projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
xfs: remove usage of is_bad_inode
[linux-2.6-block.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_format.h"
21#include "xfs_log.h"
22#include "xfs_inum.h"
23#include "xfs_trans.h"
24#include "xfs_sb.h"
25#include "xfs_ag.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_log_priv.h"
44#include "xfs_trans_priv.h"
45#include "xfs_filestream.h"
46#include "xfs_da_btree.h"
47#include "xfs_dir2_format.h"
48#include "xfs_dir2.h"
49#include "xfs_extfree_item.h"
50#include "xfs_mru_cache.h"
51#include "xfs_inode_item.h"
52#include "xfs_icache.h"
53#include "xfs_trace.h"
54#include "xfs_icreate_item.h"
55
56#include <linux/namei.h>
57#include <linux/init.h>
58#include <linux/slab.h>
59#include <linux/mount.h>
60#include <linux/mempool.h>
61#include <linux/writeback.h>
62#include <linux/kthread.h>
63#include <linux/freezer.h>
64#include <linux/parser.h>
65
66static const struct super_operations xfs_super_operations;
67static kmem_zone_t *xfs_ioend_zone;
68mempool_t *xfs_ioend_pool;
69
70#define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
71#define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
72#define MNTOPT_LOGDEV "logdev" /* log device */
73#define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
74#define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
75#define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
76#define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
77#define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
78#define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
79#define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
80#define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
81#define MNTOPT_MTPT "mtpt" /* filesystem mount point */
82#define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
83#define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
84#define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
85#define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
86#define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
87#define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
88#define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
89 * unwritten extent conversion */
90#define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
91#define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
92#define MNTOPT_32BITINODE "inode32" /* inode allocation limited to
93 * XFS_MAXINUMBER_32 */
94#define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
95#define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
96#define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
97#define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
98 * in stat(). */
99#define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
100#define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
101#define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
102#define MNTOPT_QUOTA "quota" /* disk quotas (user) */
103#define MNTOPT_NOQUOTA "noquota" /* no quotas */
104#define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
105#define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
106#define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
107#define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
108#define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
109#define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
110#define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
111#define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
112#define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
113#define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
114#define MNTOPT_DELAYLOG "delaylog" /* Delayed logging enabled */
115#define MNTOPT_NODELAYLOG "nodelaylog" /* Delayed logging disabled */
116#define MNTOPT_DISCARD "discard" /* Discard unused blocks */
117#define MNTOPT_NODISCARD "nodiscard" /* Do not discard unused blocks */
118
119/*
120 * Table driven mount option parser.
121 *
122 * Currently only used for remount, but it will be used for mount
123 * in the future, too.
124 */
125enum {
126 Opt_barrier,
127 Opt_nobarrier,
128 Opt_inode64,
129 Opt_inode32,
130 Opt_err
131};
132
133static const match_table_t tokens = {
134 {Opt_barrier, "barrier"},
135 {Opt_nobarrier, "nobarrier"},
136 {Opt_inode64, "inode64"},
137 {Opt_inode32, "inode32"},
138 {Opt_err, NULL}
139};
140
141
142STATIC unsigned long
143suffix_kstrtoint(char *s, unsigned int base, int *res)
144{
145 int last, shift_left_factor = 0, _res;
146 char *value = s;
147
148 last = strlen(value) - 1;
149 if (value[last] == 'K' || value[last] == 'k') {
150 shift_left_factor = 10;
151 value[last] = '\0';
152 }
153 if (value[last] == 'M' || value[last] == 'm') {
154 shift_left_factor = 20;
155 value[last] = '\0';
156 }
157 if (value[last] == 'G' || value[last] == 'g') {
158 shift_left_factor = 30;
159 value[last] = '\0';
160 }
161
162 if (kstrtoint(s, base, &_res))
163 return -EINVAL;
164 *res = _res << shift_left_factor;
165 return 0;
166}
167
168/*
169 * This function fills in xfs_mount_t fields based on mount args.
170 * Note: the superblock has _not_ yet been read in.
171 *
172 * Note that this function leaks the various device name allocations on
173 * failure. The caller takes care of them.
174 */
175STATIC int
176xfs_parseargs(
177 struct xfs_mount *mp,
178 char *options)
179{
180 struct super_block *sb = mp->m_super;
181 char *this_char, *value;
182 int dsunit = 0;
183 int dswidth = 0;
184 int iosize = 0;
185 __uint8_t iosizelog = 0;
186
187 /*
188 * set up the mount name first so all the errors will refer to the
189 * correct device.
190 */
191 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
192 if (!mp->m_fsname)
193 return ENOMEM;
194 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
195
196 /*
197 * Copy binary VFS mount flags we are interested in.
198 */
199 if (sb->s_flags & MS_RDONLY)
200 mp->m_flags |= XFS_MOUNT_RDONLY;
201 if (sb->s_flags & MS_DIRSYNC)
202 mp->m_flags |= XFS_MOUNT_DIRSYNC;
203 if (sb->s_flags & MS_SYNCHRONOUS)
204 mp->m_flags |= XFS_MOUNT_WSYNC;
205
206 /*
207 * Set some default flags that could be cleared by the mount option
208 * parsing.
209 */
210 mp->m_flags |= XFS_MOUNT_BARRIER;
211 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
212#if !XFS_BIG_INUMS
213 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
214#endif
215
216 /*
217 * These can be overridden by the mount option parsing.
