[XFS] merge xfs_unmount into xfs_fs_put_super / xfs_fs_fill_super
[deliverable/linux.git] / fs / xfs / linux-2.6 / xfs_super.c
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 #include "xfs.h"
19 #include "xfs_bit.h"
20 #include "xfs_log.h"
21 #include "xfs_clnt.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_ialloc.h"
40 #include "xfs_bmap.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
44 #include "xfs_fsops.h"
45 #include "xfs_rw.h"
46 #include "xfs_acl.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_vfsops.h"
52 #include "xfs_version.h"
53 #include "xfs_log_priv.h"
54 #include "xfs_trans_priv.h"
55 #include "xfs_filestream.h"
56
57 #include <linux/namei.h>
58 #include <linux/init.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
65 static struct quotactl_ops xfs_quotactl_operations;
66 static struct super_operations xfs_super_operations;
67 static kmem_zone_t *xfs_vnode_zone;
68 static kmem_zone_t *xfs_ioend_zone;
69 mempool_t *xfs_ioend_pool;
70
71 STATIC struct xfs_mount_args *
72 xfs_args_allocate(
73 struct super_block *sb,
74 int silent)
75 {
76 struct xfs_mount_args *args;
77
78 args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP);
79 args->logbufs = args->logbufsize = -1;
80 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
81
82 /* Copy the already-parsed mount(2) flags we're interested in */
83 if (sb->s_flags & MS_DIRSYNC)
84 args->flags |= XFSMNT_DIRSYNC;
85 if (sb->s_flags & MS_SYNCHRONOUS)
86 args->flags |= XFSMNT_WSYNC;
87 if (silent)
88 args->flags |= XFSMNT_QUIET;
89 args->flags |= XFSMNT_32BITINODES;
90
91 return args;
92 }
93
94 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
95 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
96 #define MNTOPT_LOGDEV "logdev" /* log device */
97 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
98 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
99 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
100 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
101 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
102 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
103 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
104 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
105 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
106 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
107 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
108 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
109 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
110 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
111 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
112 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
113 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
114 * unwritten extent conversion */
115 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
116 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
117 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
118 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
119 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
120 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
121 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
122 * in stat(). */
123 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
124 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
125 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
126 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
127 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
128 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
129 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
130 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
131 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
132 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
133 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
134 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
135 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
136 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
137 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
138 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
139 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
140 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
141
142 STATIC unsigned long
143 suffix_strtoul(char *s, char **endp, unsigned int base)
144 {
145 int last, shift_left_factor = 0;
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 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
163 }
164
165 STATIC int
166 xfs_parseargs(
167 struct xfs_mount *mp,
168 char *options,
169 struct xfs_mount_args *args,
170 int update)
171 {
172 char *this_char, *value, *eov;
173 int dsunit, dswidth, vol_dsunit, vol_dswidth;
174 int iosize;
175 int dmapi_implies_ikeep = 1;
176
177 args->flags |= XFSMNT_BARRIER;
178 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
179
180 if (!options)
181 goto done;
182
183 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
184
185 while ((this_char = strsep(&options, ",")) != NULL) {
186 if (!*this_char)
187 continue;
188 if ((value = strchr(this_char, '=')) != NULL)
189 *value++ = 0;
190
191 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
192 if (!value || !*value) {
193 cmn_err(CE_WARN,
194 "XFS: %s option requires an argument",
195 this_char);
196 return EINVAL;
197 }
198 args->logbufs = simple_strtoul(value, &eov, 10);
199 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
200 if (!value || !*value) {
201 cmn_err(CE_WARN,
202 "XFS: %s option requires an argument",
203 this_char);
204 return EINVAL;
205 }
206 args->logbufsize = suffix_strtoul(value, &eov, 10);
207 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
208 if (!value || !*value) {
209 cmn_err(CE_WARN,
210 "XFS: %s option requires an argument",
211 this_char);
212 return EINVAL;
213 }
214 strncpy(args->logname, value, MAXNAMELEN);
215 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
216 if (!value || !*value) {
217 cmn_err(CE_WARN,
218 "XFS: %s option requires an argument",
219 this_char);
220 return EINVAL;
221 }
222 strncpy(args->mtpt, value, MAXNAMELEN);
223 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
224 if (!value || !*value) {
225 cmn_err(CE_WARN,
226 "XFS: %s option requires an argument",
227 this_char);
228 return EINVAL;
229 }
230 strncpy(args->rtname, value, MAXNAMELEN);
231 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
232 if (!value || !*value) {
233 cmn_err(CE_WARN,
234 "XFS: %s option requires an argument",
235 this_char);
236 return EINVAL;
237 }
238 iosize = simple_strtoul(value, &eov, 10);
