Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
3a8a9a10 | 3 | * Copyright (C) 2004-2006 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 | ||
10 | #include <linux/sched.h> | |
11 | #include <linux/slab.h> | |
12 | #include <linux/spinlock.h> | |
13 | #include <linux/completion.h> | |
14 | #include <linux/buffer_head.h> | |
b3b94faa DT |
15 | #include <linux/statfs.h> |
16 | #include <linux/seq_file.h> | |
17 | #include <linux/mount.h> | |
18 | #include <linux/kthread.h> | |
19 | #include <linux/delay.h> | |
5c676f6d | 20 | #include <linux/gfs2_ondisk.h> |
feaa7bba | 21 | #include <linux/crc32.h> |
7d308590 | 22 | #include <linux/lm_interface.h> |
b3b94faa DT |
23 | |
24 | #include "gfs2.h" | |
5c676f6d | 25 | #include "incore.h" |
b3b94faa DT |
26 | #include "glock.h" |
27 | #include "inode.h" | |
28 | #include "lm.h" | |
29 | #include "log.h" | |
30 | #include "mount.h" | |
31 | #include "ops_super.h" | |
b3b94faa DT |
32 | #include "quota.h" |
33 | #include "recovery.h" | |
34 | #include "rgrp.h" | |
35 | #include "super.h" | |
36 | #include "sys.h" | |
5c676f6d | 37 | #include "util.h" |
feaa7bba SW |
38 | #include "trans.h" |
39 | #include "dir.h" | |
40 | #include "eattr.h" | |
41 | #include "bmap.h" | |
b3b94faa DT |
42 | |
43 | /** | |
44 | * gfs2_write_inode - Make sure the inode is stable on the disk | |
45 | * @inode: The inode | |
46 | * @sync: synchronous write flag | |
47 | * | |
48 | * Returns: errno | |
49 | */ | |
50 | ||
51 | static int gfs2_write_inode(struct inode *inode, int sync) | |
52 | { | |
feaa7bba SW |
53 | struct gfs2_inode *ip = GFS2_I(inode); |
54 | ||
55 | /* Check this is a "normal" inode */ | |
56 | if (inode->u.generic_ip) { | |
57 | if (current->flags & PF_MEMALLOC) | |
58 | return 0; | |
59 | if (sync) | |
3a8476dd | 60 | gfs2_log_flush(GFS2_SB(inode), ip->i_gl); |
feaa7bba | 61 | } |
b3b94faa DT |
62 | |
63 | return 0; | |
64 | } | |
65 | ||
66 | /** | |
67 | * gfs2_put_super - Unmount the filesystem | |
68 | * @sb: The VFS superblock | |
69 | * | |
70 | */ | |
71 | ||
72 | static void gfs2_put_super(struct super_block *sb) | |
73 | { | |
5c676f6d | 74 | struct gfs2_sbd *sdp = sb->s_fs_info; |
b3b94faa DT |
75 | int error; |
76 | ||
77 | if (!sdp) | |
78 | return; | |
79 | ||
86384605 | 80 | if (!strncmp(sb->s_type->name, "gfs2meta", 8)) |
5acd3967 | 81 | return; /* Nothing to do */ |
86384605 | 82 | |
b3b94faa DT |
83 | /* Unfreeze the filesystem, if we need to */ |
84 | ||
f55ab26a | 85 | mutex_lock(&sdp->sd_freeze_lock); |
b3b94faa DT |
86 | if (sdp->sd_freeze_count) |
87 | gfs2_glock_dq_uninit(&sdp->sd_freeze_gh); | |
f55ab26a | 88 | mutex_unlock(&sdp->sd_freeze_lock); |
b3b94faa | 89 | |
b3b94faa DT |
90 | kthread_stop(sdp->sd_quotad_process); |
91 | kthread_stop(sdp->sd_logd_process); | |
92 | kthread_stop(sdp->sd_recoverd_process); | |
93 | while (sdp->sd_glockd_num--) | |
94 | kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]); | |
95 | kthread_stop(sdp->sd_scand_process); | |
96 | ||
97 | if (!(sb->s_flags & MS_RDONLY)) { | |
98 | error = gfs2_make_fs_ro(sdp); | |
