2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
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.
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.
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
23 #include "xfs_trans.h"
27 #include "xfs_alloc.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_quota.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir2_sf.h"
35 #include "xfs_attr_sf.h"
36 #include "xfs_dinode.h"
37 #include "xfs_inode.h"
38 #include "xfs_btree.h"
39 #include "xfs_ialloc.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
44 #include "xfs_fsops.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
51 #include "xfs_vfsops.h"
52 #include "xfs_version.h"
53 #include "xfs_log_priv.h"
54 #include "xfs_trans_priv.h"
55 #include "xfs_filestream.h"
56 #include "xfs_da_btree.h"
57 #include "xfs_dir2_trace.h"
58 #include "xfs_extfree_item.h"
59 #include "xfs_mru_cache.h"
60 #include "xfs_inode_item.h"
62 #include <linux/namei.h>
63 #include <linux/init.h>
64 #include <linux/mount.h>
65 #include <linux/mempool.h>
66 #include <linux/writeback.h>
67 #include <linux/kthread.h>
68 #include <linux/freezer.h>
69 #include <linux/parser.h>
71 static struct quotactl_ops xfs_quotactl_operations;
72 static struct super_operations xfs_super_operations;
73 static kmem_zone_t *xfs_vnode_zone;
74 static kmem_zone_t *xfs_ioend_zone;
75 mempool_t *xfs_ioend_pool;
77 STATIC struct xfs_mount_args *
79 struct super_block *sb,
82 struct xfs_mount_args *args;
84 args = kzalloc(sizeof(struct xfs_mount_args), GFP_KERNEL);
88 args->logbufs = args->logbufsize = -1;
89 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
91 /* Copy the already-parsed mount(2) flags we're interested in */
92 if (sb->s_flags & MS_DIRSYNC)
93 args->flags |= XFSMNT_DIRSYNC;
94 if (sb->s_flags & MS_SYNCHRONOUS)
95 args->flags |= XFSMNT_WSYNC;
97 args->flags |= XFSMNT_QUIET;
98 args->flags |= XFSMNT_32BITINODES;
103 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
104 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
105 #define MNTOPT_LOGDEV "logdev" /* log device */
106 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
107 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
108 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
109 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
110 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
111 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
112 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
113 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
114 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
115 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
116 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
117 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
118 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
119 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
120 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
121 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
122 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
123 * unwritten extent conversion */
124 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
125 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
126 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
127 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
128 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
129 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
130 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
132 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
133 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
134 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
135 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
136 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
137 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
138 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
139 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
140 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
141 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
142 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
143 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
144 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
145 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
146 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
147 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
148 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
149 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
152 * Table driven mount option parser.
154 * Currently only used for remount, but it will be used for mount
155 * in the future, too.
158 Opt_barrier, Opt_nobarrier, Opt_err
161 static match_table_t tokens = {
162 {Opt_barrier, "barrier"},
163 {Opt_nobarrier, "nobarrier"},
169 suffix_strtoul(char *s, char **endp, unsigned int base)
171 int last, shift_left_factor = 0;
174 last = strlen(value) - 1;
175 if (value[last] == 'K' || value[last] == 'k') {
176 shift_left_factor = 10;
179 if (value[last] == 'M' || value[last] == 'm') {
180 shift_left_factor = 20;
183 if (value[last] == 'G' || value[last] == 'g') {
184 shift_left_factor = 30;
188 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
193 struct xfs_mount *mp,
195 struct xfs_mount_args *args,
198 char *this_char, *value, *eov;
199 int dsunit, dswidth, vol_dsunit, vol_dswidth;
201 int dmapi_implies_ikeep = 1;
203 args->flags |= XFSMNT_BARRIER;
204 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
209 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
211 while ((this_char = strsep(&options, ",")) != NULL) {
214 if ((value = strchr(this_char, '=')) != NULL)
217 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
218 if (!value || !*value) {
220 "XFS: %s option requires an argument",
224 args->logbufs = simple_strtoul(value, &eov, 10);
225 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
226 if (!value || !*value) {
228 "XFS: %s option requires an argument",
232 args->logbufsize = suffix_strtoul(value, &eov, 10);
233 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
234 if (!value || !*value) {
236 "XFS: %s option requires an argument",
240 strncpy(args->logname, value, MAXNAMELEN);
241 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
242 if (!value || !*value) {
244 "XFS: %s option requires an argument",
248 strncpy(args->mtpt, value, MAXNAMELEN);
249 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
250 if (!value || !*value) {
252 "XFS: %s option requires an argument",
256 strncpy(args->rtname, value, MAXNAMELEN);
257 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
258 if (!value || !*value) {
260 "XFS: %s option requires an argument",
264 iosize = simple_strtoul(value, &eov, 10);
265 args->flags |= XFSMNT_IOSIZE;
266 args->iosizelog = (uint8_t) iosize;
267 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
268 if (!value || !*value) {
270 "XFS: %s option requires an argument",
274 iosize = suffix_strtoul(value, &eov, 10);
275 args->flags |= XFSMNT_IOSIZE;
276 args->iosizelog = ffs(iosize) - 1;
277 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
278 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
279 mp->m_flags |= XFS_MOUNT_GRPID;
280 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
281 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
282 mp->m_flags &= ~XFS_MOUNT_GRPID;
283 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
284 args->flags |= XFSMNT_WSYNC;
285 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
286 args->flags |= XFSMNT_OSYNCISOSYNC;
287 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
288 args->flags |= XFSMNT_NORECOVERY;
289 } else if (!strcmp(this_char, MNTOPT_INO64)) {
290 args->flags |= XFSMNT_INO64;
293 "XFS: %s option not allowed on this system",
297 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
298 args->flags |= XFSMNT_NOALIGN;
299 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
300 args->flags |= XFSMNT_SWALLOC;
301 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
302 if (!value || !*value) {
304 "XFS: %s option requires an argument",
308 dsunit = simple_strtoul(value, &eov, 10);
309 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
310 if (!value || !*value) {
312 "XFS: %s option requires an argument",
