?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
eCryptfs: Remove mmap from directory operations
[linux/fpc-iii.git] / fs / xfs / linux-2.6 / xfs_super.c
blob77414db10dc2f44ee9b07d9741b785b6c3320a1c
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_bit.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_dir2.h"
27 #include "xfs_alloc.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_quota.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir2_sf.h"
35 #include "xfs_attr_sf.h"
36 #include "xfs_dinode.h"
37 #include "xfs_inode.h"
38 #include "xfs_btree.h"
39 #include "xfs_btree_trace.h"
40 #include "xfs_ialloc.h"
41 #include "xfs_bmap.h"
42 #include "xfs_rtalloc.h"
43 #include "xfs_error.h"
44 #include "xfs_itable.h"
45 #include "xfs_fsops.h"
46 #include "xfs_rw.h"
47 #include "xfs_attr.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
51 #include "xfs_version.h"
52 #include "xfs_log_priv.h"
53 #include "xfs_trans_priv.h"
54 #include "xfs_filestream.h"
55 #include "xfs_da_btree.h"
56 #include "xfs_extfree_item.h"
57 #include "xfs_mru_cache.h"
58 #include "xfs_inode_item.h"
59 #include "xfs_sync.h"
60 #include "xfs_trace.h"
61
62 #include <linux/namei.h>
63 #include <linux/init.h>
64 #include <linux/mount.h>
65 #include <linux/mempool.h>
66 #include <linux/writeback.h>
67 #include <linux/kthread.h>
68 #include <linux/freezer.h>
69 #include <linux/parser.h>
70
71 static const struct super_operations xfs_super_operations;
72 static kmem_zone_t *xfs_ioend_zone;
73 mempool_t *xfs_ioend_pool;
74
75 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
76 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
77 #define MNTOPT_LOGDEV "logdev" /* log device */
78 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
79 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
80 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
81 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
82 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
83 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
84 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
85 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
86 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
87 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
88 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
89 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
90 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
91 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
92 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
93 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
94 * unwritten extent conversion */
95 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
96 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
97 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
98 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
99 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
100 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
101 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
102 * in stat(). */
103 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
104 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
105 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
106 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
107 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
108 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
109 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
110 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
111 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
112 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
113 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
114 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
115 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
116 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
117 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
118 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
119 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
120 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
121
122 /*
123 * Table driven mount option parser.
124 *
125 * Currently only used for remount, but it will be used for mount
126 * in the future, too.
127 */
128 enum {
129 Opt_barrier, Opt_nobarrier, Opt_err
130 };
131
132 static const match_table_t tokens = {
133 {Opt_barrier, "barrier"},
134 {Opt_nobarrier, "nobarrier"},
135 {Opt_err, NULL}
136 };
137
138
139 STATIC unsigned long
140 suffix_strtoul(char *s, char **endp, unsigned int base)
141 {
142 int last, shift_left_factor = 0;
143 char *value = s;
144
145 last = strlen(value) - 1;
146 if (value[last] == 'K' || value[last] == 'k') {
147 shift_left_factor = 10;
148 value[last] = '\0';
149 }
150 if (value[last] == 'M' || value[last] == 'm') {
151 shift_left_factor = 20;
152 value[last] = '\0';
153 }
154 if (value[last] == 'G' || value[last] == 'g') {
155 shift_left_factor = 30;
156 value[last] = '\0';
157 }
158
159 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
160 }
161
162 /*
163 * This function fills in xfs_mount_t fields based on mount args.
164 * Note: the superblock has _not_ yet been read in.
165 *
166 * Note that this function leaks the various device name allocations on
167 * failure. The caller takes care of them.
168 */
169 STATIC int
170 xfs_parseargs(
171 struct xfs_mount *mp,
172 char *options,
173 char **mtpt)
174 {
175 struct super_block *sb = mp->m_super;
176 char *this_char, *value, *eov;
177 int dsunit = 0;
178 int dswidth = 0;
179 int iosize = 0;
180 int dmapi_implies_ikeep = 1;
181 __uint8_t iosizelog = 0;
182
183 /*
184 * Copy binary VFS mount flags we are interested in.
185 */
186 if (sb->s_flags & MS_RDONLY)
187 mp->m_flags |= XFS_MOUNT_RDONLY;
188 if (sb->s_flags & MS_DIRSYNC)
189 mp->m_flags |= XFS_MOUNT_DIRSYNC;
190 if (sb->s_flags & MS_SYNCHRONOUS)
191 mp->m_flags |= XFS_MOUNT_WSYNC;
192
193 /*
194 * Set some default flags that could be cleared by the mount option
195 * parsing.
196 */
197 mp->m_flags |= XFS_MOUNT_BARRIER;
198 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
199 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
200
201 /*
202 * These can be overridden by the mount option parsing.
203 */
204 mp->m_logbufs = -1;
205 mp->m_logbsize = -1;
206
207 if (!options)
208 goto done;
209
210 while ((this_char = strsep(&options, ",")) != NULL) {
211 if (!*this_char)
212 continue;
213 if ((value = strchr(this_char, '=')) != NULL)
214 *value++ = 0;
215
216 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
217 if (!value || !*value) {
218 cmn_err(CE_WARN,
219 "XFS: %s option requires an argument",
220 this_char);
221 return EINVAL;
222 }
223 mp->m_logbufs = simple_strtoul(value, &eov, 10);
224 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
225 if (!value || !*value) {
226 cmn_err(CE_WARN,
227 "XFS: %s option requires an argument",
228 this_char);
229 return EINVAL;
230 }
231 mp->m_logbsize = suffix_strtoul(value, &eov, 10);
232 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
233 if (!value || !*value) {
234 cmn_err(CE_WARN,
235 "XFS: %s option requires an argument",
236 this_char);
237 return EINVAL;
238 }
239 mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
240 if (!mp->m_logname)
241 return ENOMEM;
242 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
243 if (!value || !*value) {
244 cmn_err(CE_WARN,
245 "XFS: %s option requires an argument",
246 this_char);
247 return EINVAL;
248 }
249 *mtpt = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
250 if (!*mtpt)
251 return ENOMEM;
252 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
253 if (!value || !*value) {
254 cmn_err(CE_WARN,
255 "XFS: %s option requires an argument",
