?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
sched: retune wake granularity
[wrt350n-kernel.git] / fs / xfs / linux-2.6 / xfs_super.c
blob8831d95187904f0ecf4a1aa1ac5b77cc767aec77
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
56 #include <linux/namei.h>
57 #include <linux/init.h>
58 #include <linux/mount.h>
59 #include <linux/mempool.h>
60 #include <linux/writeback.h>
61 #include <linux/kthread.h>
62 #include <linux/freezer.h>
63
64 static struct quotactl_ops xfs_quotactl_operations;
65 static struct super_operations xfs_super_operations;
66 static kmem_zone_t *xfs_vnode_zone;
67 static kmem_zone_t *xfs_ioend_zone;
68 mempool_t *xfs_ioend_pool;
69
70 STATIC struct xfs_mount_args *
71 xfs_args_allocate(
72 struct super_block *sb,
73 int silent)
74 {
75 struct xfs_mount_args *args;
76
77 args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP);
78 args->logbufs = args->logbufsize = -1;
79 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
80
81 /* Copy the already-parsed mount(2) flags we're interested in */
82 if (sb->s_flags & MS_DIRSYNC)
83 args->flags |= XFSMNT_DIRSYNC;
84 if (sb->s_flags & MS_SYNCHRONOUS)
85 args->flags |= XFSMNT_WSYNC;
86 if (silent)
87 args->flags |= XFSMNT_QUIET;
88 args->flags |= XFSMNT_32BITINODES;
89
90 return args;
91 }
92
93 #define MNTOPT_LOGBUFS "logbufs" /* number of XFS log buffers */
94 #define MNTOPT_LOGBSIZE "logbsize" /* size of XFS log buffers */
95 #define MNTOPT_LOGDEV "logdev" /* log device */
96 #define MNTOPT_RTDEV "rtdev" /* realtime I/O device */
97 #define MNTOPT_BIOSIZE "biosize" /* log2 of preferred buffered io size */
98 #define MNTOPT_WSYNC "wsync" /* safe-mode nfs compatible mount */
99 #define MNTOPT_INO64 "ino64" /* force inodes into 64-bit range */
100 #define MNTOPT_NOALIGN "noalign" /* turn off stripe alignment */
101 #define MNTOPT_SWALLOC "swalloc" /* turn on stripe width allocation */
102 #define MNTOPT_SUNIT "sunit" /* data volume stripe unit */
103 #define MNTOPT_SWIDTH "swidth" /* data volume stripe width */
104 #define MNTOPT_NOUUID "nouuid" /* ignore filesystem UUID */
105 #define MNTOPT_MTPT "mtpt" /* filesystem mount point */
106 #define MNTOPT_GRPID "grpid" /* group-ID from parent directory */
107 #define MNTOPT_NOGRPID "nogrpid" /* group-ID from current process */
108 #define MNTOPT_BSDGROUPS "bsdgroups" /* group-ID from parent directory */
109 #define MNTOPT_SYSVGROUPS "sysvgroups" /* group-ID from current process */
110 #define MNTOPT_ALLOCSIZE "allocsize" /* preferred allocation size */
111 #define MNTOPT_NORECOVERY "norecovery" /* don't run XFS recovery */
112 #define MNTOPT_BARRIER "barrier" /* use writer barriers for log write and
113 * unwritten extent conversion */
114 #define MNTOPT_NOBARRIER "nobarrier" /* .. disable */
115 #define MNTOPT_OSYNCISOSYNC "osyncisosync" /* o_sync is REALLY o_sync */
116 #define MNTOPT_64BITINODE "inode64" /* inodes can be allocated anywhere */
117 #define MNTOPT_IKEEP "ikeep" /* do not free empty inode clusters */
118 #define MNTOPT_NOIKEEP "noikeep" /* free empty inode clusters */
119 #define MNTOPT_LARGEIO "largeio" /* report large I/O sizes in stat() */
120 #define MNTOPT_NOLARGEIO "nolargeio" /* do not report large I/O sizes
121 * in stat(). */
122 #define MNTOPT_ATTR2 "attr2" /* do use attr2 attribute format */
123 #define MNTOPT_NOATTR2 "noattr2" /* do not use attr2 attribute format */
124 #define MNTOPT_FILESTREAM "filestreams" /* use filestreams allocator */
125 #define MNTOPT_QUOTA "quota" /* disk quotas (user) */
126 #define MNTOPT_NOQUOTA "noquota" /* no quotas */
127 #define MNTOPT_USRQUOTA "usrquota" /* user quota enabled */
128 #define MNTOPT_GRPQUOTA "grpquota" /* group quota enabled */
129 #define MNTOPT_PRJQUOTA "prjquota" /* project quota enabled */
130 #define MNTOPT_UQUOTA "uquota" /* user quota (IRIX variant) */
131 #define MNTOPT_GQUOTA "gquota" /* group quota (IRIX variant) */
132 #define MNTOPT_PQUOTA "pquota" /* project quota (IRIX variant) */
133 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
134 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
135 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
136 #define MNTOPT_QUOTANOENF "qnoenforce" /* same as uqnoenforce */
137 #define MNTOPT_DMAPI "dmapi" /* DMI enabled (DMAPI / XDSM) */
138 #define MNTOPT_XDSM "xdsm" /* DMI enabled (DMAPI / XDSM) */
139 #define MNTOPT_DMI "dmi" /* DMI enabled (DMAPI / XDSM) */
140
141 STATIC unsigned long
142 suffix_strtoul(char *s, char **endp, unsigned int base)
143 {
144 int last, shift_left_factor = 0;
145 char *value = s;
146
147 last = strlen(value) - 1;
148 if (value[last] == 'K' || value[last] == 'k') {
149 shift_left_factor = 10;
150 value[last] = '\0';
151 }
152 if (value[last] == 'M' || value[last] == 'm') {
153 shift_left_factor = 20;
154 value[last] = '\0';
155 }
156 if (value[last] == 'G' || value[last] == 'g') {
157 shift_left_factor = 30;
158 value[last] = '\0';
159 }
160
161 return simple_strtoul((const char *)s, endp, base) << shift_left_factor;
162 }
163
164 STATIC int
165 xfs_parseargs(
166 struct xfs_mount *mp,
167 char *options,
168 struct xfs_mount_args *args,
169 int update)
170 {
171 char *this_char, *value, *eov;
172 int dsunit, dswidth, vol_dsunit, vol_dswidth;
173 int iosize;
174 int dmapi_implies_ikeep = 1;
175
176 args->flags |= XFSMNT_BARRIER;
177 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
178
179 if (!options)
180 goto done;
181
182 iosize = dsunit = dswidth = vol_dsunit = vol_dswidth = 0;
183
184 while ((this_char = strsep(&options, ",")) != NULL) {
185 if (!*this_char)
186 continue;
187 if ((value = strchr(this_char, '=')) != NULL)
188 *value++ = 0;
189
190 if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
191 if (!value || !*value) {
192 cmn_err(CE_WARN,
193 "XFS: %s option requires an argument",
194 this_char);
195 return EINVAL;
196 }