218 */
219 mp->m_logbufs = -1;
220 mp->m_logbsize = -1;
221
222 if (!options)
223 goto done;
224
225 while ((this_char = strsep(&options, ",")) != NULL) {
226 if (!*this_char)
227 continue;
228 if ((value = strchr(this_char, '=')) != NULL)
229 *value++ = 0;
230
231 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
232 if (!value || !*value) {
233 xfs_warn(mp, "%s option requires an argument",
234 this_char);
235 return EINVAL;
236 }
237 if (kstrtoint(value, 10, &mp->m_logbufs))
238 return EINVAL;
239 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
240 if (!value || !*value) {
241 xfs_warn(mp, "%s option requires an argument",
242 this_char);
243 return EINVAL;
244 }
245 if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
246 return EINVAL;
247 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
248 if (!value || !*value) {
249 xfs_warn(mp, "%s option requires an argument",
250 this_char);
251 return EINVAL;
252 }
253 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
254 if (!mp->m_logname)
255 return ENOMEM;
256 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
257 xfs_warn(mp, "%s option not allowed on this system",
258 this_char);
259 return EINVAL;
260 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
261 if (!value || !*value) {
262 xfs_warn(mp, "%s option requires an argument",
263 this_char);
264 return EINVAL;
265 }
266 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
267 if (!mp->m_rtname)
268 return ENOMEM;
269 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
270 if (!value || !*value) {
271 xfs_warn(mp, "%s option requires an argument",
272 this_char);
273 return EINVAL;
274 }
275 if (kstrtoint(value, 10, &iosize))
276 return EINVAL;
277 iosizelog = ffs(iosize) - 1;
278 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
279 if (!value || !*value) {
280 xfs_warn(mp, "%s option requires an argument",
281 this_char);
282 return EINVAL;
283 }
284 if (suffix_kstrtoint(value, 10, &iosize))
285 return EINVAL;
286 iosizelog = ffs(iosize) - 1;
287 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
288 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
289 mp->m_flags |= XFS_MOUNT_GRPID;
290 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
291 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
292 mp->m_flags &= ~XFS_MOUNT_GRPID;
293 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
294 mp->m_flags |= XFS_MOUNT_WSYNC;
295 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
296 mp->m_flags |= XFS_MOUNT_NORECOVERY;
297 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
298 mp->m_flags |= XFS_MOUNT_NOALIGN;
299 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
300 mp->m_flags |= XFS_MOUNT_SWALLOC;
301 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
302 if (!value || !*value) {
303 xfs_warn(mp, "%s option requires an argument",
304 this_char);
305 return EINVAL;
306 }
307 if (kstrtoint(value, 10, &dsunit))
308 return EINVAL;
309 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
310 if (!value || !*value) {
311 xfs_warn(mp, "%s option requires an argument",
312 this_char);
313 return EINVAL;
314 }
315 if (kstrtoint(value, 10, &dswidth))
316 return EINVAL;
317 } else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
318 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
319 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
320 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
321#if !XFS_BIG_INUMS
322 xfs_warn(mp, "%s option not allowed on this system",
323 this_char);
324 return EINVAL;
325#endif
326 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
327 mp->m_flags |= XFS_MOUNT_NOUUID;
328 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
329 mp->m_flags |= XFS_MOUNT_BARRIER;
330 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
331 mp->m_flags &= ~XFS_MOUNT_BARRIER;
332 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
333 mp->m_flags |= XFS_MOUNT_IKEEP;
334 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
335 mp->m_flags &= ~XFS_MOUNT_IKEEP;
336 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
337 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
338 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
339 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
340 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
341 mp->m_flags |= XFS_MOUNT_ATTR2;
342 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
343 mp->m_flags &= ~XFS_MOUNT_ATTR2;
344 mp->m_flags |= XFS_MOUNT_NOATTR2;
345 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
346 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
347 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
348 mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
349 mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
350 mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
351 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
352 !strcmp(this_char, MNTOPT_UQUOTA) ||
353 !strcmp(this_char, MNTOPT_USRQUOTA)) {
354 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
355 XFS_UQUOTA_ENFD);
356 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
357 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
358 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
359 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
360 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
361 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
362 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
363 XFS_PQUOTA_ENFD);
364 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
365 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
366 mp->m_qflags &= ~XFS_PQUOTA_ENFD;
367 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
368 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
369 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
370 XFS_GQUOTA_ENFD);
371 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
372 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
373 mp->m_qflags &= ~XFS_GQUOTA_ENFD;
374 } else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
375 xfs_warn(mp,
376 "delaylog is the default now, option is deprecated.");
377 } else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
378 xfs_warn(mp,
379 "nodelaylog support has been removed, option is deprecated.");
380 } else if (!strcmp(this_char, MNTOPT_DISCARD)) {
381 mp->m_flags |= XFS_MOUNT_DISCARD;
382 } else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
383 mp->m_flags &= ~XFS_MOUNT_DISCARD;
384 } else if (!strcmp(this_char, "ihashsize")) {
385 xfs_warn(mp,
386 "ihashsize no longer used, option is deprecated.");
387 } else if (!strcmp(this_char, "osyncisdsync")) {
388 xfs_warn(mp,
389 "osyncisdsync has no effect, option is deprecated.");
390 } else if (!strcmp(this_char, "osyncisosync")) {
391 xfs_warn(mp,
392 "osyncisosync has no effect, option is deprecated.");
393 } else if (!strcmp(this_char, "irixsgid")) {
394 xfs_warn(mp,
395 "irixsgid is now a sysctl(2) variable, option is deprecated.");
396 } else {
397 xfs_warn(mp, "unknown mount option [%s].", this_char);
398 return EINVAL;
399 }
400 }
401
402 /*
403 * no recovery flag requires a read-only mount
404 */
405 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
406 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
407 xfs_warn(mp, "no-recovery mounts must be read-only.");
408 return EINVAL;
409 }
410
411 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
412 xfs_warn(mp,
413 "sunit and swidth options incompatible with the noalign option");
414 return EINVAL;
415 }
416
417#ifndef CONFIG_XFS_QUOTA
418 if (XFS_IS_QUOTA_RUNNING(mp)) {
419 xfs_warn(mp, "quota support not available in this kernel.");
420 return EINVAL;
421 }
422#endif
423
424 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
425 xfs_warn(mp, "sunit and swidth must be specified together");
426 return EINVAL;
427 }
428
429 if (dsunit && (dswidth % dsunit != 0)) {
430 xfs_warn(mp,
431 "stripe width (%d) must be a multiple of the stripe unit (%d)",
432 dswidth, dsunit);
433 return EINVAL;
434 }
435
436done:
437 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
438 /*
439 * At this point the superblock has not been read
440 * in, therefore we do not know the block size.
441 * Before the mount call ends we will convert
442 * these to FSBs.
