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"
55 #include <linux/namei.h>
56 #include <linux/init.h>
57 #include <linux/mount.h>
58 #include <linux/mempool.h>
59 #include <linux/writeback.h>
60 #include <linux/kthread.h>
61 #include <linux/freezer.h>
63 static struct quotactl_ops xfs_quotactl_operations;
64 static struct super_operations xfs_super_operations;
65 static kmem_zone_t *xfs_vnode_zone;
66 static kmem_zone_t *xfs_ioend_zone;
67 mempool_t *xfs_ioend_pool;
69 STATIC struct xfs_mount_args *
71 struct super_block *sb,
74 struct xfs_mount_args *args;
76 args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP);
77 args->logbufs = args->logbufsize = -1;
78 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
80 /* Copy the already-parsed mount(2) flags we're interested in */
81 if (sb->s_flags & MS_DIRSYNC)
82 args->flags |= XFSMNT_DIRSYNC;
83 if (sb->s_flags & MS_SYNCHRONOUS)
84 args->flags |= XFSMNT_WSYNC;
86 args->flags |= XFSMNT_QUIET;
87 args->flags |= XFSMNT_32BITINODES;
92 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
93 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
94 #define MNTOPT_LOGDEV "logdev" /* log device */
95 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
96 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
97 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
98 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
99 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
100 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
101 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
102 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
103 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
104 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
105 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
106 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
107 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
108 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
109 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
110 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
111 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
112 * unwritten extent conversion */
113 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
114 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
115 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
116 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
117 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
118 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
119 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
121 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
122 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
123 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
124 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
125 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
126 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
127 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
128 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
129 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
130 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
131 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
132 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
133 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
134 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
135 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
136 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
137 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
138 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
141 suffix_strtoul(char *s, char **endp, unsigned int base)
143 int last, shift_left_factor = 0;
146 last = strlen(value) - 1;
147 if (value[last] == 'K' || value[last] == 'k') {
148 shift_left_factor = 10;
151 if (value[last] == 'M' || value[last] == 'm') {
152 shift_left_factor = 20;
155 if (value[last] == 'G' || value[last] == 'g') {
156 shift_left_factor = 30;
160 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
165 struct xfs_mount *mp,
167 struct xfs_mount_args *args,
170 char *this_char, *value, *eov;
171 int dsunit, dswidth, vol_dsunit, vol_dswidth;
175 args->flags |= XFSMNT_BARRIER;
176 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
181 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
183 while ((this_char = strsep(&options, ",")) != NULL) {
186 if ((value = strchr(this_char, '=')) != NULL)
189 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
190 if (!value || !*value) {
192 "XFS: %s option requires an argument",
196 args->logbufs = simple_strtoul(value, &eov, 10);
197 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
198 if (!value || !*value) {
200 "XFS: %s option requires an argument",
204 args->logbufsize = suffix_strtoul(value, &eov, 10);
205 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
206 if (!value || !*value) {
208 "XFS: %s option requires an argument",
212 strncpy(args->logname, value, MAXNAMELEN);
213 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
214 if (!value || !*value) {
216 "XFS: %s option requires an argument",
220 strncpy(args->mtpt, value, MAXNAMELEN);
221 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
222 if (!value || !*value) {
224 "XFS: %s option requires an argument",
228 strncpy(args->rtname, value, MAXNAMELEN);
229 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
230 if (!value || !*value) {
232 "XFS: %s option requires an argument",
236 iosize = simple_strtoul(value, &eov, 10);
237 args->flags |= XFSMNT_IOSIZE;
238 args->iosizelog = (uint8_t) iosize;
