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(
619 struct inode *inode = vn_to_inode(vp);
621 inode->i_mode = ip->i_d.di_mode;
622 inode->i_nlink = ip->i_d.di_nlink;
623 inode->i_uid = ip->i_d.di_uid;
624 inode->i_gid = ip->i_d.di_gid;
626 switch (inode->i_mode & S_IFMT) {
630 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
631 sysv_minor(ip->i_df.if_u2.if_rdev));
638 inode->i_generation = ip->i_d.di_gen;
639 i_size_write(inode, ip->i_d.di_size);
640 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
641 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
642 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
643 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
644 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
645 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
646 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
647 inode->i_flags |= S_IMMUTABLE;
649 inode->i_flags &= ~S_IMMUTABLE;
650 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
651 inode->i_flags |= S_APPEND;
653 inode->i_flags &= ~S_APPEND;
654 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
655 inode->i_flags |= S_SYNC;
657 inode->i_flags &= ~S_SYNC;
658 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
659 inode->i_flags |= S_NOATIME;
661 inode->i_flags &= ~S_NOATIME;
662 xfs_iflags_clear(ip, XFS_IMODIFIED);
666 xfs_initialize_vnode(
667 struct xfs_mount *mp,
669 struct xfs_inode *ip)
671 struct inode *inode = vn_to_inode(vp);
675 inode->i_private = ip;
679 * We need to set the ops vectors, and unlock the inode, but if
680 * we have been called during the new inode create process, it is
681 * too early to fill in the Linux inode. We will get called a
682 * second time once the inode is properly set up, and then we can
685 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
686 xfs_revalidate_inode(mp, vp, ip);
687 xfs_set_inodeops(inode);
689 xfs_iflags_clear(ip, XFS_INEW);
692 unlock_new_inode(inode);
700 struct block_device **bdevp)
704 *bdevp = open_bdev_excl(name, 0, mp);
705 if (IS_ERR(*bdevp)) {
706 error = PTR_ERR(*bdevp);
707 printk("XFS: Invalid device [%s], error=%d\n", name, error);
715 struct block_device *bdev)
718 close_bdev_excl(bdev);
722 * Try to write out the superblock using barriers.
728 xfs_buf_t *sbp = xfs_getsb(mp, 0);
733 XFS_BUF_UNDELAYWRITE(sbp);
735 XFS_BUF_UNASYNC(sbp);
736 XFS_BUF_ORDERED(sbp);
739 error = xfs_iowait(sbp);
742 * Clear all the flags we set and possible error state in the
743 * buffer. We only did the write to try out whether barriers
744 * worked and shouldn't leave any traces in the superblock
748 XFS_BUF_ERROR(sbp, 0);
749 XFS_BUF_UNORDERED(sbp);
756 xfs_mountfs_check_barriers(xfs_mount_t *mp)
760 if (mp->m_logdev_targp != mp->m_ddev_targp) {
761 xfs_fs_cmn_err(CE_NOTE, mp,
762 "Disabling barriers, not supported with external log device");
763 mp->m_flags &= ~XFS_MOUNT_BARRIER;
767 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
768 xfs_fs_cmn_err(CE_NOTE, mp,
769 "Disabling barriers, underlying device is readonly");
770 mp->m_flags &= ~XFS_MOUNT_BARRIER;
774 error = xfs_barrier_test(mp);
776 xfs_fs_cmn_err(CE_NOTE, mp,
777 "Disabling barriers, trial barrier write failed");
778 mp->m_flags &= ~XFS_MOUNT_BARRIER;
784 xfs_blkdev_issue_flush(
785 xfs_buftarg_t *buftarg)
787 blkdev_issue_flush(buftarg->bt_bdev, NULL);
792 struct xfs_mount *mp)
794 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
795 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
796 xfs_free_buftarg(mp->m_logdev_targp);
797 xfs_blkdev_put(logdev);
799 if (mp->m_rtdev_targp) {
800 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
801 xfs_free_buftarg(mp->m_rtdev_targp);
802 xfs_blkdev_put(rtdev);
804 xfs_free_buftarg(mp->m_ddev_targp);
808 * The file system configurations are:
809 * (1) device (partition) with data and internal log
810 * (2) logical volume with data and log subvolumes.
811 * (3) logical volume with data, log, and realtime subvolumes.
813 * We only have to handle opening the log and realtime volumes here if
814 * they are present. The data subvolume has already been opened by
815 * get_sb_bdev() and is stored in sb->s_bdev.
819 struct xfs_mount *mp,
820 struct xfs_mount_args *args)
822 struct block_device *ddev = mp->m_super->s_bdev;
823 struct block_device *logdev = NULL, *rtdev = NULL;
827 * Open real time and log devices - order is important.