256 this_char);
257 return EINVAL;
258 }
259 mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
260 if (!mp->m_rtname)
261 return ENOMEM;
262 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
263 if (!value || !*value) {
264 cmn_err(CE_WARN,
265 "XFS: %s option requires an argument",
266 this_char);
267 return EINVAL;
268 }
269 iosize = simple_strtoul(value, &eov, 10);
270 iosizelog = ffs(iosize) - 1;
271 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
272 if (!value || !*value) {
273 cmn_err(CE_WARN,
274 "XFS: %s option requires an argument",
275 this_char);
276 return EINVAL;
277 }
278 iosize = suffix_strtoul(value, &eov, 10);
279 iosizelog = ffs(iosize) - 1;
280 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
281 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
282 mp->m_flags |= XFS_MOUNT_GRPID;
283 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
284 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
285 mp->m_flags &= ~XFS_MOUNT_GRPID;
286 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
287 mp->m_flags |= XFS_MOUNT_WSYNC;
288 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
289 mp->m_flags |= XFS_MOUNT_OSYNCISOSYNC;
290 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
291 mp->m_flags |= XFS_MOUNT_NORECOVERY;
292 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
293 mp->m_flags |= XFS_MOUNT_NOALIGN;
294 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
295 mp->m_flags |= XFS_MOUNT_SWALLOC;
296 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
297 if (!value || !*value) {
298 cmn_err(CE_WARN,
299 "XFS: %s option requires an argument",
300 this_char);
301 return EINVAL;
302 }
303 dsunit = simple_strtoul(value, &eov, 10);
304 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
305 if (!value || !*value) {
306 cmn_err(CE_WARN,
307 "XFS: %s option requires an argument",
308 this_char);
309 return EINVAL;
310 }
311 dswidth = simple_strtoul(value, &eov, 10);
312 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
313 mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
314 #if !XFS_BIG_INUMS
315 cmn_err(CE_WARN,
316 "XFS: %s option not allowed on this system",
317 this_char);
318 return EINVAL;
319 #endif
320 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
321 mp->m_flags |= XFS_MOUNT_NOUUID;
322 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
323 mp->m_flags |= XFS_MOUNT_BARRIER;
324 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
325 mp->m_flags &= ~XFS_MOUNT_BARRIER;
326 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
327 mp->m_flags |= XFS_MOUNT_IKEEP;
328 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
329 dmapi_implies_ikeep = 0;
330 mp->m_flags &= ~XFS_MOUNT_IKEEP;
331 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
332 mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
333 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
334 mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
335 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
336 mp->m_flags |= XFS_MOUNT_ATTR2;
337 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
338 mp->m_flags &= ~XFS_MOUNT_ATTR2;
339 mp->m_flags |= XFS_MOUNT_NOATTR2;
340 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
341 mp->m_flags |= XFS_MOUNT_FILESTREAMS;
342 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
343 mp->m_qflags &= ~(XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
344 XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
345 XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
346 XFS_UQUOTA_ENFD | XFS_OQUOTA_ENFD);
347 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
348 !strcmp(this_char, MNTOPT_UQUOTA) ||
349 !strcmp(this_char, MNTOPT_USRQUOTA)) {
350 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
351 XFS_UQUOTA_ENFD);
352 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
353 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
354 mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
355 mp->m_qflags &= ~XFS_UQUOTA_ENFD;
356 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
357 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
358 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
359 XFS_OQUOTA_ENFD);
360 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
361 mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
362 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
363 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
364 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
365 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
366 XFS_OQUOTA_ENFD);
367 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
368 mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
369 mp->m_qflags &= ~XFS_OQUOTA_ENFD;
370 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
371 mp->m_flags |= XFS_MOUNT_DMAPI;
372 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
373 mp->m_flags |= XFS_MOUNT_DMAPI;
374 } else if (!strcmp(this_char, MNTOPT_DMI)) {
375 mp->m_flags |= XFS_MOUNT_DMAPI;
376 } else if (!strcmp(this_char, "ihashsize")) {
377 cmn_err(CE_WARN,
378 "XFS: ihashsize no longer used, option is deprecated.");
379 } else if (!strcmp(this_char, "osyncisdsync")) {
380 /* no-op, this is now the default */
381 cmn_err(CE_WARN,
382 "XFS: osyncisdsync is now the default, option is deprecated.");
383 } else if (!strcmp(this_char, "irixsgid")) {
384 cmn_err(CE_WARN,
385 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
386 } else {
387 cmn_err(CE_WARN,
388 "XFS: unknown mount option [%s].", this_char);
389 return EINVAL;
390 }
391 }
392
393 /*
394 * no recovery flag requires a read-only mount
395 */
396 if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
397 !(mp->m_flags & XFS_MOUNT_RDONLY)) {
398 cmn_err(CE_WARN, "XFS: no-recovery mounts must be read-only.");
399 return EINVAL;
400 }
401
402 if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
403 cmn_err(CE_WARN,
404 "XFS: sunit and swidth options incompatible with the noalign option");
405 return EINVAL;
406 }
407
408 #ifndef CONFIG_XFS_QUOTA
409 if (XFS_IS_QUOTA_RUNNING(mp)) {
410 cmn_err(CE_WARN,
411 "XFS: quota support not available in this kernel.");
412 return EINVAL;
413 }
414 #endif
415
416 if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
417 (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE))) {
418 cmn_err(CE_WARN,
419 "XFS: cannot mount with both project and group quota");
420 return EINVAL;
421 }
422
423 if ((mp->m_flags & XFS_MOUNT_DMAPI) && (!*mtpt || *mtpt[0] == '\0')) {
424 printk("XFS: %s option needs the mount point option as well\n",
425 MNTOPT_DMAPI);
426 return EINVAL;
427 }
428
429 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
430 cmn_err(CE_WARN,
431 "XFS: sunit and swidth must be specified together");
432 return EINVAL;
433 }
434
435 if (dsunit && (dswidth % dsunit != 0)) {
436 cmn_err(CE_WARN,
437 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
438 dswidth, dsunit);
439 return EINVAL;
440 }
441
442 /*
443 * Applications using DMI filesystems often expect the
444 * inode generation number to be monotonically increasing.
445 * If we delete inode chunks we break this assumption, so
446 * keep unused inode chunks on disk for DMI filesystems
447 * until we come up with a better solution.
448 * Note that if "ikeep" or "noikeep" mount options are
449 * supplied, then they are honored.
450 */
451 if ((mp->m_flags & XFS_MOUNT_DMAPI) && dmapi_implies_ikeep)
452 mp->m_flags |= XFS_MOUNT_IKEEP;
453
454 done:
455 if (!(mp->m_flags & XFS_MOUNT_NOALIGN)) {
456 /*
457 * At this point the superblock has not been read
458 * in, therefore we do not know the block size.
459 * Before the mount call ends we will convert
460 * these to FSBs.
