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"
56 #include <linux/namei.h>
57 #include <linux/init.h>
58 #include <linux/mount.h>
59 #include <linux/mempool.h>
60 #include <linux/writeback.h>
61 #include <linux/kthread.h>
62 #include <linux/freezer.h>
64 static struct quotactl_ops xfs_quotactl_operations;
65 static struct super_operations xfs_super_operations;
66 static kmem_zone_t *xfs_vnode_zone;
67 static kmem_zone_t *xfs_ioend_zone;
68 mempool_t *xfs_ioend_pool;
70 STATIC struct xfs_mount_args *
72 struct super_block *sb,
75 struct xfs_mount_args *args;
77 args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP);
78 args->logbufs = args->logbufsize = -1;
79 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
81 /* Copy the already-parsed mount(2) flags we're interested in */
82 if (sb->s_flags & MS_DIRSYNC)
83 args->flags |= XFSMNT_DIRSYNC;
84 if (sb->s_flags & MS_SYNCHRONOUS)
85 args->flags |= XFSMNT_WSYNC;
87 args->flags |= XFSMNT_QUIET;
88 args->flags |= XFSMNT_32BITINODES;
93 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
94 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
95 #define MNTOPT_LOGDEV "logdev" /* log device */
96 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
97 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
98 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
99 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
100 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
101 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
102 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
103 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
104 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
105 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
106 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
107 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
108 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
109 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
110 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
111 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
112 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
113 * unwritten extent conversion */
114 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
115 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
116 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
117 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
118 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
119 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
120 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
122 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
123 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
124 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
125 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
126 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
127 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
128 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
129 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
130 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
131 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
132 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
133 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
134 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
135 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
136 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
137 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
138 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
139 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
142 suffix_strtoul(char *s, char **endp, unsigned int base)
144 int last, shift_left_factor = 0;
147 last = strlen(value) - 1;
148 if (value[last] == 'K' || value[last] == 'k') {
149 shift_left_factor = 10;
152 if (value[last] == 'M' || value[last] == 'm') {
153 shift_left_factor = 20;
156 if (value[last] == 'G' || value[last] == 'g') {
157 shift_left_factor = 30;
161 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
166 struct xfs_mount *mp,
168 struct xfs_mount_args *args,
171 char *this_char, *value, *eov;
172 int dsunit, dswidth, vol_dsunit, vol_dswidth;
176 args->flags |= XFSMNT_BARRIER;
177 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
182 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
184 while ((this_char = strsep(&options, ",")) != NULL) {
187 if ((value = strchr(this_char, '=')) != NULL)
190 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
191 if (!value || !*value) {
193 "XFS: %s option requires an argument",
197 args->logbufs = simple_strtoul(value, &eov, 10);
198 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
199 if (!value || !*value) {
201 "XFS: %s option requires an argument",
205 args->logbufsize = suffix_strtoul(value, &eov, 10);
206 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
207 if (!value || !*value) {
209 "XFS: %s option requires an argument",
213 strncpy(args->logname, value, MAXNAMELEN);
214 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
215 if (!value || !*value) {
217 "XFS: %s option requires an argument",
221 strncpy(args->mtpt, value, MAXNAMELEN);
222 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
223 if (!value || !*value) {
225 "XFS: %s option requires an argument",
229 strncpy(args->rtname, value, MAXNAMELEN);
230 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
231 if (!value || !*value) {
233 "XFS: %s option requires an argument",
237 iosize = simple_strtoul(value, &eov, 10);
238 args->flags |= XFSMNT_IOSIZE;
239 args->iosizelog = (uint8_t) iosize;
