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_shared.h" 21 #include "xfs_format.h" 22 #include "xfs_log_format.h" 23 #include "xfs_trans_resv.h" 24 #include "xfs_inum.h" 25 #include "xfs_sb.h" 26 #include "xfs_ag.h" 27 #include "xfs_mount.h" 28 #include "xfs_da_format.h" 29 #include "xfs_inode.h" 30 #include "xfs_btree.h" 31 #include "xfs_bmap.h" 32 #include "xfs_alloc.h" 33 #include "xfs_error.h" 34 #include "xfs_fsops.h" 35 #include "xfs_trans.h" 36 #include "xfs_buf_item.h" 37 #include "xfs_log.h" 38 #include "xfs_log_priv.h" 39 #include "xfs_da_btree.h" 40 #include "xfs_dir2.h" 41 #include "xfs_extfree_item.h" 42 #include "xfs_mru_cache.h" 43 #include "xfs_inode_item.h" 44 #include "xfs_icache.h" 45 #include "xfs_trace.h" 46 #include "xfs_icreate_item.h" 47 #include "xfs_dinode.h" 48 #include "xfs_filestream.h" 49 #include "xfs_quota.h" 50 51 #include <linux/namei.h> 52 #include <linux/init.h> 53 #include <linux/slab.h> 54 #include <linux/mount.h> 55 #include <linux/mempool.h> 56 #include <linux/writeback.h> 57 #include <linux/kthread.h> 58 #include <linux/freezer.h> 59 #include <linux/parser.h> 60 61 static const struct super_operations xfs_super_operations; 62 static kmem_zone_t *xfs_ioend_zone; 63 mempool_t *xfs_ioend_pool; 64 65 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */ 66 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */ 67 #define MNTOPT_LOGDEV "logdev" /* log device */ 68 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */ 69 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */ 70 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */ 71 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */ 72 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */ 73 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */ 74 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */ 75 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */ 76 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */ 77 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */ 78 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */ 79 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */ 80 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */ 81 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */ 82 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */ 83 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and 84 * unwritten extent conversion */ 85 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */ 86 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */ 87 #define MNTOPT_32BITINODE "inode32" /* inode allocation limited to 88 * XFS_MAXINUMBER_32 */ 89 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */ 90 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */ 91 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */ 92 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes 93 * in stat(). */ 94 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */ 95 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */ 96 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */ 97 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */ 98 #define MNTOPT_NOQUOTA "noquota" /* no quotas */ 99 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */ 100 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */ 101 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */ 102 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */ 103 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */ 104 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */ 105 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */ 106 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */ 107 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */ 108 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */ 109 #define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */ 110 #define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */ 111 #define MNTOPT_DISCARD "discard" /* Discard unused blocks */ 112 #define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */ 113 114 /* 115 * Table driven mount option parser. 116 * 117 * Currently only used for remount, but it will be used for mount 118 * in the future, too. 119 */ 120 enum { 121 Opt_barrier, 122 Opt_nobarrier, 123 Opt_inode64, 124 Opt_inode32, 125 Opt_err 126 }; 127 128 static const match_table_t tokens = { 129 {Opt_barrier, "barrier"}, 130 {Opt_nobarrier, "nobarrier"}, 131 {Opt_inode64, "inode64"}, 132 {Opt_inode32, "inode32"}, 133 {Opt_err, NULL} 134 }; 135 136 137 STATIC unsigned long 138 suffix_kstrtoint(char *s, unsigned int base, int *res) 139 { 140 int last, shift_left_factor = 0, _res; 141 char *value = s; 142 143 last = strlen(value) - 1; 144 if (value[last] == 'K' || value[last] == 'k') { 145 shift_left_factor = 10; 146 value[last] = '\0'; 147 } 148 if (value[last] == 'M' || value[last] == 'm') { 149 shift_left_factor = 20; 150 value[last] = '\0'; 151 } 152 if (value[last] == 'G' || value[last] == 'g') { 153 shift_left_factor = 30; 154 value[last] = '\0'; 155 } 156 157 if (kstrtoint(s, base, &_res)) 158 return -EINVAL; 159 *res = _res << shift_left_factor; 160 return 0; 161 } 162 163 /* 164 * This function fills in xfs_mount_t fields based on mount args. 165 * Note: the superblock has _not_ yet been read in. 166 * 167 * Note that this function leaks the various device name allocations on 168 * failure. The caller takes care of them. 169 */ 170 STATIC int 171 xfs_parseargs( 172 struct xfs_mount *mp, 173 char *options) 174 { 175 struct super_block *sb = mp->m_super; 176 char *this_char, *value; 177 int dsunit = 0; 178 int dswidth = 0; 179 int iosize = 0; 180 __uint8_t iosizelog = 0; 181 182 /* 183 * set up the mount name first so all the errors will refer to the 184 * correct device. 185 */ 186 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL); 187 if (!mp->m_fsname) 188 return ENOMEM; 189 mp->m_fsname_len = strlen(mp->m_fsname) + 1; 190 191 /* 192 * Copy binary VFS mount flags we are interested in. 193 */ 194 if (sb->s_flags & MS_RDONLY) 195 mp->m_flags |= XFS_MOUNT_RDONLY; 196 if (sb->s_flags & MS_DIRSYNC) 197 mp->m_flags |= XFS_MOUNT_DIRSYNC; 198 if (sb->s_flags & MS_SYNCHRONOUS) 199 mp->m_flags |= XFS_MOUNT_WSYNC; 200 201 /* 202 * Set some default flags that could be cleared by the mount option 203 * parsing. 204 */ 205 mp->m_flags |= XFS_MOUNT_BARRIER; 206 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE; 207 #if !XFS_BIG_INUMS 208 mp->m_flags |= XFS_MOUNT_SMALL_INUMS; 209 #endif 210 211 /* 212 * These can be overridden by the mount option parsing. 