projects / deliverable / linux.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
xfs: xfs_quiesce_attr() should quiesce the log like unmount
[deliverable/linux.git] / fs / xfs / xfs_super.c
... / ...
CommitLineData
1/*
2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19#include "xfs.h"
20#include "xfs_log.h"
21#include "xfs_inum.h"
22#include "xfs_trans.h"
23#include "xfs_sb.h"
24#include "xfs_ag.h"
25#include "xfs_dir2.h"
26#include "xfs_alloc.h"
27#include "xfs_quota.h"
28#include "xfs_mount.h"
29#include "xfs_bmap_btree.h"
30#include "xfs_alloc_btree.h"
31#include "xfs_ialloc_btree.h"
32#include "xfs_dinode.h"
33#include "xfs_inode.h"
34#include "xfs_btree.h"
35#include "xfs_ialloc.h"
36#include "xfs_bmap.h"
37#include "xfs_rtalloc.h"
38#include "xfs_error.h"
39#include "xfs_itable.h"
40#include "xfs_fsops.h"
41#include "xfs_attr.h"
42#include "xfs_buf_item.h"
43#include "xfs_utils.h"
44#include "xfs_vnodeops.h"
45#include "xfs_log_priv.h"
46#include "xfs_trans_priv.h"
47#include "xfs_filestream.h"
48#include "xfs_da_btree.h"
49#include "xfs_extfree_item.h"
50#include "xfs_mru_cache.h"
51#include "xfs_inode_item.h"
52#include "xfs_sync.h"
53#include "xfs_trace.h"
54
55#include <linux/namei.h>
56#include <linux/init.h>
57#include <linux/slab.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>
63#include <linux/parser.h>
64
65static const struct super_operations xfs_super_operations;
66static kmem_zone_t *xfs_ioend_zone;
67mempool_t *xfs_ioend_pool;
68
69#define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
70#define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
71#define MNTOPT_LOGDEV "logdev" /* log device */
72#define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
73#define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
74#define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
75#define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
76#define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
77#define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
78#define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
79#define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
80#define MNTOPT_MTPT "mtpt" /* filesystem mount point */
81#define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
82#define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
83#define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
84#define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
85#define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
86#define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
87#define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
88 * unwritten extent conversion */
89#define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
90#define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
91#define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
92 * XFS_MAXINUMBER_32 */
93#define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
94#define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
95#define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
96#define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
97 * in stat(). */
98#define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
99#define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
100#define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
101#define MNTOPT_QUOTA "quota" /* disk quotas (user) */
102#define MNTOPT_NOQUOTA "noquota" /* no quotas */
103#define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
104#define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
105#define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
106#define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
107#define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
108#define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
109#define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
110#define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
111#define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
112#define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
113#define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
114#define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
115#define MNTOPT_DISCARD "discard" /* Discard unused blocks */
116#define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
117
118/*
119 * Table driven mount option parser.
120 *
121 * Currently only used for remount, but it will be used for mount
122 * in the future, too.
123 */
124enum {
125 Opt_barrier,
126 Opt_nobarrier,
127 Opt_inode64,
128 Opt_inode32,
129 Opt_err
130};
131
132static const match_table_t tokens = {
133 {Opt_barrier, "barrier"},
134 {Opt_nobarrier, "nobarrier"},
135 {Opt_inode64, "inode64"},
136 {Opt_inode32, "inode32"},
137 {Opt_err, NULL}
138};
139
140
141STATIC unsigned long
142suffix_strtoul(char *s, char **endp, unsigned int base)
143{
144 int last, shift_left_factor = 0;
145 char *value = s;
146
147 last = strlen(value) - 1;
148 if (value[last] == 'K' || value[last] == 'k') {
149 shift_left_factor = 10;
150 value[last] = '\0';
151 }
152 if (value[last] == 'M' || value[last] == 'm') {
153 shift_left_factor = 20;
154 value[last] = '\0';
155 }
156 if (value[last] == 'G' || value[last] == 'g') {
157 shift_left_factor = 30;
158 value[last] = '\0';
159 }
160
161 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
162}
163
164/*
165 * This function fills in xfs_mount_t fields based on mount args.
166 * Note: the superblock has _not_ yet been read in.
167 *
168 * Note that this function leaks the various device name allocations on
169 * failure. The caller takes care of them.
170 */
171STATIC int
172xfs_parseargs(
173 struct xfs_mount *mp,
174 char *options)
175{
176 struct super_block *sb = mp->m_super;
177 char *this_char, *value, *eov;
178 int dsunit = 0;
179 int dswidth = 0;
180 int iosize = 0;
181 __uint8_t iosizelog = 0;
182
183 /*
184 * set up the mount name first so all the errors will refer to the
185 * correct device.
186 */
187 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
188 if (!mp->m_fsname)
189 return ENOMEM;
190 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
191
192 /*
193 * Copy binary VFS mount flags we are interested in.
194 */
195 if (sb->s_flags & MS_RDONLY)
196 mp->m_flags |= XFS_MOUNT_RDONLY;
197 if (sb->s_flags & MS_DIRSYNC)
198 mp->m_flags |= XFS_MOUNT_DIRSYNC;
199 if (sb->s_flags & MS_SYNCHRONOUS)
200 mp->m_flags |= XFS_MOUNT_WSYNC;
201
202 /*
203 * Set some default flags that could be cleared by the mount option
204 * parsing.
205 */
206 mp->m_flags |= XFS_MOUNT_BARRIER;
207 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
208#if !XFS_BIG_INUMS
209 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
210#endif
211
212 /*
213 * These can be overridden by the mount option parsing.