240 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
241 if (!value || !*value) {
243 "XFS: %s option requires an argument",
247 iosize = suffix_strtoul(value, &eov, 10);
248 args->flags |= XFSMNT_IOSIZE;
249 args->iosizelog = ffs(iosize) - 1;
250 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
251 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
252 mp->m_flags |= XFS_MOUNT_GRPID;
253 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
254 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
255 mp->m_flags &= ~XFS_MOUNT_GRPID;
256 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
257 args->flags |= XFSMNT_WSYNC;
258 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
259 args->flags |= XFSMNT_OSYNCISOSYNC;
260 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
261 args->flags |= XFSMNT_NORECOVERY;
262 } else if (!strcmp(this_char, MNTOPT_INO64)) {
263 args->flags |= XFSMNT_INO64;
266 "XFS: %s option not allowed on this system",
270 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
271 args->flags |= XFSMNT_NOALIGN;
272 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
273 args->flags |= XFSMNT_SWALLOC;
274 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
275 if (!value || !*value) {
277 "XFS: %s option requires an argument",
281 dsunit = simple_strtoul(value, &eov, 10);
282 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
283 if (!value || !*value) {
285 "XFS: %s option requires an argument",
289 dswidth = simple_strtoul(value, &eov, 10);
290 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
291 args->flags &= ~XFSMNT_32BITINODES;
294 "XFS: %s option not allowed on this system",
298 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
299 args->flags |= XFSMNT_NOUUID;
300 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
301 args->flags |= XFSMNT_BARRIER;
302 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
303 args->flags &= ~XFSMNT_BARRIER;
304 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
306 args->flags &= ~XFSMNT_IDELETE;
307 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
308 args->flags |= XFSMNT_IDELETE;
309 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
310 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
311 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
312 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
313 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
314 args->flags |= XFSMNT_ATTR2;
315 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
316 args->flags &= ~XFSMNT_ATTR2;
317 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
318 args->flags2 |= XFSMNT2_FILESTREAMS;
319 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
320 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
321 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
322 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
323 !strcmp(this_char, MNTOPT_UQUOTA) ||
324 !strcmp(this_char, MNTOPT_USRQUOTA)) {
325 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
326 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
327 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
328 args->flags |= XFSMNT_UQUOTA;
329 args->flags &= ~XFSMNT_UQUOTAENF;
330 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
331 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
332 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
333 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
334 args->flags |= XFSMNT_PQUOTA;
335 args->flags &= ~XFSMNT_PQUOTAENF;
336 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
337 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
338 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
339 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
340 args->flags |= XFSMNT_GQUOTA;
341 args->flags &= ~XFSMNT_GQUOTAENF;
342 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
343 args->flags |= XFSMNT_DMAPI;
344 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
345 args->flags |= XFSMNT_DMAPI;
346 } else if (!strcmp(this_char, MNTOPT_DMI)) {
347 args->flags |= XFSMNT_DMAPI;
348 } else if (!strcmp(this_char, "ihashsize")) {
350 "XFS: ihashsize no longer used, option is deprecated.");
351 } else if (!strcmp(this_char, "osyncisdsync")) {
352 /* no-op, this is now the default */
354 "XFS: osyncisdsync is now the default, option is deprecated.");
355 } else if (!strcmp(this_char, "irixsgid")) {
357 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
360 "XFS: unknown mount option [%s].", this_char);
365 if (args->flags & XFSMNT_NORECOVERY) {
366 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
368 "XFS: no-recovery mounts must be read-only.");
373 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
375 "XFS: sunit and swidth options incompatible with the noalign option");
379 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
381 "XFS: cannot mount with both project and group quota");
385 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
386 printk("XFS: %s option needs the mount point option as well\n",
391 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
393 "XFS: sunit and swidth must be specified together");
397 if (dsunit && (dswidth % dsunit != 0)) {
399 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
405 * Applications using DMI filesystems often expect the
406 * inode generation number to be monotonically increasing.
407 * If we delete inode chunks we break this assumption, so
408 * keep unused inode chunks on disk for DMI filesystems
409 * until we come up with a better solution.
410 * Note that if "ikeep" or "noikeep" mount options are
411 * supplied, then they are honored.
