projects / deliverable / linux.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
xfs: Initialize all quota inodes to be NULLFSINO
[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_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 ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
425 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
426 xfs_warn(mp, "cannot mount with both project and group quota");
427 return EINVAL;
428 }
429
430 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
431 xfs_warn(mp, "sunit and swidth must be specified together");
432 return EINVAL;
433 }
434
435 if (dsunit && (dswidth % dsunit != 0)) {
436 xfs_warn(mp,
437 "stripe width (%d) must be a multiple of the stripe unit (%d)",
438 dswidth, dsunit);
439 return EINVAL;
440 }
441
442done:
443 if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
444 /*
445 * At this point the superblock has not been read
446 * in, therefore we do not know the block size.
447 * Before the mount call ends we will convert
448 * these to FSBs.
449 */
450 mp->m_dalign = dsunit;
451 mp->m_swidth = dswidth;
452 }
453
454 if (mp->m_logbufs != -1 &&
455 mp->m_logbufs != 0 &&
456 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
457 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
458 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
459 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
460 return XFS_ERROR(EINVAL);
461 }
462 if (mp->m_logbsize != -1 &&
463 mp->m_logbsize != 0 &&
464 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
465 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
466 !is_power_of_2(mp->m_logbsize))) {
467 xfs_warn(mp,
468 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
469 mp->m_logbsize);
470 return XFS_ERROR(EINVAL);
471 }
472
473 if (iosizelog) {
474 if (iosizelog > XFS_MAX_IO_LOG ||
475 iosizelog < XFS_MIN_IO_LOG) {
476 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
477 iosizelog, XFS_MIN_IO_LOG,
478 XFS_MAX_IO_LOG);
479 return XFS_ERROR(EINVAL);
480 }
481
482 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
483 mp->m_readio_log = iosizelog;
484 mp->m_writeio_log = iosizelog;
485 }
486
487 return 0;
488}
489
490struct proc_xfs_info {
491 int flag;
492 char *str;
493};
494
495STATIC int
496xfs_showargs(
497 struct xfs_mount *mp,
498 struct seq_file *m)
499{
500 static struct proc_xfs_info xfs_info_set[] = {
501 /* the few simple ones we can get from the mount struct */
502 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
503 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
504 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
505 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
506 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
507 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
508 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
509 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
510 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
511 { XFS_MOUNT_DISCARD, "," MNTOPT_DISCARD },
512 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_32BITINODE },
513 { 0, NULL }
514 };
515 static struct proc_xfs_info xfs_info_unset[] = {
516 /* the few simple ones we can get from the mount struct */
517 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
518 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
519 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
520 { 0, NULL }
521 };
522 struct proc_xfs_info *xfs_infop;
523
524 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
525 if (mp->m_flags & xfs_infop->flag)
526 seq_puts(m, xfs_infop->str);
527 }
528 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
529 if (!(mp->m_flags & xfs_infop->flag))
530 seq_puts(m, xfs_infop->str);
531 }
532
533 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
534 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
535 (int)(1 << mp->m_writeio_log) >> 10);
536
537 if (mp->m_logbufs > 0)
538 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
539 if (mp->m_logbsize > 0)
540 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
541
542 if (mp->m_logname)
543 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
544 if (mp->m_rtname)
545 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
546
547 if (mp->m_dalign > 0)
548 seq_printf(m, "," MNTOPT_SUNIT "=%d",
549 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
550 if (mp->m_swidth > 0)
551 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
552 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
553
554 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
555 seq_puts(m, "," MNTOPT_USRQUOTA);
556 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
557 seq_puts(m, "," MNTOPT_UQUOTANOENF);
558
559 /* Either project or group quotas can be active, not both */
560
561 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
562 if (mp->m_qflags & XFS_PQUOTA_ENFD)
563 seq_puts(m, "," MNTOPT_PRJQUOTA);
564 else
565 seq_puts(m, "," MNTOPT_PQUOTANOENF);
566 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
567 if (mp->m_qflags & XFS_GQUOTA_ENFD)
568 seq_puts(m, "," MNTOPT_GRPQUOTA);
569 else
570 seq_puts(m, "," MNTOPT_GQUOTANOENF);
571 }
572
573 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
574 seq_puts(m, "," MNTOPT_NOQUOTA);
575
576 return 0;
577}
578__uint64_t
579xfs_max_file_offset(
580 unsigned int blockshift)
581{
582 unsigned int pagefactor = 1;
583 unsigned int bitshift = BITS_PER_LONG - 1;
584
585 /* Figure out maximum filesize, on Linux this can depend on
586 * the filesystem blocksize (on 32 bit platforms).
587 * __block_write_begin does this in an [unsigned] long...
588 * page->index << (PAGE_CACHE_SHIFT - bbits)
589 * So, for page sized blocks (4K on 32 bit platforms),
590 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
591 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
592 * but for smaller blocksizes it is less (bbits = log2 bsize).
593 * Note1: get_block_t takes a long (implicit cast from above)
594 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
595 * can optionally convert the [unsigned] long from above into
596 * an [unsigned] long long.
