev = mp->m_super->s_bdev; 746 struct block_device *logdev = NULL, *rtdev = NULL; 747 int error; 748 749 /* 750 * Open real time and log devices - order is important. 751 */ 752 if (mp->m_logname) { 753 error = xfs_blkdev_get(mp, mp->m_logname, &logdev); 754 if (error) 755 goto out; 756 } 757 758 if (mp->m_rtname) { 759 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev); 760 if (error) 761 goto out_close_logdev; 762 763 if (rtdev == ddev || rtdev == logdev) { 764 xfs_warn(mp, 765 "Cannot mount filesystem with identical rtdev and ddev/logdev."); 766 error = EINVAL; 767 goto out_close_rtdev; 768 } 769 } 770 771 /* 772 * Setup xfs_mount buffer target pointers 773 */ 774 error = ENOMEM; 775 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname); 776 if (!mp->m_ddev_targp) 777 goto out_close_rtdev; 778 779 if (rtdev) { 780 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1, 781 mp->m_fsname); 782 if (!mp->m_rtdev_targp) 783 goto out_free_ddev_targ; 784 } 785 786 if (logdev && logdev != ddev) { 787 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1, 788 mp->m_fsname); 789 if (!mp->m_logdev_targp) 790 goto out_free_rtdev_targ; 791 } else { 792 mp->m_logdev_targp = mp->m_ddev_targp; 793 } 794 795 return 0; 796 797 out_free_rtdev_targ: 798 if (mp->m_rtdev_targp) 799 xfs_free_buftarg(mp, mp->m_rtdev_targp); 800 out_free_ddev_targ: 801 xfs_free_buftarg(mp, mp->m_ddev_targp); 802 out_close_rtdev: 803 if (rtdev) 804 xfs_blkdev_put(rtdev); 805 out_close_logdev: 806 if (logdev && logdev != ddev) 807 xfs_blkdev_put(logdev); 808 out: 809 return error; 810 } 811 812 /* 813 * Setup xfs_mount buffer target pointers based on superblock 814 */ 815 STATIC int 816 xfs_setup_devices( 817 struct xfs_mount *mp) 818 { 819 int error; 820 821 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize, 822 mp->m_sb.sb_sectsize); 823 if (error) 824 return error; 825 826 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { 827 unsigned int log_sector_size = BBSIZE; 828 829 if (xfs_sb_version_hassector(&mp->m_sb)) 830 log_sector_size = mp->m_sb.sb_logsectsize; 831 error = xfs_setsize_buftarg(mp->m_logdev_targp, 832 mp->m_sb.sb_blocksize, 833 log_sector_size); 834 if (error) 835 return error; 836 } 837 if (mp->m_rtdev_targp) { 838 error = xfs_setsize_buftarg(mp->m_rtdev_targp, 839 mp->m_sb.sb_blocksize, 840 mp->m_sb.sb_sectsize); 841 if (error) 842 return error; 843 } 844 845 return 0; 846 } 847 848 STATIC int 849 xfs_init_mount_workqueues( 850 struct xfs_mount *mp) 851 { 852 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s", 853 WQ_MEM_RECLAIM, 0, mp->m_fsname); 854 if (!mp->m_data_workqueue) 855 goto out; 856 857 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s", 858 WQ_MEM_RECLAIM, 0, mp->m_fsname); 859 if (!mp->m_unwritten_workqueue) 860 goto out_destroy_data_iodone_queue; 861 862 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s", 863 WQ_MEM_RECLAIM, 0, mp->m_fsname); 864 if (!mp->m_cil_workqueue) 865 goto out_destroy_unwritten; 866 867 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s", 868 WQ_NON_REENTRANT, 0, mp->m_fsname); 869 if (!mp->m_reclaim_workqueue) 870 goto out_destroy_cil; 871 872 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s", 873 WQ_NON_REENTRANT, 0, mp->m_fsname); 874 if (!mp->m_log_workqueue) 875 goto out_destroy_reclaim; 876 877 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s", 878 WQ_NON_REENTRANT, 0, mp->m_fsname); 879 if (!mp->m_eofblocks_workqueue) 880 goto out_destroy_log; 881 882 return 0; 883 884 out_destroy_log: 885 destroy_workqueue(mp->m_log_workqueue); 886 out_destroy_reclaim: 