| blob | df0d4572d70a8a7b18fdfbbd62d4bfb2caccfc32 |
1 /*
2 * Copyright (c) 2000-2003,2005 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 */
60 /*
61 * Used in xfs_itruncate(). This is the maximum number of extents
62 * freed from a file in a single transaction.
63 */
64 #define XFS_ITRUNC_MAX_EXTENTS 2
72 #ifdef DEBUG
73 /*
74 * Make sure that the extents in the given memory buffer
75 * are valid.
76 */
77 STATIC void
82 xfs_exntfmt_t fmt)
83 {
84 xfs_bmbt_irec_t irec;
85 xfs_bmbt_rec_t rec;
93 else
97 ep++;
98 }
99 }
101 #define xfs_validate_extents(ep, nrecs, disk, fmt)
104 /*
105 * Check that none of the inode's in the buffer have a next
106 * unlinked field of 0.
107 */
108 #if defined(DEBUG)
109 void
113 {
125 "Detected a bogus zero next_unlinked field in incore inode buffer 0x%p. About to pop an ASSERT.",
126 bp);
128 }
129 }
130 }
131 #endif
133 /*
134 * This routine is called to map an inode number within a file
135 * system to the buffer containing the on-disk version of the
136 * inode. It returns a pointer to the buffer containing the
137 * on-disk inode in the bpp parameter, and in the dip parameter
138 * it returns a pointer to the on-disk inode within that buffer.
139 *
140 * If a non-zero error is returned, then the contents of bpp and
141 * dipp are undefined.
142 *
143 * Use xfs_imap() to determine the size and location of the
144 * buffer to read from disk.
145 */
146 STATIC int
150 xfs_ino_t ino,
154 {
156 xfs_imap_t imap;
161 /*
162 * Call the space managment code to find the location of the
163 * inode on disk.
164 */
169 "xfs_inotobp: xfs_imap() returned an "
172 }
174 /*
175 * If the inode number maps to a block outside the bounds of the
176 * file system then return NULL rather than calling read_buf
177 * and panicing when we get an error from the driver.
178 */
182 "xfs_inotobp: inode number (%llu + %d) maps to a block outside the bounds "
187 }
189 /*
190 * Read in the buffer. If tp is NULL, xfs_trans_read_buf() will
191 * default to just a read_buf() call.
192 */
198 "xfs_inotobp: xfs_trans_read_buf() returned an "
201 }
203 di_ok =
207 XFS_RANDOM_ITOBP_INOTOBP))) {
211 "xfs_inotobp: XFS_TEST_ERROR() returned an "
214 }
218 /*
219 * Set *dipp to point to the on-disk inode in the buffer.
220 */
225 }
228 /*
229 * This routine is called to map an inode to the buffer containing
230 * the on-disk version of the inode. It returns a pointer to the
231 * buffer containing the on-disk inode in the bpp parameter, and in
232 * the dip parameter it returns a pointer to the on-disk inode within
233 * that buffer.
234 *
235 * If a non-zero error is returned, then the contents of bpp and
236 * dipp are undefined.
237 *
238 * If the inode is new and has not yet been initialized, use xfs_imap()
239 * to determine the size and location of the buffer to read from disk.
240 * If the inode has already been mapped to its buffer and read in once,
241 * then use the mapping information stored in the inode rather than
242 * calling xfs_imap(). This allows us to avoid the overhead of looking
243 * at the inode btree for small block file systems (see xfs_dilocate()).
244 * We can tell whether the inode has been mapped in before by comparing
245 * its disk block address to 0. Only uninitialized inodes will have
246 * 0 for the disk block address.
247 */
248 int
255 xfs_daddr_t bno)
256 {
259 xfs_imap_t imap;
260 #ifdef __KERNEL__
263 #endif
266 /*
267 * Call the space management code to find the location of the
268 * inode on disk.
269 */
274 }
276 /*
277 * If the inode number maps to a block outside the bounds
278 * of the file system then return NULL rather than calling
279 * read_buf and panicing when we get an error from the
280 * driver.
281 */
284 #ifdef DEBUG
286 "(imap.im_blkno (0x%llx) "
287 "+ imap.im_len (0x%llx)) > "
288 " XFS_FSB_TO_BB(mp, "
295 }
297 /*
298 * Fill in the fields in the inode that will be used to
299 * map the inode to its buffer from now on.
300 */
305 /*
306 * We've already mapped the inode once, so just use the
307 * mapping that we saved the first time.
308 */
312 }
315 /*
316 * Read in the buffer. If tp is NULL, xfs_trans_read_buf() will
317 * default to just a read_buf() call.
318 */
323 #ifdef DEBUG
325 "xfs_trans_read_buf() returned error %d, "
331 }
332 #ifdef __KERNEL__
333 /*
334 * Validate the magic number and version of every inode in the buffer
335 * (if DEBUG kernel) or the first inode in the buffer, otherwise.
336 */
337 #ifdef DEBUG
339 #else
341 #endif
351 XFS_RANDOM_ITOBP_INOTOBP))) {
352 #ifdef DEBUG
357 #endif
362 }
363 }
368 /*
369 * Mark the buffer as an inode buffer now that it looks good
370 */
373 /*
374 * Set *dipp to point to the on-disk inode in the buffer.
375 */
379 }
381 /*
382 * Move inode type and inode format specific information from the
383 * on-disk inode to the in-core inode. For fifos, devs, and sockets
384 * this means set if_rdev to the proper value. For files, directories,
385 * and symlinks this means to bring in the in-line data or extent
386 * pointers. For a file in B-tree format, only the root is immediately
387 * brought in-core. The rest will be in-lined in if_extents when it
388 * is first referenced (see xfs_iread_extents()).
389 */
390 STATIC int
394 {
398 xfs_fsize_t di_size;
408 "corrupt dinode %Lu, extent total = %d, nblocks = %Lu."
418 }
420 if (unlikely(INT_GET(dip->di_core.di_forkoff, ARCH_CONVERT) > ip->i_mount->m_sb.sb_inodesize)) {
422 "corrupt dinode %Lu, forkoff = 0x%x."
429 }
440 }
450 /*
451 * no local regular files yet
452 */
458 XFS_ERRLEVEL_LOW,
461 }
470 XFS_ERRLEVEL_LOW,
473 }
488 }
494 }
497 }
519 }
524 }
526 }
528 /*
529 * The file is in-lined in the on-disk inode.
