Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
cf45b752 | 3 | * Copyright (C) 2004-2008 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/blkdev.h> |
16 | #include <linux/kthread.h> | |
86384605 AD |
17 | #include <linux/namei.h> |
18 | #include <linux/mount.h> | |
5c676f6d | 19 | #include <linux/gfs2_ondisk.h> |
fe64d517 | 20 | #include <linux/slow-work.h> |
cc632e7f | 21 | #include <linux/quotaops.h> |
b3b94faa DT |
22 | |
23 | #include "gfs2.h" | |
5c676f6d | 24 | #include "incore.h" |
da6dd40d | 25 | #include "bmap.h" |
b3b94faa DT |
26 | #include "glock.h" |
27 | #include "glops.h" | |
28 | #include "inode.h" | |
b3b94faa DT |
29 | #include "recovery.h" |
30 | #include "rgrp.h" | |
31 | #include "super.h" | |
b3b94faa | 32 | #include "sys.h" |
5c676f6d | 33 | #include "util.h" |
bb3b0e3d | 34 | #include "log.h" |
9ac1b4d9 | 35 | #include "quota.h" |
b5289681 | 36 | #include "dir.h" |
63997775 | 37 | #include "trace_gfs2.h" |
b3b94faa DT |
38 | |
39 | #define DO 0 | |
40 | #define UNDO 1 | |
41 | ||
9b8df98f SW |
42 | static const u32 gfs2_old_fs_formats[] = { |
43 | 0 | |
44 | }; | |
45 | ||
46 | static const u32 gfs2_old_multihost_formats[] = { | |
47 | 0 | |
48 | }; | |
49 | ||
50 | /** | |
51 | * gfs2_tune_init - Fill a gfs2_tune structure with default values | |
52 | * @gt: tune | |
53 | * | |
54 | */ | |
55 | ||
56 | static void gfs2_tune_init(struct gfs2_tune *gt) | |
57 | { | |
58 | spin_lock_init(>->gt_spin); | |
59 | ||
9b8df98f | 60 | gt->gt_incore_log_blocks = 1024; |
9b8df98f | 61 | gt->gt_logd_secs = 1; |
9b8df98f SW |
62 | gt->gt_quota_simul_sync = 64; |
63 | gt->gt_quota_warn_period = 10; | |
64 | gt->gt_quota_scale_num = 1; | |
65 | gt->gt_quota_scale_den = 1; | |
9b8df98f SW |
66 | gt->gt_new_files_jdata = 0; |
67 | gt->gt_max_readahead = 1 << 18; | |
9b8df98f | 68 | gt->gt_complain_secs = 10; |
9b8df98f SW |
69 | } |
70 | ||
b3b94faa DT |
71 | static struct gfs2_sbd *init_sbd(struct super_block *sb) |
72 | { | |
73 | struct gfs2_sbd *sdp; | |
b3b94faa | 74 | |
85d1da67 | 75 | sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL); |
b3b94faa DT |
76 | if (!sdp) |
77 | return NULL; | |
78 | ||
5c676f6d | 79 | sb->s_fs_info = sdp; |
b3b94faa DT |
80 | sdp->sd_vfs = sb; |
81 | ||
82 | gfs2_tune_init(&sdp->sd_tune); | |
83 | ||
e402746a SW |
84 | init_waitqueue_head(&sdp->sd_glock_wait); |
85 | atomic_set(&sdp->sd_glock_disposal, 0); | |
b3b94faa | 86 | spin_lock_init(&sdp->sd_statfs_spin); |
b3b94faa DT |
87 | |
88 | spin_lock_init(&sdp->sd_rindex_spin); | |
f55ab26a | 89 | mutex_init(&sdp->sd_rindex_mutex); |
b3b94faa DT |
90 | INIT_LIST_HEAD(&sdp->sd_rindex_list); |
91 | INIT_LIST_HEAD(&sdp->sd_rindex_mru_list); | |
b3b94faa DT |
92 | |
93 | INIT_LIST_HEAD(&sdp->sd_jindex_list); | |
94 | spin_lock_init(&sdp->sd_jindex_spin); | |
f55ab26a | 95 | mutex_init(&sdp->sd_jindex_mutex); |
b3b94faa | 96 | |
b3b94faa | 97 | INIT_LIST_HEAD(&sdp->sd_quota_list); |
f55ab26a | 98 | mutex_init(&sdp->sd_quota_mutex); |
37b2c837 | 99 | init_waitqueue_head(&sdp->sd_quota_wait); |
813e0c46 SW |
100 | INIT_LIST_HEAD(&sdp->sd_trunc_list); |
101 | spin_lock_init(&sdp->sd_trunc_lock); | |
b3b94faa DT |
102 | |
103 | spin_lock_init(&sdp->sd_log_lock); | |
b3b94faa | 104 | |
b3b94faa DT |
105 | INIT_LIST_HEAD(&sdp->sd_log_le_buf); |
106 | INIT_LIST_HEAD(&sdp->sd_log_le_revoke); | |
107 | INIT_LIST_HEAD(&sdp->sd_log_le_rg); | |
108 | INIT_LIST_HEAD(&sdp->sd_log_le_databuf); | |
d7b616e2 | 109 | INIT_LIST_HEAD(&sdp->sd_log_le_ordered); |
b3b94faa | 110 | |
71b86f56 | 111 | mutex_init(&sdp->sd_log_reserve_mutex); |
b3b94faa DT |
112 | INIT_LIST_HEAD(&sdp->sd_ail1_list); |
113 | INIT_LIST_HEAD(&sdp->sd_ail2_list); | |
114 | ||
484adff8 | 115 | init_rwsem(&sdp->sd_log_flush_lock); |
16615be1 SW |
116 | atomic_set(&sdp->sd_log_in_flight, 0); |
117 | init_waitqueue_head(&sdp->sd_log_flush_wait); | |
b3b94faa DT |
118 | |
119 | INIT_LIST_HEAD(&sdp->sd_revoke_list); | |
120 | ||
f55ab26a | 121 | mutex_init(&sdp->sd_freeze_lock); |
b3b94faa DT |
122 | |
123 | return sdp; | |
124 | } | |
125 | ||
b3b94faa | 126 | |
9b8df98f SW |
127 | /** |
128 | * gfs2_check_sb - Check superblock | |
129 | * @sdp: the filesystem | |
130 | * @sb: The superblock | |
131 | * @silent: Don't print a message if the check fails | |
132 | * | |
133 | * Checks the version code of the FS is one that we understand how to | |
134 | * read and that the sizes of the various on-disk structures have not | |
135 | * changed. | |
136 | */ | |
137 | ||
138 | static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent) | |
139 | { | |
140 | unsigned int x; | |
141 | ||
142 | if (sb->sb_magic != GFS2_MAGIC || | |
143 | sb->sb_type != GFS2_METATYPE_SB) { | |
144 | if (!silent) | |
145 | printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n"); | |
146 | return -EINVAL; | |
147 | } | |
148 | ||
149 | /* If format numbers match exactly, we're done. */ | |
150 | ||
151 | if (sb->sb_fs_format == GFS2_FORMAT_FS && | |
152 | sb->sb_multihost_format == GFS2_FORMAT_MULTI) | |
153 | return 0; | |
154 | ||
155 | if (sb->sb_fs_format != GFS2_FORMAT_FS) { | |
156 | for (x = 0; gfs2_old_fs_formats[x]; x++) | |
157 | if (gfs2_old_fs_formats[x] == sb->sb_fs_format) | |
158 | break; | |
159 | ||
160 | if (!gfs2_old_fs_formats[x]) { | |
161 | printk(KERN_WARNING | |
162 | "GFS2: code version (%u, %u) is incompatible " | |
163 | "with ondisk format (%u, %u)\n", | |
164 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
165 | sb->sb_fs_format, sb->sb_multihost_format); | |
166 | printk(KERN_WARNING | |
167 | "GFS2: I don't know how to upgrade this FS\n"); | |
168 | return -EINVAL; | |
169 | } | |
170 | } | |
171 | ||
172 | if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) { | |
173 | for (x = 0; gfs2_old_multihost_formats[x]; x++) | |
174 | if (gfs2_old_multihost_formats[x] == | |
175 | sb->sb_multihost_format) | |
176 | break; | |
177 | ||
178 | if (!gfs2_old_multihost_formats[x]) { | |
179 | printk(KERN_WARNING | |
180 | "GFS2: code version (%u, %u) is incompatible " | |
181 | "with ondisk format (%u, %u)\n", | |
182 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
183 | sb->sb_fs_format, sb->sb_multihost_format); | |
184 | printk(KERN_WARNING | |
185 | "GFS2: I don't know how to upgrade this FS\n"); | |
186 | return -EINVAL; | |
187 | } | |
188 | } | |
189 | ||
190 | if (!sdp->sd_args.ar_upgrade) { | |
191 | printk(KERN_WARNING | |
192 | "GFS2: code version (%u, %u) is incompatible " | |
193 | "with ondisk format (%u, %u)\n", | |