197 args->logbufs = simple_strtoul(value, &eov, 10);
198 } else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
199 if (!value || !*value) {
200 cmn_err(CE_WARN,
201 "XFS: %s option requires an argument",
202 this_char);
203 return EINVAL;
204 }
205 args->logbufsize = suffix_strtoul(value, &eov, 10);
206 } else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
207 if (!value || !*value) {
208 cmn_err(CE_WARN,
209 "XFS: %s option requires an argument",
210 this_char);
211 return EINVAL;
212 }
213 strncpy(args->logname, value, MAXNAMELEN);
214 } else if (!strcmp(this_char, MNTOPT_MTPT)) {
215 if (!value || !*value) {
216 cmn_err(CE_WARN,
217 "XFS: %s option requires an argument",
218 this_char);
219 return EINVAL;
220 }
221 strncpy(args->mtpt, value, MAXNAMELEN);
222 } else if (!strcmp(this_char, MNTOPT_RTDEV)) {
223 if (!value || !*value) {
224 cmn_err(CE_WARN,
225 "XFS: %s option requires an argument",
226 this_char);
227 return EINVAL;
228 }
229 strncpy(args->rtname, value, MAXNAMELEN);
230 } else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
231 if (!value || !*value) {
232 cmn_err(CE_WARN,
233 "XFS: %s option requires an argument",
234 this_char);
235 return EINVAL;
236 }
237 iosize = simple_strtoul(value, &eov, 10);
238 args->flags |= XFSMNT_IOSIZE;
239 args->iosizelog = (uint8_t) iosize;
240 } else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
241 if (!value || !*value) {
242 cmn_err(CE_WARN,
243 "XFS: %s option requires an argument",
244 this_char);
245 return EINVAL;
246 }
247 iosize = suffix_strtoul(value, &eov, 10);
248 args->flags |= XFSMNT_IOSIZE;
249 args->iosizelog = ffs(iosize) - 1;
250 } else if (!strcmp(this_char, MNTOPT_GRPID) ||
251 !strcmp(this_char, MNTOPT_BSDGROUPS)) {
252 mp->m_flags |= XFS_MOUNT_GRPID;
253 } else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
254 !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
255 mp->m_flags &= ~XFS_MOUNT_GRPID;
256 } else if (!strcmp(this_char, MNTOPT_WSYNC)) {
257 args->flags |= XFSMNT_WSYNC;
258 } else if (!strcmp(this_char, MNTOPT_OSYNCISOSYNC)) {
259 args->flags |= XFSMNT_OSYNCISOSYNC;
260 } else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
261 args->flags |= XFSMNT_NORECOVERY;
262 } else if (!strcmp(this_char, MNTOPT_INO64)) {
263 args->flags |= XFSMNT_INO64;
264 #if !XFS_BIG_INUMS
265 cmn_err(CE_WARN,
266 "XFS: %s option not allowed on this system",
267 this_char);
268 return EINVAL;
269 #endif
270 } else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
271 args->flags |= XFSMNT_NOALIGN;
272 } else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
273 args->flags |= XFSMNT_SWALLOC;
274 } else if (!strcmp(this_char, MNTOPT_SUNIT)) {
275 if (!value || !*value) {
276 cmn_err(CE_WARN,
277 "XFS: %s option requires an argument",
278 this_char);
279 return EINVAL;
280 }
281 dsunit = simple_strtoul(value, &eov, 10);
282 } else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
283 if (!value || !*value) {
284 cmn_err(CE_WARN,
285 "XFS: %s option requires an argument",
286 this_char);
287 return EINVAL;
288 }
289 dswidth = simple_strtoul(value, &eov, 10);
290 } else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
291 args->flags &= ~XFSMNT_32BITINODES;
292 #if !XFS_BIG_INUMS
293 cmn_err(CE_WARN,
294 "XFS: %s option not allowed on this system",
295 this_char);
296 return EINVAL;
297 #endif
298 } else if (!strcmp(this_char, MNTOPT_NOUUID)) {
299 args->flags |= XFSMNT_NOUUID;
300 } else if (!strcmp(this_char, MNTOPT_BARRIER)) {
301 args->flags |= XFSMNT_BARRIER;
302 } else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
303 args->flags &= ~XFSMNT_BARRIER;
304 } else if (!strcmp(this_char, MNTOPT_IKEEP)) {
305 args->flags |= XFSMNT_IKEEP;
306 } else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
307 dmapi_implies_ikeep = 0;
308 args->flags &= ~XFSMNT_IKEEP;
309 } else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
310 args->flags2 &= ~XFSMNT2_COMPAT_IOSIZE;
311 } else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
312 args->flags2 |= XFSMNT2_COMPAT_IOSIZE;
313 } else if (!strcmp(this_char, MNTOPT_ATTR2)) {
314 args->flags |= XFSMNT_ATTR2;
315 } else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
316 args->flags &= ~XFSMNT_ATTR2;
317 } else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
318 args->flags2 |= XFSMNT2_FILESTREAMS;
319 } else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
320 args->flags &= ~(XFSMNT_UQUOTAENF|XFSMNT_UQUOTA);
321 args->flags &= ~(XFSMNT_GQUOTAENF|XFSMNT_GQUOTA);
322 } else if (!strcmp(this_char, MNTOPT_QUOTA) ||
323 !strcmp(this_char, MNTOPT_UQUOTA) ||
324 !strcmp(this_char, MNTOPT_USRQUOTA)) {
325 args->flags |= XFSMNT_UQUOTA | XFSMNT_UQUOTAENF;
326 } else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
327 !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
328 args->flags |= XFSMNT_UQUOTA;
329 args->flags &= ~XFSMNT_UQUOTAENF;
330 } else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
331 !strcmp(this_char, MNTOPT_PRJQUOTA)) {
332 args->flags |= XFSMNT_PQUOTA | XFSMNT_PQUOTAENF;
333 } else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
334 args->flags |= XFSMNT_PQUOTA;
335 args->flags &= ~XFSMNT_PQUOTAENF;
336 } else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
337 !strcmp(this_char, MNTOPT_GRPQUOTA)) {
338 args->flags |= XFSMNT_GQUOTA | XFSMNT_GQUOTAENF;
339 } else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
340 args->flags |= XFSMNT_GQUOTA;
341 args->flags &= ~XFSMNT_GQUOTAENF;
342 } else if (!strcmp(this_char, MNTOPT_DMAPI)) {
343 args->flags |= XFSMNT_DMAPI;
344 } else if (!strcmp(this_char, MNTOPT_XDSM)) {
345 args->flags |= XFSMNT_DMAPI;
346 } else if (!strcmp(this_char, MNTOPT_DMI)) {
347 args->flags |= XFSMNT_DMAPI;
348 } else if (!strcmp(this_char, "ihashsize")) {
349 cmn_err(CE_WARN,
350 "XFS: ihashsize no longer used, option is deprecated.");