597 */
598
599#if BITS_PER_LONG == 32
600# if defined(CONFIG_LBDAF)
601 ASSERT(sizeof(sector_t) == 8);
602 pagefactor = PAGE_CACHE_SIZE;
603 bitshift = BITS_PER_LONG;
604# else
605 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
606# endif
607#endif
608
609 return (((__uint64_t)pagefactor) << bitshift) - 1;
610}
611
612xfs_agnumber_t
613xfs_set_inode32(struct xfs_mount *mp)
614{
615 xfs_agnumber_t index = 0;
616 xfs_agnumber_t maxagi = 0;
617 xfs_sb_t *sbp = &mp->m_sb;
618 xfs_agnumber_t max_metadata;
619 xfs_agino_t agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks -1, 0);
620 xfs_ino_t ino = XFS_AGINO_TO_INO(mp, sbp->sb_agcount -1, agino);
621 xfs_perag_t *pag;
622
623 /* Calculate how much should be reserved for inodes to meet
624 * the max inode percentage.
625 */
626 if (mp->m_maxicount) {
627 __uint64_t icount;
628
629 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
630 do_div(icount, 100);
631 icount += sbp->sb_agblocks - 1;
632 do_div(icount, sbp->sb_agblocks);
633 max_metadata = icount;
634 } else {
635 max_metadata = sbp->sb_agcount;
636 }
637
638 for (index = 0; index < sbp->sb_agcount; index++) {
639 ino = XFS_AGINO_TO_INO(mp, index, agino);
640
641 if (ino > XFS_MAXINUMBER_32) {
642 pag = xfs_perag_get(mp, index);
643 pag->pagi_inodeok = 0;
644 pag->pagf_metadata = 0;
645 xfs_perag_put(pag);
646 continue;
647 }
648
649 pag = xfs_perag_get(mp, index);
650 pag->pagi_inodeok = 1;
651 maxagi++;
652 if (index < max_metadata)
653 pag->pagf_metadata = 1;
654 xfs_perag_put(pag);
655 }
656 mp->m_flags |= (XFS_MOUNT_32BITINODES |
657 XFS_MOUNT_SMALL_INUMS);
658
659 return maxagi;
660}
661
662xfs_agnumber_t
663xfs_set_inode64(struct xfs_mount *mp)
664{
665 xfs_agnumber_t index = 0;
666
667 for (index = 0; index < mp->m_sb.sb_agcount; index++) {
668 struct xfs_perag *pag;
669
670 pag = xfs_perag_get(mp, index);
671 pag->pagi_inodeok = 1;
672 pag->pagf_metadata = 0;
673 xfs_perag_put(pag);
674 }
675
676 /* There is no need for lock protection on m_flags,
677 * the rw_semaphore of the VFS superblock is locked
678 * during mount/umount/remount operations, so this is
679 * enough to avoid concurency on the m_flags field
680 */
681 mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
682 XFS_MOUNT_SMALL_INUMS);
683 return index;
684}
685
686STATIC int
687xfs_blkdev_get(
688 xfs_mount_t *mp,
689 const char *name,
690 struct block_device **bdevp)
691{
692 int error = 0;
693
694 *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
695 mp);
696 if (IS_ERR(*bdevp)) {
697 error = PTR_ERR(*bdevp);
698 xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
699 }
700
701 return -error;
702}
703
704STATIC void
705xfs_blkdev_put(
706 struct block_device *bdev)
707{
708 if (bdev)
709 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
710}
711
712void
713xfs_blkdev_issue_flush(
714 xfs_buftarg_t *buftarg)
715{
716 blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
717}
718
719STATIC void
720xfs_close_devices(
721 struct xfs_mount *mp)
722{
723 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
724 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
725 xfs_free_buftarg(mp, mp->m_logdev_targp);
726 xfs_blkdev_put(logdev);
727 }
728 if (mp->m_rtdev_targp) {
729 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
730 xfs_free_buftarg(mp, mp->m_rtdev_targp);
731 xfs_blkdev_put(rtdev);
732 }
733 xfs_free_buftarg(mp, mp->m_ddev_targp);
734}
735
736/*
737 * The file system configurations are:
738 * (1) device (partition) with data and internal log
739 * (2) logical volume with data and log subvolumes.
740 * (3) logical volume with data, log, and realtime subvolumes.
741 *
742 * We only have to handle opening the log and realtime volumes here if
743 * they are present. The data subvolume has already been opened by
744 * get_sb_bdev() and is stored in sb->s_bdev.
745 */
746STATIC int
747xfs_open_devices(
748 struct xfs_mount *mp)
749{
750 struct block_device *ddev = mp->m_super->s_bdev;
751 struct block_device *logdev = NULL, *rtdev = NULL;
752 int error;
753
754 /*
755 * Open real time and log devices - order is important.