530 * If it fits into if_inline_data, then copy
531 * it there, otherwise allocate a buffer for it
532 * and copy the data there. Either way, set
533 * if_data to point at the data.
534 * If we allocate a buffer for the data, make
535 * sure that its size is a multiple of 4 and
536 * record the real size in i_real_bytes.
537 */
538 STATIC int
544 {
548 /*
549 * If the size is unreasonable, then something
550 * is wrong and we just bail out rather than crash in
551 * kmem_alloc() or memcpy() below.
552 */
561 }
571 }
579 }
581 /*
582 * The file consists of a set of extents all
583 * of which fit into the on-disk inode.
584 * If there are few enough extents to fit into
585 * the if_inline_ext, then copy them there.
586 * Otherwise allocate a buffer for them and copy
587 * them into it. Either way, set if_extents
588 * to point at the extents.
589 */
590 STATIC int
595 {
607 /*
608 * If the number of extents is unreasonable, then something
609 * is wrong and we just bail out rather than crash in
610 * kmem_alloc() or memcpy() below.
611 */
619 }
630 }
639 ARCH_CONVERT);
641 ARCH_CONVERT);
642 }
644 whichfork);
650 XFS_ERRLEVEL_LOW,
653 }
654 }
657 }
659 /*
660 * The file has too many extents to fit into
661 * the inode, so they are in B-tree format.
662 * Allocate a buffer for the root of the B-tree
663 * and copy the root into it. The i_extents
664 * field will remain NULL until all of the
665 * extents are read in (when they are needed).
666 */
667 STATIC int
672 {
675 /* REFERENCED */
684 /*
685 * blow out if -- fork has less extents than can fit in
686 * fork (fork shouldn't be a btree format), root btree
687 * block has more records than can fit into the fork,
688 * or the number of extents is greater than the number of
689 * blocks.
690 */
701 }
706 /*
707 * Copy and convert from the on-disk structure
708 * to the in-memory structure.
709 */
716 }
718 /*
719 * xfs_xlate_dinode_core - translate an xfs_inode_core_t between ondisk
720 * and native format
721 *
722 * buf = on-disk representation
723 * dip = native representation
724 * dir = direction - +ve -> disk to native
725 * -ve -> native to disk
726 */
727 void
729 xfs_caddr_t buf,
732 {
755 }
782 }
784 STATIC uint
787 __uint16_t di_flags)
788 {
812 }
815 }
817 uint
820 {
825 }
827 uint
830 {
833 }
835 /*
836 * Given a mount structure and an inode number, return a pointer
837 * to a newly allocated in-core inode coresponding to the given
838 * inode number.
839 *
840 * Initialize the inode's attributes and extent pointers if it
841 * already has them (it will not if the inode has no links).
842 */
843 int
847 xfs_ino_t ino,
849 xfs_daddr_t bno)
850 {
862 /*
863 * Get pointer's to the on-disk inode and the buffer containing it.
864 * If the inode number refers to a block outside the file system
865 * then xfs_itobp() will return NULL. In this case we should
866 * return NULL as well. Set i_blkno to 0 so that xfs_itobp() will
867 * know that this is a new incore inode.
868 */
874 }
876 /*
877 * Initialize inode's trace buffers.
878 * Do this before xfs_iformat in case it adds entries.
879 */
880 #ifdef XFS_BMAP_TRACE
882 #endif
883 #ifdef XFS_BMBT_TRACE
885 #endif
886 #ifdef XFS_RW_TRACE
888 #endif
889 #ifdef XFS_ILOCK_TRACE
891 #endif
892 #ifdef XFS_DIR2_TRACE
894 #endif
896 /*
897 * If we got something that isn't an inode it means someone
898 * (nfs or dmi) has a stale handle.
899 */
903 #ifdef DEBUG
905 "dip->di_core.di_magic (0x%x) != "
908 XFS_DINODE_MAGIC);
911 }
913 /*
914 * If the on-disk inode is already linked to a directory
915 * entry, copy all of the inode into the in-core inode.
916 * xfs_iformat() handles copying in the inode format
917 * specific information.
918 * Otherwise, just get the truly permanent information.
919 */
927 #ifdef DEBUG
930 error);
933 }
939 /*
940 * Make sure to pull in the mode here as well in
941 * case the inode is released without being used.
942 * This ensures that xfs_inactive() will see that
943 * the inode is already free and not try to mess
944 * with the uninitialized part of it.
945 */
947 /*
948 * Initialize the per-fork minima and maxima for a new
949 * inode here. xfs_iformat will do it for old inodes.
950 */
953 }
957 /*
958 * The inode format changed when we moved the link count and
959 * made it 32 bits long. If this is an old format inode,
960 * convert it in memory to look like a new one. If it gets
961 * flushed to disk we will convert back before flushing or
962 * logging it. We zero out the new projid field and the old link
963 * count field. We'll handle clearing the pad field (the remains
964 * of the old uuid field) when we actually convert the inode to
965 * the new format. We don't change the version number so that we
966 * can distinguish this from a real new format inode.
967 */
972 }
976 /*
977 * Mark the buffer containing the inode as something to keep
978 * around for a while. This helps to keep recently accessed
979 * meta-data in-core longer.
980 */
983 /*
984 * Use xfs_trans_brelse() to release the buffer containing the
985 * on-disk inode, because it was acquired with xfs_trans_read_buf()
986 * in xfs_itobp() above. If tp is NULL, this is just a normal
987 * brelse(). If we're within a transaction, then xfs_trans_brelse()
988 * will only release the buffer if it is not dirty within the
989 * transaction. It will be OK to release the buffer in this case,
990 * because inodes on disk are never destroyed and we will be
991 * locking the new in-core inode before putting it in the hash
992 * table where other processes can find it. Thus we don't have
993 * to worry about the inode being changed just because we released
994 * the buffer.
995 */
999 }
1001 /*
1002 * Read in extents from a btree-format inode.
1003 * Allocate and fill in if_extents. Real work is done in xfs_bmap.c.
1004 */
1005 int
1010 {
1019 }
1022 /*
1023 * We know that the size is valid (it's checked in iformat_btree)
1024 */
1037 }
1041 }
1043 /*
1044 * Allocate an inode on disk and return a copy of its in-core version.
1045 * The in-core inode is locked exclusively. Set mode, nlink, and rdev
1046 * appropriately within the inode. The uid and gid for the inode are
1047 * set according to the contents of the given cred structure.