194 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
195 | sb->sb_fs_format, sb->sb_multihost_format); | |
196 | printk(KERN_INFO | |
197 | "GFS2: Use the \"upgrade\" mount option to upgrade " | |
198 | "the FS\n"); | |
199 | printk(KERN_INFO "GFS2: See the manual for more details\n"); | |
200 | return -EINVAL; | |
201 | } | |
202 | ||
203 | return 0; | |
204 | } | |
205 | ||
206 | static void end_bio_io_page(struct bio *bio, int error) | |
207 | { | |
208 | struct page *page = bio->bi_private; | |
209 | ||
210 | if (!error) | |
211 | SetPageUptodate(page); | |
212 | else | |
213 | printk(KERN_WARNING "gfs2: error %d reading superblock\n", error); | |
214 | unlock_page(page); | |
215 | } | |
216 | ||
217 | static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf) | |
218 | { | |
219 | const struct gfs2_sb *str = buf; | |
220 | ||
221 | sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic); | |
222 | sb->sb_type = be32_to_cpu(str->sb_header.mh_type); | |
223 | sb->sb_format = be32_to_cpu(str->sb_header.mh_format); | |
224 | sb->sb_fs_format = be32_to_cpu(str->sb_fs_format); | |
225 | sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format); | |
226 | sb->sb_bsize = be32_to_cpu(str->sb_bsize); | |
227 | sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift); | |
228 | sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr); | |
229 | sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino); | |
230 | sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr); | |
231 | sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino); | |
232 | ||
233 | memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN); | |
234 | memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN); | |
02e3cc70 | 235 | memcpy(sb->sb_uuid, str->sb_uuid, 16); |
9b8df98f SW |
236 | } |
237 | ||
238 | /** | |
239 | * gfs2_read_super - Read the gfs2 super block from disk | |
240 | * @sdp: The GFS2 super block | |
241 | * @sector: The location of the super block | |
242 | * @error: The error code to return | |
243 | * | |
244 | * This uses the bio functions to read the super block from disk | |
245 | * because we want to be 100% sure that we never read cached data. | |
246 | * A super block is read twice only during each GFS2 mount and is | |
247 | * never written to by the filesystem. The first time its read no | |
248 | * locks are held, and the only details which are looked at are those | |
249 | * relating to the locking protocol. Once locking is up and working, | |
250 | * the sb is read again under the lock to establish the location of | |
251 | * the master directory (contains pointers to journals etc) and the | |
252 | * root directory. | |
253 | * | |
254 | * Returns: 0 on success or error | |
255 | */ | |
256 | ||
257 | static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector) | |
258 | { | |
259 | struct super_block *sb = sdp->sd_vfs; | |
260 | struct gfs2_sb *p; | |
261 | struct page *page; | |
262 | struct bio *bio; | |
263 | ||
264 | page = alloc_page(GFP_NOFS); | |
265 | if (unlikely(!page)) | |
266 | return -ENOBUFS; | |
267 | ||
268 | ClearPageUptodate(page); | |
269 | ClearPageDirty(page); | |
270 | lock_page(page); | |
271 | ||
272 | bio = bio_alloc(GFP_NOFS, 1); | |
9b8df98f SW |
273 | bio->bi_sector = sector * (sb->s_blocksize >> 9); |
274 | bio->bi_bdev = sb->s_bdev; | |
275 | bio_add_page(bio, page, PAGE_SIZE, 0); | |
276 | ||
277 | bio->bi_end_io = end_bio_io_page; | |
278 | bio->bi_private = page; | |
279 | submit_bio(READ_SYNC | (1 << BIO_RW_META), bio); | |
280 | wait_on_page_locked(page); | |
281 | bio_put(bio); | |
282 | if (!PageUptodate(page)) { | |
283 | __free_page(page); | |
284 | return -EIO; | |
285 | } | |
286 | p = kmap(page); | |
287 | gfs2_sb_in(&sdp->sd_sb, p); | |
288 | kunmap(page); | |
289 | __free_page(page); | |
290 | return 0; | |
291 | } | |
6bac243f | 292 | |
9b8df98f SW |
293 | /** |
294 | * gfs2_read_sb - Read super block | |
295 | * @sdp: The GFS2 superblock | |
9b8df98f SW |
296 | * @silent: Don't print message if mount fails |
297 | * | |
298 | */ | |
299 | ||
6bac243f | 300 | static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent) |
9b8df98f SW |
301 | { |
302 | u32 hash_blocks, ind_blocks, leaf_blocks; | |
303 | u32 tmp_blocks; | |
304 | unsigned int x; | |
305 | int error; | |
306 | ||
307 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); | |
308 | if (error) { | |
309 | if (!silent) | |
310 | fs_err(sdp, "can't read superblock\n"); | |
311 | return error; | |
312 | } | |
313 | ||
314 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
315 | if (error) | |
316 | return error; | |
317 | ||
318 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
319 | GFS2_BASIC_BLOCK_SHIFT; | |
320 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
321 | sdp->sd_diptrs = (sdp->sd_sb.sb_bsize - | |
322 | sizeof(struct gfs2_dinode)) / sizeof(u64); | |
323 | sdp->sd_inptrs = (sdp->sd_sb.sb_bsize - | |
324 | sizeof(struct gfs2_meta_header)) / sizeof(u64); | |
325 | sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header); | |
326 | sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2; | |
327 | sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1; | |
328 | sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64); | |
329 | sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize - | |
330 | sizeof(struct gfs2_meta_header)) / | |
331 | sizeof(struct gfs2_quota_change); | |
332 | ||
333 | /* Compute maximum reservation required to add a entry to a directory */ | |
334 | ||
335 | hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH), | |
336 | sdp->sd_jbsize); | |
337 | ||
338 | ind_blocks = 0; | |
339 | for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) { | |
340 | tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs); | |
341 | ind_blocks += tmp_blocks; | |
342 | } | |
343 | ||
344 | leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH; | |
345 | ||
346 | sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks; | |
347 | ||
348 | sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize - | |
349 | sizeof(struct gfs2_dinode); | |
350 | sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs; | |
351 | for (x = 2;; x++) { | |
352 | u64 space, d; | |
353 | u32 m; | |
354 | ||
355 | space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs; | |
356 | d = space; | |
357 | m = do_div(d, sdp->sd_inptrs); | |
358 | ||
359 | if (d != sdp->sd_heightsize[x - 1] || m) | |
360 | break; | |
361 | sdp->sd_heightsize[x] = space; | |
