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> |
b3b94faa DT |
21 | |
22 | #include "gfs2.h" | |
5c676f6d | 23 | #include "incore.h" |
da6dd40d | 24 | #include "bmap.h" |
b3b94faa DT |
25 | #include "glock.h" |
26 | #include "glops.h" | |
27 | #include "inode.h" | |
b3b94faa DT |
28 | #include "recovery.h" |
29 | #include "rgrp.h" | |
30 | #include "super.h" | |
b3b94faa | 31 | #include "sys.h" |
5c676f6d | 32 | #include "util.h" |
bb3b0e3d | 33 | #include "log.h" |
9ac1b4d9 | 34 | #include "quota.h" |
b5289681 | 35 | #include "dir.h" |
63997775 | 36 | #include "trace_gfs2.h" |
b3b94faa DT |
37 | |
38 | #define DO 0 | |
39 | #define UNDO 1 | |
40 | ||
9b8df98f SW |
41 | static const u32 gfs2_old_fs_formats[] = { |
42 | 0 | |
43 | }; | |
44 | ||
45 | static const u32 gfs2_old_multihost_formats[] = { | |
46 | 0 | |
47 | }; | |
48 | ||
49 | /** | |
50 | * gfs2_tune_init - Fill a gfs2_tune structure with default values | |
51 | * @gt: tune | |
52 | * | |
53 | */ | |
54 | ||
55 | static void gfs2_tune_init(struct gfs2_tune *gt) | |
56 | { | |
57 | spin_lock_init(>->gt_spin); | |
58 | ||
9b8df98f | 59 | gt->gt_incore_log_blocks = 1024; |
9b8df98f | 60 | gt->gt_logd_secs = 1; |
9b8df98f SW |
61 | gt->gt_quota_simul_sync = 64; |
62 | gt->gt_quota_warn_period = 10; | |
63 | gt->gt_quota_scale_num = 1; | |
64 | gt->gt_quota_scale_den = 1; | |
9b8df98f | 65 | gt->gt_quota_quantum = 60; |
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; | |
70 | gt->gt_statfs_quantum = 30; | |
71 | gt->gt_statfs_slow = 0; | |
72 | } | |
73 | ||
b3b94faa DT |
74 | static struct gfs2_sbd *init_sbd(struct super_block *sb) |
75 | { | |
76 | struct gfs2_sbd *sdp; | |
b3b94faa | 77 | |
85d1da67 | 78 | sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL); |
b3b94faa DT |
79 | if (!sdp) |
80 | return NULL; | |
81 | ||
5c676f6d | 82 | sb->s_fs_info = sdp; |
b3b94faa DT |
83 | sdp->sd_vfs = sb; |
84 | ||
85 | gfs2_tune_init(&sdp->sd_tune); | |
86 | ||
f55ab26a | 87 | mutex_init(&sdp->sd_inum_mutex); |
b3b94faa | 88 | spin_lock_init(&sdp->sd_statfs_spin); |
b3b94faa DT |
89 | |
90 | spin_lock_init(&sdp->sd_rindex_spin); | |
f55ab26a | 91 | mutex_init(&sdp->sd_rindex_mutex); |
b3b94faa DT |
92 | INIT_LIST_HEAD(&sdp->sd_rindex_list); |
93 | INIT_LIST_HEAD(&sdp->sd_rindex_mru_list); | |
b3b94faa DT |
94 | |
95 | INIT_LIST_HEAD(&sdp->sd_jindex_list); | |
96 | spin_lock_init(&sdp->sd_jindex_spin); | |
f55ab26a | 97 | mutex_init(&sdp->sd_jindex_mutex); |
b3b94faa | 98 | |
b3b94faa | 99 | INIT_LIST_HEAD(&sdp->sd_quota_list); |
f55ab26a | 100 | mutex_init(&sdp->sd_quota_mutex); |
37b2c837 | 101 | init_waitqueue_head(&sdp->sd_quota_wait); |
813e0c46 SW |
102 | INIT_LIST_HEAD(&sdp->sd_trunc_list); |
103 | spin_lock_init(&sdp->sd_trunc_lock); | |
b3b94faa DT |
104 | |
105 | spin_lock_init(&sdp->sd_log_lock); | |
b3b94faa | 106 | |
b3b94faa DT |
107 | INIT_LIST_HEAD(&sdp->sd_log_le_buf); |
108 | INIT_LIST_HEAD(&sdp->sd_log_le_revoke); | |
109 | INIT_LIST_HEAD(&sdp->sd_log_le_rg); | |
110 | INIT_LIST_HEAD(&sdp->sd_log_le_databuf); | |
d7b616e2 | 111 | INIT_LIST_HEAD(&sdp->sd_log_le_ordered); |
b3b94faa | 112 | |
71b86f56 | 113 | mutex_init(&sdp->sd_log_reserve_mutex); |
b3b94faa DT |
114 | INIT_LIST_HEAD(&sdp->sd_ail1_list); |
115 | INIT_LIST_HEAD(&sdp->sd_ail2_list); | |
116 | ||
484adff8 | 117 | init_rwsem(&sdp->sd_log_flush_lock); |
16615be1 SW |
118 | atomic_set(&sdp->sd_log_in_flight, 0); |
119 | init_waitqueue_head(&sdp->sd_log_flush_wait); | |
b3b94faa DT |
120 | |
121 | INIT_LIST_HEAD(&sdp->sd_revoke_list); | |
122 | ||
f55ab26a | 123 | mutex_init(&sdp->sd_freeze_lock); |
b3b94faa DT |
124 | |
125 | return sdp; | |
126 | } | |
127 | ||
b3b94faa | 128 | |
9b8df98f SW |
129 | /** |
130 | * gfs2_check_sb - Check superblock | |
131 | * @sdp: the filesystem | |
132 | * @sb: The superblock | |
133 | * @silent: Don't print a message if the check fails | |
134 | * | |
135 | * Checks the version code of the FS is one that we understand how to | |
136 | * read and that the sizes of the various on-disk structures have not | |
137 | * changed. | |
138 | */ | |
139 | ||
140 | static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent) | |
141 | { | |
142 | unsigned int x; | |
143 | ||
144 | if (sb->sb_magic != GFS2_MAGIC || | |
145 | sb->sb_type != GFS2_METATYPE_SB) { | |
146 | if (!silent) | |
147 | printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n"); | |
148 | return -EINVAL; | |
149 | } | |
150 | ||
151 | /* If format numbers match exactly, we're done. */ | |
152 | ||
153 | if (sb->sb_fs_format == GFS2_FORMAT_FS && | |
154 | sb->sb_multihost_format == GFS2_FORMAT_MULTI) | |
155 | return 0; | |
156 | ||
157 | if (sb->sb_fs_format != GFS2_FORMAT_FS) { | |
158 | for (x = 0; gfs2_old_fs_formats[x]; x++) | |
159 | if (gfs2_old_fs_formats[x] == sb->sb_fs_format) | |
160 | break; | |
161 | ||
162 | if (!gfs2_old_fs_formats[x]) { | |
163 | printk(KERN_WARNING | |
164 | "GFS2: code version (%u, %u) is incompatible " | |
165 | "with ondisk format (%u, %u)\n", | |
166 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
167 | sb->sb_fs_format, sb->sb_multihost_format); | |
168 | printk(KERN_WARNING | |
169 | "GFS2: I don't know how to upgrade this FS\n"); | |
170 | return -EINVAL; | |
171 | } | |
172 | } | |
173 | ||
174 | if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) { | |
175 | for (x = 0; gfs2_old_multihost_formats[x]; x++) | |
176 | if (gfs2_old_multihost_formats[x] == | |
177 | sb->sb_multihost_format) | |
178 | break; | |
179 | ||
180 | if (!gfs2_old_multihost_formats[x]) { | |
181 | printk(KERN_WARNING | |
182 | "GFS2: code version (%u, %u) is incompatible " | |
183 | "with ondisk format (%u, %u)\n", | |
184 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
185 | sb->sb_fs_format, sb->sb_multihost_format); | |
186 | printk(KERN_WARNING | |
187 | "GFS2: I don't know how to upgrade this FS\n"); | |
188 | return -EINVAL; | |
189 | } | |
190 | } | |
191 | ||
192 | if (!sdp->sd_args.ar_upgrade) { | |
193 | printk(KERN_WARNING | |
