Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
da6dd40d | 3 | * Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
9e6e0a12 | 10 | #include <linux/bio.h> |
b3b94faa DT |
11 | #include <linux/sched.h> |
12 | #include <linux/slab.h> | |
13 | #include <linux/spinlock.h> | |
14 | #include <linux/completion.h> | |
15 | #include <linux/buffer_head.h> | |
9e6e0a12 SW |
16 | #include <linux/statfs.h> |
17 | #include <linux/seq_file.h> | |
18 | #include <linux/mount.h> | |
19 | #include <linux/kthread.h> | |
20 | #include <linux/delay.h> | |
5c676f6d | 21 | #include <linux/gfs2_ondisk.h> |
9e6e0a12 SW |
22 | #include <linux/crc32.h> |
23 | #include <linux/time.h> | |
b3b94faa DT |
24 | |
25 | #include "gfs2.h" | |
5c676f6d | 26 | #include "incore.h" |
b3b94faa DT |
27 | #include "bmap.h" |
28 | #include "dir.h" | |
b3b94faa DT |
29 | #include "glock.h" |
30 | #include "glops.h" | |
31 | #include "inode.h" | |
32 | #include "log.h" | |
33 | #include "meta_io.h" | |
34 | #include "quota.h" | |
35 | #include "recovery.h" | |
36 | #include "rgrp.h" | |
37 | #include "super.h" | |
38 | #include "trans.h" | |
5c676f6d | 39 | #include "util.h" |
9e6e0a12 SW |
40 | #include "sys.h" |
41 | #include "eattr.h" | |
42 | ||
43 | #define args_neq(a1, a2, x) ((a1)->ar_##x != (a2)->ar_##x) | |
44 | ||
45 | enum { | |
46 | Opt_lockproto, | |
47 | Opt_locktable, | |
48 | Opt_hostdata, | |
49 | Opt_spectator, | |
50 | Opt_ignore_local_fs, | |
51 | Opt_localflocks, | |
52 | Opt_localcaching, | |
53 | Opt_debug, | |
54 | Opt_nodebug, | |
55 | Opt_upgrade, | |
56 | Opt_acl, | |
57 | Opt_noacl, | |
58 | Opt_quota_off, | |
59 | Opt_quota_account, | |
60 | Opt_quota_on, | |
61 | Opt_quota, | |
62 | Opt_noquota, | |
63 | Opt_suiddir, | |
64 | Opt_nosuiddir, | |
65 | Opt_data_writeback, | |
66 | Opt_data_ordered, | |
67 | Opt_meta, | |
68 | Opt_discard, | |
69 | Opt_nodiscard, | |
70 | Opt_commit, | |
d34843d0 BP |
71 | Opt_err_withdraw, |
72 | Opt_err_panic, | |
9e6e0a12 SW |
73 | Opt_error, |
74 | }; | |
75 | ||
76 | static const match_table_t tokens = { | |
77 | {Opt_lockproto, "lockproto=%s"}, | |
78 | {Opt_locktable, "locktable=%s"}, | |
79 | {Opt_hostdata, "hostdata=%s"}, | |
80 | {Opt_spectator, "spectator"}, | |
81 | {Opt_ignore_local_fs, "ignore_local_fs"}, | |
82 | {Opt_localflocks, "localflocks"}, | |
83 | {Opt_localcaching, "localcaching"}, | |
84 | {Opt_debug, "debug"}, | |
85 | {Opt_nodebug, "nodebug"}, | |
86 | {Opt_upgrade, "upgrade"}, | |
87 | {Opt_acl, "acl"}, | |
88 | {Opt_noacl, "noacl"}, | |
89 | {Opt_quota_off, "quota=off"}, | |
90 | {Opt_quota_account, "quota=account"}, | |
91 | {Opt_quota_on, "quota=on"}, | |
92 | {Opt_quota, "quota"}, | |
93 | {Opt_noquota, "noquota"}, | |
94 | {Opt_suiddir, "suiddir"}, | |
95 | {Opt_nosuiddir, "nosuiddir"}, | |
96 | {Opt_data_writeback, "data=writeback"}, | |
97 | {Opt_data_ordered, "data=ordered"}, | |
98 | {Opt_meta, "meta"}, | |
99 | {Opt_discard, "discard"}, | |
100 | {Opt_nodiscard, "nodiscard"}, | |
101 | {Opt_commit, "commit=%d"}, | |
d34843d0 BP |
102 | {Opt_err_withdraw, "errors=withdraw"}, |
103 | {Opt_err_panic, "errors=panic"}, | |
9e6e0a12 SW |
104 | {Opt_error, NULL} |
105 | }; | |
106 | ||
107 | /** | |
108 | * gfs2_mount_args - Parse mount options | |
109 | * @sdp: | |
110 | * @data: | |
111 | * | |
112 | * Return: errno | |
113 | */ | |
114 | ||
115 | int gfs2_mount_args(struct gfs2_sbd *sdp, struct gfs2_args *args, char *options) | |
116 | { | |
117 | char *o; | |
118 | int token; | |
119 | substring_t tmp[MAX_OPT_ARGS]; | |
120 | int rv; | |
121 | ||
122 | /* Split the options into tokens with the "," character and | |
123 | process them */ | |
124 | ||
125 | while (1) { | |
126 | o = strsep(&options, ","); | |
127 | if (o == NULL) | |
128 | break; | |
129 | if (*o == '\0') | |
130 | continue; | |
131 | ||
132 | token = match_token(o, tokens, tmp); | |
133 | switch (token) { | |
134 | case Opt_lockproto: | |
135 | match_strlcpy(args->ar_lockproto, &tmp[0], | |
136 | GFS2_LOCKNAME_LEN); | |
137 | break; | |
138 | case Opt_locktable: | |
139 | match_strlcpy(args->ar_locktable, &tmp[0], | |
140 | GFS2_LOCKNAME_LEN); | |
141 | break; | |
142 | case Opt_hostdata: | |
143 | match_strlcpy(args->ar_hostdata, &tmp[0], | |
144 | GFS2_LOCKNAME_LEN); | |
145 | break; | |
146 | case Opt_spectator: | |
147 | args->ar_spectator = 1; | |
148 | break; | |
149 | case Opt_ignore_local_fs: | |
150 | args->ar_ignore_local_fs = 1; | |
151 | break; | |
152 | case Opt_localflocks: | |
153 | args->ar_localflocks = 1; | |
154 | break; | |
155 | case Opt_localcaching: | |
156 | args->ar_localcaching = 1; | |
157 | break; | |
158 | case Opt_debug: | |
d34843d0 BP |
159 | if (args->ar_errors == GFS2_ERRORS_PANIC) { |
160 | fs_info(sdp, "-o debug and -o errors=panic " | |
161 | "are mutually exclusive.\n"); | |
162 | return -EINVAL; | |
163 | } | |
9e6e0a12 SW |
164 | args->ar_debug = 1; |
165 | break; | |
166 | case Opt_nodebug: | |
167 | args->ar_debug = 0; | |
168 | break; | |
169 | case Opt_upgrade: | |
170 | args->ar_upgrade = 1; | |
171 | break; | |
172 | case Opt_acl: | |
173 | args->ar_posix_acl = 1; | |
174 | break; | |
175 | case Opt_noacl: | |
176 | args->ar_posix_acl = 0; | |
177 | break; | |
178 | case Opt_quota_off: | |
179 | case Opt_noquota: | |
180 | args->ar_quota = GFS2_QUOTA_OFF; | |
181 | break; | |
182 | case Opt_quota_account: | |
183 | args->ar_quota = GFS2_QUOTA_ACCOUNT; | |
184 | break; | |
185 | case Opt_quota_on: | |
186 | case Opt_quota: | |
187 | args->ar_quota = GFS2_QUOTA_ON; | |
188 | break; | |
189 | case Opt_suiddir: | |