887 destroy_workqueue(mp->m_reclaim_workqueue); 888 out_destroy_cil: 889 destroy_workqueue(mp->m_cil_workqueue); 890 out_destroy_unwritten: 891 destroy_workqueue(mp->m_unwritten_workqueue); 892 out_destroy_data_iodone_queue: 893 destroy_workqueue(mp->m_data_workqueue); 894 out: 895 return -ENOMEM; 896 } 897 898 STATIC void 899 xfs_destroy_mount_workqueues( 900 struct xfs_mount *mp) 901 { 902 destroy_workqueue(mp->m_eofblocks_workqueue); 903 destroy_workqueue(mp->m_log_workqueue); 904 destroy_workqueue(mp->m_reclaim_workqueue); 905 destroy_workqueue(mp->m_cil_workqueue); 906 destroy_workqueue(mp->m_data_workqueue); 907 destroy_workqueue(mp->m_unwritten_workqueue); 908 } 909 910 /* 911 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK 912 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting 913 * for IO to complete so that we effectively throttle multiple callers to the 914 * rate at which IO is completing. 915 */ 916 void 917 xfs_flush_inodes( 918 struct xfs_mount *mp) 919 { 920 struct super_block *sb = mp->m_super; 921 922 if (down_read_trylock(&sb->s_umount)) { 923 sync_inodes_sb(sb); 924 up_read(&sb->s_umount); 925 } 926 } 927 928 /* Catch misguided souls that try to use this interface on XFS */ 929 STATIC struct inode * 930 xfs_fs_alloc_inode( 931 struct super_block *sb) 932 { 933 BUG(); 934 return NULL; 935 } 936 937 /* 938 * Now that the generic code is guaranteed not to be accessing 939 * the linux inode, we can reclaim the inode. 940 */ 941 STATIC void 942 xfs_fs_destroy_inode( 943 struct inode *inode) 944 { 945 struct xfs_inode *ip = XFS_I(inode); 946 947 trace_xfs_destroy_inode(ip); 948 949 XFS_STATS_INC(vn_reclaim); 950 951 /* bad inode, get out here ASAP */ 952 if (is_bad_inode(inode)) 953 goto out_reclaim; 954 955 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0); 956 957 /* 958 * We should never get here with one of the reclaim flags already set. 959 */ 960 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE)); 961 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM)); 962 963 /* 964 * We always use background reclaim here because even if the 965 * inode is clean, it still may be under IO and hence we have 966 * to take the flush lock. The background reclaim path handles 967 * this more efficiently than we can here, so simply let background 968 * reclaim tear down all inodes. 969 */ 970 out_reclaim: 971 xfs_inode_set_reclaim_tag(ip); 972 } 973 974 /* 975 * Slab object creation initialisation for the XFS inode. 976 * This covers only the idempotent fields in the XFS inode; 977 * all other fields need to be initialised on allocation 978 * from the slab. This avoids the need to repeatedly initialise 979 * fields in the xfs inode that left in the initialise state 980 * when freeing the inode. 981 */ 982 STATIC void 983 xfs_fs_inode_init_once( 984 void *inode) 985 { 986 struct xfs_inode *ip = inode; 987 988 memset(ip, 0, sizeof(struct xfs_inode)); 989 990 /* vfs inode */ 991 inode_init_once(VFS_I(ip)); 992 993 /* xfs inode */ 994 atomic_set(&ip->i_pincount, 0); 995 spin_lock_init(&ip->i_flags_lock); 996 997 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, 998 "xfsino", ip->i_ino); 999 } 1000 1001 STATIC void 1002 xfs_fs_evict_inode( 1003 struct inode *inode) 1004 { 1005 xfs_inode_t *ip = XFS_I(inode); 1006 1007 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock)); 1008 1009 trace_xfs_evict_inode(ip); 1010 1011 truncate_inode_pages(&inode->i_data, 0); 1012 clear_inode(inode); 1013 XFS_STATS_INC(vn_rele); 1014 XFS_STATS_INC(vn_remove); 1015 XFS_STATS_DEC(vn_active); 1016 1017 xfs_inactive(ip); 