413 if (!(args->flags & XFSMNT_DMAPI) && !ikeep)
414 args->flags |= XFSMNT_IDELETE;
416 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
418 args->sunit = dsunit;
419 args->flags |= XFSMNT_RETERR;
421 args->sunit = vol_dsunit;
423 dswidth ? (args->swidth = dswidth) :
424 (args->swidth = vol_dswidth);
426 args->sunit = args->swidth = 0;
430 if (args->flags & XFSMNT_32BITINODES)
431 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
433 args->flags |= XFSMNT_FLAGS2;
437 struct proc_xfs_info {
444 struct xfs_mount *mp,
447 static struct proc_xfs_info xfs_info_set[] = {
448 /* the few simple ones we can get from the mount struct */
449 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
450 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
451 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
452 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
453 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
454 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
455 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
456 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
457 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
458 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
459 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
462 static struct proc_xfs_info xfs_info_unset[] = {
463 /* the few simple ones we can get from the mount struct */
464 { XFS_MOUNT_IDELETE, "," MNTOPT_IKEEP },
465 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
466 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
467 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
470 struct proc_xfs_info *xfs_infop;
472 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
473 if (mp->m_flags & xfs_infop->flag)
474 seq_puts(m, xfs_infop->str);
476 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
477 if (!(mp->m_flags & xfs_infop->flag))
478 seq_puts(m, xfs_infop->str);
481 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
482 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
483 (int)(1 << mp->m_writeio_log) >> 10);
485 if (mp->m_logbufs > 0)
486 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
487 if (mp->m_logbsize > 0)
488 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
491 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
493 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
495 if (mp->m_dalign > 0)
496 seq_printf(m, "," MNTOPT_SUNIT "=%d",
497 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
498 if (mp->m_swidth > 0)
499 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
500 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
502 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
503 seq_puts(m, "," MNTOPT_USRQUOTA);
504 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
505 seq_puts(m, "," MNTOPT_UQUOTANOENF);
507 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
508 seq_puts(m, "," MNTOPT_PRJQUOTA);
509 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
510 seq_puts(m, "," MNTOPT_PQUOTANOENF);
512 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
513 seq_puts(m, "," MNTOPT_GRPQUOTA);
514 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
515 seq_puts(m, "," MNTOPT_GQUOTANOENF);
517 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
518 seq_puts(m, "," MNTOPT_NOQUOTA);
524 unsigned int blockshift)
526 unsigned int pagefactor = 1;
527 unsigned int bitshift = BITS_PER_LONG - 1;
529 /* Figure out maximum filesize, on Linux this can depend on
530 * the filesystem blocksize (on 32 bit platforms).
531 * __block_prepare_write does this in an [unsigned] long...
532 * page->index << (PAGE_CACHE_SHIFT - bbits)
533 * So, for page sized blocks (4K on 32 bit platforms),
534 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
535 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
536 * but for smaller blocksizes it is less (bbits = log2 bsize).
537 * Note1: get_block_t takes a long (implicit cast from above)
538 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
539 * can optionally convert the [unsigned] long from above into
540 * an [unsigned] long long.
543 #if BITS_PER_LONG == 32
544 # if defined(CONFIG_LBD)
545 ASSERT(sizeof(sector_t) == 8);
546 pagefactor = PAGE_CACHE_SIZE;
547 bitshift = BITS_PER_LONG;
549 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
553 return (((__uint64_t)pagefactor) << bitshift) - 1;
560 switch (inode->i_mode & S_IFMT) {
562 inode->i_op = &xfs_inode_operations;
563 inode->i_fop = &xfs_file_operations;
564 inode->i_mapping->a_ops = &xfs_address_space_operations;
567 inode->i_op = &xfs_dir_inode_operations;
568 inode->i_fop = &xfs_dir_file_operations;
571 inode->i_op = &xfs_symlink_inode_operations;
572 if (!(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE))
573 inode->i_mapping->a_ops = &xfs_address_space_operations;
576 inode->i_op = &xfs_inode_operations;
577 init_special_inode(inode, inode->i_mode, inode->i_rdev);
583 xfs_revalidate_inode(
588 struct inode *inode = vn_to_inode(vp);
590 inode->i_mode = ip->i_d.di_mode;
591 inode->i_nlink = ip->i_d.di_nlink;
592 inode->i_uid = ip->i_d.di_uid;
593 inode->i_gid = ip->i_d.di_gid;
595 switch (inode->i_mode & S_IFMT) {
599 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
600 sysv_minor(ip->i_df.if_u2.if_rdev));
607 inode->i_generation = ip->i_d.di_gen;
608 i_size_write(inode, ip->i_d.di_size);
609 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
610 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
611 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
612 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
613 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
614 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
615 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
616 inode->i_flags |= S_IMMUTABLE;
618 inode->i_flags &= ~S_IMMUTABLE;
619 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
620 inode->i_flags |= S_APPEND;
622 inode->i_flags &= ~S_APPEND;
623 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
624 inode->i_flags |= S_SYNC;
626 inode->i_flags &= ~S_SYNC;
627 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
628 inode->i_flags |= S_NOATIME;
630 inode->i_flags &= ~S_NOATIME;
631 xfs_iflags_clear(ip, XFS_IMODIFIED);
635 xfs_initialize_vnode(
636 struct xfs_mount *mp,
638 struct xfs_inode *ip)
640 struct inode *inode = vn_to_inode(vp);
644 inode->i_private = ip;
648 * We need to set the ops vectors, and unlock the inode, but if
649 * we have been called during the new inode create process, it is
650 * too early to fill in the Linux inode. We will get called a
651 * second time once the inode is properly set up, and then we can
654 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
655 xfs_revalidate_inode(mp, vp, ip);
656 xfs_set_inodeops(inode);
658 xfs_iflags_clear(ip, XFS_INEW);
661 unlock_new_inode(inode);
669 struct block_device **bdevp)
673 *bdevp = open_bdev_excl(name, 0, mp);
674 if (IS_ERR(*bdevp)) {
675 error = PTR_ERR(*bdevp);
676 printk("XFS: Invalid device [%s], error=%d\n", name, error);
684 struct block_device *bdev)
687 close_bdev_excl(bdev);
691 * Try to write out the superblock using barriers.