190 | args->ar_suiddir = 1; | |
191 | break; | |
192 | case Opt_nosuiddir: | |
193 | args->ar_suiddir = 0; | |
194 | break; | |
195 | case Opt_data_writeback: | |
196 | args->ar_data = GFS2_DATA_WRITEBACK; | |
197 | break; | |
198 | case Opt_data_ordered: | |
199 | args->ar_data = GFS2_DATA_ORDERED; | |
200 | break; | |
201 | case Opt_meta: | |
202 | args->ar_meta = 1; | |
203 | break; | |
204 | case Opt_discard: | |
205 | args->ar_discard = 1; | |
206 | break; | |
207 | case Opt_nodiscard: | |
208 | args->ar_discard = 0; | |
209 | break; | |
210 | case Opt_commit: | |
211 | rv = match_int(&tmp[0], &args->ar_commit); | |
212 | if (rv || args->ar_commit <= 0) { | |
213 | fs_info(sdp, "commit mount option requires a positive numeric argument\n"); | |
214 | return rv ? rv : -EINVAL; | |
215 | } | |
216 | break; | |
d34843d0 BP |
217 | case Opt_err_withdraw: |
218 | args->ar_errors = GFS2_ERRORS_WITHDRAW; | |
219 | break; | |
220 | case Opt_err_panic: | |
221 | if (args->ar_debug) { | |
222 | fs_info(sdp, "-o debug and -o errors=panic " | |
223 | "are mutually exclusive.\n"); | |
224 | return -EINVAL; | |
225 | } | |
226 | args->ar_errors = GFS2_ERRORS_PANIC; | |
227 | break; | |
9e6e0a12 SW |
228 | case Opt_error: |
229 | default: | |
230 | fs_info(sdp, "invalid mount option: %s\n", o); | |
231 | return -EINVAL; | |
232 | } | |
233 | } | |
234 | ||
235 | return 0; | |
236 | } | |
b3b94faa | 237 | |
fefc03bf SW |
238 | /** |
239 | * gfs2_jindex_free - Clear all the journal index information | |
240 | * @sdp: The GFS2 superblock | |
241 | * | |
242 | */ | |
243 | ||
244 | void gfs2_jindex_free(struct gfs2_sbd *sdp) | |
245 | { | |
246 | struct list_head list, *head; | |
247 | struct gfs2_jdesc *jd; | |
248 | struct gfs2_journal_extent *jext; | |
249 | ||
250 | spin_lock(&sdp->sd_jindex_spin); | |
251 | list_add(&list, &sdp->sd_jindex_list); | |
252 | list_del_init(&sdp->sd_jindex_list); | |
253 | sdp->sd_journals = 0; | |
254 | spin_unlock(&sdp->sd_jindex_spin); | |
255 | ||
256 | while (!list_empty(&list)) { | |
257 | jd = list_entry(list.next, struct gfs2_jdesc, jd_list); | |
258 | head = &jd->extent_list; | |
259 | while (!list_empty(head)) { | |
260 | jext = list_entry(head->next, | |
261 | struct gfs2_journal_extent, | |
262 | extent_list); | |
263 | list_del(&jext->extent_list); | |
264 | kfree(jext); | |
265 | } | |
266 | list_del(&jd->jd_list); | |
267 | iput(jd->jd_inode); | |
268 | kfree(jd); | |
269 | } | |
270 | } | |
271 | ||
b3b94faa DT |
272 | static struct gfs2_jdesc *jdesc_find_i(struct list_head *head, unsigned int jid) |
273 | { | |
274 | struct gfs2_jdesc *jd; | |
275 | int found = 0; | |
276 | ||
277 | list_for_each_entry(jd, head, jd_list) { | |
278 | if (jd->jd_jid == jid) { | |
279 | found = 1; | |
280 | break; | |
281 | } | |
282 | } | |
283 | ||
284 | if (!found) | |
285 | jd = NULL; | |
286 | ||
287 | return jd; | |
288 | } | |
289 | ||
290 | struct gfs2_jdesc *gfs2_jdesc_find(struct gfs2_sbd *sdp, unsigned int jid) | |
291 | { | |
292 | struct gfs2_jdesc *jd; | |
293 | ||
294 | spin_lock(&sdp->sd_jindex_spin); | |
295 | jd = jdesc_find_i(&sdp->sd_jindex_list, jid); | |
296 | spin_unlock(&sdp->sd_jindex_spin); | |
297 | ||
298 | return jd; | |
299 | } | |
300 | ||
b3b94faa DT |
301 | int gfs2_jdesc_check(struct gfs2_jdesc *jd) |
302 | { | |
feaa7bba SW |
303 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); |
304 | struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode); | |
b3b94faa DT |
305 | int ar; |
306 | int error; | |
307 | ||
c9e98886 SW |
308 | if (ip->i_disksize < (8 << 20) || ip->i_disksize > (1 << 30) || |
309 | (ip->i_disksize & (sdp->sd_sb.sb_bsize - 1))) { | |
b3b94faa DT |
310 | gfs2_consist_inode(ip); |
311 | return -EIO; | |
312 | } | |
c9e98886 | 313 | jd->jd_blocks = ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; |
b3b94faa | 314 | |
c9e98886 | 315 | error = gfs2_write_alloc_required(ip, 0, ip->i_disksize, &ar); |
b3b94faa DT |
316 | if (!error && ar) { |
317 | gfs2_consist_inode(ip); | |
318 | error = -EIO; | |
319 | } | |
320 | ||
321 | return error; | |
322 | } | |
323 | ||
b3b94faa DT |
324 | /** |
325 | * gfs2_make_fs_rw - Turn a Read-Only FS into a Read-Write one | |
326 | * @sdp: the filesystem | |
327 | * | |
328 | * Returns: errno | |
329 | */ | |
330 | ||
331 | int gfs2_make_fs_rw(struct gfs2_sbd *sdp) | |
332 | { | |
feaa7bba | 333 | struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
5c676f6d | 334 | struct gfs2_glock *j_gl = ip->i_gl; |
b3b94faa | 335 | struct gfs2_holder t_gh; |
55167622 | 336 | struct gfs2_log_header_host head; |
b3b94faa DT |
337 | int error; |
338 | ||
1c0f4872 | 339 | error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, 0, &t_gh); |
b3b94faa DT |
340 | if (error) |
341 | return error; | |
342 | ||
1a14d3a6 | 343 | j_gl->gl_ops->go_inval(j_gl, DIO_METADATA); |
b3b94faa DT |
344 | |
345 | error = gfs2_find_jhead(sdp->sd_jdesc, &head); | |
346 | if (error) | |
347 | goto fail; | |
348 | ||
349 | if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) { | |
350 | gfs2_consist(sdp); | |
351 | error = -EIO; | |
352 | goto fail; | |
353 | } | |
354 | ||
355 | /* Initialize some head of the log stuff */ | |
356 | sdp->sd_log_sequence = head.lh_sequence + 1; | |
357 | gfs2_log_pointers_init(sdp, head.lh_blkno); | |
358 | ||
b3b94faa DT |
359 | error = gfs2_quota_init(sdp); |
360 | if (error) | |
a91ea69f | 361 | goto fail; |
b3b94faa DT |
362 | |
363 | set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); | |
364 | ||
365 | gfs2_glock_dq_uninit(&t_gh); | |