213 */ 214 mp->m_logbufs = -1; 215 mp->m_logbsize = -1; 216 217 if (!options) 218 goto done; 219 220 while ((this_char = strsep(&options, ",")) != NULL) { 221 if (!*this_char) 222 continue; 223 if ((value = strchr(this_char, '=')) != NULL) 224 *value++ = 0; 225 226 if (!strcmp(this_char, MNTOPT_LOGBUFS)) { 227 if (!value || !*value) { 228 xfs_warn(mp, "%s option requires an argument", 229 this_char); 230 return EINVAL; 231 } 232 if (kstrtoint(value, 10, &mp->m_logbufs)) 233 return EINVAL; 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 if (suffix_kstrtoint(value, 10, &mp->m_logbsize)) 241 return EINVAL; 242 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) { 243 if (!value || !*value) { 244 xfs_warn(mp, "%s option requires an argument", 245 this_char); 246 return EINVAL; 247 } 248 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL); 249 if (!mp->m_logname) 250 return ENOMEM; 251 } else if (!strcmp(this_char, MNTOPT_MTPT)) { 252 xfs_warn(mp, "%s option not allowed on this system", 253 this_char); 254 return EINVAL; 255 } else if (!strcmp(this_char, MNTOPT_RTDEV)) { 256 if (!value || !*value) { 257 xfs_warn(mp, "%s option requires an argument", 258 this_char); 259 return EINVAL; 260 } 261 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL); 262 if (!mp->m_rtname) 263 return ENOMEM; 264 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) { 265 if (!value || !*value) { 266 xfs_warn(mp, "%s option requires an argument", 267 this_char); 268 return EINVAL; 269 } 270 if (kstrtoint(value, 10, &iosize)) 271 return EINVAL; 272 iosizelog = ffs(iosize) - 1; 273 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) { 274 if (!value || !*value) { 275 xfs_warn(mp, "%s option requires an argument", 276 this_char); 277 return EINVAL; 278 } 279 if (suffix_kstrtoint(value, 10, &iosize)) 280 return EINVAL; 281 iosizelog = ffs(iosize) - 1; 282 } else if (!strcmp(this_char, MNTOPT_GRPID) || 283 !strcmp(this_char, MNTOPT_BSDGROUPS)) { 284 mp->m_flags |= XFS_MOUNT_GRPID; 285 } else if (!strcmp(this_char, MNTOPT_NOGRPID) || 286 !strcmp(this_char, MNTOPT_SYSVGROUPS)) { 287 mp->m_flags &= ~XFS_MOUNT_GRPID; 288 } else if (!strcmp(this_char, MNTOPT_WSYNC)) { 289 mp->m_flags |= XFS_MOUNT_WSYNC; 290 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) { 291 mp->m_flags |= XFS_MOUNT_NORECOVERY; 292 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) { 293 mp->m_flags |= XFS_MOUNT_NOALIGN; 294 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) { 295 mp->m_flags |= XFS_MOUNT_SWALLOC; 296 } else if (!strcmp(this_char, MNTOPT_SUNIT)) { 297 if (!value || !*value) { 298 xfs_warn(mp, "%s option requires an argument", 299 this_char); 300 return EINVAL; 301 } 302 if (kstrtoint(value, 10, &dsunit)) 303 return EINVAL; 304 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) { 305 if (!value || !*value) { 306 xfs_warn(mp, "%s option requires an argument", 307 this_char); 308 return EINVAL; 309 } 310 if (kstrtoint(value, 10, &dswidth)) 311 return EINVAL; 312 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) { 313 mp->m_flags |= XFS_MOUNT_SMALL_INUMS; 314 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) { 315 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; 316 #if !XFS_BIG_INUMS 317 xfs_warn(mp, "%s option not allowed on this system", 318 this_char); 319 return EINVAL; 320 #endif 321 } else if (!strcmp(this_char, MNTOPT_NOUUID)) { 322 mp->m_flags |= XFS_MOUNT_NOUUID; 323 } else if (!strcmp(this_char, MNTOPT_BARRIER)) { 324 mp->m_flags |= XFS_MOUNT_BARRIER; 325 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) { 326 mp->m_flags &= ~XFS_MOUNT_BARRIER; 327 } else if (!strcmp(this_char, MNTOPT_IKEEP)) { 328 mp->m_flags |= XFS_MOUNT_IKEEP; 329 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) { 330 mp->m_flags &= ~XFS_MOUNT_IKEEP; 331 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) { 332 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE; 333 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) { 334 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE; 335 } else if (!strcmp(this_char, MNTOPT_ATTR2)) { 336 mp->m_flags |= XFS_MOUNT_ATTR2; 337 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) { 338 mp->m_flags &= ~XFS_MOUNT_ATTR2; 339 mp->m_flags |= XFS_MOUNT_NOATTR2; 340 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) { 341 mp->m_flags |= XFS_MOUNT_FILESTREAMS; 342 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) { 343 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT; 344 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD; 345 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE; 346 } else if (!strcmp(this_char, MNTOPT_QUOTA) || 347 !strcmp(this_char, MNTOPT_UQUOTA) || 348 !strcmp(this_char, MNTOPT_USRQUOTA)) { 349 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE | 350 XFS_UQUOTA_ENFD); 351 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) || 352 !strcmp(this_char, MNTOPT_UQUOTANOENF)) { 353 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE); 354 mp->m_qflags &= ~XFS_UQUOTA_ENFD; 355 } else if (!strcmp(this_char, MNTOPT_PQUOTA) || 356 !strcmp(this_char, MNTOPT_PRJQUOTA)) { 357 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE | 358 XFS_PQUOTA_ENFD); 359 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) { 360 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE); 361 mp->m_qflags &= ~XFS_PQUOTA_ENFD; 362 } else if (!strcmp(this_char, MNTOPT_GQUOTA) || 363 !strcmp(this_char, MNTOPT_GRPQUOTA)) { 364 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE | 365 XFS_GQUOTA_ENFD); 366 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) { 367 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE); 368 mp->m_qflags &= ~XFS_GQUOTA_ENFD; 369 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) { 370 xfs_warn(mp, 371 "delaylog is the default now, option is deprecated."); 372 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) { 373 xfs_warn(mp, 374 "nodelaylog support has been removed, option is deprecated."); 375 } else if (!strcmp(this_char, MNTOPT_DISCARD)) { 376 mp->m_flags |= XFS_MOUNT_DISCARD; 377 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) { 378 mp->m_flags &= ~XFS_MOUNT_DISCARD; 379 } else if (!strcmp(this_char, "ihashsize")) { 380 xfs_warn(mp, 381 "ihashsize no longer used, option is deprecated."); 382 } else if (!strcmp(this_char, "osyncisdsync")) { 383 xfs_warn(mp, 384 "osyncisdsync has no effect, option is deprecated."); 385 } else if (!strcmp(this_char, "osyncisosync")) { 386 xfs_warn(mp, 387 "osyncisosync has no effect, option is deprecated."); 388 } else if (!strcmp(this_char, "irixsgid")) { 389 xfs_warn(mp, 390 "irixsgid is now a sysctl(2) variable, option is deprecated."); 391 } else { 392 xfs_warn(mp, "unknown mount option [%s].", this_char); 393 return EINVAL; 394 } 395 } 396 397 /* 398 * no recovery flag requires a read-only mount 399 */ 400 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) && 401 !