1018 } 1019 1020 /* 1021 * We do an unlocked check for XFS_IDONTCACHE here because we are already 1022 * serialised against cache hits here via the inode->i_lock and igrab() in 1023 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be 1024 * racing with us, and it avoids needing to grab a spinlock here for every inode 1025 * we drop the final reference on. 1026 */ 1027 STATIC int 1028 xfs_fs_drop_inode( 1029 struct inode *inode) 1030 { 1031 struct xfs_inode *ip = XFS_I(inode); 1032 1033 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE); 1034 } 1035 1036 STATIC void 1037 xfs_free_fsname( 1038 struct xfs_mount *mp) 1039 { 1040 kfree(mp->m_fsname); 1041 kfree(mp->m_rtname); 1042 kfree(mp->m_logname); 1043 } 1044 1045 STATIC void 1046 xfs_fs_put_super( 1047 struct super_block *sb) 1048 { 1049 struct xfs_mount *mp = XFS_M(sb); 1050 1051 xfs_filestream_unmount(mp); 1052 xfs_unmountfs(mp); 1053 1054 xfs_freesb(mp); 1055 xfs_icsb_destroy_counters(mp); 1056 xfs_destroy_mount_workqueues(mp); 1057 xfs_close_devices(mp); 1058 xfs_free_fsname(mp); 1059 kfree(mp); 1060 } 1061 1062 STATIC int 1063 xfs_fs_sync_fs( 1064 struct super_block *sb, 1065 int wait) 1066 { 1067 struct xfs_mount *mp = XFS_M(sb); 1068 1069 /* 1070 * Doing anything during the async pass would be counterproductive. 1071 */ 1072 if (!wait) 1073 return 0; 1074 1075 xfs_log_force(mp, XFS_LOG_SYNC); 1076 if (laptop_mode) { 1077 /* 1078 * The disk must be active because we're syncing. 1079 * We schedule log work now (now that the disk is 1080 * active) instead of later (when it might not be). 1081 */ 1082 flush_delayed_work(&mp->m_log->l_work); 1083 } 1084 1085 return 0; 1086 } 1087 1088 STATIC int 1089 xfs_fs_statfs( 1090 struct dentry *dentry, 1091 struct kstatfs *statp) 1092 { 1093 struct xfs_mount *mp = XFS_M(dentry->d_sb); 1094 xfs_sb_t *sbp = &mp->m_sb; 1095 struct xfs_inode *ip = XFS_I(dentry->d_inode); 1096 __uint64_t fakeinos, id; 1097 xfs_extlen_t lsize; 1098 __int64_t ffree; 1099 1100 statp->f_type = XFS_SB_MAGIC; 1101 statp->f_namelen = MAXNAMELEN - 1; 1102 1103 id = huge_encode_dev(mp->m_ddev_targp->bt_dev); 1104 statp->f_fsid.val[0] = (u32)id; 1105 statp->f_fsid.val[1] = (u32)(id >> 32); 1106 1107 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); 1108 1109 spin_lock(&mp->m_sb_lock); 1110 statp->f_bsize = sbp->sb_blocksize; 1111 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; 1112 statp->f_blocks = sbp->sb_dblocks - lsize; 1113 statp->f_bfree = statp->f_bavail = 1114 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); 1115 fakeinos = statp->f_bfree << sbp->sb_inopblog; 1116 statp->f_files = 1117 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER); 1118 if (mp->m_maxicount) 1119 statp->f_files = min_t(typeof(statp->f_files), 1120 statp->f_files, 1121 mp->m_maxicount); 1122 1123 /* make sure statp->f_ffree does not underflow */ 1124 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree); 1125 statp->f_ffree = max_t(__int64_t, ffree, 0); 1126 1127 spin_unlock(&mp->m_sb_lock); 1128 1129 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && 1130 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) == 1131 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD)) 1132 xfs_qm_statvfs(ip, statp); 1133 return 0; 1134 } 1135 1136 STATIC void 1137 xfs_save_resvblks(struct xfs_mount *mp) 1138 { 1139 __uint64_t resblks = 0; 1140 1141 mp->m_resblks_save = mp->m_resblks; 1142 xfs_reserve_blocks(mp, &resblks, NULL); 1143 } 1144 1145 STATIC void 1146 xfs_restore_resvblks(struct xfs_mount *mp) 1147 { 1148 __uint64_t resblks; 1149 1150 if (mp->m_resblks_save) { 1151 resblks = mp->m_resblks_save; 1152 mp->m_resblks_save = 0; 1153 } else 1154 resblks = xfs_default_resblks(mp); 1155 1156 xfs_reserve_blocks(mp, &resblks, NULL); 1157 } 1158 1159 /* 1160 * Trigger writeback of all the dirty metadata in the file system. 