443 */
444 mp->m_dalign = dsunit;
445 mp->m_swidth = dswidth;
446 }
447
448 if (mp->m_logbufs != -1 &&
449 mp->m_logbufs != 0 &&
450 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
451 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
452 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
453 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
454 return XFS_ERROR(EINVAL);
455 }
456 if (mp->m_logbsize != -1 &&
457 mp->m_logbsize != 0 &&
458 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
459 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
460 !is_power_of_2(mp->m_logbsize))) {
461 xfs_warn(mp,
462 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
463 mp->m_logbsize);
464 return XFS_ERROR(EINVAL);
465 }
466
467 if (iosizelog) {
468 if (iosizelog > XFS_MAX_IO_LOG ||
469 iosizelog < XFS_MIN_IO_LOG) {
470 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
471 iosizelog, XFS_MIN_IO_LOG,
472 XFS_MAX_IO_LOG);
473 return XFS_ERROR(EINVAL);
474 }
475
476 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
477 mp->m_readio_log = iosizelog;
478 mp->m_writeio_log = iosizelog;
479 }
480
481 return 0;
482}
483
484struct proc_xfs_info {
485 int flag;
486 char *str;
487};
488
489STATIC int
490xfs_showargs(
491 struct xfs_mount *mp,
492 struct seq_file *m)
493{
494 static struct proc_xfs_info xfs_info_set[] = {
495 /* the few simple ones we can get from the mount struct */
496 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
497 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
498 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
499 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
500 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
501 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
502 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
503 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
504 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
505 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
506 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
507 { 0, NULL }
508 };
509 static struct proc_xfs_info xfs_info_unset[] = {
510 /* the few simple ones we can get from the mount struct */
511 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
512 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
513 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
514 { 0, NULL }
515 };
516 struct proc_xfs_info *xfs_infop;
517
518 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
519 if (mp->m_flags & xfs_infop->flag)
520 seq_puts(m, xfs_infop->str);
521 }
522 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
523 if (!(mp->m_flags & xfs_infop->flag))
524 seq_puts(m, xfs_infop->str);
525 }
526
527 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
528 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
529 (int)(1 << mp->m_writeio_log) >> 10);
530
531 if (mp->m_logbufs > 0)
532 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
533 if (mp->m_logbsize > 0)
534 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
535
536 if (mp->m_logname)
537 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
538 if (mp->m_rtname)
539 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
540
541 if (mp->m_dalign > 0)
542 seq_printf(m, "," MNTOPT_SUNIT "=%d",
543 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
544 if (mp->m_swidth > 0)
545 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
546 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
547
548 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
549 seq_puts(m, "," MNTOPT_USRQUOTA);
550 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
551 seq_puts(m, "," MNTOPT_UQUOTANOENF);
552
553 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
554 if (mp->m_qflags & XFS_PQUOTA_ENFD)
555 seq_puts(m, "," MNTOPT_PRJQUOTA);
556 else
557 seq_puts(m, "," MNTOPT_PQUOTANOENF);
558 }
559 if (mp->m_qflags & XFS_GQUOTA_ACCT) {
560 if (mp->m_qflags & XFS_GQUOTA_ENFD)
561 seq_puts(m, "," MNTOPT_GRPQUOTA);
562 else
563 seq_puts(m, "," MNTOPT_GQUOTANOENF);
564 }
565
566 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
567 seq_puts(m, "," MNTOPT_NOQUOTA);
568
569 return 0;
570}
571__uint64_t
572xfs_max_file_offset(
573 unsigned int blockshift)
574{
575 unsigned int pagefactor = 1;
576 unsigned int bitshift = BITS_PER_LONG - 1;
577
578 /* Figure out maximum filesize, on Linux this can depend on
579 * the filesystem blocksize (on 32 bit platforms).
580 * __block_write_begin does this in an [unsigned] long...
581 * page->index << (PAGE_CACHE_SHIFT - bbits)
582 * So, for page sized blocks (4K on 32 bit platforms),
583 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
584 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
585 * but for smaller blocksizes it is less (bbits = log2 bsize).
586 * Note1: get_block_t takes a long (implicit cast from above)
587 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
588 * can optionally convert the [unsigned] long from above into
589 * an [unsigned] long long.
590 */
591
592#if BITS_PER_LONG == 32
593# if defined(CONFIG_LBDAF)
594 ASSERT(sizeof(sector_t) == 8);
595 pagefactor = PAGE_CACHE_SIZE;
596 bitshift = BITS_PER_LONG;
597# else
598 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
599# endif
600#endif
601
602 return (((__uint64_t)pagefactor) << bitshift) - 1;
603}
604
605xfs_agnumber_t
606xfs_set_inode32(struct xfs_mount *mp)
607{
608 xfs_agnumber_t index = 0;
609 xfs_agnumber_t maxagi = 0;
610 xfs_sb_t *sbp = &mp->m_sb;
611 xfs_agnumber_t max_metadata;
612 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0);
613 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino);
614 xfs_perag_t *pag;
615
616 /* Calculate how much should be reserved for inodes to meet
617 * the max inode percentage.
618 */
619 if (mp->m_maxicount) {
620 __uint64_t icount;
621
622 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
623 do_div(icount, 100);
624 icount += sbp->sb_agblocks - 1;
625 do_div(icount, sbp->sb_agblocks);
626 max_metadata = icount;
627 } else {
628 max_metadata = sbp->sb_agcount;
629 }
630
631 for (index = 0; index < sbp->sb_agcount; index++) {
632 ino = XFS_AGINO_TO_INO(mp, index, agino);
633
634 if (ino > XFS_MAXINUMBER_32) {
635 pag = xfs_perag_get(mp, index);
636 pag->pagi_inodeok = 0;
637 pag->pagf_metadata = 0;
638 xfs_perag_put(pag);
639 continue;
640 }
641
642 pag = xfs_perag_get(mp, index);
643 pag->pagi_inodeok = 1;
644 maxagi++;
645 if (index < max_metadata)
646 pag->pagf_metadata = 1;
647 xfs_perag_put(pag);
648 }
649 mp->m_flags |= (XFS_MOUNT_32BITINODES |
650 XFS_MOUNT_SMALL_INUMS);
651
652 return maxagi;
653}
654
655xfs_agnumber_t
656xfs_set_inode64(struct xfs_mount *mp)
657{
658 xfs_agnumber_t index = 0;
659
660 for (index = 0; index < mp->m_sb.sb_agcount; index++) {
661 struct xfs_perag *pag;
662
663 pag = xfs_perag_get(mp, index);
664 pag->pagi_inodeok = 1;
665 pag->pagf_metadata = 0;
666 xfs_perag_put(pag);
667 }
668
669 /* There is no need for lock protection on m_flags,
670 * the rw_semaphore of the VFS superblock is locked
671 * during mount/umount/remount operations, so this is
672 * enough to avoid concurency on the m_flags field
673 */
674 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
675 XFS_MOUNT_SMALL_INUMS);
676 return index;
677}
678
679STATIC int
680xfs_blkdev_get(
681 xfs_mount_t *mp,
682 const char *name,
683 struct block_device **bdevp)
684{
685 int error = 0;
686
687 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
688 mp);
689 if (IS_ERR(*bdevp)) {
690 error = PTR_ERR(*bdevp);
691 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
692 }
693
694 return -error;
695}
696
697STATIC void
698xfs_blkdev_put(
699 struct block_device *bdev)
700{
701 if (bdev)
702 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
703}
704
705void
706xfs_blkdev_issue_flush(
707 xfs_buftarg_t *buftarg)
708{
709 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
710}
711
712STATIC void
713xfs_close_devices(
714 struct xfs_mount *mp)
715{
716 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
717 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
718 xfs_free_buftarg(mp, mp->m_logdev_targp);
719 xfs_blkdev_put(logdev);
720 }
721 if (mp->m_rtdev_targp) {
722 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
723 xfs_free_buftarg(mp, mp->m_rtdev_targp);
724 xfs_blkdev_put(rtdev);
725 }
726 xfs_free_buftarg(mp, mp->m_ddev_targp);
727}
728
729/*
730 * The file system configurations are:
731 * (1) device (partition) with data and internal log
732 * (2) logical volume with data and log subvolumes.