756 */
757 if (mp->m_logname) {
758 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
759 if (error)
760 goto out;
761 }
762
763 if (mp->m_rtname) {
764 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
765 if (error)
766 goto out_close_logdev;
767
768 if (rtdev == ddev || rtdev == logdev) {
769 xfs_warn(mp,
770 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
771 error = EINVAL;
772 goto out_close_rtdev;
773 }
774 }
775
776 /*
777 * Setup xfs_mount buffer target pointers
778 */
779 error = ENOMEM;
780 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, 0, mp->m_fsname);
781 if (!mp->m_ddev_targp)
782 goto out_close_rtdev;
783
784 if (rtdev) {
785 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, 1,
786 mp->m_fsname);
787 if (!mp->m_rtdev_targp)
788 goto out_free_ddev_targ;
789 }
790
791 if (logdev && logdev != ddev) {
792 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, 1,
793 mp->m_fsname);
794 if (!mp->m_logdev_targp)
795 goto out_free_rtdev_targ;
796 } else {
797 mp->m_logdev_targp = mp->m_ddev_targp;
798 }
799
800 return 0;
801
802 out_free_rtdev_targ:
803 if (mp->m_rtdev_targp)
804 xfs_free_buftarg(mp, mp->m_rtdev_targp);
805 out_free_ddev_targ:
806 xfs_free_buftarg(mp, mp->m_ddev_targp);
807 out_close_rtdev:
808 if (rtdev)
809 xfs_blkdev_put(rtdev);
810 out_close_logdev:
811 if (logdev && logdev != ddev)
812 xfs_blkdev_put(logdev);
813 out:
814 return error;
815}
816
817/*
818 * Setup xfs_mount buffer target pointers based on superblock
819 */
820STATIC int
821xfs_setup_devices(
822 struct xfs_mount *mp)
823{
824 int error;
825
826 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
827 mp->m_sb.sb_sectsize);
828 if (error)
829 return error;
830
831 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
832 unsigned int log_sector_size = BBSIZE;
833
834 if (xfs_sb_version_hassector(&mp->m_sb))
835 log_sector_size = mp->m_sb.sb_logsectsize;
836 error = xfs_setsize_buftarg(mp->m_logdev_targp,
837 mp->m_sb.sb_blocksize,
838 log_sector_size);
839 if (error)
840 return error;
841 }
842 if (mp->m_rtdev_targp) {
843 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
844 mp->m_sb.sb_blocksize,
845 mp->m_sb.sb_sectsize);
846 if (error)
847 return error;
848 }
849
850 return 0;
851}
852
853STATIC int
854xfs_init_mount_workqueues(
855 struct xfs_mount *mp)
856{
857 mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
858 WQ_MEM_RECLAIM, 0, mp->m_fsname);
859 if (!mp->m_data_workqueue)
860 goto out;
861
862 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
863 WQ_MEM_RECLAIM, 0, mp->m_fsname);
864 if (!mp->m_unwritten_workqueue)
865 goto out_destroy_data_iodone_queue;
866
867 mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
868 WQ_MEM_RECLAIM, 0, mp->m_fsname);
869 if (!mp->m_cil_workqueue)
870 goto out_destroy_unwritten;
871
872 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
873 WQ_NON_REENTRANT, 0, mp->m_fsname);
874 if (!mp->m_reclaim_workqueue)
875 goto out_destroy_cil;
876
877 mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
878 WQ_NON_REENTRANT, 0, mp->m_fsname);
879 if (!mp->m_log_workqueue)
880 goto out_destroy_reclaim;
881
882 mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
883 WQ_NON_REENTRANT, 0, mp->m_fsname);
884 if (!mp->m_eofblocks_workqueue)
885 goto out_destroy_log;
886
887 return 0;
888
889out_destroy_log:
890 destroy_workqueue(mp->m_log_workqueue);
891out_destroy_reclaim:
892 destroy_workqueue(mp->m_reclaim_workqueue);
893out_destroy_cil:
894 destroy_workqueue(mp->m_cil_workqueue);
895out_destroy_unwritten:
896 destroy_workqueue(mp->m_unwritten_workqueue);
897out_destroy_data_iodone_queue:
898 destroy_workqueue(mp->m_data_workqueue);
899out:
900 return -ENOMEM;
901}
902
903STATIC void
904xfs_destroy_mount_workqueues(
905 struct xfs_mount *mp)
906{
907 destroy_workqueue(mp->m_eofblocks_workqueue);
908 destroy_workqueue(mp->m_log_workqueue);
909 destroy_workqueue(mp->m_reclaim_workqueue);
910 destroy_workqueue(mp->m_cil_workqueue);
911 destroy_workqueue(mp->m_data_workqueue);
912 destroy_workqueue(mp->m_unwritten_workqueue);
913}
914
915/*
916 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
917 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
918 * for IO to complete so that we effectively throttle multiple callers to the
919 * rate at which IO is completing.
920 */
921void
922xfs_flush_inodes(
923 struct xfs_mount *mp)
924{
925 struct super_block *sb = mp->m_super;
926
927 if (down_read_trylock(&sb->s_umount)) {
928 sync_inodes_sb(sb);
929 up_read(&sb->s_umount);
930 }
931}
932
933/* Catch misguided souls that try to use this interface on XFS */
934STATIC struct inode *
935xfs_fs_alloc_inode(
936 struct super_block *sb)
937{
938 BUG();
939 return NULL;
940}
941
942/*
943 * Now that the generic code is guaranteed not to be accessing
944 * the linux inode, we can reclaim the inode.
945 */
946STATIC void
947xfs_fs_destroy_inode(
948 struct inode *inode)
949{
950 struct xfs_inode *ip = XFS_I(inode);
951
952 trace_xfs_destroy_inode(ip);
953
954 XFS_STATS_INC(vn_reclaim);
955
956 /* bad inode, get out here ASAP */
957 if (is_bad_inode(inode))
958 goto out_reclaim;
959
960 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
961
962 /*
963 * We should never get here with one of the reclaim flags already set.
964 */
965 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
966 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
967
968 /*
969 * We always use background reclaim here because even if the
970 * inode is clean, it still may be under IO and hence we have
971 * to take the flush lock. The background reclaim path handles
972 * this more efficiently than we can here, so simply let background
973 * reclaim tear down all inodes.
974 */
975out_reclaim:
976 xfs_inode_set_reclaim_tag(ip);
977}
978
979/*
980 * Slab object creation initialisation for the XFS inode.