697 xfs_buf_t *sbp = xfs_getsb(mp, 0);
702 XFS_BUF_UNDELAYWRITE(sbp);
704 XFS_BUF_UNASYNC(sbp);
705 XFS_BUF_ORDERED(sbp);
708 error = xfs_iowait(sbp);
711 * Clear all the flags we set and possible error state in the
712 * buffer. We only did the write to try out whether barriers
713 * worked and shouldn't leave any traces in the superblock
717 XFS_BUF_ERROR(sbp, 0);
718 XFS_BUF_UNORDERED(sbp);
725 xfs_mountfs_check_barriers(xfs_mount_t *mp)
729 if (mp->m_logdev_targp != mp->m_ddev_targp) {
730 xfs_fs_cmn_err(CE_NOTE, mp,
731 "Disabling barriers, not supported with external log device");
732 mp->m_flags &= ~XFS_MOUNT_BARRIER;
736 if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered ==
737 QUEUE_ORDERED_NONE) {
738 xfs_fs_cmn_err(CE_NOTE, mp,
739 "Disabling barriers, not supported by the underlying device");
740 mp->m_flags &= ~XFS_MOUNT_BARRIER;
744 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
745 xfs_fs_cmn_err(CE_NOTE, mp,
746 "Disabling barriers, underlying device is readonly");
747 mp->m_flags &= ~XFS_MOUNT_BARRIER;
751 error = xfs_barrier_test(mp);
753 xfs_fs_cmn_err(CE_NOTE, mp,
754 "Disabling barriers, trial barrier write failed");
755 mp->m_flags &= ~XFS_MOUNT_BARRIER;
761 xfs_blkdev_issue_flush(
762 xfs_buftarg_t *buftarg)
764 blkdev_issue_flush(buftarg->bt_bdev, NULL);
768 * XFS AIL push thread support
773 xfs_lsn_t threshold_lsn)
775 mp->m_ail.xa_target = threshold_lsn;
776 wake_up_process(mp->m_ail.xa_task);
783 xfs_mount_t *mp = (xfs_mount_t *)data;
784 xfs_lsn_t last_pushed_lsn = 0;
787 while (!kthread_should_stop()) {
789 schedule_timeout_interruptible(msecs_to_jiffies(tout));
796 if (XFS_FORCED_SHUTDOWN(mp))
799 tout = xfsaild_push(mp, &last_pushed_lsn);
809 mp->m_ail.xa_target = 0;
810 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
811 if (IS_ERR(mp->m_ail.xa_task))
812 return -PTR_ERR(mp->m_ail.xa_task);
820 kthread_stop(mp->m_ail.xa_task);
825 STATIC struct inode *
827 struct super_block *sb)
831 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
834 return vn_to_inode(vp);
838 xfs_fs_destroy_inode(
841 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
845 xfs_fs_inode_init_once(
849 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
855 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
856 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
858 xfs_fs_inode_init_once);
862 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
864 goto out_destroy_vnode_zone;
866 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
869 goto out_free_ioend_zone;
873 kmem_zone_destroy(xfs_ioend_zone);
874 out_destroy_vnode_zone:
875 kmem_zone_destroy(xfs_vnode_zone);
881 xfs_destroy_zones(void)
883 mempool_destroy(xfs_ioend_pool);
884 kmem_zone_destroy(xfs_vnode_zone);
885 kmem_zone_destroy(xfs_ioend_zone);
889 * Attempt to flush the inode, this will actually fail
890 * if the inode is pinned, but we dirty the inode again
891 * at the point when it is unpinned after a log write,
892 * since this is when the inode itself becomes flushable.