733 * (3) logical volume with data, log, and realtime subvolumes.
734 *
735 * We only have to handle opening the log and realtime volumes here if
736 * they are present. The data subvolume has already been opened by
737 * get_sb_bdev() and is stored in sb->s_bdev.
738 */
739STATIC int
740xfs_open_devices(
741 struct xfs_mount *mp)
742{
743 struct block_device *ddev = mp->m_super->s_bdev;
744 struct block_device *logdev = NULL, *rtdev = NULL;
745 int error;
746
747 /*
748 * Open real time and log devices - order is important.
749 */
750 if (mp->m_logname) {
751 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
752 if (error)
753 goto out;
754 }
755
756 if (mp->m_rtname) {
757 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
758 if (error)
759 goto out_close_logdev;
760
761 if (rtdev == ddev || rtdev == logdev) {
762 xfs_warn(mp,
763 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
764 error = EINVAL;
765 goto out_close_rtdev;
766 }
767 }
768
769 /*
770 * Setup xfs_mount buffer target pointers
771 */
772 error = ENOMEM;
773 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
774 if (!mp->m_ddev_targp)
775 goto out_close_rtdev;
776
777 if (rtdev) {
778 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
779 mp->m_fsname);
780 if (!mp->m_rtdev_targp)
781 goto out_free_ddev_targ;
782 }
783
784 if (logdev && logdev != ddev) {
785 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
786 mp->m_fsname);
787 if (!mp->m_logdev_targp)
788 goto out_free_rtdev_targ;
789 } else {
790 mp->m_logdev_targp = mp->m_ddev_targp;
791 }
792
793 return 0;
794
795 out_free_rtdev_targ:
796 if (mp->m_rtdev_targp)
797 xfs_free_buftarg(mp, mp->m_rtdev_targp);
798 out_free_ddev_targ:
799 xfs_free_buftarg(mp, mp->m_ddev_targp);
800 out_close_rtdev:
801 if (rtdev)
802 xfs_blkdev_put(rtdev);
803 out_close_logdev:
804 if (logdev && logdev != ddev)
805 xfs_blkdev_put(logdev);
806 out:
807 return error;
808}
809
810/*
811 * Setup xfs_mount buffer target pointers based on superblock
812 */
813STATIC int
814xfs_setup_devices(
815 struct xfs_mount *mp)
816{
817 int error;
818
819 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
820 mp->m_sb.sb_sectsize);
821 if (error)
822 return error;
823
824 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
825 unsigned int log_sector_size = BBSIZE;
826
827 if (xfs_sb_version_hassector(&mp->m_sb))
828 log_sector_size = mp->m_sb.sb_logsectsize;
829 error = xfs_setsize_buftarg(mp->m_logdev_targp,
830 mp->m_sb.sb_blocksize,
831 log_sector_size);
832 if (error)
833 return error;
834 }
835 if (mp->m_rtdev_targp) {
836 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
837 mp->m_sb.sb_blocksize,
838 mp->m_sb.sb_sectsize);
839 if (error)
840 return error;
841 }
842
843 return 0;
844}
845
846STATIC int
847xfs_init_mount_workqueues(
848 struct xfs_mount *mp)
849{
850 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
851 WQ_MEM_RECLAIM, 0, mp->m_fsname);
852 if (!mp->m_data_workqueue)
853 goto out;
854
855 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
856 WQ_MEM_RECLAIM, 0, mp->m_fsname);
857 if (!mp->m_unwritten_workqueue)
858 goto out_destroy_data_iodone_queue;
859
860 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
861 WQ_MEM_RECLAIM, 0, mp->m_fsname);
862 if (!mp->m_cil_workqueue)
863 goto out_destroy_unwritten;
864
865 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
866 0, 0, mp->m_fsname);
867 if (!mp->m_reclaim_workqueue)
868 goto out_destroy_cil;
869
870 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
871 0, 0, mp->m_fsname);
872 if (!mp->m_log_workqueue)
873 goto out_destroy_reclaim;
874
875 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
876 0, 0, mp->m_fsname);
877 if (!mp->m_eofblocks_workqueue)
878 goto out_destroy_log;
879
880 return 0;
881
882out_destroy_log:
883 destroy_workqueue(mp->m_log_workqueue);
884out_destroy_reclaim:
885 destroy_workqueue(mp->m_reclaim_workqueue);
886out_destroy_cil:
887 destroy_workqueue(mp->m_cil_workqueue);
888out_destroy_unwritten:
889 destroy_workqueue(mp->m_unwritten_workqueue);
890out_destroy_data_iodone_queue:
891 destroy_workqueue(mp->m_data_workqueue);
892out:
893 return -ENOMEM;
894}
895
896STATIC void
897xfs_destroy_mount_workqueues(
898 struct xfs_mount *mp)
899{
900 destroy_workqueue(mp->m_eofblocks_workqueue);
901 destroy_workqueue(mp->m_log_workqueue);
902 destroy_workqueue(mp->m_reclaim_workqueue);
903 destroy_workqueue(mp->m_cil_workqueue);
904 destroy_workqueue(mp->m_data_workqueue);
905 destroy_workqueue(mp->m_unwritten_workqueue);
906}
907
908/*
909 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
910 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
911 * for IO to complete so that we effectively throttle multiple callers to the
912 * rate at which IO is completing.
913 */
914void
915xfs_flush_inodes(
916 struct xfs_mount *mp)
917{
918 struct super_block *sb = mp->m_super;
919
920 if (down_read_trylock(&sb->s_umount)) {
921 sync_inodes_sb(sb);
922 up_read(&sb->s_umount);
923 }
924}
925
926/* Catch misguided souls that try to use this interface on XFS */
927STATIC struct inode *
928xfs_fs_alloc_inode(
929 struct super_block *sb)
930{
931 BUG();
932 return NULL;
933}
934
935/*
936 * Now that the generic code is guaranteed not to be accessing
937 * the linux inode, we can reclaim the inode.
938 */
939STATIC void
940xfs_fs_destroy_inode(
941 struct inode *inode)
942{
943 struct xfs_inode *ip = XFS_I(inode);
944
945 trace_xfs_destroy_inode(ip);
946
947 XFS_STATS_INC(vn_reclaim);
948
949 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
950
951 /*
952 * We should never get here with one of the reclaim flags already set.
953 */
954 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
955 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
956
957 /*
958 * We always use background reclaim here because even if the
959 * inode is clean, it still may be under IO and hence we have
960 * to take the flush lock. The background reclaim path handles
961 * this more efficiently than we can here, so simply let background
962 * reclaim tear down all inodes.