239 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
240 if (!value || !*value) {
242 "XFS: %s option requires an argument",
246 iosize = suffix_strtoul(value, &eov, 10);
247 args->flags |= XFSMNT_IOSIZE;
248 args->iosizelog = ffs(iosize) - 1;
249 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
250 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
251 mp->m_flags |= XFS_MOUNT_GRPID;
252 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
253 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
254 mp->m_flags &= ~XFS_MOUNT_GRPID;
255 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
256 args->flags |= XFSMNT_WSYNC;
257 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
258 args->flags |= XFSMNT_OSYNCISOSYNC;
259 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
260 args->flags |= XFSMNT_NORECOVERY;
261 } else if (!strcmp(this_char, MNTOPT_INO64)) {
262 args->flags |= XFSMNT_INO64;
265 "XFS: %s option not allowed on this system",
269 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
270 args->flags |= XFSMNT_NOALIGN;
271 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
272 args->flags |= XFSMNT_SWALLOC;
273 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
274 if (!value || !*value) {
276 "XFS: %s option requires an argument",
280 dsunit = simple_strtoul(value, &eov, 10);
281 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
282 if (!value || !*value) {
284 "XFS: %s option requires an argument",
288 dswidth = simple_strtoul(value, &eov, 10);
289 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
290 args->flags &= ~XFSMNT_32BITINODES;
293 "XFS: %s option not allowed on this system",
297 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
298 args->flags |= XFSMNT_NOUUID;
299 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
300 args->flags |= XFSMNT_BARRIER;
301 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
302 args->flags &= ~XFSMNT_BARRIER;
303 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
305 args->flags &= ~XFSMNT_IDELETE;
306 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
307 args->flags |= XFSMNT_IDELETE;
308 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
309 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
310 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
311 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
312 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
313 args->flags |= XFSMNT_ATTR2;
314 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
315 args->flags &= ~XFSMNT_ATTR2;
316 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
317 args->flags2 |= XFSMNT2_FILESTREAMS;
318 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
319 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
320 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
321 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
322 !strcmp(this_char, MNTOPT_UQUOTA) ||
323 !strcmp(this_char, MNTOPT_USRQUOTA)) {
324 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
325 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
326 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
327 args->flags |= XFSMNT_UQUOTA;
328 args->flags &= ~XFSMNT_UQUOTAENF;
329 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
330 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
331 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
332 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
333 args->flags |= XFSMNT_PQUOTA;
334 args->flags &= ~XFSMNT_PQUOTAENF;
335 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
336 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
337 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
338 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
339 args->flags |= XFSMNT_GQUOTA;
340 args->flags &= ~XFSMNT_GQUOTAENF;
341 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
342 args->flags |= XFSMNT_DMAPI;
343 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
344 args->flags |= XFSMNT_DMAPI;
345 } else if (!strcmp(this_char, MNTOPT_DMI)) {
346 args->flags |= XFSMNT_DMAPI;
347 } else if (!strcmp(this_char, "ihashsize")) {
349 "XFS: ihashsize no longer used, option is deprecated.");
350 } else if (!strcmp(this_char, "osyncisdsync")) {
351 /* no-op, this is now the default */
353 "XFS: osyncisdsync is now the default, option is deprecated.");
354 } else if (!strcmp(this_char, "irixsgid")) {
356 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
359 "XFS: unknown mount option [%s].", this_char);
364 if (args->flags & XFSMNT_NORECOVERY) {
365 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
367 "XFS: no-recovery mounts must be read-only.");
372 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
374 "XFS: sunit and swidth options incompatible with the noalign option");
378 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
380 "XFS: cannot mount with both project and group quota");
384 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
385 printk("XFS: %s option needs the mount point option as well\n",
390 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
392 "XFS: sunit and swidth must be specified together");
396 if (dsunit && (dswidth % dsunit != 0)) {
398 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
404 * Applications using DMI filesystems often expect the
405 * inode generation number to be monotonically increasing.
406 * If we delete inode chunks we break this assumption, so
407 * keep unused inode chunks on disk for DMI filesystems
408 * until we come up with a better solution.
409 * Note that if "ikeep" or "noikeep" mount options are
410 * supplied, then they are honored.