899 int error = 0, flags = FLUSH_INODE;
901 xfs_itrace_entry(XFS_I(inode));
903 filemap_fdatawait(inode->i_mapping);
906 error = xfs_inode_flush(XFS_I(inode), flags);
908 * if we failed to write out the inode then mark
909 * it dirty again so we'll try again later.
912 mark_inode_dirty_sync(inode);
921 xfs_inode_t *ip = XFS_I(inode);
924 * ip can be null when xfs_iget_core calls xfs_idestroy if we
925 * find an inode with di_mode == 0 but without IGET_CREATE set.
928 xfs_itrace_entry(ip);
929 XFS_STATS_INC(vn_rele);
930 XFS_STATS_INC(vn_remove);
931 XFS_STATS_INC(vn_reclaim);
932 XFS_STATS_DEC(vn_active);
935 xfs_iflags_clear(ip, XFS_IMODIFIED);
937 panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, inode);
940 ASSERT(XFS_I(inode) == NULL);
944 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
945 * Doing this has two advantages:
946 * - It saves on stack space, which is tight in certain situations
947 * - It can be used (with care) as a mechanism to avoid deadlocks.
948 * Flushing while allocating in a full filesystem requires both.
951 xfs_syncd_queue_work(
952 struct xfs_mount *mp,
954 void (*syncer)(struct xfs_mount *, void *))
956 struct bhv_vfs_sync_work *work;
958 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
959 INIT_LIST_HEAD(&work->w_list);
960 work->w_syncer = syncer;
963 spin_lock(&mp->m_sync_lock);
964 list_add_tail(&work->w_list, &mp->m_sync_list);
965 spin_unlock(&mp->m_sync_lock);
966 wake_up_process(mp->m_sync_task);
970 * Flush delayed allocate data, attempting to free up reserved space
971 * from existing allocations. At this point a new allocation attempt
972 * has failed with ENOSPC and we are in the process of scratching our
973 * heads, looking about for more room...
976 xfs_flush_inode_work(
977 struct xfs_mount *mp,
980 struct inode *inode = arg;
981 filemap_flush(inode->i_mapping);
989 struct inode *inode = ip->i_vnode;
992 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
993 delay(msecs_to_jiffies(500));
997 * This is the "bigger hammer" version of xfs_flush_inode_work...
998 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
1001 xfs_flush_device_work(
1002 struct xfs_mount *mp,
1005 struct inode *inode = arg;
1006 sync_blockdev(mp->m_super->s_bdev);
1014 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
1017 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1018 delay(msecs_to_jiffies(500));
1019 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1024 struct xfs_mount *mp,
1029 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1030 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR |
1031 SYNC_REFCACHE | SYNC_SUPER);
1033 wake_up(&mp->m_wait_single_sync_task);
1040 struct xfs_mount *mp = arg;
1042 bhv_vfs_sync_work_t *work, *n;
1046 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1048 timeleft = schedule_timeout_interruptible(timeleft);
1051 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1054 spin_lock(&mp->m_sync_lock);
1056 * We can get woken by laptop mode, to do a sync -
1057 * that's the (only!) case where the list would be
1058 * empty with time remaining.