1048 *
1049 * Use xfs_dialloc() to allocate the on-disk inode. If xfs_dialloc()
1050 * has a free inode available, call xfs_iget()
1051 * to obtain the in-core version of the allocated inode. Finally,
1052 * fill in the inode and log its initial contents. In this case,
1053 * ialloc_context would be set to NULL and call_again set to false.
1054 *
1055 * If xfs_dialloc() does not have an available inode,
1056 * it will replenish its supply by doing an allocation. Since we can
1057 * only do one allocation within a transaction without deadlocks, we
1058 * must commit the current transaction before returning the inode itself.
1059 * In this case, therefore, we will set call_again to true and return.
1060 * The caller should then commit the current transaction, start a new
1061 * transaction, and call xfs_ialloc() again to actually get the inode.
1062 *
1063 * To ensure that some other process does not grab the inode that
1064 * was allocated during the first call to xfs_ialloc(), this routine
1065 * also returns the [locked] bp pointing to the head of the freelist
1066 * as ialloc_context. The caller should hold this buffer across
1067 * the commit and pass it back into this routine on the second call.
1068 */
1069 int
1073 mode_t mode,
1074 xfs_nlink_t nlink,
1075 xfs_dev_t rdev,
1077 xfs_prid_t prid,
1082 {
1083 xfs_ino_t ino;
1086 uint flags;
1089 /*
1090 * Call the space management code to pick
1091 * the on-disk inode to be allocated.
1092 */
1097 }
1101 }
1104 /*
1105 * Get the in-core inode with the lock held exclusively.
1106 * This is because we're setting fields here we need
1107 * to prevent others from looking at until we're done.
1108 */
1113 }
1126 /*
1127 * If the superblock version is up to where we support new format
1128 * inodes and this is currently an old format inode, then change
1129 * the inode version number now. This way we only do the conversion
1130 * here rather than here and in the flush/logging code.
1131 */
1135 /*
1136 * We've already zeroed the old link count, the projid field,
1137 * and the pad field.
1138 */
1139 }
1141 /*
1142 * Project ids won't be stored on disk if we are using a version 1 inode.
1143 */
1151 }
1152 }
1154 /*
1155 * If the group ID of the new file does not match the effective group
1156 * ID or one of the supplementary group IDs, the S_ISGID bit is cleared
1157 * (and only if the irix_sgid_inherit compatibility variable is set).
1158 */
1163 }
1169 /*
1170 * di_gen will have been taken care of in xfs_iread.
1171 */
1199 }
1200 }
1202 xfs_inherit_noatime)
1205 xfs_inherit_nodump)
1208 xfs_inherit_sync)
1211 xfs_inherit_nosymlinks)
1216 }
1217 /* FALLTHROUGH */
1226 }
1227 /*
1228 * Attribute fork settings for new inode.
1229 */
1233 /*
1234 * Log the new values stuffed into the inode.
1235 */
1238 /* now that we have an i_mode we can set Linux inode ops (& unlock) */
1243 }
1245 /*
1246 * Check to make sure that there are no blocks allocated to the
1247 * file beyond the size of the file. We don't check this for
1248 * files with fixed size extents or real time extents, but we
1249 * at least do it for regular files.
1250 */
1251 #ifdef DEBUG
1252 void
1256 xfs_fsize_t isize)
1257 {
1258 xfs_fileoff_t map_first;
1270 /*
1271 * The filesystem could be shutting down, so bmapi may return
1272 * an error.
1273 */
1277 map_first),
1279 NULL))
1283 }
1286 /*
1287 * Calculate the last possible buffered byte in a file. This must
1288 * include data that was buffered beyond the EOF by the write code.
1289 * This also needs to deal with overflowing the xfs_fsize_t type
1290 * which can happen for sizes near the limit.
1291 *
1292 * We also need to take into account any blocks beyond the EOF. It
1293 * may be the case that they were buffered by a write which failed.
1294 * In that case the pages will still be in memory, but the inode size
1295 * will never have been updated.
1296 */
1297 xfs_fsize_t
1300 {
1302 xfs_fsize_t last_byte;
1303 xfs_fileoff_t last_block;
1304 xfs_fileoff_t size_last_block;
1310 /*
1311 * Only check for blocks beyond the EOF if the extents have
1312 * been read in. This eliminates the need for the inode lock,
1313 * and it also saves us from looking when it really isn't
1314 * necessary.
1315 */
1318 XFS_DATA_FORK);
1321 }
1324 }
1331 }
1335 }
1337 }
1339 #if defined(XFS_RW_TRACE)
1340 STATIC void
1345 xfs_fsize_t new_size,
1346 xfs_off_t toss_start,
1347 xfs_off_t toss_finish)
1348 {
1351 }
1370 }
1371 #else
1372 #define xfs_itrunc_trace(tag, ip, flag, new_size, toss_start, toss_finish)
1373 #endif
1375 /*
1376 * Start the truncation of the file to new_size. The new size
1377 * must be smaller than the current size. This routine will
1378 * clear the buffer and page caches of file data in the removed
1379 * range, and xfs_itruncate_finish() will remove the underlying
1380 * disk blocks.
1381 *
1382 * The inode must have its I/O lock locked EXCLUSIVELY, and it
1383 * must NOT have the inode lock held at all. This is because we're
1384 * calling into the buffer/page cache code and we can't hold the
1385 * inode lock when we do so.
1386 *
1387 * The flags parameter can have either the value XFS_ITRUNC_DEFINITE
1388 * or XFS_ITRUNC_MAYBE. The XFS_ITRUNC_MAYBE value should be used
1389 * in the case that the caller is locking things out of order and
1390 * may not be able to call xfs_itruncate_finish() with the inode lock
1391 * held without dropping the I/O lock. If the caller must drop the
1392 * I/O lock before calling xfs_itruncate_finish(), then xfs_itruncate_start()
1393 * must be called again with all the same restrictions as the initial
1394 * call.
1395 */
1396 void
1399 uint flags,
1400 xfs_fsize_t new_size)
1401 {
1402 xfs_fsize_t last_byte;
1403 xfs_off_t toss_start;
1414 /*
1415 * Call VOP_TOSS_PAGES() or VOP_FLUSHINVAL_PAGES() to get rid of pages and buffers
1416 * overlapping the region being removed. We have to use
1417 * the less efficient VOP_FLUSHINVAL_PAGES() in the case that the
1418 * caller may not be able to finish the truncate without
1419 * dropping the inode's I/O lock. Make sure
1420 * to catch any pages brought in by buffers overlapping
1421 * the EOF by searching out beyond the isize by our
1422 * block size. We round new_size up to a block boundary
1423 * so that we don't toss things on the same block as
1424 * new_size but before it.