99 | if (error) | |
100 | gfs2_io_error(sdp); | |
101 | } | |
b3b94faa DT |
102 | /* At this point, we're through modifying the disk */ |
103 | ||
104 | /* Release stuff */ | |
105 | ||
f42faf4f SW |
106 | iput(sdp->sd_master_dir); |
107 | iput(sdp->sd_jindex); | |
108 | iput(sdp->sd_inum_inode); | |
109 | iput(sdp->sd_statfs_inode); | |
110 | iput(sdp->sd_rindex); | |
111 | iput(sdp->sd_quota_inode); | |
b3b94faa DT |
112 | |
113 | gfs2_glock_put(sdp->sd_rename_gl); | |
114 | gfs2_glock_put(sdp->sd_trans_gl); | |
115 | ||
116 | if (!sdp->sd_args.ar_spectator) { | |
117 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
118 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
119 | gfs2_glock_dq_uninit(&sdp->sd_ir_gh); | |
120 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); | |
b3b94faa | 121 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); |
f42faf4f SW |
122 | iput(sdp->sd_ir_inode); |
123 | iput(sdp->sd_sc_inode); | |
f42faf4f | 124 | iput(sdp->sd_qc_inode); |
b3b94faa DT |
125 | } |
126 | ||
127 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); | |
b3b94faa DT |
128 | gfs2_clear_rgrpd(sdp); |
129 | gfs2_jindex_free(sdp); | |
b3b94faa DT |
130 | /* Take apart glock structures and buffer lists */ |
131 | gfs2_gl_hash_clear(sdp, WAIT); | |
b3b94faa DT |
132 | /* Unmount the locking protocol */ |
133 | gfs2_lm_unmount(sdp); | |
134 | ||
135 | /* At this point, we're through participating in the lockspace */ | |
b3b94faa | 136 | gfs2_sys_fs_del(sdp); |
ff6af411 | 137 | kfree(sdp); |
b3b94faa DT |
138 | } |
139 | ||
140 | /** | |
141 | * gfs2_write_super - disk commit all incore transactions | |
142 | * @sb: the filesystem | |
143 | * | |
144 | * This function is called every time sync(2) is called. | |
b09e593d | 145 | * After this exits, all dirty buffers are synced. |
b3b94faa DT |
146 | */ |
147 | ||
148 | static void gfs2_write_super(struct super_block *sb) | |
149 | { | |
5acd3967 | 150 | gfs2_log_flush(sb->s_fs_info, NULL); |
b3b94faa DT |
151 | } |
152 | ||
153 | /** | |
154 | * gfs2_write_super_lockfs - prevent further writes to the filesystem | |
155 | * @sb: the VFS structure for the filesystem | |
156 | * | |
157 | */ | |
158 | ||
159 | static void gfs2_write_super_lockfs(struct super_block *sb) | |
160 | { | |
5c676f6d | 161 | struct gfs2_sbd *sdp = sb->s_fs_info; |
b3b94faa DT |
162 | int error; |
163 | ||
b3b94faa DT |
164 | for (;;) { |
165 | error = gfs2_freeze_fs(sdp); | |
166 | if (!error) | |
167 | break; | |
168 | ||
169 | switch (error) { | |
170 | case -EBUSY: | |
171 | fs_err(sdp, "waiting for recovery before freeze\n"); | |
172 | break; | |
173 | ||
174 | default: | |
175 | fs_err(sdp, "error freezing FS: %d\n", error); | |
176 | break; | |
177 | } | |
178 | ||
179 | fs_err(sdp, "retrying...\n"); | |
180 | msleep(1000); | |
181 | } | |
182 | } | |
183 | ||
184 | /** | |
185 | * gfs2_unlockfs - reallow writes to the filesystem | |
186 | * @sb: the VFS structure for the filesystem | |
187 | * | |
188 | */ | |
189 | ||
190 | static void gfs2_unlockfs(struct super_block *sb) | |