829 if (args->logname[0]) {
830 error = xfs_blkdev_get(mp, args->logname, &logdev);
835 if (args->rtname[0]) {
836 error = xfs_blkdev_get(mp, args->rtname, &rtdev);
838 goto out_close_logdev;
840 if (rtdev == ddev || rtdev == logdev) {
842 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
844 goto out_close_rtdev;
849 * Setup xfs_mount buffer target pointers
852 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
853 if (!mp->m_ddev_targp)
854 goto out_close_rtdev;
857 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
858 if (!mp->m_rtdev_targp)
859 goto out_free_ddev_targ;
862 if (logdev && logdev != ddev) {
863 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
864 if (!mp->m_logdev_targp)
865 goto out_free_rtdev_targ;
867 mp->m_logdev_targp = mp->m_ddev_targp;
873 if (mp->m_rtdev_targp)
874 xfs_free_buftarg(mp->m_rtdev_targp);
876 xfs_free_buftarg(mp->m_ddev_targp);
879 xfs_blkdev_put(rtdev);
881 if (logdev && logdev != ddev)
882 xfs_blkdev_put(logdev);
888 * Setup xfs_mount buffer target pointers based on superblock
892 struct xfs_mount *mp)
896 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
897 mp->m_sb.sb_sectsize);
901 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
902 unsigned int log_sector_size = BBSIZE;
904 if (xfs_sb_version_hassector(&mp->m_sb))
905 log_sector_size = mp->m_sb.sb_logsectsize;
906 error = xfs_setsize_buftarg(mp->m_logdev_targp,
907 mp->m_sb.sb_blocksize,
912 if (mp->m_rtdev_targp) {
913 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
914 mp->m_sb.sb_blocksize,
915 mp->m_sb.sb_sectsize);
924 * XFS AIL push thread support
929 xfs_lsn_t threshold_lsn)
931 mp->m_ail.xa_target = threshold_lsn;
932 wake_up_process(mp->m_ail.xa_task);
939 xfs_mount_t *mp = (xfs_mount_t *)data;
940 xfs_lsn_t last_pushed_lsn = 0;
943 while (!kthread_should_stop()) {
945 schedule_timeout_interruptible(msecs_to_jiffies(tout));
952 if (XFS_FORCED_SHUTDOWN(mp))
955 tout = xfsaild_push(mp, &last_pushed_lsn);
965 mp->m_ail.xa_target = 0;
966 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
967 if (IS_ERR(mp->m_ail.xa_task))
968 return -PTR_ERR(mp->m_ail.xa_task);
976 kthread_stop(mp->m_ail.xa_task);
981 STATIC struct inode *
983 struct super_block *sb)
987 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
990 return vn_to_inode(vp);
994 xfs_fs_destroy_inode(
997 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
1001 xfs_fs_inode_init_once(
1004 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
1008 * Attempt to flush the inode, this will actually fail
1009 * if the inode is pinned, but we dirty the inode again
1010 * at the point when it is unpinned after a log write,
1011 * since this is when the inode itself becomes flushable.
1015 struct inode *inode,
1021 xfs_itrace_entry(XFS_I(inode));
1023 filemap_fdatawait(inode->i_mapping);
1024 flags |= FLUSH_SYNC;
1026 error = xfs_inode_flush(XFS_I(inode), flags);
1028 * if we failed to write out the inode then mark
1029 * it dirty again so we'll try again later.
1032 mark_inode_dirty_sync(inode);
1039 struct inode *inode)
1041 xfs_inode_t *ip = XFS_I(inode);
1044 * ip can be null when xfs_iget_core calls xfs_idestroy if we
1045 * find an inode with di_mode == 0 but without IGET_CREATE set.
1048 xfs_itrace_entry(ip);
1049 XFS_STATS_INC(vn_rele);
1050 XFS_STATS_INC(vn_remove);
1051 XFS_STATS_INC(vn_reclaim);
1052 XFS_STATS_DEC(vn_active);
1055 xfs_iflags_clear(ip, XFS_IMODIFIED);
1056 if (xfs_reclaim(ip))
1057 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
1060 ASSERT(XFS_I(inode) == NULL);
1064 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
1065 * Doing this has two advantages:
1066 * - It saves on stack space, which is tight in certain situations
1067 * - It can be used (with care) as a mechanism to avoid deadlocks.
1068 * Flushing while allocating in a full filesystem requires both.
1071 xfs_syncd_queue_work(
1072 struct xfs_mount *mp,
1074 void (*syncer)(struct xfs_mount *, void *))
1076 struct bhv_vfs_sync_work *work;
1078 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
1079 INIT_LIST_HEAD(&work->w_list);
1080 work->w_syncer = syncer;
1081 work->w_data = data;
1083 spin_lock(&mp->m_sync_lock);
1084 list_add_tail(&work->w_list, &mp->m_sync_list);
1085 spin_unlock(&mp->m_sync_lock);
1086 wake_up_process(mp->m_sync_task);
1090 * Flush delayed allocate data, attempting to free up reserved space
1091 * from existing allocations. At this point a new allocation attempt
1092 * has failed with ENOSPC and we are in the process of scratching our
1093 * heads, looking about for more room...
1096 xfs_flush_inode_work(
1097 struct xfs_mount *mp,
1100 struct inode *inode = arg;
1101 filemap_flush(inode->i_mapping);
1109 struct inode *inode = ip->i_vnode;
1112 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
1113 delay(msecs_to_jiffies(500));
1117 * This is the "bigger hammer" version of xfs_flush_inode_work...