1425 *
1426 * Before calling VOP_TOSS_PAGES() or VOP_FLUSHINVAL_PAGES(), make sure to
1427 * call remapf() over the same region if the file is mapped.
1428 * This frees up mapped file references to the pages in the
1429 * given range and for the VOP_FLUSHINVAL_PAGES() case it ensures
1430 * that we get the latest mapped changes flushed out.
1431 */
1435 /*
1436 * The place to start tossing is beyond our maximum
1437 * file size, so there is no way that the data extended
1438 * out there.
1439 */
1441 }
1444 last_byte);
1450 }
1451 }
1453 #ifdef DEBUG
1456 }
1457 #endif
1458 }
1460 /*
1461 * Shrink the file to the given new_size. The new
1462 * size must be smaller than the current size.
1463 * This will free up the underlying blocks
1464 * in the removed range after a call to xfs_itruncate_start()
1465 * or xfs_atruncate_start().
1466 *
1467 * The transaction passed to this routine must have made
1468 * a permanent log reservation of at least XFS_ITRUNCATE_LOG_RES.
1469 * This routine may commit the given transaction and
1470 * start new ones, so make sure everything involved in
1471 * the transaction is tidy before calling here.
1472 * Some transaction will be returned to the caller to be
1473 * committed. The incoming transaction must already include
1474 * the inode, and both inode locks must be held exclusively.
1475 * The inode must also be "held" within the transaction. On
1476 * return the inode will be "held" within the returned transaction.
1477 * This routine does NOT require any disk space to be reserved
1478 * for it within the transaction.
1479 *
1480 * The fork parameter must be either xfs_attr_fork or xfs_data_fork,
1481 * and it indicates the fork which is to be truncated. For the
1482 * attribute fork we only support truncation to size 0.
1483 *
1484 * We use the sync parameter to indicate whether or not the first
1485 * transaction we perform might have to be synchronous. For the attr fork,
1486 * it needs to be so if the unlink of the inode is not yet known to be
1487 * permanent in the log. This keeps us from freeing and reusing the
1488 * blocks of the attribute fork before the unlink of the inode becomes
1489 * permanent.
1490 *
1491 * For the data fork, we normally have to run synchronously if we're
1492 * being called out of the inactive path or we're being called
1493 * out of the create path where we're truncating an existing file.
1494 * Either way, the truncate needs to be sync so blocks don't reappear
1495 * in the file with altered data in case of a crash. wsync filesystems
1496 * can run the first case async because anything that shrinks the inode
1497 * has to run sync so by the time we're called here from inactive, the
1498 * inode size is permanently set to 0.
1499 *
1500 * Calls from the truncate path always need to be sync unless we're
1501 * in a wsync filesystem and the file has already been unlinked.
1502 *
1503 * The caller is responsible for correctly setting the sync parameter.
1504 * It gets too hard for us to guess here which path we're being called
1505 * out of just based on inode state.
1506 */
1507 int
1511 xfs_fsize_t new_size,
1514 {
1515 xfs_fsblock_t first_block;
1516 xfs_fileoff_t first_unmap_block;
1517 xfs_fileoff_t last_block;
1523 xfs_bmap_free_t free_list;
1540 /*
1541 * We only support truncating the entire attribute fork.
1542 */
1545 }
1548 /*
1549 * The first thing we do is set the size to new_size permanently
1550 * on disk. This way we don't have to worry about anyone ever
1551 * being able to look at the data being freed even in the face
1552 * of a crash. What we're getting around here is the case where
1553 * we free a block, it is allocated to another file, it is written
1554 * to, and then we crash. If the new data gets written to the
1555 * file but the log buffers containing the free and reallocation
1556 * don't, then we'd end up with garbage in the blocks being freed.
1557 * As long as we make the new_size permanent before actually
1558 * freeing any blocks it doesn't matter if they get writtten to.
1559 *
1560 * The callers must signal into us whether or not the size
1561 * setting here must be synchronous. There are a few cases
1562 * where it doesn't have to be synchronous. Those cases
1563 * occur if the file is unlinked and we know the unlink is
1564 * permanent or if the blocks being truncated are guaranteed
1565 * to be beyond the inode eof (regardless of the link count)
1566 * and the eof value is permanent. Both of these cases occur
1567 * only on wsync-mounted filesystems. In those cases, we're
1568 * guaranteed that no user will ever see the data in the blocks
1569 * that are being truncated so the truncate can run async.
1570 * In the free beyond eof case, the file may wind up with
1571 * more blocks allocated to it than it needs if we crash
1572 * and that won't get fixed until the next time the file
1573 * is re-opened and closed but that's ok as that shouldn't
1574 * be too many blocks.
1575 *
1576 * However, we can't just make all wsync xactions run async
1577 * because there's one call out of the create path that needs
1578 * to run sync where it's truncating an existing file to size
1579 * 0 whose size is > 0.
1580 *
1581 * It's probably possible to come up with a test in this
1582 * routine that would correctly distinguish all the above
1583 * cases from the values of the function parameters and the
1584 * inode state but for sanity's sake, I've decided to let the
1585 * layers above just tell us. It's simpler to correctly figure
1586 * out in the layer above exactly under what conditions we
1587 * can run async and I think it's easier for others read and
1588 * follow the logic in case something has to be changed.
1589 * cscope is your friend -- rcc.
1590 *
1591 * The attribute fork is much simpler.
1592 *
1593 * For the attribute fork we allow the caller to tell us whether
1594 * the unlink of the inode that led to this call is yet permanent
1595 * in the on disk log. If it is not and we will be freeing extents
1596 * in this inode then we make the first transaction synchronous
1597 * to make sure that the unlink is permanent by the time we free
1598 * the blocks.
1599 */
1604 }
1609 }
1615 /*
1616 * Since it is possible for space to become allocated beyond
1617 * the end of the file (in a crash where the space is allocated
1618 * but the inode size is not yet updated), simply remove any
1619 * blocks which show up between the new EOF and the maximum
1620 * possible file size. If the first block to be removed is
1621 * beyond the maximum file size (ie it is the same as last_block),
1622 * then there is nothing to do.