239 args->flags |= XFSMNT_IOSIZE;
240 args->iosizelog = (uint8_t) iosize;
241 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
242 if (!value || !*value) {
243 cmn_err(CE_WARN,
244 "XFS: %s option requires an argument",
245 this_char);
246 return EINVAL;
247 }
248 iosize = suffix_strtoul(value, &eov, 10);
249 args->flags |= XFSMNT_IOSIZE;
250 args->iosizelog = ffs(iosize) - 1;
251 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
252 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
253 mp->m_flags |= XFS_MOUNT_GRPID;
254 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
255 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
256 mp->m_flags &= ~XFS_MOUNT_GRPID;
257 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
258 args->flags |= XFSMNT_WSYNC;
259 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
260 args->flags |= XFSMNT_OSYNCISOSYNC;
261 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
262 args->flags |= XFSMNT_NORECOVERY;
263 } else if (!strcmp(this_char, MNTOPT_INO64)) {
264 args->flags |= XFSMNT_INO64;
265 #if !XFS_BIG_INUMS
266 cmn_err(CE_WARN,
267 "XFS: %s option not allowed on this system",
268 this_char);
269 return EINVAL;
270 #endif
271 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
272 args->flags |= XFSMNT_NOALIGN;
273 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
274 args->flags |= XFSMNT_SWALLOC;
275 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
276 if (!value || !*value) {
277 cmn_err(CE_WARN,
278 "XFS: %s option requires an argument",
279 this_char);
280 return EINVAL;
281 }
282 dsunit = simple_strtoul(value, &eov, 10);
283 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
284 if (!value || !*value) {
285 cmn_err(CE_WARN,
286 "XFS: %s option requires an argument",
287 this_char);
288 return EINVAL;
289 }
290 dswidth = simple_strtoul(value, &eov, 10);
291 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
292 args->flags &= ~XFSMNT_32BITINODES;
293 #if !XFS_BIG_INUMS
294 cmn_err(CE_WARN,
295 "XFS: %s option not allowed on this system",
296 this_char);
297 return EINVAL;
298 #endif
299 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
300 args->flags |= XFSMNT_NOUUID;
301 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
302 args->flags |= XFSMNT_BARRIER;
303 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
304 args->flags &= ~XFSMNT_BARRIER;
305 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
306 args->flags |= XFSMNT_IKEEP;
307 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
308 dmapi_implies_ikeep = 0;
309 args->flags &= ~XFSMNT_IKEEP;
310 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
311 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
312 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
313 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
314 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
315 args->flags |= XFSMNT_ATTR2;
316 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
317 args->flags &= ~XFSMNT_ATTR2;
318 args->flags |= XFSMNT_NOATTR2;
319 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
320 args->flags2 |= XFSMNT2_FILESTREAMS;
321 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
322 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
323 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
324 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
325 !strcmp(this_char, MNTOPT_UQUOTA) ||
326 !strcmp(this_char, MNTOPT_USRQUOTA)) {
327 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
328 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
329 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
330 args->flags |= XFSMNT_UQUOTA;
331 args->flags &= ~XFSMNT_UQUOTAENF;
332 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
333 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
334 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
335 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
336 args->flags |= XFSMNT_PQUOTA;
337 args->flags &= ~XFSMNT_PQUOTAENF;
338 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
339 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
340 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
341 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
342 args->flags |= XFSMNT_GQUOTA;
343 args->flags &= ~XFSMNT_GQUOTAENF;
344 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
345 args->flags |= XFSMNT_DMAPI;
346 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
347 args->flags |= XFSMNT_DMAPI;
348 } else if (!strcmp(this_char, MNTOPT_DMI)) {
349 args->flags |= XFSMNT_DMAPI;
350 } else if (!strcmp(this_char, "ihashsize")) {
351 cmn_err(CE_WARN,
352 "XFS: ihashsize no longer used, option is deprecated.");
353 } else if (!strcmp(this_char, "osyncisdsync")) {
354 /* no-op, this is now the default */
355 cmn_err(CE_WARN,
356 "XFS: osyncisdsync is now the default, option is deprecated.");
357 } else if (!strcmp(this_char, "irixsgid")) {
358 cmn_err(CE_WARN,
359 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
360 } else {
361 cmn_err(CE_WARN,
362 "XFS: unknown mount option [%s].", this_char);
363 return EINVAL;
364 }
365 }
366
367 if (args->flags & XFSMNT_NORECOVERY) {
368 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
369 cmn_err(CE_WARN,
370 "XFS: no-recovery mounts must be read-only.");
371 return EINVAL;
372 }
373 }
374
375 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
376 cmn_err(CE_WARN,
377 "XFS: sunit and swidth options incompatible with the noalign option");
378 return EINVAL;
379 }
380
381 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
382 cmn_err(CE_WARN,
383 "XFS: cannot mount with both project and group quota");
384 return EINVAL;
385 }
386
387 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
388 printk("XFS: %s option needs the mount point option as well\n",
389 MNTOPT_DMAPI);
390 return EINVAL;
391 }
392
393 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
394 cmn_err(CE_WARN,
395 "XFS: sunit and swidth must be specified together");
396 return EINVAL;
397 }
398
399 if (dsunit && (dswidth % dsunit != 0)) {
400 cmn_err(CE_WARN,
401 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
402 dswidth, dsunit);
403 return EINVAL;
404 }
405
406 /*
407 * Applications using DMI filesystems often expect the
408 * inode generation number to be monotonically increasing.
409 * If we delete inode chunks we break this assumption, so
410 * keep unused inode chunks on disk for DMI filesystems
411 * until we come up with a better solution.
412 * Note that if "ikeep" or "noikeep" mount options are
413 * supplied, then they are honored.