316 dswidth = simple_strtoul(value, &eov, 10);
317 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
318 args->flags &= ~XFSMNT_32BITINODES;
321 "XFS: %s option not allowed on this system",
325 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
326 args->flags |= XFSMNT_NOUUID;
327 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
328 args->flags |= XFSMNT_BARRIER;
329 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
330 args->flags &= ~XFSMNT_BARRIER;
331 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
332 args->flags |= XFSMNT_IKEEP;
333 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
334 dmapi_implies_ikeep = 0;
335 args->flags &= ~XFSMNT_IKEEP;
336 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
337 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
338 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
339 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
340 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
341 args->flags |= XFSMNT_ATTR2;
342 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
343 args->flags &= ~XFSMNT_ATTR2;
344 args->flags |= XFSMNT_NOATTR2;
345 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
346 args->flags2 |= XFSMNT2_FILESTREAMS;
347 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
348 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
349 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
350 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
351 !strcmp(this_char, MNTOPT_UQUOTA) ||
352 !strcmp(this_char, MNTOPT_USRQUOTA)) {
353 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
354 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
355 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
356 args->flags |= XFSMNT_UQUOTA;
357 args->flags &= ~XFSMNT_UQUOTAENF;
358 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
359 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
360 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
361 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
362 args->flags |= XFSMNT_PQUOTA;
363 args->flags &= ~XFSMNT_PQUOTAENF;
364 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
365 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
366 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
367 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
368 args->flags |= XFSMNT_GQUOTA;
369 args->flags &= ~XFSMNT_GQUOTAENF;
370 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
371 args->flags |= XFSMNT_DMAPI;
372 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
373 args->flags |= XFSMNT_DMAPI;
374 } else if (!strcmp(this_char, MNTOPT_DMI)) {
375 args->flags |= XFSMNT_DMAPI;
376 } else if (!strcmp(this_char, "ihashsize")) {
378 "XFS: ihashsize no longer used, option is deprecated.");
379 } else if (!strcmp(this_char, "osyncisdsync")) {
380 /* no-op, this is now the default */
382 "XFS: osyncisdsync is now the default, option is deprecated.");
383 } else if (!strcmp(this_char, "irixsgid")) {
385 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
388 "XFS: unknown mount option [%s].", this_char);
393 if (args->flags & XFSMNT_NORECOVERY) {
394 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
396 "XFS: no-recovery mounts must be read-only.");
401 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
403 "XFS: sunit and swidth options incompatible with the noalign option");
407 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
409 "XFS: cannot mount with both project and group quota");
413 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
414 printk("XFS: %s option needs the mount point option as well\n",
419 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
421 "XFS: sunit and swidth must be specified together");
425 if (dsunit && (dswidth % dsunit != 0)) {
427 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
433 * Applications using DMI filesystems often expect the
434 * inode generation number to be monotonically increasing.
435 * If we delete inode chunks we break this assumption, so
436 * keep unused inode chunks on disk for DMI filesystems
437 * until we come up with a better solution.
438 * Note that if "ikeep" or "noikeep" mount options are
439 * supplied, then they are honored.
441 if ((args->flags & XFSMNT_DMAPI) && dmapi_implies_ikeep)
442 args->flags |= XFSMNT_IKEEP;
444 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
446 args->sunit = dsunit;
447 args->flags |= XFSMNT_RETERR;
449 args->sunit = vol_dsunit;
451 dswidth ? (args->swidth = dswidth) :
452 (args->swidth = vol_dswidth);
454 args->sunit = args->swidth = 0;
458 if (args->flags & XFSMNT_32BITINODES)
459 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
461 args->flags |= XFSMNT_FLAGS2;
465 struct proc_xfs_info {
472 struct xfs_mount *mp,
475 static struct proc_xfs_info xfs_info_set[] = {
476 /* the few simple ones we can get from the mount struct */
477 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
478 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
479 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
480 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
481 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
482 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
483 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
484 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
485 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
486 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
487 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
488 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
491 static struct proc_xfs_info xfs_info_unset[] = {
492 /* the few simple ones we can get from the mount struct */
493 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
494 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
495 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
498 struct proc_xfs_info *xfs_infop;
500 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
501 if (mp->m_flags & xfs_infop->flag)
502 seq_puts(m, xfs_infop->str);
504 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
505 if (!(mp->m_flags & xfs_infop->flag))
506 seq_puts(m, xfs_infop->str);
509 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
510 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
511 (int)(1 << mp->m_writeio_log) >> 10);
513 if (mp->m_logbufs > 0)
514 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
515 if (mp->m_logbsize > 0)
516 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
519 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
521 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
523 if (mp->m_dalign > 0)
524 seq_printf(m, "," MNTOPT_SUNIT "=%d",
525 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
526 if (mp->m_swidth > 0)
527 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
528 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
530 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
531 seq_puts(m, "," MNTOPT_USRQUOTA);
532 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
533 seq_puts(m, "," MNTOPT_UQUOTANOENF);
535 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
536 seq_puts(m, "," MNTOPT_PRJQUOTA);
537 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
538 seq_puts(m, "," MNTOPT_PQUOTANOENF);
540 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
541 seq_puts(m, "," MNTOPT_GRPQUOTA);
542 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
543 seq_puts(m, "," MNTOPT_GQUOTANOENF);
545 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
546 seq_puts(m, "," MNTOPT_NOQUOTA);
552 unsigned int blockshift)
554 unsigned int pagefactor = 1;
555 unsigned int bitshift = BITS_PER_LONG - 1;
557 /* Figure out maximum filesize, on Linux this can depend on
558 * the filesystem blocksize (on 32 bit platforms).
559 * __block_prepare_write does this in an [unsigned] long...
560 * page->index << (PAGE_CACHE_SHIFT - bbits)
561 * So, for page sized blocks (4K on 32 bit platforms),
562 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
563 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
564 * but for smaller blocksizes it is less (bbits = log2 bsize).
565 * Note1: get_block_t takes a long (implicit cast from above)
566 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
567 * can optionally convert the [unsigned] long from above into
568 * an [unsigned] long long.