1623 */
1631 }
1633 /*
1634 * Free up up to XFS_ITRUNC_MAX_EXTENTS. xfs_bunmapi()
1635 * will tell us whether it freed the entire range or
1636 * not. If this is a synchronous mount (wsync),
1637 * then we can tell bunmapi to keep all the
1638 * transactions asynchronous since the unlink
1639 * transaction that made this inode inactive has
1640 * already hit the disk. There's no danger of
1641 * the freed blocks being reused, there being a
1642 * crash, and the reused blocks suddenly reappearing
1643 * in this file with garbage in them once recovery
1644 * runs.
1645 */
1648 unmap_len,
1651 XFS_ITRUNC_MAX_EXTENTS,
1654 /*
1655 * If the bunmapi call encounters an error,
1656 * return to the caller where the transaction
1657 * can be properly aborted. We just need to
1658 * make sure we're not holding any resources
1659 * that we were not when we came in.
1660 */
1663 }
1665 /*
1666 * Duplicate the transaction that has the permanent
1667 * reservation and commit the old transaction.
1668 */
1673 /*
1674 * If the bmap finish call encounters an error,
1675 * return to the caller where the transaction
1676 * can be properly aborted. We just need to
1677 * make sure we're not holding any resources
1678 * that we were not when we came in.
1679 *
1680 * Aborting from this point might lose some
1681 * blocks in the file system, but oh well.
1682 */
1685 /*
1686 * If the passed in transaction committed
1687 * in xfs_bmap_finish(), then we want to
1688 * add the inode to this one before returning.
1689 * This keeps things simple for the higher
1690 * level code, because it always knows that
1691 * the inode is locked and held in the
1692 * transaction that returns to it whether
1693 * errors occur or not. We don't mark the
1694 * inode dirty so that this transaction can
1695 * be easily aborted if possible.
1696 */
1700 }
1702 }
1705 /*
1706 * The first xact was committed,
1707 * so add the inode to the new one.
1708 * Mark it dirty so it will be logged
1709 * and moved forward in the log as
1710 * part of every commit.
1711 */
1716 }
1721 XFS_TRANS_PERM_LOG_RES,
1722 XFS_ITRUNCATE_LOG_COUNT);
1723 /*
1724 * Add the inode being truncated to the next chained
1725 * transaction.
1726 */
1731 }
1732 /*
1733 * Only update the size in the case of the data fork, but
1734 * always re-log the inode so that our permanent transaction
1735 * can keep on rolling it forward in the log.
1736 */
1740 }
1750 }
1753 /*
1754 * xfs_igrow_start
1755 *
1756 * Do the first part of growing a file: zero any data in the last
1757 * block that is beyond the old EOF. We need to do this before
1758 * the inode is joined to the transaction to modify the i_size.
1759 * That way we can drop the inode lock and call into the buffer
1760 * cache to get the buffer mapping the EOF.
1761 */
1762 int
1765 xfs_fsize_t new_size,
1767 {
1768 xfs_fsize_t isize;
1777 /*
1778 * Zero any pages that may have been created by
1779 * xfs_write_file() beyond the end of the file
1780 * and any blocks between the old and new file sizes.
1781 */
1783 new_size);
1785 }
1787 /*
1788 * xfs_igrow_finish
1789 *
1790 * This routine is called to extend the size of a file.
1791 * The inode must have both the iolock and the ilock locked
1792 * for update and it must be a part of the current transaction.
1793 * The xfs_igrow_start() function must have been called previously.
1794 * If the change_flag is not zero, the inode change timestamp will
1795 * be updated.
1796 */
1797 void
1801 xfs_fsize_t new_size,
1803 {
1809 /*
1810 * Update the file size. Update the inode change timestamp
1811 * if change_flag set.
1812 */
1818 }
1821 /*
1822 * This is called when the inode's link count goes to 0.
1823 * We place the on-disk inode on a list in the AGI. It
1824 * will be pulled from this list when the inode is freed.
1825 */
1826 int
1830 {
1836 xfs_agnumber_t agno;
1837 xfs_daddr_t agdaddr;
1838 xfs_agino_t agino;
1853 /*
1854 * Get the agi buffer first. It ensures lock ordering
1855 * on the list.
1856 */
1861 }
1862 /*
1863 * Validate the magic number of the agi block.
1864 */
1866 agi_ok =
1870 XFS_RANDOM_IUNLINK))) {
1874 }
1875 /*
1876 * Get the index into the agi hash table for the
1877 * list this inode will go on.
1878 */
1886 /*
1887 * There is already another inode in the bucket we need
1888 * to add ourselves to. Add us at the front of the list.
1889 * Here we put the head pointer into our next pointer,
1890 * and then we fall through to point the head at us.
1891 */
1895 }
1898 /* both on-disk, don't endian flip twice */
1906 }
1908 /*
1909 * Point the bucket head pointer at the inode being inserted.
1910 */
1918 }
1920 /*
1921 * Pull the on-disk inode from the AGI unlinked list.
1922 */
1923 STATIC int
1927 {
1928 xfs_ino_t next_ino;
1934 xfs_agnumber_t agno;
1935 xfs_daddr_t agdaddr;
1936 xfs_agino_t agino;
1937 xfs_agino_t next_agino;
1945 /*
1946 * First pull the on-disk inode from the AGI unlinked list.
1947 */
1953 /*
1954 * Get the agi buffer first. It ensures lock ordering
1955 * on the list.
1956 */
1964 }
1965 /*
1966 * Validate the magic number of the agi block.
1967 */
1969 agi_ok =
1973 XFS_RANDOM_IUNLINK_REMOVE))) {
1981 }
1982 /*
1983 * Get the index into the agi hash table for the
1984 * list this inode will go on.
1985 */
1993 /*
1994 * We're at the head of the list. Get the inode's
1995 * on-disk buffer to see if there is anyone after us
1996 * on the list. Only modify our next pointer if it
1997 * is not already NULLAGINO. This saves us the overhead
1998 * of dealing with the buffer when there is no need to
1999 * change it.
2000 */
2007 }
2020 }
2021 /*
2022 * Point the bucket head pointer at the next inode.
2023 */
2032 /*
2033 * We need to search the list for the inode being freed.
2034 */
2038 /*
2039 * If the last inode wasn't the one pointing to
2040 * us, then release its buffer since we're not
2041 * going to do anything with it.
2042 */
2045 }
2054 }
2058 }
2059 /*
2060 * Now last_ibp points to the buffer previous to us on
2061 * the unlinked list. Pull us from the list.
2062 */
2069 }
2083 }
2084 /*
2085 * Point the previous inode on the list to the next inode.