191 | { | |
2bdbc5d7 | 192 | gfs2_unfreeze_fs(sb->s_fs_info); |
b3b94faa DT |
193 | } |
194 | ||
195 | /** | |
196 | * gfs2_statfs - Gather and return stats about the filesystem | |
197 | * @sb: The superblock | |
198 | * @statfsbuf: The buffer | |
199 | * | |
200 | * Returns: 0 on success or error code | |
201 | */ | |
202 | ||
0c0834a3 | 203 | static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf) |
b3b94faa | 204 | { |
0c0834a3 | 205 | struct super_block *sb = dentry->d_inode->i_sb; |
5c676f6d | 206 | struct gfs2_sbd *sdp = sb->s_fs_info; |
b3b94faa DT |
207 | struct gfs2_statfs_change sc; |
208 | int error; | |
209 | ||
b3b94faa DT |
210 | if (gfs2_tune_get(sdp, gt_statfs_slow)) |
211 | error = gfs2_statfs_slow(sdp, &sc); | |
212 | else | |
213 | error = gfs2_statfs_i(sdp, &sc); | |
214 | ||
215 | if (error) | |
216 | return error; | |
217 | ||
b3b94faa DT |
218 | buf->f_type = GFS2_MAGIC; |
219 | buf->f_bsize = sdp->sd_sb.sb_bsize; | |
220 | buf->f_blocks = sc.sc_total; | |
221 | buf->f_bfree = sc.sc_free; | |
222 | buf->f_bavail = sc.sc_free; | |
223 | buf->f_files = sc.sc_dinodes + sc.sc_free; | |
224 | buf->f_ffree = sc.sc_free; | |
225 | buf->f_namelen = GFS2_FNAMESIZE; | |
226 | ||
227 | return 0; | |
228 | } | |
229 | ||
230 | /** | |
231 | * gfs2_remount_fs - called when the FS is remounted | |
232 | * @sb: the filesystem | |
233 | * @flags: the remount flags | |
234 | * @data: extra data passed in (not used right now) | |
235 | * | |
236 | * Returns: errno | |
237 | */ | |
238 | ||
239 | static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data) | |
240 | { | |
5c676f6d | 241 | struct gfs2_sbd *sdp = sb->s_fs_info; |
b3b94faa DT |
242 | int error; |
243 | ||
b3b94faa DT |
244 | error = gfs2_mount_args(sdp, data, 1); |
245 | if (error) | |
246 | return error; | |
247 | ||
248 | if (sdp->sd_args.ar_spectator) | |
249 | *flags |= MS_RDONLY; | |
250 | else { | |
251 | if (*flags & MS_RDONLY) { | |
252 | if (!(sb->s_flags & MS_RDONLY)) | |
253 | error = gfs2_make_fs_ro(sdp); | |
254 | } else if (!(*flags & MS_RDONLY) && | |
255 | (sb->s_flags & MS_RDONLY)) { | |
256 | error = gfs2_make_fs_rw(sdp); | |
257 | } | |
258 | } | |
259 | ||
260 | if (*flags & (MS_NOATIME | MS_NODIRATIME)) | |
261 | set_bit(SDF_NOATIME, &sdp->sd_flags); | |
262 | else | |
263 | clear_bit(SDF_NOATIME, &sdp->sd_flags); | |
264 | ||
265 | /* Don't let the VFS update atimes. GFS2 handles this itself. */ | |
266 | *flags |= MS_NOATIME | MS_NODIRATIME; | |
267 | ||
268 | return error; | |
269 | } | |
270 | ||
271 | /** | |
272 | * gfs2_clear_inode - Deallocate an inode when VFS is done with it | |
273 | * @inode: The VFS inode | |
274 | * | |
275 | */ | |
276 | ||
277 | static void gfs2_clear_inode(struct inode *inode) | |
278 | { | |
feaa7bba SW |
279 | /* This tells us its a "real" inode and not one which only |
280 | * serves to contain an address space (see rgrp.c, meta_io.c) | |
281 | * which therefore doesn't have its own glocks. | |
282 | */ | |
283 | if (inode->u.generic_ip) { | |
284 | struct gfs2_inode *ip = GFS2_I(inode); | |