461 */
462 if (dsunit) {
463 mp->m_dalign = dsunit;
464 mp->m_flags |= XFS_MOUNT_RETERR;
465 }
466
467 if (dswidth)
468 mp->m_swidth = dswidth;
469 }
470
471 if (mp->m_logbufs != -1 &&
472 mp->m_logbufs != 0 &&
473 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
474 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
475 cmn_err(CE_WARN,
476 "XFS: invalid logbufs value: %d [not %d-%d]",
477 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
478 return XFS_ERROR(EINVAL);
479 }
480 if (mp->m_logbsize != -1 &&
481 mp->m_logbsize != 0 &&
482 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
483 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
484 !is_power_of_2(mp->m_logbsize))) {
485 cmn_err(CE_WARN,
486 "XFS: invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
487 mp->m_logbsize);
488 return XFS_ERROR(EINVAL);
489 }
490
491 mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
492 if (!mp->m_fsname)
493 return ENOMEM;
494 mp->m_fsname_len = strlen(mp->m_fsname) + 1;
495
496 if (iosizelog) {
497 if (iosizelog > XFS_MAX_IO_LOG ||
498 iosizelog < XFS_MIN_IO_LOG) {
499 cmn_err(CE_WARN,
500 "XFS: invalid log iosize: %d [not %d-%d]",
501 iosizelog, XFS_MIN_IO_LOG,
502 XFS_MAX_IO_LOG);
503 return XFS_ERROR(EINVAL);
504 }
505
506 mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
507 mp->m_readio_log = iosizelog;
508 mp->m_writeio_log = iosizelog;
509 }
510
511 return 0;
512 }
513
514 struct proc_xfs_info {
515 int flag;
516 char *str;
517 };
518
519 STATIC int
520 xfs_showargs(
521 struct xfs_mount *mp,
522 struct seq_file *m)
523 {
524 static struct proc_xfs_info xfs_info_set[] = {
525 /* the few simple ones we can get from the mount struct */
526 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
527 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
528 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
529 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
530 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
531 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
532 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
533 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
534 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
535 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
536 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
537 { 0, NULL }
538 };
539 static struct proc_xfs_info xfs_info_unset[] = {
540 /* the few simple ones we can get from the mount struct */
541 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
542 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
543 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
544 { 0, NULL }
545 };
546 struct proc_xfs_info *xfs_infop;
547
548 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
549 if (mp->m_flags & xfs_infop->flag)
550 seq_puts(m, xfs_infop->str);
551 }
552 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
553 if (!(mp->m_flags & xfs_infop->flag))
554 seq_puts(m, xfs_infop->str);
555 }
556
557 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
558 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
559 (int)(1 << mp->m_writeio_log) >> 10);
560
561 if (mp->m_logbufs > 0)
562 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
563 if (mp->m_logbsize > 0)
564 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
565
566 if (mp->m_logname)
567 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
568 if (mp->m_rtname)
569 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
570
571 if (mp->m_dalign > 0)
572 seq_printf(m, "," MNTOPT_SUNIT "=%d",
573 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
574 if (mp->m_swidth > 0)
575 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
576 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
577
578 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
579 seq_puts(m, "," MNTOPT_USRQUOTA);
580 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
581 seq_puts(m, "," MNTOPT_UQUOTANOENF);
582
583 /* Either project or group quotas can be active, not both */
584
585 if (mp->m_qflags & XFS_PQUOTA_ACCT) {
586 if (mp->m_qflags & XFS_OQUOTA_ENFD)
587 seq_puts(m, "," MNTOPT_PRJQUOTA);
588 else
589 seq_puts(m, "," MNTOPT_PQUOTANOENF);
590 } else if (mp->m_qflags & XFS_GQUOTA_ACCT) {
591 if (mp->m_qflags & XFS_OQUOTA_ENFD)
592 seq_puts(m, "," MNTOPT_GRPQUOTA);
593 else
594 seq_puts(m, "," MNTOPT_GQUOTANOENF);
595 }
596
597 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
598 seq_puts(m, "," MNTOPT_NOQUOTA);
599
600 return 0;
601 }
602 __uint64_t
603 xfs_max_file_offset(
604 unsigned int blockshift)
605 {
606 unsigned int pagefactor = 1;
607 unsigned int bitshift = BITS_PER_LONG - 1;
608
609 /* Figure out maximum filesize, on Linux this can depend on
610 * the filesystem blocksize (on 32 bit platforms).
611 * __block_prepare_write does this in an [unsigned] long...
612 * page->index << (PAGE_CACHE_SHIFT - bbits)
613 * So, for page sized blocks (4K on 32 bit platforms),
614 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
615 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
616 * but for smaller blocksizes it is less (bbits = log2 bsize).
617 * Note1: get_block_t takes a long (implicit cast from above)
618 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
619 * can optionally convert the [unsigned] long from above into
620 * an [unsigned] long long.
621 */
622
623 #if BITS_PER_LONG == 32
624 # if defined(CONFIG_LBDAF)
625 ASSERT(sizeof(sector_t) == 8);
626 pagefactor = PAGE_CACHE_SIZE;
627 bitshift = BITS_PER_LONG;
628 # else
629 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
630 # endif
631 #endif
632
633 return (((__uint64_t)pagefactor) << bitshift) - 1;
634 }
635
636 STATIC int
637 xfs_blkdev_get(
638 xfs_mount_t *mp,
639 const char *name,
640 struct block_device **bdevp)
641 {
642 int error = 0;
643
644 *bdevp = open_bdev_exclusive(name, FMODE_READ|FMODE_WRITE, mp);
645 if (IS_ERR(*bdevp)) {
646 error = PTR_ERR(*bdevp);
647 printk("XFS: Invalid device [%s], error=%d\n", name, error);
648 }
649
650 return -error;
651 }
652
653 STATIC void
654 xfs_blkdev_put(
655 struct block_device *bdev)
656 {
657 if (bdev)
658 close_bdev_exclusive(bdev, FMODE_READ|FMODE_WRITE);
659 }
660
661 /*
662 * Try to write out the superblock using barriers.
663 */
664 STATIC int
665 xfs_barrier_test(
666 xfs_mount_t *mp)
667 {
668 xfs_buf_t *sbp = xfs_getsb(mp, 0);
669 int error;
670
671 XFS_BUF_UNDONE(sbp);
672 XFS_BUF_UNREAD(sbp);
673 XFS_BUF_UNDELAYWRITE(sbp);
674 XFS_BUF_WRITE(sbp);
675 XFS_BUF_UNASYNC(sbp);
676 XFS_BUF_ORDERED(sbp);
677
678 xfsbdstrat(mp, sbp);
679 error = xfs_iowait(sbp);
680
681 /*
682 * Clear all the flags we set and possible error state in the
683 * buffer. We only did the write to try out whether barriers
684 * worked and shouldn't leave any traces in the superblock
685 * buffer.
