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> |
7d308590 | 20 | #include <linux/lm_interface.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 | 28 | #include "mount.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" |
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 SW |
59 | gt->gt_incore_log_blocks = 1024; |
60 | gt->gt_log_flush_secs = 60; | |
61 | gt->gt_recoverd_secs = 60; | |
62 | gt->gt_logd_secs = 1; | |
9b8df98f SW |
63 | gt->gt_quota_simul_sync = 64; |
64 | gt->gt_quota_warn_period = 10; | |
65 | gt->gt_quota_scale_num = 1; | |
66 | gt->gt_quota_scale_den = 1; | |
67 | gt->gt_quota_cache_secs = 300; | |
68 | gt->gt_quota_quantum = 60; | |
9b8df98f SW |
69 | gt->gt_new_files_jdata = 0; |
70 | gt->gt_max_readahead = 1 << 18; | |
71 | gt->gt_stall_secs = 600; | |
72 | gt->gt_complain_secs = 10; | |
73 | gt->gt_statfs_quantum = 30; | |
74 | gt->gt_statfs_slow = 0; | |
75 | } | |
76 | ||
b3b94faa DT |
77 | static struct gfs2_sbd *init_sbd(struct super_block *sb) |
78 | { | |
79 | struct gfs2_sbd *sdp; | |
b3b94faa | 80 | |
85d1da67 | 81 | sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL); |
b3b94faa DT |
82 | if (!sdp) |
83 | return NULL; | |
84 | ||
5c676f6d | 85 | sb->s_fs_info = sdp; |
b3b94faa DT |
86 | sdp->sd_vfs = sb; |
87 | ||
88 | gfs2_tune_init(&sdp->sd_tune); | |
89 | ||
f55ab26a | 90 | mutex_init(&sdp->sd_inum_mutex); |
b3b94faa | 91 | spin_lock_init(&sdp->sd_statfs_spin); |
b3b94faa DT |
92 | |
93 | spin_lock_init(&sdp->sd_rindex_spin); | |
f55ab26a | 94 | mutex_init(&sdp->sd_rindex_mutex); |
b3b94faa DT |
95 | INIT_LIST_HEAD(&sdp->sd_rindex_list); |
96 | INIT_LIST_HEAD(&sdp->sd_rindex_mru_list); | |
b3b94faa DT |
97 | |
98 | INIT_LIST_HEAD(&sdp->sd_jindex_list); | |
99 | spin_lock_init(&sdp->sd_jindex_spin); | |
f55ab26a | 100 | mutex_init(&sdp->sd_jindex_mutex); |
b3b94faa | 101 | |
b3b94faa DT |
102 | INIT_LIST_HEAD(&sdp->sd_quota_list); |
103 | spin_lock_init(&sdp->sd_quota_spin); | |
f55ab26a | 104 | mutex_init(&sdp->sd_quota_mutex); |
37b2c837 | 105 | init_waitqueue_head(&sdp->sd_quota_wait); |
813e0c46 SW |
106 | INIT_LIST_HEAD(&sdp->sd_trunc_list); |
107 | spin_lock_init(&sdp->sd_trunc_lock); | |
b3b94faa DT |
108 | |
109 | spin_lock_init(&sdp->sd_log_lock); | |
b3b94faa | 110 | |
b3b94faa DT |
111 | INIT_LIST_HEAD(&sdp->sd_log_le_buf); |
112 | INIT_LIST_HEAD(&sdp->sd_log_le_revoke); | |
113 | INIT_LIST_HEAD(&sdp->sd_log_le_rg); | |
114 | INIT_LIST_HEAD(&sdp->sd_log_le_databuf); | |
d7b616e2 | 115 | INIT_LIST_HEAD(&sdp->sd_log_le_ordered); |
b3b94faa | 116 | |
71b86f56 | 117 | mutex_init(&sdp->sd_log_reserve_mutex); |
b3b94faa DT |
118 | INIT_LIST_HEAD(&sdp->sd_ail1_list); |
119 | INIT_LIST_HEAD(&sdp->sd_ail2_list); | |
120 | ||
484adff8 | 121 | init_rwsem(&sdp->sd_log_flush_lock); |
16615be1 SW |
122 | atomic_set(&sdp->sd_log_in_flight, 0); |
123 | init_waitqueue_head(&sdp->sd_log_flush_wait); | |
b3b94faa DT |
124 | |
125 | INIT_LIST_HEAD(&sdp->sd_revoke_list); | |
126 | ||
f55ab26a | 127 | mutex_init(&sdp->sd_freeze_lock); |
b3b94faa DT |
128 | |
129 | return sdp; | |
130 | } | |
131 | ||
b3b94faa | 132 | |
9b8df98f SW |
133 | /** |
134 | * gfs2_check_sb - Check superblock | |
135 | * @sdp: the filesystem | |
136 | * @sb: The superblock | |
137 | * @silent: Don't print a message if the check fails | |
138 | * | |
139 | * Checks the version code of the FS is one that we understand how to | |
140 | * read and that the sizes of the various on-disk structures have not | |
141 | * changed. | |
142 | */ | |
143 | ||
144 | static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent) | |
145 | { | |
146 | unsigned int x; | |
147 | ||
148 | if (sb->sb_magic != GFS2_MAGIC || | |
149 | sb->sb_type != GFS2_METATYPE_SB) { | |
150 | if (!silent) | |
151 | printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n"); | |
152 | return -EINVAL; | |
153 | } | |
154 | ||
155 | /* If format numbers match exactly, we're done. */ | |
156 | ||
157 | if (sb->sb_fs_format == GFS2_FORMAT_FS && | |
158 | sb->sb_multihost_format == GFS2_FORMAT_MULTI) | |