414 */
415 if ((args->flags & XFSMNT_DMAPI) && dmapi_implies_ikeep)
416 args->flags |= XFSMNT_IKEEP;
417
418 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
419 if (dsunit) {
420 args->sunit = dsunit;
421 args->flags |= XFSMNT_RETERR;
422 } else {
423 args->sunit = vol_dsunit;
424 }
425 dswidth ? (args->swidth = dswidth) :
426 (args->swidth = vol_dswidth);
427 } else {
428 args->sunit = args->swidth = 0;
429 }
430
431 done:
432 if (args->flags & XFSMNT_32BITINODES)
433 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
434 if (args->flags2)
435 args->flags |= XFSMNT_FLAGS2;
436 return 0;
437 }
438
439 struct proc_xfs_info {
440 int flag;
441 char *str;
442 };
443
444 STATIC int
445 xfs_showargs(
446 struct xfs_mount *mp,
447 struct seq_file *m)
448 {
449 static struct proc_xfs_info xfs_info_set[] = {
450 /* the few simple ones we can get from the mount struct */
451 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
452 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
453 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
454 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
455 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
456 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
457 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
458 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
459 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
460 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
461 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
462 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
463 { 0, NULL }
464 };
465 static struct proc_xfs_info xfs_info_unset[] = {
466 /* the few simple ones we can get from the mount struct */
467 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
468 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
469 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
470 { 0, NULL }
471 };
472 struct proc_xfs_info *xfs_infop;
473
474 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
475 if (mp->m_flags & xfs_infop->flag)
476 seq_puts(m, xfs_infop->str);
477 }
478 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
479 if (!(mp->m_flags & xfs_infop->flag))
480 seq_puts(m, xfs_infop->str);
481 }
482
483 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
484 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
485 (int)(1 << mp->m_writeio_log) >> 10);
486
487 if (mp->m_logbufs > 0)
488 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
489 if (mp->m_logbsize > 0)
490 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
491
492 if (mp->m_logname)
493 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
494 if (mp->m_rtname)
495 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
496
497 if (mp->m_dalign > 0)
498 seq_printf(m, "," MNTOPT_SUNIT "=%d",
499 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
500 if (mp->m_swidth > 0)
501 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
502 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
503
504 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
505 seq_puts(m, "," MNTOPT_USRQUOTA);
506 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
507 seq_puts(m, "," MNTOPT_UQUOTANOENF);
508
509 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
510 seq_puts(m, "," MNTOPT_PRJQUOTA);
511 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
512 seq_puts(m, "," MNTOPT_PQUOTANOENF);
513
514 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
515 seq_puts(m, "," MNTOPT_GRPQUOTA);
516 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
517 seq_puts(m, "," MNTOPT_GQUOTANOENF);
518
519 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
520 seq_puts(m, "," MNTOPT_NOQUOTA);
521
522 return 0;
523 }
524 __uint64_t
525 xfs_max_file_offset(
526 unsigned int blockshift)
527 {
528 unsigned int pagefactor = 1;
529 unsigned int bitshift = BITS_PER_LONG - 1;
530
531 /* Figure out maximum filesize, on Linux this can depend on
532 * the filesystem blocksize (on 32 bit platforms).
533 * __block_prepare_write does this in an [unsigned] long...
534 * page->index << (PAGE_CACHE_SHIFT - bbits)
535 * So, for page sized blocks (4K on 32 bit platforms),
536 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
537 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
538 * but for smaller blocksizes it is less (bbits = log2 bsize).
539 * Note1: get_block_t takes a long (implicit cast from above)
540 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
541 * can optionally convert the [unsigned] long from above into
542 * an [unsigned] long long.
543 */
544
545 #if BITS_PER_LONG == 32
546 # if defined(CONFIG_LBD)
547 ASSERT(sizeof(sector_t) == 8);
548 pagefactor = PAGE_CACHE_SIZE;
549 bitshift = BITS_PER_LONG;
550 # else
551 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
552 # endif
553 #endif
554
555 return (((__uint64_t)pagefactor) << bitshift) - 1;
556 }
557
558 STATIC_INLINE void
559 xfs_set_inodeops(
560 struct inode *inode)
561 {
562 switch (inode->i_mode & S_IFMT) {
563 case S_IFREG:
564 inode->i_op = &xfs_inode_operations;
565 inode->i_fop = &xfs_file_operations;
566 inode->i_mapping->a_ops = &xfs_address_space_operations;
567 break;
568 case S_IFDIR:
569 inode->i_op = &xfs_dir_inode_operations;
570 inode->i_fop = &xfs_dir_file_operations;
571 break;
572 case S_IFLNK:
573 inode->i_op = &xfs_symlink_inode_operations;
574 if (!(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE))
575 inode->i_mapping->a_ops = &xfs_address_space_operations;
576 break;
577 default:
578 inode->i_op = &xfs_inode_operations;
579 init_special_inode(inode, inode->i_mode, inode->i_rdev);
580 break;
581 }
582 }
583
584 STATIC_INLINE void
585 xfs_revalidate_inode(
586 xfs_mount_t *mp,
587 bhv_vnode_t *vp,
588 xfs_inode_t *ip)
589 {
590 struct inode *inode = vn_to_inode(vp);
591
592 inode->i_mode = ip->i_d.di_mode;
593 inode->i_nlink = ip->i_d.di_nlink;
594 inode->i_uid = ip->i_d.di_uid;
595 inode->i_gid = ip->i_d.di_gid;
596
597 switch (inode->i_mode & S_IFMT) {
598 case S_IFBLK:
599 case S_IFCHR:
600 inode->i_rdev =
601 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
602 sysv_minor(ip->i_df.if_u2.if_rdev));
603 break;
604 default:
605 inode->i_rdev = 0;
606 break;
607 }
608
609 inode->i_generation = ip->i_d.di_gen;
610 i_size_write(inode, ip->i_d.di_size);
611 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
612 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
613 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
614 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
615 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
616 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
617 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
618 inode->i_flags |= S_IMMUTABLE;
619 else
620 inode->i_flags &= ~S_IMMUTABLE;
621 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
622 inode->i_flags |= S_APPEND;
623 else
624 inode->i_flags &= ~S_APPEND;
625 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
626 inode->i_flags |= S_SYNC;
627 else
628 inode->i_flags &= ~S_SYNC;
629 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
630 inode->i_flags |= S_NOATIME;
631 else
632 inode->i_flags &= ~S_NOATIME;
633 xfs_iflags_clear(ip, XFS_IMODIFIED);
634 }
635
636 void
637 xfs_initialize_vnode(
638 struct xfs_mount *mp,
639 bhv_vnode_t *vp,
640 struct xfs_inode *ip)
641 {
642 struct inode *inode = vn_to_inode(vp);
643
644 if (!ip->i_vnode) {
645 ip->i_vnode = vp;
646 inode->i_private = ip;
647 }
648
649 /*
650 * We need to set the ops vectors, and unlock the inode, but if
651 * we have been called during the new inode create process, it is
652 * too early to fill in the Linux inode. We will get called a
653 * second time once the inode is properly set up, and then we can
654 * finish our work.