214 */
215 mp->m_logbufs = -1;
216 mp->m_logbsize = -1;
217
218 if (!options)
219 goto done;
220
221 while ((this_char = strsep(&options, ",")) != NULL) {
222 if (!*this_char)
223 continue;
224 if ((value = strchr(this_char, '=')) != NULL)
225 *value++ = 0;
226
227 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
228 if (!value || !*value) {
229 xfs_warn(mp, "%s option requires an argument",
230 this_char);
231 return EINVAL;
232 }
233 mp->m_logbufs = simple_strtoul(value, &eov, 10);
234 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
235 if (!value || !*value) {
236 xfs_warn(mp, "%s option requires an argument",
237 this_char);
238 return EINVAL;
239 }
240 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
241 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
242 if (!value || !*value) {
243 xfs_warn(mp, "%s option requires an argument",
244 this_char);
245 return EINVAL;
246 }
247 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
248 if (!mp->m_logname)
249 return ENOMEM;
250 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
251 xfs_warn(mp, "%s option not allowed on this system",
252 this_char);
253 return EINVAL;
254 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
255 if (!value || !*value) {
256 xfs_warn(mp, "%s option requires an argument",
257 this_char);
258 return EINVAL;
259 }
260 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
261 if (!mp->m_rtname)
262 return ENOMEM;
263 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
264 if (!value || !*value) {
265 xfs_warn(mp, "%s option requires an argument",
266 this_char);
267 return EINVAL;
268 }
269 iosize = simple_strtoul(value, &eov, 10);
270 iosizelog = ffs(iosize) - 1;
271 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
272 if (!value || !*value) {
273 xfs_warn(mp, "%s option requires an argument",
274 this_char);
275 return EINVAL;
276 }
277 iosize = suffix_strtoul(value, &eov, 10);
278 iosizelog = ffs(iosize) - 1;
279 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
280 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
281 mp->m_flags |= XFS_MOUNT_GRPID;
282 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
283 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
284 mp->m_flags &= ~XFS_MOUNT_GRPID;
285 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
286 mp->m_flags |= XFS_MOUNT_WSYNC;
287 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
288 mp->m_flags |= XFS_MOUNT_NORECOVERY;
289 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
290 mp->m_flags |= XFS_MOUNT_NOALIGN;
291 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
292 mp->m_flags |= XFS_MOUNT_SWALLOC;
293 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
294 if (!value || !*value) {
295 xfs_warn(mp, "%s option requires an argument",
296 this_char);
297 return EINVAL;
298 }
299 dsunit = simple_strtoul(value, &eov, 10);
300 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
301 if (!value || !*value) {
302 xfs_warn(mp, "%s option requires an argument",
303 this_char);
304 return EINVAL;
305 }
306 dswidth = simple_strtoul(value, &eov, 10);
307 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
308 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
309 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
310 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
311#if !XFS_BIG_INUMS
312 xfs_warn(mp, "%s option not allowed on this system",
313 this_char);
314 return EINVAL;
315#endif
316 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
317 mp->m_flags |= XFS_MOUNT_NOUUID;
318 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
319 mp->m_flags |= XFS_MOUNT_BARRIER;
320 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
321 mp->m_flags &= ~XFS_MOUNT_BARRIER;
322 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
323 mp->m_flags |= XFS_MOUNT_IKEEP;
324 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
325 mp->m_flags &= ~XFS_MOUNT_IKEEP;
326 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
327 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
328 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
329 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
330 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
331 mp->m_flags |= XFS_MOUNT_ATTR2;
332 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
333 mp->m_flags &= ~XFS_MOUNT_ATTR2;
334 mp->m_flags |= XFS_MOUNT_NOATTR2;
335 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
336 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
337 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
338 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
339 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
340 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
341 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
342 !strcmp(this_char, MNTOPT_UQUOTA) ||
343 !strcmp(this_char, MNTOPT_USRQUOTA)) {
344 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
345 XFS_UQUOTA_ENFD);
346 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
347 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
348 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
349 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
350 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
351 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
352 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
353 XFS_OQUOTA_ENFD);
354 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
355 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
356 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
357 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
358 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
359 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
360 XFS_OQUOTA_ENFD);
361 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
362 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
363 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
364 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
365 xfs_warn(mp,
366 "delaylog is the default now, option is deprecated.");
367 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
368 xfs_warn(mp,
369 "nodelaylog support has been removed, option is deprecated.");
370 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
371 mp->m_flags |= XFS_MOUNT_DISCARD;
372 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
373 mp->m_flags &= ~XFS_MOUNT_DISCARD;
374 } else if (!strcmp(this_char, "ihashsize")) {
375 xfs_warn(mp,
376 "ihashsize no longer used, option is deprecated.");
377 } else if (!strcmp(this_char, "osyncisdsync")) {
378 xfs_warn(mp,
379 "osyncisdsync has no effect, option is deprecated.");
380 } else if (!strcmp(this_char, "osyncisosync")) {
381 xfs_warn(mp,
382 "osyncisosync has no effect, option is deprecated.");
383 } else if (!strcmp(this_char, "irixsgid")) {
384 xfs_warn(mp,
385 "irixsgid is now a sysctl(2) variable, option is deprecated.");
386 } else {
387 xfs_warn(mp, "unknown mount option [%s].", this_char);
388 return EINVAL;
389 }
390 }
391
392 /*
393 * no recovery flag requires a read-only mount
394 */
395 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
396 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
397 xfs_warn(mp, "no-recovery mounts must be read-only.");
398 return EINVAL;
399 }
400
401 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
402 xfs_warn(mp,
403 "sunit and swidth options incompatible with the noalign option");
404 return EINVAL;
405 }
406
407#ifndef CONFIG_XFS_QUOTA
408 if (XFS_IS_QUOTA_RUNNING(mp)) {
409 xfs_warn(mp, "quota support not available in this kernel.");
410 return EINVAL;
411 }
412#endif
413
414 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
415 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
416 xfs_warn(mp, "cannot mount with both project and group quota");
417 return EINVAL;
418 }
419
420 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
421 xfs_warn(mp, "sunit and swidth must be specified together");
422 return EINVAL;
423 }
424
425 if (dsunit && (dswidth % dsunit != 0)) {
426 xfs_warn(mp,
427 "stripe width (%d) must be a multiple of the stripe unit (%d)",
428 dswidth, dsunit);
429 return EINVAL;
430 }
431
432done:
433 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
434 /*
435 * At this point the superblock has not been read
436 * in, therefore we do not know the block size.
437 * Before the mount call ends we will convert
438 * these to FSBs.