194 | "GFS2: code version (%u, %u) is incompatible " | |
195 | "with ondisk format (%u, %u)\n", | |
196 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
197 | sb->sb_fs_format, sb->sb_multihost_format); | |
198 | printk(KERN_INFO | |
199 | "GFS2: Use the \"upgrade\" mount option to upgrade " | |
200 | "the FS\n"); | |
201 | printk(KERN_INFO "GFS2: See the manual for more details\n"); | |
202 | return -EINVAL; | |
203 | } | |
204 | ||
205 | return 0; | |
206 | } | |
207 | ||
208 | static void end_bio_io_page(struct bio *bio, int error) | |
209 | { | |
210 | struct page *page = bio->bi_private; | |
211 | ||
212 | if (!error) | |
213 | SetPageUptodate(page); | |
214 | else | |
215 | printk(KERN_WARNING "gfs2: error %d reading superblock\n", error); | |
216 | unlock_page(page); | |
217 | } | |
218 | ||
219 | static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf) | |
220 | { | |
221 | const struct gfs2_sb *str = buf; | |
222 | ||
223 | sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic); | |
224 | sb->sb_type = be32_to_cpu(str->sb_header.mh_type); | |
225 | sb->sb_format = be32_to_cpu(str->sb_header.mh_format); | |
226 | sb->sb_fs_format = be32_to_cpu(str->sb_fs_format); | |
227 | sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format); | |
228 | sb->sb_bsize = be32_to_cpu(str->sb_bsize); | |
229 | sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift); | |
230 | sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr); | |
231 | sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino); | |
232 | sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr); | |
233 | sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino); | |
234 | ||
235 | memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN); | |
236 | memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN); | |
02e3cc70 | 237 | memcpy(sb->sb_uuid, str->sb_uuid, 16); |
9b8df98f SW |
238 | } |
239 | ||
240 | /** | |
241 | * gfs2_read_super - Read the gfs2 super block from disk | |
242 | * @sdp: The GFS2 super block | |
243 | * @sector: The location of the super block | |
244 | * @error: The error code to return | |
245 | * | |
246 | * This uses the bio functions to read the super block from disk | |
247 | * because we want to be 100% sure that we never read cached data. | |
248 | * A super block is read twice only during each GFS2 mount and is | |
249 | * never written to by the filesystem. The first time its read no | |
250 | * locks are held, and the only details which are looked at are those | |
251 | * relating to the locking protocol. Once locking is up and working, | |
252 | * the sb is read again under the lock to establish the location of | |
253 | * the master directory (contains pointers to journals etc) and the | |
254 | * root directory. | |
255 | * | |
256 | * Returns: 0 on success or error | |
257 | */ | |
258 | ||
259 | static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector) | |
260 | { | |
261 | struct super_block *sb = sdp->sd_vfs; | |
262 | struct gfs2_sb *p; | |
263 | struct page *page; | |
264 | struct bio *bio; | |
265 | ||
266 | page = alloc_page(GFP_NOFS); | |
267 | if (unlikely(!page)) | |
268 | return -ENOBUFS; | |
269 | ||
270 | ClearPageUptodate(page); | |
271 | ClearPageDirty(page); | |
272 | lock_page(page); | |
273 | ||
274 | bio = bio_alloc(GFP_NOFS, 1); | |
9b8df98f SW |
275 | bio->bi_sector = sector * (sb->s_blocksize >> 9); |
276 | bio->bi_bdev = sb->s_bdev; | |
277 | bio_add_page(bio, page, PAGE_SIZE, 0); | |
278 | ||
279 | bio->bi_end_io = end_bio_io_page; | |
280 | bio->bi_private = page; | |
281 | submit_bio(READ_SYNC | (1 << BIO_RW_META), bio); | |
282 | wait_on_page_locked(page); | |
283 | bio_put(bio); | |
284 | if (!PageUptodate(page)) { | |
285 | __free_page(page); | |
286 | return -EIO; | |
287 | } | |
288 | p = kmap(page); | |
289 | gfs2_sb_in(&sdp->sd_sb, p); | |
290 | kunmap(page); | |
291 | __free_page(page); | |
292 | return 0; | |
293 | } | |
6bac243f | 294 | |
9b8df98f SW |
295 | /** |
296 | * gfs2_read_sb - Read super block | |
297 | * @sdp: The GFS2 superblock | |
9b8df98f SW |
298 | * @silent: Don't print message if mount fails |
299 | * | |
300 | */ | |
301 | ||
6bac243f | 302 | static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent) |
9b8df98f SW |
303 | { |
304 | u32 hash_blocks, ind_blocks, leaf_blocks; | |
305 | u32 tmp_blocks; | |
306 | unsigned int x; | |
307 | int error; | |
308 | ||
309 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); | |
310 | if (error) { | |
311 | if (!silent) | |
312 | fs_err(sdp, "can't read superblock\n"); | |
313 | return error; | |
314 | } | |
315 | ||
316 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
317 | if (error) | |
318 | return error; | |
319 | ||
320 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
321 | GFS2_BASIC_BLOCK_SHIFT; | |
322 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
323 | sdp->sd_diptrs = (sdp->sd_sb.sb_bsize - | |
324 | sizeof(struct gfs2_dinode)) / sizeof(u64); | |
325 | sdp->sd_inptrs = (sdp->sd_sb.sb_bsize - | |
326 | sizeof(struct gfs2_meta_header)) / sizeof(u64); | |
327 | sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header); | |
328 | sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2; | |
329 | sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1; | |
330 | sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64); | |
331 | sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize - | |
332 | sizeof(struct gfs2_meta_header)) / | |
333 | sizeof(struct gfs2_quota_change); | |
334 | ||
335 | /* Compute maximum reservation required to add a entry to a directory */ | |
336 | ||
337 | hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH), | |
338 | sdp->sd_jbsize); | |
339 | ||
340 | ind_blocks = 0; | |
341 | for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) { | |
342 | tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs); | |
343 | ind_blocks += tmp_blocks; | |
344 | } | |
345 | ||
346 | leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH; | |