655 */
656 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
657 xfs_revalidate_inode(mp, vp, ip);
658 xfs_set_inodeops(inode);
659
660 xfs_iflags_clear(ip, XFS_INEW);
661 barrier();
662
663 unlock_new_inode(inode);
664 }
665 }
666
667 int
668 xfs_blkdev_get(
669 xfs_mount_t *mp,
670 const char *name,
671 struct block_device **bdevp)
672 {
673 int error = 0;
674
675 *bdevp = open_bdev_excl(name, 0, mp);
676 if (IS_ERR(*bdevp)) {
677 error = PTR_ERR(*bdevp);
678 printk("XFS: Invalid device [%s], error=%d\n", name, error);
679 }
680
681 return -error;
682 }
683
684 void
685 xfs_blkdev_put(
686 struct block_device *bdev)
687 {
688 if (bdev)
689 close_bdev_excl(bdev);
690 }
691
692 /*
693 * Try to write out the superblock using barriers.
694 */
695 STATIC int
696 xfs_barrier_test(
697 xfs_mount_t *mp)
698 {
699 xfs_buf_t *sbp = xfs_getsb(mp, 0);
700 int error;
701
702 XFS_BUF_UNDONE(sbp);
703 XFS_BUF_UNREAD(sbp);
704 XFS_BUF_UNDELAYWRITE(sbp);
705 XFS_BUF_WRITE(sbp);
706 XFS_BUF_UNASYNC(sbp);
707 XFS_BUF_ORDERED(sbp);
708
709 xfsbdstrat(mp, sbp);
710 error = xfs_iowait(sbp);
711
712 /*
713 * Clear all the flags we set and possible error state in the
714 * buffer. We only did the write to try out whether barriers
715 * worked and shouldn't leave any traces in the superblock
716 * buffer.
717 */
718 XFS_BUF_DONE(sbp);
719 XFS_BUF_ERROR(sbp, 0);
720 XFS_BUF_UNORDERED(sbp);
721
722 xfs_buf_relse(sbp);
723 return error;
724 }
725
726 void
727 xfs_mountfs_check_barriers(xfs_mount_t *mp)
728 {
729 int error;
730
731 if (mp->m_logdev_targp != mp->m_ddev_targp) {
732 xfs_fs_cmn_err(CE_NOTE, mp,
733 "Disabling barriers, not supported with external log device");
734 mp->m_flags &= ~XFS_MOUNT_BARRIER;
735 return;
736 }
737
738 if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered ==
739 QUEUE_ORDERED_NONE) {
740 xfs_fs_cmn_err(CE_NOTE, mp,
741 "Disabling barriers, not supported by the underlying device");
742 mp->m_flags &= ~XFS_MOUNT_BARRIER;
743 return;
744 }
745
746 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
747 xfs_fs_cmn_err(CE_NOTE, mp,
748 "Disabling barriers, underlying device is readonly");
749 mp->m_flags &= ~XFS_MOUNT_BARRIER;
750 return;
751 }
752
753 error = xfs_barrier_test(mp);
754 if (error) {
755 xfs_fs_cmn_err(CE_NOTE, mp,
756 "Disabling barriers, trial barrier write failed");
757 mp->m_flags &= ~XFS_MOUNT_BARRIER;
758 return;
759 }
760 }
761
762 void
763 xfs_blkdev_issue_flush(
764 xfs_buftarg_t *buftarg)
765 {
766 blkdev_issue_flush(buftarg->bt_bdev, NULL);
767 }
768
769 /*
770 * XFS AIL push thread support
771 */
772 void
773 xfsaild_wakeup(
774 xfs_mount_t *mp,
775 xfs_lsn_t threshold_lsn)
776 {
777 mp->m_ail.xa_target = threshold_lsn;
778 wake_up_process(mp->m_ail.xa_task);
779 }
780
781 int
782 xfsaild(
783 void *data)
784 {
785 xfs_mount_t *mp = (xfs_mount_t *)data;
786 xfs_lsn_t last_pushed_lsn = 0;
787 long tout = 0;
788
789 while (!kthread_should_stop()) {
790 if (tout)
791 schedule_timeout_interruptible(msecs_to_jiffies(tout));
792 tout = 1000;
793
794 /* swsusp */
795 try_to_freeze();
796
797 ASSERT(mp->m_log);
798 if (XFS_FORCED_SHUTDOWN(mp))
799 continue;
800
801 tout = xfsaild_push(mp, &last_pushed_lsn);
802 }
803
804 return 0;
805 } /* xfsaild */
806
807 int
808 xfsaild_start(
809 xfs_mount_t *mp)
810 {
811 mp->m_ail.xa_target = 0;
812 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
813 if (IS_ERR(mp->m_ail.xa_task))
814 return -PTR_ERR(mp->m_ail.xa_task);
815 return 0;
816 }
817
818 void
819 xfsaild_stop(
820 xfs_mount_t *mp)
821 {
822 kthread_stop(mp->m_ail.xa_task);
823 }
824
825
826
827 STATIC struct inode *
828 xfs_fs_alloc_inode(
829 struct super_block *sb)
830 {
831 bhv_vnode_t *vp;
832
833 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
834 if (unlikely(!vp))
835 return NULL;
836 return vn_to_inode(vp);
837 }
838
839 STATIC void
840 xfs_fs_destroy_inode(
841 struct inode *inode)
842 {
843 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
844 }
845
846 STATIC void
847 xfs_fs_inode_init_once(
848 void *vnode)
849 {
850 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
851 }
852
853 STATIC int __init
854 xfs_init_zones(void)
855 {
856 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
857 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
858 KM_ZONE_SPREAD,
859 xfs_fs_inode_init_once);
860 if (!xfs_vnode_zone)
861 goto out;
862
863 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
864 if (!xfs_ioend_zone)
865 goto out_destroy_vnode_zone;
866
867 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
868 xfs_ioend_zone);
869 if (!xfs_ioend_pool)
870 goto out_free_ioend_zone;
871 return 0;
872
873 out_free_ioend_zone:
874 kmem_zone_destroy(xfs_ioend_zone);
875 out_destroy_vnode_zone:
876 kmem_zone_destroy(xfs_vnode_zone);
877 out:
878 return -ENOMEM;
879 }
880
881 STATIC void
882 xfs_destroy_zones(void)
883 {
884 mempool_destroy(xfs_ioend_pool);
885 kmem_zone_destroy(xfs_vnode_zone);
886 kmem_zone_destroy(xfs_ioend_zone);
887 }
888
889 /*
890 * Attempt to flush the inode, this will actually fail
891 * if the inode is pinned, but we dirty the inode again
892 * at the point when it is unpinned after a log write,
893 * since this is when the inode itself becomes flushable.