686 */
687 XFS_BUF_DONE(sbp);
688 XFS_BUF_ERROR(sbp, 0);
689 XFS_BUF_UNORDERED(sbp);
690
691 xfs_buf_relse(sbp);
692 return error;
693 }
694
695 STATIC void
696 xfs_mountfs_check_barriers(xfs_mount_t *mp)
697 {
698 int error;
699
700 if (mp->m_logdev_targp != mp->m_ddev_targp) {
701 xfs_fs_cmn_err(CE_NOTE, mp,
702 "Disabling barriers, not supported with external log device");
703 mp->m_flags &= ~XFS_MOUNT_BARRIER;
704 return;
705 }
706
707 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
708 xfs_fs_cmn_err(CE_NOTE, mp,
709 "Disabling barriers, underlying device is readonly");
710 mp->m_flags &= ~XFS_MOUNT_BARRIER;
711 return;
712 }
713
714 error = xfs_barrier_test(mp);
715 if (error) {
716 xfs_fs_cmn_err(CE_NOTE, mp,
717 "Disabling barriers, trial barrier write failed");
718 mp->m_flags &= ~XFS_MOUNT_BARRIER;
719 return;
720 }
721 }
722
723 void
724 xfs_blkdev_issue_flush(
725 xfs_buftarg_t *buftarg)
726 {
727 blkdev_issue_flush(buftarg->bt_bdev, NULL);
728 }
729
730 STATIC void
731 xfs_close_devices(
732 struct xfs_mount *mp)
733 {
734 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
735 struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
736 xfs_free_buftarg(mp, mp->m_logdev_targp);
737 xfs_blkdev_put(logdev);
738 }
739 if (mp->m_rtdev_targp) {
740 struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
741 xfs_free_buftarg(mp, mp->m_rtdev_targp);
742 xfs_blkdev_put(rtdev);
743 }
744 xfs_free_buftarg(mp, mp->m_ddev_targp);
745 }
746
747 /*
748 * The file system configurations are:
749 * (1) device (partition) with data and internal log
750 * (2) logical volume with data and log subvolumes.
751 * (3) logical volume with data, log, and realtime subvolumes.
752 *
753 * We only have to handle opening the log and realtime volumes here if
754 * they are present. The data subvolume has already been opened by
755 * get_sb_bdev() and is stored in sb->s_bdev.
756 */
757 STATIC int
758 xfs_open_devices(
759 struct xfs_mount *mp)
760 {
761 struct block_device *ddev = mp->m_super->s_bdev;
762 struct block_device *logdev = NULL, *rtdev = NULL;
763 int error;
764
765 /*
766 * Open real time and log devices - order is important.
767 */
768 if (mp->m_logname) {
769 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
770 if (error)
771 goto out;
772 }
773
774 if (mp->m_rtname) {
775 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
776 if (error)
777 goto out_close_logdev;
778
779 if (rtdev == ddev || rtdev == logdev) {
780 cmn_err(CE_WARN,
781 "XFS: Cannot mount filesystem with identical rtdev and ddev/logdev.");
782 error = EINVAL;
783 goto out_close_rtdev;
784 }
785 }
786
787 /*
788 * Setup xfs_mount buffer target pointers
789 */
790 error = ENOMEM;
791 mp->m_ddev_targp = xfs_alloc_buftarg(ddev, 0);
792 if (!mp->m_ddev_targp)
793 goto out_close_rtdev;
794
795 if (rtdev) {
796 mp->m_rtdev_targp = xfs_alloc_buftarg(rtdev, 1);
797 if (!mp->m_rtdev_targp)
798 goto out_free_ddev_targ;
799 }
800
801 if (logdev && logdev != ddev) {
802 mp->m_logdev_targp = xfs_alloc_buftarg(logdev, 1);
803 if (!mp->m_logdev_targp)
804 goto out_free_rtdev_targ;
805 } else {
806 mp->m_logdev_targp = mp->m_ddev_targp;
807 }
808
809 return 0;
810
811 out_free_rtdev_targ:
812 if (mp->m_rtdev_targp)
813 xfs_free_buftarg(mp, mp->m_rtdev_targp);
814 out_free_ddev_targ:
815 xfs_free_buftarg(mp, mp->m_ddev_targp);
816 out_close_rtdev:
817 if (rtdev)
818 xfs_blkdev_put(rtdev);
819 out_close_logdev:
820 if (logdev && logdev != ddev)
821 xfs_blkdev_put(logdev);
822 out:
823 return error;
824 }
825
826 /*
827 * Setup xfs_mount buffer target pointers based on superblock
828 */
829 STATIC int
830 xfs_setup_devices(
831 struct xfs_mount *mp)
832 {
833 int error;
834
835 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_blocksize,
836 mp->m_sb.sb_sectsize);
837 if (error)
838 return error;
839
840 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
841 unsigned int log_sector_size = BBSIZE;
842
843 if (xfs_sb_version_hassector(&mp->m_sb))
844 log_sector_size = mp->m_sb.sb_logsectsize;
845 error = xfs_setsize_buftarg(mp->m_logdev_targp,
846 mp->m_sb.sb_blocksize,
847 log_sector_size);
848 if (error)
849 return error;
850 }
851 if (mp->m_rtdev_targp) {
852 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
853 mp->m_sb.sb_blocksize,
854 mp->m_sb.sb_sectsize);
855 if (error)
856 return error;
857 }
858
859 return 0;
860 }
861
862 /*
863 * XFS AIL push thread support
864 */
865 void
866 xfsaild_wakeup(
867 struct xfs_ail *ailp,
868 xfs_lsn_t threshold_lsn)
869 {
870 ailp->xa_target = threshold_lsn;
871 wake_up_process(ailp->xa_task);
872 }
873
874 STATIC int
875 xfsaild(
876 void *data)
877 {
878 struct xfs_ail *ailp = data;
879 xfs_lsn_t last_pushed_lsn = 0;
880 long tout = 0;
881
882 while (!kthread_should_stop()) {
883 if (tout)
884 schedule_timeout_interruptible(msecs_to_jiffies(tout));
885 tout = 1000;
886
887 /* swsusp */
888 try_to_freeze();
889
890 ASSERT(ailp->xa_mount->m_log);
891 if (XFS_FORCED_SHUTDOWN(ailp->xa_mount))
892 continue;
893
894 tout = xfsaild_push(ailp, &last_pushed_lsn);
895 }
896
897 return 0;
898 } /* xfsaild */
899
900 int
901 xfsaild_start(
902 struct xfs_ail *ailp)
903 {
904 ailp->xa_target = 0;
905 ailp->xa_task = kthread_run(xfsaild, ailp, "xfsaild");
906 if (IS_ERR(ailp->xa_task))
907 return -PTR_ERR(ailp->xa_task);
908 return 0;
909 }
910
911 void
912 xfsaild_stop(
913 struct xfs_ail *ailp)
914 {
915 kthread_stop(ailp->xa_task);
916 }
917
918
919 /* Catch misguided souls that try to use this interface on XFS */
920 STATIC struct inode *
921 xfs_fs_alloc_inode(
922 struct super_block *sb)
923 {
924 BUG();
925 return NULL;
926 }
927
928 /*
929 * Now that the generic code is guaranteed not to be accessing
930 * the linux inode, we can reclaim the inode.
931 */
932 STATIC void
933 xfs_fs_destroy_inode(
934 struct inode *inode)
935 {
936 struct xfs_inode *ip = XFS_I(inode);
937
938 xfs_itrace_entry(ip);
939
940 XFS_STATS_INC(vn_reclaim);
941
942 /* bad inode, get out here ASAP */
943 if (is_bad_inode(inode))
944 goto out_reclaim;
945
946 xfs_ioend_wait(ip);
947
948 ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
949
950 /*
951 * We should never get here with one of the reclaim flags already set.
952 */
953 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
954 ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
955
956 /*
957 * We always use background reclaim here because even if the
958 * inode is clean, it still may be under IO and hence we have
959 * to take the flush lock. The background reclaim path handles
960 * this more efficiently than we can here, so simply let background
961 * reclaim tear down all inodes.