412 if (!(args->flags & XFSMNT_DMAPI) && !ikeep)
413 args->flags |= XFSMNT_IDELETE;
415 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
417 args->sunit = dsunit;
418 args->flags |= XFSMNT_RETERR;
420 args->sunit = vol_dsunit;
422 dswidth ? (args->swidth = dswidth) :
423 (args->swidth = vol_dswidth);
425 args->sunit = args->swidth = 0;
429 if (args->flags & XFSMNT_32BITINODES)
430 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
432 args->flags |= XFSMNT_FLAGS2;
436 struct proc_xfs_info {
443 struct xfs_mount *mp,
446 static struct proc_xfs_info xfs_info_set[] = {
447 /* the few simple ones we can get from the mount struct */
448 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
449 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
450 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
451 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
452 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
453 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
454 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
455 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
456 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
457 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
458 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
461 static struct proc_xfs_info xfs_info_unset[] = {
462 /* the few simple ones we can get from the mount struct */
463 { XFS_MOUNT_IDELETE, "," MNTOPT_IKEEP },
464 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
465 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
466 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
469 struct proc_xfs_info *xfs_infop;
471 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
472 if (mp->m_flags & xfs_infop->flag)
473 seq_puts(m, xfs_infop->str);
475 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
476 if (!(mp->m_flags & xfs_infop->flag))
477 seq_puts(m, xfs_infop->str);
480 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
481 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
482 (int)(1 << mp->m_writeio_log) >> 10);
484 if (mp->m_logbufs > 0)
485 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
486 if (mp->m_logbsize > 0)
487 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
490 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
492 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
494 if (mp->m_dalign > 0)
495 seq_printf(m, "," MNTOPT_SUNIT "=%d",
496 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
497 if (mp->m_swidth > 0)
498 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
499 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
501 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
502 seq_puts(m, "," MNTOPT_USRQUOTA);
503 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
504 seq_puts(m, "," MNTOPT_UQUOTANOENF);
506 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
507 seq_puts(m, "," MNTOPT_PRJQUOTA);
508 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
509 seq_puts(m, "," MNTOPT_PQUOTANOENF);
511 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
512 seq_puts(m, "," MNTOPT_GRPQUOTA);
513 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
514 seq_puts(m, "," MNTOPT_GQUOTANOENF);
516 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
517 seq_puts(m, "," MNTOPT_NOQUOTA);
523 unsigned int blockshift)
525 unsigned int pagefactor = 1;
526 unsigned int bitshift = BITS_PER_LONG - 1;
528 /* Figure out maximum filesize, on Linux this can depend on
529 * the filesystem blocksize (on 32 bit platforms).
530 * __block_prepare_write does this in an [unsigned] long...
531 * page->index << (PAGE_CACHE_SHIFT - bbits)
532 * So, for page sized blocks (4K on 32 bit platforms),
533 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
534 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
535 * but for smaller blocksizes it is less (bbits = log2 bsize).
536 * Note1: get_block_t takes a long (implicit cast from above)
537 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
538 * can optionally convert the [unsigned] long from above into
539 * an [unsigned] long long.
542 #if BITS_PER_LONG == 32
543 # if defined(CONFIG_LBD)
544 ASSERT(sizeof(sector_t) == 8);
545 pagefactor = PAGE_CACHE_SIZE;
546 bitshift = BITS_PER_LONG;
548 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
552 return (((__uint64_t)pagefactor) << bitshift) - 1;
559 switch (inode->i_mode & S_IFMT) {
561 inode->i_op = &xfs_inode_operations;
562 inode->i_fop = &xfs_file_operations;
563 inode->i_mapping->a_ops = &xfs_address_space_operations;
566 inode->i_op = &xfs_dir_inode_operations;
567 inode->i_fop = &xfs_dir_file_operations;
570 inode->i_op = &xfs_symlink_inode_operations;
572 inode->i_mapping->a_ops = &xfs_address_space_operations;
575 inode->i_op = &xfs_inode_operations;