351 } else if (!strcmp(this_char, "osyncisdsync")) {
352 /* no-op, this is now the default */
353 cmn_err(CE_WARN,
354 "XFS: osyncisdsync is now the default, option is deprecated.");
355 } else if (!strcmp(this_char, "irixsgid")) {
356 cmn_err(CE_WARN,
357 "XFS: irixsgid is now a sysctl(2) variable, option is deprecated.");
358 } else {
359 cmn_err(CE_WARN,
360 "XFS: unknown mount option [%s].", this_char);
361 return EINVAL;
362 }
363 }
364
365 if (args->flags & XFSMNT_NORECOVERY) {
366 if ((mp->m_flags & XFS_MOUNT_RDONLY) == 0) {
367 cmn_err(CE_WARN,
368 "XFS: no-recovery mounts must be read-only.");
369 return EINVAL;
370 }
371 }
372
373 if ((args->flags & XFSMNT_NOALIGN) && (dsunit || dswidth)) {
374 cmn_err(CE_WARN,
375 "XFS: sunit and swidth options incompatible with the noalign option");
376 return EINVAL;
377 }
378
379 if ((args->flags & XFSMNT_GQUOTA) && (args->flags & XFSMNT_PQUOTA)) {
380 cmn_err(CE_WARN,
381 "XFS: cannot mount with both project and group quota");
382 return EINVAL;
383 }
384
385 if ((args->flags & XFSMNT_DMAPI) && *args->mtpt == '\0') {
386 printk("XFS: %s option needs the mount point option as well\n",
387 MNTOPT_DMAPI);
388 return EINVAL;
389 }
390
391 if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
392 cmn_err(CE_WARN,
393 "XFS: sunit and swidth must be specified together");
394 return EINVAL;
395 }
396
397 if (dsunit && (dswidth % dsunit != 0)) {
398 cmn_err(CE_WARN,
399 "XFS: stripe width (%d) must be a multiple of the stripe unit (%d)",
400 dswidth, dsunit);
401 return EINVAL;
402 }
403
404 /*
405 * Applications using DMI filesystems often expect the
406 * inode generation number to be monotonically increasing.
407 * If we delete inode chunks we break this assumption, so
408 * keep unused inode chunks on disk for DMI filesystems
409 * until we come up with a better solution.
410 * Note that if "ikeep" or "noikeep" mount options are
411 * supplied, then they are honored.
412 */
413 if ((args->flags & XFSMNT_DMAPI) && dmapi_implies_ikeep)
414 args->flags |= XFSMNT_IKEEP;
415
416 if ((args->flags & XFSMNT_NOALIGN) != XFSMNT_NOALIGN) {
417 if (dsunit) {
418 args->sunit = dsunit;
419 args->flags |= XFSMNT_RETERR;
420 } else {
421 args->sunit = vol_dsunit;
422 }
423 dswidth ? (args->swidth = dswidth) :
424 (args->swidth = vol_dswidth);
425 } else {
426 args->sunit = args->swidth = 0;
427 }
428
429 done:
430 if (args->flags & XFSMNT_32BITINODES)
431 mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
432 if (args->flags2)
433 args->flags |= XFSMNT_FLAGS2;
434 return 0;
435 }
436
437 struct proc_xfs_info {
438 int flag;
439 char *str;
440 };
441
442 STATIC int
443 xfs_showargs(
444 struct xfs_mount *mp,
445 struct seq_file *m)
446 {
447 static struct proc_xfs_info xfs_info_set[] = {
448 /* the few simple ones we can get from the mount struct */
449 { XFS_MOUNT_IKEEP, "," MNTOPT_IKEEP },
450 { XFS_MOUNT_WSYNC, "," MNTOPT_WSYNC },
451 { XFS_MOUNT_INO64, "," MNTOPT_INO64 },
452 { XFS_MOUNT_NOALIGN, "," MNTOPT_NOALIGN },
453 { XFS_MOUNT_SWALLOC, "," MNTOPT_SWALLOC },
454 { XFS_MOUNT_NOUUID, "," MNTOPT_NOUUID },
455 { XFS_MOUNT_NORECOVERY, "," MNTOPT_NORECOVERY },
456 { XFS_MOUNT_OSYNCISOSYNC, "," MNTOPT_OSYNCISOSYNC },
457 { XFS_MOUNT_ATTR2, "," MNTOPT_ATTR2 },
458 { XFS_MOUNT_FILESTREAMS, "," MNTOPT_FILESTREAM },
459 { XFS_MOUNT_DMAPI, "," MNTOPT_DMAPI },
460 { XFS_MOUNT_GRPID, "," MNTOPT_GRPID },
461 { 0, NULL }
462 };
463 static struct proc_xfs_info xfs_info_unset[] = {
464 /* the few simple ones we can get from the mount struct */
465 { XFS_MOUNT_COMPAT_IOSIZE, "," MNTOPT_LARGEIO },
466 { XFS_MOUNT_BARRIER, "," MNTOPT_NOBARRIER },
467 { XFS_MOUNT_SMALL_INUMS, "," MNTOPT_64BITINODE },
468 { 0, NULL }
469 };
470 struct proc_xfs_info *xfs_infop;
471
472 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
473 if (mp->m_flags & xfs_infop->flag)
474 seq_puts(m, xfs_infop->str);
475 }
476 for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
477 if (!(mp->m_flags & xfs_infop->flag))
478 seq_puts(m, xfs_infop->str);
479 }
480
481 if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
482 seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
483 (int)(1 << mp->m_writeio_log) >> 10);
484
485 if (mp->m_logbufs > 0)
486 seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
487 if (mp->m_logbsize > 0)
488 seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
489
490 if (mp->m_logname)
491 seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
492 if (mp->m_rtname)
493 seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
494
495 if (mp->m_dalign > 0)
496 seq_printf(m, "," MNTOPT_SUNIT "=%d",
497 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
498 if (mp->m_swidth > 0)
499 seq_printf(m, "," MNTOPT_SWIDTH "=%d",
500 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
501
502 if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
503 seq_puts(m, "," MNTOPT_USRQUOTA);
504 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
505 seq_puts(m, "," MNTOPT_UQUOTANOENF);
506
507 if (mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_OQUOTA_ENFD))
508 seq_puts(m, "," MNTOPT_PRJQUOTA);
509 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
510 seq_puts(m, "," MNTOPT_PQUOTANOENF);
511
512 if (mp->m_qflags & (XFS_GQUOTA_ACCT|XFS_OQUOTA_ENFD))
513 seq_puts(m, "," MNTOPT_GRPQUOTA);
514 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
515 seq_puts(m, "," MNTOPT_GQUOTANOENF);
516
517 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
518 seq_puts(m, "," MNTOPT_NOQUOTA);
519
520 return 0;
521 }
522 __uint64_t
523 xfs_max_file_offset(
524 unsigned int blockshift)
525 {
526 unsigned int pagefactor = 1;
527 unsigned int bitshift = BITS_PER_LONG - 1;
528
529 /* Figure out maximum filesize, on Linux this can depend on
530 * the filesystem blocksize (on 32 bit platforms).