894 */
895 STATIC int
896 xfs_fs_write_inode(
897 struct inode *inode,
898 int sync)
899 {
900 int error = 0;
901 int flags = 0;
902
903 xfs_itrace_entry(XFS_I(inode));
904 if (sync) {
905 filemap_fdatawait(inode->i_mapping);
906 flags |= FLUSH_SYNC;
907 }
908 error = xfs_inode_flush(XFS_I(inode), flags);
909 /*
910 * if we failed to write out the inode then mark
911 * it dirty again so we'll try again later.
912 */
913 if (error)
914 mark_inode_dirty_sync(inode);
915
916 return -error;
917 }
918
919 STATIC void
920 xfs_fs_clear_inode(
921 struct inode *inode)
922 {
923 xfs_inode_t *ip = XFS_I(inode);
924
925 /*
926 * ip can be null when xfs_iget_core calls xfs_idestroy if we
927 * find an inode with di_mode == 0 but without IGET_CREATE set.
928 */
929 if (ip) {
930 xfs_itrace_entry(ip);
931 XFS_STATS_INC(vn_rele);
932 XFS_STATS_INC(vn_remove);
933 XFS_STATS_INC(vn_reclaim);
934 XFS_STATS_DEC(vn_active);
935
936 xfs_inactive(ip);
937 xfs_iflags_clear(ip, XFS_IMODIFIED);
938 if (xfs_reclaim(ip))
939 panic("%s: cannot reclaim 0x%p\n", __func__, inode);
940 }
941
942 ASSERT(XFS_I(inode) == NULL);
943 }
944
945 /*
946 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
947 * Doing this has two advantages:
948 * - It saves on stack space, which is tight in certain situations
949 * - It can be used (with care) as a mechanism to avoid deadlocks.
950 * Flushing while allocating in a full filesystem requires both.
951 */
952 STATIC void
953 xfs_syncd_queue_work(
954 struct xfs_mount *mp,
955 void *data,
956 void (*syncer)(struct xfs_mount *, void *))
957 {
958 struct bhv_vfs_sync_work *work;
959
960 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
961 INIT_LIST_HEAD(&work->w_list);
962 work->w_syncer = syncer;
963 work->w_data = data;
964 work->w_mount = mp;
965 spin_lock(&mp->m_sync_lock);
966 list_add_tail(&work->w_list, &mp->m_sync_list);
967 spin_unlock(&mp->m_sync_lock);
968 wake_up_process(mp->m_sync_task);
969 }
970
971 /*
972 * Flush delayed allocate data, attempting to free up reserved space
973 * from existing allocations. At this point a new allocation attempt
974 * has failed with ENOSPC and we are in the process of scratching our
975 * heads, looking about for more room...
976 */
977 STATIC void
978 xfs_flush_inode_work(
979 struct xfs_mount *mp,
980 void *arg)
981 {
982 struct inode *inode = arg;
983 filemap_flush(inode->i_mapping);
984 iput(inode);
985 }
986
987 void
988 xfs_flush_inode(
989 xfs_inode_t *ip)
990 {
991 struct inode *inode = ip->i_vnode;
992
993 igrab(inode);
994 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
995 delay(msecs_to_jiffies(500));
996 }
997
998 /*
999 * This is the "bigger hammer" version of xfs_flush_inode_work...
1000 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
1001 */
1002 STATIC void
1003 xfs_flush_device_work(
1004 struct xfs_mount *mp,
1005 void *arg)
1006 {
1007 struct inode *inode = arg;
1008 sync_blockdev(mp->m_super->s_bdev);
1009 iput(inode);
1010 }
1011
1012 void
1013 xfs_flush_device(
1014 xfs_inode_t *ip)
1015 {
1016 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
1017
1018 igrab(inode);
1019 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1020 delay(msecs_to_jiffies(500));
1021 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1022 }
1023
1024 STATIC void
1025 xfs_sync_worker(
1026 struct xfs_mount *mp,
1027 void *unused)
1028 {
1029 int error;
1030
1031 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1032 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR);
1033 mp->m_sync_seq++;
1034 wake_up(&mp->m_wait_single_sync_task);
1035 }
1036
1037 STATIC int
1038 xfssyncd(
1039 void *arg)
1040 {
1041 struct xfs_mount *mp = arg;
1042 long timeleft;
1043 bhv_vfs_sync_work_t *work, *n;
1044 LIST_HEAD (tmp);
1045
1046 set_freezable();
1047 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1048 for (;;) {
1049 timeleft = schedule_timeout_interruptible(timeleft);
1050 /* swsusp */
1051 try_to_freeze();
1052 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1053 break;
1054
1055 spin_lock(&mp->m_sync_lock);
1056 /*
1057 * We can get woken by laptop mode, to do a sync -
1058 * that's the (only!) case where the list would be
1059 * empty with time remaining.