571 #if BITS_PER_LONG == 32
572 # if defined(CONFIG_LBD)
573 ASSERT(sizeof(sector_t) == 8);
574 pagefactor = PAGE_CACHE_SIZE;
575 bitshift = BITS_PER_LONG;
577 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
581 return (((__uint64_t)pagefactor) << bitshift) - 1;
588 switch (inode->i_mode & S_IFMT) {
590 inode->i_op = &xfs_inode_operations;
591 inode->i_fop = &xfs_file_operations;
592 inode->i_mapping->a_ops = &xfs_address_space_operations;
595 if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
596 inode->i_op = &xfs_dir_ci_inode_operations;
598 inode->i_op = &xfs_dir_inode_operations;
599 inode->i_fop = &xfs_dir_file_operations;
602 inode->i_op = &xfs_symlink_inode_operations;
603 if (!(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE))
604 inode->i_mapping->a_ops = &xfs_address_space_operations;
607 inode->i_op = &xfs_inode_operations;
608 init_special_inode(inode, inode->i_mode, inode->i_rdev);
614 xfs_revalidate_inode(
620 inode->i_mode = ip->i_d.di_mode;
621 inode->i_nlink = ip->i_d.di_nlink;
622 inode->i_uid = ip->i_d.di_uid;
623 inode->i_gid = ip->i_d.di_gid;
625 switch (inode->i_mode & S_IFMT) {
629 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
630 sysv_minor(ip->i_df.if_u2.if_rdev));
637 inode->i_generation = ip->i_d.di_gen;
638 i_size_write(inode, ip->i_d.di_size);
639 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
640 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
641 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
642 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
643 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
644 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
645 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
646 inode->i_flags |= S_IMMUTABLE;
648 inode->i_flags &= ~S_IMMUTABLE;
649 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
650 inode->i_flags |= S_APPEND;
652 inode->i_flags &= ~S_APPEND;
653 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
654 inode->i_flags |= S_SYNC;
656 inode->i_flags &= ~S_SYNC;
657 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
658 inode->i_flags |= S_NOATIME;
660 inode->i_flags &= ~S_NOATIME;
661 xfs_iflags_clear(ip, XFS_IMODIFIED);
665 xfs_initialize_vnode(
666 struct xfs_mount *mp,
668 struct xfs_inode *ip)
673 inode->i_private = ip;
677 * We need to set the ops vectors, and unlock the inode, but if
678 * we have been called during the new inode create process, it is
679 * too early to fill in the Linux inode. We will get called a
680 * second time once the inode is properly set up, and then we can
683 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
684 xfs_revalidate_inode(mp, inode, ip);
685 xfs_set_inodeops(inode);
687 xfs_iflags_clear(ip, XFS_INEW);
690 unlock_new_inode(inode);
698 struct block_device **bdevp)
702 *bdevp = open_bdev_excl(name, 0, mp);
703 if (IS_ERR(*bdevp)) {
704 error = PTR_ERR(*bdevp);
705 printk("XFS: Invalid device [%s], error=%d\n", name, error);
713 struct block_device *bdev)
716 close_bdev_excl(bdev);
720 * Try to write out the superblock using barriers.
726 xfs_buf_t *sbp = xfs_getsb(mp, 0);
731 XFS_BUF_UNDELAYWRITE(sbp);
733 XFS_BUF_UNASYNC(sbp);
734 XFS_BUF_ORDERED(sbp);
737 error = xfs_iowait(sbp);
740 * Clear all the flags we set and possible error state in the
741 * buffer. We only did the write to try out whether barriers
742 * worked and shouldn't leave any traces in the superblock
746 XFS_BUF_ERROR(sbp, 0);
747 XFS_BUF_UNORDERED(sbp);
754 xfs_mountfs_check_barriers(xfs_mount_t *mp)
758 if (mp->m_logdev_targp != mp->m_ddev_targp) {
759 xfs_fs_cmn_err(CE_NOTE, mp,
760 "Disabling barriers, not supported with external log device");
761 mp->m_flags &= ~XFS_MOUNT_BARRIER;
765 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
766 xfs_fs_cmn_err(CE_NOTE, mp,
767 "Disabling barriers, underlying device is readonly");
768 mp->m_flags &= ~XFS_MOUNT_BARRIER;
772 error = xfs_barrier_test(mp);
774 xfs_fs_cmn_err(CE_NOTE, mp,
775 "Disabling barriers, trial barrier write failed");
776 mp->m_flags &= ~XFS_MOUNT_BARRIER;
782 xfs_blkdev_issue_flush(
783 xfs_buftarg_t *buftarg)
785 blkdev_issue_flush(buftarg->bt_bdev, NULL);
790 struct xfs_mount *mp)
792 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
793 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
794 xfs_free_buftarg(mp->m_logdev_targp);
795 xfs_blkdev_put(logdev);
797 if (mp->m_rtdev_targp) {
798 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
799 xfs_free_buftarg(mp->m_rtdev_targp);
800 xfs_blkdev_put(rtdev);
802 xfs_free_buftarg(mp->m_ddev_targp);
806 * The file system configurations are:
807 * (1) device (partition) with data and internal log
808 * (2) logical volume with data and log subvolumes.
809 * (3) logical volume with data, log, and realtime subvolumes.
811 * We only have to handle opening the log and realtime volumes here if
812 * they are present. The data subvolume has already been opened by
813 * get_sb_bdev() and is stored in sb->s_bdev.
817 struct xfs_mount *mp,
818 struct xfs_mount_args *args)
820 struct block_device *ddev = mp->m_super->s_bdev;
821 struct block_device *logdev = NULL, *rtdev = NULL;
825 * Open real time and log devices - order is important.
827 if (args->logname[0]) {
828 error = xfs_blkdev_get(mp, args->logname, &logdev);
833 if (args->rtname[0]) {
834 error = xfs_blkdev_get(mp, args->rtname, &rtdev);
836 goto out_close_logdev;
838 if (rtdev == ddev || rtdev == logdev) {
840 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
842 goto out_close_rtdev;
847 * Setup xfs_mount buffer target pointers
850 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
851 if (!mp->m_ddev_targp)
852 goto out_close_rtdev;
855 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
856 if (!mp->m_rtdev_targp)
857 goto out_free_ddev_targ;
860 if (logdev && logdev != ddev) {
861 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
862 if (!mp->m_logdev_targp)
863 goto out_free_rtdev_targ;
865 mp->m_logdev_targp = mp->m_ddev_targp;
871 if (mp->m_rtdev_targp)
872 xfs_free_buftarg(mp->m_rtdev_targp);
874 xfs_free_buftarg(mp->m_ddev_targp);
877 xfs_blkdev_put(rtdev);
879 if (logdev && logdev != ddev)
880 xfs_blkdev_put(logdev);
886 * Setup xfs_mount buffer target pointers based on superblock
890 struct xfs_mount *mp)
894 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
895 mp->m_sb.sb_sectsize);
899 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
900 unsigned int log_sector_size = BBSIZE;
902 if (xfs_sb_version_hassector(&mp->m_sb))
903 log_sector_size = mp->m_sb.sb_logsectsize;
904 error = xfs_setsize_buftarg(mp->m_logdev_targp,
905 mp->m_sb.sb_blocksize,
910 if (mp->m_rtdev_targp) {
911 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
912 mp->m_sb.sb_blocksize,
913 mp->m_sb.sb_sectsize);
922 * XFS AIL push thread support
927 xfs_lsn_t threshold_lsn)
929 mp->m_ail.xa_target = threshold_lsn;
930 wake_up_process(mp->m_ail.xa_task);
937 xfs_mount_t *mp = (xfs_mount_t *)data;
938 xfs_lsn_t last_pushed_lsn = 0;
941 while (!kthread_should_stop()) {
943 schedule_timeout_interruptible(msecs_to_jiffies(tout));
950 if (XFS_FORCED_SHUTDOWN(mp))
953 tout = xfsaild_push(mp, &last_pushed_lsn);
963 mp->m_ail.xa_target = 0;
964 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
965 if (IS_ERR(mp->m_ail.xa_task))
966 return -PTR_ERR(mp->m_ail.xa_task);
974 kthread_stop(mp->m_ail.xa_task);
979 STATIC struct inode *
981 struct super_block *sb)
985 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
992 xfs_fs_destroy_inode(
995 kmem_zone_free(xfs_vnode_zone, inode);
999 xfs_fs_inode_init_once(
1002 inode_init_once((struct inode *)vnode);
1006 * Attempt to flush the inode, this will actually fail
1007 * if the inode is pinned, but we dirty the inode again
1008 * at the point when it is unpinned after a log write,
1009 * since this is when the inode itself becomes flushable.