(mp->m_flags & XFS_MOUNT_RDONLY)) { 402 xfs_warn(mp, "no-recovery mounts must be read-only."); 403 return EINVAL; 404 } 405 406 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) { 407 xfs_warn(mp, 408 "sunit and swidth options incompatible with the noalign option"); 409 return EINVAL; 410 } 411 412 #ifndef CONFIG_XFS_QUOTA 413 if (XFS_IS_QUOTA_RUNNING(mp)) { 414 xfs_warn(mp, "quota support not available in this kernel."); 415 return EINVAL; 416 } 417 #endif 418 419 if ((dsunit && !dswidth) || (!dsunit && dswidth)) { 420 xfs_warn(mp, "sunit and swidth must be specified together"); 421 return EINVAL; 422 } 423 424 if (dsunit && (dswidth % dsunit != 0)) { 425 xfs_warn(mp, 426 "stripe width (%d) must be a multiple of the stripe unit (%d)", 427 dswidth, dsunit); 428 return EINVAL; 429 } 430 431 done: 432 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) { 433 /* 434 * At this point the superblock has not been read 435 * in, therefore we do not know the block size. 436 * Before the mount call ends we will convert 437 * these to FSBs. 438 */ 439 mp->m_dalign = dsunit; 440 mp->m_swidth = dswidth; 441 } 442 443 if (mp->m_logbufs != -1 && 444 mp->m_logbufs != 0 && 445 (mp->m_logbufs < XLOG_MIN_ICLOGS || 446 mp->m_logbufs > XLOG_MAX_ICLOGS)) { 447 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]", 448 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS); 449 return XFS_ERROR(EINVAL); 450 } 451 if (mp->m_logbsize != -1 && 452 mp->m_logbsize != 0 && 453 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE || 454 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE || 455 !is_power_of_2(mp->m_logbsize))) { 456 xfs_warn(mp, 457 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]", 458 mp->m_logbsize); 459 return XFS_ERROR(EINVAL); 460 } 461 462 if (iosizelog) { 463 if (iosizelog > XFS_MAX_IO_LOG || 464 iosizelog < XFS_MIN_IO_LOG) { 465 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]", 466 iosizelog, XFS_MIN_IO_LOG, 467 XFS_MAX_IO_LOG); 468 return XFS_ERROR(EINVAL); 469 } 470 471 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE; 472 mp->m_readio_log = iosizelog; 473 mp->m_writeio_log = iosizelog; 474 } 475 476 return 0; 477 } 478 479 struct proc_xfs_info { 480 int flag; 481 char *str; 482 }; 483 484 STATIC int 485 xfs_showargs( 486 struct xfs_mount *mp, 487 struct seq_file *m) 488 { 489 static struct proc_xfs_info xfs_info_set[] = { 490 /* the few simple ones we can get from the mount struct */ 491 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP }, 492 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC }, 493 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN }, 494 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC }, 495 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID }, 496 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY }, 497 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 }, 498 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM }, 499 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID }, 500 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD }, 501 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE }, 502 { 0, NULL } 503 }; 504 static struct proc_xfs_info xfs_info_unset[] = { 505 /* the few simple ones we can get from the mount struct */ 506 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO }, 507 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER }, 508 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE }, 509 { 0, NULL } 510 }; 511 struct proc_xfs_info *xfs_infop; 512 513 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) { 514 if (mp->m_flags & xfs_infop->flag) 515 seq_puts(m, xfs_infop->str); 516 } 517 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) { 518 if (!(mp->m_flags & xfs_infop->flag)) 519 seq_puts(m, xfs_infop->str); 520 } 521 522 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE) 523 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk", 524 (int)(1 << mp->m_writeio_log) >> 10); 525 526 if (mp->m_logbufs > 0) 527 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs); 528 if (mp->m_logbsize > 0) 529 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10); 530 531 if (mp->m_logname) 532 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname); 533 if (mp->m_rtname) 534 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname); 535 536 if (mp->m_dalign > 0) 537 seq_printf(m, "," MNTOPT_SUNIT "=%d", 538 (int)XFS_FSB_TO_BB(mp, mp->m_dalign)); 539 if (mp->m_swidth > 0) 540 seq_printf(m, "," MNTOPT_SWIDTH "=%d", 541 (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); 542 543 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD)) 544 seq_puts(m, "," MNTOPT_USRQUOTA); 545 else if (mp->m_qflags & XFS_UQUOTA_ACCT) 546 seq_puts(m, "," MNTOPT_UQUOTANOENF); 547 548 if (mp->m_qflags & XFS_PQUOTA_ACCT) { 549 if (mp->m_qflags & XFS_PQUOTA_ENFD) 550 seq_puts(m, "," MNTOPT_PRJQUOTA); 551 else 552 seq_puts(m, "," MNTOPT_PQUOTANOENF); 553 } 554 if (mp->m_qflags & XFS_GQUOTA_ACCT) { 555 if (mp->m_qflags & XFS_GQUOTA_ENFD) 556 seq_puts(m, "," MNTOPT_GRPQUOTA); 557 else 558 seq_puts(m, "," MNTOPT_GQUOTANOENF); 559 } 560 561 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT)) 562 seq_puts(m, "," MNTOPT_NOQUOTA); 563 564 return 0; 565 } 566 __uint64_t 567 xfs_max_file_offset( 568 unsigned int blockshift) 569 { 570 unsigned int pagefactor = 1; 571 unsigned int bitshift = BITS_PER_LONG - 1; 572 573 /* Figure out maximum filesize, on Linux this can depend on 574 * the filesystem blocksize (on 32 bit platforms). 575 * __block_write_begin does this in an [unsigned] long... 576 * page->index << (PAGE_CACHE_SHIFT - bbits) 577 * So, for page sized blocks (4K on 32 bit platforms), 578 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is 579 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) 580 * but for smaller blocksizes it is less (bbits = log2 bsize). 