347 | ||
348 | sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks; | |
349 | ||
350 | sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize - | |
351 | sizeof(struct gfs2_dinode); | |
352 | sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs; | |
353 | for (x = 2;; x++) { | |
354 | u64 space, d; | |
355 | u32 m; | |
356 | ||
357 | space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs; | |
358 | d = space; | |
359 | m = do_div(d, sdp->sd_inptrs); | |
360 | ||
361 | if (d != sdp->sd_heightsize[x - 1] || m) | |
362 | break; | |
363 | sdp->sd_heightsize[x] = space; | |
364 | } | |
365 | sdp->sd_max_height = x; | |
366 | sdp->sd_heightsize[x] = ~0; | |
367 | gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT); | |
368 | ||
369 | sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize - | |
370 | sizeof(struct gfs2_dinode); | |
371 | sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs; | |
372 | for (x = 2;; x++) { | |
373 | u64 space, d; | |
374 | u32 m; | |
375 | ||
376 | space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs; | |
377 | d = space; | |
378 | m = do_div(d, sdp->sd_inptrs); | |
379 | ||
380 | if (d != sdp->sd_jheightsize[x - 1] || m) | |
381 | break; | |
382 | sdp->sd_jheightsize[x] = space; | |
383 | } | |
384 | sdp->sd_max_jheight = x; | |
385 | sdp->sd_jheightsize[x] = ~0; | |
386 | gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT); | |
387 | ||
388 | return 0; | |
389 | } | |
390 | ||
b3b94faa DT |
391 | static int init_names(struct gfs2_sbd *sdp, int silent) |
392 | { | |
b3b94faa DT |
393 | char *proto, *table; |
394 | int error = 0; | |
395 | ||
396 | proto = sdp->sd_args.ar_lockproto; | |
397 | table = sdp->sd_args.ar_locktable; | |
398 | ||
399 | /* Try to autodetect */ | |
400 | ||
401 | if (!proto[0] || !table[0]) { | |
bb8d8a6f SW |
402 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); |
403 | if (error) | |
404 | return error; | |
3cf1e7be SW |
405 | |
406 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
b3b94faa DT |
407 | if (error) |
408 | goto out; | |
409 | ||
410 | if (!proto[0]) | |
3cf1e7be | 411 | proto = sdp->sd_sb.sb_lockproto; |
b3b94faa | 412 | if (!table[0]) |
3cf1e7be | 413 | table = sdp->sd_sb.sb_locktable; |
b3b94faa DT |
414 | } |
415 | ||
416 | if (!table[0]) | |
417 | table = sdp->sd_vfs->s_id; | |
418 | ||
00377d8e JD |
419 | strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN); |
420 | strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN); | |
b3b94faa | 421 | |
5d35e31f DC |
422 | table = sdp->sd_table_name; |
423 | while ((table = strchr(table, '/'))) | |
b35997d4 RP |
424 | *table = '_'; |
425 | ||
a91ea69f | 426 | out: |
b3b94faa DT |
427 | return error; |
428 | } | |
429 | ||
430 | static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh, | |
431 | int undo) | |
432 | { | |
b3b94faa DT |
433 | int error = 0; |
434 | ||
435 | if (undo) | |
436 | goto fail_trans; | |
437 | ||
b3b94faa DT |
438 | error = gfs2_glock_nq_num(sdp, |
439 | GFS2_MOUNT_LOCK, &gfs2_nondisk_glops, | |
440 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE, | |
441 | mount_gh); | |
442 | if (error) { | |
443 | fs_err(sdp, "can't acquire mount glock: %d\n", error); | |
444 | goto fail; | |
445 | } | |
446 | ||
447 | error = gfs2_glock_nq_num(sdp, | |
448 | GFS2_LIVE_LOCK, &gfs2_nondisk_glops, | |
449 | LM_ST_SHARED, | |
579b78a4 | 450 | LM_FLAG_NOEXP | GL_EXACT, |
b3b94faa DT |
451 | &sdp->sd_live_gh); |
452 | if (error) { | |
453 | fs_err(sdp, "can't acquire live glock: %d\n", error); | |
454 | goto fail_mount; | |
455 | } | |
456 | ||
457 | error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops, | |
458 | CREATE, &sdp->sd_rename_gl); | |
459 | if (error) { | |
460 | fs_err(sdp, "can't create rename glock: %d\n", error); | |
461 | goto fail_live; | |
462 | } | |
463 | ||
464 | error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops, | |
465 | CREATE, &sdp->sd_trans_gl); | |
466 | if (error) { | |
467 | fs_err(sdp, "can't create transaction glock: %d\n", error); | |
468 | goto fail_rename; | |
469 | } | |
b3b94faa DT |
470 | |
471 | return 0; | |
472 | ||
feaa7bba | 473 | fail_trans: |
b3b94faa | 474 | gfs2_glock_put(sdp->sd_trans_gl); |
feaa7bba | 475 | fail_rename: |
b3b94faa | 476 | gfs2_glock_put(sdp->sd_rename_gl); |
feaa7bba | 477 | fail_live: |
b3b94faa | 478 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); |
feaa7bba | 479 | fail_mount: |
b3b94faa | 480 | gfs2_glock_dq_uninit(mount_gh); |
feaa7bba | 481 | fail: |
b3b94faa DT |
482 | return error; |
483 | } | |
484 | ||
9b8df98f SW |
485 | static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr, |
486 | u64 no_addr, const char *name) | |
f42faf4f | 487 | { |
9b8df98f SW |
488 | struct gfs2_sbd *sdp = sb->s_fs_info; |
489 | struct dentry *dentry; | |
490 | struct inode *inode; | |
491 | ||
492 | inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0); | |
493 | if (IS_ERR(inode)) { | |
494 | fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode)); | |
495 | return PTR_ERR(inode); | |
496 | } | |
497 | dentry = d_alloc_root(inode); | |
498 | if (!dentry) { | |
499 | fs_err(sdp, "can't alloc %s dentry\n", name); | |
500 | iput(inode); | |
501 | return -ENOMEM; | |
502 | } | |
503 | dentry->d_op = &gfs2_dops; | |
504 | *dptr = dentry; | |
505 | return 0; | |
f42faf4f SW |
506 | } |
507 | ||
9b8df98f | 508 | static int init_sb(struct gfs2_sbd *sdp, int silent) |
b3b94faa DT |
509 | { |
510 | struct super_block *sb = sdp->sd_vfs; | |
511 | struct gfs2_holder sb_gh; | |
dbb7cae2 | 512 | u64 no_addr; |
9b8df98f | 513 | int ret; |
b3b94faa | 514 | |
9b8df98f SW |
515 | ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops, |
516 | LM_ST_SHARED, 0, &sb_gh); | |
517 | if (ret) { | |
518 | fs_err(sdp, "can't acquire superblock glock: %d\n", ret); | |
519 | return ret; | |
b3b94faa | 520 | } |
907b9bce | 521 | |
6bac243f | 522 | ret = gfs2_read_sb(sdp, silent); |
9b8df98f SW |
523 | if (ret) { |