1060 */
1061 if (!timeleft || list_empty(&mp->m_sync_list)) {
1062 if (!timeleft)
1063 timeleft = xfs_syncd_centisecs *
1064 msecs_to_jiffies(10);
1065 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1066 list_add_tail(&mp->m_sync_work.w_list,
1067 &mp->m_sync_list);
1068 }
1069 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1070 list_move(&work->w_list, &tmp);
1071 spin_unlock(&mp->m_sync_lock);
1072
1073 list_for_each_entry_safe(work, n, &tmp, w_list) {
1074 (*work->w_syncer)(mp, work->w_data);
1075 list_del(&work->w_list);
1076 if (work == &mp->m_sync_work)
1077 continue;
1078 kmem_free(work);
1079 }
1080 }
1081
1082 return 0;
1083 }
1084
1085 STATIC void
1086 xfs_fs_put_super(
1087 struct super_block *sb)
1088 {
1089 struct xfs_mount *mp = XFS_M(sb);
1090 struct xfs_inode *rip = mp->m_rootip;
1091 int unmount_event_flags = 0;
1092 int error;
1093
1094 kthread_stop(mp->m_sync_task);
1095
1096 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1097
1098 #ifdef HAVE_DMAPI
1099 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1100 unmount_event_flags =
1101 (mp->m_dmevmask & (1 << DM_EVENT_UNMOUNT)) ?
1102 0 : DM_FLAGS_UNWANTED;
1103 /*
1104 * Ignore error from dmapi here, first unmount is not allowed
1105 * to fail anyway, and second we wouldn't want to fail a
1106 * unmount because of dmapi.
1107 */
1108 XFS_SEND_PREUNMOUNT(mp, rip, DM_RIGHT_NULL, rip, DM_RIGHT_NULL,
1109 NULL, NULL, 0, 0, unmount_event_flags);
1110 }
1111 #endif
1112
1113 /*
1114 * Blow away any referenced inode in the filestreams cache.
1115 * This can and will cause log traffic as inodes go inactive
1116 * here.
1117 */
1118 xfs_filestream_unmount(mp);
1119
1120 XFS_bflush(mp->m_ddev_targp);
1121 error = xfs_unmount_flush(mp, 0);
1122 WARN_ON(error);
1123
1124 IRELE(rip);
1125
1126 /*
1127 * If we're forcing a shutdown, typically because of a media error,
1128 * we want to make sure we invalidate dirty pages that belong to
1129 * referenced vnodes as well.
1130 */
1131 if (XFS_FORCED_SHUTDOWN(mp)) {
1132 error = xfs_sync(mp, SYNC_WAIT | SYNC_CLOSE);
1133 ASSERT(error != EFSCORRUPTED);
1134 }
1135
1136 if (mp->m_flags & XFS_MOUNT_DMAPI) {
1137 XFS_SEND_UNMOUNT(mp, rip, DM_RIGHT_NULL, 0, 0,
1138 unmount_event_flags);
1139 }
1140
1141 xfs_unmountfs(mp, NULL);
1142 xfs_qmops_put(mp);
1143 xfs_dmops_put(mp);
1144 kmem_free(mp);
1145 }
1146
1147 STATIC void
1148 xfs_fs_write_super(
1149 struct super_block *sb)
1150 {
1151 if (!(sb->s_flags & MS_RDONLY))
1152 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1153 sb->s_dirt = 0;
1154 }
1155
1156 STATIC int
1157 xfs_fs_sync_super(
1158 struct super_block *sb,
1159 int wait)
1160 {
1161 struct xfs_mount *mp = XFS_M(sb);
1162 int error;
1163 int flags;
1164
1165 /*
1166 * Treat a sync operation like a freeze. This is to work
1167 * around a race in sync_inodes() which works in two phases
1168 * - an asynchronous flush, which can write out an inode
1169 * without waiting for file size updates to complete, and a
1170 * synchronous flush, which wont do anything because the
1171 * async flush removed the inode's dirty flag. Also
1172 * sync_inodes() will not see any files that just have
1173 * outstanding transactions to be flushed because we don't
1174 * dirty the Linux inode until after the transaction I/O
1175 * completes.
1176 */
1177 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1178 /*
1179 * First stage of freeze - no more writers will make progress
1180 * now we are here, so we flush delwri and delalloc buffers
1181 * here, then wait for all I/O to complete. Data is frozen at
1182 * that point. Metadata is not frozen, transactions can still
1183 * occur here so don't bother flushing the buftarg (i.e
1184 * SYNC_QUIESCE) because it'll just get dirty again.
1185 */
1186 flags = SYNC_DATA_QUIESCE;
1187 } else
1188 flags = SYNC_FSDATA;
1189
1190 error = xfs_sync(mp, flags);
1191 sb->s_dirt = 0;
1192
1193 if (unlikely(laptop_mode)) {
1194 int prev_sync_seq = mp->m_sync_seq;
1195
1196 /*
1197 * The disk must be active because we're syncing.
1198 * We schedule xfssyncd now (now that the disk is
1199 * active) instead of later (when it might not be).
1200 */
1201 wake_up_process(mp->m_sync_task);
1202 /*
1203 * We have to wait for the sync iteration to complete.
1204 * If we don't, the disk activity caused by the sync
1205 * will come after the sync is completed, and that
1206 * triggers another sync from laptop mode.