1013 struct inode *inode,
1019 xfs_itrace_entry(XFS_I(inode));
1021 filemap_fdatawait(inode->i_mapping);
1022 flags |= FLUSH_SYNC;
1024 error = xfs_inode_flush(XFS_I(inode), flags);
1026 * if we failed to write out the inode then mark
1027 * it dirty again so we'll try again later.
1030 mark_inode_dirty_sync(inode);
1037 struct inode *inode)
1039 xfs_inode_t *ip = XFS_I(inode);
1042 * ip can be null when xfs_iget_core calls xfs_idestroy if we
1043 * find an inode with di_mode == 0 but without IGET_CREATE set.
1046 xfs_itrace_entry(ip);
1047 XFS_STATS_INC(vn_rele);
1048 XFS_STATS_INC(vn_remove);
1049 XFS_STATS_INC(vn_reclaim);
1050 XFS_STATS_DEC(vn_active);
1053 xfs_iflags_clear(ip, XFS_IMODIFIED);
1054 if (xfs_reclaim(ip))
1055 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
1058 ASSERT(XFS_I(inode) == NULL);
1062 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
1063 * Doing this has two advantages:
1064 * - It saves on stack space, which is tight in certain situations
1065 * - It can be used (with care) as a mechanism to avoid deadlocks.
1066 * Flushing while allocating in a full filesystem requires both.
1069 xfs_syncd_queue_work(
1070 struct xfs_mount *mp,
1072 void (*syncer)(struct xfs_mount *, void *))
1074 struct bhv_vfs_sync_work *work;
1076 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
1077 INIT_LIST_HEAD(&work->w_list);
1078 work->w_syncer = syncer;
1079 work->w_data = data;
1081 spin_lock(&mp->m_sync_lock);
1082 list_add_tail(&work->w_list, &mp->m_sync_list);
1083 spin_unlock(&mp->m_sync_lock);
1084 wake_up_process(mp->m_sync_task);
1088 * Flush delayed allocate data, attempting to free up reserved space
1089 * from existing allocations. At this point a new allocation attempt
1090 * has failed with ENOSPC and we are in the process of scratching our
1091 * heads, looking about for more room...
1094 xfs_flush_inode_work(
1095 struct xfs_mount *mp,
1098 struct inode *inode = arg;
1099 filemap_flush(inode->i_mapping);
1107 struct inode *inode = VFS_I(ip);
1110 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
1111 delay(msecs_to_jiffies(500));
1115 * This is the "bigger hammer" version of xfs_flush_inode_work...
1116 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
1119 xfs_flush_device_work(
1120 struct xfs_mount *mp,
1123 struct inode *inode = arg;
1124 sync_blockdev(mp->m_super->s_bdev);
1132 struct inode *inode = VFS_I(ip);
1135 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1136 delay(msecs_to_jiffies(500));
1137 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1142 struct xfs_mount *mp,
1147 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1148 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR);
1150 wake_up(&mp->m_wait_single_sync_task);
1157 struct xfs_mount *mp = arg;
1159 bhv_vfs_sync_work_t *work, *n;
1163 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1165 timeleft = schedule_timeout_interruptible(timeleft);
1168 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1171 spin_lock(&mp->m_sync_lock);
1173 * We can get woken by laptop mode, to do a sync -
1174 * that's the (only!) case where the list would be
1175 * empty with time remaining.
1177 if (!timeleft || list_empty(&mp->m_sync_list)) {
1179 timeleft = xfs_syncd_centisecs *
1180 msecs_to_jiffies(10);
1181 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1182 list_add_tail(&mp->m_sync_work.w_list,
1185 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1186 list_move(&work->w_list, &tmp);
1187 spin_unlock(&mp->m_sync_lock);
1189 list_for_each_entry_safe(work, n, &tmp, w_list) {
1190 (*work->w_syncer)(mp, work->w_data);
1191 list_del(&work->w_list);
1192 if (work == &mp->m_sync_work)
1203 struct xfs_mount *mp)
1205 kfree(mp->m_fsname);
1206 kfree(mp->m_rtname);
1207 kfree(mp->m_logname);
1212 struct super_block *sb)
1214 struct xfs_mount *mp = XFS_M(sb);
1215 struct xfs_inode *rip = mp->m_rootip;
1216 int unmount_event_flags = 0;
1219 kthread_stop(mp->m_sync_task);
1221 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1224 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1225 unmount_event_flags =
1226 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1227 0 : DM_FLAGS_UNWANTED;
1229 * Ignore error from dmapi here, first unmount is not allowed
1230 * to fail anyway, and second we wouldn't want to fail a
1231 * unmount because of dmapi.
1233 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1234 NULL, NULL, 0, 0, unmount_event_flags);
1239 * Blow away any referenced inode in the filestreams cache.
1240 * This can and will cause log traffic as inodes go inactive
1243 xfs_filestream_unmount(mp);
1245 XFS_bflush(mp->m_ddev_targp);
1246 error = xfs_unmount_flush(mp, 0);
1252 * If we're forcing a shutdown, typically because of a media error,
1253 * we want to make sure we invalidate dirty pages that belong to
1254 * referenced vnodes as well.
1256 if (XFS_FORCED_SHUTDOWN(mp)) {
1257 error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
1258 ASSERT(error != EFSCORRUPTED);
1261 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1262 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1263 unmount_event_flags);
1267 xfs_icsb_destroy_counters(mp);
1268 xfs_close_devices(mp);
1271 xfs_free_fsname(mp);
1277 struct super_block *sb)
1279 if (!(sb->s_flags & MS_RDONLY))
1280 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1286 struct super_block *sb,
1289 struct xfs_mount *mp = XFS_M(sb);
1294 * Treat a sync operation like a freeze. This is to work
1295 * around a race in sync_inodes() which works in two phases
1296 * - an asynchronous flush, which can write out an inode
1297 * without waiting for file size updates to complete, and a
1298 * synchronous flush, which wont do anything because the
1299 * async flush removed the inode's dirty flag. Also
1300 * sync_inodes() will not see any files that just have
1301 * outstanding transactions to be flushed because we don't
1302 * dirty the Linux inode until after the transaction I/O
1305 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1307 * First stage of freeze - no more writers will make progress
1308 * now we are here, so we flush delwri and delalloc buffers
1309 * here, then wait for all I/O to complete. Data is frozen at
1310 * that point. Metadata is not frozen, transactions can still
1311 * occur here so don't bother flushing the buftarg (i.e
1312 * SYNC_QUIESCE) because it'll just get dirty again.