962 */
963 out_reclaim:
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 intialise
972 * fields in the xfs inode that left in the initialise state
973 * when freeing the inode.
974 */
975 STATIC void
976 xfs_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_iocount, 0);
988 atomic_set(&ip->i_pincount, 0);
989 spin_lock_init(&ip->i_flags_lock);
990 init_waitqueue_head(&ip->i_ipin_wait);
991 /*
992 * Because we want to use a counting completion, complete
993 * the flush completion once to allow a single access to
994 * the flush completion without blocking.
995 */
996 init_completion(&ip->i_flush);
997 complete(&ip->i_flush);
998
999 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
1000 "xfsino", ip->i_ino);
1001 }
1002
1003 /*
1004 * Dirty the XFS inode when mark_inode_dirty_sync() is called so that
1005 * we catch unlogged VFS level updates to the inode. Care must be taken
1006 * here - the transaction code calls mark_inode_dirty_sync() to mark the
1007 * VFS inode dirty in a transaction and clears the i_update_core field;
1008 * it must clear the field after calling mark_inode_dirty_sync() to
1009 * correctly indicate that the dirty state has been propagated into the
1010 * inode log item.
1011 *
1012 * We need the barrier() to maintain correct ordering between unlogged
1013 * updates and the transaction commit code that clears the i_update_core
1014 * field. This requires all updates to be completed before marking the
1015 * inode dirty.
1016 */
1017 STATIC void
1018 xfs_fs_dirty_inode(
1019 struct inode *inode)
1020 {
1021 barrier();
1022 XFS_I(inode)->i_update_core = 1;
1023 }
1024
1025 /*
1026 * Attempt to flush the inode, this will actually fail
1027 * if the inode is pinned, but we dirty the inode again
1028 * at the point when it is unpinned after a log write,
1029 * since this is when the inode itself becomes flushable.
1030 */
1031 STATIC int
1032 xfs_fs_write_inode(
1033 struct inode *inode,
1034 int sync)
1035 {
1036 struct xfs_inode *ip = XFS_I(inode);
1037 struct xfs_mount *mp = ip->i_mount;
1038 int error = 0;
1039
1040 xfs_itrace_entry(ip);
1041
1042 if (XFS_FORCED_SHUTDOWN(mp))
1043 return XFS_ERROR(EIO);
1044
1045 if (sync) {
1046 error = xfs_wait_on_pages(ip, 0, -1);
1047 if (error)
1048 goto out;
1049 }
1050
1051 /*
1052 * Bypass inodes which have already been cleaned by
1053 * the inode flush clustering code inside xfs_iflush
1054 */
1055 if (xfs_inode_clean(ip))
1056 goto out;
1057
1058 /*
1059 * We make this non-blocking if the inode is contended, return
1060 * EAGAIN to indicate to the caller that they did not succeed.
1061 * This prevents the flush path from blocking on inodes inside
1062 * another operation right now, they get caught later by xfs_sync.
1063 */
1064 if (sync) {
1065 xfs_ilock(ip, XFS_ILOCK_SHARED);
1066 xfs_iflock(ip);
1067
1068 error = xfs_iflush(ip, XFS_IFLUSH_SYNC);
1069 } else {
1070 error = EAGAIN;
1071 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1072 goto out;
1073 if (xfs_ipincount(ip) || !xfs_iflock_nowait(ip))
1074 goto out_unlock;
1075
1076 error = xfs_iflush(ip, XFS_IFLUSH_ASYNC_NOBLOCK);
1077 }
1078
1079 out_unlock:
1080 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1081 out:
1082 /*
1083 * if we failed to write out the inode then mark
1084 * it dirty again so we'll try again later.
1085 */
1086 if (error)
1087 xfs_mark_inode_dirty_sync(ip);
1088 return -error;
1089 }
1090
1091 STATIC void
1092 xfs_fs_clear_inode(
1093 struct inode *inode)
1094 {
1095 xfs_inode_t *ip = XFS_I(inode);
1096
1097 xfs_itrace_entry(ip);
1098 XFS_STATS_INC(vn_rele);
1099 XFS_STATS_INC(vn_remove);
1100 XFS_STATS_DEC(vn_active);
1101
1102 /*
1103 * The iolock is used by the file system to coordinate reads,
1104 * writes, and block truncates. Up to this point the lock
1105 * protected concurrent accesses by users of the inode. But
1106 * from here forward we're doing some final processing of the
1107 * inode because we're done with it, and although we reuse the
1108 * iolock for protection it is really a distinct lock class
1109 * (in the lockdep sense) from before. To keep lockdep happy
1110 * (and basically indicate what we are doing), we explicitly
1111 * re-init the iolock here.
1112 */
1113 ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1114 mrlock_init(&ip->i_iolock, MRLOCK_BARRIER, "xfsio", ip->i_ino);
1115
1116 xfs_inactive(ip);
1117 }
1118
1119 STATIC void
1120 xfs_free_fsname(
1121 struct xfs_mount *mp)
1122 {
1123 kfree(mp->m_fsname);
1124 kfree(mp->m_rtname);
1125 kfree(mp->m_logname);
1126 }
1127
1128 STATIC void
1129 xfs_fs_put_super(
1130 struct super_block *sb)
1131 {
1132 struct xfs_mount *mp = XFS_M(sb);
1133
1134 xfs_syncd_stop(mp);
1135
1136 if (!(sb->s_flags & MS_RDONLY)) {
1137 /*
1138 * XXX(hch): this should be SYNC_WAIT.
1139 *
1140 * Or more likely not needed at all because the VFS is already
1141 * calling ->sync_fs after shutting down all filestem
1142 * operations and just before calling ->put_super.
1143 */
1144 xfs_sync_data(mp, 0);
1145 xfs_sync_attr(mp, 0);
1146 }
1147
1148 XFS_SEND_PREUNMOUNT(mp);
1149
1150 /*
1151 * Blow away any referenced inode in the filestreams cache.
1152 * This can and will cause log traffic as inodes go inactive
1153 * here.
1154 */
1155 xfs_filestream_unmount(mp);
1156
1157 XFS_bflush(mp->m_ddev_targp);
1158
1159 XFS_SEND_UNMOUNT(mp);
1160
1161 xfs_unmountfs(mp);
1162 xfs_freesb(mp);
1163 xfs_icsb_destroy_counters(mp);
1164 xfs_close_devices(mp);
1165 xfs_dmops_put(mp);
1166 xfs_free_fsname(mp);
1167 kfree(mp);
1168 }
1169
1170 STATIC int
1171 xfs_fs_sync_fs(
1172 struct super_block *sb,
1173 int wait)
1174 {
1175 struct xfs_mount *mp = XFS_M(sb);
1176 int error;
1177
1178 /*
1179 * Not much we can do for the first async pass. Writing out the
1180 * superblock would be counter-productive as we are going to redirty
1181 * when writing out other data and metadata (and writing out a single
1182 * block is quite fast anyway).