963 */
964 xfs_inode_set_reclaim_tag(ip);
965}
966
967/*
968 * Slab object creation initialisation for the XFS inode.
969 * This covers only the idempotent fields in the XFS inode;
970 * all other fields need to be initialised on allocation
971 * from the slab. This avoids the need to repeatedly initialise
972 * fields in the xfs inode that left in the initialise state
973 * when freeing the inode.
974 */
975STATIC void
976xfs_fs_inode_init_once(
977 void *inode)
978{
979 struct xfs_inode *ip = inode;
980
981 memset(ip, 0, sizeof(struct xfs_inode));
982
983 /* vfs inode */
984 inode_init_once(VFS_I(ip));
985
986 /* xfs inode */
987 atomic_set(&ip->i_pincount, 0);
988 spin_lock_init(&ip->i_flags_lock);
989
990 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
991 "xfsino", ip->i_ino);
992}
993
994STATIC void
995xfs_fs_evict_inode(
996 struct inode *inode)
997{
998 xfs_inode_t *ip = XFS_I(inode);
999
1000 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1001
1002 trace_xfs_evict_inode(ip);
1003
1004 truncate_inode_pages(&inode->i_data, 0);
1005 clear_inode(inode);
1006 XFS_STATS_INC(vn_rele);
1007 XFS_STATS_INC(vn_remove);
1008 XFS_STATS_DEC(vn_active);
1009
1010 xfs_inactive(ip);
1011}
1012
1013/*
1014 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1015 * serialised against cache hits here via the inode->i_lock and igrab() in
1016 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1017 * racing with us, and it avoids needing to grab a spinlock here for every inode
1018 * we drop the final reference on.
1019 */
1020STATIC int
1021xfs_fs_drop_inode(
1022 struct inode *inode)
1023{
1024 struct xfs_inode *ip = XFS_I(inode);
1025
1026 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1027}
1028
1029STATIC void
1030xfs_free_fsname(
1031 struct xfs_mount *mp)
1032{
1033 kfree(mp->m_fsname);
1034 kfree(mp->m_rtname);
1035 kfree(mp->m_logname);
1036}
1037
1038STATIC void
1039xfs_fs_put_super(
1040 struct super_block *sb)
1041{
1042 struct xfs_mount *mp = XFS_M(sb);
1043
1044 xfs_filestream_unmount(mp);
1045 xfs_unmountfs(mp);
1046
1047 xfs_freesb(mp);
1048 xfs_icsb_destroy_counters(mp);
1049 xfs_destroy_mount_workqueues(mp);
1050 xfs_close_devices(mp);
1051 xfs_free_fsname(mp);
1052 kfree(mp);
1053}
1054
1055STATIC int
1056xfs_fs_sync_fs(
1057 struct super_block *sb,
1058 int wait)
1059{
1060 struct xfs_mount *mp = XFS_M(sb);
1061
1062 /*
1063 * Doing anything during the async pass would be counterproductive.
1064 */
1065 if (!wait)
1066 return 0;
1067
1068 xfs_log_force(mp, XFS_LOG_SYNC);
1069 if (laptop_mode) {
1070 /*
1071 * The disk must be active because we're syncing.
1072 * We schedule log work now (now that the disk is
1073 * active) instead of later (when it might not be).
1074 */
1075 flush_delayed_work(&mp->m_log->l_work);
1076 }
1077
1078 return 0;
1079}
1080
1081STATIC int
1082xfs_fs_statfs(
1083 struct dentry *dentry,
1084 struct kstatfs *statp)
1085{
1086 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1087 xfs_sb_t *sbp = &mp->m_sb;
1088 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1089 __uint64_t fakeinos, id;
1090 xfs_extlen_t lsize;
1091 __int64_t ffree;
1092
1093 statp->f_type = XFS_SB_MAGIC;
1094 statp->f_namelen = MAXNAMELEN - 1;
1095
1096 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1097 statp->f_fsid.val[0] = (u32)id;
1098 statp->f_fsid.val[1] = (u32)(id >> 32);
1099
1100 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1101
1102 spin_lock(&mp->m_sb_lock);
1103 statp->f_bsize = sbp->sb_blocksize;
1104 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1105 statp->f_blocks = sbp->sb_dblocks - lsize;
1106 statp->f_bfree = statp->f_bavail =
1107 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1108 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1109 statp->f_files =
1110 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1111 if (mp->m_maxicount)
1112 statp->f_files = min_t(typeof(statp->f_files),
1113 statp->f_files,
1114 mp->m_maxicount);
1115
1116 /* make sure statp->f_ffree does not underflow */
1117 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1118 statp->f_ffree = max_t(__int64_t, ffree, 0);
1119
1120 spin_unlock(&mp->m_sb_lock);
1121
1122 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1123 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1124 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1125 xfs_qm_statvfs(ip, statp);
1126 return 0;
1127}
1128
1129STATIC void
1130xfs_save_resvblks(struct xfs_mount *mp)
1131{
1132 __uint64_t resblks = 0;
1133
1134 mp->m_resblks_save = mp->m_resblks;
1135 xfs_reserve_blocks(mp, &resblks, NULL);
1136}
1137
1138STATIC void
1139xfs_restore_resvblks(struct xfs_mount *mp)
1140{
1141 __uint64_t resblks;
1142
1143 if (mp->m_resblks_save) {
1144 resblks = mp->m_resblks_save;
1145 mp->m_resblks_save = 0;
1146 } else
1147 resblks = xfs_default_resblks(mp);
1148
1149 xfs_reserve_blocks(mp, &resblks, NULL);
1150}
1151
1152/*
1153 * Trigger writeback of all the dirty metadata in the file system.
1154 *
1155 * This ensures that the metadata is written to their location on disk rather
1156 * than just existing in transactions in the log. This means after a quiesce
1157 * there is no log replay required to write the inodes to disk - this is the
1158 * primary difference between a sync and a quiesce.
1159 *
1160 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1161 * it is started again when appropriate.
1162 */
1163void
1164xfs_quiesce_attr(
1165 struct xfs_mount *mp)
1166{
1167 int error = 0;
1168
1169 /* wait for all modifications to complete */
1170 while (atomic_read(&mp->m_active_trans) > 0)
1171 delay(100);
1172
1173 /* force the log to unpin objects from the now complete transactions */
1174 xfs_log_force(mp, XFS_LOG_SYNC);
1175
1176 /* reclaim inodes to do any IO before the freeze completes */
1177 xfs_reclaim_inodes(mp, 0);
1178 xfs_reclaim_inodes(mp, SYNC_WAIT);
1179
1180 /* Push the superblock and write an unmount record */
1181 error = xfs_log_sbcount(mp);
1182 if (error)
1183 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1184 "Frozen image may not be consistent.");
1185 /*
1186 * Just warn here till VFS can correctly support
1187 * read-only remount without racing.