1161 * 1162 * This ensures that the metadata is written to their location on disk rather 1163 * than just existing in transactions in the log. This means after a quiesce 1164 * there is no log replay required to write the inodes to disk - this is the 1165 * primary difference between a sync and a quiesce. 1166 * 1167 * Note: xfs_log_quiesce() stops background log work - the callers must ensure 1168 * it is started again when appropriate. 1169 */ 1170 void 1171 xfs_quiesce_attr( 1172 struct xfs_mount *mp) 1173 { 1174 int error = 0; 1175 1176 /* wait for all modifications to complete */ 1177 while (atomic_read(&mp->m_active_trans) > 0) 1178 delay(100); 1179 1180 /* force the log to unpin objects from the now complete transactions */ 1181 xfs_log_force(mp, XFS_LOG_SYNC); 1182 1183 /* reclaim inodes to do any IO before the freeze completes */ 1184 xfs_reclaim_inodes(mp, 0); 1185 xfs_reclaim_inodes(mp, SYNC_WAIT); 1186 1187 /* Push the superblock and write an unmount record */ 1188 error = xfs_log_sbcount(mp); 1189 if (error) 1190 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. " 1191 "Frozen image may not be consistent."); 1192 /* 1193 * Just warn here till VFS can correctly support 1194 * read-only remount without racing. 1195 */ 1196 WARN_ON(atomic_read(&mp->m_active_trans) != 0); 1197 1198 xfs_log_quiesce(mp); 1199 } 1200 1201 STATIC int 1202 xfs_fs_remount( 1203 struct super_block *sb, 1204 int *flags, 1205 char *options) 1206 { 1207 struct xfs_mount *mp = XFS_M(sb); 1208 substring_t args[MAX_OPT_ARGS]; 1209 char *p; 1210 int error; 1211 1212 while ((p = strsep(&options, ",")) != NULL) { 1213 int token; 1214 1215 if (!*p) 1216 continue; 1217 1218 token = match_token(p, tokens, args); 1219 switch (token) { 1220 case Opt_barrier: 1221 mp->m_flags |= XFS_MOUNT_BARRIER; 1222 break; 1223 case Opt_nobarrier: 1224 mp->m_flags &= ~XFS_MOUNT_BARRIER; 1225 break; 1226 case Opt_inode64: 1227 mp->m_maxagi = xfs_set_inode64(mp); 1228 break; 1229 case Opt_inode32: 1230 mp->m_maxagi = xfs_set_inode32(mp); 1231 break; 1232 default: 1233 /* 1234 * Logically we would return an error here to prevent 1235 * users from believing they might have changed 1236 * mount options using remount which can't be changed. 1237 * 1238 * But unfortunately mount(8) adds all options from 1239 * mtab and fstab to the mount arguments in some cases 1240 * so we can't blindly reject options, but have to 1241 * check for each specified option if it actually 1242 * differs from the currently set option and only 1243 * reject it if that's the case. 1244 * 1245 * Until that is implemented we return success for 1246 * every remount request, and silently ignore all 1247 * options that we can't actually change. 1248 */ 1249 #if 0 1250 xfs_info(mp, 1251 "mount option \"%s\" not supported for remount\n", p); 1252 return -EINVAL; 1253 #else 1254 break; 1255 #endif 1256 } 1257 } 1258 1259 /* ro -> rw */ 1260 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) { 1261 mp->m_flags &= ~XFS_MOUNT_RDONLY; 1262 1263 /* 1264 * If this is the first remount to writeable state we 1265 * might have some superblock changes to update. 