362 | } | |
363 | sdp->sd_max_height = x; | |
364 | sdp->sd_heightsize[x] = ~0; | |
365 | gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT); | |
366 | ||
367 | sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize - | |
368 | sizeof(struct gfs2_dinode); | |
369 | sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs; | |
370 | for (x = 2;; x++) { | |
371 | u64 space, d; | |
372 | u32 m; | |
373 | ||
374 | space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs; | |
375 | d = space; | |
376 | m = do_div(d, sdp->sd_inptrs); | |
377 | ||
378 | if (d != sdp->sd_jheightsize[x - 1] || m) | |
379 | break; | |
380 | sdp->sd_jheightsize[x] = space; | |
381 | } | |
382 | sdp->sd_max_jheight = x; | |
383 | sdp->sd_jheightsize[x] = ~0; | |
384 | gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT); | |
385 | ||
386 | return 0; | |
387 | } | |
388 | ||
b3b94faa DT |
389 | static int init_names(struct gfs2_sbd *sdp, int silent) |
390 | { | |
b3b94faa DT |
391 | char *proto, *table; |
392 | int error = 0; | |
393 | ||
394 | proto = sdp->sd_args.ar_lockproto; | |
395 | table = sdp->sd_args.ar_locktable; | |
396 | ||
397 | /* Try to autodetect */ | |
398 | ||
399 | if (!proto[0] || !table[0]) { | |
bb8d8a6f SW |
400 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); |
401 | if (error) | |
402 | return error; | |
3cf1e7be SW |
403 | |
404 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
b3b94faa DT |
405 | if (error) |
406 | goto out; | |
407 | ||
408 | if (!proto[0]) | |
3cf1e7be | 409 | proto = sdp->sd_sb.sb_lockproto; |
b3b94faa | 410 | if (!table[0]) |
3cf1e7be | 411 | table = sdp->sd_sb.sb_locktable; |
b3b94faa DT |
412 | } |
413 | ||
414 | if (!table[0]) | |
415 | table = sdp->sd_vfs->s_id; | |
416 | ||
00377d8e JD |
417 | strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN); |
418 | strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN); | |
b3b94faa | 419 | |
5d35e31f DC |
420 | table = sdp->sd_table_name; |
421 | while ((table = strchr(table, '/'))) | |
b35997d4 RP |
422 | *table = '_'; |
423 | ||
a91ea69f | 424 | out: |
b3b94faa DT |
425 | return error; |
426 | } | |
427 | ||
428 | static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh, | |
429 | int undo) | |
430 | { | |
b3b94faa DT |
431 | int error = 0; |
432 | ||
433 | if (undo) | |
434 | goto fail_trans; | |
435 | ||
b3b94faa DT |
436 | error = gfs2_glock_nq_num(sdp, |
437 | GFS2_MOUNT_LOCK, &gfs2_nondisk_glops, | |
438 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE, | |
439 | mount_gh); | |
440 | if (error) { | |
441 | fs_err(sdp, "can't acquire mount glock: %d\n", error); | |
442 | goto fail; | |
443 | } | |
444 | ||
445 | error = gfs2_glock_nq_num(sdp, | |
446 | GFS2_LIVE_LOCK, &gfs2_nondisk_glops, | |
447 | LM_ST_SHARED, | |
579b78a4 | 448 | LM_FLAG_NOEXP | GL_EXACT, |
b3b94faa DT |
449 | &sdp->sd_live_gh); |
450 | if (error) { | |
451 | fs_err(sdp, "can't acquire live glock: %d\n", error); | |
452 | goto fail_mount; | |
453 | } | |
454 | ||
455 | error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops, | |
456 | CREATE, &sdp->sd_rename_gl); | |
457 | if (error) { | |
458 | fs_err(sdp, "can't create rename glock: %d\n", error); | |
459 | goto fail_live; | |
460 | } | |
461 | ||
462 | error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops, | |
463 | CREATE, &sdp->sd_trans_gl); | |
464 | if (error) { | |
465 | fs_err(sdp, "can't create transaction glock: %d\n", error); | |
466 | goto fail_rename; | |
467 | } | |
b3b94faa DT |
468 | |
469 | return 0; | |
470 | ||
feaa7bba | 471 | fail_trans: |
b3b94faa | 472 | gfs2_glock_put(sdp->sd_trans_gl); |
feaa7bba | 473 | fail_rename: |
b3b94faa | 474 | gfs2_glock_put(sdp->sd_rename_gl); |
feaa7bba | 475 | fail_live: |
b3b94faa | 476 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); |
feaa7bba | 477 | fail_mount: |
b3b94faa | 478 | gfs2_glock_dq_uninit(mount_gh); |
feaa7bba | 479 | fail: |
b3b94faa DT |
480 | return error; |
481 | } | |
482 | ||
9b8df98f SW |
483 | static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr, |
484 | u64 no_addr, const char *name) | |
f42faf4f | 485 | { |
9b8df98f SW |
486 | struct gfs2_sbd *sdp = sb->s_fs_info; |
487 | struct dentry *dentry; | |
488 | struct inode *inode; | |
489 | ||
490 | inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0); | |
491 | if (IS_ERR(inode)) { | |
492 | fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode)); | |
493 | return PTR_ERR(inode); | |
494 | } | |
495 | dentry = d_alloc_root(inode); | |
496 | if (!dentry) { | |
497 | fs_err(sdp, "can't alloc %s dentry\n", name); | |
498 | iput(inode); | |
499 | return -ENOMEM; | |
500 | } | |
501 | dentry->d_op = &gfs2_dops; | |
502 | *dptr = dentry; | |
503 | return 0; | |
f42faf4f SW |
504 | } |
505 | ||
9b8df98f | 506 | static int init_sb(struct gfs2_sbd *sdp, int silent) |
b3b94faa DT |
507 | { |
508 | struct super_block *sb = sdp->sd_vfs; | |
509 | struct gfs2_holder sb_gh; | |
dbb7cae2 | 510 | u64 no_addr; |
9b8df98f | 511 | int ret; |
b3b94faa | 512 | |
9b8df98f SW |
513 | ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops, |
514 | LM_ST_SHARED, 0, &sb_gh); | |
515 | if (ret) { | |
516 | fs_err(sdp, "can't acquire superblock glock: %d\n", ret); | |
517 | return ret; | |
b3b94faa | 518 | } |
907b9bce | 519 | |
6bac243f | 520 | ret = gfs2_read_sb(sdp, silent); |
9b8df98f SW |
521 | if (ret) { |
522 | fs_err(sdp, "can't read superblock: %d\n", ret); | |
b3b94faa DT |
523 | goto out; |
524 | } | |
525 | ||
526 | /* Set up the buffer cache and SB for real */ | |
e1defc4f | 527 | if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) { |
9b8df98f | 528 | ret = -EINVAL; |
b3b94faa DT |
529 | fs_err(sdp, "FS block size (%u) is too small for device " |
530 | "block size (%u)\n", | |
e1defc4f | 531 | sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev)); |
b3b94faa DT |
532 | goto out; |
533 | } | |
534 | if (sdp->sd_sb.sb_bsize > PAGE_SIZE) { | |
9b8df98f | 535 | ret = -EINVAL; |
b3b94faa DT |
536 | fs_err(sdp, "FS block size (%u) is too big for machine " |
537 | "page size (%u)\n", | |
538 | sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE); | |
539 | goto out; | |
540 | } | |
b3b94faa DT |
541 | sb_set_blocksize(sb, sdp->sd_sb.sb_bsize); |
542 | ||
f42faf4f | 543 | /* Get the root inode */ |
dbb7cae2 | 544 | no_addr = sdp->sd_sb.sb_root_dir.no_addr; |
9b8df98f SW |
545 | ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root"); |
546 | if (ret) | |
f42faf4f | 547 | goto out; |