159 | return 0; | |
160 | ||
161 | if (sb->sb_fs_format != GFS2_FORMAT_FS) { | |
162 | for (x = 0; gfs2_old_fs_formats[x]; x++) | |
163 | if (gfs2_old_fs_formats[x] == sb->sb_fs_format) | |
164 | break; | |
165 | ||
166 | if (!gfs2_old_fs_formats[x]) { | |
167 | printk(KERN_WARNING | |
168 | "GFS2: code version (%u, %u) is incompatible " | |
169 | "with ondisk format (%u, %u)\n", | |
170 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
171 | sb->sb_fs_format, sb->sb_multihost_format); | |
172 | printk(KERN_WARNING | |
173 | "GFS2: I don't know how to upgrade this FS\n"); | |
174 | return -EINVAL; | |
175 | } | |
176 | } | |
177 | ||
178 | if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) { | |
179 | for (x = 0; gfs2_old_multihost_formats[x]; x++) | |
180 | if (gfs2_old_multihost_formats[x] == | |
181 | sb->sb_multihost_format) | |
182 | break; | |
183 | ||
184 | if (!gfs2_old_multihost_formats[x]) { | |
185 | printk(KERN_WARNING | |
186 | "GFS2: code version (%u, %u) is incompatible " | |
187 | "with ondisk format (%u, %u)\n", | |
188 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
189 | sb->sb_fs_format, sb->sb_multihost_format); | |
190 | printk(KERN_WARNING | |
191 | "GFS2: I don't know how to upgrade this FS\n"); | |
192 | return -EINVAL; | |
193 | } | |
194 | } | |
195 | ||
196 | if (!sdp->sd_args.ar_upgrade) { | |
197 | printk(KERN_WARNING | |
198 | "GFS2: code version (%u, %u) is incompatible " | |
199 | "with ondisk format (%u, %u)\n", | |
200 | GFS2_FORMAT_FS, GFS2_FORMAT_MULTI, | |
201 | sb->sb_fs_format, sb->sb_multihost_format); | |
202 | printk(KERN_INFO | |
203 | "GFS2: Use the \"upgrade\" mount option to upgrade " | |
204 | "the FS\n"); | |
205 | printk(KERN_INFO "GFS2: See the manual for more details\n"); | |
206 | return -EINVAL; | |
207 | } | |
208 | ||
209 | return 0; | |
210 | } | |
211 | ||
212 | static void end_bio_io_page(struct bio *bio, int error) | |
213 | { | |
214 | struct page *page = bio->bi_private; | |
215 | ||
216 | if (!error) | |
217 | SetPageUptodate(page); | |
218 | else | |
219 | printk(KERN_WARNING "gfs2: error %d reading superblock\n", error); | |
220 | unlock_page(page); | |
221 | } | |
222 | ||
223 | static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf) | |
224 | { | |
225 | const struct gfs2_sb *str = buf; | |
226 | ||
227 | sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic); | |
228 | sb->sb_type = be32_to_cpu(str->sb_header.mh_type); | |
229 | sb->sb_format = be32_to_cpu(str->sb_header.mh_format); | |
230 | sb->sb_fs_format = be32_to_cpu(str->sb_fs_format); | |
231 | sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format); | |
232 | sb->sb_bsize = be32_to_cpu(str->sb_bsize); | |
233 | sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift); | |
234 | sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr); | |
235 | sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino); | |
236 | sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr); | |
237 | sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino); | |
238 | ||
239 | memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN); | |
240 | memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN); | |
241 | } | |
242 | ||
243 | /** | |
244 | * gfs2_read_super - Read the gfs2 super block from disk | |
245 | * @sdp: The GFS2 super block | |
246 | * @sector: The location of the super block | |
247 | * @error: The error code to return | |
248 | * | |
249 | * This uses the bio functions to read the super block from disk | |
250 | * because we want to be 100% sure that we never read cached data. | |
251 | * A super block is read twice only during each GFS2 mount and is | |
252 | * never written to by the filesystem. The first time its read no | |
253 | * locks are held, and the only details which are looked at are those | |
254 | * relating to the locking protocol. Once locking is up and working, | |
255 | * the sb is read again under the lock to establish the location of | |
256 | * the master directory (contains pointers to journals etc) and the | |