1266 */ 1267 if (mp->m_update_flags) { 1268 error = xfs_mount_log_sb(mp, mp->m_update_flags); 1269 if (error) { 1270 xfs_warn(mp, "failed to write sb changes"); 1271 return error; 1272 } 1273 mp->m_update_flags = 0; 1274 } 1275 1276 /* 1277 * Fill out the reserve pool if it is empty. Use the stashed 1278 * value if it is non-zero, otherwise go with the default. 1279 */ 1280 xfs_restore_resvblks(mp); 1281 xfs_log_work_queue(mp); 1282 } 1283 1284 /* rw -> ro */ 1285 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) { 1286 /* 1287 * Before we sync the metadata, we need to free up the reserve 1288 * block pool so that the used block count in the superblock on 1289 * disk is correct at the end of the remount. Stash the current 1290 * reserve pool size so that if we get remounted rw, we can 1291 * return it to the same size. 1292 */ 1293 xfs_save_resvblks(mp); 1294 xfs_quiesce_attr(mp); 1295 mp->m_flags |= XFS_MOUNT_RDONLY; 1296 } 1297 1298 return 0; 1299 } 1300 1301 /* 1302 * Second stage of a freeze. The data is already frozen so we only 1303 * need to take care of the metadata. Once that's done write a dummy 1304 * record to dirty the log in case of a crash while frozen. 1305 */ 1306 STATIC int 1307 xfs_fs_freeze( 1308 struct super_block *sb) 1309 { 1310 struct xfs_mount *mp = XFS_M(sb); 1311 1312 xfs_save_resvblks(mp); 1313 xfs_quiesce_attr(mp); 1314 return -xfs_fs_log_dummy(mp); 1315 } 1316 1317 STATIC int 1318 xfs_fs_unfreeze( 1319 struct super_block *sb) 1320 { 1321 struct xfs_mount *mp = XFS_M(sb); 1322 1323 xfs_restore_resvblks(mp); 1324 xfs_log_work_queue(mp); 1325 return 0; 1326 } 1327 1328 STATIC int 1329 xfs_fs_show_options( 1330 struct seq_file *m, 1331 struct dentry *root) 1332 { 1333 return -xfs_showargs(XFS_M(root->d_sb), m); 1334 } 1335 1336 /* 1337 * This function fills in xfs_mount_t fields based on mount args. 1338 * Note: the superblock _has_ now been read in. 1339 */ 1340 STATIC int 1341 xfs_finish_flags( 1342 struct xfs_mount *mp) 1343 { 1344 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY); 1345 1346 /* Fail a mount where the logbuf is smaller than the log stripe */ 1347 if (xfs_sb_version_haslogv2(&mp->m_sb)) { 1348 if (mp->m_logbsize <= 0 && 1349 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) { 1350 mp->m_logbsize = mp->m_sb.sb_logsunit; 1351 } else if (mp->m_logbsize > 0 && 1352 mp->m_logbsize < mp->m_sb.sb_logsunit) { 1353 xfs_warn(mp, 1354 "logbuf size must be greater than or equal to log stripe size"); 1355 return XFS_ERROR(EINVAL); 1356 } 1357 } else { 1358 /* Fail a mount if the logbuf is larger than 32K */ 1359 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) { 1360 xfs_warn(mp, 1361 "logbuf size for version 1 logs must be 16K or 32K"); 1362 return XFS_ERROR(EINVAL); 1363 } 1364 } 1365 1366 /* 1367 * mkfs'ed attr2 will turn on attr2 mount unless explicitly 1368 * told by noattr2 to turn it off 1369 */ 1370 if (xfs_sb_version_hasattr2(&mp->m_sb) && 1371 !(mp->m_flags & XFS_MOUNT_NOATTR2)) 1372 mp->m_flags |= XFS_MOUNT_ATTR2; 1373 1374 /* 1375 * prohibit r/w mounts of read-only filesystems 1376 */ 1377 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) { 1378 xfs_warn(mp, 1379 "cannot mount a read-only filesystem as read-write"); 1380 return XFS_ERROR(EROFS); 1381 } 1382 1383 return 0; 1384 } 1385 1386 STATIC int 1387 xfs_fs_fill_super( 1388 struct super_block *sb, 1389 void *data, 1390 int silent) 1391 { 1392 struct inode *root; 1393 struct xfs_mount *mp = NULL; 1394 int flags = 0, error = ENOMEM; 1395 1396 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL); 1397 if (!mp) 1398 goto out; 1399 1400 spin_lock_init(&mp->m_sb_lock); 1401 mutex_init(&mp->m_growlock); 1402 atomic_set(&mp->m_active_trans, 0); 1403 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); 