1314 flags = SYNC_DATA_QUIESCE;
1316 flags = SYNC_FSDATA;
1318 error = xfs_sync(mp, flags);
1321 if (unlikely(laptop_mode)) {
1322 int prev_sync_seq = mp->m_sync_seq;
1325 * The disk must be active because we're syncing.
1326 * We schedule xfssyncd now (now that the disk is
1327 * active) instead of later (when it might not be).
1329 wake_up_process(mp->m_sync_task);
1331 * We have to wait for the sync iteration to complete.
1332 * If we don't, the disk activity caused by the sync
1333 * will come after the sync is completed, and that
1334 * triggers another sync from laptop mode.
1336 wait_event(mp->m_wait_single_sync_task,
1337 mp->m_sync_seq != prev_sync_seq);
1345 struct dentry *dentry,
1346 struct kstatfs *statp)
1348 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1349 xfs_sb_t *sbp = &mp->m_sb;
1350 __uint64_t fakeinos, id;
1353 statp->f_type = XFS_SB_MAGIC;
1354 statp->f_namelen = MAXNAMELEN - 1;
1356 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1357 statp->f_fsid.val[0] = (u32)id;
1358 statp->f_fsid.val[1] = (u32)(id >> 32);
1360 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1362 spin_lock(&mp->m_sb_lock);
1363 statp->f_bsize = sbp->sb_blocksize;
1364 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1365 statp->f_blocks = sbp->sb_dblocks - lsize;
1366 statp->f_bfree = statp->f_bavail =
1367 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1368 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1370 fakeinos += mp->m_inoadd;
1373 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1374 if (mp->m_maxicount)
1378 statp->f_files = min_t(typeof(statp->f_files),
1381 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1382 spin_unlock(&mp->m_sb_lock);
1384 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1390 struct super_block *sb,
1394 struct xfs_mount *mp = XFS_M(sb);
1395 substring_t args[MAX_OPT_ARGS];
1398 while ((p = strsep(&options, ",")) != NULL) {
1404 token = match_token(p, tokens, args);
1407 mp->m_flags |= XFS_MOUNT_BARRIER;
1410 * Test if barriers are actually working if we can,
1411 * else delay this check until the filesystem is
1414 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1415 xfs_mountfs_check_barriers(mp);
1418 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1422 "XFS: mount option \"%s\" not supported for remount\n", p);
1428 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1429 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1430 if (mp->m_flags & XFS_MOUNT_BARRIER)
1431 xfs_mountfs_check_barriers(mp);
1435 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1436 xfs_filestream_flush(mp);
1437 xfs_sync(mp, SYNC_DATA_QUIESCE);
1438 xfs_attr_quiesce(mp);
1439 mp->m_flags |= XFS_MOUNT_RDONLY;
1446 * Second stage of a freeze. The data is already frozen so we only
1447 * need to take care of themetadata. Once that's done write a dummy
1448 * record to dirty the log in case of a crash while frozen.
1452 struct super_block *sb)
1454 struct xfs_mount *mp = XFS_M(sb);
1456 xfs_attr_quiesce(mp);
1457 xfs_fs_log_dummy(mp);
1461 xfs_fs_show_options(
1463 struct vfsmount *mnt)
1465 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1470 struct super_block *sb,
1473 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1478 struct super_block *sb,
1479 struct fs_quota_stat *fqs)
1481 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1486 struct super_block *sb,
1490 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1495 struct super_block *sb,
1498 struct fs_disk_quota *fdq)
1500 return -XFS_QM_QUOTACTL(XFS_M(sb),
1501 (type == USRQUOTA) ? Q_XGETQUOTA :
1502 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1503 Q_XGETPQUOTA), id, (caddr_t)fdq);
1508 struct super_block *sb,
1511 struct fs_disk_quota *fdq)
1513 return -XFS_QM_QUOTACTL(XFS_M(sb),
1514 (type == USRQUOTA) ? Q_XSETQLIM :
1515 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1516 Q_XSETPQLIM), id, (caddr_t)fdq);
1520 * This function fills in xfs_mount_t fields based on mount args.
1521 * Note: the superblock has _not_ yet been read in.
1525 struct xfs_mount_args *ap,
1526 struct xfs_mount *mp)
1530 /* Values are in BBs */
1531 if ((ap->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
1533 * At this point the superblock has not been read
1534 * in, therefore we do not know the block size.