581 * Note1: get_block_t takes a long (implicit cast from above) 582 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch 583 * can optionally convert the [unsigned] long from above into 584 * an [unsigned] long long. 585 */ 586 587 #if BITS_PER_LONG == 32 588 # if defined(CONFIG_LBDAF) 589 ASSERT(sizeof(sector_t) == 8); 590 pagefactor = PAGE_CACHE_SIZE; 591 bitshift = BITS_PER_LONG; 592 # else 593 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); 594 # endif 595 #endif 596 597 return (((__uint64_t)pagefactor) << bitshift) - 1; 598 } 599 600 xfs_agnumber_t 601 xfs_set_inode32(struct xfs_mount *mp) 602 { 603 xfs_agnumber_t index = 0; 604 xfs_agnumber_t maxagi = 0; 605 xfs_sb_t *sbp = &mp->m_sb; 606 xfs_agnumber_t max_metadata; 607 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0); 608 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino); 609 xfs_perag_t *pag; 610 611 /* Calculate how much should be reserved for inodes to meet 612 * the max inode percentage. 613 */ 614 if (mp->m_maxicount) { 615 __uint64_t icount; 616 617 icount = sbp->sb_dblocks * sbp->sb_imax_pct; 618 do_div(icount, 100); 619 icount += sbp->sb_agblocks - 1; 620 do_div(icount, sbp->sb_agblocks); 621 max_metadata = icount; 622 } else { 623 max_metadata = sbp->sb_agcount; 624 } 625 626 for (index = 0; index < sbp->sb_agcount; index++) { 627 ino = XFS_AGINO_TO_INO(mp, index, agino); 628 629 if (ino > XFS_MAXINUMBER_32) { 630 pag = xfs_perag_get(mp, index); 631 pag->pagi_inodeok = 0; 632 pag->pagf_metadata = 0; 633 xfs_perag_put(pag); 634 continue; 635 } 636 637 pag = xfs_perag_get(mp, index); 638 pag->pagi_inodeok = 1; 639 maxagi++; 640 if (index < max_metadata) 641 pag->pagf_metadata = 1; 642 xfs_perag_put(pag); 643 } 644 mp->m_flags |= (XFS_MOUNT_32BITINODES | 645 XFS_MOUNT_SMALL_INUMS); 646 647 return maxagi; 648 } 649 650 xfs_agnumber_t 651 xfs_set_inode64(struct xfs_mount *mp) 652 { 653 xfs_agnumber_t index = 0; 654 655 for (index = 0; index < mp->m_sb.sb_agcount; index++) { 656 struct xfs_perag *pag; 657 658 pag = xfs_perag_get(mp, index); 659 pag->pagi_inodeok = 1; 660 pag->pagf_metadata = 0; 661 xfs_perag_put(pag); 662 } 663 664 /* There is no need for lock protection on m_flags, 665 * the rw_semaphore of the VFS superblock is locked 666 * during mount/umount/remount operations, so this is 667 * enough to avoid concurency on the m_flags field 668 */ 669 mp->m_flags &= ~(XFS_MOUNT_32BITINODES | 670 XFS_MOUNT_SMALL_INUMS); 671 return index; 672 } 673 674 STATIC int 675 xfs_blkdev_get( 676 xfs_mount_t *mp, 677 const char *name, 678 struct block_device **bdevp) 679 { 680 int error = 0; 681 682 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL, 683 mp); 684 if (IS_ERR(*bdevp)) { 685 error = PTR_ERR(*bdevp); 686 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error); 687 } 688 689 return -error; 690 } 691 692 STATIC void 693 xfs_blkdev_put( 694 struct block_device *bdev) 695 { 696 if (bdev) 697 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 698 } 699 700 void 701 xfs_blkdev_issue_flush( 702 xfs_buftarg_t *buftarg) 703 { 704 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL); 705 } 706 707 STATIC void 708 xfs_close_devices( 709 struct xfs_mount *mp) 710 { 711 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { 712 struct block_device *logdev = mp->m_logdev_targp->bt_bdev; 713 xfs_free_buftarg(mp, mp->m_logdev_targp); 714 xfs_blkdev_put(logdev); 715 } 716 if (mp->m_rtdev_targp) { 717 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev; 718 xfs_free_buftarg(mp, mp->m_rtdev_targp); 719 xfs_blkdev_put(rtdev); 720 } 721 xfs_free_buftarg(mp, mp->m_ddev_targp); 722 } 723 724 /* 725 * The file system configurations are: 726 * (1) device (partition) with data and internal log 727 * (2) logical volume with data and log subvolumes. 728 * (3) logical volume with data, log, and realtime subvolumes. 729 * 730 * We only have to handle opening the log and realtime volumes here if 731 * they are present. The data subvolume has already been opened by 732 * get_sb_bdev() and is stored in sb->s_bdev. 733 */ 734 STATIC int 735 xfs_open_devices( 736 struct xfs_mount *mp) 737 { 738 struct block_device *ddev = mp->m_super->s_bdev; 739 struct block_device *logdev = NULL, *rtdev = NULL; 740 int error; 741 742 /* 743 * Open real time and log devices - order is important. 744 */ 745 if (mp->m_logname) { 746 error = xfs_blkdev_get(mp, mp->m_logname, &logdev); 747 if (error) 748 goto out; 749 } 750 751 if (mp->m_rtname) { 752 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev); 753 if (error) 754 goto out_close_logdev; 755 756 if (rtdev == ddev || rtdev == logdev) { 757 xfs_warn(mp, 758 "Cannot mount filesystem with identical rtdev and ddev/logdev."); 759 error = EINVAL; 760 goto out_close_rtdev; 761 } 762 } 763 764 /* 765 * Setup xfs_mount buffer target pointers 766 */ 767 error = ENOMEM; 768 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev); 769 if (!mp->m_ddev_targp) 770 goto out_close_rtdev; 771 772 if (rtdev) { 773 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev); 774 if (!mp->m_rtdev_targp) 775 goto out_free_ddev_targ; 776 } 777 778 if (logdev && logdev != ddev) { 779 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev); 780 if (!mp->m_logdev_targp) 781 goto out_free_rtdev_targ; 782 } else { 783 mp->m_logdev_targp = mp->m_ddev_targp; 784 } 785 786 return 0; 787 788 out_free_rtdev_targ: 789 if (mp->m_rtdev_targp) 790 xfs_free_buftarg(mp, mp->m_rtdev_targp); 791 out_free_ddev_targ: 792 xfs_free_buftarg(mp, mp->m_ddev_targp); 793 out_close_rtdev: 794 if (rtdev) 795 xfs_blkdev_put(rtdev); 796 out_close_logdev: 797 if (logdev && logdev != ddev) 798 xfs_blkdev_put(logdev); 799 out: 800 return error; 801 } 802 803 /* 804 * Setup xfs_mount buffer target pointers based on superblock 805 */ 806 STATIC int 807 xfs_setup_devices( 808 struct xfs_mount *mp) 809 { 810 int error; 811 812 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize); 813 if (error) 814 return error; 815 816 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { 817 unsigned int log_sector_size = BBSIZE; 818 819 if (xfs_sb_version_hassector(&mp->m_sb)) 820 log_sector_size = mp->m_sb.sb_logsectsize; 821 error = xfs_setsize_buftarg(mp->m_logdev_targp, 822 log_sector_size); 823 if (error) 824 return error; 825 } 826 if (mp->m_rtdev_targp) { 827 error = xfs_setsize_buftarg(mp->m_rtdev_targp, 828 mp->m_sb.sb_sectsize); 829 if (error) 830 return error; 831 } 832 833 return 0; 834 } 835 836 STATIC int 837 xfs_init_mount_workqueues( 838 struct xfs_mount *mp) 839 { 840 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s", 841 WQ_MEM_RECLAIM, 0, mp->m_fsname); 842 if (!mp->m_data_workqueue) 843 goto out; 844 845 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s", 846 WQ_MEM_RECLAIM, 0, mp->m_fsname); 847 if (!mp->m_unwritten_workqueue) 848 goto out_destroy_data_iodone_queue; 849 850 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s", 851 WQ_MEM_RECLAIM, 0, mp->m_fsname); 852 if (!mp->m_cil_workqueue) 853 goto out_destroy_unwritten; 854 855 