1404 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker); 1405 1406 mp->m_super = sb; 1407 sb->s_fs_info = mp; 1408 1409 error = xfs_parseargs(mp, (char *)data); 1410 if (error) 1411 goto out_free_fsname; 1412 1413 sb_min_blocksize(sb, BBSIZE); 1414 sb->s_xattr = xfs_xattr_handlers; 1415 sb->s_export_op = &xfs_export_operations; 1416 #ifdef CONFIG_XFS_QUOTA 1417 sb->s_qcop = &xfs_quotactl_operations; 1418 #endif 1419 sb->s_op = &xfs_super_operations; 1420 1421 if (silent) 1422 flags |= XFS_MFSI_QUIET; 1423 1424 error = xfs_open_devices(mp); 1425 if (error) 1426 goto out_free_fsname; 1427 1428 error = xfs_init_mount_workqueues(mp); 1429 if (error) 1430 goto out_close_devices; 1431 1432 error = xfs_icsb_init_counters(mp); 1433 if (error) 1434 goto out_destroy_workqueues; 1435 1436 error = xfs_readsb(mp, flags); 1437 if (error) 1438 goto out_destroy_counters; 1439 1440 error = xfs_finish_flags(mp); 1441 if (error) 1442 goto out_free_sb; 1443 1444 error = xfs_setup_devices(mp); 1445 if (error) 1446 goto out_free_sb; 1447 1448 error = xfs_filestream_mount(mp); 1449 if (error) 1450 goto out_free_sb; 1451 1452 /* 1453 * we must configure the block size in the superblock before we run the 1454 * full mount process as the mount process can lookup and cache inodes. 1455 */ 1456 sb->s_magic = XFS_SB_MAGIC; 1457 sb->s_blocksize = mp->m_sb.sb_blocksize; 1458 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1; 1459 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); 1460 sb->s_max_links = XFS_MAXLINK; 1461 sb->s_time_gran = 1; 1462 set_posix_acl_flag(sb); 1463 1464 error = xfs_mountfs(mp); 1465 if (error) 1466 goto out_filestream_unmount; 1467 1468 root = igrab(VFS_I(mp->m_rootip)); 1469 if (!root) { 1470 error = ENOENT; 1471 goto out_unmount; 1472 } 1473 if (is_bad_inode(root)) { 1474 error = EINVAL; 1475 goto out_unmount; 1476 } 1477 sb->s_root = d_make_root(root); 1478 if (!sb->s_root) { 1479 error = ENOMEM; 1480 goto out_unmount; 1481 } 1482 1483 return 0; 1484 1485 out_filestream_unmount: 1486 xfs_filestream_unmount(mp); 1487 out_free_sb: 1488 xfs_freesb(mp); 1489 out_destroy_counters: 1490 xfs_icsb_destroy_counters(mp); 1491 out_destroy_workqueues: 1492 xfs_destroy_mount_workqueues(mp); 1493 out_close_devices: 1494 xfs_close_devices(mp); 1495 out_free_fsname: 1496 xfs_free_fsname(mp); 1497 kfree(mp); 1498 out: 1499 return -error; 1500 1501 out_unmount: 1502 xfs_filestream_unmount(mp); 1503 xfs_unmountfs(mp); 1504 goto out_free_sb; 1505 } 1506 1507 STATIC struct dentry * 1508 xfs_fs_mount( 1509 struct file_system_type *fs_type, 1510 int flags, 1511 const char *dev_name, 1512 void *data) 1513 { 1514 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super); 1515 } 1516 1517 static int 1518 xfs_fs_nr_cached_objects( 1519 struct super_block *sb) 1520 { 1521 return xfs_reclaim_inodes_count(XFS_M(sb)); 1522 } 1523 1524 static void 1525 xfs_fs_free_cached_objects( 1526 struct super_block *sb, 1527 int nr_to_scan) 1528 { 1529 xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan); 1530 } 1531 1532 static const struct super_operations xfs_super_operations = { 1533 .alloc_inode = xfs_fs_alloc_inode, 1534 .destroy_inode = xfs_fs_destroy_inode, 1535 .evict_inode = xfs_fs_evict_inode, 1536 .drop_inode = xfs_fs_drop_inode, 1537 .put_super = xfs_fs_put_super, 1538 .sync_fs = xfs_fs_sync_fs, 1539 .freeze_fs = xfs_fs_freeze, 1540 .unfreeze_fs = xfs_fs_unfreeze, 1541 .statfs = xfs_fs_statfs, 1542 .remount_fs = xfs_fs_remount, 1543 .show_options = xfs_fs_show_options, 1544 .nr_cached_objects = xfs_fs_nr_cached_objects, 