981 * This covers only the idempotent fields in the XFS inode;
982 * all other fields need to be initialised on allocation
983 * from the slab. This avoids the need to repeatedly initialise
984 * fields in the xfs inode that left in the initialise state
985 * when freeing the inode.
986 */
987STATIC void
988xfs_fs_inode_init_once(
989 void *inode)
990{
991 struct xfs_inode *ip = inode;
992
993 memset(ip, 0, sizeof(struct xfs_inode));
994
995 /* vfs inode */
996 inode_init_once(VFS_I(ip));
997
998 /* xfs inode */
999 atomic_set(&ip->i_pincount, 0);
1000 spin_lock_init(&ip->i_flags_lock);
1001
1002 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
1003 "xfsino", ip->i_ino);
1004}
1005
1006STATIC void
1007xfs_fs_evict_inode(
1008 struct inode *inode)
1009{
1010 xfs_inode_t *ip = XFS_I(inode);
1011
1012 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1013
1014 trace_xfs_evict_inode(ip);
1015
1016 truncate_inode_pages(&inode->i_data, 0);
1017 clear_inode(inode);
1018 XFS_STATS_INC(vn_rele);
1019 XFS_STATS_INC(vn_remove);
1020 XFS_STATS_DEC(vn_active);
1021
1022 xfs_inactive(ip);
1023}
1024
1025/*
1026 * We do an unlocked check for XFS_IDONTCACHE here because we are already
1027 * serialised against cache hits here via the inode->i_lock and igrab() in
1028 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1029 * racing with us, and it avoids needing to grab a spinlock here for every inode
1030 * we drop the final reference on.
1031 */
1032STATIC int
1033xfs_fs_drop_inode(
1034 struct inode *inode)
1035{
1036 struct xfs_inode *ip = XFS_I(inode);
1037
1038 return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1039}
1040
1041STATIC void
1042xfs_free_fsname(
1043 struct xfs_mount *mp)
1044{
1045 kfree(mp->m_fsname);
1046 kfree(mp->m_rtname);
1047 kfree(mp->m_logname);
1048}
1049
1050STATIC void
1051xfs_fs_put_super(
1052 struct super_block *sb)
1053{
1054 struct xfs_mount *mp = XFS_M(sb);
1055
1056 xfs_filestream_unmount(mp);
1057 xfs_unmountfs(mp);
1058
1059 xfs_freesb(mp);
1060 xfs_icsb_destroy_counters(mp);
1061 xfs_destroy_mount_workqueues(mp);
1062 xfs_close_devices(mp);
1063 xfs_free_fsname(mp);
1064 kfree(mp);
1065}
1066
1067STATIC int
1068xfs_fs_sync_fs(
1069 struct super_block *sb,
1070 int wait)
1071{
1072 struct xfs_mount *mp = XFS_M(sb);
1073
1074 /*
1075 * Doing anything during the async pass would be counterproductive.
1076 */
1077 if (!wait)
1078 return 0;
1079
1080 xfs_log_force(mp, XFS_LOG_SYNC);
1081 if (laptop_mode) {
1082 /*
1083 * The disk must be active because we're syncing.
1084 * We schedule log work now (now that the disk is
1085 * active) instead of later (when it might not be).
1086 */
1087 flush_delayed_work(&mp->m_log->l_work);
1088 }
1089
1090 return 0;
1091}
1092
1093STATIC int
1094xfs_fs_statfs(
1095 struct dentry *dentry,
1096 struct kstatfs *statp)
1097{
1098 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1099 xfs_sb_t *sbp = &mp->m_sb;
1100 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1101 __uint64_t fakeinos, id;
1102 xfs_extlen_t lsize;
1103 __int64_t ffree;
1104
1105 statp->f_type = XFS_SB_MAGIC;
1106 statp->f_namelen = MAXNAMELEN - 1;
1107
1108 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1109 statp->f_fsid.val[0] = (u32)id;
1110 statp->f_fsid.val[1] = (u32)(id >> 32);
1111
1112 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1113
1114 spin_lock(&mp->m_sb_lock);
1115 statp->f_bsize = sbp->sb_blocksize;
1116 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1117 statp->f_blocks = sbp->sb_dblocks - lsize;
1118 statp->f_bfree = statp->f_bavail =
1119 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1120 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1121 statp->f_files =
1122 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1123 if (mp->m_maxicount)
1124 statp->f_files = min_t(typeof(statp->f_files),
1125 statp->f_files,
1126 mp->m_maxicount);
1127
1128 /* make sure statp->f_ffree does not underflow */
1129 ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1130 statp->f_ffree = max_t(__int64_t, ffree, 0);
1131
1132 spin_unlock(&mp->m_sb_lock);
1133
1134 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1135 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1136 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1137 xfs_qm_statvfs(ip, statp);
1138 return 0;
1139}
1140
1141STATIC void
1142xfs_save_resvblks(struct xfs_mount *mp)
1143{
1144 __uint64_t resblks = 0;
1145
1146 mp->m_resblks_save = mp->m_resblks;
1147 xfs_reserve_blocks(mp, &resblks, NULL);
1148}
1149
1150STATIC void
1151xfs_restore_resvblks(struct xfs_mount *mp)
1152{
1153 __uint64_t resblks;
1154
1155 if (mp->m_resblks_save) {
1156 resblks = mp->m_resblks_save;
1157 mp->m_resblks_save = 0;
1158 } else
1159 resblks = xfs_default_resblks(mp);
1160
1161 xfs_reserve_blocks(mp, &resblks, NULL);
1162}
1163
1164/*
1165 * Trigger writeback of all the dirty metadata in the file system.