366 | ||
367 | return 0; | |
368 | ||
a91ea69f | 369 | fail: |
b3b94faa DT |
370 | t_gh.gh_flags |= GL_NOCACHE; |
371 | gfs2_glock_dq_uninit(&t_gh); | |
372 | ||
373 | return error; | |
374 | } | |
375 | ||
1946f70a | 376 | void gfs2_statfs_change_in(struct gfs2_statfs_change_host *sc, const void *buf) |
bb8d8a6f SW |
377 | { |
378 | const struct gfs2_statfs_change *str = buf; | |
379 | ||
380 | sc->sc_total = be64_to_cpu(str->sc_total); | |
381 | sc->sc_free = be64_to_cpu(str->sc_free); | |
382 | sc->sc_dinodes = be64_to_cpu(str->sc_dinodes); | |
383 | } | |
384 | ||
385 | static void gfs2_statfs_change_out(const struct gfs2_statfs_change_host *sc, void *buf) | |
386 | { | |
387 | struct gfs2_statfs_change *str = buf; | |
388 | ||
389 | str->sc_total = cpu_to_be64(sc->sc_total); | |
390 | str->sc_free = cpu_to_be64(sc->sc_free); | |
391 | str->sc_dinodes = cpu_to_be64(sc->sc_dinodes); | |
392 | } | |
393 | ||
b3b94faa DT |
394 | int gfs2_statfs_init(struct gfs2_sbd *sdp) |
395 | { | |
feaa7bba | 396 | struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode); |
bd209cc0 | 397 | struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master; |
feaa7bba | 398 | struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode); |
bd209cc0 | 399 | struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local; |
b3b94faa DT |
400 | struct buffer_head *m_bh, *l_bh; |
401 | struct gfs2_holder gh; | |
402 | int error; | |
403 | ||
404 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE, | |
405 | &gh); | |
406 | if (error) | |
407 | return error; | |
408 | ||
409 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); | |
410 | if (error) | |
411 | goto out; | |
412 | ||
413 | if (sdp->sd_args.ar_spectator) { | |
414 | spin_lock(&sdp->sd_statfs_spin); | |
415 | gfs2_statfs_change_in(m_sc, m_bh->b_data + | |
416 | sizeof(struct gfs2_dinode)); | |
417 | spin_unlock(&sdp->sd_statfs_spin); | |
418 | } else { | |
419 | error = gfs2_meta_inode_buffer(l_ip, &l_bh); | |
420 | if (error) | |
421 | goto out_m_bh; | |
422 | ||
423 | spin_lock(&sdp->sd_statfs_spin); | |
424 | gfs2_statfs_change_in(m_sc, m_bh->b_data + | |
425 | sizeof(struct gfs2_dinode)); | |
426 | gfs2_statfs_change_in(l_sc, l_bh->b_data + | |
427 | sizeof(struct gfs2_dinode)); | |
428 | spin_unlock(&sdp->sd_statfs_spin); | |
429 | ||
430 | brelse(l_bh); | |
431 | } | |
432 | ||
a91ea69f | 433 | out_m_bh: |
b3b94faa | 434 | brelse(m_bh); |
a91ea69f | 435 | out: |
b3b94faa | 436 | gfs2_glock_dq_uninit(&gh); |
b3b94faa DT |
437 | return 0; |
438 | } | |
439 | ||
cd915493 SW |
440 | void gfs2_statfs_change(struct gfs2_sbd *sdp, s64 total, s64 free, |
441 | s64 dinodes) | |
b3b94faa | 442 | { |
feaa7bba | 443 | struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode); |
bd209cc0 | 444 | struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local; |
b3b94faa DT |
445 | struct buffer_head *l_bh; |
446 | int error; | |
447 | ||
448 | error = gfs2_meta_inode_buffer(l_ip, &l_bh); | |
449 | if (error) | |
450 | return; | |
451 | ||
d4e9c4c3 | 452 | gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1); |
b3b94faa DT |
453 | |
454 | spin_lock(&sdp->sd_statfs_spin); | |
455 | l_sc->sc_total += total; | |
456 | l_sc->sc_free += free; | |
457 | l_sc->sc_dinodes += dinodes; | |
907b9bce | 458 | gfs2_statfs_change_out(l_sc, l_bh->b_data + sizeof(struct gfs2_dinode)); |
b3b94faa DT |
459 | spin_unlock(&sdp->sd_statfs_spin); |
460 | ||
461 | brelse(l_bh); | |
462 | } | |
463 | ||
1946f70a BM |
464 | void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh, |
465 | struct buffer_head *l_bh) | |
466 | { | |
467 | struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode); | |
468 | struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode); | |
469 | struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master; | |
470 | struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local; | |
471 | ||
472 | gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1); | |
473 | ||
474 | spin_lock(&sdp->sd_statfs_spin); | |
475 | m_sc->sc_total += l_sc->sc_total; | |
476 | m_sc->sc_free += l_sc->sc_free; | |
477 | m_sc->sc_dinodes += l_sc->sc_dinodes; | |
478 | memset(l_sc, 0, sizeof(struct gfs2_statfs_change)); | |
479 | memset(l_bh->b_data + sizeof(struct gfs2_dinode), | |
480 | 0, sizeof(struct gfs2_statfs_change)); | |
481 | spin_unlock(&sdp->sd_statfs_spin); | |
482 | ||
483 | gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1); | |
484 | gfs2_statfs_change_out(m_sc, m_bh->b_data + sizeof(struct gfs2_dinode)); | |
485 | } | |
486 | ||
b3b94faa DT |
487 | int gfs2_statfs_sync(struct gfs2_sbd *sdp) |
488 | { | |
feaa7bba SW |
489 | struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode); |
490 | struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode); | |
bd209cc0 AV |
491 | struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master; |
492 | struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local; | |
b3b94faa DT |
493 | struct gfs2_holder gh; |
494 | struct buffer_head *m_bh, *l_bh; | |
495 | int error; | |
496 | ||
497 | error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE, | |
498 | &gh); | |
499 | if (error) | |
500 | return error; | |
501 | ||
502 | error = gfs2_meta_inode_buffer(m_ip, &m_bh); | |
503 | if (error) | |
504 | goto out; | |
505 | ||
506 | spin_lock(&sdp->sd_statfs_spin); | |
507 | gfs2_statfs_change_in(m_sc, m_bh->b_data + | |
907b9bce | 508 | sizeof(struct gfs2_dinode)); |
b3b94faa DT |
509 | if (!l_sc->sc_total && !l_sc->sc_free && !l_sc->sc_dinodes) { |