2086 */
2094 }
2096 }
2099 {
2103 }
2105 STATIC void
2109 xfs_ino_t inum)
2110 {
2116 xfs_daddr_t blkno;
2133 }
2142 /*
2143 * Look for each inode in memory and attempt to lock it,
2144 * we can be racing with flush and tail pushing here.
2145 * any inode we get the locks on, add to an array of
2146 * inode items to process later.
2147 *
2148 * The get the buffer lock, we could beat a flush
2149 * or tail pushing thread to the lock here, in which
2150 * case they will go looking for the inode buffer
2151 * and fail, we need some other form of interlock
2152 * here.
2153 */
2161 }
2163 /* Inode not in memory or we found it already,
2164 * nothing to do
2165 */
2169 }
2174 }
2176 /* If we can get the locks then add it to the
2177 * list, otherwise by the time we get the bp lock
2178 * below it will already be attached to the
2179 * inode buffer.
2180 */
2182 /* This inode will already be locked - by us, lets
2183 * keep it that way.
2184 */
2194 }
2195 }
2198 }
2209 }
2212 }
2213 }
2216 }
2220 XFS_BUF_LOCK);
2233 pre_flushed++;
2234 }
2236 }
2247 }
2261 }
2262 }
2267 }
2270 }
2272 /*
2273 * This is called to return an inode to the inode free list.
2274 * The inode should already be truncated to 0 length and have
2275 * no pages associated with it. This routine also assumes that
2276 * the inode is already a part of the transaction.
2277 *
2278 * The on-disk copy of the inode will have been added to the list
2279 * of unlinked inodes in the AGI. We need to remove the inode from
2280 * that list atomically with respect to freeing it here.
2281 */
2282 int
2287 {
2290 xfs_ino_t first_ino;
2301 /*
2302 * Pull the on-disk inode from the AGI unlinked list.
2303 */
2307 }
2312 }
2321 /*
2322 * Bump the generation count so no one will be confused
2323 * by reincarnations of this inode.
2324 */
2330 }
2333 }
2335 /*
2336 * Reallocate the space for if_broot based on the number of records
2337 * being added or deleted as indicated in rec_diff. Move the records
2338 * and pointers in if_broot to fit the new size. When shrinking this
2339 * will eliminate holes between the records and pointers created by
2340 * the caller. When growing this will create holes to be filled in
2341 * by the caller.
2342 *
2343 * The caller must not request to add more records than would fit in
2344 * the on-disk inode root. If the if_broot is currently NULL, then
2345 * if we adding records one will be allocated. The caller must also
2346 * not request that the number of records go below zero, although
2347 * it can go to zero.
2348 *
2349 * ip -- the inode whose if_broot area is changing
2350 * ext_diff -- the change in the number of records, positive or negative,
2351 * requested for the if_broot array.
2352 */
2353 void
2358 {
2367 /*
2368 * Handle the degenerate case quietly.
2369 */
2372 }
2376 /*
2377 * If there wasn't any memory allocated before, just
2378 * allocate it now and get out.
2379 */
2383 KM_SLEEP);
2386 }
2388 /*
2389 * If there is already an existing if_broot, then we need
2390 * to realloc() it and shift the pointers to their new
2391 * location. The records don't change location because
2392 * they are kept butted up against the btree block header.
2393 */
2399 new_size,
2401 KM_SLEEP);
2411 }
2413 /*
2414 * rec_diff is less than 0. In this case, we are shrinking the
2415 * if_broot buffer. It must already exist. If we go to zero
2416 * records, just get rid of the root and clear the status bit.
2417 */
2424 else
2428 /*
2429 * First copy over the btree block header.
2430 */
2435 }
2437 /*
2438 * Only copy the records and pointers if there are any.
2439 */
2441 /*
2442 * First copy the records.
2443 */
2450 /*
2451 * Then copy the pointers.
2452 */
2458 }
2465 }
2468 /*
2469 * This is called when the amount of space needed for if_extents
2470 * is increased or decreased. The change in size is indicated by
2471 * the number of extents that need to be added or deleted in the
2472 * ext_diff parameter.
2473 *
2474 * If the amount of space needed has decreased below the size of the
2475 * inline buffer, then switch to using the inline buffer. Otherwise,
2476 * use kmem_realloc() or kmem_alloc() to adjust the size of the buffer
2477 * to what is needed.
2478 *
2479 * ip -- the inode whose if_extents area is changing
2480 * ext_diff -- the change in the number of extents, positive or negative,
2481 * requested for the if_extents array.
2482 */
2483 void
2488 {
2492 uint rnew_size;
2496 }
2507 }
2511 /*
2512 * If the valid extents can fit in if_inline_ext,
2513 * copy them from the malloc'd vector and free it.
2514 */
2516 /*
2517 * For now, empty files are format EXTENTS,
2518 * so the if_extents pointer is null.
2519 */
2525 }
2527 }
2533 /*
2534 * Stuck with malloc/realloc.
2535 */
2544 rnew_size,
2546 KM_NOFS);
2547 }
2548 }
2551 }
2554 /*
2555 * This is called when the amount of space needed for if_data
2556 * is increased or decreased. The change in size is indicated by
2557 * the number of bytes that need to be added or deleted in the
2558 * byte_diff parameter.
2559 *
2560 * If the amount of space needed has decreased below the size of the
2561 * inline buffer, then switch to using the inline buffer. Otherwise,
2562 * use kmem_realloc() or kmem_alloc() to adjust the size of the buffer
2563 * to what is needed.
2564 *
2565 * ip -- the inode whose if_data area is changing
2566 * byte_diff -- the change in the number of bytes, positive or negative,
2567 * requested for the if_data array.
2568 */
2569 void
2574 {
2581 }
2590 }
2594 /*
2595 * If the valid extents/data can fit in if_inline_ext/data,
2596 * copy them from the malloc'd vector and free it.
2597 */
2603 new_size);
2606 }
2609 /*
2610 * Stuck with malloc/realloc.
2611 * For inline data, the underlying buffer must be
2612 * a multiple of 4 bytes in size so that it can be
2613 * logged and stay on word boundaries. We enforce
2614 * that here.
2615 */
2621 /*
2622 * Only do the realloc if the underlying size
2623 * is really changing.
2624 */
2628 real_size,
2630 KM_SLEEP);
2631 }
2637 }
2638 }
2642 }
2647 /*
2648 * Map inode to disk block and offset.