1183 *
1184 * Try to asynchronously kick off quota syncing at least.
1185 */
1186 if (!wait) {
1187 xfs_qm_sync(mp, SYNC_TRYLOCK);
1188 return 0;
1189 }
1190
1191 error = xfs_quiesce_data(mp);
1192 if (error)
1193 return -error;
1194
1195 if (laptop_mode) {
1196 int prev_sync_seq = mp->m_sync_seq;
1197
1198 /*
1199 * The disk must be active because we're syncing.
1200 * We schedule xfssyncd now (now that the disk is
1201 * active) instead of later (when it might not be).
1202 */
1203 wake_up_process(mp->m_sync_task);
1204 /*
1205 * We have to wait for the sync iteration to complete.
1206 * If we don't, the disk activity caused by the sync
1207 * will come after the sync is completed, and that
1208 * triggers another sync from laptop mode.
1209 */
1210 wait_event(mp->m_wait_single_sync_task,
1211 mp->m_sync_seq != prev_sync_seq);
1212 }
1213
1214 return 0;
1215 }
1216
1217 STATIC int
1218 xfs_fs_statfs(
1219 struct dentry *dentry,
1220 struct kstatfs *statp)
1221 {
1222 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1223 xfs_sb_t *sbp = &mp->m_sb;
1224 struct xfs_inode *ip = XFS_I(dentry->d_inode);
1225 __uint64_t fakeinos, id;
1226 xfs_extlen_t lsize;
1227
1228 statp->f_type = XFS_SB_MAGIC;
1229 statp->f_namelen = MAXNAMELEN - 1;
1230
1231 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1232 statp->f_fsid.val[0] = (u32)id;
1233 statp->f_fsid.val[1] = (u32)(id >> 32);
1234
1235 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1236
1237 spin_lock(&mp->m_sb_lock);
1238 statp->f_bsize = sbp->sb_blocksize;
1239 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1240 statp->f_blocks = sbp->sb_dblocks - lsize;
1241 statp->f_bfree = statp->f_bavail =
1242 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1243 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1244 statp->f_files =
1245 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1246 if (mp->m_maxicount)
1247 statp->f_files = min_t(typeof(statp->f_files),
1248 statp->f_files,
1249 mp->m_maxicount);
1250 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1251 spin_unlock(&mp->m_sb_lock);
1252
1253 if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) ||
1254 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))) ==
1255 (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
1256 xfs_qm_statvfs(ip, statp);
1257 return 0;
1258 }
1259
1260 STATIC int
1261 xfs_fs_remount(
1262 struct super_block *sb,
1263 int *flags,
1264 char *options)
1265 {
1266 struct xfs_mount *mp = XFS_M(sb);
1267 substring_t args[MAX_OPT_ARGS];
1268 char *p;
1269 int error;
1270
1271 while ((p = strsep(&options, ",")) != NULL) {
1272 int token;
1273
1274 if (!*p)
1275 continue;
1276
1277 token = match_token(p, tokens, args);
1278 switch (token) {
1279 case Opt_barrier:
1280 mp->m_flags |= XFS_MOUNT_BARRIER;
1281
1282 /*
1283 * Test if barriers are actually working if we can,
1284 * else delay this check until the filesystem is
1285 * marked writeable.
1286 */
1287 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1288 xfs_mountfs_check_barriers(mp);
1289 break;
1290 case Opt_nobarrier:
1291 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1292 break;
1293 default:
1294 /*
1295 * Logically we would return an error here to prevent
1296 * users from believing they might have changed
1297 * mount options using remount which can't be changed.
1298 *
1299 * But unfortunately mount(8) adds all options from
1300 * mtab and fstab to the mount arguments in some cases
1301 * so we can't blindly reject options, but have to
1302 * check for each specified option if it actually
1303 * differs from the currently set option and only
1304 * reject it if that's the case.
1305 *
1306 * Until that is implemented we return success for
1307 * every remount request, and silently ignore all
1308 * options that we can't actually change.
1309 */
1310 #if 0
1311 printk(KERN_INFO
1312 "XFS: mount option \"%s\" not supported for remount\n", p);
1313 return -EINVAL;
1314 #else
1315 break;
1316 #endif
1317 }
1318 }
1319
1320 /* ro -> rw */
1321 if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1322 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1323 if (mp->m_flags & XFS_MOUNT_BARRIER)
1324 xfs_mountfs_check_barriers(mp);
1325
1326 /*
1327 * If this is the first remount to writeable state we
1328 * might have some superblock changes to update.
1329 */
1330 if (mp->m_update_flags) {
1331 error = xfs_mount_log_sb(mp, mp->m_update_flags);
1332 if (error) {
1333 cmn_err(CE_WARN,
1334 "XFS: failed to write sb changes");
1335 return error;
1336 }
1337 mp->m_update_flags = 0;
1338 }
1339 }
1340
1341 /* rw -> ro */
1342 if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1343 xfs_quiesce_data(mp);
1344 xfs_quiesce_attr(mp);
1345 mp->m_flags |= XFS_MOUNT_RDONLY;
1346 }
1347
1348 return 0;
1349 }
1350
1351 /*
1352 * Second stage of a freeze. The data is already frozen so we only
1353 * need to take care of the metadata. Once that's done write a dummy
1354 * record to dirty the log in case of a crash while frozen.
1355 */
1356 STATIC int
1357 xfs_fs_freeze(
1358 struct super_block *sb)
1359 {
1360 struct xfs_mount *mp = XFS_M(sb);
1361
1362 xfs_quiesce_attr(mp);
1363 return -xfs_fs_log_dummy(mp);
1364 }
1365
1366 STATIC int
1367 xfs_fs_show_options(
1368 struct seq_file *m,
1369 struct vfsmount *mnt)
1370 {
1371 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1372 }
1373
1374 /*
1375 * This function fills in xfs_mount_t fields based on mount args.
1376 * Note: the superblock _has_ now been read in.