1188 */
1189 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1190
1191 xfs_log_quiesce(mp);
1192}
1193
1194STATIC int
1195xfs_fs_remount(
1196 struct super_block *sb,
1197 int *flags,
1198 char *options)
1199{
1200 struct xfs_mount *mp = XFS_M(sb);
1201 substring_t args[MAX_OPT_ARGS];
1202 char *p;
1203 int error;
1204
1205 while ((p = strsep(&options, ",")) != NULL) {
1206 int token;
1207
1208 if (!*p)
1209 continue;
1210
1211 token = match_token(p, tokens, args);
1212 switch (token) {
1213 case Opt_barrier:
1214 mp->m_flags |= XFS_MOUNT_BARRIER;
1215 break;
1216 case Opt_nobarrier:
1217 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1218 break;
1219 case Opt_inode64:
1220 mp->m_maxagi = xfs_set_inode64(mp);
1221 break;
1222 case Opt_inode32:
1223 mp->m_maxagi = xfs_set_inode32(mp);
1224 break;
1225 default:
1226 /*
1227 * Logically we would return an error here to prevent
1228 * users from believing they might have changed
1229 * mount options using remount which can't be changed.
1230 *
1231 * But unfortunately mount(8) adds all options from
1232 * mtab and fstab to the mount arguments in some cases
1233 * so we can't blindly reject options, but have to
1234 * check for each specified option if it actually
1235 * differs from the currently set option and only
1236 * reject it if that's the case.
1237 *
1238 * Until that is implemented we return success for
1239 * every remount request, and silently ignore all
1240 * options that we can't actually change.
1241 */
1242#if 0
1243 xfs_info(mp,
1244 "mount option \"%s\" not supported for remount\n", p);
1245 return -EINVAL;
1246#else
1247 break;
1248#endif
1249 }
1250 }
1251
1252 /* ro -> rw */
1253 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1254 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1255
1256 /*
1257 * If this is the first remount to writeable state we
1258 * might have some superblock changes to update.
1259 */
1260 if (mp->m_update_flags) {
1261 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1262 if (error) {
1263 xfs_warn(mp, "failed to write sb changes");
1264 return error;
1265 }
1266 mp->m_update_flags = 0;
1267 }
1268
1269 /*
1270 * Fill out the reserve pool if it is empty. Use the stashed
1271 * value if it is non-zero, otherwise go with the default.
1272 */
1273 xfs_restore_resvblks(mp);
1274 xfs_log_work_queue(mp);
1275 }
1276
1277 /* rw -> ro */
1278 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1279 /*
1280 * Before we sync the metadata, we need to free up the reserve
1281 * block pool so that the used block count in the superblock on
1282 * disk is correct at the end of the remount. Stash the current
1283 * reserve pool size so that if we get remounted rw, we can
1284 * return it to the same size.
1285 */
1286 xfs_save_resvblks(mp);
1287 xfs_quiesce_attr(mp);
1288 mp->m_flags |= XFS_MOUNT_RDONLY;
1289 }
1290
1291 return 0;
1292}
1293
1294/*
1295 * Second stage of a freeze. The data is already frozen so we only
1296 * need to take care of the metadata. Once that's done write a dummy
1297 * record to dirty the log in case of a crash while frozen.
1298 */
1299STATIC int
1300xfs_fs_freeze(
1301 struct super_block *sb)
1302{
1303 struct xfs_mount *mp = XFS_M(sb);
1304
1305 xfs_save_resvblks(mp);
1306 xfs_quiesce_attr(mp);
1307 return -xfs_fs_log_dummy(mp);
1308}
1309
1310STATIC int
1311xfs_fs_unfreeze(
1312 struct super_block *sb)
1313{
1314 struct xfs_mount *mp = XFS_M(sb);
1315
1316 xfs_restore_resvblks(mp);
1317 xfs_log_work_queue(mp);
1318 return 0;
1319}
1320
1321STATIC int
1322xfs_fs_show_options(
1323 struct seq_file *m,
1324 struct dentry *root)
1325{
1326 return -xfs_showargs(XFS_M(root->d_sb), m);
1327}
1328
1329/*
1330 * This function fills in xfs_mount_t fields based on mount args.
1331 * Note: the superblock _has_ now been read in.
1332 */
1333STATIC int
1334xfs_finish_flags(
1335 struct xfs_mount *mp)
1336{
1337 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1338
1339 /* Fail a mount where the logbuf is smaller than the log stripe */
1340 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1341 if (mp->m_logbsize <= 0 &&
1342 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1343 mp->m_logbsize = mp->m_sb.sb_logsunit;
1344 } else if (mp->m_logbsize > 0 &&
1345 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1346 xfs_warn(mp,
1347 "logbuf size must be greater than or equal to log stripe size");
1348 return XFS_ERROR(EINVAL);
1349 }
1350 } else {
1351 /* Fail a mount if the logbuf is larger than 32K */
1352 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1353 xfs_warn(mp,
1354 "logbuf size for version 1 logs must be 16K or 32K");
1355 return XFS_ERROR(EINVAL);
1356 }
1357 }
1358
1359 /*
1360 * V5 filesystems always use attr2 format for attributes.
1361 */
1362 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1363 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1364 xfs_warn(mp,
1365"Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1366 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1367 return XFS_ERROR(EINVAL);
1368 }
1369
1370 /*
1371 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1372 * told by noattr2 to turn it off
1373 */
1374 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1375 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1376 mp->m_flags |= XFS_MOUNT_ATTR2;
1377
1378 /*
1379 * prohibit r/w mounts of read-only filesystems
1380 */
1381 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1382 xfs_warn(mp,
1383 "cannot mount a read-only filesystem as read-write");
1384 return XFS_ERROR(EROFS);
1385 }
1386
1387 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1388 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1389 !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1390 xfs_warn(mp,
1391 "Super block does not support project and group quota together");
1392 return XFS_ERROR(EINVAL);
1393 }
1394
1395 return 0;
1396}
1397
1398STATIC int
1399xfs_fs_fill_super(
1400 struct super_block *sb,
1401 void *data,
1402 int silent)
1403{
1404 struct inode *root;
1405 struct xfs_mount *mp = NULL;
1406 int flags = 0, error = ENOMEM;
1407
1408 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1409 if (!mp)
1410 goto out;
1411
1412 spin_lock_init(&mp->m_sb_lock);
1413 mutex_init(&mp->m_growlock);
1414 atomic_set(&mp->m_active_trans, 0);
1415 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1416 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1417
1418 mp->m_super = sb;
1419 sb->s_fs_info = mp;
1420
1421 error = xfs_parseargs(mp, (char *)data);
1422 if (error)
1423 goto out_free_fsname;
1424
1425 sb_min_blocksize(sb, BBSIZE);
1426 sb->s_xattr = xfs_xattr_handlers;
1427 sb->s_export_op = &xfs_export_operations;
1428#ifdef CONFIG_XFS_QUOTA
1429 sb->s_qcop = &xfs_quotactl_operations;
1430#endif
1431 sb->s_op = &xfs_super_operations;
1432
1433 if (silent)
1434 flags |= XFS_MFSI_QUIET;
1435
1436 error = xfs_open_devices(mp);
1437 if (error)
1438 goto out_free_fsname;
1439
1440 error = xfs_init_mount_workqueues(mp);
1441 if (error)
1442 goto out_close_devices;
1443
1444 error = xfs_icsb_init_counters(mp);
1445 if (error)
1446 goto out_destroy_workqueues;
1447
1448 error = xfs_readsb(mp, flags);
1449 if (error)
1450 goto out_destroy_counters;
1451
1452 error = xfs_finish_flags(mp);
1453 if (error)
1454 goto out_free_sb;
1455
1456 error = xfs_setup_devices(mp);
1457 if (error)
1458 goto out_free_sb;
1459
1460 error = xfs_filestream_mount(mp);
1461 if (error)
1462 goto out_free_sb;
1463
1464 /*
1465 * we must configure the block size in the superblock before we run the
1466 * full mount process as the mount process can lookup and cache inodes.