b3b94faa | 548 | |
9b8df98f SW |
549 | /* Get the master inode */ |
550 | no_addr = sdp->sd_sb.sb_master_dir.no_addr; | |
551 | ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master"); | |
552 | if (ret) { | |
553 | dput(sdp->sd_root_dir); | |
554 | goto out; | |
555 | } | |
556 | sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir); | |
f42faf4f | 557 | out: |
b3b94faa | 558 | gfs2_glock_dq_uninit(&sb_gh); |
9b8df98f | 559 | return ret; |
b3b94faa DT |
560 | } |
561 | ||
da6dd40d BP |
562 | /** |
563 | * map_journal_extents - create a reusable "extent" mapping from all logical | |
564 | * blocks to all physical blocks for the given journal. This will save | |
565 | * us time when writing journal blocks. Most journals will have only one | |
566 | * extent that maps all their logical blocks. That's because gfs2.mkfs | |
567 | * arranges the journal blocks sequentially to maximize performance. | |
568 | * So the extent would map the first block for the entire file length. | |
569 | * However, gfs2_jadd can happen while file activity is happening, so | |
570 | * those journals may not be sequential. Less likely is the case where | |
571 | * the users created their own journals by mounting the metafs and | |
572 | * laying it out. But it's still possible. These journals might have | |
573 | * several extents. | |
574 | * | |
575 | * TODO: This should be done in bigger chunks rather than one block at a time, | |
576 | * but since it's only done at mount time, I'm not worried about the | |
577 | * time it takes. | |
578 | */ | |
579 | static int map_journal_extents(struct gfs2_sbd *sdp) | |
580 | { | |
581 | struct gfs2_jdesc *jd = sdp->sd_jdesc; | |
582 | unsigned int lb; | |
583 | u64 db, prev_db; /* logical block, disk block, prev disk block */ | |
584 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); | |
585 | struct gfs2_journal_extent *jext = NULL; | |
586 | struct buffer_head bh; | |
587 | int rc = 0; | |
588 | ||
da6dd40d BP |
589 | prev_db = 0; |
590 | ||
c9e98886 | 591 | for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) { |
da6dd40d BP |
592 | bh.b_state = 0; |
593 | bh.b_blocknr = 0; | |
594 | bh.b_size = 1 << ip->i_inode.i_blkbits; | |
595 | rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0); | |
596 | db = bh.b_blocknr; | |
597 | if (rc || !db) { | |
598 | printk(KERN_INFO "GFS2 journal mapping error %d: lb=" | |
599 | "%u db=%llu\n", rc, lb, (unsigned long long)db); | |
600 | break; | |
601 | } | |
602 | if (!prev_db || db != prev_db + 1) { | |
603 | jext = kzalloc(sizeof(struct gfs2_journal_extent), | |
604 | GFP_KERNEL); | |
605 | if (!jext) { | |
606 | printk(KERN_INFO "GFS2 error: out of memory " | |
607 | "mapping journal extents.\n"); | |
608 | rc = -ENOMEM; | |
609 | break; | |
610 | } | |
611 | jext->dblock = db; | |
612 | jext->lblock = lb; | |
613 | jext->blocks = 1; | |
614 | list_add_tail(&jext->extent_list, &jd->extent_list); | |
615 | } else { | |
616 | jext->blocks++; | |
617 | } | |
618 | prev_db = db; | |
619 | } | |
620 | return rc; | |
621 | } | |
622 | ||
f057f6cd | 623 | static void gfs2_others_may_mount(struct gfs2_sbd *sdp) |
da755fdb | 624 | { |
f057f6cd SW |
625 | char *message = "FIRSTMOUNT=Done"; |
626 | char *envp[] = { message, NULL }; | |
627 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
628 | ls->ls_first_done = 1; | |
629 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp); | |
da755fdb SW |
630 | } |
631 | ||
b5289681 SW |
632 | /** |
633 | * gfs2_jindex_hold - Grab a lock on the jindex | |
634 | * @sdp: The GFS2 superblock | |
635 | * @ji_gh: the holder for the jindex glock | |
636 | * | |
637 | * Returns: errno | |
638 | */ | |
639 | ||
640 | static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh) | |
641 | { | |
642 | struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex); | |
643 | struct qstr name; | |
644 | char buf[20]; | |
645 | struct gfs2_jdesc *jd; | |
646 | int error; | |
647 | ||
648 | name.name = buf; | |
649 | ||
650 | mutex_lock(&sdp->sd_jindex_mutex); | |
651 | ||
652 | for (;;) { | |
653 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh); | |
654 | if (error) | |
655 | break; | |
656 | ||
657 | name.len = sprintf(buf, "journal%u", sdp->sd_journals); | |
658 | name.hash = gfs2_disk_hash(name.name, name.len); | |
659 | ||
660 | error = gfs2_dir_check(sdp->sd_jindex, &name, NULL); | |
661 | if (error == -ENOENT) { | |
662 | error = 0; | |
663 | break; | |
664 | } | |
665 | ||
666 | gfs2_glock_dq_uninit(ji_gh); | |
667 | ||
668 | if (error) | |
669 | break; | |
670 | ||
671 | error = -ENOMEM; | |
672 | jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL); | |
673 | if (!jd) | |
674 | break; | |
675 | ||
676 | INIT_LIST_HEAD(&jd->extent_list); | |
fe64d517 | 677 | slow_work_init(&jd->jd_work, &gfs2_recover_ops); |
b5289681 SW |
678 | jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1); |
679 | if (!jd->jd_inode || IS_ERR(jd->jd_inode)) { | |
680 | if (!jd->jd_inode) | |
681 | error = -ENOENT; | |
682 | else | |
683 | error = PTR_ERR(jd->jd_inode); | |
684 | kfree(jd); | |
685 | break; | |
686 | } | |
687 | ||
688 | spin_lock(&sdp->sd_jindex_spin); | |
689 | jd->jd_jid = sdp->sd_journals++; | |
690 | list_add_tail(&jd->jd_list, &sdp->sd_jindex_list); | |
691 | spin_unlock(&sdp->sd_jindex_spin); | |
692 | } | |
693 | ||
694 | mutex_unlock(&sdp->sd_jindex_mutex); | |
695 | ||
696 | return error; | |
697 | } | |
698 | ||
b3b94faa DT |
699 | static int init_journal(struct gfs2_sbd *sdp, int undo) |
700 | { | |
9b8df98f | 701 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa | 702 | struct gfs2_holder ji_gh; |
5c676f6d | 703 | struct gfs2_inode *ip; |
b3b94faa DT |
704 | int jindex = 1; |
705 | int error = 0; | |
706 | ||
707 | if (undo) { | |
708 | jindex = 0; | |
fe64d517 | 709 | goto fail_jinode_gh; |
b3b94faa DT |
710 | } |
711 | ||
9b8df98f | 712 | sdp->sd_jindex = gfs2_lookup_simple(master, "jindex"); |
c752666c | 713 | if (IS_ERR(sdp->sd_jindex)) { |
b3b94faa | 714 | fs_err(sdp, "can't lookup journal index: %d\n", error); |
c752666c | 715 | return PTR_ERR(sdp->sd_jindex); |
b3b94faa | 716 | } |
feaa7bba | 717 | ip = GFS2_I(sdp->sd_jindex); |
b3b94faa DT |
718 | |
719 | /* Load in the journal index special file */ | |
720 | ||
721 | error = gfs2_jindex_hold(sdp, &ji_gh); | |
722 | if (error) { | |
723 | fs_err(sdp, "can't read journal index: %d\n", error); | |
724 | goto fail; | |
725 | } | |
726 | ||
0e5a9fb0 | 727 | error = -EUSERS; |
b3b94faa DT |
728 | if (!gfs2_jindex_size(sdp)) { |