524 | fs_err(sdp, "can't read superblock: %d\n", ret); | |
b3b94faa DT |
525 | goto out; |
526 | } | |
527 | ||
528 | /* Set up the buffer cache and SB for real */ | |
e1defc4f | 529 | if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) { |
9b8df98f | 530 | ret = -EINVAL; |
b3b94faa DT |
531 | fs_err(sdp, "FS block size (%u) is too small for device " |
532 | "block size (%u)\n", | |
e1defc4f | 533 | sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev)); |
b3b94faa DT |
534 | goto out; |
535 | } | |
536 | if (sdp->sd_sb.sb_bsize > PAGE_SIZE) { | |
9b8df98f | 537 | ret = -EINVAL; |
b3b94faa DT |
538 | fs_err(sdp, "FS block size (%u) is too big for machine " |
539 | "page size (%u)\n", | |
540 | sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE); | |
541 | goto out; | |
542 | } | |
b3b94faa DT |
543 | sb_set_blocksize(sb, sdp->sd_sb.sb_bsize); |
544 | ||
f42faf4f | 545 | /* Get the root inode */ |
dbb7cae2 | 546 | no_addr = sdp->sd_sb.sb_root_dir.no_addr; |
9b8df98f SW |
547 | ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root"); |
548 | if (ret) | |
f42faf4f | 549 | goto out; |
b3b94faa | 550 | |
9b8df98f SW |
551 | /* Get the master inode */ |
552 | no_addr = sdp->sd_sb.sb_master_dir.no_addr; | |
553 | ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master"); | |
554 | if (ret) { | |
555 | dput(sdp->sd_root_dir); | |
556 | goto out; | |
557 | } | |
558 | sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir); | |
f42faf4f | 559 | out: |
b3b94faa | 560 | gfs2_glock_dq_uninit(&sb_gh); |
9b8df98f | 561 | return ret; |
b3b94faa DT |
562 | } |
563 | ||
da6dd40d BP |
564 | /** |
565 | * map_journal_extents - create a reusable "extent" mapping from all logical | |
566 | * blocks to all physical blocks for the given journal. This will save | |
567 | * us time when writing journal blocks. Most journals will have only one | |
568 | * extent that maps all their logical blocks. That's because gfs2.mkfs | |
569 | * arranges the journal blocks sequentially to maximize performance. | |
570 | * So the extent would map the first block for the entire file length. | |
571 | * However, gfs2_jadd can happen while file activity is happening, so | |
572 | * those journals may not be sequential. Less likely is the case where | |
573 | * the users created their own journals by mounting the metafs and | |
574 | * laying it out. But it's still possible. These journals might have | |
575 | * several extents. | |
576 | * | |
577 | * TODO: This should be done in bigger chunks rather than one block at a time, | |
578 | * but since it's only done at mount time, I'm not worried about the | |
579 | * time it takes. | |
580 | */ | |
581 | static int map_journal_extents(struct gfs2_sbd *sdp) | |
582 | { | |
583 | struct gfs2_jdesc *jd = sdp->sd_jdesc; | |
584 | unsigned int lb; | |
585 | u64 db, prev_db; /* logical block, disk block, prev disk block */ | |
586 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); | |
587 | struct gfs2_journal_extent *jext = NULL; | |
588 | struct buffer_head bh; | |
589 | int rc = 0; | |
590 | ||
da6dd40d BP |
591 | prev_db = 0; |
592 | ||
c9e98886 | 593 | for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) { |
da6dd40d BP |
594 | bh.b_state = 0; |
595 | bh.b_blocknr = 0; | |
596 | bh.b_size = 1 << ip->i_inode.i_blkbits; | |
597 | rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0); | |
598 | db = bh.b_blocknr; | |
599 | if (rc || !db) { | |
600 | printk(KERN_INFO "GFS2 journal mapping error %d: lb=" | |
601 | "%u db=%llu\n", rc, lb, (unsigned long long)db); | |
602 | break; | |
603 | } | |
604 | if (!prev_db || db != prev_db + 1) { | |
605 | jext = kzalloc(sizeof(struct gfs2_journal_extent), | |
606 | GFP_KERNEL); | |
607 | if (!jext) { | |
608 | printk(KERN_INFO "GFS2 error: out of memory " | |
609 | "mapping journal extents.\n"); | |
610 | rc = -ENOMEM; | |
611 | break; | |
612 | } | |
613 | jext->dblock = db; | |
614 | jext->lblock = lb; | |
615 | jext->blocks = 1; | |
616 | list_add_tail(&jext->extent_list, &jd->extent_list); | |
617 | } else { | |
618 | jext->blocks++; | |
619 | } | |
620 | prev_db = db; | |
621 | } | |
622 | return rc; | |
623 | } | |
624 | ||
f057f6cd | 625 | static void gfs2_others_may_mount(struct gfs2_sbd *sdp) |
da755fdb | 626 | { |
f057f6cd SW |
627 | char *message = "FIRSTMOUNT=Done"; |
628 | char *envp[] = { message, NULL }; | |
629 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
630 | ls->ls_first_done = 1; | |
631 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp); | |
da755fdb SW |
632 | } |
633 | ||
b5289681 SW |
634 | /** |
635 | * gfs2_jindex_hold - Grab a lock on the jindex | |
636 | * @sdp: The GFS2 superblock | |
637 | * @ji_gh: the holder for the jindex glock | |
638 | * | |
639 | * Returns: errno | |
640 | */ | |
641 | ||
642 | static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh) | |
643 | { | |
644 | struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex); | |
645 | struct qstr name; | |
646 | char buf[20]; | |
647 | struct gfs2_jdesc *jd; | |
648 | int error; | |
649 | ||
650 | name.name = buf; | |
651 | ||
652 | mutex_lock(&sdp->sd_jindex_mutex); | |
653 | ||
654 | for (;;) { | |
655 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh); | |
656 | if (error) | |
657 | break; | |
658 | ||
659 | name.len = sprintf(buf, "journal%u", sdp->sd_journals); | |
660 | name.hash = gfs2_disk_hash(name.name, name.len); | |
661 | ||
662 | error = gfs2_dir_check(sdp->sd_jindex, &name, NULL); | |
663 | if (error == -ENOENT) { | |
664 | error = 0; | |
665 | break; | |
666 | } | |
667 | ||
668 | gfs2_glock_dq_uninit(ji_gh); | |
669 | ||
670 | if (error) | |
671 | break; | |
672 | ||
673 | error = -ENOMEM; | |
674 | jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL); | |
675 | if (!jd) | |
676 | break; | |
677 | ||
678 | INIT_LIST_HEAD(&jd->extent_list); | |
fe64d517 | 679 | slow_work_init(&jd->jd_work, &gfs2_recover_ops); |
b5289681 SW |
680 | jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1); |