1467 */
1468 sb->s_magic = XFS_SB_MAGIC;
1469 sb->s_blocksize = mp->m_sb.sb_blocksize;
1470 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1471 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1472 sb->s_max_links = XFS_MAXLINK;
1473 sb->s_time_gran = 1;
1474 set_posix_acl_flag(sb);
1475
1476 /* version 5 superblocks support inode version counters. */
1477 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1478 sb->s_flags |= MS_I_VERSION;
1479
1480 error = xfs_mountfs(mp);
1481 if (error)
1482 goto out_filestream_unmount;
1483
1484 root = igrab(VFS_I(mp->m_rootip));
1485 if (!root) {
1486 error = ENOENT;
1487 goto out_unmount;
1488 }
1489 sb->s_root = d_make_root(root);
1490 if (!sb->s_root) {
1491 error = ENOMEM;
1492 goto out_unmount;
1493 }
1494
1495 return 0;
1496
1497 out_filestream_unmount:
1498 xfs_filestream_unmount(mp);
1499 out_free_sb:
1500 xfs_freesb(mp);
1501 out_destroy_counters:
1502 xfs_icsb_destroy_counters(mp);
1503out_destroy_workqueues:
1504 xfs_destroy_mount_workqueues(mp);
1505 out_close_devices:
1506 xfs_close_devices(mp);
1507 out_free_fsname:
1508 xfs_free_fsname(mp);
1509 kfree(mp);
1510 out:
1511 return -error;
1512
1513 out_unmount:
1514 xfs_filestream_unmount(mp);
1515 xfs_unmountfs(mp);
1516 goto out_free_sb;
1517}
1518
1519STATIC struct dentry *
1520xfs_fs_mount(
1521 struct file_system_type *fs_type,
1522 int flags,
1523 const char *dev_name,
1524 void *data)
1525{
1526 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1527}
1528
1529static long
1530xfs_fs_nr_cached_objects(
1531 struct super_block *sb,
1532 int nid)
1533{
1534 return xfs_reclaim_inodes_count(XFS_M(sb));
1535}
1536
1537static long
1538xfs_fs_free_cached_objects(
1539 struct super_block *sb,
1540 long nr_to_scan,
1541 int nid)
1542{
1543 return xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1544}
1545
1546static const struct super_operations xfs_super_operations = {
1547 .alloc_inode = xfs_fs_alloc_inode,
1548 .destroy_inode = xfs_fs_destroy_inode,
1549 .evict_inode = xfs_fs_evict_inode,
1550 .drop_inode = xfs_fs_drop_inode,
1551 .put_super = xfs_fs_put_super,
1552 .sync_fs = xfs_fs_sync_fs,
1553 .freeze_fs = xfs_fs_freeze,
1554 .unfreeze_fs = xfs_fs_unfreeze,
1555 .statfs = xfs_fs_statfs,
1556 .remount_fs = xfs_fs_remount,
1557 .show_options = xfs_fs_show_options,
1558 .nr_cached_objects = xfs_fs_nr_cached_objects,
1559 .free_cached_objects = xfs_fs_free_cached_objects,
1560};
1561
1562static struct file_system_type xfs_fs_type = {
1563 .owner = THIS_MODULE,
1564 .name = "xfs",
1565 .mount = xfs_fs_mount,
1566 .kill_sb = kill_block_super,
1567 .fs_flags = FS_REQUIRES_DEV,
1568};
1569MODULE_ALIAS_FS("xfs");
1570
1571STATIC int __init
1572xfs_init_zones(void)
1573{
1574
1575 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1576 if (!xfs_ioend_zone)
1577 goto out;
1578
1579 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1580 xfs_ioend_zone);
1581 if (!xfs_ioend_pool)
1582 goto out_destroy_ioend_zone;
1583
1584 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1585 "xfs_log_ticket");
1586 if (!xfs_log_ticket_zone)
1587 goto out_destroy_ioend_pool;
1588
1589 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1590 "xfs_bmap_free_item");
1591 if (!xfs_bmap_free_item_zone)
1592 goto out_destroy_log_ticket_zone;
1593
1594 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1595 "xfs_btree_cur");
1596 if (!xfs_btree_cur_zone)
1597 goto out_destroy_bmap_free_item_zone;
1598
1599 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1600 "xfs_da_state");
1601 if (!xfs_da_state_zone)
1602 goto out_destroy_btree_cur_zone;
1603
1604 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1605 if (!xfs_ifork_zone)
1606 goto out_destroy_da_state_zone;
1607
1608 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1609 if (!xfs_trans_zone)
1610 goto out_destroy_ifork_zone;
1611
1612 xfs_log_item_desc_zone =
1613 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1614 "xfs_log_item_desc");
1615 if (!xfs_log_item_desc_zone)
1616 goto out_destroy_trans_zone;
1617
1618 /*
1619 * The size of the zone allocated buf log item is the maximum
1620 * size possible under XFS. This wastes a little bit of memory,
1621 * but it is much faster.