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s", 856 0, 0, mp->m_fsname); 857 if (!mp->m_reclaim_workqueue) 858 goto out_destroy_cil; 859 860 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s", 861 0, 0, mp->m_fsname); 862 if (!mp->m_log_workqueue) 863 goto out_destroy_reclaim; 864 865 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s", 866 0, 0, mp->m_fsname); 867 if (!mp->m_eofblocks_workqueue) 868 goto out_destroy_log; 869 870 return 0; 871 872 out_destroy_log: 873 destroy_workqueue(mp->m_log_workqueue); 874 out_destroy_reclaim: 875 destroy_workqueue(mp->m_reclaim_workqueue); 876 out_destroy_cil: 877 destroy_workqueue(mp->m_cil_workqueue); 878 out_destroy_unwritten: 879 destroy_workqueue(mp->m_unwritten_workqueue); 880 out_destroy_data_iodone_queue: 881 destroy_workqueue(mp->m_data_workqueue); 882 out: 883 return -ENOMEM; 884 } 885 886 STATIC void 887 xfs_destroy_mount_workqueues( 888 struct xfs_mount *mp) 889 { 890 destroy_workqueue(mp->m_eofblocks_workqueue); 891 destroy_workqueue(mp->m_log_workqueue); 892 destroy_workqueue(mp->m_reclaim_workqueue); 893 destroy_workqueue(mp->m_cil_workqueue); 894 destroy_workqueue(mp->m_data_workqueue); 895 destroy_workqueue(mp->m_unwritten_workqueue); 896 } 897 898 /* 899 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK 900 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting 901 * for IO to complete so that we effectively throttle multiple callers to the 902 * rate at which IO is completing. 903 */ 904 void 905 xfs_flush_inodes( 906 struct xfs_mount *mp) 907 { 908 struct super_block *sb = mp->m_super; 909 910 if (down_read_trylock(&sb->s_umount)) { 911 sync_inodes_sb(sb); 912 up_read(&sb->s_umount); 913 } 914 } 915 916 /* Catch misguided souls that try to use this interface on XFS */ 917 STATIC struct inode * 918 xfs_fs_alloc_inode( 919 struct super_block *sb) 920 { 921 BUG(); 922 return NULL; 923 } 924 925 /* 926 * Now that the generic code is guaranteed not to be accessing 927 * the linux inode, we can reclaim the inode. 928 */ 929 STATIC void 930 xfs_fs_destroy_inode( 931 struct inode *inode) 932 { 933 struct xfs_inode *ip = XFS_I(inode); 934 935 trace_xfs_destroy_inode(ip); 936 937 XFS_STATS_INC(vn_reclaim); 938 939 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0); 940 941 /* 942 * We should never get here with one of the reclaim flags already set. 943 */ 944 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE)); 945 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM)); 946 947 /* 948 * We always use background reclaim here because even if the 949 * inode is clean, it still may be under IO and hence we have 950 * to take the flush lock. The background reclaim path handles 951 * this more efficiently than we can here, so simply let background 952 * reclaim tear down all inodes. 953 */ 954 xfs_inode_set_reclaim_tag(ip); 955 } 956 957 /* 958 * Slab object creation initialisation for the XFS inode. 959 * This covers only the idempotent fields in the XFS inode; 960 * all other fields need to be initialised on allocation 961 * from the slab. This avoids the need to repeatedly initialise 962 * fields in the xfs inode that left in the initialise state 963 * when freeing the inode. 964 */ 965 STATIC void 966 xfs_fs_inode_init_once( 967 void *inode) 968 { 969 struct xfs_inode *ip = inode; 970 971 memset(ip, 0, sizeof(struct xfs_inode)); 972 973 /* vfs inode */ 974 inode_init_once(VFS_I(ip)); 975 976 /* xfs inode */ 977 atomic_set(&ip->i_pincount, 0); 978 spin_lock_init(&ip->i_flags_lock); 979 980 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, 981 "xfsino", ip->i_ino); 982 } 983 984 STATIC void 985 xfs_fs_evict_inode( 986 struct inode *inode) 987 { 988 xfs_inode_t *ip = XFS_I(inode); 989 990 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock)); 991 992 trace_xfs_evict_inode(ip); 993 994 truncate_inode_pages_final(&inode->i_data); 995 clear_inode(inode); 996 XFS_STATS_INC(vn_rele); 997 XFS_STATS_INC(vn_remove); 998 XFS_STATS_DEC(vn_active); 999 1000 xfs_inactive(ip); 1001 } 1002 1003 /* 1004 * We do an unlocked check for XFS_IDONTCACHE here because we are already 1005 * serialised against cache hits here via the inode->i_lock and igrab() in 1006 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be 1007 * racing with us, and it avoids needing to grab a spinlock here for every inode 1008 * we drop the final reference on. 1009 */ 1010 STATIC int 1011 xfs_fs_drop_inode( 1012 struct inode *inode) 1013 { 1014 struct xfs_inode *ip = XFS_I(inode); 1015 1016 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE); 1017 } 1018 1019 STATIC void 1020 xfs_free_fsname( 1021 struct xfs_mount *mp) 1022 { 1023 kfree(mp->m_fsname); 1024 kfree(mp->m_rtname); 1025 kfree(mp->m_logname); 1026 } 1027 1028 STATIC void 1029 xfs_fs_put_super( 1030 struct super_block *sb) 1031 { 1032 struct xfs_mount *mp = XFS_M(sb); 1033 1034 xfs_filestream_unmount(mp); 1035 xfs_unmountfs(mp); 1036 1037 xfs_freesb(mp); 1038 xfs_icsb_destroy_counters(mp); 1039 xfs_destroy_mount_workqueues(mp); 1040 xfs_close_devices(mp); 1041 xfs_free_fsname(mp); 1042 kfree(mp); 1043 } 1044 1045 STATIC int 1046 xfs_fs_sync_fs( 1047 struct super_block *sb, 1048 int wait) 1049 { 1050 struct xfs_mount *mp = XFS_M(sb); 1051 1052 /* 1053 * Doing anything during the async pass would be counterproductive. 1054 */ 1055 if (!wait) 1056 return 0; 1057 1058 xfs_log_force(mp, XFS_LOG_SYNC); 1059 if (laptop_mode) { 1060 /* 1061 * The disk must be active because we're syncing. 1062 * We schedule log work now (now that the disk is 1063 * active) instead of later (when it might not be). 1064 */ 1065 flush_delayed_work(&mp->m_log->l_work); 1066 } 1067 1068 return 0; 1069 } 1070 1071 STATIC int 1072 xfs_fs_statfs( 1073 struct dentry *dentry, 1074 struct kstatfs *statp) 1075 { 1076 struct xfs_mount *mp = XFS_M(dentry->d_sb); 1077 xfs_sb_t *sbp = &mp->m_sb; 1078 struct xfs_inode *ip = XFS_I(dentry->d_inode); 1079 __uint64_t fakeinos, id; 1080 xfs_extlen_t lsize; 1081 __int64_t ffree; 1082 1083 statp->f_type = XFS_SB_MAGIC; 1084 statp->f_namelen = MAXNAMELEN - 1; 1085 1086 id = huge_encode_dev(mp->m_ddev_targp->bt_dev); 1087 statp->f_fsid.val[0] = (u32)id; 1088 statp->f_fsid.val[1] = (u32)(id >> 32); 1089 1090 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT); 1091 1092 spin_lock(&mp->m_sb_lock); 1093 statp->f_bsize = sbp->sb_blocksize; 1094 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; 1095 statp->f_blocks = sbp->sb_dblocks - lsize; 1096 statp->f_bfree = statp->f_bavail = 1097 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp); 1098 fakeinos = statp->f_bfree << sbp->sb_inopblog; 