681 | if (!jd->jd_inode || IS_ERR(jd->jd_inode)) { | |
682 | if (!jd->jd_inode) | |
683 | error = -ENOENT; | |
684 | else | |
685 | error = PTR_ERR(jd->jd_inode); | |
686 | kfree(jd); | |
687 | break; | |
688 | } | |
689 | ||
690 | spin_lock(&sdp->sd_jindex_spin); | |
691 | jd->jd_jid = sdp->sd_journals++; | |
692 | list_add_tail(&jd->jd_list, &sdp->sd_jindex_list); | |
693 | spin_unlock(&sdp->sd_jindex_spin); | |
694 | } | |
695 | ||
696 | mutex_unlock(&sdp->sd_jindex_mutex); | |
697 | ||
698 | return error; | |
699 | } | |
700 | ||
b3b94faa DT |
701 | static int init_journal(struct gfs2_sbd *sdp, int undo) |
702 | { | |
9b8df98f | 703 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa | 704 | struct gfs2_holder ji_gh; |
5c676f6d | 705 | struct gfs2_inode *ip; |
b3b94faa DT |
706 | int jindex = 1; |
707 | int error = 0; | |
708 | ||
709 | if (undo) { | |
710 | jindex = 0; | |
fe64d517 | 711 | goto fail_jinode_gh; |
b3b94faa DT |
712 | } |
713 | ||
9b8df98f | 714 | sdp->sd_jindex = gfs2_lookup_simple(master, "jindex"); |
c752666c | 715 | if (IS_ERR(sdp->sd_jindex)) { |
b3b94faa | 716 | fs_err(sdp, "can't lookup journal index: %d\n", error); |
c752666c | 717 | return PTR_ERR(sdp->sd_jindex); |
b3b94faa | 718 | } |
feaa7bba | 719 | ip = GFS2_I(sdp->sd_jindex); |
b3b94faa DT |
720 | |
721 | /* Load in the journal index special file */ | |
722 | ||
723 | error = gfs2_jindex_hold(sdp, &ji_gh); | |
724 | if (error) { | |
725 | fs_err(sdp, "can't read journal index: %d\n", error); | |
726 | goto fail; | |
727 | } | |
728 | ||
729 | error = -EINVAL; | |
730 | if (!gfs2_jindex_size(sdp)) { | |
731 | fs_err(sdp, "no journals!\n"); | |
907b9bce | 732 | goto fail_jindex; |
b3b94faa DT |
733 | } |
734 | ||
735 | if (sdp->sd_args.ar_spectator) { | |
736 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0); | |
fd041f0b | 737 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
b3b94faa DT |
738 | } else { |
739 | if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) { | |
740 | fs_err(sdp, "can't mount journal #%u\n", | |
741 | sdp->sd_lockstruct.ls_jid); | |
742 | fs_err(sdp, "there are only %u journals (0 - %u)\n", | |
743 | gfs2_jindex_size(sdp), | |
744 | gfs2_jindex_size(sdp) - 1); | |
745 | goto fail_jindex; | |
746 | } | |
747 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid); | |
748 | ||
feaa7bba | 749 | error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid, |
b3b94faa DT |
750 | &gfs2_journal_glops, |
751 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP, | |
752 | &sdp->sd_journal_gh); | |
753 | if (error) { | |
754 | fs_err(sdp, "can't acquire journal glock: %d\n", error); | |
755 | goto fail_jindex; | |
756 | } | |
757 | ||
feaa7bba SW |
758 | ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
759 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, | |
75be73a8 | 760 | LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE, |
b3b94faa DT |
761 | &sdp->sd_jinode_gh); |
762 | if (error) { | |
763 | fs_err(sdp, "can't acquire journal inode glock: %d\n", | |
764 | error); | |
765 | goto fail_journal_gh; | |
766 | } | |
767 | ||
768 | error = gfs2_jdesc_check(sdp->sd_jdesc); | |
769 | if (error) { | |
770 | fs_err(sdp, "my journal (%u) is bad: %d\n", | |
771 | sdp->sd_jdesc->jd_jid, error); | |
772 | goto fail_jinode_gh; | |
773 | } | |
fd041f0b | 774 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
da6dd40d BP |
775 | |
776 | /* Map the extents for this journal's blocks */ | |
777 | map_journal_extents(sdp); | |
b3b94faa | 778 | } |
63997775 | 779 | trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free)); |
b3b94faa DT |
780 | |
781 | if (sdp->sd_lockstruct.ls_first) { | |
782 | unsigned int x; | |
783 | for (x = 0; x < sdp->sd_journals; x++) { | |
c63e31c2 | 784 | error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x)); |
b3b94faa DT |
785 | if (error) { |
786 | fs_err(sdp, "error recovering journal %u: %d\n", | |
787 | x, error); | |
788 | goto fail_jinode_gh; | |
789 | } | |
790 | } | |
791 | ||
f057f6cd | 792 | gfs2_others_may_mount(sdp); |
b3b94faa | 793 | } else if (!sdp->sd_args.ar_spectator) { |
c63e31c2 | 794 | error = gfs2_recover_journal(sdp->sd_jdesc); |
b3b94faa DT |
795 | if (error) { |
796 | fs_err(sdp, "error recovering my journal: %d\n", error); | |
797 | goto fail_jinode_gh; | |
798 | } | |
799 | } | |
800 | ||
801 | set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags); | |
802 | gfs2_glock_dq_uninit(&ji_gh); | |
803 | jindex = 0; | |
804 | ||
b3b94faa DT |
805 | return 0; |
806 | ||
a91ea69f | 807 | fail_jinode_gh: |
b3b94faa DT |
808 | if (!sdp->sd_args.ar_spectator) |
809 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
a91ea69f | 810 | fail_journal_gh: |
b3b94faa DT |
811 | if (!sdp->sd_args.ar_spectator) |
812 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
a91ea69f | 813 | fail_jindex: |
b3b94faa DT |
814 | gfs2_jindex_free(sdp); |
815 | if (jindex) | |
816 | gfs2_glock_dq_uninit(&ji_gh); | |
a91ea69f | 817 | fail: |
f42faf4f | 818 | iput(sdp->sd_jindex); |
b3b94faa DT |
819 | return error; |
820 | } | |
821 | ||
b3b94faa DT |
822 | |
823 | static int init_inodes(struct gfs2_sbd *sdp, int undo) | |
824 | { | |
b3b94faa | 825 | int error = 0; |
5c676f6d | 826 | struct gfs2_inode *ip; |
9b8df98f | 827 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
828 | |
829 | if (undo) | |
f42faf4f SW |
830 | goto fail_qinode; |
831 | ||
f42faf4f SW |
832 | error = init_journal(sdp, undo); |
833 | if (error) | |
9b8df98f | 834 | goto fail; |
b3b94faa DT |
835 | |
836 | /* Read in the master inode number inode */ | |
9b8df98f | 837 | sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum"); |
c752666c SW |
838 | if (IS_ERR(sdp->sd_inum_inode)) { |
839 | error = PTR_ERR(sdp->sd_inum_inode); | |
b3b94faa | 840 | fs_err(sdp, "can't read in inum inode: %d\n", error); |
f42faf4f | 841 | goto fail_journal; |
b3b94faa DT |
842 | } |
843 | ||
f42faf4f | 844 | |
b3b94faa | 845 | /* Read in the master statfs inode */ |
9b8df98f | 846 | sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs"); |