531 * __block_prepare_write does this in an [unsigned] long...
532 * page->index << (PAGE_CACHE_SHIFT - bbits)
533 * So, for page sized blocks (4K on 32 bit platforms),
534 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
535 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
536 * but for smaller blocksizes it is less (bbits = log2 bsize).
537 * Note1: get_block_t takes a long (implicit cast from above)
538 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
539 * can optionally convert the [unsigned] long from above into
540 * an [unsigned] long long.
541 */
542
543 #if BITS_PER_LONG == 32
544 # if defined(CONFIG_LBD)
545 ASSERT(sizeof(sector_t) == 8);
546 pagefactor = PAGE_CACHE_SIZE;
547 bitshift = BITS_PER_LONG;
548 # else
549 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
550 # endif
551 #endif
552
553 return (((__uint64_t)pagefactor) << bitshift) - 1;
554 }
555
556 STATIC_INLINE void
557 xfs_set_inodeops(
558 struct inode *inode)
559 {
560 switch (inode->i_mode & S_IFMT) {
561 case S_IFREG:
562 inode->i_op = &xfs_inode_operations;
563 inode->i_fop = &xfs_file_operations;
564 inode->i_mapping->a_ops = &xfs_address_space_operations;
565 break;
566 case S_IFDIR:
567 inode->i_op = &xfs_dir_inode_operations;
568 inode->i_fop = &xfs_dir_file_operations;
569 break;
570 case S_IFLNK:
571 inode->i_op = &xfs_symlink_inode_operations;
572 if (!(XFS_I(inode)->i_df.if_flags & XFS_IFINLINE))
573 inode->i_mapping->a_ops = &xfs_address_space_operations;
574 break;
575 default:
576 inode->i_op = &xfs_inode_operations;
577 init_special_inode(inode, inode->i_mode, inode->i_rdev);
578 break;
579 }
580 }
581
582 STATIC_INLINE void
583 xfs_revalidate_inode(
584 xfs_mount_t *mp,
585 bhv_vnode_t *vp,
586 xfs_inode_t *ip)
587 {
588 struct inode *inode = vn_to_inode(vp);
589
590 inode->i_mode = ip->i_d.di_mode;
591 inode->i_nlink = ip->i_d.di_nlink;
592 inode->i_uid = ip->i_d.di_uid;
593 inode->i_gid = ip->i_d.di_gid;
594
595 switch (inode->i_mode & S_IFMT) {
596 case S_IFBLK:
597 case S_IFCHR:
598 inode->i_rdev =
599 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
600 sysv_minor(ip->i_df.if_u2.if_rdev));
601 break;
602 default:
603 inode->i_rdev = 0;
604 break;
605 }
606
607 inode->i_generation = ip->i_d.di_gen;
608 i_size_write(inode, ip->i_d.di_size);
609 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
610 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
611 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
612 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
613 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
614 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
615 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
616 inode->i_flags |= S_IMMUTABLE;
617 else
618 inode->i_flags &= ~S_IMMUTABLE;
619 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
620 inode->i_flags |= S_APPEND;
621 else
622 inode->i_flags &= ~S_APPEND;
623 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
624 inode->i_flags |= S_SYNC;
625 else
626 inode->i_flags &= ~S_SYNC;
627 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
628 inode->i_flags |= S_NOATIME;
629 else
630 inode->i_flags &= ~S_NOATIME;
631 xfs_iflags_clear(ip, XFS_IMODIFIED);
632 }
633
634 void
635 xfs_initialize_vnode(
636 struct xfs_mount *mp,
637 bhv_vnode_t *vp,
638 struct xfs_inode *ip)
639 {
640 struct inode *inode = vn_to_inode(vp);
641
642 if (!ip->i_vnode) {
643 ip->i_vnode = vp;
644 inode->i_private = ip;
645 }
646
647 /*
648 * We need to set the ops vectors, and unlock the inode, but if
649 * we have been called during the new inode create process, it is
650 * too early to fill in the Linux inode. We will get called a
651 * second time once the inode is properly set up, and then we can
652 * finish our work.
653 */
654 if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) {
655 xfs_revalidate_inode(mp, vp, ip);
656 xfs_set_inodeops(inode);
657
658 xfs_iflags_clear(ip, XFS_INEW);
659 barrier();
660
661 unlock_new_inode(inode);
662 }
663 }
664
665 int
666 xfs_blkdev_get(
667 xfs_mount_t *mp,
668 const char *name,
669 struct block_device **bdevp)
670 {
671 int error = 0;
672
673 *bdevp = open_bdev_excl(name, 0, mp);
674 if (IS_ERR(*bdevp)) {
675 error = PTR_ERR(*bdevp);
676 printk("XFS: Invalid device [%s], error=%d\n", name, error);
677 }
678
679 return -error;
680 }
681
682 void
683 xfs_blkdev_put(
684 struct block_device *bdev)
685 {
686 if (bdev)
687 close_bdev_excl(bdev);
688 }
689
690 /*
691 * Try to write out the superblock using barriers.