285 | gfs2_glock_inode_squish(inode); | |
286 | gfs2_assert(inode->i_sb->s_fs_info, ip->i_gl->gl_state == LM_ST_UNLOCKED); | |
287 | ip->i_gl->gl_object = NULL; | |
b3b94faa | 288 | gfs2_glock_schedule_for_reclaim(ip->i_gl); |
feaa7bba SW |
289 | gfs2_glock_put(ip->i_gl); |
290 | ip->i_gl = NULL; | |
291 | if (ip->i_iopen_gh.gh_gl) | |
292 | gfs2_glock_dq_uninit(&ip->i_iopen_gh); | |
b3b94faa DT |
293 | } |
294 | } | |
295 | ||
296 | /** | |
297 | * gfs2_show_options - Show mount options for /proc/mounts | |
298 | * @s: seq_file structure | |
299 | * @mnt: vfsmount | |
300 | * | |
301 | * Returns: 0 on success or error code | |
302 | */ | |
303 | ||
304 | static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt) | |
305 | { | |
5c676f6d | 306 | struct gfs2_sbd *sdp = mnt->mnt_sb->s_fs_info; |
b3b94faa DT |
307 | struct gfs2_args *args = &sdp->sd_args; |
308 | ||
b3b94faa DT |
309 | if (args->ar_lockproto[0]) |
310 | seq_printf(s, ",lockproto=%s", args->ar_lockproto); | |
311 | if (args->ar_locktable[0]) | |
312 | seq_printf(s, ",locktable=%s", args->ar_locktable); | |
313 | if (args->ar_hostdata[0]) | |
314 | seq_printf(s, ",hostdata=%s", args->ar_hostdata); | |
315 | if (args->ar_spectator) | |
316 | seq_printf(s, ",spectator"); | |
317 | if (args->ar_ignore_local_fs) | |
318 | seq_printf(s, ",ignore_local_fs"); | |
319 | if (args->ar_localflocks) | |
320 | seq_printf(s, ",localflocks"); | |
321 | if (args->ar_localcaching) | |
322 | seq_printf(s, ",localcaching"); | |
323 | if (args->ar_debug) | |
324 | seq_printf(s, ",debug"); | |
325 | if (args->ar_upgrade) | |
326 | seq_printf(s, ",upgrade"); | |
327 | if (args->ar_num_glockd != GFS2_GLOCKD_DEFAULT) | |
328 | seq_printf(s, ",num_glockd=%u", args->ar_num_glockd); | |
329 | if (args->ar_posix_acl) | |
330 | seq_printf(s, ",acl"); | |
331 | if (args->ar_quota != GFS2_QUOTA_DEFAULT) { | |
332 | char *state; | |
333 | switch (args->ar_quota) { | |
334 | case GFS2_QUOTA_OFF: | |
335 | state = "off"; | |
336 | break; | |
337 | case GFS2_QUOTA_ACCOUNT: | |
338 | state = "account"; | |
339 | break; | |
340 | case GFS2_QUOTA_ON: | |
341 | state = "on"; | |
342 | break; | |
343 | default: | |
344 | state = "unknown"; | |
345 | break; | |
346 | } | |
347 | seq_printf(s, ",quota=%s", state); | |
348 | } | |
349 | if (args->ar_suiddir) | |
350 | seq_printf(s, ",suiddir"); | |
351 | if (args->ar_data != GFS2_DATA_DEFAULT) { | |
352 | char *state; | |
353 | switch (args->ar_data) { | |
354 | case GFS2_DATA_WRITEBACK: | |
355 | state = "writeback"; | |
356 | break; | |
357 | case GFS2_DATA_ORDERED: | |
358 | state = "ordered"; | |
359 | break; | |
360 | default: | |
361 | state = "unknown"; | |
362 | break; | |
363 | } | |
364 | seq_printf(s, ",data=%s", state); | |
365 | } | |
366 | ||
367 | return 0; | |
368 | } | |
369 | ||
feaa7bba SW |
370 | /* |
371 | * We have to (at the moment) hold the inodes main lock to cover | |
372 | * the gap between unlocking the shared lock on the iopen lock and | |
373 | * taking the exclusive lock. I'd rather do a shared -> exclusive | |