439 */
440 if (dsunit) {
441 mp->m_dalign = dsunit;
442 mp->m_flags |= XFS_MOUNT_RETERR;
443 }
444
445 if (dswidth)
446 mp->m_swidth = dswidth;
447 }
448
449 if (mp->m_logbufs != -1 &&
450 mp->m_logbufs != 0 &&
451 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
452 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
453 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
454 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
455 return XFS_ERROR(EINVAL);
456 }
457 if (mp->m_logbsize != -1 &&
458 mp->m_logbsize != 0 &&
459 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
460 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
461 !is_power_of_2(mp->m_logbsize))) {
462 xfs_warn(mp,
463 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
464 mp->m_logbsize);
465 return XFS_ERROR(EINVAL);
466 }
467
468 if (iosizelog) {
469 if (iosizelog > XFS_MAX_IO_LOG ||
470 iosizelog < XFS_MIN_IO_LOG) {
471 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
472 iosizelog, XFS_MIN_IO_LOG,
473 XFS_MAX_IO_LOG);
474 return XFS_ERROR(EINVAL);
475 }
476
477 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
478 mp->m_readio_log = iosizelog;
479 mp->m_writeio_log = iosizelog;
480 }
481
482 return 0;
483}
484
485struct proc_xfs_info {
486 int flag;
487 char *str;
488};
489
490STATIC int
491xfs_showargs(
492 struct xfs_mount *mp,
493 struct seq_file *m)
494{
495 static struct proc_xfs_info xfs_info_set[] = {
496 /* the few simple ones we can get from the mount struct */
497 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
498 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
499 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
500 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
501 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
502 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
503 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
504 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
505 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
506 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
507 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
508 { 0, NULL }
509 };
510 static struct proc_xfs_info xfs_info_unset[] = {
511 /* the few simple ones we can get from the mount struct */
512 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
513 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
514 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
515 { 0, NULL }
516 };
517 struct proc_xfs_info *xfs_infop;
518
519 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
520 if (mp->m_flags & xfs_infop->flag)
521 seq_puts(m, xfs_infop->str);
522 }
523 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
524 if (!(mp->m_flags & xfs_infop->flag))
525 seq_puts(m, xfs_infop->str);
526 }
527
528 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
529 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
530 (int)(1 << mp->m_writeio_log) >> 10);
531
532 if (mp->m_logbufs > 0)
533 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
534 if (mp->m_logbsize > 0)
535 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
536
537 if (mp->m_logname)
538 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
539 if (mp->m_rtname)
540 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
541
542 if (mp->m_dalign > 0)
543 seq_printf(m, "," MNTOPT_SUNIT "=%d",
544 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
545 if (mp->m_swidth > 0)
546 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
547 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
548
549 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
550 seq_puts(m, "," MNTOPT_USRQUOTA);
551 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
552 seq_puts(m, "," MNTOPT_UQUOTANOENF);
553
554 /* Either project or group quotas can be active, not both */
555
556 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
557 if (mp->m_qflags & XFS_OQUOTA_ENFD)
558 seq_puts(m, "," MNTOPT_PRJQUOTA);
559 else
560 seq_puts(m, "," MNTOPT_PQUOTANOENF);
561 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
562 if (mp->m_qflags & XFS_OQUOTA_ENFD)
563 seq_puts(m, "," MNTOPT_GRPQUOTA);
564 else
565 seq_puts(m, "," MNTOPT_GQUOTANOENF);
566 }
567
568 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
569 seq_puts(m, "," MNTOPT_NOQUOTA);
570
571 return 0;
572}
573__uint64_t
574xfs_max_file_offset(
575 unsigned int blockshift)
576{
577 unsigned int pagefactor = 1;
578 unsigned int bitshift = BITS_PER_LONG - 1;
579
580 /* Figure out maximum filesize, on Linux this can depend on
581 * the filesystem blocksize (on 32 bit platforms).
582 * __block_write_begin does this in an [unsigned] long...
583 * page->index << (PAGE_CACHE_SHIFT - bbits)
584 * So, for page sized blocks (4K on 32 bit platforms),
585 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
586 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
587 * but for smaller blocksizes it is less (bbits = log2 bsize).
588 * Note1: get_block_t takes a long (implicit cast from above)
589 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
590 * can optionally convert the [unsigned] long from above into
591 * an [unsigned] long long.
592 */
593
594#if BITS_PER_LONG == 32
595# if defined(CONFIG_LBDAF)
596 ASSERT(sizeof(sector_t) == 8);
597 pagefactor = PAGE_CACHE_SIZE;
598 bitshift = BITS_PER_LONG;
599# else
600 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
601# endif
602#endif
603
604 return (((__uint64_t)pagefactor) << bitshift) - 1;
605}
606
607xfs_agnumber_t
608xfs_set_inode32(struct xfs_mount *mp)
609{
610 xfs_agnumber_t index = 0;
611 xfs_agnumber_t maxagi = 0;
612 xfs_sb_t *sbp = &mp->m_sb;
613 xfs_agnumber_t max_metadata;
614 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0);
615 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino);
616 xfs_perag_t *pag;
617
618 /* Calculate how much should be reserved for inodes to meet
619 * the max inode percentage.
620 */
621 if (mp->m_maxicount) {
622 __uint64_t icount;
623
624 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
625 do_div(icount, 100);
626 icount += sbp->sb_agblocks - 1;
627 do_div(icount, sbp->sb_agblocks);
628 max_metadata = icount;
629 } else {
630 max_metadata = sbp->sb_agcount;
631 }
632
633 for (index = 0; index < sbp->sb_agcount; index++) {
634 ino = XFS_AGINO_TO_INO(mp, index, agino);
635
636 if (ino > XFS_MAXINUMBER_32) {
637 pag = xfs_perag_get(mp, index);
638 pag->pagi_inodeok = 0;
639 pag->pagf_metadata = 0;
640 xfs_perag_put(pag);
641 continue;
642 }
643
644 pag = xfs_perag_get(mp, index);
645 pag->pagi_inodeok = 1;
646 maxagi++;
647 if (index < max_metadata)
648 pag->pagf_metadata = 1;
649 xfs_perag_put(pag);
650 }
651 mp->m_flags |= (XFS_MOUNT_32BITINODES |
652 XFS_MOUNT_SMALL_INUMS);
653
654 return maxagi;
655}
656
657xfs_agnumber_t
658xfs_set_inode64(struct xfs_mount *mp)
659{
660 xfs_agnumber_t index = 0;
661
662 for (index = 0; index < mp->m_sb.sb_agcount; index++) {
663 struct xfs_perag *pag;
664
665 pag = xfs_perag_get(mp, index);
666 pag->pagi_inodeok = 1;
667 pag->pagf_metadata = 0;
668 xfs_perag_put(pag);
669 }
670
671 /* There is no need for lock protection on m_flags,
672 * the rw_semaphore of the VFS superblock is locked
673 * during mount/umount/remount operations, so this is
674 * enough to avoid concurency on the m_flags field
675 */
676 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
677 XFS_MOUNT_SMALL_INUMS);
678 return index;
679}
680
681STATIC int
682xfs_blkdev_get(
683 xfs_mount_t *mp,
684 const char *name,
685 struct block_device **bdevp)
686{
687 int error = 0;
688
689 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
690 mp);
691 if (IS_ERR(*bdevp)) {
692 error = PTR_ERR(*bdevp);
693 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
694 }
695
696 return -error;
697}
698
699STATIC void
700xfs_blkdev_put(
701 struct block_device *bdev)
702{
703 if (bdev)
704 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
705}
706
707void
708xfs_blkdev_issue_flush(
709 xfs_buftarg_t *buftarg)
710{
711 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
712}
713
714STATIC void
715xfs_close_devices(
716 struct xfs_mount *mp)
717{
718 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
719 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
720 xfs_free_buftarg(mp, mp->m_logdev_targp);
721 xfs_blkdev_put(logdev);
722 }
723 if (mp->m_rtdev_targp) {
724 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
725 xfs_free_buftarg(mp, mp->m_rtdev_targp);
726 xfs_blkdev_put(rtdev);
727 }
728 xfs_free_buftarg(mp, mp->m_ddev_targp);
729}
730
731/*
732 * The file system configurations are:
733 * (1) device (partition) with data and internal log
734 * (2) logical volume with data and log subvolumes.