1207 */
1208 wait_event(mp->m_wait_single_sync_task,
1209 mp->m_sync_seq != prev_sync_seq);
1210 }
1211
1212 return -error;
1213 }
1214
1215 STATIC int
1216 xfs_fs_statfs(
1217 struct dentry *dentry,
1218 struct kstatfs *statp)
1219 {
1220 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1221 xfs_sb_t *sbp = &mp->m_sb;
1222 __uint64_t fakeinos, id;
1223 xfs_extlen_t lsize;
1224
1225 statp->f_type = XFS_SB_MAGIC;
1226 statp->f_namelen = MAXNAMELEN - 1;
1227
1228 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1229 statp->f_fsid.val[0] = (u32)id;
1230 statp->f_fsid.val[1] = (u32)(id >> 32);
1231
1232 xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1233
1234 spin_lock(&mp->m_sb_lock);
1235 statp->f_bsize = sbp->sb_blocksize;
1236 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1237 statp->f_blocks = sbp->sb_dblocks - lsize;
1238 statp->f_bfree = statp->f_bavail =
1239 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1240 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1241 #if XFS_BIG_INUMS
1242 fakeinos += mp->m_inoadd;
1243 #endif
1244 statp->f_files =
1245 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1246 if (mp->m_maxicount)
1247 #if XFS_BIG_INUMS
1248 if (!mp->m_inoadd)
1249 #endif
1250 statp->f_files = min_t(typeof(statp->f_files),
1251 statp->f_files,
1252 mp->m_maxicount);
1253 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1254 spin_unlock(&mp->m_sb_lock);
1255
1256 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), 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 struct xfs_mount_args *args = xfs_args_allocate(sb, 0);
1268 int error;
1269
1270 error = xfs_parseargs(mp, options, args, 1);
1271 if (error)
1272 goto out_free_args;
1273
1274 if (!(*flags & MS_RDONLY)) { /* rw/ro -> rw */
1275 if (mp->m_flags & XFS_MOUNT_RDONLY)
1276 mp->m_flags &= ~XFS_MOUNT_RDONLY;
1277 if (args->flags & XFSMNT_BARRIER) {
1278 mp->m_flags |= XFS_MOUNT_BARRIER;
1279 xfs_mountfs_check_barriers(mp);
1280 } else {
1281 mp->m_flags &= ~XFS_MOUNT_BARRIER;
1282 }
1283 } else if (!(mp->m_flags & XFS_MOUNT_RDONLY)) { /* rw -> ro */
1284 xfs_filestream_flush(mp);
1285 xfs_sync(mp, SYNC_DATA_QUIESCE);
1286 xfs_attr_quiesce(mp);
1287 mp->m_flags |= XFS_MOUNT_RDONLY;
1288 }
1289
1290 out_free_args:
1291 kmem_free(args);
1292 return -error;
1293 }
1294
1295 /*
1296 * Second stage of a freeze. The data is already frozen so we only
1297 * need to take care of themetadata. Once that's done write a dummy
1298 * record to dirty the log in case of a crash while frozen.
1299 */
1300 STATIC void
1301 xfs_fs_lockfs(
1302 struct super_block *sb)
1303 {
1304 struct xfs_mount *mp = XFS_M(sb);
1305
1306 xfs_attr_quiesce(mp);
1307 xfs_fs_log_dummy(mp);
1308 }
1309
1310 STATIC int
1311 xfs_fs_show_options(
1312 struct seq_file *m,
1313 struct vfsmount *mnt)
1314 {
1315 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1316 }
1317
1318 STATIC int
1319 xfs_fs_quotasync(
1320 struct super_block *sb,
1321 int type)
1322 {
1323 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1324 }
1325
1326 STATIC int
1327 xfs_fs_getxstate(
1328 struct super_block *sb,
1329 struct fs_quota_stat *fqs)
1330 {
1331 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1332 }
1333
1334 STATIC int
1335 xfs_fs_setxstate(
1336 struct super_block *sb,
1337 unsigned int flags,
1338 int op)
1339 {
1340 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1341 }
1342
1343 STATIC int
1344 xfs_fs_getxquota(
1345 struct super_block *sb,
1346 int type,
1347 qid_t id,
1348 struct fs_disk_quota *fdq)
1349 {
1350 return -XFS_QM_QUOTACTL(XFS_M(sb),
1351 (type == USRQUOTA) ? Q_XGETQUOTA :
1352 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1353 Q_XGETPQUOTA), id, (caddr_t)fdq);
1354 }
1355
1356 STATIC int
1357 xfs_fs_setxquota(
1358 struct super_block *sb,
1359 int type,
1360 qid_t id,
1361 struct fs_disk_quota *fdq)
1362 {
1363 return -XFS_QM_QUOTACTL(XFS_M(sb),
1364 (type == USRQUOTA) ? Q_XSETQLIM :
1365 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1366 Q_XSETPQLIM), id, (caddr_t)fdq);
1367 }
1368
1369 STATIC int
1370 xfs_fs_fill_super(
1371 struct super_block *sb,
1372 void *data,
1373 int silent)
1374 {
1375 struct inode *root;
1376 struct xfs_mount *mp = NULL;
1377 struct xfs_mount_args *args = xfs_args_allocate(sb, silent);
1378 int error;
1379
1380 mp = xfs_mount_init();
1381
1382 INIT_LIST_HEAD(&mp->m_sync_list);
1383 spin_lock_init(&mp->m_sync_lock);
1384 init_waitqueue_head(&mp->m_wait_single_sync_task);
1385
1386 mp->m_super = sb;