c752666c SW |
847 | if (IS_ERR(sdp->sd_statfs_inode)) { |
848 | error = PTR_ERR(sdp->sd_statfs_inode); | |
b3b94faa | 849 | fs_err(sdp, "can't read in statfs inode: %d\n", error); |
f42faf4f | 850 | goto fail_inum; |
b3b94faa DT |
851 | } |
852 | ||
853 | /* Read in the resource index inode */ | |
9b8df98f | 854 | sdp->sd_rindex = gfs2_lookup_simple(master, "rindex"); |
c752666c SW |
855 | if (IS_ERR(sdp->sd_rindex)) { |
856 | error = PTR_ERR(sdp->sd_rindex); | |
b3b94faa DT |
857 | fs_err(sdp, "can't get resource index inode: %d\n", error); |
858 | goto fail_statfs; | |
859 | } | |
feaa7bba | 860 | ip = GFS2_I(sdp->sd_rindex); |
cf45b752 | 861 | sdp->sd_rindex_uptodate = 0; |
b3b94faa DT |
862 | |
863 | /* Read in the quota inode */ | |
9b8df98f | 864 | sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota"); |
c752666c SW |
865 | if (IS_ERR(sdp->sd_quota_inode)) { |
866 | error = PTR_ERR(sdp->sd_quota_inode); | |
b3b94faa DT |
867 | fs_err(sdp, "can't get quota file inode: %d\n", error); |
868 | goto fail_rindex; | |
869 | } | |
b3b94faa DT |
870 | return 0; |
871 | ||
f42faf4f SW |
872 | fail_qinode: |
873 | iput(sdp->sd_quota_inode); | |
f42faf4f | 874 | fail_rindex: |
b3b94faa | 875 | gfs2_clear_rgrpd(sdp); |
f42faf4f | 876 | iput(sdp->sd_rindex); |
f42faf4f SW |
877 | fail_statfs: |
878 | iput(sdp->sd_statfs_inode); | |
f42faf4f SW |
879 | fail_inum: |
880 | iput(sdp->sd_inum_inode); | |
881 | fail_journal: | |
882 | init_journal(sdp, UNDO); | |
f42faf4f | 883 | fail: |
b3b94faa DT |
884 | return error; |
885 | } | |
886 | ||
887 | static int init_per_node(struct gfs2_sbd *sdp, int undo) | |
888 | { | |
f42faf4f | 889 | struct inode *pn = NULL; |
b3b94faa DT |
890 | char buf[30]; |
891 | int error = 0; | |
5c676f6d | 892 | struct gfs2_inode *ip; |
9b8df98f | 893 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
894 | |
895 | if (sdp->sd_args.ar_spectator) | |
896 | return 0; | |
897 | ||
898 | if (undo) | |
899 | goto fail_qc_gh; | |
900 | ||
9b8df98f | 901 | pn = gfs2_lookup_simple(master, "per_node"); |
c752666c SW |
902 | if (IS_ERR(pn)) { |
903 | error = PTR_ERR(pn); | |
b3b94faa DT |
904 | fs_err(sdp, "can't find per_node directory: %d\n", error); |
905 | return error; | |
906 | } | |
907 | ||
908 | sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid); | |
c752666c SW |
909 | sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf); |
910 | if (IS_ERR(sdp->sd_ir_inode)) { | |
911 | error = PTR_ERR(sdp->sd_ir_inode); | |
b3b94faa DT |
912 | fs_err(sdp, "can't find local \"ir\" file: %d\n", error); |
913 | goto fail; | |
914 | } | |
915 | ||
916 | sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid); | |
c752666c SW |
917 | sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf); |
918 | if (IS_ERR(sdp->sd_sc_inode)) { | |
919 | error = PTR_ERR(sdp->sd_sc_inode); | |
b3b94faa DT |
920 | fs_err(sdp, "can't find local \"sc\" file: %d\n", error); |
921 | goto fail_ir_i; | |
922 | } | |
923 | ||
b3b94faa | 924 | sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
925 | sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf); |
926 | if (IS_ERR(sdp->sd_qc_inode)) { | |
927 | error = PTR_ERR(sdp->sd_qc_inode); | |
b3b94faa DT |
928 | fs_err(sdp, "can't find local \"qc\" file: %d\n", error); |
929 | goto fail_ut_i; | |
930 | } | |
931 | ||
f42faf4f | 932 | iput(pn); |
b3b94faa DT |
933 | pn = NULL; |
934 | ||
feaa7bba | 935 | ip = GFS2_I(sdp->sd_ir_inode); |
5c676f6d | 936 | error = gfs2_glock_nq_init(ip->i_gl, |
579b78a4 | 937 | LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
938 | &sdp->sd_ir_gh); |
939 | if (error) { | |
940 | fs_err(sdp, "can't lock local \"ir\" file: %d\n", error); | |
941 | goto fail_qc_i; | |
942 | } | |
943 | ||
feaa7bba | 944 | ip = GFS2_I(sdp->sd_sc_inode); |
5c676f6d | 945 | error = gfs2_glock_nq_init(ip->i_gl, |
579b78a4 | 946 | LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
947 | &sdp->sd_sc_gh); |
948 | if (error) { | |
949 | fs_err(sdp, "can't lock local \"sc\" file: %d\n", error); | |
950 | goto fail_ir_gh; | |
951 | } | |
952 | ||
feaa7bba | 953 | ip = GFS2_I(sdp->sd_qc_inode); |
5c676f6d | 954 | error = gfs2_glock_nq_init(ip->i_gl, |
579b78a4 | 955 | LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
956 | &sdp->sd_qc_gh); |
957 | if (error) { | |
958 | fs_err(sdp, "can't lock local \"qc\" file: %d\n", error); | |
959 | goto fail_ut_gh; | |
960 | } | |
961 | ||
962 | return 0; | |
963 | ||
a91ea69f | 964 | fail_qc_gh: |
b3b94faa | 965 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); |
a91ea69f | 966 | fail_ut_gh: |
b3b94faa | 967 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); |
a91ea69f | 968 | fail_ir_gh: |
b3b94faa | 969 | gfs2_glock_dq_uninit(&sdp->sd_ir_gh); |
a91ea69f | 970 | fail_qc_i: |
f42faf4f | 971 | iput(sdp->sd_qc_inode); |
a91ea69f | 972 | fail_ut_i: |
f42faf4f | 973 | iput(sdp->sd_sc_inode); |
a91ea69f | 974 | fail_ir_i: |
f42faf4f | 975 | iput(sdp->sd_ir_inode); |
a91ea69f | 976 | fail: |
b3b94faa | 977 | if (pn) |
f42faf4f | 978 | iput(pn); |
b3b94faa DT |
979 | return error; |
980 | } | |
981 | ||
982 | static int init_threads(struct gfs2_sbd *sdp, int undo) | |
983 | { | |
984 | struct task_struct *p; | |
985 | int error = 0; | |
986 | ||
987 | if (undo) | |
feaa7bba | 988 | goto fail_quotad; |
b3b94faa DT |
989 | |
990 | sdp->sd_log_flush_time = jiffies; | |
b3b94faa DT |
991 | |
992 | p = kthread_run(gfs2_logd, sdp, "gfs2_logd"); | |
993 | error = IS_ERR(p); | |
994 | if (error) { | |
995 | fs_err(sdp, "can't start logd thread: %d\n", error); | |
996 | return error; | |
997 | } | |
998 | sdp->sd_logd_process = p; | |
999 | ||
b3b94faa DT |
1000 | p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad"); |
1001 | error = IS_ERR(p); | |
1002 | if (error) { | |
1003 | fs_err(sdp, "can't start quotad thread: %d\n", error); | |
1004 | goto fail; | |
1005 | } | |
1006 | sdp->sd_quotad_process = p; | |
1007 | ||
b3b94faa DT |
1008 | return 0; |
1009 | ||