1118 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
1121 xfs_flush_device_work(
1122 struct xfs_mount *mp,
1125 struct inode *inode = arg;
1126 sync_blockdev(mp->m_super->s_bdev);
1134 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
1137 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1138 delay(msecs_to_jiffies(500));
1139 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1144 struct xfs_mount *mp,
1149 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1150 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR);
1152 wake_up(&mp->m_wait_single_sync_task);
1159 struct xfs_mount *mp = arg;
1161 bhv_vfs_sync_work_t *work, *n;
1165 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1167 timeleft = schedule_timeout_interruptible(timeleft);
1170 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1173 spin_lock(&mp->m_sync_lock);
1175 * We can get woken by laptop mode, to do a sync -
1176 * that's the (only!) case where the list would be
1177 * empty with time remaining.
1179 if (!timeleft || list_empty(&mp->m_sync_list)) {
1181 timeleft = xfs_syncd_centisecs *
1182 msecs_to_jiffies(10);
1183 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1184 list_add_tail(&mp->m_sync_work.w_list,
1187 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1188 list_move(&work->w_list, &tmp);
1189 spin_unlock(&mp->m_sync_lock);
1191 list_for_each_entry_safe(work, n, &tmp, w_list) {
1192 (*work->w_syncer)(mp, work->w_data);
1193 list_del(&work->w_list);
1194 if (work == &mp->m_sync_work)
1205 struct super_block *sb)
1207 struct xfs_mount *mp = XFS_M(sb);
1208 struct xfs_inode *rip = mp->m_rootip;
1209 int unmount_event_flags = 0;
1212 kthread_stop(mp->m_sync_task);
1214 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1217 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1218 unmount_event_flags =
1219 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1220 0 : DM_FLAGS_UNWANTED;
1222 * Ignore error from dmapi here, first unmount is not allowed
1223 * to fail anyway, and second we wouldn't want to fail a
1224 * unmount because of dmapi.
1226 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1227 NULL, NULL, 0, 0, unmount_event_flags);
1232 * Blow away any referenced inode in the filestreams cache.
1233 * This can and will cause log traffic as inodes go inactive
1236 xfs_filestream_unmount(mp);
1238 XFS_bflush(mp->m_ddev_targp);
1239 error = xfs_unmount_flush(mp, 0);
1245 * If we're forcing a shutdown, typically because of a media error,
1246 * we want to make sure we invalidate dirty pages that belong to
1247 * referenced vnodes as well.
1249 if (XFS_FORCED_SHUTDOWN(mp)) {
1250 error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
1251 ASSERT(error != EFSCORRUPTED);
1254 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1255 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1256 unmount_event_flags);
1260 xfs_icsb_destroy_counters(mp);
1261 xfs_close_devices(mp);
1269 struct super_block *sb)
1271 if (!(sb->s_flags & MS_RDONLY))
1272 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1278 struct super_block *sb,
1281 struct xfs_mount *mp = XFS_M(sb);
1286 * Treat a sync operation like a freeze. This is to work
1287 * around a race in sync_inodes() which works in two phases
1288 * - an asynchronous flush, which can write out an inode
1289 * without waiting for file size updates to complete, and a
1290 * synchronous flush, which wont do anything because the
1291 * async flush removed the inode's dirty flag. Also
1292 * sync_inodes() will not see any files that just have
1293 * outstanding transactions to be flushed because we don't
1294 * dirty the Linux inode until after the transaction I/O
1297 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1299 * First stage of freeze - no more writers will make progress
1300 * now we are here, so we flush delwri and delalloc buffers
1301 * here, then wait for all I/O to complete. Data is frozen at
1302 * that point. Metadata is not frozen, transactions can still
1303 * occur here so don't bother flushing the buftarg (i.e
1304 * SYNC_QUIESCE) because it'll just get dirty again.
1306 flags = SYNC_DATA_QUIESCE;
1308 flags = SYNC_FSDATA;
1310 error = xfs_sync(mp, flags);
1313 if (unlikely(laptop_mode)) {
1314 int prev_sync_seq = mp->m_sync_seq;
1317 * The disk must be active because we're syncing.
1318 * We schedule xfssyncd now (now that the disk is
1319 * active) instead of later (when it might not be).
1321 wake_up_process(mp->m_sync_task);
1323 * We have to wait for the sync iteration to complete.
1324 * If we don't, the disk activity caused by the sync
1325 * will come after the sync is completed, and that
1326 * triggers another sync from laptop mode.