1166 *
1167 * This ensures that the metadata is written to their location on disk rather
1168 * than just existing in transactions in the log. This means after a quiesce
1169 * there is no log replay required to write the inodes to disk - this is the
1170 * primary difference between a sync and a quiesce.
1171 *
1172 * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1173 * it is started again when appropriate.
1174 */
1175void
1176xfs_quiesce_attr(
1177 struct xfs_mount *mp)
1178{
1179 int error = 0;
1180
1181 /* wait for all modifications to complete */
1182 while (atomic_read(&mp->m_active_trans) > 0)
1183 delay(100);
1184
1185 /* force the log to unpin objects from the now complete transactions */
1186 xfs_log_force(mp, XFS_LOG_SYNC);
1187
1188 /* reclaim inodes to do any IO before the freeze completes */
1189 xfs_reclaim_inodes(mp, 0);
1190 xfs_reclaim_inodes(mp, SYNC_WAIT);
1191
1192 /* Push the superblock and write an unmount record */
1193 error = xfs_log_sbcount(mp);
1194 if (error)
1195 xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1196 "Frozen image may not be consistent.");
1197 /*
1198 * Just warn here till VFS can correctly support
1199 * read-only remount without racing.
1200 */
1201 WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1202
1203 xfs_log_quiesce(mp);
1204}
1205
1206STATIC int
1207xfs_fs_remount(
1208 struct super_block *sb,
1209 int *flags,
1210 char *options)
1211{
1212 struct xfs_mount *mp = XFS_M(sb);
1213 substring_t args[MAX_OPT_ARGS];
1214 char *p;
1215 int error;
1216
1217 while ((p = strsep(&options, ",")) != NULL) {
1218 int token;
1219
1220 if (!*p)
1221 continue;
1222
1223 token = match_token(p, tokens, args);
1224 switch (token) {
1225 case Opt_barrier:
1226 mp->m_flags |= XFS_MOUNT_BARRIER;
1227 break;
1228 case Opt_nobarrier:
1229 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1230 break;
1231 case Opt_inode64:
1232 mp->m_maxagi = xfs_set_inode64(mp);
1233 break;
1234 case Opt_inode32:
1235 mp->m_maxagi = xfs_set_inode32(mp);
1236 break;
1237 default:
1238 /*
1239 * Logically we would return an error here to prevent
1240 * users from believing they might have changed
1241 * mount options using remount which can't be changed.
1242 *
1243 * But unfortunately mount(8) adds all options from
1244 * mtab and fstab to the mount arguments in some cases
1245 * so we can't blindly reject options, but have to
1246 * check for each specified option if it actually
1247 * differs from the currently set option and only
1248 * reject it if that's the case.
1249 *
1250 * Until that is implemented we return success for
1251 * every remount request, and silently ignore all
1252 * options that we can't actually change.
1253 */
1254#if 0
1255 xfs_info(mp,
1256 "mount option \"%s\" not supported for remount\n", p);
1257 return -EINVAL;
1258#else
1259 break;
1260#endif
1261 }
1262 }
1263
1264 /* ro -> rw */
1265 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1266 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1267
1268 /*
1269 * If this is the first remount to writeable state we
1270 * might have some superblock changes to update.
1271 */
1272 if (mp->m_update_flags) {
1273 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1274 if (error) {
1275 xfs_warn(mp, "failed to write sb changes");
1276 return error;
1277 }
1278 mp->m_update_flags = 0;
1279 }
1280
1281 /*
1282 * Fill out the reserve pool if it is empty. Use the stashed
1283 * value if it is non-zero, otherwise go with the default.
1284 */
1285 xfs_restore_resvblks(mp);
1286 xfs_log_work_queue(mp);
1287 }
1288
1289 /* rw -> ro */
1290 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1291 /*
1292 * Before we sync the metadata, we need to free up the reserve
1293 * block pool so that the used block count in the superblock on
1294 * disk is correct at the end of the remount. Stash the current
1295 * reserve pool size so that if we get remounted rw, we can
1296 * return it to the same size.
1297 */
1298 xfs_save_resvblks(mp);
1299 xfs_quiesce_attr(mp);
1300 mp->m_flags |= XFS_MOUNT_RDONLY;
1301 }
1302
1303 return 0;
1304}
1305
1306/*
1307 * Second stage of a freeze. The data is already frozen so we only
1308 * need to take care of the metadata. Once that's done write a dummy
1309 * record to dirty the log in case of a crash while frozen.
1310 */
1311STATIC int
1312xfs_fs_freeze(
1313 struct super_block *sb)
1314{
1315 struct xfs_mount *mp = XFS_M(sb);
1316
1317 xfs_save_resvblks(mp);
1318 xfs_quiesce_attr(mp);
1319 return -xfs_fs_log_dummy(mp);
1320}
1321
1322STATIC int
1323xfs_fs_unfreeze(
1324 struct super_block *sb)
1325{
1326 struct xfs_mount *mp = XFS_M(sb);
1327
1328 xfs_restore_resvblks(mp);
1329 xfs_log_work_queue(mp);
1330 return 0;
1331}
1332
1333STATIC int
1334xfs_fs_show_options(
1335 struct seq_file *m,
1336 struct dentry *root)
1337{
1338 return -xfs_showargs(XFS_M(root->d_sb), m);
1339}
1340
1341/*
1342 * This function fills in xfs_mount_t fields based on mount args.