1545 .free_cached_objects = xfs_fs_free_cached_objects, 1546 }; 1547 1548 static struct file_system_type xfs_fs_type = { 1549 .owner = THIS_MODULE, 1550 .name = "xfs", 1551 .mount = xfs_fs_mount, 1552 .kill_sb = kill_block_super, 1553 .fs_flags = FS_REQUIRES_DEV, 1554 }; 1555 1556 STATIC int __init 1557 xfs_init_zones(void) 1558 { 1559 1560 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); 1561 if (!xfs_ioend_zone) 1562 goto out; 1563 1564 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE, 1565 xfs_ioend_zone); 1566 if (!xfs_ioend_pool) 1567 goto out_destroy_ioend_zone; 1568 1569 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t), 1570 "xfs_log_ticket"); 1571 if (!xfs_log_ticket_zone) 1572 goto out_destroy_ioend_pool; 1573 1574 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t), 1575 "xfs_bmap_free_item"); 1576 if (!xfs_bmap_free_item_zone) 1577 goto out_destroy_log_ticket_zone; 1578 1579 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t), 1580 "xfs_btree_cur"); 1581 if (!xfs_btree_cur_zone) 1582 goto out_destroy_bmap_free_item_zone; 1583 1584 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t), 1585 "xfs_da_state"); 1586 if (!xfs_da_state_zone) 1587 goto out_destroy_btree_cur_zone; 1588 1589 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork"); 1590 if (!xfs_ifork_zone) 1591 goto out_destroy_da_state_zone; 1592 1593 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans"); 1594 if (!xfs_trans_zone) 1595 goto out_destroy_ifork_zone; 1596 1597 xfs_log_item_desc_zone = 1598 kmem_zone_init(sizeof(struct xfs_log_item_desc), 1599 "xfs_log_item_desc"); 1600 if (!xfs_log_item_desc_zone) 1601 goto out_destroy_trans_zone; 1602 1603 /* 1604 * The size of the zone allocated buf log item is the maximum 1605 * size possible under XFS. This wastes a little bit of memory, 1606 * but it is much faster. 1607 */ 1608 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item), 1609 "xfs_buf_item"); 1610 if (!xfs_buf_item_zone) 1611 goto out_destroy_log_item_desc_zone; 1612 1613 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) + 1614 ((XFS_EFD_MAX_FAST_EXTENTS - 1) * 1615 sizeof(xfs_extent_t))), "xfs_efd_item"); 1616 if (!xfs_efd_zone) 1617 goto out_destroy_buf_item_zone; 1618 1619 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) + 1620 ((XFS_EFI_MAX_FAST_EXTENTS - 1) * 1621 sizeof(xfs_extent_t))), "xfs_efi_item"); 1622 if (!xfs_efi_zone) 1623 goto out_destroy_efd_zone; 1624 1625 xfs_inode_zone = 1626 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode", 1627 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD, 1628 xfs_fs_inode_init_once); 1629 if (!xfs_inode_zone) 1630 goto out_destroy_efi_zone; 1631 1632 xfs_ili_zone = 1633 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili", 1634 KM_ZONE_SPREAD, NULL); 1635 if (!xfs_ili_zone) 1636 goto out_destroy_inode_zone; 1637 1638 return 0; 1639 1640 out_destroy_inode_zone: 1641 kmem_zone_destroy(xfs_inode_zone); 1642 out_destroy_efi_zone: 1643 kmem_zone_destroy(xfs_efi_zone); 1644 out_destroy_efd_zone: 1645 kmem_zone_destroy(xfs_efd_zone); 1646 out_destroy_buf_item_zone: 1647 kmem_zone_destroy(xfs_buf_item_zone); 1648 out_destroy_log_item_desc_zone: 1649 kmem_zone_destroy(xfs_log_item_desc_zone); 1650 out_destroy_trans_zone: 1651 kmem_zone_destroy(xfs_trans_zone); 1652 out_destroy_ifork_zone: 1653 kmem_zone_destroy(xfs_ifork_zone); 1654 out_destroy_da_state_zone: 1655 kmem_zone_destroy(xfs_da_state_zone); 1656 out_destroy_btree_cur_zone: 1657 kmem_zone_destroy(xfs_btree_cur_zone); 1658 out_destroy_bmap_free_item_zone: 1659 kmem_zone_destroy(xfs_bmap_free_item_zone); 1660 out_destroy_log_ticket_zone: 1661 kmem_zone_destroy(xfs_log_ticket_zone); 1662 out_destroy_ioend_pool: 1663 mempool_destroy(xfs_ioend_pool); 1664 out_destroy_ioend_zone: 1665 kmem_zone_destroy(xfs_ioend_zone); 1666 out: 1667 return -ENOMEM; 1668 } 1669 1670 STATIC void 1671 xfs_destroy_zones(void) 1672 { 1673 /* 1674 * Make sure all delayed rcu free are flushed before we 1675 * destroy caches. 