576 init_special_inode(inode, inode->i_mode, inode->i_rdev);
582 xfs_revalidate_inode(
587 struct inode *inode = vn_to_inode(vp);
589 inode->i_mode = ip->i_d.di_mode;
590 inode->i_nlink = ip->i_d.di_nlink;
591 inode->i_uid = ip->i_d.di_uid;
592 inode->i_gid = ip->i_d.di_gid;
594 switch (inode->i_mode & S_IFMT) {
598 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
599 sysv_minor(ip->i_df.if_u2.if_rdev));
606 inode->i_generation = ip->i_d.di_gen;
607 i_size_write(inode, ip->i_d.di_size);
609 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
610 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
611 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
612 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
613 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
614 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
615 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
616 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
617 inode->i_flags |= S_IMMUTABLE;
619 inode->i_flags &= ~S_IMMUTABLE;
620 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
621 inode->i_flags |= S_APPEND;
623 inode->i_flags &= ~S_APPEND;
624 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
625 inode->i_flags |= S_SYNC;
627 inode->i_flags &= ~S_SYNC;
628 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
629 inode->i_flags |= S_NOATIME;
631 inode->i_flags &= ~S_NOATIME;
632 xfs_iflags_clear(ip, XFS_IMODIFIED);
636 xfs_initialize_vnode(
637 struct xfs_mount *mp,
639 struct xfs_inode *ip)
641 struct inode *inode = vn_to_inode(vp);
645 inode->i_private = ip;
649 * We need to set the ops vectors, and unlock the inode, but if
650 * we have been called during the new inode create process, it is
651 * too early to fill in the Linux inode. We will get called a
652 * second time once the inode is properly set up, and then we can
655 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
656 xfs_revalidate_inode(mp, vp, ip);
657 xfs_set_inodeops(inode);
659 xfs_iflags_clear(ip, XFS_INEW);
662 unlock_new_inode(inode);
670 struct block_device **bdevp)
674 *bdevp = open_bdev_excl(name, 0, mp);
675 if (IS_ERR(*bdevp)) {
676 error = PTR_ERR(*bdevp);
677 printk("XFS: Invalid device [%s], error=%d\n", name, error);
685 struct block_device *bdev)
688 close_bdev_excl(bdev);
692 * Try to write out the superblock using barriers.
698 xfs_buf_t *sbp = xfs_getsb(mp, 0);
703 XFS_BUF_UNDELAYWRITE(sbp);
705 XFS_BUF_UNASYNC(sbp);
706 XFS_BUF_ORDERED(sbp);
709 error = xfs_iowait(sbp);
712 * Clear all the flags we set and possible error state in the
713 * buffer. We only did the write to try out whether barriers
714 * worked and shouldn't leave any traces in the superblock
718 XFS_BUF_ERROR(sbp, 0);
719 XFS_BUF_UNORDERED(sbp);
726 xfs_mountfs_check_barriers(xfs_mount_t *mp)
730 if (mp->m_logdev_targp != mp->m_ddev_targp) {
731 xfs_fs_cmn_err(CE_NOTE, mp,
732 "Disabling barriers, not supported with external log device");
733 mp->m_flags &= ~XFS_MOUNT_BARRIER;
737 if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered ==
738 QUEUE_ORDERED_NONE) {
739 xfs_fs_cmn_err(CE_NOTE, mp,
740 "Disabling barriers, not supported by the underlying device");
741 mp->m_flags &= ~XFS_MOUNT_BARRIER;
745 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
746 xfs_fs_cmn_err(CE_NOTE, mp,
747 "Disabling barriers, underlying device is readonly");
748 mp->m_flags &= ~XFS_MOUNT_BARRIER;
752 error = xfs_barrier_test(mp);
754 xfs_fs_cmn_err(CE_NOTE, mp,
755 "Disabling barriers, trial barrier write failed");
756 mp->m_flags &= ~XFS_MOUNT_BARRIER;
762 xfs_blkdev_issue_flush(
763 xfs_buftarg_t *buftarg)
765 blkdev_issue_flush(buftarg->bt_bdev, NULL);
768 STATIC struct inode *
770 struct super_block *sb)
774 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
777 return vn_to_inode(vp);
781 xfs_fs_destroy_inode(
784 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
788 xfs_fs_inode_init_once(
792 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
798 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
799 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
801 xfs_fs_inode_init_once);
805 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
807 goto out_destroy_vnode_zone;
809 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
812 goto out_free_ioend_zone;
816 kmem_zone_destroy(xfs_ioend_zone);
817 out_destroy_vnode_zone:
818 kmem_zone_destroy(xfs_vnode_zone);
824 xfs_destroy_zones(void)
826 mempool_destroy(xfs_ioend_pool);
827 kmem_zone_destroy(xfs_vnode_zone);
828 kmem_zone_destroy(xfs_ioend_zone);
832 * Attempt to flush the inode, this will actually fail
833 * if the inode is pinned, but we dirty the inode again
834 * at the point when it is unpinned after a log write,
835 * since this is when the inode itself becomes flushable.