1535 * Before the mount call ends we will convert
1538 mp->m_dalign = ap->sunit;
1539 mp->m_swidth = ap->swidth;
1542 if (ap->logbufs != -1 &&
1544 (ap->logbufs < XLOG_MIN_ICLOGS ||
1545 ap->logbufs > XLOG_MAX_ICLOGS)) {
1547 "XFS: invalid logbufs value: %d [not %d-%d]",
1548 ap->logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
1549 return XFS_ERROR(EINVAL);
1551 mp->m_logbufs = ap->logbufs;
1552 if (ap->logbufsize != -1 &&
1553 ap->logbufsize != 0 &&
1554 (ap->logbufsize < XLOG_MIN_RECORD_BSIZE ||
1555 ap->logbufsize > XLOG_MAX_RECORD_BSIZE ||
1556 !is_power_of_2(ap->logbufsize))) {
1558 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1560 return XFS_ERROR(EINVAL);
1565 mp->m_logbsize = ap->logbufsize;
1566 mp->m_fsname_len = strlen(ap->fsname) + 1;
1568 mp->m_fsname = kstrdup(ap->fsname, GFP_KERNEL);
1572 if (ap->rtname[0]) {
1573 mp->m_rtname = kstrdup(ap->rtname, GFP_KERNEL);
1575 goto out_free_fsname;
1579 if (ap->logname[0]) {
1580 mp->m_logname = kstrdup(ap->logname, GFP_KERNEL);
1582 goto out_free_rtname;
1585 if (ap->flags & XFSMNT_WSYNC)
1586 mp->m_flags |= XFS_MOUNT_WSYNC;
1588 if (ap->flags & XFSMNT_INO64) {
1589 mp->m_flags |= XFS_MOUNT_INO64;
1590 mp->m_inoadd = XFS_INO64_OFFSET;
1593 if (ap->flags & XFSMNT_RETERR)
1594 mp->m_flags |= XFS_MOUNT_RETERR;
1595 if (ap->flags & XFSMNT_NOALIGN)
1596 mp->m_flags |= XFS_MOUNT_NOALIGN;
1597 if (ap->flags & XFSMNT_SWALLOC)
1598 mp->m_flags |= XFS_MOUNT_SWALLOC;
1599 if (ap->flags & XFSMNT_OSYNCISOSYNC)
1600 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
1601 if (ap->flags & XFSMNT_32BITINODES)
1602 mp->m_flags |= XFS_MOUNT_32BITINODES;
1604 if (ap->flags & XFSMNT_IOSIZE) {
1605 if (ap->iosizelog > XFS_MAX_IO_LOG ||
1606 ap->iosizelog < XFS_MIN_IO_LOG) {
1608 "XFS: invalid log iosize: %d [not %d-%d]",
1609 ap->iosizelog, XFS_MIN_IO_LOG,
1611 return XFS_ERROR(EINVAL);
1614 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
1615 mp->m_readio_log = mp->m_writeio_log = ap->iosizelog;
1618 if (ap->flags & XFSMNT_IKEEP)
1619 mp->m_flags |= XFS_MOUNT_IKEEP;
1620 if (ap->flags & XFSMNT_DIRSYNC)
1621 mp->m_flags |= XFS_MOUNT_DIRSYNC;
1622 if (ap->flags & XFSMNT_ATTR2)
1623 mp->m_flags |= XFS_MOUNT_ATTR2;
1624 if (ap->flags & XFSMNT_NOATTR2)
1625 mp->m_flags |= XFS_MOUNT_NOATTR2;
1627 if (ap->flags2 & XFSMNT2_COMPAT_IOSIZE)
1628 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
1631 * no recovery flag requires a read-only mount
1633 if (ap->flags & XFSMNT_NORECOVERY) {
1634 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
1636 "XFS: tried to mount a FS read-write without recovery!");
1637 return XFS_ERROR(EINVAL);
1639 mp->m_flags |= XFS_MOUNT_NORECOVERY;
1642 if (ap->flags & XFSMNT_NOUUID)
1643 mp->m_flags |= XFS_MOUNT_NOUUID;
1644 if (ap->flags & XFSMNT_BARRIER)
1645 mp->m_flags |= XFS_MOUNT_BARRIER;
1647 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1649 if (ap->flags2 & XFSMNT2_FILESTREAMS)
1650 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
1652 if (ap->flags & XFSMNT_DMAPI)
1653 mp->m_flags |= XFS_MOUNT_DMAPI;
1658 kfree(mp->m_rtname);
1660 kfree(mp->m_fsname);
1666 * This function fills in xfs_mount_t fields based on mount args.
1667 * Note: the superblock _has_ now been read in.
1671 struct xfs_mount_args *ap,
1672 struct xfs_mount *mp)
1674 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1676 /* Fail a mount where the logbuf is smaller then the log stripe */
1677 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1678 if ((ap->logbufsize <= 0) &&
1679 (mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE)) {
1680 mp->m_logbsize = mp->m_sb.sb_logsunit;
1681 } else if (ap->logbufsize > 0 &&
1682 ap->logbufsize < mp->m_sb.sb_logsunit) {
1684 "XFS: logbuf size must be greater than or equal to log stripe size");
1685 return XFS_ERROR(EINVAL);
1688 /* Fail a mount if the logbuf is larger than 32K */
1689 if (ap->logbufsize > XLOG_BIG_RECORD_BSIZE) {
1691 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1692 return XFS_ERROR(EINVAL);
1697 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1698 * told by noattr2 to turn it off
1700 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1701 !(ap->flags & XFSMNT_NOATTR2))
1702 mp->m_flags |= XFS_MOUNT_ATTR2;
1705 * prohibit r/w mounts of read-only filesystems
1707 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1709 "XFS: cannot mount a read-only filesystem as read-write");
1710 return XFS_ERROR(EROFS);
1714 * check for shared mount.
1716 if (ap->flags & XFSMNT_SHARED) {
1717 if (!xfs_sb_version_hasshared(&mp->m_sb))
1718 return XFS_ERROR(EINVAL);
1721 * For IRIX 6.5, shared mounts must have the shared
1722 * version bit set, have the persistent readonly
1723 * field set, must be version 0 and can only be mounted
1726 if (!ronly || !(mp->m_sb.sb_flags & XFS_SBF_READONLY) ||
1727 (mp->m_sb.sb_shared_vn != 0))
1728 return XFS_ERROR(EINVAL);
1730 mp->m_flags |= XFS_MOUNT_SHARED;
1733 * Shared XFS V0 can't deal with DMI. Return EINVAL.