1377 */
1378 STATIC int
1379 xfs_finish_flags(
1380 struct xfs_mount *mp)
1381 {
1382 int ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1383
1384 /* Fail a mount where the logbuf is smaller than the log stripe */
1385 if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1386 if (mp->m_logbsize <= 0 &&
1387 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1388 mp->m_logbsize = mp->m_sb.sb_logsunit;
1389 } else if (mp->m_logbsize > 0 &&
1390 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1391 cmn_err(CE_WARN,
1392 "XFS: logbuf size must be greater than or equal to log stripe size");
1393 return XFS_ERROR(EINVAL);
1394 }
1395 } else {
1396 /* Fail a mount if the logbuf is larger than 32K */
1397 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1398 cmn_err(CE_WARN,
1399 "XFS: logbuf size for version 1 logs must be 16K or 32K");
1400 return XFS_ERROR(EINVAL);
1401 }
1402 }
1403
1404 /*
1405 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1406 * told by noattr2 to turn it off
1407 */
1408 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1409 !(mp->m_flags & XFS_MOUNT_NOATTR2))
1410 mp->m_flags |= XFS_MOUNT_ATTR2;
1411
1412 /*
1413 * prohibit r/w mounts of read-only filesystems
1414 */
1415 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1416 cmn_err(CE_WARN,
1417 "XFS: cannot mount a read-only filesystem as read-write");
1418 return XFS_ERROR(EROFS);
1419 }
1420
1421 return 0;
1422 }
1423
1424 STATIC int
1425 xfs_fs_fill_super(
1426 struct super_block *sb,
1427 void *data,
1428 int silent)
1429 {
1430 struct inode *root;
1431 struct xfs_mount *mp = NULL;
1432 int flags = 0, error = ENOMEM;
1433 char *mtpt = NULL;
1434
1435 mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1436 if (!mp)
1437 goto out;
1438
1439 spin_lock_init(&mp->m_sb_lock);
1440 mutex_init(&mp->m_growlock);
1441 atomic_set(&mp->m_active_trans, 0);
1442 INIT_LIST_HEAD(&mp->m_sync_list);
1443 spin_lock_init(&mp->m_sync_lock);
1444 init_waitqueue_head(&mp->m_wait_single_sync_task);
1445
1446 mp->m_super = sb;
1447 sb->s_fs_info = mp;
1448
1449 error = xfs_parseargs(mp, (char *)data, &mtpt);
1450 if (error)
1451 goto out_free_fsname;
1452
1453 sb_min_blocksize(sb, BBSIZE);
1454 sb->s_xattr = xfs_xattr_handlers;
1455 sb->s_export_op = &xfs_export_operations;
1456 #ifdef CONFIG_XFS_QUOTA
1457 sb->s_qcop = &xfs_quotactl_operations;
1458 #endif
1459 sb->s_op = &xfs_super_operations;
1460
1461 error = xfs_dmops_get(mp);
1462 if (error)
1463 goto out_free_fsname;
1464
1465 if (silent)
1466 flags |= XFS_MFSI_QUIET;
1467
1468 error = xfs_open_devices(mp);
1469 if (error)
1470 goto out_put_dmops;
1471
1472 if (xfs_icsb_init_counters(mp))
1473 mp->m_flags |= XFS_MOUNT_NO_PERCPU_SB;
1474
1475 error = xfs_readsb(mp, flags);
1476 if (error)
1477 goto out_destroy_counters;
1478
1479 error = xfs_finish_flags(mp);
1480 if (error)
1481 goto out_free_sb;
1482
1483 error = xfs_setup_devices(mp);
1484 if (error)
1485 goto out_free_sb;
1486
1487 if (mp->m_flags & XFS_MOUNT_BARRIER)
1488 xfs_mountfs_check_barriers(mp);
1489
1490 error = xfs_filestream_mount(mp);
1491 if (error)
1492 goto out_free_sb;
1493
1494 error = xfs_mountfs(mp);
1495 if (error)
1496 goto out_filestream_unmount;
1497
1498 XFS_SEND_MOUNT(mp, DM_RIGHT_NULL, mtpt, mp->m_fsname);
1499
1500 sb->s_magic = XFS_SB_MAGIC;
1501 sb->s_blocksize = mp->m_sb.sb_blocksize;
1502 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1503 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1504 sb->s_time_gran = 1;
1505 set_posix_acl_flag(sb);
1506
1507 root = igrab(VFS_I(mp->m_rootip));
1508 if (!root) {
1509 error = ENOENT;
1510 goto fail_unmount;
1511 }
1512 if (is_bad_inode(root)) {
1513 error = EINVAL;
1514 goto fail_vnrele;
1515 }
1516 sb->s_root = d_alloc_root(root);
1517 if (!sb->s_root) {
1518 error = ENOMEM;
1519 goto fail_vnrele;
1520 }
1521
1522 error = xfs_syncd_init(mp);
1523 if (error)
1524 goto fail_vnrele;
1525
1526 kfree(mtpt);
1527 return 0;
1528
1529 out_filestream_unmount:
1530 xfs_filestream_unmount(mp);
1531 out_free_sb:
1532 xfs_freesb(mp);
1533 out_destroy_counters:
1534 xfs_icsb_destroy_counters(mp);
1535 xfs_close_devices(mp);
1536 out_put_dmops:
1537 xfs_dmops_put(mp);
1538 out_free_fsname:
1539 xfs_free_fsname(mp);
1540 kfree(mtpt);
1541 kfree(mp);
1542 out:
1543 return -error;
1544
1545 fail_vnrele:
1546 if (sb->s_root) {
1547 dput(sb->s_root);
1548 sb->s_root = NULL;
1549 } else {
1550 iput(root);
1551 }
1552
1553 fail_unmount:
1554 /*
1555 * Blow away any referenced inode in the filestreams cache.
1556 * This can and will cause log traffic as inodes go inactive
1557 * here.
1558 */
1559 xfs_filestream_unmount(mp);
1560
1561 XFS_bflush(mp->m_ddev_targp);
1562
1563 xfs_unmountfs(mp);
1564 goto out_free_sb;
1565 }
1566
1567 STATIC int
1568 xfs_fs_get_sb(
1569 struct file_system_type *fs_type,
1570 int flags,
1571 const char *dev_name,
1572 void *data,
1573 struct vfsmount *mnt)
1574 {
1575 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1576 mnt);
1577 }
1578
1579 static const struct super_operations xfs_super_operations = {
1580 .alloc_inode = xfs_fs_alloc_inode,
1581 .destroy_inode = xfs_fs_destroy_inode,
1582 .dirty_inode = xfs_fs_dirty_inode,
1583 .write_inode = xfs_fs_write_inode,
1584 .clear_inode = xfs_fs_clear_inode,
1585 .put_super = xfs_fs_put_super,
1586 .sync_fs = xfs_fs_sync_fs,
1587 .freeze_fs = xfs_fs_freeze,
1588 .statfs = xfs_fs_statfs,
1589 .remount_fs = xfs_fs_remount,
1590 .show_options = xfs_fs_show_options,
1591 };
1592
1593 static struct file_system_type xfs_fs_type = {
1594 .owner = THIS_MODULE,
1595 .name = "xfs",
1596 .get_sb = xfs_fs_get_sb,
1597 .kill_sb = kill_block_super,
1598 .fs_flags = FS_REQUIRES_DEV,
1599 };
1600
1601 STATIC int __init
1602 xfs_init_zones(void)
1603 {
1604
1605 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1606 if (!xfs_ioend_zone)
1607 goto out;
1608
1609 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1610 xfs_ioend_zone);
1611 if (!xfs_ioend_pool)
1612 goto out_destroy_ioend_zone;
1613
1614 xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1615 "xfs_log_ticket");
1616 if (!xfs_log_ticket_zone)
1617 goto out_destroy_ioend_pool;
1618
1619 xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1620 "xfs_bmap_free_item");
1621 if (!xfs_bmap_free_item_zone)
1622 goto out_destroy_log_ticket_zone;
1623
1624 xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1625 "xfs_btree_cur");
1626 if (!xfs_btree_cur_zone)
1627 goto out_destroy_bmap_free_item_zone;
1628
1629 xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1630 "xfs_da_state");
1631 if (!xfs_da_state_zone)
1632 goto out_destroy_btree_cur_zone;
1633
1634 xfs_dabuf_zone = kmem_zone_init(sizeof(xfs_dabuf_t), "xfs_dabuf");
1635 if (!xfs_dabuf_zone)
1636 goto out_destroy_da_state_zone;
1637
1638 xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1639 if (!xfs_ifork_zone)
1640 goto out_destroy_dabuf_zone;
1641
1642 xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1643 if (!xfs_trans_zone)
1644 goto out_destroy_ifork_zone;
1645
1646 /*
1647 * The size of the zone allocated buf log item is the maximum
1648 * size possible under XFS. This wastes a little bit of memory,
1649 * but it is much faster.