1343 * Note: the superblock _has_ now been read in.
1344 */
1345STATIC int
1346xfs_finish_flags(
1347 struct xfs_mount *mp)
1348{
1349 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1350
1351 /* Fail a mount where the logbuf is smaller than the log stripe */
1352 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1353 if (mp->m_logbsize <= 0 &&
1354 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1355 mp->m_logbsize = mp->m_sb.sb_logsunit;
1356 } else if (mp->m_logbsize > 0 &&
1357 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1358 xfs_warn(mp,
1359 "logbuf size must be greater than or equal to log stripe size");
1360 return XFS_ERROR(EINVAL);
1361 }
1362 } else {
1363 /* Fail a mount if the logbuf is larger than 32K */
1364 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1365 xfs_warn(mp,
1366 "logbuf size for version 1 logs must be 16K or 32K");
1367 return XFS_ERROR(EINVAL);
1368 }
1369 }
1370
1371 /*
1372 * V5 filesystems always use attr2 format for attributes.
1373 */
1374 if (xfs_sb_version_hascrc(&mp->m_sb) &&
1375 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1376 xfs_warn(mp,
1377"Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1378 MNTOPT_NOATTR2, MNTOPT_ATTR2);
1379 return XFS_ERROR(EINVAL);
1380 }
1381
1382 /*
1383 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1384 * told by noattr2 to turn it off
1385 */
1386 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1387 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1388 mp->m_flags |= XFS_MOUNT_ATTR2;
1389
1390 /*
1391 * prohibit r/w mounts of read-only filesystems
1392 */
1393 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1394 xfs_warn(mp,
1395 "cannot mount a read-only filesystem as read-write");
1396 return XFS_ERROR(EROFS);
1397 }
1398
1399 return 0;
1400}
1401
1402STATIC int
1403xfs_fs_fill_super(
1404 struct super_block *sb,
1405 void *data,
1406 int silent)
1407{
1408 struct inode *root;
1409 struct xfs_mount *mp = NULL;
1410 int flags = 0, error = ENOMEM;
1411
1412 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1413 if (!mp)
1414 goto out;
1415
1416 spin_lock_init(&mp->m_sb_lock);
1417 mutex_init(&mp->m_growlock);
1418 atomic_set(&mp->m_active_trans, 0);
1419 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1420 INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1421
1422 mp->m_super = sb;
1423 sb->s_fs_info = mp;
1424
1425 error = xfs_parseargs(mp, (char *)data);
1426 if (error)
1427 goto out_free_fsname;
1428
1429 sb_min_blocksize(sb, BBSIZE);
1430 sb->s_xattr = xfs_xattr_handlers;
1431 sb->s_export_op = &xfs_export_operations;
1432#ifdef CONFIG_XFS_QUOTA
1433 sb->s_qcop = &xfs_quotactl_operations;
1434#endif
1435 sb->s_op = &xfs_super_operations;
1436
1437 if (silent)
1438 flags |= XFS_MFSI_QUIET;
1439
1440 error = xfs_open_devices(mp);
1441 if (error)
1442 goto out_free_fsname;
1443
1444 error = xfs_init_mount_workqueues(mp);
1445 if (error)
1446 goto out_close_devices;
1447
1448 error = xfs_icsb_init_counters(mp);
1449 if (error)
1450 goto out_destroy_workqueues;
1451
1452 error = xfs_readsb(mp, flags);
1453 if (error)
1454 goto out_destroy_counters;
1455
1456 error = xfs_finish_flags(mp);
1457 if (error)
1458 goto out_free_sb;
1459
1460 error = xfs_setup_devices(mp);
1461 if (error)
1462 goto out_free_sb;
1463
1464 error = xfs_filestream_mount(mp);
1465 if (error)
1466 goto out_free_sb;
1467
1468 /*
1469 * we must configure the block size in the superblock before we run the
1470 * full mount process as the mount process can lookup and cache inodes.