2649 *
2650 * mp -- the mount point structure for the current file system
2651 * tp -- the current transaction
2652 * ino -- the inode number of the inode to be located
2653 * imap -- this structure is filled in with the information necessary
2654 * to retrieve the given inode from disk
2655 * flags -- flags to pass to xfs_dilocate indicating whether or not
2656 * lookups in the inode btree were OK or not
2657 */
2658 int
2662 xfs_ino_t ino,
2664 uint flags)
2665 {
2666 xfs_fsblock_t fsbno;
2676 }
2683 }
2685 void
2689 {
2696 }
2698 /*
2699 * If the format is local, then we can't have an extents
2700 * array so just look for an inline data array. If we're
2701 * not local then we may or may not have an extents list,
2702 * so check and free it up if we do.
2703 */
2711 }
2719 }
2726 }
2727 }
2729 /*
2730 * This is called free all the memory associated with an inode.
2731 * It must free the inode itself and any buffers allocated for
2732 * if_extents/if_data and if_broot. It must also free the lock
2733 * associated with the inode.
2734 */
2735 void
2738 {
2746 }
2752 #ifdef XFS_BMAP_TRACE
2754 #endif
2755 #ifdef XFS_BMBT_TRACE
2757 #endif
2758 #ifdef XFS_RW_TRACE
2760 #endif
2761 #ifdef XFS_ILOCK_TRACE
2763 #endif
2764 #ifdef XFS_DIR2_TRACE
2766 #endif
2768 /* XXXdpd should be able to assert this but shutdown
2769 * is leaving the AIL behind. */
2773 }
2775 }
2778 /*
2779 * Increment the pin count of the given buffer.
2780 * This value is protected by ipinlock spinlock in the mount structure.
2781 */
2782 void
2785 {
2789 }
2791 /*
2792 * Decrement the pin count of the given inode, and wake up
2793 * anyone in xfs_iwait_unpin() if the count goes to 0. The
2794 * inode must have been previoulsy pinned with a call to xfs_ipin().
2795 */
2796 void
2799 {
2805 /* make sync come back and flush this inode */
2811 }
2814 }
2815 }
2817 /*
2818 * This is called to wait for the given inode to be unpinned.
2819 * It will sleep until this happens. The caller must have the
2820 * inode locked in at least shared mode so that the buffer cannot
2821 * be subsequently pinned once someone is waiting for it to be
2822 * unpinned.
2823 */
2824 STATIC void
2827 {
2829 xfs_lsn_t lsn;
2835 }
2842 }
2844 /*
2845 * Give the log a push so we don't wait here too long.
2846 */
2850 }
2853 /*
2854 * xfs_iextents_copy()
2855 *
2856 * This is called to copy the REAL extents (as opposed to the delayed
2857 * allocation extents) from the inode into the given buffer. It
2858 * returns the number of bytes copied into the buffer.
2859 *
2860 * If there are no delayed allocation extents, then we can just
2861 * memcpy() the extents into the buffer. Otherwise, we need to
2862 * examine each extent in turn and skip those which are delayed.
2863 */
2864 int
2869 {
2873 #ifdef XFS_BMAP_TRACE
2875 #endif
2879 xfs_fsblock_t start_block;
2889 /*
2890 * There are some delayed allocation extents in the
2891 * inode, so copy the extents one at a time and skip
2892 * the delayed ones. There must be at least one
2893 * non-delayed extent.
2894 */
2901 /*
2902 * It's a delayed allocation extent, so skip it.
2903 */
2904 ep++;
2906 }
2908 /* Translate to on disk format */
2913 dest_ep++;
2914 ep++;
2915 copied++;
2916 }
2921 }
2923 /*
2924 * Each of the following cases stores data into the same region
2925 * of the on-disk inode, so only one of them can be valid at
2926 * any given time. While it is possible to have conflicting formats
2927 * and log flags, e.g. having XFS_ILOG_?DATA set when the fork is
2928 * in EXTENTS format, this can only happen when the fork has
2929 * changed formats after being modified but before being flushed.
2930 * In these cases, the format always takes precedence, because the
2931 * format indicates the current state of the fork.
2932 */
2933 /*ARGSUSED*/
2934 STATIC int
2941 {
2945 #ifdef XFS_TRANS_DEBUG
2947 #endif
2958 /*
2959 * This can happen if we gave up in iformat in an error path,
2960 * for the attribute fork.
2961 */
2965 }
2975 }
2981 }
2982 }
2994 whichfork);
2995 }
3004 XFS_BROOT_SIZE_ADJ));
3008 }
3015 }
3023 }
3029 }
3032 }
3034 /*
3035 * xfs_iflush() will write a modified inode's changes out to the
3036 * inode's on disk home. The caller must have the inode lock held
3037 * in at least shared mode and the inode flush semaphore must be
3038 * held as well. The inode lock will still be held upon return from
3039 * the call and the caller is free to unlock it.
3040 * The inode flush lock will be unlocked when the inode reaches the disk.
3041 * The flags indicate how the inode's buffer should be written out.
3042 */
3043 int
3046 uint flags)
3047 {
3053 /* REFERENCED */
3071 /*
3072 * If the inode isn't dirty, then just release the inode
3073 * flush lock and do nothing.
3074 */
3081 }
3083 /*
3084 * We can't flush the inode until it is unpinned, so
3085 * wait for it. We know noone new can pin it, because
3086 * we are holding the inode lock shared and you need
3087 * to hold it exclusively to pin the inode.
3088 */
3091 /*
3092 * This may have been unpinned because the filesystem is shutting
3093 * down forcibly. If that's the case we must not write this inode
3094 * to disk, because the log record didn't make it to disk!
3095 */
3102 }
3104 /*
3105 * Get the buffer containing the on-disk inode.
3106 */
3111 }
3113 /*
3114 * Decide how buffer will be flushed out. This is done before
3115 * the call to xfs_iflush_int because this field is zeroed by it.
3116 */
3118 /*
3119 * Flush out the inode buffer according to the directions
3120 * of the caller. In the cases where the caller has given
3121 * us a choice choose the non-delwri case. This is because
3122 * the inode is in the AIL and we need to get it out soon.
3123 */
3140 }
3158 }
3159 }
3161 /*
3162 * First flush out the inode that xfs_iflush was called with.
3163 */
3167 }
3169 /*
3170 * inode clustering:
3171 * see if other inodes can be gathered into this write
3172 */
3181 /*
3182 * Do an un-protected check to see if the inode is dirty and
3183 * is a candidate for flushing. These checks will be repeated
3184 * later after the appropriate locks are acquired.
3185 */
3192 }
3194 /*
3195 * Try to get locks. If any are unavailable,
3196 * then this inode cannot be flushed and is skipped.