257 | * root directory. | |
258 | * | |
259 | * Returns: 0 on success or error | |
260 | */ | |
261 | ||
262 | static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector) | |
263 | { | |
264 | struct super_block *sb = sdp->sd_vfs; | |
265 | struct gfs2_sb *p; | |
266 | struct page *page; | |
267 | struct bio *bio; | |
268 | ||
269 | page = alloc_page(GFP_NOFS); | |
270 | if (unlikely(!page)) | |
271 | return -ENOBUFS; | |
272 | ||
273 | ClearPageUptodate(page); | |
274 | ClearPageDirty(page); | |
275 | lock_page(page); | |
276 | ||
277 | bio = bio_alloc(GFP_NOFS, 1); | |
278 | if (unlikely(!bio)) { | |
279 | __free_page(page); | |
280 | return -ENOBUFS; | |
281 | } | |
282 | ||
283 | bio->bi_sector = sector * (sb->s_blocksize >> 9); | |
284 | bio->bi_bdev = sb->s_bdev; | |
285 | bio_add_page(bio, page, PAGE_SIZE, 0); | |
286 | ||
287 | bio->bi_end_io = end_bio_io_page; | |
288 | bio->bi_private = page; | |
289 | submit_bio(READ_SYNC | (1 << BIO_RW_META), bio); | |
290 | wait_on_page_locked(page); | |
291 | bio_put(bio); | |
292 | if (!PageUptodate(page)) { | |
293 | __free_page(page); | |
294 | return -EIO; | |
295 | } | |
296 | p = kmap(page); | |
297 | gfs2_sb_in(&sdp->sd_sb, p); | |
298 | kunmap(page); | |
299 | __free_page(page); | |
300 | return 0; | |
301 | } | |
302 | /** | |
303 | * gfs2_read_sb - Read super block | |
304 | * @sdp: The GFS2 superblock | |
305 | * @gl: the glock for the superblock (assumed to be held) | |
306 | * @silent: Don't print message if mount fails | |
307 | * | |
308 | */ | |
309 | ||
310 | static int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent) | |
311 | { | |
312 | u32 hash_blocks, ind_blocks, leaf_blocks; | |
313 | u32 tmp_blocks; | |
314 | unsigned int x; | |
315 | int error; | |
316 | ||
317 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); | |
318 | if (error) { | |
319 | if (!silent) | |
320 | fs_err(sdp, "can't read superblock\n"); | |
321 | return error; | |
322 | } | |
323 | ||
324 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
325 | if (error) | |
326 | return error; | |
327 | ||
328 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
329 | GFS2_BASIC_BLOCK_SHIFT; | |
330 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
331 | sdp->sd_diptrs = (sdp->sd_sb.sb_bsize - | |
332 | sizeof(struct gfs2_dinode)) / sizeof(u64); | |
333 | sdp->sd_inptrs = (sdp->sd_sb.sb_bsize - | |
334 | sizeof(struct gfs2_meta_header)) / sizeof(u64); | |
335 | sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header); | |
336 | sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2; | |
337 | sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1; | |
338 | sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64); | |
339 | sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize - | |
340 | sizeof(struct gfs2_meta_header)) / | |
341 | sizeof(struct gfs2_quota_change); | |
342 | ||
343 | /* Compute maximum reservation required to add a entry to a directory */ | |
344 | ||
345 | hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH), | |
346 | sdp->sd_jbsize); | |
347 | ||
348 | ind_blocks = 0; | |
349 | for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) { | |
350 | tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs); | |
351 | ind_blocks += tmp_blocks; | |
352 | } | |
353 | ||
354 | leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH; | |
355 | ||
356 | sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks; | |
357 | ||
358 | sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize - | |
359 | sizeof(struct gfs2_dinode); | |
360 | sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs; | |
361 | for (x = 2;; x++) { | |
362 | u64 space, d; | |
363 | u32 m; | |
364 | ||
365 | space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs; | |
366 | d = space; | |
367 | m = do_div(d, sdp->sd_inptrs); | |
368 | ||
369 | if (d != sdp->sd_heightsize[x - 1] || m) | |
370 | break; | |
371 | sdp->sd_heightsize[x] = space; | |
372 | } | |
373 | sdp->sd_max_height = x; | |
374 | sdp->sd_heightsize[x] = ~0; | |