1471 */
1472 sb->s_magic = XFS_SB_MAGIC;
1473 sb->s_blocksize = mp->m_sb.sb_blocksize;
1474 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1475 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1476 sb->s_max_links = XFS_MAXLINK;
1477 sb->s_time_gran = 1;
1478 set_posix_acl_flag(sb);
1479
1480 /* version 5 superblocks support inode version counters. */
1481 if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1482 sb->s_flags |= MS_I_VERSION;
1483
1484 error = xfs_mountfs(mp);
1485 if (error)
1486 goto out_filestream_unmount;
1487
1488 root = igrab(VFS_I(mp->m_rootip));
1489 if (!root) {
1490 error = ENOENT;
1491 goto out_unmount;
1492 }
1493 if (is_bad_inode(root)) {
1494 error = EINVAL;
1495 goto out_unmount;
1496 }
1497 sb->s_root = d_make_root(root);
1498 if (!sb->s_root) {
1499 error = ENOMEM;
1500 goto out_unmount;
1501 }
1502
1503 return 0;
1504
1505 out_filestream_unmount:
1506 xfs_filestream_unmount(mp);
1507 out_free_sb:
1508 xfs_freesb(mp);
1509 out_destroy_counters:
1510 xfs_icsb_destroy_counters(mp);
1511out_destroy_workqueues:
1512 xfs_destroy_mount_workqueues(mp);
1513 out_close_devices:
1514 xfs_close_devices(mp);
1515 out_free_fsname:
1516 xfs_free_fsname(mp);
1517 kfree(mp);
1518 out:
1519 return -error;
1520
1521 out_unmount:
1522 xfs_filestream_unmount(mp);
1523 xfs_unmountfs(mp);
1524 goto out_free_sb;
1525}
1526
1527STATIC struct dentry *
1528xfs_fs_mount(
1529 struct file_system_type *fs_type,
1530 int flags,
1531 const char *dev_name,
1532 void *data)
1533{
1534 return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1535}
1536
1537static int
1538xfs_fs_nr_cached_objects(
1539 struct super_block *sb)
1540{
1541 return xfs_reclaim_inodes_count(XFS_M(sb));
1542}
1543
1544static void
1545xfs_fs_free_cached_objects(
1546 struct super_block *sb,
1547 int nr_to_scan)
1548{
1549 xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1550}
1551
1552static const struct super_operations xfs_super_operations = {
1553 .alloc_inode = xfs_fs_alloc_inode,
1554 .destroy_inode = xfs_fs_destroy_inode,
1555 .evict_inode = xfs_fs_evict_inode,
1556 .drop_inode = xfs_fs_drop_inode,
1557 .put_super = xfs_fs_put_super,
1558 .sync_fs = xfs_fs_sync_fs,
1559 .freeze_fs = xfs_fs_freeze,
1560 .unfreeze_fs = xfs_fs_unfreeze,
1561 .statfs = xfs_fs_statfs,
1562 .remount_fs = xfs_fs_remount,
1563 .show_options = xfs_fs_show_options,
1564 .nr_cached_objects = xfs_fs_nr_cached_objects,
1565 .free_cached_objects = xfs_fs_free_cached_objects,
1566};
1567
1568static struct file_system_type xfs_fs_type = {
1569 .owner = THIS_MODULE,
1570 .name = "xfs",
1571 .mount = xfs_fs_mount,
1572 .kill_sb = kill_block_super,
1573 .fs_flags = FS_REQUIRES_DEV,
1574};
1575MODULE_ALIAS_FS("xfs");
1576
1577STATIC int __init
1578xfs_init_zones(void)
1579{
1580
1581 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1582 if (!xfs_ioend_zone)
1583 goto out;
1584
1585 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1586 xfs_ioend_zone);
1587 if (!xfs_ioend_pool)
1588 goto out_destroy_ioend_zone;
1589
1590 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1591 "xfs_log_ticket");
1592 if (!xfs_log_ticket_zone)
1593 goto out_destroy_ioend_pool;
1594
1595 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1596 "xfs_bmap_free_item");
1597 if (!xfs_bmap_free_item_zone)
1598 goto out_destroy_log_ticket_zone;
1599
1600 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1601 "xfs_btree_cur");
1602 if (!xfs_btree_cur_zone)
1603 goto out_destroy_bmap_free_item_zone;
1604
1605 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1606 "xfs_da_state");
1607 if (!xfs_da_state_zone)
1608 goto out_destroy_btree_cur_zone;
1609
1610 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1611 if (!xfs_ifork_zone)
1612 goto out_destroy_da_state_zone;
1613
1614 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1615 if (!xfs_trans_zone)
1616 goto out_destroy_ifork_zone;
1617
1618 xfs_log_item_desc_zone =
1619 kmem_zone_init(sizeof(struct xfs_log_item_desc),
1620 "xfs_log_item_desc");
1621 if (!xfs_log_item_desc_zone)
1622 goto out_destroy_trans_zone;
1623
1624 /*
1625 * The size of the zone allocated buf log item is the maximum
1626 * size possible under XFS. This wastes a little bit of memory,
1627 * but it is much faster.
1628 */
1629 xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1630 "xfs_buf_item");
1631 if (!xfs_buf_item_zone)
1632 goto out_destroy_log_item_desc_zone;
1633
1634 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1635 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1636 sizeof(xfs_extent_t))), "xfs_efd_item");
1637 if (!xfs_efd_zone)
1638 goto out_destroy_buf_item_zone;
1639
1640 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1641 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1642 sizeof(xfs_extent_t))), "xfs_efi_item");
1643 if (!xfs_efi_zone)
1644 goto out_destroy_efd_zone;
1645
1646 xfs_inode_zone =
1647 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1648 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1649 xfs_fs_inode_init_once);
1650 if (!xfs_inode_zone)
1651 goto out_destroy_efi_zone;
1652
1653 xfs_ili_zone =
1654 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1655 KM_ZONE_SPREAD, NULL);
1656 if (!xfs_ili_zone)
1657 goto out_destroy_inode_zone;
1658 xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1659 "xfs_icr");
1660 if (!xfs_icreate_zone)
1661 goto out_destroy_ili_zone;
1662
1663 return 0;
1664
1665 out_destroy_ili_zone:
1666 kmem_zone_destroy(xfs_ili_zone);
1667 out_destroy_inode_zone:
1668 kmem_zone_destroy(xfs_inode_zone);
1669 out_destroy_efi_zone:
1670 kmem_zone_destroy(xfs_efi_zone);
1671 out_destroy_efd_zone:
1672 kmem_zone_destroy(xfs_efd_zone);
1673 out_destroy_buf_item_zone:
1674 kmem_zone_destroy(xfs_buf_item_zone);
1675 out_destroy_log_item_desc_zone:
1676 kmem_zone_destroy(xfs_log_item_desc_zone);
1677 out_destroy_trans_zone:
1678 kmem_zone_destroy(xfs_trans_zone);
1679 out_destroy_ifork_zone:
1680 kmem_zone_destroy(xfs_ifork_zone);
1681 out_destroy_da_state_zone:
1682 kmem_zone_destroy(xfs_da_state_zone);
1683 out_destroy_btree_cur_zone:
1684 kmem_zone_destroy(xfs_btree_cur_zone);
1685 out_destroy_bmap_free_item_zone:
1686 kmem_zone_destroy(xfs_bmap_free_item_zone);
1687 out_destroy_log_ticket_zone:
1688 kmem_zone_destroy(xfs_log_ticket_zone);
1689 out_destroy_ioend_pool:
1690 mempool_destroy(xfs_ioend_pool);
1691 out_destroy_ioend_zone:
1692 kmem_zone_destroy(xfs_ioend_zone);
1693 out:
1694 return -ENOMEM;
1695}
1696
1697STATIC void
1698xfs_destroy_zones(void)
1699{
1700 /*
1701 * Make sure all delayed rcu free are flushed before we
1702 * destroy caches.