510 | spin_unlock(&sdp->sd_statfs_spin); | |
511 | goto out_bh; | |
512 | } | |
513 | spin_unlock(&sdp->sd_statfs_spin); | |
514 | ||
515 | error = gfs2_meta_inode_buffer(l_ip, &l_bh); | |
516 | if (error) | |
517 | goto out_bh; | |
518 | ||
519 | error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0); | |
520 | if (error) | |
521 | goto out_bh2; | |
522 | ||
1946f70a | 523 | update_statfs(sdp, m_bh, l_bh); |
b3b94faa DT |
524 | |
525 | gfs2_trans_end(sdp); | |
526 | ||
a91ea69f | 527 | out_bh2: |
b3b94faa | 528 | brelse(l_bh); |
a91ea69f | 529 | out_bh: |
b3b94faa | 530 | brelse(m_bh); |
a91ea69f | 531 | out: |
b3b94faa | 532 | gfs2_glock_dq_uninit(&gh); |
b3b94faa DT |
533 | return error; |
534 | } | |
535 | ||
b3b94faa DT |
536 | struct lfcc { |
537 | struct list_head list; | |
538 | struct gfs2_holder gh; | |
539 | }; | |
540 | ||
541 | /** | |
542 | * gfs2_lock_fs_check_clean - Stop all writes to the FS and check that all | |
543 | * journals are clean | |
544 | * @sdp: the file system | |
545 | * @state: the state to put the transaction lock into | |
546 | * @t_gh: the hold on the transaction lock | |
547 | * | |
548 | * Returns: errno | |
549 | */ | |
550 | ||
08bc2dbc AB |
551 | static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp, |
552 | struct gfs2_holder *t_gh) | |
b3b94faa | 553 | { |
5c676f6d | 554 | struct gfs2_inode *ip; |
b3b94faa DT |
555 | struct gfs2_jdesc *jd; |
556 | struct lfcc *lfcc; | |
557 | LIST_HEAD(list); | |
55167622 | 558 | struct gfs2_log_header_host lh; |
b3b94faa DT |
559 | int error; |
560 | ||
b3b94faa DT |
561 | list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { |
562 | lfcc = kmalloc(sizeof(struct lfcc), GFP_KERNEL); | |
563 | if (!lfcc) { | |
564 | error = -ENOMEM; | |
565 | goto out; | |
566 | } | |
feaa7bba SW |
567 | ip = GFS2_I(jd->jd_inode); |
568 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &lfcc->gh); | |
b3b94faa DT |
569 | if (error) { |
570 | kfree(lfcc); | |
571 | goto out; | |
572 | } | |
573 | list_add(&lfcc->list, &list); | |
574 | } | |
575 | ||
576 | error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_DEFERRED, | |
6802e340 | 577 | GL_NOCACHE, t_gh); |
b3b94faa DT |
578 | |
579 | list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { | |
580 | error = gfs2_jdesc_check(jd); | |
581 | if (error) | |
582 | break; | |
583 | error = gfs2_find_jhead(jd, &lh); | |
584 | if (error) | |
585 | break; | |
586 | if (!(lh.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) { | |
587 | error = -EBUSY; | |
588 | break; | |
589 | } | |
590 | } | |
591 | ||
592 | if (error) | |
593 | gfs2_glock_dq_uninit(t_gh); | |
594 | ||
a91ea69f | 595 | out: |
b3b94faa DT |
596 | while (!list_empty(&list)) { |
597 | lfcc = list_entry(list.next, struct lfcc, list); | |
598 | list_del(&lfcc->list); | |
599 | gfs2_glock_dq_uninit(&lfcc->gh); | |
600 | kfree(lfcc); | |
601 | } | |
b3b94faa DT |
602 | return error; |
603 | } | |
604 | ||
605 | /** | |
606 | * gfs2_freeze_fs - freezes the file system | |
607 | * @sdp: the file system | |
608 | * | |
609 | * This function flushes data and meta data for all machines by | |
610 | * aquiring the transaction log exclusively. All journals are | |
611 | * ensured to be in a clean state as well. | |
612 | * | |
613 | * Returns: errno | |
614 | */ | |
615 | ||
616 | int gfs2_freeze_fs(struct gfs2_sbd *sdp) | |
617 | { | |
618 | int error = 0; | |
619 | ||
f55ab26a | 620 | mutex_lock(&sdp->sd_freeze_lock); |
b3b94faa DT |
621 | |
622 | if (!sdp->sd_freeze_count++) { | |
623 | error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh); | |
624 | if (error) | |
625 | sdp->sd_freeze_count--; | |
626 | } | |
627 | ||
f55ab26a | 628 | mutex_unlock(&sdp->sd_freeze_lock); |
b3b94faa DT |
629 | |
630 | return error; | |
631 | } | |
632 | ||
633 | /** | |
634 | * gfs2_unfreeze_fs - unfreezes the file system | |
635 | * @sdp: the file system | |
636 | * | |
637 | * This function allows the file system to proceed by unlocking | |
638 | * the exclusively held transaction lock. Other GFS2 nodes are | |
639 | * now free to acquire the lock shared and go on with their lives. | |
640 | * | |
641 | */ | |
642 | ||
643 | void gfs2_unfreeze_fs(struct gfs2_sbd *sdp) | |
644 | { | |
f55ab26a | 645 | mutex_lock(&sdp->sd_freeze_lock); |
b3b94faa DT |
646 | |
647 | if (sdp->sd_freeze_count && !--sdp->sd_freeze_count) | |
648 | gfs2_glock_dq_uninit(&sdp->sd_freeze_gh); | |
649 | ||
f55ab26a | 650 | mutex_unlock(&sdp->sd_freeze_lock); |
b3b94faa DT |
651 | } |
652 | ||
9e6e0a12 SW |
653 | |
654 | /** | |
655 | * gfs2_write_inode - Make sure the inode is stable on the disk | |
656 | * @inode: The inode | |
657 | * @sync: synchronous write flag | |
658 | * | |
659 | * Returns: errno | |
660 | */ | |
661 | ||
662 | static int gfs2_write_inode(struct inode *inode, int sync) | |
663 | { | |
664 | struct gfs2_inode *ip = GFS2_I(inode); | |
665 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
666 | struct gfs2_holder gh; | |
667 | struct buffer_head *bh; | |
668 | struct timespec atime; | |
669 | struct gfs2_dinode *di; | |
670 | int ret = 0; | |
671 | ||
672 | /* Check this is a "normal" inode, etc */ | |
673 | if (!test_bit(GIF_USER, &ip->i_flags) || | |
674 | (current->flags & PF_MEMALLOC)) | |
675 | return 0; | |
676 | ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
677 | if (ret) | |
678 | goto do_flush; | |
679 | ret = gfs2_trans_begin(sdp, RES_DINODE, 0); | |
680 | if (ret) | |
681 | goto do_unlock; | |
682 | ret = gfs2_meta_inode_buffer(ip, &bh); | |
683 | if (ret == 0) { | |