1328 wait_event(mp->m_wait_single_sync_task,
1329 mp->m_sync_seq != prev_sync_seq);
1337 struct dentry *dentry,
1338 struct kstatfs *statp)
1340 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1341 xfs_sb_t *sbp = &mp->m_sb;
1342 __uint64_t fakeinos, id;
1345 statp->f_type = XFS_SB_MAGIC;
1346 statp->f_namelen = MAXNAMELEN - 1;
1348 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1349 statp->f_fsid.val[0] = (u32)id;
1350 statp->f_fsid.val[1] = (u32)(id >> 32);
1352 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1354 spin_lock(&mp->m_sb_lock);
1355 statp->f_bsize = sbp->sb_blocksize;
1356 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1357 statp->f_blocks = sbp->sb_dblocks - lsize;
1358 statp->f_bfree = statp->f_bavail =
1359 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1360 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1362 fakeinos += mp->m_inoadd;
1365 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1366 if (mp->m_maxicount)
1370 statp->f_files = min_t(typeof(statp->f_files),
1373 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1374 spin_unlock(&mp->m_sb_lock);
1376 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1382 struct super_block *sb,
1386 struct xfs_mount *mp = XFS_M(sb);
1387 substring_t args[MAX_OPT_ARGS];
1390 while ((p = strsep(&options, ",")) != NULL) {
1396 token = match_token(p, tokens, args);
1399 mp->m_flags |= XFS_MOUNT_BARRIER;
1402 * Test if barriers are actually working if we can,
1403 * else delay this check until the filesystem is
1406 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1407 xfs_mountfs_check_barriers(mp);
1410 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1414 "XFS: mount option \"%s\" not supported for remount\n", p);
1420 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1421 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1422 if (mp->m_flags & XFS_MOUNT_BARRIER)
1423 xfs_mountfs_check_barriers(mp);
1427 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1428 xfs_filestream_flush(mp);
1429 xfs_sync(mp, SYNC_DATA_QUIESCE);
1430 xfs_attr_quiesce(mp);
1431 mp->m_flags |= XFS_MOUNT_RDONLY;
1438 * Second stage of a freeze. The data is already frozen so we only
1439 * need to take care of themetadata. Once that's done write a dummy
1440 * record to dirty the log in case of a crash while frozen.
1444 struct super_block *sb)
1446 struct xfs_mount *mp = XFS_M(sb);
1448 xfs_attr_quiesce(mp);
1449 xfs_fs_log_dummy(mp);
1453 xfs_fs_show_options(
1455 struct vfsmount *mnt)
1457 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1462 struct super_block *sb,
1465 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1470 struct super_block *sb,
1471 struct fs_quota_stat *fqs)
1473 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1478 struct super_block *sb,
1482 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1487 struct super_block *sb,
1490 struct fs_disk_quota *fdq)
1492 return -XFS_QM_QUOTACTL(XFS_M(sb),
1493 (type == USRQUOTA) ? Q_XGETQUOTA :
1494 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1495 Q_XGETPQUOTA), id, (caddr_t)fdq);
1500 struct super_block *sb,
1503 struct fs_disk_quota *fdq)
1505 return -XFS_QM_QUOTACTL(XFS_M(sb),
1506 (type == USRQUOTA) ? Q_XSETQLIM :
1507 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1508 Q_XSETPQLIM), id, (caddr_t)fdq);
1512 * This function fills in xfs_mount_t fields based on mount args.
1513 * Note: the superblock has _not_ yet been read in.
1517 struct xfs_mount_args *ap,
1518 struct xfs_mount *mp)
1520 /* Values are in BBs */
1521 if ((ap->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
1523 * At this point the superblock has not been read
1524 * in, therefore we do not know the block size.
1525 * Before the mount call ends we will convert
1528 mp->m_dalign = ap->sunit;
1529 mp->m_swidth = ap->swidth;
1532 if (ap->logbufs != -1 &&
1534 (ap->logbufs < XLOG_MIN_ICLOGS ||
1535 ap->logbufs > XLOG_MAX_ICLOGS)) {
1537 "XFS: invalid logbufs value: %d [not %d-%d]",
1538 ap->logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
1539 return XFS_ERROR(EINVAL);
1541 mp->m_logbufs = ap->logbufs;
1542 if (ap->logbufsize != -1 &&
1543 ap->logbufsize != 0 &&
1544 (ap->logbufsize < XLOG_MIN_RECORD_BSIZE ||
1545 ap->logbufsize > XLOG_MAX_RECORD_BSIZE ||
1546 !is_power_of_2(ap->logbufsize))) {
1548 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1550 return XFS_ERROR(EINVAL);