1099 statp->f_files = 1100 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER); 1101 if (mp->m_maxicount) 1102 statp->f_files = min_t(typeof(statp->f_files), 1103 statp->f_files, 1104 mp->m_maxicount); 1105 1106 /* make sure statp->f_ffree does not underflow */ 1107 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree); 1108 statp->f_ffree = max_t(__int64_t, ffree, 0); 1109 1110 spin_unlock(&mp->m_sb_lock); 1111 1112 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && 1113 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) == 1114 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD)) 1115 xfs_qm_statvfs(ip, statp); 1116 return 0; 1117 } 1118 1119 STATIC void 1120 xfs_save_resvblks(struct xfs_mount *mp) 1121 { 1122 __uint64_t resblks = 0; 1123 1124 mp->m_resblks_save = mp->m_resblks; 1125 xfs_reserve_blocks(mp, &resblks, NULL); 1126 } 1127 1128 STATIC void 1129 xfs_restore_resvblks(struct xfs_mount *mp) 1130 { 1131 __uint64_t resblks; 1132 1133 if (mp->m_resblks_save) { 1134 resblks = mp->m_resblks_save; 1135 mp->m_resblks_save = 0; 1136 } else 1137 resblks = xfs_default_resblks(mp); 1138 1139 xfs_reserve_blocks(mp, &resblks, NULL); 1140 } 1141 1142 /* 1143 * Trigger writeback of all the dirty metadata in the file system. 1144 * 1145 * This ensures that the metadata is written to their location on disk rather 1146 * than just existing in transactions in the log. This means after a quiesce 1147 * there is no log replay required to write the inodes to disk - this is the 1148 * primary difference between a sync and a quiesce. 1149 * 1150 * Note: xfs_log_quiesce() stops background log work - the callers must ensure 1151 * it is started again when appropriate. 1152 */ 1153 static void 1154 xfs_quiesce_attr( 1155 struct xfs_mount *mp) 1156 { 1157 int error = 0; 1158 1159 /* wait for all modifications to complete */ 1160 while (atomic_read(&mp->m_active_trans) > 0) 1161 delay(100); 1162 1163 /* force the log to unpin objects from the now complete transactions */ 1164 xfs_log_force(mp, XFS_LOG_SYNC); 1165 1166 /* reclaim inodes to do any IO before the freeze completes */ 1167 xfs_reclaim_inodes(mp, 0); 1168 xfs_reclaim_inodes(mp, SYNC_WAIT); 1169 1170 /* Push the superblock and write an unmount record */ 1171 error = xfs_log_sbcount(mp); 1172 if (error) 1173 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. " 1174 "Frozen image may not be consistent."); 1175 /* 1176 * Just warn here till VFS can correctly support 1177 * read-only remount without racing. 1178 */ 1179 WARN_ON(atomic_read(&mp->m_active_trans) != 0); 1180 1181 xfs_log_quiesce(mp); 1182 } 1183 1184 STATIC int 1185 xfs_fs_remount( 1186 struct super_block *sb, 1187 int *flags, 1188 char *options) 1189 { 1190 struct xfs_mount *mp = XFS_M(sb); 1191 substring_t args[MAX_OPT_ARGS]; 1192 char *p; 1193 int error; 1194 1195 sync_filesystem(sb); 1196 while ((p = strsep(&options, ",")) != NULL) { 1197 int token; 1198 1199 if (!*p) 1200 continue; 1201 1202 token = match_token(p, tokens, args); 1203 switch (token) { 1204 case Opt_barrier: 1205 mp->m_flags |= XFS_MOUNT_BARRIER; 1206 break; 1207 case Opt_nobarrier: 1208 mp->m_flags &= ~XFS_MOUNT_BARRIER; 1209 break; 1210 case Opt_inode64: 1211 mp->m_maxagi = xfs_set_inode64(mp); 1212 break; 1213 case Opt_inode32: 1214 mp->m_maxagi = xfs_set_inode32(mp); 1215 break; 1216 default: 1217 /* 1218 * Logically we would return an error here to prevent 1219 * users from believing they might have changed 1220 * mount options using remount which can't be changed. 1221 * 1222 * But unfortunately mount(8) adds all options from 1223 * mtab and fstab to the mount arguments in some cases 1224 * so we can't blindly reject options, but have to 1225 * check for each specified option if it actually 1226 * differs from the currently set option and only 1227 * reject it if that's the case. 1228 * 1229 * Until that is implemented we return success for 1230 * every remount request, and silently ignore all 1231 * options that we can't actually change. 1232 */ 1233 #if 0 1234 xfs_info(mp, 1235 "mount option \"%s\" not supported for remount", p); 1236 return -EINVAL; 1237 #else 1238 break; 1239 #endif 1240 } 1241 } 1242 1243 /* ro -> rw */ 1244 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) { 1245 mp->m_flags &= ~XFS_MOUNT_RDONLY; 1246 1247 /* 1248 * If this is the first remount to writeable state we 1249 * might have some superblock changes to update. 1250 */ 1251 if (mp->m_update_flags) { 1252 error = xfs_mount_log_sb(mp, mp->m_update_flags); 1253 if (error) { 1254 xfs_warn(mp, "failed to write sb changes"); 1255 return error; 1256 } 1257 mp->m_update_flags = 0; 1258 } 1259 1260 /* 1261 * Fill out the reserve pool if it is empty. Use the stashed 1262 * value if it is non-zero, otherwise go with the default. 1263 */ 1264 xfs_restore_resvblks(mp); 1265 xfs_log_work_queue(mp); 1266 } 1267 1268 /* rw -> ro */ 1269 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) { 1270 /* 1271 * Before we sync the metadata, we need to free up the reserve 1272 * block pool so that the used block count in the superblock on 1273 * disk is correct at the end of the remount. Stash the current 1274 * reserve pool size so that if we get remounted rw, we can 1275 * return it to the same size. 1276 */ 1277 xfs_save_resvblks(mp); 1278 xfs_quiesce_attr(mp); 1279 mp->m_flags |= XFS_MOUNT_RDONLY; 1280 } 1281 1282 return 0; 1283 } 1284 1285 /* 1286 * Second stage of a freeze. The data is already frozen so we only 1287 * need to take care of the metadata. Once that's done write a dummy 1288 * record to dirty the log in case of a crash while frozen. 1289 */ 1290 STATIC int 1291 xfs_fs_freeze( 1292 struct super_block *sb) 1293 { 1294 struct xfs_mount *mp = XFS_M(sb); 1295 1296 xfs_save_resvblks(mp); 1297 xfs_quiesce_attr(mp); 1298 return -xfs_fs_log_dummy(mp); 1299 } 1300 1301 STATIC int 1302 xfs_fs_unfreeze( 1303 struct super_block *sb) 1304 { 1305 struct xfs_mount *mp = XFS_M(sb); 1306 1307 xfs_restore_resvblks(mp); 1308 xfs_log_work_queue(mp); 1309 return 0; 1310 } 1311 1312 STATIC int 1313 xfs_fs_show_options( 1314 struct seq_file *m, 1315 struct dentry *root) 1316 { 1317 return -xfs_showargs(XFS_M(root->d_sb), m); 1318 } 1319 1320 /* 1321 * This function fills in xfs_mount_t fields based on mount args. 1322 * Note: the superblock _has_ now been read in. 1323 */ 1324 STATIC int 1325 xfs_finish_flags( 1326 struct xfs_mount *mp) 1327 { 1328 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY); 1329 1330 /* Fail a mount where the logbuf is smaller than the log stripe */ 1331 if (xfs_sb_version_haslogv2(&mp->m_sb)) { 1332 if (mp->m_logbsize <= 0 && 1333 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) { 1334 mp->m_logbsize = mp->m_sb.sb_logsunit; 1335 } else if (mp->m_logbsize > 0 && 1336 mp->m_logbsize < mp->m_sb.sb_logsunit) { 1337 xfs_warn(mp, 1338 "logbuf size must be greater than or equal to log stripe size"); 1339 return XFS_ERROR(EINVAL); 1340 } 1341 } else { 1342 /* Fail a mount if the logbuf is larger than 32K */ 1343 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) { 1344 xfs_warn(mp, 1345 "logbuf size for version 1 logs must be 16K or 32K"); 1346 return XFS_ERROR(EINVAL); 1347 } 1348 } 1349 1350 /* 1351 * V5 filesystems always use attr2 format for attributes. 