735 * (3) logical volume with data, log, and realtime subvolumes.
736 *
737 * We only have to handle opening the log and realtime volumes here if
738 * they are present. The data subvolume has already been opened by
739 * get_sb_bdev() and is stored in sb->s_bdev.
740 */
741STATIC int
742xfs_open_devices(
743 struct xfs_mount *mp)
744{
745 struct block_device *ddev = mp->m_super->s_bdev;
746 struct block_device *logdev = NULL, *rtdev = NULL;
747 int error;
748
749 /*
750 * Open real time and log devices - order is important.
751 */
752 if (mp->m_logname) {
753 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
754 if (error)
755 goto out;
756 }
757
758 if (mp->m_rtname) {
759 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
760 if (error)
761 goto out_close_logdev;
762
763 if (rtdev == ddev || rtdev == logdev) {
764 xfs_warn(mp,
765 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
766 error = EINVAL;
767 goto out_close_rtdev;
768 }
769 }
770
771 /*
772 * Setup xfs_mount buffer target pointers
773 */
774 error = ENOMEM;
775 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
776 if (!mp->m_ddev_targp)
777 goto out_close_rtdev;
778
779 if (rtdev) {
780 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
781 mp->m_fsname);
782 if (!mp->m_rtdev_targp)
783 goto out_free_ddev_targ;
784 }
785
786 if (logdev && logdev != ddev) {
787 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
788 mp->m_fsname);
789 if (!mp->m_logdev_targp)
790 goto out_free_rtdev_targ;
791 } else {
792 mp->m_logdev_targp = mp->m_ddev_targp;
793 }
794
795 return 0;
796
797 out_free_rtdev_targ:
798 if (mp->m_rtdev_targp)
799 xfs_free_buftarg(mp, mp->m_rtdev_targp);
800 out_free_ddev_targ:
801 xfs_free_buftarg(mp, mp->m_ddev_targp);
802 out_close_rtdev:
803 if (rtdev)
804 xfs_blkdev_put(rtdev);
805 out_close_logdev:
806 if (logdev && logdev != ddev)
807 xfs_blkdev_put(logdev);
808 out:
809 return error;
810}
811
812/*
813 * Setup xfs_mount buffer target pointers based on superblock
814 */
815STATIC int
816xfs_setup_devices(
817 struct xfs_mount *mp)
818{
819 int error;
820
821 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
822 mp->m_sb.sb_sectsize);
823 if (error)
824 return error;
825
826 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
827 unsigned int log_sector_size = BBSIZE;
828
829 if (xfs_sb_version_hassector(&mp->m_sb))
830 log_sector_size = mp->m_sb.sb_logsectsize;
831 error = xfs_setsize_buftarg(mp->m_logdev_targp,
832 mp->m_sb.sb_blocksize,
833 log_sector_size);
834 if (error)
835 return error;
836 }
837 if (mp->m_rtdev_targp) {
838 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
839 mp->m_sb.sb_blocksize,
840 mp->m_sb.sb_sectsize);
841 if (error)
842 return error;
843 }
844
845 return 0;
846}
847
848STATIC int
849xfs_init_mount_workqueues(
850 struct xfs_mount *mp)
851{
852 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
853 WQ_MEM_RECLAIM, 0, mp->m_fsname);
854 if (!mp->m_data_workqueue)
855 goto out;
856
857 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
858 WQ_MEM_RECLAIM, 0, mp->m_fsname);
859 if (!mp->m_unwritten_workqueue)
860 goto out_destroy_data_iodone_queue;
861
862 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
863 WQ_MEM_RECLAIM, 0, mp->m_fsname);
864 if (!mp->m_cil_workqueue)
865 goto out_destroy_unwritten;
866
867 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
868 WQ_NON_REENTRANT, 0, mp->m_fsname);
869 if (!mp->m_reclaim_workqueue)
870 goto out_destroy_cil;
871
872 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
873 WQ_NON_REENTRANT, 0, mp->m_fsname);
874 if (!mp->m_log_workqueue)
875 goto out_destroy_reclaim;
876
877 return 0;
878
879out_destroy_reclaim:
880 destroy_workqueue(mp->m_reclaim_workqueue);
881out_destroy_cil:
882 destroy_workqueue(mp->m_cil_workqueue);
883out_destroy_unwritten:
884 destroy_workqueue(mp->m_unwritten_workqueue);
885out_destroy_data_iodone_queue:
886 destroy_workqueue(mp->m_data_workqueue);
887out:
888 return -ENOMEM;
889}
890
891STATIC void
892xfs_destroy_mount_workqueues(
893 struct xfs_mount *mp)
894{
895 destroy_workqueue(mp->m_log_workqueue);
896 destroy_workqueue(mp->m_reclaim_workqueue);
897 destroy_workqueue(mp->m_cil_workqueue);
898 destroy_workqueue(mp->m_data_workqueue);
899 destroy_workqueue(mp->m_unwritten_workqueue);
900}
901
902/*
903 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
904 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
905 * for IO to complete so that we effectively throttle multiple callers to the
906 * rate at which IO is completing.
907 */
908void
909xfs_flush_inodes(
910 struct xfs_mount *mp)
911{
912 struct super_block *sb = mp->m_super;
913
914 if (down_read_trylock(&sb->s_umount)) {
915 sync_inodes_sb(sb);
916 up_read(&sb->s_umount);
917 }
918}
919
920/* Catch misguided souls that try to use this interface on XFS */
921STATIC struct inode *
922xfs_fs_alloc_inode(
923 struct super_block *sb)
924{
925 BUG();
926 return NULL;
927}
928
929/*
930 * Now that the generic code is guaranteed not to be accessing
931 * the linux inode, we can reclaim the inode.
932 */
933STATIC void
934xfs_fs_destroy_inode(
935 struct inode *inode)
936{
937 struct xfs_inode *ip = XFS_I(inode);
938
939 trace_xfs_destroy_inode(ip);
940
941 XFS_STATS_INC(vn_reclaim);
942
943 /* bad inode, get out here ASAP */
944 if (is_bad_inode(inode))
945 goto out_reclaim;
946
947 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
948
949 /*
950 * We should never get here with one of the reclaim flags already set.