692 */
693 STATIC int
694 xfs_barrier_test(
695 xfs_mount_t *mp)
696 {
697 xfs_buf_t *sbp = xfs_getsb(mp, 0);
698 int error;
699
700 XFS_BUF_UNDONE(sbp);
701 XFS_BUF_UNREAD(sbp);
702 XFS_BUF_UNDELAYWRITE(sbp);
703 XFS_BUF_WRITE(sbp);
704 XFS_BUF_UNASYNC(sbp);
705 XFS_BUF_ORDERED(sbp);
706
707 xfsbdstrat(mp, sbp);
708 error = xfs_iowait(sbp);
709
710 /*
711 * Clear all the flags we set and possible error state in the
712 * buffer. We only did the write to try out whether barriers
713 * worked and shouldn't leave any traces in the superblock
714 * buffer.
715 */
716 XFS_BUF_DONE(sbp);
717 XFS_BUF_ERROR(sbp, 0);
718 XFS_BUF_UNORDERED(sbp);
719
720 xfs_buf_relse(sbp);
721 return error;
722 }
723
724 void
725 xfs_mountfs_check_barriers(xfs_mount_t *mp)
726 {
727 int error;
728
729 if (mp->m_logdev_targp != mp->m_ddev_targp) {
730 xfs_fs_cmn_err(CE_NOTE, mp,
731 "Disabling barriers, not supported with external log device");
732 mp->m_flags &= ~XFS_MOUNT_BARRIER;
733 return;
734 }
735
736 if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered ==
737 QUEUE_ORDERED_NONE) {
738 xfs_fs_cmn_err(CE_NOTE, mp,
739 "Disabling barriers, not supported by the underlying device");
740 mp->m_flags &= ~XFS_MOUNT_BARRIER;
741 return;
742 }
743
744 if (xfs_readonly_buftarg(mp->m_ddev_targp)) {
745 xfs_fs_cmn_err(CE_NOTE, mp,
746 "Disabling barriers, underlying device is readonly");
747 mp->m_flags &= ~XFS_MOUNT_BARRIER;
748 return;
749 }
750
751 error = xfs_barrier_test(mp);
752 if (error) {
753 xfs_fs_cmn_err(CE_NOTE, mp,
754 "Disabling barriers, trial barrier write failed");
755 mp->m_flags &= ~XFS_MOUNT_BARRIER;
756 return;
757 }
758 }
759
760 void
761 xfs_blkdev_issue_flush(
762 xfs_buftarg_t *buftarg)
763 {
764 blkdev_issue_flush(buftarg->bt_bdev, NULL);
765 }
766
767 /*
768 * XFS AIL push thread support
769 */
770 void
771 xfsaild_wakeup(
772 xfs_mount_t *mp,
773 xfs_lsn_t threshold_lsn)
774 {
775 mp->m_ail.xa_target = threshold_lsn;
776 wake_up_process(mp->m_ail.xa_task);
777 }
778
779 int
780 xfsaild(
781 void *data)
782 {
783 xfs_mount_t *mp = (xfs_mount_t *)data;
784 xfs_lsn_t last_pushed_lsn = 0;
785 long tout = 0;
786
787 while (!kthread_should_stop()) {
788 if (tout)
789 schedule_timeout_interruptible(msecs_to_jiffies(tout));
790 tout = 1000;
791
792 /* swsusp */
793 try_to_freeze();
794
795 ASSERT(mp->m_log);
796 if (XFS_FORCED_SHUTDOWN(mp))
797 continue;
798
799 tout = xfsaild_push(mp, &last_pushed_lsn);
800 }
801
802 return 0;
803 } /* xfsaild */
804
805 int
806 xfsaild_start(
807 xfs_mount_t *mp)
808 {
809 mp->m_ail.xa_target = 0;
810 mp->m_ail.xa_task = kthread_run(xfsaild, mp, "xfsaild");
811 if (IS_ERR(mp->m_ail.xa_task))
812 return -PTR_ERR(mp->m_ail.xa_task);
813 return 0;
814 }
815
816 void
817 xfsaild_stop(
818 xfs_mount_t *mp)
819 {
820 kthread_stop(mp->m_ail.xa_task);
821 }
822
823
824
825 STATIC struct inode *
826 xfs_fs_alloc_inode(
827 struct super_block *sb)
828 {
829 bhv_vnode_t *vp;
830
831 vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP);
832 if (unlikely(!vp))
833 return NULL;
834 return vn_to_inode(vp);
835 }
836
837 STATIC void
838 xfs_fs_destroy_inode(
839 struct inode *inode)
840 {
841 kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode));
842 }
843
844 STATIC void
845 xfs_fs_inode_init_once(
846 kmem_zone_t *zonep,
847 void *vnode)
848 {
849 inode_init_once(vn_to_inode((bhv_vnode_t *)vnode));
850 }
851
852 STATIC int __init
853 xfs_init_zones(void)
854 {
855 xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode",
856 KM_ZONE_HWALIGN | KM_ZONE_RECLAIM |
857 KM_ZONE_SPREAD,
858 xfs_fs_inode_init_once);
859 if (!xfs_vnode_zone)
860 goto out;
861
862 xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
863 if (!xfs_ioend_zone)
864 goto out_destroy_vnode_zone;
865
866 xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
867 xfs_ioend_zone);
868 if (!xfs_ioend_pool)
869 goto out_free_ioend_zone;
870 return 0;
871
872 out_free_ioend_zone:
873 kmem_zone_destroy(xfs_ioend_zone);
874 out_destroy_vnode_zone:
875 kmem_zone_destroy(xfs_vnode_zone);
876 out:
877 return -ENOMEM;
878 }
879
880 STATIC void
881 xfs_destroy_zones(void)
882 {
883 mempool_destroy(xfs_ioend_pool);
884 kmem_zone_destroy(xfs_vnode_zone);
885 kmem_zone_destroy(xfs_ioend_zone);
886 }
887
888 /*
889 * Attempt to flush the inode, this will actually fail
890 * if the inode is pinned, but we dirty the inode again
891 * at the point when it is unpinned after a log write,
892 * since this is when the inode itself becomes flushable.