1352 */ 1353 if (xfs_sb_version_hascrc(&mp->m_sb) && 1354 (mp->m_flags & XFS_MOUNT_NOATTR2)) { 1355 xfs_warn(mp, 1356 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.", 1357 MNTOPT_NOATTR2, MNTOPT_ATTR2); 1358 return XFS_ERROR(EINVAL); 1359 } 1360 1361 /* 1362 * mkfs'ed attr2 will turn on attr2 mount unless explicitly 1363 * told by noattr2 to turn it off 1364 */ 1365 if (xfs_sb_version_hasattr2(&mp->m_sb) && 1366 !(mp->m_flags & XFS_MOUNT_NOATTR2)) 1367 mp->m_flags |= XFS_MOUNT_ATTR2; 1368 1369 /* 1370 * prohibit r/w mounts of read-only filesystems 1371 */ 1372 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) { 1373 xfs_warn(mp, 1374 "cannot mount a read-only filesystem as read-write"); 1375 return XFS_ERROR(EROFS); 1376 } 1377 1378 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) && 1379 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) && 1380 !xfs_sb_version_has_pquotino(&mp->m_sb)) { 1381 xfs_warn(mp, 1382 "Super block does not support project and group quota together"); 1383 return XFS_ERROR(EINVAL); 1384 } 1385 1386 return 0; 1387 } 1388 1389 STATIC int 1390 xfs_fs_fill_super( 1391 struct super_block *sb, 1392 void *data, 1393 int silent) 1394 { 1395 struct inode *root; 1396 struct xfs_mount *mp = NULL; 1397 int flags = 0, error = ENOMEM; 1398 1399 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL); 1400 if (!mp) 1401 goto out; 1402 1403 spin_lock_init(&mp->m_sb_lock); 1404 mutex_init(&mp->m_growlock); 1405 atomic_set(&mp->m_active_trans, 0); 1406 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); 1407 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker); 1408 1409 mp->m_super = sb; 1410 sb->s_fs_info = mp; 1411 1412 error = xfs_parseargs(mp, (char *)data); 1413 if (error) 1414 goto out_free_fsname; 1415 1416 sb_min_blocksize(sb, BBSIZE); 1417 sb->s_xattr = xfs_xattr_handlers; 1418 sb->s_export_op = &xfs_export_operations; 1419 #ifdef CONFIG_XFS_QUOTA 1420 sb->s_qcop = &xfs_quotactl_operations; 1421 #endif 1422 sb->s_op = &xfs_super_operations; 1423 1424 if (silent) 1425 flags |= XFS_MFSI_QUIET; 1426 1427 error = xfs_open_devices(mp); 1428 if (error) 1429 goto out_free_fsname; 1430 1431 error = -xfs_init_mount_workqueues(mp); 1432 if (error) 1433 goto out_close_devices; 1434 1435 error = -xfs_icsb_init_counters(mp); 1436 if (error) 1437 goto out_destroy_workqueues; 1438 1439 error = xfs_readsb(mp, flags); 1440 if (error) 1441 goto out_destroy_counters; 1442 1443 error = xfs_finish_flags(mp); 1444 if (error) 1445 goto out_free_sb; 1446 1447 error = xfs_setup_devices(mp); 1448 if (error) 1449 goto out_free_sb; 1450 1451 error = xfs_filestream_mount(mp); 1452 if (error) 1453 goto out_free_sb; 1454 1455 /* 1456 * we must configure the block size in the superblock before we run the 1457 * full mount process as the mount process can lookup and cache inodes. 1458 */ 1459 sb->s_magic = XFS_SB_MAGIC; 1460 sb->s_blocksize = mp->m_sb.sb_blocksize; 1461 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1; 1462 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); 1463 sb->s_max_links = XFS_MAXLINK; 1464 sb->s_time_gran = 1; 1465 set_posix_acl_flag(sb); 1466 1467 /* version 5 superblocks support inode version counters. */ 1468 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5) 1469 sb->s_flags |= MS_I_VERSION; 1470 1471 error = xfs_mountfs(mp); 1472 if (error) 1473 goto out_filestream_unmount; 1474 1475 root = igrab(VFS_I(mp->m_rootip)); 1476 if (!root) { 1477 error = ENOENT; 1478 goto out_unmount; 1479 } 1480 sb->s_root = d_make_root(root); 1481 if (!sb->s_root) { 1482 error = ENOMEM; 1483 goto out_unmount; 1484 } 1485 1486 return 0; 1487 1488 out_filestream_unmount: 1489 xfs_filestream_unmount(mp); 1490 out_free_sb: 1491 xfs_freesb(mp); 1492 out_destroy_counters: 1493 xfs_icsb_destroy_counters(mp); 1494 out_destroy_workqueues: 1495 xfs_destroy_mount_workqueues(mp); 1496 out_close_devices: 1497 xfs_close_devices(mp); 1498 out_free_fsname: 1499 xfs_free_fsname(mp); 1500 kfree(mp); 1501 out: 1502 return -error; 1503 1504 out_unmount: 1505 xfs_filestream_unmount(mp); 1506 xfs_unmountfs(mp); 1507 goto out_free_sb; 1508 } 1509 1510 STATIC struct dentry * 1511 xfs_fs_mount( 1512 struct file_system_type *fs_type, 1513 int flags, 1514 const char *dev_name, 1515 void *data) 1516 { 1517 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super); 1518 } 1519 1520 static long 1521 xfs_fs_nr_cached_objects( 1522 struct super_block *sb, 1523 int nid) 1524 { 1525 return xfs_reclaim_inodes_count(XFS_M(sb)); 1526 } 1527 1528 static long 1529 xfs_fs_free_cached_objects( 1530 struct super_block *sb, 1531 long nr_to_scan, 1532 int nid) 1533 { 1534 return xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan); 1535 } 1536 1537 static const struct super_operations xfs_super_operations = { 1538 .alloc_inode = xfs_fs_alloc_inode, 1539 .destroy_inode = xfs_fs_destroy_inode, 1540 .evict_inode = xfs_fs_evict_inode, 1541 .drop_inode = xfs_fs_drop_inode, 1542 .put_super = xfs_fs_put_super, 1543 .sync_fs = xfs_fs_sync_fs, 1544 .freeze_fs = xfs_fs_freeze, 1545 .unfreeze_fs = xfs_fs_unfreeze, 1546 .statfs = xfs_fs_statfs, 1547 .remount_fs = xfs_fs_remount, 1548 .show_options = xfs_fs_show_options, 1549 .nr_cached_objects = xfs_fs_nr_cached_objects, 1550 .free_cached_objects = xfs_fs_free_cached_objects, 1551 }; 1552 1553 static struct file_system_type xfs_fs_type = { 1554 .owner = THIS_MODULE, 1555 .name = "xfs", 1556 .mount = xfs_fs_mount, 1557 .kill_sb = kill_block_super, 1558 .fs_flags = FS_REQUIRES_DEV, 1559 }; 1560 MODULE_ALIAS_FS("xfs"); 1561 1562 STATIC int __init 1563 xfs_init_zones(void) 1564 { 1565 1566 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); 1567 if (!xfs_ioend_zone) 1568 goto out; 1569 1570 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE, 1571 xfs_ioend_zone); 1572 if (!xfs_ioend_pool) 1573 goto out_destroy_ioend_zone; 1574 1575 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t), 