1735 if (mp->m_sb.sb_shared_vn == 0 && (ap->flags & XFSMNT_DMAPI))
1736 return XFS_ERROR(EINVAL);
1739 if (ap->flags & XFSMNT_UQUOTA) {
1740 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
1741 if (ap->flags & XFSMNT_UQUOTAENF)
1742 mp->m_qflags |= XFS_UQUOTA_ENFD;
1745 if (ap->flags & XFSMNT_GQUOTA) {
1746 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
1747 if (ap->flags & XFSMNT_GQUOTAENF)
1748 mp->m_qflags |= XFS_OQUOTA_ENFD;
1749 } else if (ap->flags & XFSMNT_PQUOTA) {
1750 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
1751 if (ap->flags & XFSMNT_PQUOTAENF)
1752 mp->m_qflags |= XFS_OQUOTA_ENFD;
1760 struct super_block *sb,
1765 struct xfs_mount *mp = NULL;
1766 struct xfs_mount_args *args;
1767 int flags = 0, error = ENOMEM;
1769 args = xfs_args_allocate(sb, silent);
1773 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1777 spin_lock_init(&mp->m_sb_lock);
1778 mutex_init(&mp->m_ilock);
1779 mutex_init(&mp->m_growlock);
1780 atomic_set(&mp->m_active_trans, 0);
1781 INIT_LIST_HEAD(&mp->m_sync_list);
1782 spin_lock_init(&mp->m_sync_lock);
1783 init_waitqueue_head(&mp->m_wait_single_sync_task);
1788 if (sb->s_flags & MS_RDONLY)
1789 mp->m_flags |= XFS_MOUNT_RDONLY;
1791 error = xfs_parseargs(mp, (char *)data, args, 0);
1795 sb_min_blocksize(sb, BBSIZE);
1796 sb->s_xattr = xfs_xattr_handlers;
1797 sb->s_export_op = &xfs_export_operations;
1798 sb->s_qcop = &xfs_quotactl_operations;
1799 sb->s_op = &xfs_super_operations;
1801 error = xfs_dmops_get(mp, args);
1804 error = xfs_qmops_get(mp, args);
1808 if (args->flags & XFSMNT_QUIET)
1809 flags |= XFS_MFSI_QUIET;
1811 error = xfs_open_devices(mp, args);
1815 if (xfs_icsb_init_counters(mp))
1816 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1819 * Setup flags based on mount(2) options and then the superblock
1821 error = xfs_start_flags(args, mp);
1823 goto out_free_fsname;
1824 error = xfs_readsb(mp, flags);
1826 goto out_free_fsname;
1827 error = xfs_finish_flags(args, mp);
1831 error = xfs_setup_devices(mp);
1835 if (mp->m_flags & XFS_MOUNT_BARRIER)
1836 xfs_mountfs_check_barriers(mp);
1838 error = xfs_filestream_mount(mp);
1842 error = xfs_mountfs(mp, flags);
1844 goto out_filestream_unmount;
1846 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, args->mtpt, args->fsname);
1849 sb->s_magic = XFS_SB_MAGIC;
1850 sb->s_blocksize = mp->m_sb.sb_blocksize;
1851 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1852 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1853 sb->s_time_gran = 1;
1854 set_posix_acl_flag(sb);
1856 root = igrab(VFS_I(mp->m_rootip));
1861 if (is_bad_inode(root)) {
1865 sb->s_root = d_alloc_root(root);
1871 mp->m_sync_work.w_syncer = xfs_sync_worker;
1872 mp->m_sync_work.w_mount = mp;
1873 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1874 if (IS_ERR(mp->m_sync_task)) {
1875 error = -PTR_ERR(mp->m_sync_task);
1879 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1884 out_filestream_unmount:
1885 xfs_filestream_unmount(mp);
1889 xfs_free_fsname(mp);
1890 xfs_icsb_destroy_counters(mp);
1891 xfs_close_devices(mp);
1912 * Blow away any referenced inode in the filestreams cache.
1913 * This can and will cause log traffic as inodes go inactive
1916 xfs_filestream_unmount(mp);
1918 XFS_bflush(mp->m_ddev_targp);
1919 error = xfs_unmount_flush(mp, 0);
1922 IRELE(mp->m_rootip);
1925 goto out_free_fsname;
1930 struct file_system_type *fs_type,
1932 const char *dev_name,
1934 struct vfsmount *mnt)
1936 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1940 static struct super_operations xfs_super_operations = {
1941 .alloc_inode = xfs_fs_alloc_inode,
1942 .destroy_inode = xfs_fs_destroy_inode,
1943 .write_inode = xfs_fs_write_inode,
1944 .clear_inode = xfs_fs_clear_inode,
1945 .put_super = xfs_fs_put_super,
1946 .write_super = xfs_fs_write_super,
1947 .sync_fs = xfs_fs_sync_super,
1948 .write_super_lockfs = xfs_fs_lockfs,
1949 .statfs = xfs_fs_statfs,
1950 .remount_fs = xfs_fs_remount,
1951 .show_options = xfs_fs_show_options,
1954 static struct quotactl_ops xfs_quotactl_operations = {
1955 .quota_sync = xfs_fs_quotasync,
1956 .get_xstate = xfs_fs_getxstate,
1957 .set_xstate = xfs_fs_setxstate,
1958 .get_xquota = xfs_fs_getxquota,
1959 .set_xquota = xfs_fs_setxquota,
1962 static struct file_system_type xfs_fs_type = {
1963 .owner = THIS_MODULE,
1965 .get_sb = xfs_fs_get_sb,
1966 .kill_sb = kill_block_super,
1967 .fs_flags = FS_REQUIRES_DEV,
1971 xfs_alloc_trace_bufs(void)
1973 #ifdef XFS_ALLOC_TRACE
1974 xfs_alloc_trace_buf = ktrace_alloc(XFS_ALLOC_TRACE_SIZE, KM_MAYFAIL);
1975 if (!xfs_alloc_trace_buf)
1978 #ifdef XFS_BMAP_TRACE
1979 xfs_bmap_trace_buf = ktrace_alloc(XFS_BMAP_TRACE_SIZE, KM_MAYFAIL);
1980 if (!xfs_bmap_trace_buf)
1981 goto out_free_alloc_trace;
1983 #ifdef XFS_BMBT_TRACE
1984 xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_MAYFAIL);
1985 if (!xfs_bmbt_trace_buf)
1986 goto out_free_bmap_trace;
1988 #ifdef XFS_ATTR_TRACE
1989 xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_MAYFAIL);
1990 if (!xfs_attr_trace_buf)
1991 goto out_free_bmbt_trace;
1993 #ifdef XFS_DIR2_TRACE
1994 xfs_dir2_trace_buf = ktrace_alloc(XFS_DIR2_GTRACE_SIZE, KM_MAYFAIL);
1995 if (!xfs_dir2_trace_buf)
1996 goto out_free_attr_trace;
2001 #ifdef XFS_DIR2_TRACE
2002 out_free_attr_trace:
2004 #ifdef XFS_ATTR_TRACE
2005 ktrace_free(xfs_attr_trace_buf);
2006 out_free_bmbt_trace:
2008 #ifdef XFS_BMBT_TRACE
2009 ktrace_free(xfs_bmbt_trace_buf);
2010 out_free_bmap_trace:
2012 #ifdef XFS_BMAP_TRACE
2013 ktrace_free(xfs_bmap_trace_buf);
2014 out_free_alloc_trace:
2016 #ifdef XFS_ALLOC_TRACE
2017 ktrace_free(xfs_alloc_trace_buf);
2024 xfs_free_trace_bufs(void)
2026 #ifdef XFS_DIR2_TRACE
2027 ktrace_free(xfs_dir2_trace_buf);
2029 #ifdef XFS_ATTR_TRACE
2030 ktrace_free(xfs_attr_trace_buf);
2032 #ifdef XFS_BMBT_TRACE
2033 ktrace_free(xfs_bmbt_trace_buf);
2035 #ifdef XFS_BMAP_TRACE
2036 ktrace_free(xfs_bmap_trace_buf);
2038 #ifdef XFS_ALLOC_TRACE
2039 ktrace_free(xfs_alloc_trace_buf);
2044 xfs_init_zones(void)
2046 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
2047 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
2049 xfs_fs_inode_init_once);
2050 if (!xfs_vnode_zone)
2053 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
2054 if (!xfs_ioend_zone)
2055 goto out_destroy_vnode_zone;
2057 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
2059 if (!xfs_ioend_pool)
2060 goto out_destroy_ioend_zone;
2062 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
2064 if (!xfs_log_ticket_zone)
2065 goto out_destroy_ioend_pool;
2067 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
2068 "xfs_bmap_free_item");
2069 if (!xfs_bmap_free_item_zone)
2070 goto out_destroy_log_ticket_zone;
2071 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
2073 if (!xfs_btree_cur_zone)
2074 goto out_destroy_bmap_free_item_zone;
2076 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
2078 if (!xfs_da_state_zone)
2079 goto out_destroy_btree_cur_zone;
2081 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
2082 if (!xfs_dabuf_zone)
2083 goto out_destroy_da_state_zone;
2085 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
2086 if (!xfs_ifork_zone)
2087 goto out_destroy_dabuf_zone;
2089 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
2090 if (!xfs_trans_zone)
2091 goto out_destroy_ifork_zone;
2094 * The size of the zone allocated buf log item is the maximum
2095 * size possible under XFS. This wastes a little bit of memory,
2096 * but it is much faster.