375 | gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT); | |
376 | ||
377 | sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize - | |
378 | sizeof(struct gfs2_dinode); | |
379 | sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs; | |
380 | for (x = 2;; x++) { | |
381 | u64 space, d; | |
382 | u32 m; | |
383 | ||
384 | space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs; | |
385 | d = space; | |
386 | m = do_div(d, sdp->sd_inptrs); | |
387 | ||
388 | if (d != sdp->sd_jheightsize[x - 1] || m) | |
389 | break; | |
390 | sdp->sd_jheightsize[x] = space; | |
391 | } | |
392 | sdp->sd_max_jheight = x; | |
393 | sdp->sd_jheightsize[x] = ~0; | |
394 | gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT); | |
395 | ||
396 | return 0; | |
397 | } | |
398 | ||
b3b94faa DT |
399 | static int init_names(struct gfs2_sbd *sdp, int silent) |
400 | { | |
b3b94faa DT |
401 | char *proto, *table; |
402 | int error = 0; | |
403 | ||
404 | proto = sdp->sd_args.ar_lockproto; | |
405 | table = sdp->sd_args.ar_locktable; | |
406 | ||
407 | /* Try to autodetect */ | |
408 | ||
409 | if (!proto[0] || !table[0]) { | |
bb8d8a6f SW |
410 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); |
411 | if (error) | |
412 | return error; | |
3cf1e7be SW |
413 | |
414 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
b3b94faa DT |
415 | if (error) |
416 | goto out; | |
417 | ||
418 | if (!proto[0]) | |
3cf1e7be | 419 | proto = sdp->sd_sb.sb_lockproto; |
b3b94faa | 420 | if (!table[0]) |
3cf1e7be | 421 | table = sdp->sd_sb.sb_locktable; |
b3b94faa DT |
422 | } |
423 | ||
424 | if (!table[0]) | |
425 | table = sdp->sd_vfs->s_id; | |
426 | ||
00377d8e JD |
427 | strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN); |
428 | strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN); | |
b3b94faa | 429 | |
5d35e31f DC |
430 | table = sdp->sd_table_name; |
431 | while ((table = strchr(table, '/'))) | |
b35997d4 RP |
432 | *table = '_'; |
433 | ||
a91ea69f | 434 | out: |
b3b94faa DT |
435 | return error; |
436 | } | |
437 | ||
438 | static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh, | |
439 | int undo) | |
440 | { | |
b3b94faa DT |
441 | int error = 0; |
442 | ||
443 | if (undo) | |
444 | goto fail_trans; | |
445 | ||
b3b94faa DT |
446 | error = gfs2_glock_nq_num(sdp, |
447 | GFS2_MOUNT_LOCK, &gfs2_nondisk_glops, | |
448 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE, | |
449 | mount_gh); | |
450 | if (error) { | |
451 | fs_err(sdp, "can't acquire mount glock: %d\n", error); | |
452 | goto fail; | |
453 | } | |
454 | ||
455 | error = gfs2_glock_nq_num(sdp, | |
456 | GFS2_LIVE_LOCK, &gfs2_nondisk_glops, | |
457 | LM_ST_SHARED, | |
579b78a4 | 458 | LM_FLAG_NOEXP | GL_EXACT, |
b3b94faa DT |
459 | &sdp->sd_live_gh); |
460 | if (error) { | |
461 | fs_err(sdp, "can't acquire live glock: %d\n", error); | |
462 | goto fail_mount; | |
463 | } | |
464 | ||
465 | error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops, | |
466 | CREATE, &sdp->sd_rename_gl); | |
467 | if (error) { | |
468 | fs_err(sdp, "can't create rename glock: %d\n", error); | |
469 | goto fail_live; | |
470 | } | |
471 | ||
472 | error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops, | |
473 | CREATE, &sdp->sd_trans_gl); | |
474 | if (error) { | |
475 | fs_err(sdp, "can't create transaction glock: %d\n", error); | |
476 | goto fail_rename; | |
477 | } | |
b3b94faa DT |
478 | |
479 | return 0; | |
480 | ||
feaa7bba | 481 | fail_trans: |
b3b94faa | 482 | gfs2_glock_put(sdp->sd_trans_gl); |
feaa7bba | 483 | fail_rename: |
b3b94faa | 484 | gfs2_glock_put(sdp->sd_rename_gl); |
feaa7bba | 485 | fail_live: |
b3b94faa | 486 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); |
feaa7bba | 487 | fail_mount: |
b3b94faa | 488 | gfs2_glock_dq_uninit(mount_gh); |
feaa7bba | 489 | fail: |
b3b94faa DT |
490 | return error; |
491 | } | |
492 | ||
9b8df98f SW |
493 | static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr, |
494 | u64 no_addr, const char *name) | |
f42faf4f | 495 | { |
9b8df98f SW |
496 | struct gfs2_sbd *sdp = sb->s_fs_info; |