3197 */
3199 /* get inode locks (just i_lock) */
3201 /* get inode flush lock */
3203 /* check if pinned */
3205 /* arriving here means that
3206 * this inode can be flushed.
3207 * first re-check that it's
3208 * dirty
3209 */
3214 clcount++;
3218 XFS_ILOCK_SHARED);
3220 }
3223 }
3226 }
3227 }
3229 }
3230 }
3236 }
3238 /*
3239 * If the buffer is pinned then push on the log so we won't
3240 * get stuck waiting in the write for too long.
3241 */
3244 }
3252 }
3255 corrupt_out:
3259 /*
3260 * Unlocks the flush lock
3261 */
3264 cluster_corrupt_out:
3265 /* Corruption detected in the clustering loop. Invalidate the
3266 * inode buffer and shut down the filesystem.
3267 */
3270 /*
3271 * Clean up the buffer. If it was B_DELWRI, just release it --
3272 * brelse can handle it with no problems. If not, shut down the
3273 * filesystem before releasing the buffer.
3274 */
3277 }
3282 /*
3283 * Just like incore_relse: if we have b_iodone functions,
3284 * mark the buffer as an error and call them. Otherwise
3285 * mark it as stale and brelse.
3286 */
3297 }
3298 }
3301 /*
3302 * Unlocks the flush lock
3303 */
3305 }
3308 STATIC int
3312 {
3316 #ifdef XFS_TRANS_DEBUG
3318 #endif
3330 /*
3331 * If the inode isn't dirty, then just release the inode
3332 * flush lock and do nothing.
3333 */
3338 }
3340 /* set *dip = inode's place in the buffer */
3343 /*
3344 * Clear i_update_core before copying out the data.
3345 * This is for coordination with our timestamp updates
3346 * that don't hold the inode lock. They will always
3347 * update the timestamps BEFORE setting i_update_core,
3348 * so if we clear i_update_core after they set it we
3349 * are guaranteed to see their updates to the timestamps.
3350 * I believe that this depends on strongly ordered memory
3351 * semantics, but we have that. We use the SYNCHRONIZE
3352 * macro to make sure that the compiler does not reorder
3353 * the i_update_core access below the data copy below.
3354 */
3364 }
3371 }
3381 }
3392 }
3393 }
3396 XFS_RANDOM_IFLUSH_5)) {
3402 ip);
3404 }
3411 }
3412 /*
3413 * bump the flush iteration count, used to detect flushes which
3414 * postdate a log record during recovery.
3415 */
3419 /*
3420 * Copy the dirty parts of the inode into the on-disk
3421 * inode. We always copy out the core of the inode,
3422 * because if the inode is dirty at all the core must
3423 * be.
3424 */
3427 /* Wrap, we never let the log put out DI_MAX_FLUSH */
3431 /*
3432 * If this is really an old format inode and the superblock version
3433 * has not been updated to support only new format inodes, then
3434 * convert back to the old inode format. If the superblock version
3435 * has been updated, then make the conversion permanent.
3436 */
3441 /*
3442 * Convert it back.
3443 */
3447 /*
3448 * The superblock version has already been bumped,
3449 * so just make the conversion to the new inode
3450 * format permanent.
3451 */
3460 }
3461 }
3465 }
3468 /*
3469 * The only error from xfs_iflush_fork is on the data fork.
3470 */
3472 }
3475 /*
3476 * We've recorded everything logged in the inode, so we'd
3477 * like to clear the ilf_fields bits so we don't log and
3478 * flush things unnecessarily. However, we can't stop
3479 * logging all this information until the data we've copied
3480 * into the disk buffer is written to disk. If we did we might
3481 * overwrite the copy of the inode in the log with all the
3482 * data after re-logging only part of it, and in the face of
3483 * a crash we wouldn't have all the data we need to recover.
3484 *
3485 * What we do is move the bits to the ili_last_fields field.
3486 * When logging the inode, these bits are moved back to the
3487 * ilf_fields field. In the xfs_iflush_done() routine we
3488 * clear ili_last_fields, since we know that the information
3489 * those bits represent is permanently on disk. As long as
3490 * the flush completes before the inode is logged again, then
3491 * both ilf_fields and ili_last_fields will be cleared.
3492 *
3493 * We can play with the ilf_fields bits here, because the inode
3494 * lock must be held exclusively in order to set bits there
3495 * and the flush lock protects the ili_last_fields bits.
3496 * Set ili_logged so the flush done
3497 * routine can tell whether or not to look in the AIL.
3498 * Also, store the current LSN of the inode so that we can tell
3499 * whether the item has moved in the AIL from xfs_iflush_done().
3500 * In order to read the lsn we need the AIL lock, because
3501 * it is a 64 bit value that cannot be read atomically.
3502 */
3513 /*
3514 * Attach the function xfs_iflush_done to the inode's
3515 * buffer. This will remove the inode from the AIL
3516 * and unlock the inode's flush lock when the inode is
3517 * completely written to disk.
3518 */
3525 /*
3526 * We're flushing an inode which is not in the AIL and has
3527 * not been logged but has i_update_core set. For this
3528 * case we can use a B_DELWRI flush and immediately drop
3529 * the inode flush lock because we can avoid the whole
3530 * AIL state thing. It's OK to drop the flush lock now,
3531 * because we've already locked the buffer and to do anything
3532 * you really need both.
3533 */
3538 }
3540 }
3544 corrupt_out:
3546 }
3549 /*
3550 * Flush all inactive inodes in mp.
3551 */
3552 void
3555 {
3559 again:
3566 /* Make sure we skip markers inserted by sync */
3570 }
3577 }
3583 out:
3585 }
3587 /*
3588 * xfs_iaccess: check accessibility of inode for mode.
3589 */
3590 int
3593 mode_t mode,
3595 {
3609 }
3611 /*
3612 * If there's an Access Control List it's used instead of
3613 * the mode bits.
3614 */
3622 }
3624 /*
3625 * If the DACs are ok we don't need any capability check.
3626 */
3629 /*
3630 * Read/write DACs are always overridable.
3631 * Executable DACs are overridable if at least one exec bit is set.
3632 */
3642 #ifdef NOISE
3646 }
3648 }
3650 /*
3651 * xfs_iroundup: round up argument to next power of two
3652 */
3653 uint
3655 uint v)
3656 {
3658 uint m;
3671 }
3674 }
3676 #ifdef XFS_ILOCK_TRACE
3679 void
3681 {
3690 }
3691 #endif