2098 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
2099 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
2100 NBWORD) * sizeof(int))), "xfs_buf_item");
2101 if (!xfs_buf_item_zone)
2102 goto out_destroy_trans_zone;
2104 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
2105 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
2106 sizeof(xfs_extent_t))), "xfs_efd_item");
2108 goto out_destroy_buf_item_zone;
2110 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
2111 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
2112 sizeof(xfs_extent_t))), "xfs_efi_item");
2114 goto out_destroy_efd_zone;
2117 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
2118 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
2119 KM_ZONE_SPREAD, NULL);
2120 if (!xfs_inode_zone)
2121 goto out_destroy_efi_zone;
2124 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
2125 KM_ZONE_SPREAD, NULL);
2127 goto out_destroy_inode_zone;
2129 #ifdef CONFIG_XFS_POSIX_ACL
2130 xfs_acl_zone = kmem_zone_init(sizeof(xfs_acl_t), "xfs_acl");
2132 goto out_destroy_ili_zone;
2137 #ifdef CONFIG_XFS_POSIX_ACL
2138 out_destroy_ili_zone:
2140 kmem_zone_destroy(xfs_ili_zone);
2141 out_destroy_inode_zone:
2142 kmem_zone_destroy(xfs_inode_zone);
2143 out_destroy_efi_zone:
2144 kmem_zone_destroy(xfs_efi_zone);
2145 out_destroy_efd_zone:
2146 kmem_zone_destroy(xfs_efd_zone);
2147 out_destroy_buf_item_zone:
2148 kmem_zone_destroy(xfs_buf_item_zone);
2149 out_destroy_trans_zone:
2150 kmem_zone_destroy(xfs_trans_zone);
2151 out_destroy_ifork_zone:
2152 kmem_zone_destroy(xfs_ifork_zone);
2153 out_destroy_dabuf_zone:
2154 kmem_zone_destroy(xfs_dabuf_zone);
2155 out_destroy_da_state_zone:
2156 kmem_zone_destroy(xfs_da_state_zone);
2157 out_destroy_btree_cur_zone:
2158 kmem_zone_destroy(xfs_btree_cur_zone);
2159 out_destroy_bmap_free_item_zone:
2160 kmem_zone_destroy(xfs_bmap_free_item_zone);
2161 out_destroy_log_ticket_zone:
2162 kmem_zone_destroy(xfs_log_ticket_zone);
2163 out_destroy_ioend_pool:
2164 mempool_destroy(xfs_ioend_pool);
2165 out_destroy_ioend_zone:
2166 kmem_zone_destroy(xfs_ioend_zone);
2167 out_destroy_vnode_zone:
2168 kmem_zone_destroy(xfs_vnode_zone);
2174 xfs_destroy_zones(void)
2176 #ifdef CONFIG_XFS_POSIX_ACL
2177 kmem_zone_destroy(xfs_acl_zone);
2179 kmem_zone_destroy(xfs_ili_zone);
2180 kmem_zone_destroy(xfs_inode_zone);
2181 kmem_zone_destroy(xfs_efi_zone);
2182 kmem_zone_destroy(xfs_efd_zone);
2183 kmem_zone_destroy(xfs_buf_item_zone);
2184 kmem_zone_destroy(xfs_trans_zone);
2185 kmem_zone_destroy(xfs_ifork_zone);
2186 kmem_zone_destroy(xfs_dabuf_zone);
2187 kmem_zone_destroy(xfs_da_state_zone);
2188 kmem_zone_destroy(xfs_btree_cur_zone);
2189 kmem_zone_destroy(xfs_bmap_free_item_zone);
2190 kmem_zone_destroy(xfs_log_ticket_zone);
2191 mempool_destroy(xfs_ioend_pool);
2192 kmem_zone_destroy(xfs_ioend_zone);
2193 kmem_zone_destroy(xfs_vnode_zone);
2201 static char message[] __initdata = KERN_INFO \
2202 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
2210 error = xfs_init_zones();
2214 error = xfs_alloc_trace_bufs();
2216 goto out_destroy_zones;
2218 error = xfs_mru_cache_init();
2220 goto out_free_trace_buffers;
2222 error = xfs_filestream_init();
2224 goto out_mru_cache_uninit;
2226 error = xfs_buf_init();
2228 goto out_filestream_uninit;
2230 error = xfs_init_procfs();
2232 goto out_buf_terminate;
2234 error = xfs_sysctl_register();
2236 goto out_cleanup_procfs;
2240 error = register_filesystem(&xfs_fs_type);
2242 goto out_sysctl_unregister;
2245 out_sysctl_unregister:
2246 xfs_sysctl_unregister();
2248 xfs_cleanup_procfs();
2250 xfs_buf_terminate();
2251 out_filestream_uninit:
2252 xfs_filestream_uninit();
2253 out_mru_cache_uninit:
2254 xfs_mru_cache_uninit();
2255 out_free_trace_buffers:
2256 xfs_free_trace_bufs();
2258 xfs_destroy_zones();
2267 unregister_filesystem(&xfs_fs_type);
2268 xfs_sysctl_unregister();
2269 xfs_cleanup_procfs();
2270 xfs_buf_terminate();
2271 xfs_filestream_uninit();
2272 xfs_mru_cache_uninit();
2273 xfs_free_trace_bufs();
2274 xfs_destroy_zones();
2278 module_init(init_xfs_fs);
2279 module_exit(exit_xfs_fs);
2281 MODULE_AUTHOR("Silicon Graphics, Inc.");
2282 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
2283 MODULE_LICENSE("GPL");