1650 */
1651 xfs_buf_item_zone = kmem_zone_init((sizeof(xfs_buf_log_item_t) +
1652 (((XFS_MAX_BLOCKSIZE / XFS_BLI_CHUNK) /
1653 NBWORD) * sizeof(int))), "xfs_buf_item");
1654 if (!xfs_buf_item_zone)
1655 goto out_destroy_trans_zone;
1656
1657 xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1658 ((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1659 sizeof(xfs_extent_t))), "xfs_efd_item");
1660 if (!xfs_efd_zone)
1661 goto out_destroy_buf_item_zone;
1662
1663 xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1664 ((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1665 sizeof(xfs_extent_t))), "xfs_efi_item");
1666 if (!xfs_efi_zone)
1667 goto out_destroy_efd_zone;
1668
1669 xfs_inode_zone =
1670 kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1671 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1672 xfs_fs_inode_init_once);
1673 if (!xfs_inode_zone)
1674 goto out_destroy_efi_zone;
1675
1676 xfs_ili_zone =
1677 kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1678 KM_ZONE_SPREAD, NULL);
1679 if (!xfs_ili_zone)
1680 goto out_destroy_inode_zone;
1681
1682 return 0;
1683
1684 out_destroy_inode_zone:
1685 kmem_zone_destroy(xfs_inode_zone);
1686 out_destroy_efi_zone:
1687 kmem_zone_destroy(xfs_efi_zone);
1688 out_destroy_efd_zone:
1689 kmem_zone_destroy(xfs_efd_zone);
1690 out_destroy_buf_item_zone:
1691 kmem_zone_destroy(xfs_buf_item_zone);
1692 out_destroy_trans_zone:
1693 kmem_zone_destroy(xfs_trans_zone);
1694 out_destroy_ifork_zone:
1695 kmem_zone_destroy(xfs_ifork_zone);
1696 out_destroy_dabuf_zone:
1697 kmem_zone_destroy(xfs_dabuf_zone);
1698 out_destroy_da_state_zone:
1699 kmem_zone_destroy(xfs_da_state_zone);
1700 out_destroy_btree_cur_zone:
1701 kmem_zone_destroy(xfs_btree_cur_zone);
1702 out_destroy_bmap_free_item_zone:
1703 kmem_zone_destroy(xfs_bmap_free_item_zone);
1704 out_destroy_log_ticket_zone:
1705 kmem_zone_destroy(xfs_log_ticket_zone);
1706 out_destroy_ioend_pool:
1707 mempool_destroy(xfs_ioend_pool);
1708 out_destroy_ioend_zone:
1709 kmem_zone_destroy(xfs_ioend_zone);
1710 out:
1711 return -ENOMEM;
1712 }
1713
1714 STATIC void
1715 xfs_destroy_zones(void)
1716 {
1717 kmem_zone_destroy(xfs_ili_zone);
1718 kmem_zone_destroy(xfs_inode_zone);
1719 kmem_zone_destroy(xfs_efi_zone);
1720 kmem_zone_destroy(xfs_efd_zone);
1721 kmem_zone_destroy(xfs_buf_item_zone);
1722 kmem_zone_destroy(xfs_trans_zone);
1723 kmem_zone_destroy(xfs_ifork_zone);
1724 kmem_zone_destroy(xfs_dabuf_zone);
1725 kmem_zone_destroy(xfs_da_state_zone);
1726 kmem_zone_destroy(xfs_btree_cur_zone);
1727 kmem_zone_destroy(xfs_bmap_free_item_zone);
1728 kmem_zone_destroy(xfs_log_ticket_zone);
1729 mempool_destroy(xfs_ioend_pool);
1730 kmem_zone_destroy(xfs_ioend_zone);
1731
1732 }
1733
1734 STATIC int __init
1735 init_xfs_fs(void)
1736 {
1737 int error;
1738
1739 printk(KERN_INFO XFS_VERSION_STRING " with "
1740 XFS_BUILD_OPTIONS " enabled\n");
1741
1742 xfs_ioend_init();
1743 xfs_dir_startup();
1744
1745 error = xfs_init_zones();
1746 if (error)
1747 goto out;
1748
1749 error = xfs_mru_cache_init();
1750 if (error)
1751 goto out_destroy_zones;
1752
1753 error = xfs_filestream_init();
1754 if (error)
1755 goto out_mru_cache_uninit;
1756
1757 error = xfs_buf_init();
1758 if (error)
1759 goto out_filestream_uninit;
1760
1761 error = xfs_init_procfs();
1762 if (error)
1763 goto out_buf_terminate;
1764
1765 error = xfs_sysctl_register();
1766 if (error)
1767 goto out_cleanup_procfs;
1768
1769 vfs_initquota();
1770
1771 error = register_filesystem(&xfs_fs_type);
1772 if (error)
1773 goto out_sysctl_unregister;
1774 return 0;
1775
1776 out_sysctl_unregister:
1777 xfs_sysctl_unregister();
1778 out_cleanup_procfs:
1779 xfs_cleanup_procfs();
1780 out_buf_terminate:
1781 xfs_buf_terminate();
1782 out_filestream_uninit:
1783 xfs_filestream_uninit();
1784 out_mru_cache_uninit:
1785 xfs_mru_cache_uninit();
1786 out_destroy_zones:
1787 xfs_destroy_zones();
1788 out:
1789 return error;
1790 }
1791
1792 STATIC void __exit
1793 exit_xfs_fs(void)
1794 {
1795 vfs_exitquota();
1796 unregister_filesystem(&xfs_fs_type);
1797 xfs_sysctl_unregister();
1798 xfs_cleanup_procfs();
1799 xfs_buf_terminate();
1800 xfs_filestream_uninit();
1801 xfs_mru_cache_uninit();
1802 xfs_destroy_zones();
1803 }
1804
1805 module_init(init_xfs_fs);
1806 module_exit(exit_xfs_fs);
1807
1808 MODULE_AUTHOR("Silicon Graphics, Inc.");
1809 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1810 MODULE_LICENSE("GPL");
Linux with added FPC-III support