729 | fs_err(sdp, "no journals!\n"); | |
907b9bce | 730 | goto fail_jindex; |
b3b94faa DT |
731 | } |
732 | ||
733 | if (sdp->sd_args.ar_spectator) { | |
734 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0); | |
fd041f0b | 735 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
b3b94faa DT |
736 | } else { |
737 | if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) { | |
738 | fs_err(sdp, "can't mount journal #%u\n", | |
739 | sdp->sd_lockstruct.ls_jid); | |
740 | fs_err(sdp, "there are only %u journals (0 - %u)\n", | |
741 | gfs2_jindex_size(sdp), | |
742 | gfs2_jindex_size(sdp) - 1); | |
743 | goto fail_jindex; | |
744 | } | |
745 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid); | |
746 | ||
feaa7bba | 747 | error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid, |
b3b94faa DT |
748 | &gfs2_journal_glops, |
749 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP, | |
750 | &sdp->sd_journal_gh); | |
751 | if (error) { | |
752 | fs_err(sdp, "can't acquire journal glock: %d\n", error); | |
753 | goto fail_jindex; | |
754 | } | |
755 | ||
feaa7bba SW |
756 | ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
757 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, | |
75be73a8 | 758 | LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE, |
b3b94faa DT |
759 | &sdp->sd_jinode_gh); |
760 | if (error) { | |
761 | fs_err(sdp, "can't acquire journal inode glock: %d\n", | |
762 | error); | |
763 | goto fail_journal_gh; | |
764 | } | |
765 | ||
766 | error = gfs2_jdesc_check(sdp->sd_jdesc); | |
767 | if (error) { | |
768 | fs_err(sdp, "my journal (%u) is bad: %d\n", | |
769 | sdp->sd_jdesc->jd_jid, error); | |
770 | goto fail_jinode_gh; | |
771 | } | |
fd041f0b | 772 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
da6dd40d BP |
773 | |
774 | /* Map the extents for this journal's blocks */ | |
775 | map_journal_extents(sdp); | |
b3b94faa | 776 | } |
63997775 | 777 | trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free)); |
b3b94faa DT |
778 | |
779 | if (sdp->sd_lockstruct.ls_first) { | |
780 | unsigned int x; | |
781 | for (x = 0; x < sdp->sd_journals; x++) { | |
c63e31c2 | 782 | error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x)); |
b3b94faa DT |
783 | if (error) { |
784 | fs_err(sdp, "error recovering journal %u: %d\n", | |
785 | x, error); | |
786 | goto fail_jinode_gh; | |
787 | } | |
788 | } | |
789 | ||
f057f6cd | 790 | gfs2_others_may_mount(sdp); |
b3b94faa | 791 | } else if (!sdp->sd_args.ar_spectator) { |
c63e31c2 | 792 | error = gfs2_recover_journal(sdp->sd_jdesc); |
b3b94faa DT |
793 | if (error) { |
794 | fs_err(sdp, "error recovering my journal: %d\n", error); | |
795 | goto fail_jinode_gh; | |
796 | } | |
797 | } | |
798 | ||
799 | set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags); | |
800 | gfs2_glock_dq_uninit(&ji_gh); | |
801 | jindex = 0; | |
802 | ||
b3b94faa DT |
803 | return 0; |
804 | ||
a91ea69f | 805 | fail_jinode_gh: |
b3b94faa DT |
806 | if (!sdp->sd_args.ar_spectator) |
807 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
a91ea69f | 808 | fail_journal_gh: |
b3b94faa DT |
809 | if (!sdp->sd_args.ar_spectator) |
810 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
a91ea69f | 811 | fail_jindex: |
b3b94faa DT |
812 | gfs2_jindex_free(sdp); |
813 | if (jindex) | |
814 | gfs2_glock_dq_uninit(&ji_gh); | |
a91ea69f | 815 | fail: |
f42faf4f | 816 | iput(sdp->sd_jindex); |
b3b94faa DT |
817 | return error; |
818 | } | |
819 | ||
b3b94faa DT |
820 | |
821 | static int init_inodes(struct gfs2_sbd *sdp, int undo) | |
822 | { | |
b3b94faa | 823 | int error = 0; |
5c676f6d | 824 | struct gfs2_inode *ip; |
9b8df98f | 825 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
826 | |
827 | if (undo) | |
f42faf4f SW |
828 | goto fail_qinode; |
829 | ||
f42faf4f SW |
830 | error = init_journal(sdp, undo); |
831 | if (error) | |
9b8df98f | 832 | goto fail; |
b3b94faa | 833 | |
b3b94faa | 834 | /* Read in the master statfs inode */ |
9b8df98f | 835 | sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs"); |
c752666c SW |
836 | if (IS_ERR(sdp->sd_statfs_inode)) { |
837 | error = PTR_ERR(sdp->sd_statfs_inode); | |
b3b94faa | 838 | fs_err(sdp, "can't read in statfs inode: %d\n", error); |
8d8291ae | 839 | goto fail_journal; |
b3b94faa DT |
840 | } |
841 | ||
842 | /* Read in the resource index inode */ | |
9b8df98f | 843 | sdp->sd_rindex = gfs2_lookup_simple(master, "rindex"); |
c752666c SW |
844 | if (IS_ERR(sdp->sd_rindex)) { |
845 | error = PTR_ERR(sdp->sd_rindex); | |
b3b94faa DT |
846 | fs_err(sdp, "can't get resource index inode: %d\n", error); |
847 | goto fail_statfs; | |
848 | } | |
feaa7bba | 849 | ip = GFS2_I(sdp->sd_rindex); |
cf45b752 | 850 | sdp->sd_rindex_uptodate = 0; |
b3b94faa DT |
851 | |
852 | /* Read in the quota inode */ | |
9b8df98f | 853 | sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota"); |
c752666c SW |
854 | if (IS_ERR(sdp->sd_quota_inode)) { |
855 | error = PTR_ERR(sdp->sd_quota_inode); | |
b3b94faa DT |
856 | fs_err(sdp, "can't get quota file inode: %d\n", error); |
857 | goto fail_rindex; | |
858 | } | |
b3b94faa DT |
859 | return 0; |
860 | ||
f42faf4f SW |
861 | fail_qinode: |
862 | iput(sdp->sd_quota_inode); | |
f42faf4f | 863 | fail_rindex: |
b3b94faa | 864 | gfs2_clear_rgrpd(sdp); |
f42faf4f | 865 | iput(sdp->sd_rindex); |
f42faf4f SW |
866 | fail_statfs: |
867 | iput(sdp->sd_statfs_inode); | |
f42faf4f SW |
868 | fail_journal: |
869 | init_journal(sdp, UNDO); | |
f42faf4f | 870 | fail: |
b3b94faa DT |
871 | return error; |
872 | } | |
873 | ||
874 | static int init_per_node(struct gfs2_sbd *sdp, int undo) | |
875 | { | |
f42faf4f | 876 | struct inode *pn = NULL; |
b3b94faa DT |
877 | char buf[30]; |
878 | int error = 0; | |
5c676f6d | 879 | struct gfs2_inode *ip; |
9b8df98f | 880 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
881 | |
882 | if (sdp->sd_args.ar_spectator) | |
883 | return 0; | |
884 | ||
885 | if (undo) | |
886 | goto fail_qc_gh; | |
887 | ||
9b8df98f | 888 | pn = gfs2_lookup_simple(master, "per_node"); |
c752666c SW |
889 | if (IS_ERR(pn)) { |
890 | error = PTR_ERR(pn); | |
b3b94faa DT |
891 | fs_err(sdp, "can't find per_node directory: %d\n", error); |
892 | return error; | |
893 | } | |
894 | ||
b3b94faa | 895 | sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
896 | sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf); |
897 | if (IS_ERR(sdp->sd_sc_inode)) { | |
898 | error = PTR_ERR(sdp->sd_sc_inode); | |