893 */
894 STATIC int
895 xfs_fs_write_inode(
896 struct inode *inode,
897 int sync)
898 {
899 int error = 0, flags = FLUSH_INODE;
900
901 xfs_itrace_entry(XFS_I(inode));
902 if (sync) {
903 filemap_fdatawait(inode->i_mapping);
904 flags |= FLUSH_SYNC;
905 }
906 error = xfs_inode_flush(XFS_I(inode), flags);
907 /*
908 * if we failed to write out the inode then mark
909 * it dirty again so we'll try again later.
910 */
911 if (error)
912 mark_inode_dirty_sync(inode);
913
914 return -error;
915 }
916
917 STATIC void
918 xfs_fs_clear_inode(
919 struct inode *inode)
920 {
921 xfs_inode_t *ip = XFS_I(inode);
922
923 /*
924 * ip can be null when xfs_iget_core calls xfs_idestroy if we
925 * find an inode with di_mode == 0 but without IGET_CREATE set.
926 */
927 if (ip) {
928 xfs_itrace_entry(ip);
929 XFS_STATS_INC(vn_rele);
930 XFS_STATS_INC(vn_remove);
931 XFS_STATS_INC(vn_reclaim);
932 XFS_STATS_DEC(vn_active);
933
934 xfs_inactive(ip);
935 xfs_iflags_clear(ip, XFS_IMODIFIED);
936 if (xfs_reclaim(ip))
937 panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, inode);
938 }
939
940 ASSERT(XFS_I(inode) == NULL);
941 }
942
943 /*
944 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
945 * Doing this has two advantages:
946 * - It saves on stack space, which is tight in certain situations
947 * - It can be used (with care) as a mechanism to avoid deadlocks.
948 * Flushing while allocating in a full filesystem requires both.
949 */
950 STATIC void
951 xfs_syncd_queue_work(
952 struct xfs_mount *mp,
953 void *data,
954 void (*syncer)(struct xfs_mount *, void *))
955 {
956 struct bhv_vfs_sync_work *work;
957
958 work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP);
959 INIT_LIST_HEAD(&work->w_list);
960 work->w_syncer = syncer;
961 work->w_data = data;
962 work->w_mount = mp;
963 spin_lock(&mp->m_sync_lock);
964 list_add_tail(&work->w_list, &mp->m_sync_list);
965 spin_unlock(&mp->m_sync_lock);
966 wake_up_process(mp->m_sync_task);
967 }
968
969 /*
970 * Flush delayed allocate data, attempting to free up reserved space
971 * from existing allocations. At this point a new allocation attempt
972 * has failed with ENOSPC and we are in the process of scratching our
973 * heads, looking about for more room...
974 */
975 STATIC void
976 xfs_flush_inode_work(
977 struct xfs_mount *mp,
978 void *arg)
979 {
980 struct inode *inode = arg;
981 filemap_flush(inode->i_mapping);
982 iput(inode);
983 }
984
985 void
986 xfs_flush_inode(
987 xfs_inode_t *ip)
988 {
989 struct inode *inode = ip->i_vnode;
990
991 igrab(inode);
992 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work);
993 delay(msecs_to_jiffies(500));
994 }
995
996 /*
997 * This is the "bigger hammer" version of xfs_flush_inode_work...
998 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
999 */
1000 STATIC void
1001 xfs_flush_device_work(
1002 struct xfs_mount *mp,
1003 void *arg)
1004 {
1005 struct inode *inode = arg;
1006 sync_blockdev(mp->m_super->s_bdev);
1007 iput(inode);
1008 }
1009
1010 void
1011 xfs_flush_device(
1012 xfs_inode_t *ip)
1013 {
1014 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
1015
1016 igrab(inode);
1017 xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work);
1018 delay(msecs_to_jiffies(500));
1019 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
1020 }
1021
1022 STATIC void
1023 xfs_sync_worker(
1024 struct xfs_mount *mp,
1025 void *unused)
1026 {
1027 int error;
1028
1029 if (!(mp->m_flags & XFS_MOUNT_RDONLY))
1030 error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR |
1031 SYNC_REFCACHE | SYNC_SUPER);
1032 mp->m_sync_seq++;
1033 wake_up(&mp->m_wait_single_sync_task);
1034 }
1035
1036 STATIC int
1037 xfssyncd(
1038 void *arg)
1039 {
1040 struct xfs_mount *mp = arg;
1041 long timeleft;
1042 bhv_vfs_sync_work_t *work, *n;
1043 LIST_HEAD (tmp);
1044
1045 set_freezable();
1046 timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10);
1047 for (;;) {
1048 timeleft = schedule_timeout_interruptible(timeleft);
1049 /* swsusp */
1050 try_to_freeze();
1051 if (kthread_should_stop() && list_empty(&mp->m_sync_list))
1052 break;
1053
1054 spin_lock(&mp->m_sync_lock);
1055 /*
1056 * We can get woken by laptop mode, to do a sync -
1057 * that's the (only!) case where the list would be
1058 * empty with time remaining.
1059 */
1060 if (!timeleft || list_empty(&mp->m_sync_list)) {
1061 if (!timeleft)
1062 timeleft = xfs_syncd_centisecs *
1063 msecs_to_jiffies(10);
1064 INIT_LIST_HEAD(&mp->m_sync_work.w_list);
1065 list_add_tail(&mp->m_sync_work.w_list,
1066 &mp->m_sync_list);
1067 }
1068 list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list)
1069 list_move(&work->w_list, &tmp);
1070 spin_unlock(&mp->m_sync_lock);
1071
1072 list_for_each_entry_safe(work, n, &tmp, w_list) {
1073 (*work->w_syncer)(mp, work->w_data);
1074 list_del(&work->w_list);
1075 if (work == &mp->m_sync_work)
1076 continue;
1077 kmem_free(work, sizeof(struct bhv_vfs_sync_work));
1078 }
1079 }
1080
1081 return 0;
1082 }
1083
1084 STATIC void
1085 xfs_fs_put_super(
1086 struct super_block *sb)
1087 {
1088 struct xfs_mount *mp = XFS_M(sb);
1089 int error;
1090
1091 kthread_stop(mp->m_sync_task);
1092
1093 xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI);
1094 error = xfs_unmount(mp, 0, NULL);
1095 if (error)
1096 printk("XFS: unmount got error=%d\n", error);
1097 }
1098
1099 STATIC void
1100 xfs_fs_write_super(
1101 struct super_block *sb)
1102 {
1103 if (!(sb->s_flags & MS_RDONLY))
1104 xfs_sync(XFS_M(sb), SYNC_FSDATA);
1105 sb->s_dirt = 0;
1106 }
1107
1108 STATIC int
1109 xfs_fs_sync_super(
1110 struct super_block *sb,
1111 int wait)
1112 {
1113 struct xfs_mount *mp = XFS_M(sb);
1114 int error;
1115 int flags;
1116
1117 /*
1118 * Treat a sync operation like a freeze. This is to work
1119 * around a race in sync_inodes() which works in two phases
1120 * - an asynchronous flush, which can write out an inode
1121 * without waiting for file size updates to complete, and a
1122 * synchronous flush, which wont do anything because the
1123 * async flush removed the inode's dirty flag. Also
1124 * sync_inodes() will not see any files that just have
1125 * outstanding transactions to be flushed because we don't
1126 * dirty the Linux inode until after the transaction I/O
1127 * completes.
