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