1552 mp->m_logbsize = ap->logbufsize;
1553 mp->m_fsname_len = strlen(ap->fsname) + 1;
1554 mp->m_fsname = kmem_alloc(mp->m_fsname_len, KM_SLEEP);
1555 strcpy(mp->m_fsname, ap->fsname);
1556 if (ap->rtname[0]) {
1557 mp->m_rtname = kmem_alloc(strlen(ap->rtname) + 1, KM_SLEEP);
1558 strcpy(mp->m_rtname, ap->rtname);
1560 if (ap->logname[0]) {
1561 mp->m_logname = kmem_alloc(strlen(ap->logname) + 1, KM_SLEEP);
1562 strcpy(mp->m_logname, ap->logname);
1565 if (ap->flags & XFSMNT_WSYNC)
1566 mp->m_flags |= XFS_MOUNT_WSYNC;
1568 if (ap->flags & XFSMNT_INO64) {
1569 mp->m_flags |= XFS_MOUNT_INO64;
1570 mp->m_inoadd = XFS_INO64_OFFSET;
1573 if (ap->flags & XFSMNT_RETERR)
1574 mp->m_flags |= XFS_MOUNT_RETERR;
1575 if (ap->flags & XFSMNT_NOALIGN)
1576 mp->m_flags |= XFS_MOUNT_NOALIGN;
1577 if (ap->flags & XFSMNT_SWALLOC)
1578 mp->m_flags |= XFS_MOUNT_SWALLOC;
1579 if (ap->flags & XFSMNT_OSYNCISOSYNC)
1580 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
1581 if (ap->flags & XFSMNT_32BITINODES)
1582 mp->m_flags |= XFS_MOUNT_32BITINODES;
1584 if (ap->flags & XFSMNT_IOSIZE) {
1585 if (ap->iosizelog > XFS_MAX_IO_LOG ||
1586 ap->iosizelog < XFS_MIN_IO_LOG) {
1588 "XFS: invalid log iosize: %d [not %d-%d]",
1589 ap->iosizelog, XFS_MIN_IO_LOG,
1591 return XFS_ERROR(EINVAL);
1594 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
1595 mp->m_readio_log = mp->m_writeio_log = ap->iosizelog;
1598 if (ap->flags & XFSMNT_IKEEP)
1599 mp->m_flags |= XFS_MOUNT_IKEEP;
1600 if (ap->flags & XFSMNT_DIRSYNC)
1601 mp->m_flags |= XFS_MOUNT_DIRSYNC;
1602 if (ap->flags & XFSMNT_ATTR2)
1603 mp->m_flags |= XFS_MOUNT_ATTR2;
1604 if (ap->flags & XFSMNT_NOATTR2)
1605 mp->m_flags |= XFS_MOUNT_NOATTR2;
1607 if (ap->flags2 & XFSMNT2_COMPAT_IOSIZE)
1608 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
1611 * no recovery flag requires a read-only mount
1613 if (ap->flags & XFSMNT_NORECOVERY) {
1614 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
1616 "XFS: tried to mount a FS read-write without recovery!");
1617 return XFS_ERROR(EINVAL);
1619 mp->m_flags |= XFS_MOUNT_NORECOVERY;
1622 if (ap->flags & XFSMNT_NOUUID)
1623 mp->m_flags |= XFS_MOUNT_NOUUID;
1624 if (ap->flags & XFSMNT_BARRIER)
1625 mp->m_flags |= XFS_MOUNT_BARRIER;
1627 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1629 if (ap->flags2 & XFSMNT2_FILESTREAMS)
1630 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
1632 if (ap->flags & XFSMNT_DMAPI)
1633 mp->m_flags |= XFS_MOUNT_DMAPI;
1638 * This function fills in xfs_mount_t fields based on mount args.
1639 * Note: the superblock _has_ now been read in.
1643 struct xfs_mount_args *ap,
1644 struct xfs_mount *mp)
1646 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1648 /* Fail a mount where the logbuf is smaller then the log stripe */
1649 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1650 if ((ap->logbufsize <= 0) &&
1651 (mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE)) {
1652 mp->m_logbsize = mp->m_sb.sb_logsunit;
1653 } else if (ap->logbufsize > 0 &&
1654 ap->logbufsize < mp->m_sb.sb_logsunit) {
1656 "XFS: logbuf size must be greater than or equal to log stripe size");
1657 return XFS_ERROR(EINVAL);
1660 /* Fail a mount if the logbuf is larger than 32K */
1661 if (ap->logbufsize > XLOG_BIG_RECORD_BSIZE) {
1663 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1664 return XFS_ERROR(EINVAL);
1669 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1670 * told by noattr2 to turn it off
1672 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1673 !(ap->flags & XFSMNT_NOATTR2))
1674 mp->m_flags |= XFS_MOUNT_ATTR2;
1677 * prohibit r/w mounts of read-only filesystems
1679 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1681 "XFS: cannot mount a read-only filesystem as read-write");
1682 return XFS_ERROR(EROFS);
1686 * check for shared mount.
1688 if (ap->flags & XFSMNT_SHARED) {
1689 if (!xfs_sb_version_hasshared(&mp->m_sb))
1690 return XFS_ERROR(EINVAL);
1693 * For IRIX 6.5, shared mounts must have the shared
1694 * version bit set, have the persistent readonly
1695 * field set, must be version 0 and can only be mounted
1698 if (!ronly || !(mp->m_sb.sb_flags & XFS_SBF_READONLY) ||
1699 (mp->m_sb.sb_shared_vn != 0))
1700 return XFS_ERROR(EINVAL);
1702 mp->m_flags |= XFS_MOUNT_SHARED;
1705 * Shared XFS V0 can't deal with DMI. Return EINVAL.