1060 if (!timeleft || list_empty(&mp->m_sync_list)) {
1062 timeleft = xfs_syncd_centisecs *
1063 msecs_to_jiffies(10);
1064 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1065 list_add_tail(&mp->m_sync_work.w_list,
1068 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1069 list_move(&work->w_list, &tmp);
1070 spin_unlock(&mp->m_sync_lock);
1072 list_for_each_entry_safe(work, n, &tmp, w_list) {
1073 (*work->w_syncer)(mp, work->w_data);
1074 list_del(&work->w_list);
1075 if (work == &mp->m_sync_work)
1077 kmem_free(work, sizeof(struct bhv_vfs_sync_work));
1086 struct super_block *sb)
1088 struct xfs_mount *mp = XFS_M(sb);
1091 kthread_stop(mp->m_sync_task);
1093 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1094 error = xfs_unmount(mp, 0, NULL);
1096 printk("XFS: unmount got error=%d\n", error);
1101 struct super_block *sb)
1103 if (!(sb->s_flags & MS_RDONLY))
1104 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1110 struct super_block *sb,
1113 struct xfs_mount *mp = XFS_M(sb);
1118 * Treat a sync operation like a freeze. This is to work
1119 * around a race in sync_inodes() which works in two phases
1120 * - an asynchronous flush, which can write out an inode
1121 * without waiting for file size updates to complete, and a
1122 * synchronous flush, which wont do anything because the
1123 * async flush removed the inode's dirty flag. Also
1124 * sync_inodes() will not see any files that just have
1125 * outstanding transactions to be flushed because we don't
1126 * dirty the Linux inode until after the transaction I/O
1129 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1131 * First stage of freeze - no more writers will make progress
1132 * now we are here, so we flush delwri and delalloc buffers
1133 * here, then wait for all I/O to complete. Data is frozen at
1134 * that point. Metadata is not frozen, transactions can still
1135 * occur here so don't bother flushing the buftarg (i.e
1136 * SYNC_QUIESCE) because it'll just get dirty again.
1138 flags = SYNC_DATA_QUIESCE;
1140 flags = SYNC_FSDATA;
1142 error = xfs_sync(mp, flags);
1145 if (unlikely(laptop_mode)) {
1146 int prev_sync_seq = mp->m_sync_seq;
1149 * The disk must be active because we're syncing.
1150 * We schedule xfssyncd now (now that the disk is
1151 * active) instead of later (when it might not be).
1153 wake_up_process(mp->m_sync_task);
1155 * We have to wait for the sync iteration to complete.
1156 * If we don't, the disk activity caused by the sync
1157 * will come after the sync is completed, and that
1158 * triggers another sync from laptop mode.
1160 wait_event(mp->m_wait_single_sync_task,
1161 mp->m_sync_seq != prev_sync_seq);
1169 struct dentry *dentry,
1170 struct kstatfs *statp)
1172 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1173 xfs_sb_t *sbp = &mp->m_sb;
1174 __uint64_t fakeinos, id;
1177 statp->f_type = XFS_SB_MAGIC;
1178 statp->f_namelen = MAXNAMELEN - 1;
1180 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1181 statp->f_fsid.val[0] = (u32)id;
1182 statp->f_fsid.val[1] = (u32)(id >> 32);
1184 xfs_icsb_sync_counters_flags(mp, XFS_ICSB_LAZY_COUNT);
1186 spin_lock(&mp->m_sb_lock);
1187 statp->f_bsize = sbp->sb_blocksize;
1188 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1189 statp->f_blocks = sbp->sb_dblocks - lsize;
1190 statp->f_bfree = statp->f_bavail =
1191 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1192 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1194 fakeinos += mp->m_inoadd;
1197 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1198 if (mp->m_maxicount)
1202 statp->f_files = min_t(typeof(statp->f_files),
1205 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1206 spin_unlock(&mp->m_sb_lock);
1208 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1214 struct super_block *sb,
1218 struct xfs_mount *mp = XFS_M(sb);
1219 struct xfs_mount_args *args = xfs_args_allocate(sb, 0);
1222 error = xfs_parseargs(mp, options, args, 1);
1224 error = xfs_mntupdate(mp, flags, args);
1225 kmem_free(args, sizeof(*args));
1230 * Second stage of a freeze. The data is already frozen so we only
1231 * need to take care of themetadata. Once that's done write a dummy
1232 * record to dirty the log in case of a crash while frozen.