b3b94faa | 899 | fs_err(sdp, "can't find local \"sc\" file: %d\n", error); |
8d8291ae | 900 | goto fail; |
b3b94faa DT |
901 | } |
902 | ||
b3b94faa | 903 | sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
904 | sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf); |
905 | if (IS_ERR(sdp->sd_qc_inode)) { | |
906 | error = PTR_ERR(sdp->sd_qc_inode); | |
b3b94faa DT |
907 | fs_err(sdp, "can't find local \"qc\" file: %d\n", error); |
908 | goto fail_ut_i; | |
909 | } | |
910 | ||
f42faf4f | 911 | iput(pn); |
b3b94faa DT |
912 | pn = NULL; |
913 | ||
feaa7bba | 914 | ip = GFS2_I(sdp->sd_sc_inode); |
8d8291ae | 915 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
916 | &sdp->sd_sc_gh); |
917 | if (error) { | |
918 | fs_err(sdp, "can't lock local \"sc\" file: %d\n", error); | |
8d8291ae | 919 | goto fail_qc_i; |
b3b94faa DT |
920 | } |
921 | ||
feaa7bba | 922 | ip = GFS2_I(sdp->sd_qc_inode); |
8d8291ae | 923 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
924 | &sdp->sd_qc_gh); |
925 | if (error) { | |
926 | fs_err(sdp, "can't lock local \"qc\" file: %d\n", error); | |
927 | goto fail_ut_gh; | |
928 | } | |
929 | ||
930 | return 0; | |
931 | ||
a91ea69f | 932 | fail_qc_gh: |
b3b94faa | 933 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); |
a91ea69f | 934 | fail_ut_gh: |
b3b94faa | 935 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); |
a91ea69f | 936 | fail_qc_i: |
f42faf4f | 937 | iput(sdp->sd_qc_inode); |
a91ea69f | 938 | fail_ut_i: |
f42faf4f | 939 | iput(sdp->sd_sc_inode); |
a91ea69f | 940 | fail: |
b3b94faa | 941 | if (pn) |
f42faf4f | 942 | iput(pn); |
b3b94faa DT |
943 | return error; |
944 | } | |
945 | ||
946 | static int init_threads(struct gfs2_sbd *sdp, int undo) | |
947 | { | |
948 | struct task_struct *p; | |
949 | int error = 0; | |
950 | ||
951 | if (undo) | |
feaa7bba | 952 | goto fail_quotad; |
b3b94faa DT |
953 | |
954 | sdp->sd_log_flush_time = jiffies; | |
b3b94faa DT |
955 | |
956 | p = kthread_run(gfs2_logd, sdp, "gfs2_logd"); | |
957 | error = IS_ERR(p); | |
958 | if (error) { | |
959 | fs_err(sdp, "can't start logd thread: %d\n", error); | |
960 | return error; | |
961 | } | |
962 | sdp->sd_logd_process = p; | |
963 | ||
b3b94faa DT |
964 | p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad"); |
965 | error = IS_ERR(p); | |
966 | if (error) { | |
967 | fs_err(sdp, "can't start quotad thread: %d\n", error); | |
968 | goto fail; | |
969 | } | |
970 | sdp->sd_quotad_process = p; | |
971 | ||
b3b94faa DT |
972 | return 0; |
973 | ||
b3b94faa | 974 | |
feaa7bba | 975 | fail_quotad: |
b3b94faa | 976 | kthread_stop(sdp->sd_quotad_process); |
feaa7bba | 977 | fail: |
b3b94faa | 978 | kthread_stop(sdp->sd_logd_process); |
b3b94faa DT |
979 | return error; |
980 | } | |
da755fdb | 981 | |
f057f6cd SW |
982 | static const match_table_t nolock_tokens = { |
983 | { Opt_jid, "jid=%d\n", }, | |
984 | { Opt_err, NULL }, | |
985 | }; | |
986 | ||
8f05228e SW |
987 | static void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl) |
988 | { | |
989 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
990 | kmem_cache_free(cachep, gl); | |
991 | if (atomic_dec_and_test(&sdp->sd_glock_disposal)) | |
992 | wake_up(&sdp->sd_glock_wait); | |
993 | } | |
994 | ||
f057f6cd SW |
995 | static const struct lm_lockops nolock_ops = { |
996 | .lm_proto_name = "lock_nolock", | |
8f05228e | 997 | .lm_put_lock = nolock_put_lock, |
f057f6cd SW |
998 | .lm_tokens = &nolock_tokens, |
999 | }; | |
1000 | ||
da755fdb SW |
1001 | /** |
1002 | * gfs2_lm_mount - mount a locking protocol | |
1003 | * @sdp: the filesystem | |
3ad2f3fb | 1004 | * @args: mount arguments |
da755fdb SW |
1005 | * @silent: if 1, don't complain if the FS isn't a GFS2 fs |
1006 | * | |
1007 | * Returns: errno | |
1008 | */ | |
1009 | ||
1010 | static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent) | |
1011 | { | |
f057f6cd SW |
1012 | const struct lm_lockops *lm; |
1013 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
1014 | struct gfs2_args *args = &sdp->sd_args; | |
1015 | const char *proto = sdp->sd_proto_name; | |
1016 | const char *table = sdp->sd_table_name; | |
1017 | const char *fsname; | |
1018 | char *o, *options; | |
1019 | int ret; | |
da755fdb | 1020 | |
f057f6cd SW |
1021 | if (!strcmp("lock_nolock", proto)) { |
1022 | lm = &nolock_ops; | |
1023 | sdp->sd_args.ar_localflocks = 1; | |
1024 | sdp->sd_args.ar_localcaching = 1; | |
1025 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM | |
1026 | } else if (!strcmp("lock_dlm", proto)) { | |
1027 | lm = &gfs2_dlm_ops; | |
1028 | #endif | |
1029 | } else { | |
1030 | printk(KERN_INFO "GFS2: can't find protocol %s\n", proto); | |
1031 | return -ENOENT; | |
1032 | } | |
da755fdb SW |
1033 | |
1034 | fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table); | |
1035 | ||
f057f6cd SW |
1036 | ls->ls_ops = lm; |
1037 | ls->ls_first = 1; | |
da755fdb | 1038 | |
f057f6cd SW |
1039 | for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) { |
1040 | substring_t tmp[MAX_OPT_ARGS]; | |
1041 | int token, option; | |
1042 | ||
1043 | if (!o || !*o) | |
1044 | continue; | |
1045 | ||
1046 | token = match_token(o, *lm->lm_tokens, tmp); | |
1047 | switch (token) { | |
1048 | case Opt_jid: | |
1049 | ret = match_int(&tmp[0], &option); | |
1050 | if (ret || option < 0) | |
1051 | goto hostdata_error; | |
1052 | ls->ls_jid = option; | |
1053 | break; | |
1054 | case Opt_id: | |
2b88f7c5 | 1055 | /* Obsolete, but left for backward compat purposes */ |
f057f6cd SW |
1056 | break; |
1057 | case Opt_first: | |
1058 | ret = match_int(&tmp[0], &option); | |
1059 | if (ret || (option != 0 && option != 1)) | |
1060 | goto hostdata_error; | |
1061 | ls->ls_first = option; | |
1062 | break; | |
1063 | case Opt_nodir: | |
1064 | ret = match_int(&tmp[0], &option); | |
1065 | if (ret || (option != 0 && option != 1)) | |
1066 | goto hostdata_error; | |
1067 | ls->ls_nodir = option; | |
1068 | break; | |
1069 | case Opt_err: | |
1070 | default: | |
1071 | hostdata_error: | |
1072 | fs_info(sdp, "unknown hostdata (%s)\n", o); | |
1073 | return -EINVAL; | |
1074 | } | |
da755fdb SW |
1075 | } |
1076 | ||
1077 | if (sdp->sd_args.ar_spectator) | |
1078 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table); | |
1079 | else | |
1080 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table, | |
1081 | sdp->sd_lockstruct.ls_jid); | |
1082 | ||
f057f6cd SW |
1083 | fsname = strchr(table, ':'); |
1084 | if (fsname) | |
1085 | fsname++; | |
1086 | if (lm->lm_mount == NULL) { | |