1707 if (mp->m_sb.sb_shared_vn == 0 && (ap->flags & XFSMNT_DMAPI))
1708 return XFS_ERROR(EINVAL);
1711 if (ap->flags & XFSMNT_UQUOTA) {
1712 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
1713 if (ap->flags & XFSMNT_UQUOTAENF)
1714 mp->m_qflags |= XFS_UQUOTA_ENFD;
1717 if (ap->flags & XFSMNT_GQUOTA) {
1718 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
1719 if (ap->flags & XFSMNT_GQUOTAENF)
1720 mp->m_qflags |= XFS_OQUOTA_ENFD;
1721 } else if (ap->flags & XFSMNT_PQUOTA) {
1722 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
1723 if (ap->flags & XFSMNT_PQUOTAENF)
1724 mp->m_qflags |= XFS_OQUOTA_ENFD;
1732 struct super_block *sb,
1737 struct xfs_mount *mp = NULL;
1738 struct xfs_mount_args *args;
1739 int flags = 0, error = ENOMEM;
1741 args = xfs_args_allocate(sb, silent);
1745 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1749 spin_lock_init(&mp->m_sb_lock);
1750 mutex_init(&mp->m_ilock);
1751 mutex_init(&mp->m_growlock);
1752 atomic_set(&mp->m_active_trans, 0);
1753 INIT_LIST_HEAD(&mp->m_sync_list);
1754 spin_lock_init(&mp->m_sync_lock);
1755 init_waitqueue_head(&mp->m_wait_single_sync_task);
1760 if (sb->s_flags & MS_RDONLY)
1761 mp->m_flags |= XFS_MOUNT_RDONLY;
1763 error = xfs_parseargs(mp, (char *)data, args, 0);
1767 sb_min_blocksize(sb, BBSIZE);
1768 sb->s_xattr = xfs_xattr_handlers;
1769 sb->s_export_op = &xfs_export_operations;
1770 sb->s_qcop = &xfs_quotactl_operations;
1771 sb->s_op = &xfs_super_operations;
1773 error = xfs_dmops_get(mp, args);
1776 error = xfs_qmops_get(mp, args);
1780 if (args->flags & XFSMNT_QUIET)
1781 flags |= XFS_MFSI_QUIET;
1783 error = xfs_open_devices(mp, args);
1787 if (xfs_icsb_init_counters(mp))
1788 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1791 * Setup flags based on mount(2) options and then the superblock
1793 error = xfs_start_flags(args, mp);
1795 goto out_destroy_counters;
1796 error = xfs_readsb(mp, flags);
1798 goto out_destroy_counters;
1799 error = xfs_finish_flags(args, mp);
1803 error = xfs_setup_devices(mp);
1807 if (mp->m_flags & XFS_MOUNT_BARRIER)
1808 xfs_mountfs_check_barriers(mp);
1810 error = xfs_filestream_mount(mp);
1814 error = xfs_mountfs(mp, flags);
1816 goto out_filestream_unmount;
1818 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, args->mtpt, args->fsname);
1821 sb->s_magic = XFS_SB_MAGIC;
1822 sb->s_blocksize = mp->m_sb.sb_blocksize;
1823 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1824 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1825 sb->s_time_gran = 1;
1826 set_posix_acl_flag(sb);
1828 root = igrab(mp->m_rootip->i_vnode);
1833 if (is_bad_inode(root)) {
1837 sb->s_root = d_alloc_root(root);
1843 mp->m_sync_work.w_syncer = xfs_sync_worker;
1844 mp->m_sync_work.w_mount = mp;
1845 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1846 if (IS_ERR(mp->m_sync_task)) {
1847 error = -PTR_ERR(mp->m_sync_task);
1851 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1856 out_filestream_unmount:
1857 xfs_filestream_unmount(mp);
1860 out_destroy_counters:
1861 xfs_icsb_destroy_counters(mp);
1862 xfs_close_devices(mp);
1883 * Blow away any referenced inode in the filestreams cache.