b3b94faa | 1010 | |
feaa7bba | 1011 | fail_quotad: |
b3b94faa | 1012 | kthread_stop(sdp->sd_quotad_process); |
feaa7bba | 1013 | fail: |
b3b94faa | 1014 | kthread_stop(sdp->sd_logd_process); |
b3b94faa DT |
1015 | return error; |
1016 | } | |
da755fdb | 1017 | |
f057f6cd SW |
1018 | static const match_table_t nolock_tokens = { |
1019 | { Opt_jid, "jid=%d\n", }, | |
1020 | { Opt_err, NULL }, | |
1021 | }; | |
1022 | ||
1023 | static const struct lm_lockops nolock_ops = { | |
1024 | .lm_proto_name = "lock_nolock", | |
1025 | .lm_put_lock = kmem_cache_free, | |
1026 | .lm_tokens = &nolock_tokens, | |
1027 | }; | |
1028 | ||
da755fdb SW |
1029 | /** |
1030 | * gfs2_lm_mount - mount a locking protocol | |
1031 | * @sdp: the filesystem | |
1032 | * @args: mount arguements | |
1033 | * @silent: if 1, don't complain if the FS isn't a GFS2 fs | |
1034 | * | |
1035 | * Returns: errno | |
1036 | */ | |
1037 | ||
1038 | static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent) | |
1039 | { | |
f057f6cd SW |
1040 | const struct lm_lockops *lm; |
1041 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
1042 | struct gfs2_args *args = &sdp->sd_args; | |
1043 | const char *proto = sdp->sd_proto_name; | |
1044 | const char *table = sdp->sd_table_name; | |
1045 | const char *fsname; | |
1046 | char *o, *options; | |
1047 | int ret; | |
da755fdb | 1048 | |
f057f6cd SW |
1049 | if (!strcmp("lock_nolock", proto)) { |
1050 | lm = &nolock_ops; | |
1051 | sdp->sd_args.ar_localflocks = 1; | |
1052 | sdp->sd_args.ar_localcaching = 1; | |
1053 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM | |
1054 | } else if (!strcmp("lock_dlm", proto)) { | |
1055 | lm = &gfs2_dlm_ops; | |
1056 | #endif | |
1057 | } else { | |
1058 | printk(KERN_INFO "GFS2: can't find protocol %s\n", proto); | |
1059 | return -ENOENT; | |
1060 | } | |
da755fdb SW |
1061 | |
1062 | fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table); | |
1063 | ||
f057f6cd SW |
1064 | ls->ls_ops = lm; |
1065 | ls->ls_first = 1; | |
1066 | ls->ls_id = 0; | |
da755fdb | 1067 | |
f057f6cd SW |
1068 | for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) { |
1069 | substring_t tmp[MAX_OPT_ARGS]; | |
1070 | int token, option; | |
1071 | ||
1072 | if (!o || !*o) | |
1073 | continue; | |
1074 | ||
1075 | token = match_token(o, *lm->lm_tokens, tmp); | |
1076 | switch (token) { | |
1077 | case Opt_jid: | |
1078 | ret = match_int(&tmp[0], &option); | |
1079 | if (ret || option < 0) | |
1080 | goto hostdata_error; | |
1081 | ls->ls_jid = option; | |
1082 | break; | |
1083 | case Opt_id: | |
1084 | ret = match_int(&tmp[0], &option); | |
1085 | if (ret) | |
1086 | goto hostdata_error; | |
1087 | ls->ls_id = option; | |
1088 | break; | |
1089 | case Opt_first: | |
1090 | ret = match_int(&tmp[0], &option); | |
1091 | if (ret || (option != 0 && option != 1)) | |
1092 | goto hostdata_error; | |
1093 | ls->ls_first = option; | |
1094 | break; | |
1095 | case Opt_nodir: | |
1096 | ret = match_int(&tmp[0], &option); | |
1097 | if (ret || (option != 0 && option != 1)) | |
1098 | goto hostdata_error; | |
1099 | ls->ls_nodir = option; | |
1100 | break; | |
1101 | case Opt_err: | |
1102 | default: | |
1103 | hostdata_error: | |
1104 | fs_info(sdp, "unknown hostdata (%s)\n", o); | |
1105 | return -EINVAL; | |
1106 | } | |
da755fdb SW |
1107 | } |
1108 | ||
1109 | if (sdp->sd_args.ar_spectator) | |
1110 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table); | |
1111 | else | |
1112 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table, | |
1113 | sdp->sd_lockstruct.ls_jid); | |
1114 | ||
f057f6cd SW |
1115 | fsname = strchr(table, ':'); |
1116 | if (fsname) | |
1117 | fsname++; | |
1118 | if (lm->lm_mount == NULL) { | |
1119 | fs_info(sdp, "Now mounting FS...\n"); | |
1120 | return 0; | |
da755fdb | 1121 | } |
f057f6cd SW |
1122 | ret = lm->lm_mount(sdp, fsname); |
1123 | if (ret == 0) | |
1124 | fs_info(sdp, "Joined cluster. Now mounting FS...\n"); | |
1125 | return ret; | |
da755fdb SW |
1126 | } |
1127 | ||
1128 | void gfs2_lm_unmount(struct gfs2_sbd *sdp) | |
1129 | { | |
f057f6cd SW |
1130 | const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops; |
1131 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) && | |
1132 | lm->lm_unmount) | |
1133 | lm->lm_unmount(sdp); | |
da755fdb | 1134 | } |
b3b94faa DT |
1135 | |
1136 | /** | |
1137 | * fill_super - Read in superblock | |
1138 | * @sb: The VFS superblock | |
1139 | * @data: Mount options | |
1140 | * @silent: Don't complain if it's not a GFS2 filesystem | |
1141 | * | |
1142 | * Returns: errno | |
1143 | */ | |
1144 | ||
1145 | static int fill_super(struct super_block *sb, void *data, int silent) | |
1146 | { | |
1147 | struct gfs2_sbd *sdp; | |
1148 | struct gfs2_holder mount_gh; | |
1149 | int error; | |
1150 | ||
1151 | sdp = init_sbd(sb); | |
1152 | if (!sdp) { | |
d92a8d48 | 1153 | printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n"); |
b3b94faa DT |
1154 | return -ENOMEM; |
1155 | } | |
1156 | ||
6f04c1c7 SW |
1157 | sdp->sd_args.ar_quota = GFS2_QUOTA_DEFAULT; |
1158 | sdp->sd_args.ar_data = GFS2_DATA_DEFAULT; | |
48c2b613 | 1159 | sdp->sd_args.ar_commit = 60; |
6f04c1c7 SW |
1160 | |
1161 | error = gfs2_mount_args(sdp, &sdp->sd_args, data); | |
b3b94faa | 1162 | if (error) { |
d92a8d48 | 1163 | printk(KERN_WARNING "GFS2: can't parse mount arguments\n"); |
b3b94faa DT |
1164 | goto fail; |
1165 | } | |
1166 | ||
fe64d517 | 1167 | if (sdp->sd_args.ar_spectator) { |
6f04c1c7 | 1168 | sb->s_flags |= MS_RDONLY; |
fe64d517 SW |
1169 | set_bit(SDF_NORECOVERY, &sdp->sd_flags); |
1170 | } | |
6f04c1c7 SW |
1171 | if (sdp->sd_args.ar_posix_acl) |
1172 | sb->s_flags |= MS_POSIXACL; | |
1173 | ||
719ee344 SW |
1174 | sb->s_magic = GFS2_MAGIC; |
1175 | sb->s_op = &gfs2_super_ops; | |
1176 | sb->s_export_op = &gfs2_export_ops; | |
1177 | sb->s_time_gran = 1; | |
1178 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
419c93e0 SW |
1179 | |
1180 | /* Set up the buffer cache and fill in some fake block size values | |