684 | di = (struct gfs2_dinode *)bh->b_data; | |
685 | atime.tv_sec = be64_to_cpu(di->di_atime); | |
686 | atime.tv_nsec = be32_to_cpu(di->di_atime_nsec); | |
687 | if (timespec_compare(&inode->i_atime, &atime) > 0) { | |
688 | gfs2_trans_add_bh(ip->i_gl, bh, 1); | |
689 | gfs2_dinode_out(ip, bh->b_data); | |
690 | } | |
691 | brelse(bh); | |
692 | } | |
693 | gfs2_trans_end(sdp); | |
694 | do_unlock: | |
695 | gfs2_glock_dq_uninit(&gh); | |
696 | do_flush: | |
697 | if (sync != 0) | |
698 | gfs2_log_flush(GFS2_SB(inode), ip->i_gl); | |
699 | return ret; | |
700 | } | |
701 | ||
702 | /** | |
703 | * gfs2_make_fs_ro - Turn a Read-Write FS into a Read-Only one | |
704 | * @sdp: the filesystem | |
705 | * | |
706 | * Returns: errno | |
707 | */ | |
708 | ||
709 | static int gfs2_make_fs_ro(struct gfs2_sbd *sdp) | |
710 | { | |
711 | struct gfs2_holder t_gh; | |
712 | int error; | |
713 | ||
b94a170e | 714 | flush_workqueue(gfs2_delete_workqueue); |
9e6e0a12 SW |
715 | gfs2_quota_sync(sdp); |
716 | gfs2_statfs_sync(sdp); | |
717 | ||
718 | error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, GL_NOCACHE, | |
719 | &t_gh); | |
720 | if (error && !test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) | |
721 | return error; | |
722 | ||
723 | gfs2_meta_syncfs(sdp); | |
724 | gfs2_log_shutdown(sdp); | |
725 | ||
726 | clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); | |
727 | ||
728 | if (t_gh.gh_gl) | |
729 | gfs2_glock_dq_uninit(&t_gh); | |
730 | ||
731 | gfs2_quota_cleanup(sdp); | |
732 | ||
733 | return error; | |
734 | } | |
735 | ||
736 | static int gfs2_umount_recovery_wait(void *word) | |
737 | { | |
738 | schedule(); | |
739 | return 0; | |
740 | } | |
741 | ||
742 | /** | |
743 | * gfs2_put_super - Unmount the filesystem | |
744 | * @sb: The VFS superblock | |
745 | * | |
746 | */ | |
747 | ||
748 | static void gfs2_put_super(struct super_block *sb) | |
749 | { | |
750 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
751 | int error; | |
752 | struct gfs2_jdesc *jd; | |
753 | ||
754 | /* Unfreeze the filesystem, if we need to */ | |
755 | ||
756 | mutex_lock(&sdp->sd_freeze_lock); | |
757 | if (sdp->sd_freeze_count) | |
758 | gfs2_glock_dq_uninit(&sdp->sd_freeze_gh); | |
759 | mutex_unlock(&sdp->sd_freeze_lock); | |
760 | ||
761 | /* No more recovery requests */ | |
762 | set_bit(SDF_NORECOVERY, &sdp->sd_flags); | |
763 | smp_mb(); | |
764 | ||
765 | /* Wait on outstanding recovery */ | |
766 | restart: | |
767 | spin_lock(&sdp->sd_jindex_spin); | |
768 | list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) { | |
769 | if (!test_bit(JDF_RECOVERY, &jd->jd_flags)) | |
770 | continue; | |
771 | spin_unlock(&sdp->sd_jindex_spin); | |
772 | wait_on_bit(&jd->jd_flags, JDF_RECOVERY, | |
773 | gfs2_umount_recovery_wait, TASK_UNINTERRUPTIBLE); | |
774 | goto restart; | |
775 | } | |
776 | spin_unlock(&sdp->sd_jindex_spin); | |
777 | ||
778 | kthread_stop(sdp->sd_quotad_process); | |
779 | kthread_stop(sdp->sd_logd_process); | |
780 | ||
781 | if (!(sb->s_flags & MS_RDONLY)) { | |
782 | error = gfs2_make_fs_ro(sdp); | |
783 | if (error) | |
784 | gfs2_io_error(sdp); | |
785 | } | |
786 | /* At this point, we're through modifying the disk */ | |
787 | ||
788 | /* Release stuff */ | |
789 | ||
790 | iput(sdp->sd_jindex); | |
791 | iput(sdp->sd_inum_inode); | |
792 | iput(sdp->sd_statfs_inode); | |
793 | iput(sdp->sd_rindex); | |
794 | iput(sdp->sd_quota_inode); | |
795 | ||
796 | gfs2_glock_put(sdp->sd_rename_gl); | |
797 | gfs2_glock_put(sdp->sd_trans_gl); | |
798 | ||
799 | if (!sdp->sd_args.ar_spectator) { | |
800 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
801 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
802 | gfs2_glock_dq_uninit(&sdp->sd_ir_gh); | |
803 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); | |
804 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); | |
805 | iput(sdp->sd_ir_inode); | |
806 | iput(sdp->sd_sc_inode); | |
807 | iput(sdp->sd_qc_inode); | |
808 | } | |
809 | ||
810 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); | |
811 | gfs2_clear_rgrpd(sdp); | |
812 | gfs2_jindex_free(sdp); | |
813 | /* Take apart glock structures and buffer lists */ | |
814 | gfs2_gl_hash_clear(sdp); | |
815 | /* Unmount the locking protocol */ | |
816 | gfs2_lm_unmount(sdp); | |
817 | ||
818 | /* At this point, we're through participating in the lockspace */ | |
819 | gfs2_sys_fs_del(sdp); | |
820 | } | |
821 | ||
9e6e0a12 SW |
822 | /** |
823 | * gfs2_sync_fs - sync the filesystem | |
824 | * @sb: the superblock | |
825 | * | |
826 | * Flushes the log to disk. | |
827 | */ | |
828 | ||
829 | static int gfs2_sync_fs(struct super_block *sb, int wait) | |
830 | { | |
9e6e0a12 SW |
831 | if (wait && sb->s_fs_info) |
832 | gfs2_log_flush(sb->s_fs_info, NULL); | |
833 | return 0; | |
834 | } | |
835 | ||
836 | /** | |
837 | * gfs2_freeze - prevent further writes to the filesystem | |
838 | * @sb: the VFS structure for the filesystem | |
839 | * | |
840 | */ | |
841 | ||
842 | static int gfs2_freeze(struct super_block *sb) | |
843 | { | |
844 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
845 | int error; | |
846 | ||
847 | if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) | |
848 | return -EINVAL; | |
849 | ||
850 | for (;;) { | |
851 | error = gfs2_freeze_fs(sdp); | |
852 | if (!error) | |
853 | break; | |
854 | ||
855 | switch (error) { | |
856 | case -EBUSY: | |
857 | fs_err(sdp, "waiting for recovery before freeze\n"); | |
858 | break; | |
859 | ||
860 | default: | |
861 | fs_err(sdp, "error freezing FS: %d\n", error); | |