1884 * This can and will cause log traffic as inodes go inactive
1887 xfs_filestream_unmount(mp);
1889 XFS_bflush(mp->m_ddev_targp);
1890 error = xfs_unmount_flush(mp, 0);
1893 IRELE(mp->m_rootip);
1896 goto out_destroy_counters;
1901 struct file_system_type *fs_type,
1903 const char *dev_name,
1905 struct vfsmount *mnt)
1907 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1911 static struct super_operations xfs_super_operations = {
1912 .alloc_inode = xfs_fs_alloc_inode,
1913 .destroy_inode = xfs_fs_destroy_inode,
1914 .write_inode = xfs_fs_write_inode,
1915 .clear_inode = xfs_fs_clear_inode,
1916 .put_super = xfs_fs_put_super,
1917 .write_super = xfs_fs_write_super,
1918 .sync_fs = xfs_fs_sync_super,
1919 .write_super_lockfs = xfs_fs_lockfs,
1920 .statfs = xfs_fs_statfs,
1921 .remount_fs = xfs_fs_remount,
1922 .show_options = xfs_fs_show_options,
1925 static struct quotactl_ops xfs_quotactl_operations = {
1926 .quota_sync = xfs_fs_quotasync,
1927 .get_xstate = xfs_fs_getxstate,
1928 .set_xstate = xfs_fs_setxstate,
1929 .get_xquota = xfs_fs_getxquota,
1930 .set_xquota = xfs_fs_setxquota,
1933 static struct file_system_type xfs_fs_type = {
1934 .owner = THIS_MODULE,
1936 .get_sb = xfs_fs_get_sb,
1937 .kill_sb = kill_block_super,
1938 .fs_flags = FS_REQUIRES_DEV,
1942 xfs_alloc_trace_bufs(void)
1944 #ifdef XFS_ALLOC_TRACE
1945 xfs_alloc_trace_buf = ktrace_alloc(XFS_ALLOC_TRACE_SIZE, KM_MAYFAIL);
1946 if (!xfs_alloc_trace_buf)
1949 #ifdef XFS_BMAP_TRACE
1950 xfs_bmap_trace_buf = ktrace_alloc(XFS_BMAP_TRACE_SIZE, KM_MAYFAIL);
1951 if (!xfs_bmap_trace_buf)
1952 goto out_free_alloc_trace;
1954 #ifdef XFS_BMBT_TRACE
1955 xfs_bmbt_trace_buf = ktrace_alloc(XFS_BMBT_TRACE_SIZE, KM_MAYFAIL);
1956 if (!xfs_bmbt_trace_buf)
1957 goto out_free_bmap_trace;
1959 #ifdef XFS_ATTR_TRACE
1960 xfs_attr_trace_buf = ktrace_alloc(XFS_ATTR_TRACE_SIZE, KM_MAYFAIL);
1961 if (!xfs_attr_trace_buf)
1962 goto out_free_bmbt_trace;
1964 #ifdef XFS_DIR2_TRACE
1965 xfs_dir2_trace_buf = ktrace_alloc(XFS_DIR2_GTRACE_SIZE, KM_MAYFAIL);
1966 if (!xfs_dir2_trace_buf)
1967 goto out_free_attr_trace;
1972 #ifdef XFS_DIR2_TRACE
1973 out_free_attr_trace:
1975 #ifdef XFS_ATTR_TRACE
1976 ktrace_free(xfs_attr_trace_buf);
1977 out_free_bmbt_trace:
1979 #ifdef XFS_BMBT_TRACE
1980 ktrace_free(xfs_bmbt_trace_buf);
1981 out_free_bmap_trace:
1983 #ifdef XFS_BMAP_TRACE
1984 ktrace_free(xfs_bmap_trace_buf);
1985 out_free_alloc_trace:
1987 #ifdef XFS_ALLOC_TRACE
1988 ktrace_free(xfs_alloc_trace_buf);
1995 xfs_free_trace_bufs(void)
1997 #ifdef XFS_DIR2_TRACE
1998 ktrace_free(xfs_dir2_trace_buf);
2000 #ifdef XFS_ATTR_TRACE
2001 ktrace_free(xfs_attr_trace_buf);
2003 #ifdef XFS_BMBT_TRACE
2004 ktrace_free(xfs_bmbt_trace_buf);
2006 #ifdef XFS_BMAP_TRACE
2007 ktrace_free(xfs_bmap_trace_buf);
2009 #ifdef XFS_ALLOC_TRACE
2010 ktrace_free(xfs_alloc_trace_buf);
2015 xfs_init_zones(void)
2017 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
2018 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
2020 xfs_fs_inode_init_once);
2021 if (!xfs_vnode_zone)
2024 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
2025 if (!xfs_ioend_zone)
2026 goto out_destroy_vnode_zone;
2028 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
2030 if (!xfs_ioend_pool)
2031 goto out_destroy_ioend_zone;
2033 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
2035 if (!xfs_log_ticket_zone)
2036 goto out_destroy_ioend_pool;
2038 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
2039 "xfs_bmap_free_item");
2040 if (!xfs_bmap_free_item_zone)
2041 goto out_destroy_log_ticket_zone;
2042 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
2044 if (!xfs_btree_cur_zone)
2045 goto out_destroy_bmap_free_item_zone;
2047 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
2049 if (!xfs_da_state_zone)
2050 goto out_destroy_btree_cur_zone;
2052 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
2053 if (!xfs_dabuf_zone)
2054 goto out_destroy_da_state_zone;
2056 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
2057 if (!xfs_ifork_zone)
2058 goto out_destroy_dabuf_zone;
2060 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
2061 if (!xfs_trans_zone)
2062 goto out_destroy_ifork_zone;
2065 * The size of the zone allocated buf log item is the maximum
2066 * size possible under XFS. This wastes a little bit of memory,
2067 * but it is much faster.