1576 "xfs_log_ticket"); 1577 if (!xfs_log_ticket_zone) 1578 goto out_destroy_ioend_pool; 1579 1580 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t), 1581 "xfs_bmap_free_item"); 1582 if (!xfs_bmap_free_item_zone) 1583 goto out_destroy_log_ticket_zone; 1584 1585 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t), 1586 "xfs_btree_cur"); 1587 if (!xfs_btree_cur_zone) 1588 goto out_destroy_bmap_free_item_zone; 1589 1590 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t), 1591 "xfs_da_state"); 1592 if (!xfs_da_state_zone) 1593 goto out_destroy_btree_cur_zone; 1594 1595 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork"); 1596 if (!xfs_ifork_zone) 1597 goto out_destroy_da_state_zone; 1598 1599 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans"); 1600 if (!xfs_trans_zone) 1601 goto out_destroy_ifork_zone; 1602 1603 xfs_log_item_desc_zone = 1604 kmem_zone_init(sizeof(struct xfs_log_item_desc), 1605 "xfs_log_item_desc"); 1606 if (!xfs_log_item_desc_zone) 1607 goto out_destroy_trans_zone; 1608 1609 /* 1610 * The size of the zone allocated buf log item is the maximum 1611 * size possible under XFS. This wastes a little bit of memory, 1612 * but it is much faster. 1613 */ 1614 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item), 1615 "xfs_buf_item"); 1616 if (!xfs_buf_item_zone) 1617 goto out_destroy_log_item_desc_zone; 1618 1619 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) + 1620 ((XFS_EFD_MAX_FAST_EXTENTS - 1) * 1621 sizeof(xfs_extent_t))), "xfs_efd_item"); 1622 if (!xfs_efd_zone) 1623 goto out_destroy_buf_item_zone; 1624 1625 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) + 1626 ((XFS_EFI_MAX_FAST_EXTENTS - 1) * 1627 sizeof(xfs_extent_t))), "xfs_efi_item"); 1628 if (!xfs_efi_zone) 1629 goto out_destroy_efd_zone; 1630 1631 xfs_inode_zone = 1632 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode", 1633 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD, 1634 xfs_fs_inode_init_once); 1635 if (!xfs_inode_zone) 1636 goto out_destroy_efi_zone; 1637 1638 xfs_ili_zone = 1639 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili", 1640 KM_ZONE_SPREAD, NULL); 1641 if (!xfs_ili_zone) 1642 goto out_destroy_inode_zone; 1643 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item), 1644 "xfs_icr"); 1645 if (!xfs_icreate_zone) 1646 goto out_destroy_ili_zone; 1647 1648 return 0; 1649 1650 out_destroy_ili_zone: 1651 kmem_zone_destroy(xfs_ili_zone); 1652 out_destroy_inode_zone: 1653 kmem_zone_destroy(xfs_inode_zone); 1654 out_destroy_efi_zone: 1655 kmem_zone_destroy(xfs_efi_zone); 1656 out_destroy_efd_zone: 1657 kmem_zone_destroy(xfs_efd_zone); 1658 out_destroy_buf_item_zone: 1659 kmem_zone_destroy(xfs_buf_item_zone); 1660 out_destroy_log_item_desc_zone: 1661 kmem_zone_destroy(xfs_log_item_desc_zone); 1662 out_destroy_trans_zone: 1663 kmem_zone_destroy(xfs_trans_zone); 1664 out_destroy_ifork_zone: 1665 kmem_zone_destroy(xfs_ifork_zone); 1666 out_destroy_da_state_zone: 1667 kmem_zone_destroy(xfs_da_state_zone); 1668 out_destroy_btree_cur_zone: 1669 kmem_zone_destroy(xfs_btree_cur_zone); 1670 out_destroy_bmap_free_item_zone: 1671 kmem_zone_destroy(xfs_bmap_free_item_zone); 1672 out_destroy_log_ticket_zone: 1673 kmem_zone_destroy(xfs_log_ticket_zone); 1674 out_destroy_ioend_pool: 1675 mempool_destroy(xfs_ioend_pool); 1676 out_destroy_ioend_zone: 1677 kmem_zone_destroy(xfs_ioend_zone); 1678 out: 1679 return -ENOMEM; 1680 } 1681 1682 STATIC void 1683 xfs_destroy_zones(void) 1684 { 1685 /* 1686 * Make sure all delayed rcu free are flushed before we 1687 * destroy caches. 1688 */ 1689 rcu_barrier(); 1690 kmem_zone_destroy(xfs_icreate_zone); 1691 kmem_zone_destroy(xfs_ili_zone); 1692 kmem_zone_destroy(xfs_inode_zone); 1693 kmem_zone_destroy(xfs_efi_zone); 1694 kmem_zone_destroy(xfs_efd_zone); 1695 kmem_zone_destroy(xfs_buf_item_zone); 1696 kmem_zone_destroy(xfs_log_item_desc_zone); 1697 kmem_zone_destroy(xfs_trans_zone); 1698 kmem_zone_destroy(xfs_ifork_zone); 1699 kmem_zone_destroy(xfs_da_state_zone); 1700 kmem_zone_destroy(xfs_btree_cur_zone); 1701 kmem_zone_destroy(xfs_bmap_free_item_zone); 1702 kmem_zone_destroy(xfs_log_ticket_zone); 1703 mempool_destroy(xfs_ioend_pool); 1704 kmem_zone_destroy(xfs_ioend_zone); 1705 1706 } 1707 1708 STATIC int __init 1709 xfs_init_workqueues(void) 1710 { 1711 /* 1712 * The allocation workqueue can be used in memory reclaim situations 1713 * (writepage path), and parallelism is only limited by the number of 1714 * AGs in all the filesystems mounted. Hence use the default large 1715 * max_active value for this workqueue. 1716 */ 1717 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0); 1718 if (!xfs_alloc_wq) 1719 return -ENOMEM; 1720 1721 return 0; 1722 } 1723 1724 STATIC void 1725 xfs_destroy_workqueues(void) 1726 { 1727 destroy_workqueue(xfs_alloc_wq); 1728 } 1729 1730 STATIC int __init 1731 init_xfs_fs(void) 1732 { 1733 int error; 1734 1735 printk(KERN_INFO XFS_VERSION_STRING " with " 1736 XFS_BUILD_OPTIONS " enabled\n"); 1737 1738 xfs_dir_startup(); 1739 1740 error = xfs_init_zones(); 1741 if (error) 1742 goto out; 1743 1744 error = xfs_init_workqueues(); 1745 if (error) 1746 goto out_destroy_zones; 1747 1748 error = xfs_mru_cache_init(); 1749 if (error) 1750 goto out_destroy_wq; 1751 1752 error = xfs_buf_init(); 1753 if (error) 1754 goto out_mru_cache_uninit; 1755 1756 error = xfs_init_procfs(); 1757 if (error) 1758 goto out_buf_terminate; 1759 1760 error = xfs_sysctl_register(); 1761 if (error) 1762 goto out_cleanup_procfs; 1763 1764 error = xfs_qm_init(); 1765 if (error) 1766 goto out_sysctl_unregister; 1767 1768 error = register_filesystem(&xfs_fs_type); 1769 if (error) 1770 goto out_qm_exit; 1771 return 0; 1772 1773 out_qm_exit: 1774 xfs_qm_exit(); 1775 out_sysctl_unregister: 1776 xfs_sysctl_unregister(); 1777 out_cleanup_procfs: 1778 xfs_cleanup_procfs(); 1779 out_buf_terminate: 1780 xfs_buf_terminate(); 1781 out_mru_cache_uninit: 1782 xfs_mru_cache_uninit(); 1783 out_destroy_wq: 1784 xfs_destroy_workqueues(); 1785 out_destroy_zones: 1786 xfs_destroy_zones(); 1787 out: 1788 return error; 1789 } 1790 1791 STATIC void __exit 1792 exit_xfs_fs(void) 1793 { 1794 xfs_qm_exit(); 1795 unregister_filesystem(&xfs_fs_type); 1796 xfs_sysctl_unregister(); 1797 xfs_cleanup_procfs(); 1798 xfs_buf_terminate(); 1799 xfs_mru_cache_uninit(); 1800 xfs_destroy_workqueues(); 1801 xfs_destroy_zones(); 1802 } 1803 1804 module_init(init_xfs_fs); 1805 module_exit(exit_xfs_fs); 1806 1807 MODULE_AUTHOR("Silicon Graphics, Inc."); 1808 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); 1809 MODULE_LICENSE("GPL"); 1810