1236 struct super_block *sb)
1238 struct xfs_mount *mp = XFS_M(sb);
1240 xfs_attr_quiesce(mp);
1241 xfs_fs_log_dummy(mp);
1245 xfs_fs_show_options(
1247 struct vfsmount *mnt)
1249 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1254 struct super_block *sb,
1257 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1262 struct super_block *sb,
1263 struct fs_quota_stat *fqs)
1265 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1270 struct super_block *sb,
1274 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1279 struct super_block *sb,
1282 struct fs_disk_quota *fdq)
1284 return -XFS_QM_QUOTACTL(XFS_M(sb),
1285 (type == USRQUOTA) ? Q_XGETQUOTA :
1286 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1287 Q_XGETPQUOTA), id, (caddr_t)fdq);
1292 struct super_block *sb,
1295 struct fs_disk_quota *fdq)
1297 return -XFS_QM_QUOTACTL(XFS_M(sb),
1298 (type == USRQUOTA) ? Q_XSETQLIM :
1299 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1300 Q_XSETPQLIM), id, (caddr_t)fdq);
1305 struct super_block *sb,
1309 struct inode *rootvp;
1310 struct xfs_mount *mp = NULL;
1311 struct xfs_mount_args *args = xfs_args_allocate(sb, silent);
1314 mp = xfs_mount_init();
1316 INIT_LIST_HEAD(&mp->m_sync_list);
1317 spin_lock_init(&mp->m_sync_lock);
1318 init_waitqueue_head(&mp->m_wait_single_sync_task);
1323 if (sb->s_flags & MS_RDONLY)
1324 mp->m_flags |= XFS_MOUNT_RDONLY;
1326 error = xfs_parseargs(mp, (char *)data, args, 0);
1330 sb_min_blocksize(sb, BBSIZE);
1331 sb->s_export_op = &xfs_export_operations;
1332 sb->s_qcop = &xfs_quotactl_operations;
1333 sb->s_op = &xfs_super_operations;
1335 error = xfs_mount(mp, args, NULL);
1340 sb->s_magic = XFS_SB_MAGIC;
1341 sb->s_blocksize = mp->m_sb.sb_blocksize;
1342 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1343 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1344 sb->s_time_gran = 1;
1345 set_posix_acl_flag(sb);
1347 rootvp = igrab(mp->m_rootip->i_vnode);
1353 sb->s_root = d_alloc_root(vn_to_inode(rootvp));
1358 if (is_bad_inode(sb->s_root->d_inode)) {
1363 mp->m_sync_work.w_syncer = xfs_sync_worker;
1364 mp->m_sync_work.w_mount = mp;
1365 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1366 if (IS_ERR(mp->m_sync_task)) {
1367 error = -PTR_ERR(mp->m_sync_task);
1371 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1373 kmem_free(args, sizeof(*args));
1385 xfs_unmount(mp, 0, NULL);
1388 kmem_free(args, sizeof(*args));
1394 struct file_system_type *fs_type,
1396 const char *dev_name,
1398 struct vfsmount *mnt)
1400 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1404 static struct super_operations xfs_super_operations = {
1405 .alloc_inode = xfs_fs_alloc_inode,
1406 .destroy_inode = xfs_fs_destroy_inode,
1407 .write_inode = xfs_fs_write_inode,
1408 .clear_inode = xfs_fs_clear_inode,
1409 .put_super = xfs_fs_put_super,
1410 .write_super = xfs_fs_write_super,
1411 .sync_fs = xfs_fs_sync_super,
1412 .write_super_lockfs = xfs_fs_lockfs,
1413 .statfs = xfs_fs_statfs,
1414 .remount_fs = xfs_fs_remount,
1415 .show_options = xfs_fs_show_options,
1418 static struct quotactl_ops xfs_quotactl_operations = {
1419 .quota_sync = xfs_fs_quotasync,
1420 .get_xstate = xfs_fs_getxstate,
1421 .set_xstate = xfs_fs_setxstate,
1422 .get_xquota = xfs_fs_getxquota,
1423 .set_xquota = xfs_fs_setxquota,
1426 static struct file_system_type xfs_fs_type = {
1427 .owner = THIS_MODULE,
1429 .get_sb = xfs_fs_get_sb,
1430 .kill_sb = kill_block_super,
1431 .fs_flags = FS_REQUIRES_DEV,
1439 static char message[] __initdata = KERN_INFO \
1440 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
1446 error = xfs_init_zones();
1450 error = xfs_buf_init();
1459 error = register_filesystem(&xfs_fs_type);
1465 xfs_buf_terminate();
1468 xfs_destroy_zones();
1478 unregister_filesystem(&xfs_fs_type);
1480 xfs_buf_terminate();
1481 xfs_destroy_zones();
1485 module_init(init_xfs_fs);
1486 module_exit(exit_xfs_fs);
1488 MODULE_AUTHOR("Silicon Graphics, Inc.");
1489 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1490 MODULE_LICENSE("GPL");