2069 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
2070 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
2071 NBWORD) * sizeof(int))), "xfs_buf_item");
2072 if (!xfs_buf_item_zone)
2073 goto out_destroy_trans_zone;
2075 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
2076 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
2077 sizeof(xfs_extent_t))), "xfs_efd_item");
2079 goto out_destroy_buf_item_zone;
2081 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
2082 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
2083 sizeof(xfs_extent_t))), "xfs_efi_item");
2085 goto out_destroy_efd_zone;
2088 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
2089 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
2090 KM_ZONE_SPREAD, NULL);
2091 if (!xfs_inode_zone)
2092 goto out_destroy_efi_zone;
2095 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
2096 KM_ZONE_SPREAD, NULL);
2098 goto out_destroy_inode_zone;
2100 #ifdef CONFIG_XFS_POSIX_ACL
2101 xfs_acl_zone = kmem_zone_init(sizeof(xfs_acl_t), "xfs_acl");
2103 goto out_destroy_ili_zone;
2108 #ifdef CONFIG_XFS_POSIX_ACL
2109 out_destroy_ili_zone:
2111 kmem_zone_destroy(xfs_ili_zone);
2112 out_destroy_inode_zone:
2113 kmem_zone_destroy(xfs_inode_zone);
2114 out_destroy_efi_zone:
2115 kmem_zone_destroy(xfs_efi_zone);
2116 out_destroy_efd_zone:
2117 kmem_zone_destroy(xfs_efd_zone);
2118 out_destroy_buf_item_zone:
2119 kmem_zone_destroy(xfs_buf_item_zone);
2120 out_destroy_trans_zone:
2121 kmem_zone_destroy(xfs_trans_zone);
2122 out_destroy_ifork_zone:
2123 kmem_zone_destroy(xfs_ifork_zone);
2124 out_destroy_dabuf_zone:
2125 kmem_zone_destroy(xfs_dabuf_zone);
2126 out_destroy_da_state_zone:
2127 kmem_zone_destroy(xfs_da_state_zone);
2128 out_destroy_btree_cur_zone:
2129 kmem_zone_destroy(xfs_btree_cur_zone);
2130 out_destroy_bmap_free_item_zone:
2131 kmem_zone_destroy(xfs_bmap_free_item_zone);
2132 out_destroy_log_ticket_zone:
2133 kmem_zone_destroy(xfs_log_ticket_zone);
2134 out_destroy_ioend_pool:
2135 mempool_destroy(xfs_ioend_pool);
2136 out_destroy_ioend_zone:
2137 kmem_zone_destroy(xfs_ioend_zone);
2138 out_destroy_vnode_zone:
2139 kmem_zone_destroy(xfs_vnode_zone);
2145 xfs_destroy_zones(void)
2147 #ifdef CONFIG_XFS_POSIX_ACL
2148 kmem_zone_destroy(xfs_acl_zone);
2150 kmem_zone_destroy(xfs_ili_zone);
2151 kmem_zone_destroy(xfs_inode_zone);
2152 kmem_zone_destroy(xfs_efi_zone);
2153 kmem_zone_destroy(xfs_efd_zone);
2154 kmem_zone_destroy(xfs_buf_item_zone);
2155 kmem_zone_destroy(xfs_trans_zone);
2156 kmem_zone_destroy(xfs_ifork_zone);
2157 kmem_zone_destroy(xfs_dabuf_zone);
2158 kmem_zone_destroy(xfs_da_state_zone);
2159 kmem_zone_destroy(xfs_btree_cur_zone);
2160 kmem_zone_destroy(xfs_bmap_free_item_zone);
2161 kmem_zone_destroy(xfs_log_ticket_zone);
2162 mempool_destroy(xfs_ioend_pool);
2163 kmem_zone_destroy(xfs_ioend_zone);
2164 kmem_zone_destroy(xfs_vnode_zone);
2172 static char message[] __initdata = KERN_INFO \
2173 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
2181 error = xfs_init_zones();
2185 error = xfs_alloc_trace_bufs();
2187 goto out_destroy_zones;
2189 error = xfs_mru_cache_init();
2191 goto out_free_trace_buffers;
2193 error = xfs_filestream_init();
2195 goto out_mru_cache_uninit;
2197 error = xfs_buf_init();
2199 goto out_filestream_uninit;
2201 error = xfs_init_procfs();
2203 goto out_buf_terminate;
2205 error = xfs_sysctl_register();
2207 goto out_cleanup_procfs;
2211 error = register_filesystem(&xfs_fs_type);
2213 goto out_sysctl_unregister;
2216 out_sysctl_unregister:
2217 xfs_sysctl_unregister();
2219 xfs_cleanup_procfs();
2221 xfs_buf_terminate();
2222 out_filestream_uninit:
2223 xfs_filestream_uninit();
2224 out_mru_cache_uninit:
2225 xfs_mru_cache_uninit();
2226 out_free_trace_buffers:
2227 xfs_free_trace_bufs();
2229 xfs_destroy_zones();
2238 unregister_filesystem(&xfs_fs_type);
2239 xfs_sysctl_unregister();
2240 xfs_cleanup_procfs();
2241 xfs_buf_terminate();
2242 xfs_filestream_uninit();
2243 xfs_mru_cache_uninit();
2244 xfs_free_trace_bufs();
2245 xfs_destroy_zones();
2249 module_init(init_xfs_fs);
2250 module_exit(exit_xfs_fs);
2252 MODULE_AUTHOR("Silicon Graphics, Inc.");
2253 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
2254 MODULE_LICENSE("GPL");