951 */
952 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
953 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
954
955 /*
956 * We always use background reclaim here because even if the
957 * inode is clean, it still may be under IO and hence we have
958 * to take the flush lock. The background reclaim path handles
959 * this more efficiently than we can here, so simply let background
960 * reclaim tear down all inodes.
961 */
962out_reclaim:
963 xfs_inode_set_reclaim_tag(ip);
964}
965
966/*
967 * Slab object creation initialisation for the XFS inode.
968 * This covers only the idempotent fields in the XFS inode;
969 * all other fields need to be initialised on allocation
970 * from the slab. This avoids the need to repeatedly initialise
971 * fields in the xfs inode that left in the initialise state
972 * when freeing the inode.
973 */
974STATIC void
975xfs_fs_inode_init_once(
976 void *inode)
977{
978 struct xfs_inode *ip = inode;
979
980 memset(ip, 0, sizeof(struct xfs_inode));
981
982 /* vfs inode */
983 inode_init_once(VFS_I(ip));
984
985 /* xfs inode */
986 atomic_set(&ip->i_pincount, 0);
987 spin_lock_init(&ip->i_flags_lock);
988
989 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
990 "xfsino", ip->i_ino);
991}
992
993STATIC void
994xfs_fs_evict_inode(
995 struct inode *inode)
996{
997 xfs_inode_t *ip = XFS_I(inode);
998
999 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1000
1001 trace_xfs_evict_inode(ip);
1002
1003 truncate_inode_pages(&inode->i_data, 0);
1004 clear_inode(inode);
1005 XFS_STATS_INC(vn_rele);
1006 XFS_STATS_INC(vn_remove);
1007 XFS_STATS_DEC(vn_active);
1008
1009 xfs_inactive(ip);
1010}
1011
1012/*
1013 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1014 * serialised against cache hits here via the inode->i_lock and igrab() in
1015 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1016 * racing with us, and it avoids needing to grab a spinlock here for every inode
1017 * we drop the final reference on.
1018 */
1019STATIC int
1020xfs_fs_drop_inode(
1021 struct inode *inode)
1022{
1023 struct xfs_inode *ip = XFS_I(inode);
1024
1025 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1026}
1027
1028STATIC void
1029xfs_free_fsname(
1030 struct xfs_mount *mp)
1031{
1032 kfree(mp->m_fsname);
1033 kfree(mp->m_rtname);
1034 kfree(mp->m_logname);
1035}
1036
1037STATIC void
1038xfs_fs_put_super(
1039 struct super_block *sb)
1040{
1041 struct xfs_mount *mp = XFS_M(sb);
1042
1043 xfs_filestream_unmount(mp);
1044 xfs_unmountfs(mp);
1045
1046 xfs_freesb(mp);
1047 xfs_icsb_destroy_counters(mp);
1048 xfs_destroy_mount_workqueues(mp);
1049 xfs_close_devices(mp);
1050 xfs_free_fsname(mp);
1051 kfree(mp);
1052}
1053
1054STATIC int
1055xfs_fs_sync_fs(
1056 struct super_block *sb,
1057 int wait)
1058{
1059 struct xfs_mount *mp = XFS_M(sb);
1060
1061 /*
1062 * Doing anything during the async pass would be counterproductive.
1063 */
1064 if (!wait)
1065 return 0;
1066
1067 xfs_log_force(mp, XFS_LOG_SYNC);
1068 if (laptop_mode) {
1069 /*
1070 * The disk must be active because we're syncing.
1071 * We schedule log work now (now that the disk is
1072 * active) instead of later (when it might not be).
1073 */
1074 flush_delayed_work(&mp->m_log->l_work);
1075 }
1076
1077 return 0;
1078}
1079
1080STATIC int
1081xfs_fs_statfs(
1082 struct dentry *dentry,
1083 struct kstatfs *statp)
1084{
1085 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1086 xfs_sb_t *sbp = &mp->m_sb;
1087 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1088 __uint64_t fakeinos, id;
1089 xfs_extlen_t lsize;
1090 __int64_t ffree;
1091
1092 statp->f_type = XFS_SB_MAGIC;
1093 statp->f_namelen = MAXNAMELEN - 1;
1094
1095 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1096 statp->f_fsid.val[0] = (u32)id;
1097 statp->f_fsid.val[1] = (u32)(id >> 32);
1098
1099 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1100
1101 spin_lock(&mp->m_sb_lock);
1102 statp->f_bsize = sbp->sb_blocksize;
1103 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1104 statp->f_blocks = sbp->sb_dblocks - lsize;
1105 statp->f_bfree = statp->f_bavail =
1106 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1107 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1108 statp->f_files =
1109 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1110 if (mp->m_maxicount)
1111 statp->f_files = min_t(typeof(statp->f_files),
1112 statp->f_files,
1113 mp->m_maxicount);
1114
1115 /* make sure statp->f_ffree does not underflow */
1116 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1117 statp->f_ffree = max_t(__int64_t, ffree, 0);
1118
1119 spin_unlock(&mp->m_sb_lock);
1120
1121 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1122 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1123 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1124 xfs_qm_statvfs(ip, statp);
1125 return 0;
1126}
1127
1128STATIC void
1129xfs_save_resvblks(struct xfs_mount *mp)
1130{
1131 __uint64_t resblks = 0;
1132
1133 mp->m_resblks_save = mp->m_resblks;
1134 xfs_reserve_blocks(mp, &resblks, NULL);
1135}
1136
1137STATIC void
1138xfs_restore_resvblks(struct xfs_mount *mp)
1139{
1140 __uint64_t resblks;
1141
1142 if (mp->m_resblks_save) {
1143 resblks = mp->m_resblks_save;
1144 mp->m_resblks_save = 0;
1145 } else
1146 resblks = xfs_default_resblks(mp);
1147
1148 xfs_reserve_blocks(mp, &resblks, NULL);
1149}
1150
1151/*
1152 * Trigger writeback of all the dirty metadata in the file system.
1153 *
1154 * This ensures that the metadata is written to their location on disk rather
1155 * than just existing in transactions in the log. This means after a quiesce
1156 * there is no log replay required to write the inodes to disk - this is the
1157 * primary difference between a sync and a quiesce.
1158 *
1159 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1160 * it is started again when appropriate.