842 int error = 0, flags = FLUSH_INODE;
844 xfs_itrace_entry(XFS_I(inode));
846 filemap_fdatawait(inode->i_mapping);
849 error = xfs_inode_flush(XFS_I(inode), flags);
851 * if we failed to write out the inode then mark
852 * it dirty again so we'll try again later.
855 mark_inode_dirty_sync(inode);
864 xfs_inode_t *ip = XFS_I(inode);
867 * ip can be null when xfs_iget_core calls xfs_idestroy if we
868 * find an inode with di_mode == 0 but without IGET_CREATE set.
871 xfs_itrace_entry(ip);
872 XFS_STATS_INC(vn_rele);
873 XFS_STATS_INC(vn_remove);
874 XFS_STATS_INC(vn_reclaim);
875 XFS_STATS_DEC(vn_active);
878 xfs_iflags_clear(ip, XFS_IMODIFIED);
880 panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, inode);
883 ASSERT(XFS_I(inode) == NULL);
887 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
888 * Doing this has two advantages:
889 * - It saves on stack space, which is tight in certain situations
890 * - It can be used (with care) as a mechanism to avoid deadlocks.
891 * Flushing while allocating in a full filesystem requires both.
894 xfs_syncd_queue_work(
895 struct xfs_mount *mp,
897 void (*syncer)(struct xfs_mount *, void *))
899 struct bhv_vfs_sync_work *work;
901 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
902 INIT_LIST_HEAD(&work->w_list);
903 work->w_syncer = syncer;
906 spin_lock(&mp->m_sync_lock);
907 list_add_tail(&work->w_list, &mp->m_sync_list);
908 spin_unlock(&mp->m_sync_lock);
909 wake_up_process(mp->m_sync_task);
913 * Flush delayed allocate data, attempting to free up reserved space
914 * from existing allocations. At this point a new allocation attempt
915 * has failed with ENOSPC and we are in the process of scratching our
916 * heads, looking about for more room...
919 xfs_flush_inode_work(
920 struct xfs_mount *mp,
923 struct inode *inode = arg;
924 filemap_flush(inode->i_mapping);
932 struct inode *inode = ip->i_vnode;
935 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
936 delay(msecs_to_jiffies(500));
940 * This is the "bigger hammer" version of xfs_flush_inode_work...
941 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
944 xfs_flush_device_work(
945 struct xfs_mount *mp,
948 struct inode *inode = arg;
949 sync_blockdev(mp->m_super->s_bdev);
957 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
960 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
961 delay(msecs_to_jiffies(500));
962 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
967 struct xfs_mount *mp,
972 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
973 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR |
974 SYNC_REFCACHE | SYNC_SUPER);
976 wake_up(&mp->m_wait_single_sync_task);
983 struct xfs_mount *mp = arg;
985 bhv_vfs_sync_work_t *work, *n;
989 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
991 timeleft = schedule_timeout_interruptible(timeleft);
994 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
997 spin_lock(&mp->m_sync_lock);
999 * We can get woken by laptop mode, to do a sync -
1000 * that's the (only!) case where the list would be
1001 * empty with time remaining.