497 | struct dentry *dentry; | |
498 | struct inode *inode; | |
499 | ||
500 | inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0); | |
501 | if (IS_ERR(inode)) { | |
502 | fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode)); | |
503 | return PTR_ERR(inode); | |
504 | } | |
505 | dentry = d_alloc_root(inode); | |
506 | if (!dentry) { | |
507 | fs_err(sdp, "can't alloc %s dentry\n", name); | |
508 | iput(inode); | |
509 | return -ENOMEM; | |
510 | } | |
511 | dentry->d_op = &gfs2_dops; | |
512 | *dptr = dentry; | |
513 | return 0; | |
f42faf4f SW |
514 | } |
515 | ||
9b8df98f | 516 | static int init_sb(struct gfs2_sbd *sdp, int silent) |
b3b94faa DT |
517 | { |
518 | struct super_block *sb = sdp->sd_vfs; | |
519 | struct gfs2_holder sb_gh; | |
dbb7cae2 | 520 | u64 no_addr; |
9b8df98f | 521 | int ret; |
b3b94faa | 522 | |
9b8df98f SW |
523 | ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops, |
524 | LM_ST_SHARED, 0, &sb_gh); | |
525 | if (ret) { | |
526 | fs_err(sdp, "can't acquire superblock glock: %d\n", ret); | |
527 | return ret; | |
b3b94faa | 528 | } |
907b9bce | 529 | |
9b8df98f SW |
530 | ret = gfs2_read_sb(sdp, sb_gh.gh_gl, silent); |
531 | if (ret) { | |
532 | fs_err(sdp, "can't read superblock: %d\n", ret); | |
b3b94faa DT |
533 | goto out; |
534 | } | |
535 | ||
536 | /* Set up the buffer cache and SB for real */ | |
b3b94faa | 537 | if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) { |
9b8df98f | 538 | ret = -EINVAL; |
b3b94faa DT |
539 | fs_err(sdp, "FS block size (%u) is too small for device " |
540 | "block size (%u)\n", | |
541 | sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev)); | |
542 | goto out; | |
543 | } | |
544 | if (sdp->sd_sb.sb_bsize > PAGE_SIZE) { | |
9b8df98f | 545 | ret = -EINVAL; |
b3b94faa DT |
546 | fs_err(sdp, "FS block size (%u) is too big for machine " |
547 | "page size (%u)\n", | |
548 | sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE); | |
549 | goto out; | |
550 | } | |
b3b94faa DT |
551 | sb_set_blocksize(sb, sdp->sd_sb.sb_bsize); |
552 | ||
f42faf4f | 553 | /* Get the root inode */ |
dbb7cae2 | 554 | no_addr = sdp->sd_sb.sb_root_dir.no_addr; |
9b8df98f SW |
555 | ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root"); |
556 | if (ret) | |
f42faf4f | 557 | goto out; |
b3b94faa | 558 | |
9b8df98f SW |
559 | /* Get the master inode */ |
560 | no_addr = sdp->sd_sb.sb_master_dir.no_addr; | |
561 | ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master"); | |
562 | if (ret) { | |
563 | dput(sdp->sd_root_dir); | |
564 | goto out; | |
565 | } | |
566 | sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir); | |
f42faf4f | 567 | out: |
b3b94faa | 568 | gfs2_glock_dq_uninit(&sb_gh); |
9b8df98f | 569 | return ret; |
b3b94faa DT |
570 | } |
571 | ||
da6dd40d BP |
572 | /** |
573 | * map_journal_extents - create a reusable "extent" mapping from all logical | |
574 | * blocks to all physical blocks for the given journal. This will save | |
575 | * us time when writing journal blocks. Most journals will have only one | |
576 | * extent that maps all their logical blocks. That's because gfs2.mkfs | |
577 | * arranges the journal blocks sequentially to maximize performance. | |
578 | * So the extent would map the first block for the entire file length. | |
579 | * However, gfs2_jadd can happen while file activity is happening, so | |
580 | * those journals may not be sequential. Less likely is the case where | |
581 | * the users created their own journals by mounting the metafs and | |
582 | * laying it out. But it's still possible. These journals might have | |
583 | * several extents. | |
584 | * | |
585 | * TODO: This should be done in bigger chunks rather than one block at a time, | |
586 | * but since it's only done at mount time, I'm not worried about the | |
587 | * time it takes. | |
588 | */ | |
589 | static int map_journal_extents(struct gfs2_sbd *sdp) | |
590 | { | |
591 | struct gfs2_jdesc *jd = sdp->sd_jdesc; | |
592 | unsigned int lb; | |
593 | u64 db, prev_db; /* logical block, disk block, prev disk block */ | |