1622 */
1623 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1624 "xfs_buf_item");
1625 if (!xfs_buf_item_zone)
1626 goto out_destroy_log_item_desc_zone;
1627
1628 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1629 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1630 sizeof(xfs_extent_t))), "xfs_efd_item");
1631 if (!xfs_efd_zone)
1632 goto out_destroy_buf_item_zone;
1633
1634 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1635 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1636 sizeof(xfs_extent_t))), "xfs_efi_item");
1637 if (!xfs_efi_zone)
1638 goto out_destroy_efd_zone;
1639
1640 xfs_inode_zone =
1641 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1642 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1643 xfs_fs_inode_init_once);
1644 if (!xfs_inode_zone)
1645 goto out_destroy_efi_zone;
1646
1647 xfs_ili_zone =
1648 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1649 KM_ZONE_SPREAD, NULL);
1650 if (!xfs_ili_zone)
1651 goto out_destroy_inode_zone;
1652 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1653 "xfs_icr");
1654 if (!xfs_icreate_zone)
1655 goto out_destroy_ili_zone;
1656
1657 return 0;
1658
1659 out_destroy_ili_zone:
1660 kmem_zone_destroy(xfs_ili_zone);
1661 out_destroy_inode_zone:
1662 kmem_zone_destroy(xfs_inode_zone);
1663 out_destroy_efi_zone:
1664 kmem_zone_destroy(xfs_efi_zone);
1665 out_destroy_efd_zone:
1666 kmem_zone_destroy(xfs_efd_zone);
1667 out_destroy_buf_item_zone:
1668 kmem_zone_destroy(xfs_buf_item_zone);
1669 out_destroy_log_item_desc_zone:
1670 kmem_zone_destroy(xfs_log_item_desc_zone);
1671 out_destroy_trans_zone:
1672 kmem_zone_destroy(xfs_trans_zone);
1673 out_destroy_ifork_zone:
1674 kmem_zone_destroy(xfs_ifork_zone);
1675 out_destroy_da_state_zone:
1676 kmem_zone_destroy(xfs_da_state_zone);
1677 out_destroy_btree_cur_zone:
1678 kmem_zone_destroy(xfs_btree_cur_zone);
1679 out_destroy_bmap_free_item_zone:
1680 kmem_zone_destroy(xfs_bmap_free_item_zone);
1681 out_destroy_log_ticket_zone:
1682 kmem_zone_destroy(xfs_log_ticket_zone);
1683 out_destroy_ioend_pool:
1684 mempool_destroy(xfs_ioend_pool);
1685 out_destroy_ioend_zone:
1686 kmem_zone_destroy(xfs_ioend_zone);
1687 out:
1688 return -ENOMEM;
1689}
1690
1691STATIC void
1692xfs_destroy_zones(void)
1693{
1694 /*
1695 * Make sure all delayed rcu free are flushed before we
1696 * destroy caches.
1697 */
1698 rcu_barrier();
1699 kmem_zone_destroy(xfs_icreate_zone);
1700 kmem_zone_destroy(xfs_ili_zone);
1701 kmem_zone_destroy(xfs_inode_zone);
1702 kmem_zone_destroy(xfs_efi_zone);
1703 kmem_zone_destroy(xfs_efd_zone);
1704 kmem_zone_destroy(xfs_buf_item_zone);
1705 kmem_zone_destroy(xfs_log_item_desc_zone);
1706 kmem_zone_destroy(xfs_trans_zone);
1707 kmem_zone_destroy(xfs_ifork_zone);
1708 kmem_zone_destroy(xfs_da_state_zone);
1709 kmem_zone_destroy(xfs_btree_cur_zone);
1710 kmem_zone_destroy(xfs_bmap_free_item_zone);
1711 kmem_zone_destroy(xfs_log_ticket_zone);
1712 mempool_destroy(xfs_ioend_pool);
1713 kmem_zone_destroy(xfs_ioend_zone);
1714
1715}
1716
1717STATIC int __init
1718xfs_init_workqueues(void)
1719{
1720 /*
1721 * The allocation workqueue can be used in memory reclaim situations
1722 * (writepage path), and parallelism is only limited by the number of
1723 * AGs in all the filesystems mounted. Hence use the default large
1724 * max_active value for this workqueue.
1725 */
1726 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
1727 if (!xfs_alloc_wq)
1728 return -ENOMEM;
1729
1730 return 0;
1731}
1732
1733STATIC void
1734xfs_destroy_workqueues(void)
1735{
1736 destroy_workqueue(xfs_alloc_wq);
1737}
1738
1739STATIC int __init
1740init_xfs_fs(void)
1741{
1742 int error;
1743
1744 printk(KERN_INFO XFS_VERSION_STRING " with "
1745 XFS_BUILD_OPTIONS " enabled\n");
1746
1747 xfs_dir_startup();
1748
1749 error = xfs_init_zones();
1750 if (error)
1751 goto out;
1752
1753 error = xfs_init_workqueues();
1754 if (error)
1755 goto out_destroy_zones;
1756
1757 error = xfs_mru_cache_init();
1758 if (error)
1759 goto out_destroy_wq;
1760
1761 error = xfs_filestream_init();
1762 if (error)
1763 goto out_mru_cache_uninit;
1764
1765 error = xfs_buf_init();
1766 if (error)
1767 goto out_filestream_uninit;
1768
1769 error = xfs_init_procfs();
1770 if (error)
1771 goto out_buf_terminate;
1772
1773 error = xfs_sysctl_register();
1774 if (error)
1775 goto out_cleanup_procfs;
1776
1777 error = xfs_qm_init();
1778 if (error)
1779 goto out_sysctl_unregister;
1780
1781 error = register_filesystem(&xfs_fs_type);
1782 if (error)
1783 goto out_qm_exit;
1784 return 0;
1785
1786 out_qm_exit:
1787 xfs_qm_exit();
1788 out_sysctl_unregister:
1789 xfs_sysctl_unregister();
1790 out_cleanup_procfs:
1791 xfs_cleanup_procfs();
1792 out_buf_terminate:
1793 xfs_buf_terminate();
1794 out_filestream_uninit:
1795 xfs_filestream_uninit();
1796 out_mru_cache_uninit:
1797 xfs_mru_cache_uninit();
1798 out_destroy_wq:
1799 xfs_destroy_workqueues();
1800 out_destroy_zones:
1801 xfs_destroy_zones();
1802 out:
1803 return error;
1804}
1805
1806STATIC void __exit
1807exit_xfs_fs(void)
1808{
1809 xfs_qm_exit();
1810 unregister_filesystem(&xfs_fs_type);
1811 xfs_sysctl_unregister();
1812 xfs_cleanup_procfs();
1813 xfs_buf_terminate();
1814 xfs_filestream_uninit();
1815 xfs_mru_cache_uninit();
1816 xfs_destroy_workqueues();
1817 xfs_destroy_zones();
1818}
1819
1820module_init(init_xfs_fs);
1821module_exit(exit_xfs_fs);
1822
1823MODULE_AUTHOR("Silicon Graphics, Inc.");
1824MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1825MODULE_LICENSE("GPL");
Linux 6.x block layer and io_uring tree(s)