1003 if (!timeleft || list_empty(&mp->m_sync_list)) {
1005 timeleft = xfs_syncd_centisecs *
1006 msecs_to_jiffies(10);
1007 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1008 list_add_tail(&mp->m_sync_work.w_list,
1011 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1012 list_move(&work->w_list, &tmp);
1013 spin_unlock(&mp->m_sync_lock);
1015 list_for_each_entry_safe(work, n, &tmp, w_list) {
1016 (*work->w_syncer)(mp, work->w_data);
1017 list_del(&work->w_list);
1018 if (work == &mp->m_sync_work)
1020 kmem_free(work, sizeof(struct bhv_vfs_sync_work));
1029 struct super_block *sb)
1031 struct xfs_mount *mp = XFS_M(sb);
1034 kthread_stop(mp->m_sync_task);
1036 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1037 error = xfs_unmount(mp, 0, NULL);
1039 printk("XFS: unmount got error=%d\n", error);
1044 struct super_block *sb)
1046 if (!(sb->s_flags & MS_RDONLY))
1047 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1053 struct super_block *sb,
1056 struct xfs_mount *mp = XFS_M(sb);
1061 * Treat a sync operation like a freeze. This is to work
1062 * around a race in sync_inodes() which works in two phases
1063 * - an asynchronous flush, which can write out an inode
1064 * without waiting for file size updates to complete, and a
1065 * synchronous flush, which wont do anything because the
1066 * async flush removed the inode's dirty flag. Also
1067 * sync_inodes() will not see any files that just have
1068 * outstanding transactions to be flushed because we don't
1069 * dirty the Linux inode until after the transaction I/O
1072 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1074 * First stage of freeze - no more writers will make progress
1075 * now we are here, so we flush delwri and delalloc buffers
1076 * here, then wait for all I/O to complete. Data is frozen at
1077 * that point. Metadata is not frozen, transactions can still
1078 * occur here so don't bother flushing the buftarg (i.e
1079 * SYNC_QUIESCE) because it'll just get dirty again.
1081 flags = SYNC_DATA_QUIESCE;
1083 flags = SYNC_FSDATA;
1085 error = xfs_sync(mp, flags);
1088 if (unlikely(laptop_mode)) {
1089 int prev_sync_seq = mp->m_sync_seq;
1092 * The disk must be active because we're syncing.
1093 * We schedule xfssyncd now (now that the disk is
1094 * active) instead of later (when it might not be).
1096 wake_up_process(mp->m_sync_task);
1098 * We have to wait for the sync iteration to complete.
1099 * If we don't, the disk activity caused by the sync
1100 * will come after the sync is completed, and that
1101 * triggers another sync from laptop mode.
1103 wait_event(mp->m_wait_single_sync_task,
1104 mp->m_sync_seq != prev_sync_seq);
1112 struct dentry *dentry,
1113 struct kstatfs *statp)
1115 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1116 xfs_sb_t *sbp = &mp->m_sb;
1117 __uint64_t fakeinos, id;
1120 statp->f_type = XFS_SB_MAGIC;
1121 statp->f_namelen = MAXNAMELEN - 1;
1123 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1124 statp->f_fsid.val[0] = (u32)id;
1125 statp->f_fsid.val[1] = (u32)(id >> 32);
1127 xfs_icsb_sync_counters_flags(mp, XFS_ICSB_LAZY_COUNT);
1129 spin_lock(&mp->m_sb_lock);
1130 statp->f_bsize = sbp->sb_blocksize;
1131 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1132 statp->f_blocks = sbp->sb_dblocks - lsize;
1133 statp->f_bfree = statp->f_bavail =
1134 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1135 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1137 fakeinos += mp->m_inoadd;
1140 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1141 if (mp->m_maxicount)
1145 statp->f_files = min_t(typeof(statp->f_files),
1148 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1149 spin_unlock(&mp->m_sb_lock);
1151 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1157 struct super_block *sb,
1161 struct xfs_mount *mp = XFS_M(sb);
1162 struct xfs_mount_args *args = xfs_args_allocate(sb, 0);
1165 error = xfs_parseargs(mp, options, args, 1);
1167 error = xfs_mntupdate(mp, flags, args);
1168 kmem_free(args, sizeof(*args));
1173 * Second stage of a freeze. The data is already frozen so we only
1174 * need to take care of themetadata. Once that's done write a dummy
1175 * record to dirty the log in case of a crash while frozen.