1087 | fs_info(sdp, "Now mounting FS...\n"); | |
1088 | return 0; | |
da755fdb | 1089 | } |
f057f6cd SW |
1090 | ret = lm->lm_mount(sdp, fsname); |
1091 | if (ret == 0) | |
1092 | fs_info(sdp, "Joined cluster. Now mounting FS...\n"); | |
1093 | return ret; | |
da755fdb SW |
1094 | } |
1095 | ||
1096 | void gfs2_lm_unmount(struct gfs2_sbd *sdp) | |
1097 | { | |
f057f6cd SW |
1098 | const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops; |
1099 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) && | |
1100 | lm->lm_unmount) | |
1101 | lm->lm_unmount(sdp); | |
da755fdb | 1102 | } |
b3b94faa | 1103 | |
8633ecfa SW |
1104 | void gfs2_online_uevent(struct gfs2_sbd *sdp) |
1105 | { | |
1106 | struct super_block *sb = sdp->sd_vfs; | |
1107 | char ro[20]; | |
1108 | char spectator[20]; | |
1109 | char *envp[] = { ro, spectator, NULL }; | |
1110 | sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0); | |
1111 | sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0); | |
1112 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp); | |
1113 | } | |
1114 | ||
b3b94faa DT |
1115 | /** |
1116 | * fill_super - Read in superblock | |
1117 | * @sb: The VFS superblock | |
1118 | * @data: Mount options | |
1119 | * @silent: Don't complain if it's not a GFS2 filesystem | |
1120 | * | |
1121 | * Returns: errno | |
1122 | */ | |
1123 | ||
f55073ff | 1124 | static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent) |
b3b94faa DT |
1125 | { |
1126 | struct gfs2_sbd *sdp; | |
1127 | struct gfs2_holder mount_gh; | |
1128 | int error; | |
1129 | ||
1130 | sdp = init_sbd(sb); | |
1131 | if (!sdp) { | |
d92a8d48 | 1132 | printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n"); |
b3b94faa DT |
1133 | return -ENOMEM; |
1134 | } | |
f55073ff | 1135 | sdp->sd_args = *args; |
b3b94faa | 1136 | |
fe64d517 | 1137 | if (sdp->sd_args.ar_spectator) { |
6f04c1c7 | 1138 | sb->s_flags |= MS_RDONLY; |
fe64d517 SW |
1139 | set_bit(SDF_NORECOVERY, &sdp->sd_flags); |
1140 | } | |
6f04c1c7 SW |
1141 | if (sdp->sd_args.ar_posix_acl) |
1142 | sb->s_flags |= MS_POSIXACL; | |
f25934c5 CH |
1143 | if (sdp->sd_args.ar_nobarrier) |
1144 | set_bit(SDF_NOBARRIERS, &sdp->sd_flags); | |
6f04c1c7 | 1145 | |
719ee344 SW |
1146 | sb->s_magic = GFS2_MAGIC; |
1147 | sb->s_op = &gfs2_super_ops; | |
1148 | sb->s_export_op = &gfs2_export_ops; | |
40b78a32 | 1149 | sb->s_xattr = gfs2_xattr_handlers; |
cc632e7f SW |
1150 | sb->s_qcop = &gfs2_quotactl_ops; |
1151 | sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; | |
719ee344 SW |
1152 | sb->s_time_gran = 1; |
1153 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
419c93e0 SW |
1154 | |
1155 | /* Set up the buffer cache and fill in some fake block size values | |
1156 | to allow us to read-in the on-disk superblock. */ | |
1157 | sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK); | |
1158 | sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits; | |
1159 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
1160 | GFS2_BASIC_BLOCK_SHIFT; | |
1161 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
b3b94faa | 1162 | |
48c2b613 | 1163 | sdp->sd_tune.gt_log_flush_secs = sdp->sd_args.ar_commit; |
3d3c10f2 BM |
1164 | sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum; |
1165 | if (sdp->sd_args.ar_statfs_quantum) { | |
1166 | sdp->sd_tune.gt_statfs_slow = 0; | |
1167 | sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum; | |
1168 | } | |
1169 | else { | |
1170 | sdp->sd_tune.gt_statfs_slow = 1; | |
1171 | sdp->sd_tune.gt_statfs_quantum = 30; | |
1172 | } | |
48c2b613 | 1173 | |
b3b94faa DT |
1174 | error = init_names(sdp, silent); |
1175 | if (error) | |
1176 | goto fail; | |
1177 | ||
7c52b166 RP |
1178 | gfs2_create_debugfs_file(sdp); |
1179 | ||
b3b94faa DT |
1180 | error = gfs2_sys_fs_add(sdp); |
1181 | if (error) | |
1182 | goto fail; | |
1183 | ||
1184 | error = gfs2_lm_mount(sdp, silent); | |
1185 | if (error) | |
1186 | goto fail_sys; | |
1187 | ||
1188 | error = init_locking(sdp, &mount_gh, DO); | |
1189 | if (error) | |
1190 | goto fail_lm; | |
1191 | ||
9b8df98f | 1192 | error = init_sb(sdp, silent); |
b3b94faa DT |
1193 | if (error) |
1194 | goto fail_locking; | |
b3b94faa DT |
1195 | |
1196 | error = init_inodes(sdp, DO); | |
1197 | if (error) | |
f42faf4f | 1198 | goto fail_sb; |
b3b94faa DT |
1199 | |
1200 | error = init_per_node(sdp, DO); | |
1201 | if (error) | |
1202 | goto fail_inodes; | |
1203 | ||
1204 | error = gfs2_statfs_init(sdp); | |
1205 | if (error) { | |
1206 | fs_err(sdp, "can't initialize statfs subsystem: %d\n", error); | |
1207 | goto fail_per_node; | |
1208 | } | |
1209 | ||
1210 | error = init_threads(sdp, DO); | |
1211 | if (error) | |
1212 | goto fail_per_node; | |
1213 | ||
1214 | if (!(sb->s_flags & MS_RDONLY)) { | |
1215 | error = gfs2_make_fs_rw(sdp); | |
1216 | if (error) { | |
1217 | fs_err(sdp, "can't make FS RW: %d\n", error); | |
1218 | goto fail_threads; | |
1219 | } | |
1220 | } | |
1221 | ||
1222 | gfs2_glock_dq_uninit(&mount_gh); | |
8633ecfa | 1223 | gfs2_online_uevent(sdp); |
b3b94faa DT |
1224 | return 0; |
1225 | ||
a91ea69f | 1226 | fail_threads: |
b3b94faa | 1227 | init_threads(sdp, UNDO); |
a91ea69f | 1228 | fail_per_node: |
b3b94faa | 1229 | init_per_node(sdp, UNDO); |
a91ea69f | 1230 | fail_inodes: |
b3b94faa | 1231 | init_inodes(sdp, UNDO); |
a91ea69f | 1232 | fail_sb: |
9b8df98f SW |
1233 | if (sdp->sd_root_dir) |
1234 | dput(sdp->sd_root_dir); | |
1235 | if (sdp->sd_master_dir) | |
1236 | dput(sdp->sd_master_dir); | |
e7c8707e SW |
1237 | if (sb->s_root) |
1238 | dput(sb->s_root); | |
9b8df98f | 1239 | sb->s_root = NULL; |
a91ea69f | 1240 | fail_locking: |
b3b94faa | 1241 | init_locking(sdp, &mount_gh, UNDO); |
a91ea69f | 1242 | fail_lm: |
c1184f8a | 1243 | invalidate_inodes(sb); |
fefc03bf | 1244 | gfs2_gl_hash_clear(sdp); |
b3b94faa | 1245 | gfs2_lm_unmount(sdp); |
a91ea69f | 1246 | fail_sys: |
b3b94faa | 1247 | gfs2_sys_fs_del(sdp); |
a91ea69f | 1248 | fail: |
5f882096 | 1249 | gfs2_delete_debugfs_file(sdp); |
a2c45807 | 1250 | kfree(sdp); |
5c676f6d | 1251 | sb->s_fs_info = NULL; |
b3b94faa DT |
1252 | return error; |
1253 | } | |
1254 | ||
f55073ff | 1255 | static int set_gfs2_super(struct super_block *s, void *data) |
b3b94faa | 1256 | { |
f55073ff SW |
1257 | s->s_bdev = data; |
1258 | s->s_dev = s->s_bdev->bd_dev; | |
1259 | ||
1260 | /* | |
1261 | * We set the bdi here to the queue backing, file systems can | |
1262 | * overwrite this in ->fill_super() | |
1263 | */ | |