1703 */
1704 rcu_barrier();
1705 kmem_zone_destroy(xfs_icreate_zone);
1706 kmem_zone_destroy(xfs_ili_zone);
1707 kmem_zone_destroy(xfs_inode_zone);
1708 kmem_zone_destroy(xfs_efi_zone);
1709 kmem_zone_destroy(xfs_efd_zone);
1710 kmem_zone_destroy(xfs_buf_item_zone);
1711 kmem_zone_destroy(xfs_log_item_desc_zone);
1712 kmem_zone_destroy(xfs_trans_zone);
1713 kmem_zone_destroy(xfs_ifork_zone);
1714 kmem_zone_destroy(xfs_da_state_zone);
1715 kmem_zone_destroy(xfs_btree_cur_zone);
1716 kmem_zone_destroy(xfs_bmap_free_item_zone);
1717 kmem_zone_destroy(xfs_log_ticket_zone);
1718 mempool_destroy(xfs_ioend_pool);
1719 kmem_zone_destroy(xfs_ioend_zone);
1720
1721}
1722
1723STATIC int __init
1724xfs_init_workqueues(void)
1725{
1726 /*
1727 * The allocation workqueue can be used in memory reclaim situations
1728 * (writepage path), and parallelism is only limited by the number of
1729 * AGs in all the filesystems mounted. Hence use the default large
1730 * max_active value for this workqueue.
1731 */
1732 xfs_alloc_wq = alloc_workqueue("xfsalloc", WQ_MEM_RECLAIM, 0);
1733 if (!xfs_alloc_wq)
1734 return -ENOMEM;
1735
1736 return 0;
1737}
1738
1739STATIC void
1740xfs_destroy_workqueues(void)
1741{
1742 destroy_workqueue(xfs_alloc_wq);
1743}
1744
1745STATIC int __init
1746init_xfs_fs(void)
1747{
1748 int error;
1749
1750 printk(KERN_INFO XFS_VERSION_STRING " with "
1751 XFS_BUILD_OPTIONS " enabled\n");
1752
1753 xfs_dir_startup();
1754
1755 error = xfs_init_zones();
1756 if (error)
1757 goto out;
1758
1759 error = xfs_init_workqueues();
1760 if (error)
1761 goto out_destroy_zones;
1762
1763 error = xfs_mru_cache_init();
1764 if (error)
1765 goto out_destroy_wq;
1766
1767 error = xfs_filestream_init();
1768 if (error)
1769 goto out_mru_cache_uninit;
1770
1771 error = xfs_buf_init();
1772 if (error)
1773 goto out_filestream_uninit;
1774
1775 error = xfs_init_procfs();
1776 if (error)
1777 goto out_buf_terminate;
1778
1779 error = xfs_sysctl_register();
1780 if (error)
1781 goto out_cleanup_procfs;
1782
1783 error = xfs_qm_init();
1784 if (error)
1785 goto out_sysctl_unregister;
1786
1787 error = register_filesystem(&xfs_fs_type);
1788 if (error)
1789 goto out_qm_exit;
1790 return 0;
1791
1792 out_qm_exit:
1793 xfs_qm_exit();
1794 out_sysctl_unregister:
1795 xfs_sysctl_unregister();
1796 out_cleanup_procfs:
1797 xfs_cleanup_procfs();
1798 out_buf_terminate:
1799 xfs_buf_terminate();
1800 out_filestream_uninit:
1801 xfs_filestream_uninit();
1802 out_mru_cache_uninit:
1803 xfs_mru_cache_uninit();
1804 out_destroy_wq:
1805 xfs_destroy_workqueues();
1806 out_destroy_zones:
1807 xfs_destroy_zones();
1808 out:
1809 return error;
1810}
1811
1812STATIC void __exit
1813exit_xfs_fs(void)
1814{
1815 xfs_qm_exit();
1816 unregister_filesystem(&xfs_fs_type);
1817 xfs_sysctl_unregister();
1818 xfs_cleanup_procfs();
1819 xfs_buf_terminate();
1820 xfs_filestream_uninit();
1821 xfs_mru_cache_uninit();
1822 xfs_destroy_workqueues();
1823 xfs_destroy_zones();
1824}
1825
1826module_init(init_xfs_fs);
1827module_exit(exit_xfs_fs);
1828
1829MODULE_AUTHOR("Silicon Graphics, Inc.");
1830MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1831MODULE_LICENSE("GPL");
Linux kernel - Deliverables
This page took 0.031312 seconds and 5 git commands to generate.