594 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); | |
595 | struct gfs2_journal_extent *jext = NULL; | |
596 | struct buffer_head bh; | |
597 | int rc = 0; | |
598 | ||
da6dd40d BP |
599 | prev_db = 0; |
600 | ||
c9e98886 | 601 | for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) { |
da6dd40d BP |
602 | bh.b_state = 0; |
603 | bh.b_blocknr = 0; | |
604 | bh.b_size = 1 << ip->i_inode.i_blkbits; | |
605 | rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0); | |
606 | db = bh.b_blocknr; | |
607 | if (rc || !db) { | |
608 | printk(KERN_INFO "GFS2 journal mapping error %d: lb=" | |
609 | "%u db=%llu\n", rc, lb, (unsigned long long)db); | |
610 | break; | |
611 | } | |
612 | if (!prev_db || db != prev_db + 1) { | |
613 | jext = kzalloc(sizeof(struct gfs2_journal_extent), | |
614 | GFP_KERNEL); | |
615 | if (!jext) { | |
616 | printk(KERN_INFO "GFS2 error: out of memory " | |
617 | "mapping journal extents.\n"); | |
618 | rc = -ENOMEM; | |
619 | break; | |
620 | } | |
621 | jext->dblock = db; | |
622 | jext->lblock = lb; | |
623 | jext->blocks = 1; | |
624 | list_add_tail(&jext->extent_list, &jd->extent_list); | |
625 | } else { | |
626 | jext->blocks++; | |
627 | } | |
628 | prev_db = db; | |
629 | } | |
630 | return rc; | |
631 | } | |
632 | ||
da755fdb SW |
633 | static void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp) |
634 | { | |
048bca22 SW |
635 | if (!sdp->sd_lockstruct.ls_ops->lm_others_may_mount) |
636 | return; | |
da755fdb SW |
637 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) |
638 | sdp->sd_lockstruct.ls_ops->lm_others_may_mount( | |
639 | sdp->sd_lockstruct.ls_lockspace); | |
640 | } | |
641 | ||
b5289681 SW |
642 | /** |
643 | * gfs2_jindex_hold - Grab a lock on the jindex | |
644 | * @sdp: The GFS2 superblock | |
645 | * @ji_gh: the holder for the jindex glock | |
646 | * | |
647 | * Returns: errno | |
648 | */ | |
649 | ||
650 | static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh) | |
651 | { | |
652 | struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex); | |
653 | struct qstr name; | |
654 | char buf[20]; | |
655 | struct gfs2_jdesc *jd; | |
656 | int error; | |
657 | ||
658 | name.name = buf; | |
659 | ||
660 | mutex_lock(&sdp->sd_jindex_mutex); | |
661 | ||
662 | for (;;) { | |
663 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh); | |
664 | if (error) | |
665 | break; | |
666 | ||
667 | name.len = sprintf(buf, "journal%u", sdp->sd_journals); | |
668 | name.hash = gfs2_disk_hash(name.name, name.len); | |
669 | ||
670 | error = gfs2_dir_check(sdp->sd_jindex, &name, NULL); | |
671 | if (error == -ENOENT) { | |
672 | error = 0; | |
673 | break; | |
674 | } | |
675 | ||
676 | gfs2_glock_dq_uninit(ji_gh); | |
677 | ||
678 | if (error) | |
679 | break; | |
680 | ||
681 | error = -ENOMEM; | |
682 | jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL); | |
683 | if (!jd) | |
684 | break; | |
685 | ||
686 | INIT_LIST_HEAD(&jd->extent_list); | |
687 | jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1); | |
688 | if (!jd->jd_inode || IS_ERR(jd->jd_inode)) { | |
689 | if (!jd->jd_inode) | |
690 | error = -ENOENT; | |
691 | else | |
692 | error = PTR_ERR(jd->jd_inode); | |
693 | kfree(jd); | |
694 | break; | |
695 | } | |
696 | ||
697 | spin_lock(&sdp->sd_jindex_spin); | |
698 | jd->jd_jid = sdp->sd_journals++; | |
699 | list_add_tail(&jd->jd_list, &sdp->sd_jindex_list); | |
700 | spin_unlock(&sdp->sd_jindex_spin); | |
701 | } | |
702 | ||
703 | mutex_unlock(&sdp->sd_jindex_mutex); | |
704 | ||
705 | return error; | |
706 | } | |
707 | ||
b3b94faa DT |
708 | static int init_journal(struct gfs2_sbd *sdp, int undo) |
709 | { | |
9b8df98f | 710 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
711 | struct gfs2_holder ji_gh; |
712 | struct task_struct *p; | |
5c676f6d | 713 | struct gfs2_inode *ip; |
b3b94faa DT |
714 | int jindex = 1; |
715 | int error = 0; | |
716 | ||
717 | if (undo) { | |
718 | jindex = 0; | |
719 | goto fail_recoverd; | |
720 | } | |
721 | ||