374 | * conversion on the iopen lock, but we can change that later. This | |
375 | * is safe, just less efficient. | |
376 | */ | |
377 | static void gfs2_delete_inode(struct inode *inode) | |
378 | { | |
379 | struct gfs2_sbd *sdp = inode->i_sb->s_fs_info; | |
380 | struct gfs2_inode *ip = GFS2_I(inode); | |
381 | struct gfs2_holder gh; | |
382 | int error; | |
383 | ||
384 | if (!inode->u.generic_ip) | |
385 | goto out; | |
386 | ||
387 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &gh); | |
388 | if (unlikely(error)) { | |
389 | gfs2_glock_dq_uninit(&ip->i_iopen_gh); | |
390 | goto out; | |
391 | } | |
392 | ||
393 | gfs2_glock_dq(&ip->i_iopen_gh); | |
394 | gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh); | |
395 | error = gfs2_glock_nq(&ip->i_iopen_gh); | |
396 | if (error) | |
397 | goto out_uninit; | |
398 | ||
399 | if (S_ISDIR(ip->i_di.di_mode) && | |
400 | (ip->i_di.di_flags & GFS2_DIF_EXHASH)) { | |
401 | error = gfs2_dir_exhash_dealloc(ip); | |
402 | if (error) | |
403 | goto out_unlock; | |
404 | } | |
405 | ||
406 | if (ip->i_di.di_eattr) { | |
407 | error = gfs2_ea_dealloc(ip); | |
408 | if (error) | |
409 | goto out_unlock; | |
410 | } | |
411 | ||
412 | if (!gfs2_is_stuffed(ip)) { | |
413 | error = gfs2_file_dealloc(ip); | |
414 | if (error) | |
415 | goto out_unlock; | |
416 | } | |
417 | ||
418 | error = gfs2_dinode_dealloc(ip); | |
419 | ||
420 | out_unlock: | |
421 | gfs2_glock_dq(&ip->i_iopen_gh); | |
422 | out_uninit: | |
423 | gfs2_holder_uninit(&ip->i_iopen_gh); | |
424 | gfs2_glock_dq_uninit(&gh); | |
425 | if (error) | |
426 | fs_warn(sdp, "gfs2_delete_inode: %d\n", error); | |
427 | out: | |
428 | truncate_inode_pages(&inode->i_data, 0); | |
429 | clear_inode(inode); | |
430 | } | |
431 | ||
432 | ||
433 | ||
320dd101 SW |
434 | static struct inode *gfs2_alloc_inode(struct super_block *sb) |
435 | { | |
436 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
437 | struct gfs2_inode *ip; | |
438 | ||
439 | ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL); | |
440 | if (ip) { | |
441 | ip->i_flags = 0; | |
442 | ip->i_gl = NULL; | |
320dd101 SW |
443 | ip->i_greedy = gfs2_tune_get(sdp, gt_greedy_default); |
444 | ip->i_last_pfault = jiffies; | |
445 | } | |
446 | return &ip->i_inode; | |
447 | } | |
448 | ||
449 | static void gfs2_destroy_inode(struct inode *inode) | |
450 | { | |
451 | kmem_cache_free(gfs2_inode_cachep, inode); | |
452 | } | |
453 | ||
b3b94faa | 454 | struct super_operations gfs2_super_ops = { |
320dd101 SW |
455 | .alloc_inode = gfs2_alloc_inode, |
456 | .destroy_inode = gfs2_destroy_inode, | |
b3b94faa | 457 | .write_inode = gfs2_write_inode, |
feaa7bba | 458 | .delete_inode = gfs2_delete_inode, |
b3b94faa DT |
459 | .put_super = gfs2_put_super, |
460 | .write_super = gfs2_write_super, | |
461 | .write_super_lockfs = gfs2_write_super_lockfs, | |
462 | .unlockfs = gfs2_unlockfs, | |
463 | .statfs = gfs2_statfs, | |
464 | .remount_fs = gfs2_remount_fs, | |
465 | .clear_inode = gfs2_clear_inode, | |
466 | .show_options = gfs2_show_options, | |
467 | }; | |
468 |