862 | break; | |
863 | } | |
864 | ||
865 | fs_err(sdp, "retrying...\n"); | |
866 | msleep(1000); | |
867 | } | |
868 | return 0; | |
869 | } | |
870 | ||
871 | /** | |
872 | * gfs2_unfreeze - reallow writes to the filesystem | |
873 | * @sb: the VFS structure for the filesystem | |
874 | * | |
875 | */ | |
876 | ||
877 | static int gfs2_unfreeze(struct super_block *sb) | |
878 | { | |
879 | gfs2_unfreeze_fs(sb->s_fs_info); | |
880 | return 0; | |
881 | } | |
882 | ||
883 | /** | |
884 | * statfs_fill - fill in the sg for a given RG | |
885 | * @rgd: the RG | |
886 | * @sc: the sc structure | |
887 | * | |
888 | * Returns: 0 on success, -ESTALE if the LVB is invalid | |
889 | */ | |
890 | ||
891 | static int statfs_slow_fill(struct gfs2_rgrpd *rgd, | |
892 | struct gfs2_statfs_change_host *sc) | |
893 | { | |
894 | gfs2_rgrp_verify(rgd); | |
895 | sc->sc_total += rgd->rd_data; | |
896 | sc->sc_free += rgd->rd_free; | |
897 | sc->sc_dinodes += rgd->rd_dinodes; | |
898 | return 0; | |
899 | } | |
900 | ||
901 | /** | |
902 | * gfs2_statfs_slow - Stat a filesystem using asynchronous locking | |
903 | * @sdp: the filesystem | |
904 | * @sc: the sc info that will be returned | |
905 | * | |
906 | * Any error (other than a signal) will cause this routine to fall back | |
907 | * to the synchronous version. | |
908 | * | |
909 | * FIXME: This really shouldn't busy wait like this. | |
910 | * | |
911 | * Returns: errno | |
912 | */ | |
913 | ||
914 | static int gfs2_statfs_slow(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc) | |
915 | { | |
916 | struct gfs2_holder ri_gh; | |
917 | struct gfs2_rgrpd *rgd_next; | |
918 | struct gfs2_holder *gha, *gh; | |
919 | unsigned int slots = 64; | |
920 | unsigned int x; | |
921 | int done; | |
922 | int error = 0, err; | |
923 | ||
924 | memset(sc, 0, sizeof(struct gfs2_statfs_change_host)); | |
925 | gha = kcalloc(slots, sizeof(struct gfs2_holder), GFP_KERNEL); | |
926 | if (!gha) | |
927 | return -ENOMEM; | |
928 | ||
929 | error = gfs2_rindex_hold(sdp, &ri_gh); | |
930 | if (error) | |
931 | goto out; | |
932 | ||
933 | rgd_next = gfs2_rgrpd_get_first(sdp); | |
934 | ||
935 | for (;;) { | |
936 | done = 1; | |
937 | ||
938 | for (x = 0; x < slots; x++) { | |
939 | gh = gha + x; | |
940 | ||
941 | if (gh->gh_gl && gfs2_glock_poll(gh)) { | |
942 | err = gfs2_glock_wait(gh); | |
943 | if (err) { | |
944 | gfs2_holder_uninit(gh); | |
945 | error = err; | |
946 | } else { | |
947 | if (!error) | |
948 | error = statfs_slow_fill( | |
949 | gh->gh_gl->gl_object, sc); | |
950 | gfs2_glock_dq_uninit(gh); | |
951 | } | |
952 | } | |
953 | ||
954 | if (gh->gh_gl) | |
955 | done = 0; | |
956 | else if (rgd_next && !error) { | |
957 | error = gfs2_glock_nq_init(rgd_next->rd_gl, | |
958 | LM_ST_SHARED, | |
959 | GL_ASYNC, | |
960 | gh); | |
961 | rgd_next = gfs2_rgrpd_get_next(rgd_next); | |
962 | done = 0; | |
963 | } | |
964 | ||
965 | if (signal_pending(current)) | |
966 | error = -ERESTARTSYS; | |
967 | } | |
968 | ||
969 | if (done) | |
970 | break; | |
971 | ||
972 | yield(); | |
973 | } | |
974 | ||
975 | gfs2_glock_dq_uninit(&ri_gh); | |
976 | ||
977 | out: | |
978 | kfree(gha); | |
979 | return error; | |
980 | } | |
981 | ||
982 | /** | |
983 | * gfs2_statfs_i - Do a statfs | |
984 | * @sdp: the filesystem | |
985 | * @sg: the sg structure | |
986 | * | |
987 | * Returns: errno | |
988 | */ | |
989 | ||
990 | static int gfs2_statfs_i(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc) | |
991 | { | |
992 | struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master; | |
993 | struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local; | |
994 | ||
995 | spin_lock(&sdp->sd_statfs_spin); | |
996 | ||
997 | *sc = *m_sc; | |
998 | sc->sc_total += l_sc->sc_total; | |
999 | sc->sc_free += l_sc->sc_free; | |
1000 | sc->sc_dinodes += l_sc->sc_dinodes; | |
1001 | ||
1002 | spin_unlock(&sdp->sd_statfs_spin); | |
1003 | ||
1004 | if (sc->sc_free < 0) | |
1005 | sc->sc_free = 0; | |
1006 | if (sc->sc_free > sc->sc_total) | |
1007 | sc->sc_free = sc->sc_total; | |
1008 | if (sc->sc_dinodes < 0) | |
1009 | sc->sc_dinodes = 0; | |
1010 | ||
1011 | return 0; | |
1012 | } | |
1013 | ||
1014 | /** | |
1015 | * gfs2_statfs - Gather and return stats about the filesystem | |
1016 | * @sb: The superblock | |
1017 | * @statfsbuf: The buffer | |
1018 | * | |
1019 | * Returns: 0 on success or error code | |
1020 | */ | |
1021 | ||
1022 | static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf) | |
1023 | { | |
1024 | struct super_block *sb = dentry->d_inode->i_sb; | |
1025 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
1026 | struct gfs2_statfs_change_host sc; | |
1027 | int error; | |
1028 | ||
1029 | if (gfs2_tune_get(sdp, gt_statfs_slow)) | |
1030 | error = gfs2_statfs_slow(sdp, &sc); | |
1031 | else | |
1032 | error = gfs2_statfs_i(sdp, &sc); | |
1033 | ||
1034 | if (error) | |
1035 | return error; | |
1036 | ||
1037 | buf->f_type = GFS2_MAGIC; | |
1038 | buf->f_bsize = sdp->sd_sb.sb_bsize; | |
1039 | buf->f_blocks = sc.sc_total; | |
1040 | buf->f_bfree = sc.sc_free; | |
1041 | buf->f_bavail = sc.sc_free; | |
1042 | buf->f_files = sc.sc_dinodes + sc.sc_free; | |
1043 | buf->f_ffree = sc.sc_free; | |
1044 | buf->f_namelen = GFS2_FNAMESIZE; | |
1045 | ||
1046 | return 0; | |
1047 | } | |
1048 | ||
1049 | /** | |
1050 | * gfs2_remount_fs - called when the FS is remounted | |
1051 | * @sb: the filesystem | |
1052 | * @flags: the remount flags | |
1053 | * @data: extra data passed in (not used right now) | |
1054 | * | |