9b8df98f | 722 | sdp->sd_jindex = gfs2_lookup_simple(master, "jindex"); |
c752666c | 723 | if (IS_ERR(sdp->sd_jindex)) { |
b3b94faa | 724 | fs_err(sdp, "can't lookup journal index: %d\n", error); |
c752666c | 725 | return PTR_ERR(sdp->sd_jindex); |
b3b94faa | 726 | } |
feaa7bba | 727 | ip = GFS2_I(sdp->sd_jindex); |
b3b94faa DT |
728 | |
729 | /* Load in the journal index special file */ | |
730 | ||
731 | error = gfs2_jindex_hold(sdp, &ji_gh); | |
732 | if (error) { | |
733 | fs_err(sdp, "can't read journal index: %d\n", error); | |
734 | goto fail; | |
735 | } | |
736 | ||
737 | error = -EINVAL; | |
738 | if (!gfs2_jindex_size(sdp)) { | |
739 | fs_err(sdp, "no journals!\n"); | |
907b9bce | 740 | goto fail_jindex; |
b3b94faa DT |
741 | } |
742 | ||
743 | if (sdp->sd_args.ar_spectator) { | |
744 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0); | |
fd041f0b | 745 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
b3b94faa DT |
746 | } else { |
747 | if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) { | |
748 | fs_err(sdp, "can't mount journal #%u\n", | |
749 | sdp->sd_lockstruct.ls_jid); | |
750 | fs_err(sdp, "there are only %u journals (0 - %u)\n", | |
751 | gfs2_jindex_size(sdp), | |
752 | gfs2_jindex_size(sdp) - 1); | |
753 | goto fail_jindex; | |
754 | } | |
755 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid); | |
756 | ||
feaa7bba | 757 | error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid, |
b3b94faa DT |
758 | &gfs2_journal_glops, |
759 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP, | |
760 | &sdp->sd_journal_gh); | |
761 | if (error) { | |
762 | fs_err(sdp, "can't acquire journal glock: %d\n", error); | |
763 | goto fail_jindex; | |
764 | } | |
765 | ||
feaa7bba SW |
766 | ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
767 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, | |
75be73a8 | 768 | LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE, |
b3b94faa DT |
769 | &sdp->sd_jinode_gh); |
770 | if (error) { | |
771 | fs_err(sdp, "can't acquire journal inode glock: %d\n", | |
772 | error); | |
773 | goto fail_journal_gh; | |
774 | } | |
775 | ||
776 | error = gfs2_jdesc_check(sdp->sd_jdesc); | |
777 | if (error) { | |
778 | fs_err(sdp, "my journal (%u) is bad: %d\n", | |
779 | sdp->sd_jdesc->jd_jid, error); | |
780 | goto fail_jinode_gh; | |
781 | } | |
fd041f0b | 782 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
da6dd40d BP |
783 | |
784 | /* Map the extents for this journal's blocks */ | |
785 | map_journal_extents(sdp); | |
b3b94faa DT |
786 | } |
787 | ||
788 | if (sdp->sd_lockstruct.ls_first) { | |
789 | unsigned int x; | |
790 | for (x = 0; x < sdp->sd_journals; x++) { | |
c63e31c2 | 791 | error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x)); |
b3b94faa DT |
792 | if (error) { |
793 | fs_err(sdp, "error recovering journal %u: %d\n", | |
794 | x, error); | |
795 | goto fail_jinode_gh; | |
796 | } | |
797 | } | |
798 | ||
799 | gfs2_lm_others_may_mount(sdp); | |
800 | } else if (!sdp->sd_args.ar_spectator) { | |
c63e31c2 | 801 | error = gfs2_recover_journal(sdp->sd_jdesc); |
b3b94faa DT |
802 | if (error) { |
803 | fs_err(sdp, "error recovering my journal: %d\n", error); | |
804 | goto fail_jinode_gh; | |
805 | } | |
806 | } | |
807 | ||
808 | set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags); | |
809 | gfs2_glock_dq_uninit(&ji_gh); | |
810 | jindex = 0; | |
811 | ||
b3b94faa DT |
812 | p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd"); |
813 | error = IS_ERR(p); | |
814 | if (error) { | |
815 | fs_err(sdp, "can't start recoverd thread: %d\n", error); | |
816 | goto fail_jinode_gh; | |
817 | } | |
818 | sdp->sd_recoverd_process = p; | |
819 | ||
820 | return 0; | |
821 | ||
a91ea69f | 822 | fail_recoverd: |
b3b94faa | 823 | kthread_stop(sdp->sd_recoverd_process); |
a91ea69f | 824 | fail_jinode_gh: |
b3b94faa DT |
825 | if (!sdp->sd_args.ar_spectator) |
826 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
a91ea69f | 827 | fail_journal_gh: |
b3b94faa DT |
828 | if (!sdp->sd_args.ar_spectator) |