1161 */
1162void
1163xfs_quiesce_attr(
1164 struct xfs_mount *mp)
1165{
1166 int error = 0;
1167
1168 /* wait for all modifications to complete */
1169 while (atomic_read(&mp->m_active_trans) > 0)
1170 delay(100);
1171
1172 /* force the log to unpin objects from the now complete transactions */
1173 xfs_log_force(mp, XFS_LOG_SYNC);
1174
1175 /* reclaim inodes to do any IO before the freeze completes */
1176 xfs_reclaim_inodes(mp, 0);
1177 xfs_reclaim_inodes(mp, SYNC_WAIT);
1178
1179 /* Push the superblock and write an unmount record */
1180 error = xfs_log_sbcount(mp);
1181 if (error)
1182 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1183 "Frozen image may not be consistent.");
1184 /*
1185 * Just warn here till VFS can correctly support
1186 * read-only remount without racing.
1187 */
1188 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1189
1190 xfs_log_quiesce(mp);
1191}
1192
1193STATIC int
1194xfs_fs_remount(
1195 struct super_block *sb,
1196 int *flags,
1197 char *options)
1198{
1199 struct xfs_mount *mp = XFS_M(sb);
1200 substring_t args[MAX_OPT_ARGS];
1201 char *p;
1202 int error;
1203
1204 while ((p = strsep(&options, ",")) != NULL) {
1205 int token;
1206
1207 if (!*p)
1208 continue;
1209
1210 token = match_token(p, tokens, args);
1211 switch (token) {
1212 case Opt_barrier:
1213 mp->m_flags |= XFS_MOUNT_BARRIER;
1214 break;
1215 case Opt_nobarrier:
1216 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1217 break;
1218 case Opt_inode64:
1219 mp->m_maxagi = xfs_set_inode64(mp);
1220 break;
1221 case Opt_inode32:
1222 mp->m_maxagi = xfs_set_inode32(mp);
1223 break;
1224 default:
1225 /*
1226 * Logically we would return an error here to prevent
1227 * users from believing they might have changed
1228 * mount options using remount which can't be changed.
1229 *
1230 * But unfortunately mount(8) adds all options from
1231 * mtab and fstab to the mount arguments in some cases
1232 * so we can't blindly reject options, but have to
1233 * check for each specified option if it actually
1234 * differs from the currently set option and only
1235 * reject it if that's the case.
1236 *
1237 * Until that is implemented we return success for
1238 * every remount request, and silently ignore all
1239 * options that we can't actually change.
1240 */
1241#if 0
1242 xfs_info(mp,
1243 "mount option \"%s\" not supported for remount\n", p);
1244 return -EINVAL;
1245#else
1246 break;
1247#endif
1248 }
1249 }
1250
1251 /* ro -> rw */
1252 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1253 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1254
1255 /*
1256 * If this is the first remount to writeable state we
1257 * might have some superblock changes to update.
1258 */
1259 if (mp->m_update_flags) {
1260 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1261 if (error) {
1262 xfs_warn(mp, "failed to write sb changes");
1263 return error;
1264 }
1265 mp->m_update_flags = 0;
1266 }
1267
1268 /*
1269 * Fill out the reserve pool if it is empty. Use the stashed
1270 * value if it is non-zero, otherwise go with the default.
1271 */
1272 xfs_restore_resvblks(mp);
1273 xfs_log_work_queue(mp);
1274 }
1275
1276 /* rw -> ro */
1277 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1278 /*
1279 * Before we sync the metadata, we need to free up the reserve
1280 * block pool so that the used block count in the superblock on
1281 * disk is correct at the end of the remount. Stash the current
1282 * reserve pool size so that if we get remounted rw, we can
1283 * return it to the same size.
1284 */
1285 xfs_save_resvblks(mp);
1286 xfs_quiesce_attr(mp);
1287 mp->m_flags |= XFS_MOUNT_RDONLY;
1288 }
1289
1290 return 0;
1291}
1292
1293/*
1294 * Second stage of a freeze. The data is already frozen so we only
1295 * need to take care of the metadata. Once that's done write a dummy
1296 * record to dirty the log in case of a crash while frozen.
1297 */
1298STATIC int
1299xfs_fs_freeze(
1300 struct super_block *sb)
1301{
1302 struct xfs_mount *mp = XFS_M(sb);
1303
1304 xfs_save_resvblks(mp);
1305 xfs_quiesce_attr(mp);
1306 return -xfs_fs_log_dummy(mp);
1307}
1308
1309STATIC int
1310xfs_fs_unfreeze(
1311 struct super_block *sb)
1312{
1313 struct xfs_mount *mp = XFS_M(sb);
1314
1315 xfs_restore_resvblks(mp);
1316 xfs_log_work_queue(mp);
1317 return 0;
1318}
1319
1320STATIC int
1321xfs_fs_show_options(
1322 struct seq_file *m,
1323 struct dentry *root)
1324{
1325 return -xfs_showargs(XFS_M(root->d_sb), m);
1326}
1327
1328/*
1329 * This function fills in xfs_mount_t fields based on mount args.
1330 * Note: the superblock _has_ now been read in.
1331 */
1332STATIC int
1333xfs_finish_flags(
1334 struct xfs_mount *mp)
1335{
1336 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1337
1338 /* Fail a mount where the logbuf is smaller than the log stripe */
1339 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1340 if (mp->m_logbsize <= 0 &&
1341 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1342 mp->m_logbsize = mp->m_sb.sb_logsunit;
1343 } else if (mp->m_logbsize > 0 &&
1344 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1345 xfs_warn(mp,
1346 "logbuf size must be greater than or equal to log stripe size");
1347 return XFS_ERROR(EINVAL);
1348 }
1349 } else {
1350 /* Fail a mount if the logbuf is larger than 32K */
1351 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1352 xfs_warn(mp,
1353 "logbuf size for version 1 logs must be 16K or 32K");
1354 return XFS_ERROR(EINVAL);
1355 }
1356 }
1357
1358 /*
1359 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1360 * told by noattr2 to turn it off
1361 */
1362 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1363 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1364 mp->m_flags |= XFS_MOUNT_ATTR2;
1365
1366 /*
1367 * prohibit r/w mounts of read-only filesystems
1368 */
1369 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1370 xfs_warn(mp,
1371 "cannot mount a read-only filesystem as read-write");
1372 return XFS_ERROR(EROFS);
1373 }
1374
1375 return 0;
1376}
1377
1378STATIC int
1379xfs_fs_fill_super(
1380 struct super_block *sb,
1381 void *data,
1382 int silent)
1383{
1384 struct inode *root;
1385 struct xfs_mount *mp = NULL;
1386 int flags = 0, error = ENOMEM;
1387
1388 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1389 if (!mp)
1390 goto out;
1391
1392 spin_lock_init(&mp->m_sb_lock);
1393 mutex_init(&mp->m_growlock);
1394 atomic_set(&mp->m_active_trans, 0);
1395 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1396
1397 mp->m_super = sb;
1398 sb->s_fs_info = mp;
1399
1400 error = xfs_parseargs(mp, (char *)data);
1401 if (error)
1402 goto out_free_fsname;
1403
1404 sb_min_blocksize(sb, BBSIZE);
1405 sb->s_xattr = xfs_xattr_handlers;
1406 sb->s_export_op = &xfs_export_operations;
1407#ifdef CONFIG_XFS_QUOTA
1408 sb->s_qcop = &xfs_quotactl_operations;
1409#endif
1410 sb->s_op = &xfs_super_operations;
1411
1412 if (silent)
1413 flags |= XFS_MFSI_QUIET;
1414
1415 error = xfs_open_devices(mp);
1416 if (error)
1417 goto out_free_fsname;
1418
1419 error = xfs_init_mount_workqueues(mp);
1420 if (error)
1421 goto out_close_devices;
1422
1423 error = xfs_icsb_init_counters(mp);
1424 if (error)
1425 goto out_destroy_workqueues;
1426
1427 error = xfs_readsb(mp, flags);
1428 if (error)
1429 goto out_destroy_counters;
1430
1431 error = xfs_finish_flags(mp);
1432 if (error)
1433 goto out_free_sb;
1434
1435 error = xfs_setup_devices(mp);
1436 if (error)
1437 goto out_free_sb;
1438
1439 error = xfs_filestream_mount(mp);
1440 if (error)
1441 goto out_free_sb;
1442
1443 /*
1444 * we must configure the block size in the superblock before we run the
1445 * full mount process as the mount process can lookup and cache inodes.