1264 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1265 | return 0; | |
86384605 AD |
1266 | } |
1267 | ||
f55073ff | 1268 | static int test_gfs2_super(struct super_block *s, void *ptr) |
f6d03139 SW |
1269 | { |
1270 | struct block_device *bdev = ptr; | |
1271 | return (bdev == s->s_bdev); | |
1272 | } | |
1273 | ||
f55073ff SW |
1274 | /** |
1275 | * gfs2_get_sb - Get the GFS2 superblock | |
1276 | * @fs_type: The GFS2 filesystem type | |
1277 | * @flags: Mount flags | |
1278 | * @dev_name: The name of the device | |
1279 | * @data: The mount arguments | |
1280 | * @mnt: The vfsmnt for this mount | |
1281 | * | |
1282 | * Q. Why not use get_sb_bdev() ? | |
1283 | * A. We need to select one of two root directories to mount, independent | |
1284 | * of whether this is the initial, or subsequent, mount of this sb | |
1285 | * | |
1286 | * Returns: 0 or -ve on error | |
1287 | */ | |
1288 | ||
1289 | static int gfs2_get_sb(struct file_system_type *fs_type, int flags, | |
1290 | const char *dev_name, void *data, struct vfsmount *mnt) | |
1291 | { | |
1292 | struct block_device *bdev; | |
1293 | struct super_block *s; | |
1294 | fmode_t mode = FMODE_READ; | |
1295 | int error; | |
1296 | struct gfs2_args args; | |
1297 | struct gfs2_sbd *sdp; | |
1298 | ||
1299 | if (!(flags & MS_RDONLY)) | |
1300 | mode |= FMODE_WRITE; | |
1301 | ||
1302 | bdev = open_bdev_exclusive(dev_name, mode, fs_type); | |
1303 | if (IS_ERR(bdev)) | |
1304 | return PTR_ERR(bdev); | |
1305 | ||
1306 | /* | |
1307 | * once the super is inserted into the list by sget, s_umount | |
1308 | * will protect the lockfs code from trying to start a snapshot | |
1309 | * while we are mounting | |
1310 | */ | |
1311 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
1312 | if (bdev->bd_fsfreeze_count > 0) { | |
1313 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1314 | error = -EBUSY; | |
1315 | goto error_bdev; | |
1316 | } | |
1317 | s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev); | |
1318 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1319 | error = PTR_ERR(s); | |
1320 | if (IS_ERR(s)) | |
1321 | goto error_bdev; | |
1322 | ||
1323 | memset(&args, 0, sizeof(args)); | |
1324 | args.ar_quota = GFS2_QUOTA_DEFAULT; | |
1325 | args.ar_data = GFS2_DATA_DEFAULT; | |
1326 | args.ar_commit = 60; | |
3d3c10f2 BM |
1327 | args.ar_statfs_quantum = 30; |
1328 | args.ar_quota_quantum = 60; | |
f55073ff SW |
1329 | args.ar_errors = GFS2_ERRORS_DEFAULT; |
1330 | ||
1331 | error = gfs2_mount_args(&args, data); | |
1332 | if (error) { | |
1333 | printk(KERN_WARNING "GFS2: can't parse mount arguments\n"); | |
1334 | if (s->s_root) | |
1335 | goto error_super; | |
1336 | deactivate_locked_super(s); | |
1337 | return error; | |
1338 | } | |
1339 | ||
1340 | if (s->s_root) { | |
1341 | error = -EBUSY; | |
1342 | if ((flags ^ s->s_flags) & MS_RDONLY) | |
1343 | goto error_super; | |
1344 | close_bdev_exclusive(bdev, mode); | |
1345 | } else { | |
1346 | char b[BDEVNAME_SIZE]; | |
1347 | ||
1348 | s->s_flags = flags; | |
1349 | s->s_mode = mode; | |
1350 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | |
1351 | sb_set_blocksize(s, block_size(bdev)); | |
1352 | error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0); | |
1353 | if (error) { | |
1354 | deactivate_locked_super(s); | |
1355 | return error; | |
1356 | } | |
1357 | s->s_flags |= MS_ACTIVE; | |
1358 | bdev->bd_super = s; | |
1359 | } | |
1360 | ||
1361 | sdp = s->s_fs_info; | |
1362 | mnt->mnt_sb = s; | |
1363 | if (args.ar_meta) | |
1364 | mnt->mnt_root = dget(sdp->sd_master_dir); | |
1365 | else | |
1366 | mnt->mnt_root = dget(sdp->sd_root_dir); | |
1367 | return 0; | |
1368 | ||
1369 | error_super: | |
1370 | deactivate_locked_super(s); | |
1371 | error_bdev: | |
1372 | close_bdev_exclusive(bdev, mode); | |
1373 | return error; | |
1374 | } | |
1375 | ||
f6d03139 SW |
1376 | static int set_meta_super(struct super_block *s, void *ptr) |
1377 | { | |
1378 | return -EINVAL; | |
1379 | } | |
1380 | ||
003dec89 SW |
1381 | static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags, |
1382 | const char *dev_name, void *data, struct vfsmount *mnt) | |
86384605 | 1383 | { |
f6d03139 | 1384 | struct super_block *s; |
003dec89 | 1385 | struct gfs2_sbd *sdp; |
e24977d4 | 1386 | struct path path; |
86384605 | 1387 | int error; |
907b9bce | 1388 | |
e24977d4 | 1389 | error = kern_path(dev_name, LOOKUP_FOLLOW, &path); |
86384605 | 1390 | if (error) { |
9b8df98f SW |
1391 | printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n", |
1392 | dev_name, error); | |
003dec89 | 1393 | return error; |
86384605 | 1394 | } |
f55073ff | 1395 | s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, |
f6d03139 | 1396 | path.dentry->d_inode->i_sb->s_bdev); |
e24977d4 | 1397 | path_put(&path); |
f6d03139 | 1398 | if (IS_ERR(s)) { |
86384605 | 1399 | printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n"); |
f6d03139 | 1400 | return PTR_ERR(s); |
86384605 | 1401 | } |
f55073ff SW |
1402 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
1403 | deactivate_locked_super(s); | |
1404 | return -EBUSY; | |
1405 | } | |
f6d03139 SW |
1406 | sdp = s->s_fs_info; |
1407 | mnt->mnt_sb = s; | |
9b8df98f SW |
1408 | mnt->mnt_root = dget(sdp->sd_master_dir); |
1409 | return 0; | |
b3b94faa DT |
1410 | } |
1411 | ||
419c93e0 | 1412 | static void gfs2_kill_sb(struct super_block *sb) |
86384605 AD |
1413 | { |
1414 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
88a19ad0 SW |
1415 | |
1416 | if (sdp == NULL) { | |
1417 | kill_block_super(sb); | |
1418 | return; | |
3af165ac | 1419 | } |
88a19ad0 SW |
1420 | |
1421 | gfs2_meta_syncfs(sdp); | |
1422 | dput(sdp->sd_root_dir); | |
1423 | dput(sdp->sd_master_dir); | |
1424 | sdp->sd_root_dir = NULL; | |
1425 | sdp->sd_master_dir = NULL; | |
fefc03bf | 1426 | shrink_dcache_sb(sb); |
9b8df98f | 1427 | kill_block_super(sb); |
88a19ad0 SW |
1428 | gfs2_delete_debugfs_file(sdp); |
1429 | kfree(sdp); | |
86384605 AD |
1430 | } |
1431 | ||
b3b94faa DT |
1432 | struct file_system_type gfs2_fs_type = { |
1433 | .name = "gfs2", | |
1434 | .fs_flags = FS_REQUIRES_DEV, | |
1435 | .get_sb = gfs2_get_sb, | |
419c93e0 SW |
1436 | .kill_sb = gfs2_kill_sb, |
1437 | .owner = THIS_MODULE, | |
1438 | }; | |
1439 | ||
1440 | struct file_system_type gfs2meta_fs_type = { | |
1441 | .name = "gfs2meta", | |
1442 | .fs_flags = FS_REQUIRES_DEV, | |
86384605 | 1443 | .get_sb = gfs2_get_sb_meta, |
b3b94faa DT |
1444 | .owner = THIS_MODULE, |
1445 | }; | |
1446 |