1676 */ 1677 rcu_barrier(); 1678 kmem_zone_destroy(xfs_ili_zone); 1679 kmem_zone_destroy(xfs_inode_zone); 1680 kmem_zone_destroy(xfs_efi_zone); 1681 kmem_zone_destroy(xfs_efd_zone); 1682 kmem_zone_destroy(xfs_buf_item_zone); 1683 kmem_zone_destroy(xfs_log_item_desc_zone); 1684 kmem_zone_destroy(xfs_trans_zone); 1685 kmem_zone_destroy(xfs_ifork_zone); 1686 kmem_zone_destroy(xfs_da_state_zone); 1687 kmem_zone_destroy(xfs_btree_cur_zone); 1688 kmem_zone_destroy(xfs_bmap_free_item_zone); 1689 kmem_zone_destroy(xfs_log_ticket_zone); 1690 mempool_destroy(xfs_ioend_pool); 1691 kmem_zone_destroy(xfs_ioend_zone); 1692 1693 } 1694 1695 STATIC int __init 1696 xfs_init_workqueues(void) 1697 { 1698 /* 1699 * The allocation workqueue can be used in memory reclaim situations 1700 * (writepage path), and parallelism is only limited by the number of 1701 * AGs in all the filesystems mounted. Hence use the default large 1702 * max_active value for this workqueue. 1703 */ 1704 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0); 1705 if (!xfs_alloc_wq) 1706 return -ENOMEM; 1707 1708 return 0; 1709 } 1710 1711 STATIC void 1712 xfs_destroy_workqueues(void) 1713 { 1714 destroy_workqueue(xfs_alloc_wq); 1715 } 1716 1717 STATIC int __init 1718 init_xfs_fs(void) 1719 { 1720 int error; 1721 1722 printk(KERN_INFO XFS_VERSION_STRING " with " 1723 XFS_BUILD_OPTIONS " enabled\n"); 1724 1725 xfs_dir_startup(); 1726 1727 error = xfs_init_zones(); 1728 if (error) 1729 goto out; 1730 1731 error = xfs_init_workqueues(); 1732 if (error) 1733 goto out_destroy_zones; 1734 1735 error = xfs_mru_cache_init(); 1736 if (error) 1737 goto out_destroy_wq; 1738 1739 error = xfs_filestream_init(); 1740 if (error) 1741 goto out_mru_cache_uninit; 1742 1743 error = xfs_buf_init(); 1744 if (error) 1745 goto out_filestream_uninit; 1746 1747 error = xfs_init_procfs(); 1748 if (error) 1749 goto out_buf_terminate; 1750 1751 error = xfs_sysctl_register(); 1752 if (error) 1753 goto out_cleanup_procfs; 1754 1755 error = xfs_qm_init(); 1756 if (error) 1757 goto out_sysctl_unregister; 1758 1759 error = register_filesystem(&xfs_fs_type); 1760 if (error) 1761 goto out_qm_exit; 1762 return 0; 1763 1764 out_qm_exit: 1765 xfs_qm_exit(); 1766 out_sysctl_unregister: 1767 xfs_sysctl_unregister(); 1768 out_cleanup_procfs: 1769 xfs_cleanup_procfs(); 1770 out_buf_terminate: 1771 xfs_buf_terminate(); 1772 out_filestream_uninit: 1773 xfs_filestream_uninit(); 1774 out_mru_cache_uninit: 1775 xfs_mru_cache_uninit(); 1776 out_destroy_wq: 1777 xfs_destroy_workqueues(); 1778 out_destroy_zones: 1779 xfs_destroy_zones(); 1780 out: 1781 return error; 1782 } 1783 1784 STATIC void __exit 1785 exit_xfs_fs(void) 1786 { 1787 xfs_qm_exit(); 1788 unregister_filesystem(&xfs_fs_type); 1789 xfs_sysctl_unregister(); 1790 xfs_cleanup_procfs(); 1791 xfs_buf_terminate(); 1792 xfs_filestream_uninit(); 1793 xfs_mru_cache_uninit(); 1794 xfs_destroy_workqueues(); 1795 xfs_destroy_zones(); 1796 } 1797 1798 module_init(init_xfs_fs); 1799 module_exit(exit_xfs_fs); 1800 1801 MODULE_AUTHOR("Silicon Graphics, Inc."); 1802 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); 1803 MODULE_LICENSE("GPL"); 1804