1446 */
1447 sb->s_magic = XFS_SB_MAGIC;
1448 sb->s_blocksize = mp->m_sb.sb_blocksize;
1449 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1450 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1451 sb->s_max_links = XFS_MAXLINK;
1452 sb->s_time_gran = 1;
1453 set_posix_acl_flag(sb);
1454
1455 error = xfs_mountfs(mp);
1456 if (error)
1457 goto out_filestream_unmount;
1458
1459 root = igrab(VFS_I(mp->m_rootip));
1460 if (!root) {
1461 error = ENOENT;
1462 goto out_unmount;
1463 }
1464 if (is_bad_inode(root)) {
1465 error = EINVAL;
1466 goto out_unmount;
1467 }
1468 sb->s_root = d_make_root(root);
1469 if (!sb->s_root) {
1470 error = ENOMEM;
1471 goto out_unmount;
1472 }
1473
1474 return 0;
1475
1476 out_filestream_unmount:
1477 xfs_filestream_unmount(mp);
1478 out_free_sb:
1479 xfs_freesb(mp);
1480 out_destroy_counters:
1481 xfs_icsb_destroy_counters(mp);
1482out_destroy_workqueues:
1483 xfs_destroy_mount_workqueues(mp);
1484 out_close_devices:
1485 xfs_close_devices(mp);
1486 out_free_fsname:
1487 xfs_free_fsname(mp);
1488 kfree(mp);
1489 out:
1490 return -error;
1491
1492 out_unmount:
1493 xfs_filestream_unmount(mp);
1494 xfs_unmountfs(mp);
1495 goto out_free_sb;
1496}
1497
1498STATIC struct dentry *
1499xfs_fs_mount(
1500 struct file_system_type *fs_type,
1501 int flags,
1502 const char *dev_name,
1503 void *data)
1504{
1505 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1506}
1507
1508static int
1509xfs_fs_nr_cached_objects(
1510 struct super_block *sb)
1511{
1512 return xfs_reclaim_inodes_count(XFS_M(sb));
1513}
1514
1515static void
1516xfs_fs_free_cached_objects(
1517 struct super_block *sb,
1518 int nr_to_scan)
1519{
1520 xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1521}
1522
1523static const struct super_operations xfs_super_operations = {
1524 .alloc_inode = xfs_fs_alloc_inode,
1525 .destroy_inode = xfs_fs_destroy_inode,
1526 .evict_inode = xfs_fs_evict_inode,
1527 .drop_inode = xfs_fs_drop_inode,
1528 .put_super = xfs_fs_put_super,
1529 .sync_fs = xfs_fs_sync_fs,
1530 .freeze_fs = xfs_fs_freeze,
1531 .unfreeze_fs = xfs_fs_unfreeze,
1532 .statfs = xfs_fs_statfs,
1533 .remount_fs = xfs_fs_remount,
1534 .show_options = xfs_fs_show_options,
1535 .nr_cached_objects = xfs_fs_nr_cached_objects,
1536 .free_cached_objects = xfs_fs_free_cached_objects,
1537};
1538
1539static struct file_system_type xfs_fs_type = {
1540 .owner = THIS_MODULE,
1541 .name = "xfs",
1542 .mount = xfs_fs_mount,
1543 .kill_sb = kill_block_super,
1544 .fs_flags = FS_REQUIRES_DEV,
1545};
1546
1547STATIC int __init
1548xfs_init_zones(void)
1549{
1550
1551 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1552 if (!xfs_ioend_zone)
1553 goto out;
1554
1555 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1556 xfs_ioend_zone);
1557 if (!xfs_ioend_pool)
1558 goto out_destroy_ioend_zone;
1559
1560 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1561 "xfs_log_ticket");
1562 if (!xfs_log_ticket_zone)
1563 goto out_destroy_ioend_pool;
1564
1565 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1566 "xfs_bmap_free_item");
1567 if (!xfs_bmap_free_item_zone)
1568 goto out_destroy_log_ticket_zone;
1569
1570 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1571 "xfs_btree_cur");
1572 if (!xfs_btree_cur_zone)
1573 goto out_destroy_bmap_free_item_zone;
1574
1575 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1576 "xfs_da_state");
1577 if (!xfs_da_state_zone)
1578 goto out_destroy_btree_cur_zone;
1579
1580 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1581 if (!xfs_ifork_zone)
1582 goto out_destroy_da_state_zone;
1583
1584 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1585 if (!xfs_trans_zone)
1586 goto out_destroy_ifork_zone;
1587
1588 xfs_log_item_desc_zone =
1589 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1590 "xfs_log_item_desc");
1591 if (!xfs_log_item_desc_zone)
1592 goto out_destroy_trans_zone;
1593
1594 /*
1595 * The size of the zone allocated buf log item is the maximum
1596 * size possible under XFS. This wastes a little bit of memory,
1597 * but it is much faster.