1128 */
1129 if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) {
1130 /*
1131 * First stage of freeze - no more writers will make progress
1132 * now we are here, so we flush delwri and delalloc buffers
1133 * here, then wait for all I/O to complete. Data is frozen at
1134 * that point. Metadata is not frozen, transactions can still
1135 * occur here so don't bother flushing the buftarg (i.e
1136 * SYNC_QUIESCE) because it'll just get dirty again.
1137 */
1138 flags = SYNC_DATA_QUIESCE;
1139 } else
1140 flags = SYNC_FSDATA;
1141
1142 error = xfs_sync(mp, flags);
1143 sb->s_dirt = 0;
1144
1145 if (unlikely(laptop_mode)) {
1146 int prev_sync_seq = mp->m_sync_seq;
1147
1148 /*
1149 * The disk must be active because we're syncing.
1150 * We schedule xfssyncd now (now that the disk is
1151 * active) instead of later (when it might not be).
1152 */
1153 wake_up_process(mp->m_sync_task);
1154 /*
1155 * We have to wait for the sync iteration to complete.
1156 * If we don't, the disk activity caused by the sync
1157 * will come after the sync is completed, and that
1158 * triggers another sync from laptop mode.
1159 */
1160 wait_event(mp->m_wait_single_sync_task,
1161 mp->m_sync_seq != prev_sync_seq);
1162 }
1163
1164 return -error;
1165 }
1166
1167 STATIC int
1168 xfs_fs_statfs(
1169 struct dentry *dentry,
1170 struct kstatfs *statp)
1171 {
1172 struct xfs_mount *mp = XFS_M(dentry->d_sb);
1173 xfs_sb_t *sbp = &mp->m_sb;
1174 __uint64_t fakeinos, id;
1175 xfs_extlen_t lsize;
1176
1177 statp->f_type = XFS_SB_MAGIC;
1178 statp->f_namelen = MAXNAMELEN - 1;
1179
1180 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1181 statp->f_fsid.val[0] = (u32)id;
1182 statp->f_fsid.val[1] = (u32)(id >> 32);
1183
1184 xfs_icsb_sync_counters_flags(mp, XFS_ICSB_LAZY_COUNT);
1185
1186 spin_lock(&mp->m_sb_lock);
1187 statp->f_bsize = sbp->sb_blocksize;
1188 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1189 statp->f_blocks = sbp->sb_dblocks - lsize;
1190 statp->f_bfree = statp->f_bavail =
1191 sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1192 fakeinos = statp->f_bfree << sbp->sb_inopblog;
1193 #if XFS_BIG_INUMS
1194 fakeinos += mp->m_inoadd;
1195 #endif
1196 statp->f_files =
1197 MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1198 if (mp->m_maxicount)
1199 #if XFS_BIG_INUMS
1200 if (!mp->m_inoadd)
1201 #endif
1202 statp->f_files = min_t(typeof(statp->f_files),
1203 statp->f_files,
1204 mp->m_maxicount);
1205 statp->f_ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1206 spin_unlock(&mp->m_sb_lock);
1207
1208 XFS_QM_DQSTATVFS(XFS_I(dentry->d_inode), statp);
1209 return 0;
1210 }
1211
1212 STATIC int
1213 xfs_fs_remount(
1214 struct super_block *sb,
1215 int *flags,
1216 char *options)
1217 {
1218 struct xfs_mount *mp = XFS_M(sb);
1219 struct xfs_mount_args *args = xfs_args_allocate(sb, 0);
1220 int error;
1221
1222 error = xfs_parseargs(mp, options, args, 1);
1223 if (!error)
1224 error = xfs_mntupdate(mp, flags, args);
1225 kmem_free(args, sizeof(*args));
1226 return -error;
1227 }
1228
1229 /*
1230 * Second stage of a freeze. The data is already frozen so we only
1231 * need to take care of themetadata. Once that's done write a dummy
1232 * record to dirty the log in case of a crash while frozen.