1055 | * Returns: errno | |
1056 | */ | |
1057 | ||
1058 | static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data) | |
1059 | { | |
1060 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
1061 | struct gfs2_args args = sdp->sd_args; /* Default to current settings */ | |
1062 | struct gfs2_tune *gt = &sdp->sd_tune; | |
1063 | int error; | |
1064 | ||
1065 | spin_lock(>->gt_spin); | |
1066 | args.ar_commit = gt->gt_log_flush_secs; | |
1067 | spin_unlock(>->gt_spin); | |
1068 | error = gfs2_mount_args(sdp, &args, data); | |
1069 | if (error) | |
1070 | return error; | |
1071 | ||
1072 | /* Not allowed to change locking details */ | |
1073 | if (strcmp(args.ar_lockproto, sdp->sd_args.ar_lockproto) || | |
1074 | strcmp(args.ar_locktable, sdp->sd_args.ar_locktable) || | |
1075 | strcmp(args.ar_hostdata, sdp->sd_args.ar_hostdata)) | |
1076 | return -EINVAL; | |
1077 | ||
1078 | /* Some flags must not be changed */ | |
1079 | if (args_neq(&args, &sdp->sd_args, spectator) || | |
1080 | args_neq(&args, &sdp->sd_args, ignore_local_fs) || | |
1081 | args_neq(&args, &sdp->sd_args, localflocks) || | |
1082 | args_neq(&args, &sdp->sd_args, localcaching) || | |
1083 | args_neq(&args, &sdp->sd_args, meta)) | |
1084 | return -EINVAL; | |
1085 | ||
1086 | if (sdp->sd_args.ar_spectator) | |
1087 | *flags |= MS_RDONLY; | |
1088 | ||
1089 | if ((sb->s_flags ^ *flags) & MS_RDONLY) { | |
1090 | if (*flags & MS_RDONLY) | |
1091 | error = gfs2_make_fs_ro(sdp); | |
1092 | else | |
1093 | error = gfs2_make_fs_rw(sdp); | |
1094 | if (error) | |
1095 | return error; | |
1096 | } | |
1097 | ||
1098 | sdp->sd_args = args; | |
1099 | if (sdp->sd_args.ar_posix_acl) | |
1100 | sb->s_flags |= MS_POSIXACL; | |
1101 | else | |
1102 | sb->s_flags &= ~MS_POSIXACL; | |
1103 | spin_lock(>->gt_spin); | |
1104 | gt->gt_log_flush_secs = args.ar_commit; | |
1105 | spin_unlock(>->gt_spin); | |
1106 | ||
8633ecfa | 1107 | gfs2_online_uevent(sdp); |
9e6e0a12 SW |
1108 | return 0; |
1109 | } | |
1110 | ||
1111 | /** | |
1112 | * gfs2_drop_inode - Drop an inode (test for remote unlink) | |
1113 | * @inode: The inode to drop | |
1114 | * | |
1115 | * If we've received a callback on an iopen lock then its because a | |
1116 | * remote node tried to deallocate the inode but failed due to this node | |
1117 | * still having the inode open. Here we mark the link count zero | |
1118 | * since we know that it must have reached zero if the GLF_DEMOTE flag | |
1119 | * is set on the iopen glock. If we didn't do a disk read since the | |
1120 | * remote node removed the final link then we might otherwise miss | |
1121 | * this event. This check ensures that this node will deallocate the | |
1122 | * inode's blocks, or alternatively pass the baton on to another | |
1123 | * node for later deallocation. | |
1124 | */ | |
1125 | ||
1126 | static void gfs2_drop_inode(struct inode *inode) | |
1127 | { | |
1128 | struct gfs2_inode *ip = GFS2_I(inode); | |
1129 | ||
1130 | if (test_bit(GIF_USER, &ip->i_flags) && inode->i_nlink) { | |
1131 | struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl; | |
1132 | if (gl && test_bit(GLF_DEMOTE, &gl->gl_flags)) | |
1133 | clear_nlink(inode); | |
1134 | } | |
1135 | generic_drop_inode(inode); | |
1136 | } | |
1137 | ||
1138 | /** | |
1139 | * gfs2_clear_inode - Deallocate an inode when VFS is done with it | |
1140 | * @inode: The VFS inode | |
1141 | * | |
1142 | */ | |
1143 | ||
1144 | static void gfs2_clear_inode(struct inode *inode) | |
1145 | { | |
1146 | struct gfs2_inode *ip = GFS2_I(inode); | |
1147 | ||
1148 | /* This tells us its a "real" inode and not one which only | |
1149 | * serves to contain an address space (see rgrp.c, meta_io.c) | |
1150 | * which therefore doesn't have its own glocks. | |
1151 | */ | |
1152 | if (test_bit(GIF_USER, &ip->i_flags)) { | |
1153 | ip->i_gl->gl_object = NULL; | |
1154 | gfs2_glock_put(ip->i_gl); | |
1155 | ip->i_gl = NULL; | |
1156 | if (ip->i_iopen_gh.gh_gl) { | |
1157 | ip->i_iopen_gh.gh_gl->gl_object = NULL; | |
1158 | gfs2_glock_dq_uninit(&ip->i_iopen_gh); | |
1159 | } | |
1160 | } | |
1161 | } | |
1162 | ||
1163 | static int is_ancestor(const struct dentry *d1, const struct dentry *d2) | |
1164 | { | |
1165 | do { | |
1166 | if (d1 == d2) | |
1167 | return 1; | |
1168 | d1 = d1->d_parent; | |
1169 | } while (!IS_ROOT(d1)); | |
1170 | return 0; | |
1171 | } | |
1172 | ||
1173 | /** | |
1174 | * gfs2_show_options - Show mount options for /proc/mounts | |
1175 | * @s: seq_file structure | |
1176 | * @mnt: vfsmount | |
1177 | * | |
1178 | * Returns: 0 on success or error code | |
1179 | */ | |
1180 | ||
1181 | static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt) | |
1182 | { | |
1183 | struct gfs2_sbd *sdp = mnt->mnt_sb->s_fs_info; | |
1184 | struct gfs2_args *args = &sdp->sd_args; | |
1185 | int lfsecs; | |
1186 | ||
1187 | if (is_ancestor(mnt->mnt_root, sdp->sd_master_dir)) | |
1188 | seq_printf(s, ",meta"); | |
1189 | if (args->ar_lockproto[0]) | |
1190 | seq_printf(s, ",lockproto=%s", args->ar_lockproto); | |
1191 | if (args->ar_locktable[0]) | |
1192 | seq_printf(s, ",locktable=%s", args->ar_locktable); | |
1193 | if (args->ar_hostdata[0]) | |
1194 | seq_printf(s, ",hostdata=%s", args->ar_hostdata); | |
1195 | if (args->ar_spectator) | |
1196 | seq_printf(s, ",spectator"); | |
1197 | if (args->ar_ignore_local_fs) | |
1198 | seq_printf(s, ",ignore_local_fs"); | |
1199 | if (args->ar_localflocks) | |
1200 | seq_printf(s, ",localflocks"); | |
1201 | if (args->ar_localcaching) | |
1202 | seq_printf(s, ",localcaching"); | |