1598 */
1599 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1600 "xfs_buf_item");
1601 if (!xfs_buf_item_zone)
1602 goto out_destroy_log_item_desc_zone;
1603
1604 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1605 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1606 sizeof(xfs_extent_t))), "xfs_efd_item");
1607 if (!xfs_efd_zone)
1608 goto out_destroy_buf_item_zone;
1609
1610 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1611 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1612 sizeof(xfs_extent_t))), "xfs_efi_item");
1613 if (!xfs_efi_zone)
1614 goto out_destroy_efd_zone;
1615
1616 xfs_inode_zone =
1617 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1618 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1619 xfs_fs_inode_init_once);
1620 if (!xfs_inode_zone)
1621 goto out_destroy_efi_zone;
1622
1623 xfs_ili_zone =
1624 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1625 KM_ZONE_SPREAD, NULL);
1626 if (!xfs_ili_zone)
1627 goto out_destroy_inode_zone;
1628
1629 return 0;
1630
1631 out_destroy_inode_zone:
1632 kmem_zone_destroy(xfs_inode_zone);
1633 out_destroy_efi_zone:
1634 kmem_zone_destroy(xfs_efi_zone);
1635 out_destroy_efd_zone:
1636 kmem_zone_destroy(xfs_efd_zone);
1637 out_destroy_buf_item_zone:
1638 kmem_zone_destroy(xfs_buf_item_zone);
1639 out_destroy_log_item_desc_zone:
1640 kmem_zone_destroy(xfs_log_item_desc_zone);
1641 out_destroy_trans_zone:
1642 kmem_zone_destroy(xfs_trans_zone);
1643 out_destroy_ifork_zone:
1644 kmem_zone_destroy(xfs_ifork_zone);
1645 out_destroy_da_state_zone:
1646 kmem_zone_destroy(xfs_da_state_zone);
1647 out_destroy_btree_cur_zone:
1648 kmem_zone_destroy(xfs_btree_cur_zone);
1649 out_destroy_bmap_free_item_zone:
1650 kmem_zone_destroy(xfs_bmap_free_item_zone);
1651 out_destroy_log_ticket_zone:
1652 kmem_zone_destroy(xfs_log_ticket_zone);
1653 out_destroy_ioend_pool:
1654 mempool_destroy(xfs_ioend_pool);
1655 out_destroy_ioend_zone:
1656 kmem_zone_destroy(xfs_ioend_zone);
1657 out:
1658 return -ENOMEM;
1659}
1660
1661STATIC void
1662xfs_destroy_zones(void)
1663{
1664 /*
1665 * Make sure all delayed rcu free are flushed before we
1666 * destroy caches.
1667 */
1668 rcu_barrier();
1669 kmem_zone_destroy(xfs_ili_zone);
1670 kmem_zone_destroy(xfs_inode_zone);
1671 kmem_zone_destroy(xfs_efi_zone);
1672 kmem_zone_destroy(xfs_efd_zone);
1673 kmem_zone_destroy(xfs_buf_item_zone);
1674 kmem_zone_destroy(xfs_log_item_desc_zone);
1675 kmem_zone_destroy(xfs_trans_zone);
1676 kmem_zone_destroy(xfs_ifork_zone);
1677 kmem_zone_destroy(xfs_da_state_zone);
1678 kmem_zone_destroy(xfs_btree_cur_zone);
1679 kmem_zone_destroy(xfs_bmap_free_item_zone);
1680 kmem_zone_destroy(xfs_log_ticket_zone);
1681 mempool_destroy(xfs_ioend_pool);
1682 kmem_zone_destroy(xfs_ioend_zone);
1683
1684}
1685
1686STATIC int __init
1687xfs_init_workqueues(void)
1688{
1689 /*
1690 * The allocation workqueue can be used in memory reclaim situations
1691 * (writepage path), and parallelism is only limited by the number of
1692 * AGs in all the filesystems mounted. Hence use the default large
1693 * max_active value for this workqueue.
1694 */
1695 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
1696 if (!xfs_alloc_wq)
1697 return -ENOMEM;
1698
1699 return 0;
1700}
1701
1702STATIC void
1703xfs_destroy_workqueues(void)
1704{
1705 destroy_workqueue(xfs_alloc_wq);
1706}
1707
1708STATIC int __init
1709init_xfs_fs(void)
1710{
1711 int error;
1712
1713 printk(KERN_INFO XFS_VERSION_STRING " with "
1714 XFS_BUILD_OPTIONS " enabled\n");
1715
1716 xfs_dir_startup();
1717
1718 error = xfs_init_zones();
1719 if (error)
1720 goto out;
1721
1722 error = xfs_init_workqueues();
1723 if (error)
1724 goto out_destroy_zones;
1725
1726 error = xfs_mru_cache_init();
1727 if (error)
1728 goto out_destroy_wq;
1729
1730 error = xfs_filestream_init();
1731 if (error)
1732 goto out_mru_cache_uninit;
1733
1734 error = xfs_buf_init();
1735 if (error)
1736 goto out_filestream_uninit;
1737
1738 error = xfs_init_procfs();
1739 if (error)
1740 goto out_buf_terminate;
1741
1742 error = xfs_sysctl_register();
1743 if (error)
1744 goto out_cleanup_procfs;
1745
1746 error = xfs_qm_init();
1747 if (error)
1748 goto out_sysctl_unregister;
1749
1750 error = register_filesystem(&xfs_fs_type);
1751 if (error)
1752 goto out_qm_exit;
1753 return 0;
1754
1755 out_qm_exit:
1756 xfs_qm_exit();
1757 out_sysctl_unregister:
1758 xfs_sysctl_unregister();
1759 out_cleanup_procfs:
1760 xfs_cleanup_procfs();
1761 out_buf_terminate:
1762 xfs_buf_terminate();
1763 out_filestream_uninit:
1764 xfs_filestream_uninit();
1765 out_mru_cache_uninit:
1766 xfs_mru_cache_uninit();
1767 out_destroy_wq:
1768 xfs_destroy_workqueues();
1769 out_destroy_zones:
1770 xfs_destroy_zones();
1771 out:
1772 return error;
1773}
1774
1775STATIC void __exit
1776exit_xfs_fs(void)
1777{
1778 xfs_qm_exit();
1779 unregister_filesystem(&xfs_fs_type);
1780 xfs_sysctl_unregister();
1781 xfs_cleanup_procfs();
1782 xfs_buf_terminate();
1783 xfs_filestream_uninit();
1784 xfs_mru_cache_uninit();
1785 xfs_destroy_workqueues();
1786 xfs_destroy_zones();
1787}
1788
1789module_init(init_xfs_fs);
1790module_exit(exit_xfs_fs);
1791
1792MODULE_AUTHOR("Silicon Graphics, Inc.");
1793MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1794MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.030206 seconds and 5 git commands to generate.