1233 */
1234 STATIC void
1235 xfs_fs_lockfs(
1236 struct super_block *sb)
1237 {
1238 struct xfs_mount *mp = XFS_M(sb);
1239
1240 xfs_attr_quiesce(mp);
1241 xfs_fs_log_dummy(mp);
1242 }
1243
1244 STATIC int
1245 xfs_fs_show_options(
1246 struct seq_file *m,
1247 struct vfsmount *mnt)
1248 {
1249 return -xfs_showargs(XFS_M(mnt->mnt_sb), m);
1250 }
1251
1252 STATIC int
1253 xfs_fs_quotasync(
1254 struct super_block *sb,
1255 int type)
1256 {
1257 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL);
1258 }
1259
1260 STATIC int
1261 xfs_fs_getxstate(
1262 struct super_block *sb,
1263 struct fs_quota_stat *fqs)
1264 {
1265 return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs);
1266 }
1267
1268 STATIC int
1269 xfs_fs_setxstate(
1270 struct super_block *sb,
1271 unsigned int flags,
1272 int op)
1273 {
1274 return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags);
1275 }
1276
1277 STATIC int
1278 xfs_fs_getxquota(
1279 struct super_block *sb,
1280 int type,
1281 qid_t id,
1282 struct fs_disk_quota *fdq)
1283 {
1284 return -XFS_QM_QUOTACTL(XFS_M(sb),
1285 (type == USRQUOTA) ? Q_XGETQUOTA :
1286 ((type == GRPQUOTA) ? Q_XGETGQUOTA :
1287 Q_XGETPQUOTA), id, (caddr_t)fdq);
1288 }
1289
1290 STATIC int
1291 xfs_fs_setxquota(
1292 struct super_block *sb,
1293 int type,
1294 qid_t id,
1295 struct fs_disk_quota *fdq)
1296 {
1297 return -XFS_QM_QUOTACTL(XFS_M(sb),
1298 (type == USRQUOTA) ? Q_XSETQLIM :
1299 ((type == GRPQUOTA) ? Q_XSETGQLIM :
1300 Q_XSETPQLIM), id, (caddr_t)fdq);
1301 }
1302
1303 STATIC int
1304 xfs_fs_fill_super(
1305 struct super_block *sb,
1306 void *data,
1307 int silent)
1308 {
1309 struct inode *rootvp;
1310 struct xfs_mount *mp = NULL;
1311 struct xfs_mount_args *args = xfs_args_allocate(sb, silent);
1312 int error;
1313
1314 mp = xfs_mount_init();
1315
1316 INIT_LIST_HEAD(&mp->m_sync_list);
1317 spin_lock_init(&mp->m_sync_lock);
1318 init_waitqueue_head(&mp->m_wait_single_sync_task);
1319
1320 mp->m_super = sb;
1321 sb->s_fs_info = mp;
1322
1323 if (sb->s_flags & MS_RDONLY)
1324 mp->m_flags |= XFS_MOUNT_RDONLY;
1325
1326 error = xfs_parseargs(mp, (char *)data, args, 0);
1327 if (error)
1328 goto fail_vfsop;
1329
1330 sb_min_blocksize(sb, BBSIZE);
1331 sb->s_export_op = &xfs_export_operations;
1332 sb->s_qcop = &xfs_quotactl_operations;
1333 sb->s_op = &xfs_super_operations;
1334
1335 error = xfs_mount(mp, args, NULL);
1336 if (error)
1337 goto fail_vfsop;
1338
1339 sb->s_dirt = 1;
1340 sb->s_magic = XFS_SB_MAGIC;
1341 sb->s_blocksize = mp->m_sb.sb_blocksize;
1342 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1343 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1344 sb->s_time_gran = 1;
1345 set_posix_acl_flag(sb);
1346
1347 rootvp = igrab(mp->m_rootip->i_vnode);
1348 if (!rootvp) {
1349 error = ENOENT;
1350 goto fail_unmount;
1351 }
1352
1353 sb->s_root = d_alloc_root(vn_to_inode(rootvp));
1354 if (!sb->s_root) {
1355 error = ENOMEM;
1356 goto fail_vnrele;
1357 }
1358 if (is_bad_inode(sb->s_root->d_inode)) {
1359 error = EINVAL;
1360 goto fail_vnrele;
1361 }
1362
1363 mp->m_sync_work.w_syncer = xfs_sync_worker;
1364 mp->m_sync_work.w_mount = mp;
1365 mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd");
1366 if (IS_ERR(mp->m_sync_task)) {
1367 error = -PTR_ERR(mp->m_sync_task);
1368 goto fail_vnrele;
1369 }
1370
1371 xfs_itrace_exit(XFS_I(sb->s_root->d_inode));
1372
1373 kmem_free(args, sizeof(*args));
1374 return 0;
1375
1376 fail_vnrele:
1377 if (sb->s_root) {
1378 dput(sb->s_root);
1379 sb->s_root = NULL;
1380 } else {
1381 VN_RELE(rootvp);
1382 }
1383
1384 fail_unmount:
1385 xfs_unmount(mp, 0, NULL);
1386
1387 fail_vfsop:
1388 kmem_free(args, sizeof(*args));
1389 return -error;
1390 }
1391
1392 STATIC int
1393 xfs_fs_get_sb(
1394 struct file_system_type *fs_type,
1395 int flags,
1396 const char *dev_name,
1397 void *data,
1398 struct vfsmount *mnt)
1399 {
1400 return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super,
1401 mnt);
1402 }
1403
1404 static struct super_operations xfs_super_operations = {
1405 .alloc_inode = xfs_fs_alloc_inode,
1406 .destroy_inode = xfs_fs_destroy_inode,
1407 .write_inode = xfs_fs_write_inode,
1408 .clear_inode = xfs_fs_clear_inode,
1409 .put_super = xfs_fs_put_super,
1410 .write_super = xfs_fs_write_super,
1411 .sync_fs = xfs_fs_sync_super,
1412 .write_super_lockfs = xfs_fs_lockfs,
1413 .statfs = xfs_fs_statfs,
1414 .remount_fs = xfs_fs_remount,
1415 .show_options = xfs_fs_show_options,
1416 };
1417
1418 static struct quotactl_ops xfs_quotactl_operations = {
1419 .quota_sync = xfs_fs_quotasync,
1420 .get_xstate = xfs_fs_getxstate,
1421 .set_xstate = xfs_fs_setxstate,
1422 .get_xquota = xfs_fs_getxquota,
1423 .set_xquota = xfs_fs_setxquota,
1424 };
1425
1426 static struct file_system_type xfs_fs_type = {
1427 .owner = THIS_MODULE,
1428 .name = "xfs",
1429 .get_sb = xfs_fs_get_sb,
1430 .kill_sb = kill_block_super,
1431 .fs_flags = FS_REQUIRES_DEV,
1432 };
1433
1434
1435 STATIC int __init
1436 init_xfs_fs( void )
1437 {
1438 int error;
1439 static char message[] __initdata = KERN_INFO \
1440 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
1441
1442 printk(message);
1443
1444 ktrace_init(64);
1445
1446 error = xfs_init_zones();
1447 if (error < 0)
1448 goto undo_zones;
1449
1450 error = xfs_buf_init();
1451 if (error < 0)
1452 goto undo_buffers;
1453
1454 vn_init();
1455 xfs_init();
1456 uuid_init();
1457 vfs_initquota();
1458
1459 error = register_filesystem(&xfs_fs_type);
1460 if (error)
1461 goto undo_register;
1462 return 0;
1463
1464 undo_register:
1465 xfs_buf_terminate();
1466
1467 undo_buffers:
1468 xfs_destroy_zones();
1469
1470 undo_zones:
1471 return error;
1472 }
1473
1474 STATIC void __exit
1475 exit_xfs_fs( void )
1476 {
1477 vfs_exitquota();
1478 unregister_filesystem(&xfs_fs_type);
1479 xfs_cleanup();
1480 xfs_buf_terminate();
1481 xfs_destroy_zones();
1482 ktrace_uninit();
1483 }
1484
1485 module_init(init_xfs_fs);
1486 module_exit(exit_xfs_fs);
1487
1488 MODULE_AUTHOR("Silicon Graphics, Inc.");
1489 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1490 MODULE_LICENSE("GPL");
WRT350N Kernel Patches