1203 | if (args->ar_debug) | |
1204 | seq_printf(s, ",debug"); | |
1205 | if (args->ar_upgrade) | |
1206 | seq_printf(s, ",upgrade"); | |
1207 | if (args->ar_posix_acl) | |
1208 | seq_printf(s, ",acl"); | |
1209 | if (args->ar_quota != GFS2_QUOTA_DEFAULT) { | |
1210 | char *state; | |
1211 | switch (args->ar_quota) { | |
1212 | case GFS2_QUOTA_OFF: | |
1213 | state = "off"; | |
1214 | break; | |
1215 | case GFS2_QUOTA_ACCOUNT: | |
1216 | state = "account"; | |
1217 | break; | |
1218 | case GFS2_QUOTA_ON: | |
1219 | state = "on"; | |
1220 | break; | |
1221 | default: | |
1222 | state = "unknown"; | |
1223 | break; | |
1224 | } | |
1225 | seq_printf(s, ",quota=%s", state); | |
1226 | } | |
1227 | if (args->ar_suiddir) | |
1228 | seq_printf(s, ",suiddir"); | |
1229 | if (args->ar_data != GFS2_DATA_DEFAULT) { | |
1230 | char *state; | |
1231 | switch (args->ar_data) { | |
1232 | case GFS2_DATA_WRITEBACK: | |
1233 | state = "writeback"; | |
1234 | break; | |
1235 | case GFS2_DATA_ORDERED: | |
1236 | state = "ordered"; | |
1237 | break; | |
1238 | default: | |
1239 | state = "unknown"; | |
1240 | break; | |
1241 | } | |
1242 | seq_printf(s, ",data=%s", state); | |
1243 | } | |
1244 | if (args->ar_discard) | |
1245 | seq_printf(s, ",discard"); | |
1246 | lfsecs = sdp->sd_tune.gt_log_flush_secs; | |
1247 | if (lfsecs != 60) | |
1248 | seq_printf(s, ",commit=%d", lfsecs); | |
d34843d0 BP |
1249 | if (args->ar_errors != GFS2_ERRORS_DEFAULT) { |
1250 | const char *state; | |
1251 | ||
1252 | switch (args->ar_errors) { | |
1253 | case GFS2_ERRORS_WITHDRAW: | |
1254 | state = "withdraw"; | |
1255 | break; | |
1256 | case GFS2_ERRORS_PANIC: | |
1257 | state = "panic"; | |
1258 | break; | |
1259 | default: | |
1260 | state = "unknown"; | |
1261 | break; | |
1262 | } | |
1263 | seq_printf(s, ",errors=%s", state); | |
1264 | } | |
9e6e0a12 SW |
1265 | return 0; |
1266 | } | |
1267 | ||
1268 | /* | |
1269 | * We have to (at the moment) hold the inodes main lock to cover | |
1270 | * the gap between unlocking the shared lock on the iopen lock and | |
1271 | * taking the exclusive lock. I'd rather do a shared -> exclusive | |
1272 | * conversion on the iopen lock, but we can change that later. This | |
1273 | * is safe, just less efficient. | |
1274 | */ | |
1275 | ||
1276 | static void gfs2_delete_inode(struct inode *inode) | |
1277 | { | |
1278 | struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; | |
1279 | struct gfs2_inode *ip = GFS2_I(inode); | |
1280 | struct gfs2_holder gh; | |
1281 | int error; | |
1282 | ||
1283 | if (!test_bit(GIF_USER, &ip->i_flags)) | |
1284 | goto out; | |
1285 | ||
1286 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
1287 | if (unlikely(error)) { | |
1288 | gfs2_glock_dq_uninit(&ip->i_iopen_gh); | |
1289 | goto out; | |
1290 | } | |
1291 | ||
1292 | gfs2_glock_dq_wait(&ip->i_iopen_gh); | |
1293 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh); | |
1294 | error = gfs2_glock_nq(&ip->i_iopen_gh); | |
1295 | if (error) | |
1296 | goto out_truncate; | |
1297 | ||
1298 | if (S_ISDIR(inode->i_mode) && | |
1299 | (ip->i_diskflags & GFS2_DIF_EXHASH)) { | |
1300 | error = gfs2_dir_exhash_dealloc(ip); | |
1301 | if (error) | |
1302 | goto out_unlock; | |
1303 | } | |
1304 | ||
1305 | if (ip->i_eattr) { | |
1306 | error = gfs2_ea_dealloc(ip); | |
1307 | if (error) | |
1308 | goto out_unlock; | |
1309 | } | |
1310 | ||
1311 | if (!gfs2_is_stuffed(ip)) { | |
1312 | error = gfs2_file_dealloc(ip); | |
1313 | if (error) | |
1314 | goto out_unlock; | |
1315 | } | |
1316 | ||
1317 | error = gfs2_dinode_dealloc(ip); | |
1318 | if (error) | |
1319 | goto out_unlock; | |
1320 | ||
1321 | out_truncate: | |
1322 | error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks); | |
1323 | if (error) | |
1324 | goto out_unlock; | |
1325 | /* Needs to be done before glock release & also in a transaction */ | |
1326 | truncate_inode_pages(&inode->i_data, 0); | |
1327 | gfs2_trans_end(sdp); | |
1328 | ||
1329 | out_unlock: | |
1330 | if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags)) | |
1331 | gfs2_glock_dq(&ip->i_iopen_gh); | |
1332 | gfs2_holder_uninit(&ip->i_iopen_gh); | |
1333 | gfs2_glock_dq_uninit(&gh); | |
1334 | if (error && error != GLR_TRYFAILED && error != -EROFS) | |
1335 | fs_warn(sdp, "gfs2_delete_inode: %d\n", error); | |
1336 | out: | |
1337 | truncate_inode_pages(&inode->i_data, 0); | |
1338 | clear_inode(inode); | |
1339 | } | |
1340 | ||
1341 | static struct inode *gfs2_alloc_inode(struct super_block *sb) | |
1342 | { | |
1343 | struct gfs2_inode *ip; | |
1344 | ||
1345 | ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL); | |
1346 | if (ip) { | |
1347 | ip->i_flags = 0; | |
1348 | ip->i_gl = NULL; | |
1349 | } | |
1350 | return &ip->i_inode; | |
1351 | } | |
1352 | ||
1353 | static void gfs2_destroy_inode(struct inode *inode) | |
1354 | { | |
1355 | kmem_cache_free(gfs2_inode_cachep, inode); | |
1356 | } | |
1357 | ||
1358 | const struct super_operations gfs2_super_ops = { | |
1359 | .alloc_inode = gfs2_alloc_inode, | |
1360 | .destroy_inode = gfs2_destroy_inode, | |
1361 | .write_inode = gfs2_write_inode, | |
1362 | .delete_inode = gfs2_delete_inode, | |
1363 | .put_super = gfs2_put_super, | |
9e6e0a12 SW |
1364 | .sync_fs = gfs2_sync_fs, |
1365 | .freeze_fs = gfs2_freeze, | |
1366 | .unfreeze_fs = gfs2_unfreeze, | |
1367 | .statfs = gfs2_statfs, | |
1368 | .remount_fs = gfs2_remount_fs, | |
1369 | .clear_inode = gfs2_clear_inode, | |
1370 | .drop_inode = gfs2_drop_inode, | |
1371 | .show_options = gfs2_show_options, | |
1372 | }; | |
1373 |