1181 | to allow us to read-in the on-disk superblock. */ | |
1182 | sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK); | |
1183 | sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits; | |
1184 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
1185 | GFS2_BASIC_BLOCK_SHIFT; | |
1186 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
b3b94faa | 1187 | |
48c2b613 SW |
1188 | sdp->sd_tune.gt_log_flush_secs = sdp->sd_args.ar_commit; |
1189 | ||
b3b94faa DT |
1190 | error = init_names(sdp, silent); |
1191 | if (error) | |
1192 | goto fail; | |
1193 | ||
7c52b166 RP |
1194 | gfs2_create_debugfs_file(sdp); |
1195 | ||
b3b94faa DT |
1196 | error = gfs2_sys_fs_add(sdp); |
1197 | if (error) | |
1198 | goto fail; | |
1199 | ||
1200 | error = gfs2_lm_mount(sdp, silent); | |
1201 | if (error) | |
1202 | goto fail_sys; | |
1203 | ||
1204 | error = init_locking(sdp, &mount_gh, DO); | |
1205 | if (error) | |
1206 | goto fail_lm; | |
1207 | ||
9b8df98f | 1208 | error = init_sb(sdp, silent); |
b3b94faa DT |
1209 | if (error) |
1210 | goto fail_locking; | |
b3b94faa DT |
1211 | |
1212 | error = init_inodes(sdp, DO); | |
1213 | if (error) | |
f42faf4f | 1214 | goto fail_sb; |
b3b94faa DT |
1215 | |
1216 | error = init_per_node(sdp, DO); | |
1217 | if (error) | |
1218 | goto fail_inodes; | |
1219 | ||
1220 | error = gfs2_statfs_init(sdp); | |
1221 | if (error) { | |
1222 | fs_err(sdp, "can't initialize statfs subsystem: %d\n", error); | |
1223 | goto fail_per_node; | |
1224 | } | |
1225 | ||
1226 | error = init_threads(sdp, DO); | |
1227 | if (error) | |
1228 | goto fail_per_node; | |
1229 | ||
1230 | if (!(sb->s_flags & MS_RDONLY)) { | |
1231 | error = gfs2_make_fs_rw(sdp); | |
1232 | if (error) { | |
1233 | fs_err(sdp, "can't make FS RW: %d\n", error); | |
1234 | goto fail_threads; | |
1235 | } | |
1236 | } | |
1237 | ||
1238 | gfs2_glock_dq_uninit(&mount_gh); | |
1239 | ||
1240 | return 0; | |
1241 | ||
a91ea69f | 1242 | fail_threads: |
b3b94faa | 1243 | init_threads(sdp, UNDO); |
a91ea69f | 1244 | fail_per_node: |
b3b94faa | 1245 | init_per_node(sdp, UNDO); |
a91ea69f | 1246 | fail_inodes: |
b3b94faa | 1247 | init_inodes(sdp, UNDO); |
a91ea69f | 1248 | fail_sb: |
9b8df98f SW |
1249 | if (sdp->sd_root_dir) |
1250 | dput(sdp->sd_root_dir); | |
1251 | if (sdp->sd_master_dir) | |
1252 | dput(sdp->sd_master_dir); | |
e7c8707e SW |
1253 | if (sb->s_root) |
1254 | dput(sb->s_root); | |
9b8df98f | 1255 | sb->s_root = NULL; |
a91ea69f | 1256 | fail_locking: |
b3b94faa | 1257 | init_locking(sdp, &mount_gh, UNDO); |
a91ea69f | 1258 | fail_lm: |
fefc03bf | 1259 | gfs2_gl_hash_clear(sdp); |
b3b94faa DT |
1260 | gfs2_lm_unmount(sdp); |
1261 | while (invalidate_inodes(sb)) | |
1262 | yield(); | |
a91ea69f | 1263 | fail_sys: |
b3b94faa | 1264 | gfs2_sys_fs_del(sdp); |
a91ea69f | 1265 | fail: |
5f882096 | 1266 | gfs2_delete_debugfs_file(sdp); |
a2c45807 | 1267 | kfree(sdp); |
5c676f6d | 1268 | sb->s_fs_info = NULL; |
b3b94faa DT |
1269 | return error; |
1270 | } | |
1271 | ||
ccd6efd0 | 1272 | static int gfs2_get_sb(struct file_system_type *fs_type, int flags, |
9b8df98f | 1273 | const char *dev_name, void *data, struct vfsmount *mnt) |
b3b94faa | 1274 | { |
9b8df98f | 1275 | return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt); |
86384605 AD |
1276 | } |
1277 | ||
f6d03139 SW |
1278 | static int test_meta_super(struct super_block *s, void *ptr) |
1279 | { | |
1280 | struct block_device *bdev = ptr; | |
1281 | return (bdev == s->s_bdev); | |
1282 | } | |
1283 | ||
1284 | static int set_meta_super(struct super_block *s, void *ptr) | |
1285 | { | |
1286 | return -EINVAL; | |
1287 | } | |
1288 | ||
003dec89 SW |
1289 | static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags, |
1290 | const char *dev_name, void *data, struct vfsmount *mnt) | |
86384605 | 1291 | { |
f6d03139 | 1292 | struct super_block *s; |
003dec89 | 1293 | struct gfs2_sbd *sdp; |
e24977d4 | 1294 | struct path path; |
86384605 | 1295 | int error; |
907b9bce | 1296 | |
e24977d4 | 1297 | error = kern_path(dev_name, LOOKUP_FOLLOW, &path); |
86384605 | 1298 | if (error) { |
9b8df98f SW |
1299 | printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n", |
1300 | dev_name, error); | |
003dec89 | 1301 | return error; |
86384605 | 1302 | } |
f6d03139 SW |
1303 | s = sget(&gfs2_fs_type, test_meta_super, set_meta_super, |
1304 | path.dentry->d_inode->i_sb->s_bdev); | |
e24977d4 | 1305 | path_put(&path); |
f6d03139 | 1306 | if (IS_ERR(s)) { |
86384605 | 1307 | printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n"); |
f6d03139 | 1308 | return PTR_ERR(s); |
86384605 | 1309 | } |
f6d03139 SW |
1310 | sdp = s->s_fs_info; |
1311 | mnt->mnt_sb = s; | |
9b8df98f SW |
1312 | mnt->mnt_root = dget(sdp->sd_master_dir); |
1313 | return 0; | |
b3b94faa DT |
1314 | } |
1315 | ||
419c93e0 | 1316 | static void gfs2_kill_sb(struct super_block *sb) |
86384605 AD |
1317 | { |
1318 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
88a19ad0 SW |
1319 | |
1320 | if (sdp == NULL) { | |
1321 | kill_block_super(sb); | |
1322 | return; | |
3af165ac | 1323 | } |
88a19ad0 SW |
1324 | |
1325 | gfs2_meta_syncfs(sdp); | |
1326 | dput(sdp->sd_root_dir); | |
1327 | dput(sdp->sd_master_dir); | |
1328 | sdp->sd_root_dir = NULL; | |
1329 | sdp->sd_master_dir = NULL; | |
fefc03bf | 1330 | shrink_dcache_sb(sb); |
9b8df98f | 1331 | kill_block_super(sb); |
88a19ad0 SW |
1332 | gfs2_delete_debugfs_file(sdp); |
1333 | kfree(sdp); | |
86384605 AD |
1334 | } |
1335 | ||
b3b94faa DT |
1336 | struct file_system_type gfs2_fs_type = { |
1337 | .name = "gfs2", | |
1338 | .fs_flags = FS_REQUIRES_DEV, | |
1339 | .get_sb = gfs2_get_sb, | |
419c93e0 SW |
1340 | .kill_sb = gfs2_kill_sb, |
1341 | .owner = THIS_MODULE, | |
1342 | }; | |
1343 | ||
1344 | struct file_system_type gfs2meta_fs_type = { | |
1345 | .name = "gfs2meta", | |
1346 | .fs_flags = FS_REQUIRES_DEV, | |
86384605 | 1347 | .get_sb = gfs2_get_sb_meta, |
b3b94faa DT |
1348 | .owner = THIS_MODULE, |
1349 | }; | |
1350 |