1387 sb->s_fs_info = mp;
1388
1389 if (sb->s_flags & MS_RDONLY)
1390 mp->m_flags |= XFS_MOUNT_RDONLY;
1391
1392 error = xfs_parseargs(mp, (char *)data, args, 0);
1393 if (error)
1394 goto fail_vfsop;
1395
1396 sb_min_blocksize(sb, BBSIZE);
1397 sb->s_export_op = &xfs_export_operations;
1398 sb->s_qcop = &xfs_quotactl_operations;
1399 sb->s_op = &xfs_super_operations;
1400
1401 error = xfs_mount(mp, args, NULL);
1402 if (error)
1403 goto fail_vfsop;
1404
1405 sb->s_dirt = 1;
1406 sb->s_magic = XFS_SB_MAGIC;
1407 sb->s_blocksize = mp->m_sb.sb_blocksize;
1408 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1409 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1410 sb->s_time_gran = 1;
1411 set_posix_acl_flag(sb);
1412
1413 root = igrab(mp->m_rootip->i_vnode);
1414 if (!root) {
1415 error = ENOENT;
1416 goto fail_unmount;
1417 }
1418 if (is_bad_inode(root)) {
1419 error = EINVAL;
1420 goto fail_vnrele;
1421 }
1422 sb->s_root = d_alloc_root(root);
1423 if (!sb->s_root) {
1424 error = ENOMEM;
1425 goto fail_vnrele;
1426 }
1427
1428 mp->m_sync_work.w_syncer = xfs_sync_worker;
1429 mp->m_sync_work.w_mount = mp;
1430 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1431 if (IS_ERR(mp->m_sync_task)) {
1432 error = -PTR_ERR(mp->m_sync_task);
1433 goto fail_vnrele;
1434 }
1435
1436 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1437
1438 kmem_free(args);
1439 return 0;
1440
1441 fail_vnrele:
1442 if (sb->s_root) {
1443 dput(sb->s_root);
1444 sb->s_root = NULL;
1445 } else {
1446 iput(root);
1447 }
1448
1449 fail_unmount:
1450 /*
1451 * Blow away any referenced inode in the filestreams cache.
1452 * This can and will cause log traffic as inodes go inactive
1453 * here.
1454 */
1455 xfs_filestream_unmount(mp);
1456
1457 XFS_bflush(mp->m_ddev_targp);
1458 error = xfs_unmount_flush(mp, 0);
1459 WARN_ON(error);
1460
1461 IRELE(mp->m_rootip);
1462
1463 xfs_unmountfs(mp, NULL);
1464 xfs_qmops_put(mp);
1465 xfs_dmops_put(mp);
1466 kmem_free(mp);
1467
1468 fail_vfsop:
1469 kmem_free(args);
1470 return -error;
1471 }
1472
1473 STATIC int
1474 xfs_fs_get_sb(
1475 struct file_system_type *fs_type,
1476 int flags,
1477 const char *dev_name,
1478 void *data,
1479 struct vfsmount *mnt)
1480 {
1481 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1482 mnt);
1483 }
1484
1485 static struct super_operations xfs_super_operations = {
1486 .alloc_inode = xfs_fs_alloc_inode,
1487 .destroy_inode = xfs_fs_destroy_inode,
1488 .write_inode = xfs_fs_write_inode,
1489 .clear_inode = xfs_fs_clear_inode,
1490 .put_super = xfs_fs_put_super,
1491 .write_super = xfs_fs_write_super,
1492 .sync_fs = xfs_fs_sync_super,
1493 .write_super_lockfs = xfs_fs_lockfs,
1494 .statfs = xfs_fs_statfs,
1495 .remount_fs = xfs_fs_remount,
1496 .show_options = xfs_fs_show_options,
1497 };
1498
1499 static struct quotactl_ops xfs_quotactl_operations = {
1500 .quota_sync = xfs_fs_quotasync,
1501 .get_xstate = xfs_fs_getxstate,
1502 .set_xstate = xfs_fs_setxstate,
1503 .get_xquota = xfs_fs_getxquota,
1504 .set_xquota = xfs_fs_setxquota,
1505 };
1506
1507 static struct file_system_type xfs_fs_type = {
1508 .owner = THIS_MODULE,
1509 .name = "xfs",
1510 .get_sb = xfs_fs_get_sb,
1511 .kill_sb = kill_block_super,
1512 .fs_flags = FS_REQUIRES_DEV,
1513 };
1514
1515
1516 STATIC int __init
1517 init_xfs_fs( void )
1518 {
1519 int error;
1520 static char message[] __initdata = KERN_INFO \
1521 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
1522
1523 printk(message);
1524
1525 ktrace_init(64);
1526
1527 error = xfs_init_zones();
1528 if (error < 0)
1529 goto undo_zones;
1530
1531 error = xfs_buf_init();
1532 if (error < 0)
1533 goto undo_buffers;
1534
1535 vn_init();
1536 xfs_init();
1537 uuid_init();
1538 vfs_initquota();
1539
1540 error = register_filesystem(&xfs_fs_type);
1541 if (error)
1542 goto undo_register;
1543 return 0;
1544
1545 undo_register:
1546 xfs_buf_terminate();
1547
1548 undo_buffers:
1549 xfs_destroy_zones();
1550
1551 undo_zones:
1552 return error;
1553 }
1554
1555 STATIC void __exit
1556 exit_xfs_fs( void )
1557 {
1558 vfs_exitquota();
1559 unregister_filesystem(&xfs_fs_type);
1560 xfs_cleanup();
1561 xfs_buf_terminate();
1562 xfs_destroy_zones();
1563 ktrace_uninit();
1564 }
1565
1566 module_init(init_xfs_fs);
1567 module_exit(exit_xfs_fs);
1568
1569 MODULE_AUTHOR("Silicon Graphics, Inc.");
1570 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1571 MODULE_LICENSE("GPL");
This page took 0.06394 seconds and 6 git commands to generate.