1179 struct super_block *sb)
1181 struct xfs_mount *mp = XFS_M(sb);
1183 xfs_attr_quiesce(mp);
1184 xfs_fs_log_dummy(mp);
1188 xfs_fs_show_options(
1190 struct vfsmount *mnt)
1192 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1197 struct super_block *sb,
1200 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1205 struct super_block *sb,
1206 struct fs_quota_stat *fqs)
1208 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1213 struct super_block *sb,
1217 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1222 struct super_block *sb,
1225 struct fs_disk_quota *fdq)
1227 return -XFS_QM_QUOTACTL(XFS_M(sb),
1228 (type == USRQUOTA) ? Q_XGETQUOTA :
1229 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1230 Q_XGETPQUOTA), id, (caddr_t)fdq);
1235 struct super_block *sb,
1238 struct fs_disk_quota *fdq)
1240 return -XFS_QM_QUOTACTL(XFS_M(sb),
1241 (type == USRQUOTA) ? Q_XSETQLIM :
1242 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1243 Q_XSETPQLIM), id, (caddr_t)fdq);
1248 struct super_block *sb,
1252 struct inode *rootvp;
1253 struct xfs_mount *mp = NULL;
1254 struct xfs_mount_args *args = xfs_args_allocate(sb, silent);
1257 mp = xfs_mount_init();
1259 INIT_LIST_HEAD(&mp->m_sync_list);
1260 spin_lock_init(&mp->m_sync_lock);
1261 init_waitqueue_head(&mp->m_wait_single_sync_task);
1266 if (sb->s_flags & MS_RDONLY)
1267 mp->m_flags |= XFS_MOUNT_RDONLY;
1269 error = xfs_parseargs(mp, (char *)data, args, 0);
1273 sb_min_blocksize(sb, BBSIZE);
1274 sb->s_export_op = &xfs_export_operations;
1275 sb->s_qcop = &xfs_quotactl_operations;
1276 sb->s_op = &xfs_super_operations;
1278 error = xfs_mount(mp, args, NULL);
1283 sb->s_magic = XFS_SB_MAGIC;
1284 sb->s_blocksize = mp->m_sb.sb_blocksize;
1285 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1286 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1287 sb->s_time_gran = 1;
1288 set_posix_acl_flag(sb);
1290 error = xfs_root(mp, &rootvp);
1294 sb->s_root = d_alloc_root(vn_to_inode(rootvp));
1299 if (is_bad_inode(sb->s_root->d_inode)) {
1304 mp->m_sync_work.w_syncer = xfs_sync_worker;
1305 mp->m_sync_work.w_mount = mp;
1306 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1307 if (IS_ERR(mp->m_sync_task)) {
1308 error = -PTR_ERR(mp->m_sync_task);
1312 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1314 kmem_free(args, sizeof(*args));
1326 xfs_unmount(mp, 0, NULL);
1329 kmem_free(args, sizeof(*args));
1335 struct file_system_type *fs_type,
1337 const char *dev_name,
1339 struct vfsmount *mnt)
1341 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1345 static struct super_operations xfs_super_operations = {
1346 .alloc_inode = xfs_fs_alloc_inode,
1347 .destroy_inode = xfs_fs_destroy_inode,
1348 .write_inode = xfs_fs_write_inode,
1349 .clear_inode = xfs_fs_clear_inode,
1350 .put_super = xfs_fs_put_super,
1351 .write_super = xfs_fs_write_super,
1352 .sync_fs = xfs_fs_sync_super,
1353 .write_super_lockfs = xfs_fs_lockfs,
1354 .statfs = xfs_fs_statfs,
1355 .remount_fs = xfs_fs_remount,
1356 .show_options = xfs_fs_show_options,
1359 static struct quotactl_ops xfs_quotactl_operations = {
1360 .quota_sync = xfs_fs_quotasync,
1361 .get_xstate = xfs_fs_getxstate,
1362 .set_xstate = xfs_fs_setxstate,
1363 .get_xquota = xfs_fs_getxquota,
1364 .set_xquota = xfs_fs_setxquota,
1367 static struct file_system_type xfs_fs_type = {
1368 .owner = THIS_MODULE,
1370 .get_sb = xfs_fs_get_sb,
1371 .kill_sb = kill_block_super,
1372 .fs_flags = FS_REQUIRES_DEV,
1380 static char message[] __initdata = KERN_INFO \
1381 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
1387 error = xfs_init_zones();
1391 error = xfs_buf_init();
1400 error = register_filesystem(&xfs_fs_type);
1406 xfs_buf_terminate();
1409 xfs_destroy_zones();
1419 unregister_filesystem(&xfs_fs_type);
1421 xfs_buf_terminate();
1422 xfs_destroy_zones();
1426 module_init(init_xfs_fs);
1427 module_exit(exit_xfs_fs);
1429 MODULE_AUTHOR("Silicon Graphics, Inc.");
1430 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1431 MODULE_LICENSE("GPL");