829 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
a91ea69f | 830 | fail_jindex: |
b3b94faa DT |
831 | gfs2_jindex_free(sdp); |
832 | if (jindex) | |
833 | gfs2_glock_dq_uninit(&ji_gh); | |
a91ea69f | 834 | fail: |
f42faf4f | 835 | iput(sdp->sd_jindex); |
b3b94faa DT |
836 | return error; |
837 | } | |
838 | ||
b3b94faa DT |
839 | |
840 | static int init_inodes(struct gfs2_sbd *sdp, int undo) | |
841 | { | |
b3b94faa | 842 | int error = 0; |
5c676f6d | 843 | struct gfs2_inode *ip; |
9b8df98f | 844 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
845 | |
846 | if (undo) | |
f42faf4f SW |
847 | goto fail_qinode; |
848 | ||
f42faf4f SW |
849 | error = init_journal(sdp, undo); |
850 | if (error) | |
9b8df98f | 851 | goto fail; |
b3b94faa DT |
852 | |
853 | /* Read in the master inode number inode */ | |
9b8df98f | 854 | sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum"); |
c752666c SW |
855 | if (IS_ERR(sdp->sd_inum_inode)) { |
856 | error = PTR_ERR(sdp->sd_inum_inode); | |
b3b94faa | 857 | fs_err(sdp, "can't read in inum inode: %d\n", error); |
f42faf4f | 858 | goto fail_journal; |
b3b94faa DT |
859 | } |
860 | ||
f42faf4f | 861 | |
b3b94faa | 862 | /* Read in the master statfs inode */ |
9b8df98f | 863 | sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs"); |
c752666c SW |
864 | if (IS_ERR(sdp->sd_statfs_inode)) { |
865 | error = PTR_ERR(sdp->sd_statfs_inode); | |
b3b94faa | 866 | fs_err(sdp, "can't read in statfs inode: %d\n", error); |
f42faf4f | 867 | goto fail_inum; |
b3b94faa DT |
868 | } |
869 | ||
870 | /* Read in the resource index inode */ | |
9b8df98f | 871 | sdp->sd_rindex = gfs2_lookup_simple(master, "rindex"); |
c752666c SW |
872 | if (IS_ERR(sdp->sd_rindex)) { |
873 | error = PTR_ERR(sdp->sd_rindex); | |
b3b94faa DT |
874 | fs_err(sdp, "can't get resource index inode: %d\n", error); |
875 | goto fail_statfs; | |
876 | } | |
feaa7bba | 877 | ip = GFS2_I(sdp->sd_rindex); |
cf45b752 | 878 | sdp->sd_rindex_uptodate = 0; |
b3b94faa DT |
879 | |
880 | /* Read in the quota inode */ | |
9b8df98f | 881 | sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota"); |
c752666c SW |
882 | if (IS_ERR(sdp->sd_quota_inode)) { |
883 | error = PTR_ERR(sdp->sd_quota_inode); | |
b3b94faa DT |
884 | fs_err(sdp, "can't get quota file inode: %d\n", error); |
885 | goto fail_rindex; | |
886 | } | |
b3b94faa DT |
887 | return 0; |
888 | ||
f42faf4f SW |
889 | fail_qinode: |
890 | iput(sdp->sd_quota_inode); | |
f42faf4f | 891 | fail_rindex: |
b3b94faa | 892 | gfs2_clear_rgrpd(sdp); |
f42faf4f | 893 | iput(sdp->sd_rindex); |
f42faf4f SW |
894 | fail_statfs: |
895 | iput(sdp->sd_statfs_inode); | |
f42faf4f SW |
896 | fail_inum: |
897 | iput(sdp->sd_inum_inode); | |
898 | fail_journal: | |
899 | init_journal(sdp, UNDO); | |
f42faf4f | 900 | fail: |
b3b94faa DT |
901 | return error; |
902 | } | |
903 | ||
904 | static int init_per_node(struct gfs2_sbd *sdp, int undo) | |
905 | { | |
f42faf4f | 906 | struct inode *pn = NULL; |
b3b94faa DT |
907 | char buf[30]; |
908 | int error = 0; | |
5c676f6d | 909 | struct gfs2_inode *ip; |
9b8df98f | 910 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
911 | |
912 | if (sdp->sd_args.ar_spectator) | |
913 | return 0; | |
914 | ||
915 | if (undo) | |
916 | goto fail_qc_gh; | |
917 | ||
9b8df98f | 918 | pn = gfs2_lookup_simple(master, "per_node"); |
c752666c SW |
919 | if (IS_ERR(pn)) { |
920 | error = PTR_ERR(pn); | |
b3b94faa DT |
921 | fs_err(sdp, "can't find per_node directory: %d\n", error); |
922 | return error; | |
923 | } | |
924 | ||
925 | sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid); | |
c752666c SW |
926 | sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf); |
927 | if (IS_ERR(sdp->sd_ir_inode)) { | |
928 | error = PTR_ERR(sdp->sd_ir_inode); | |
b3b94faa DT |
929 | fs_err(sdp, "can't find local \"ir\" file: %d\n", error); |
930 | goto fail; | |
931 | } | |
932 | ||
933 | sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid); | |
c752666c SW |
934 | sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf); |
935 | if (IS_ERR(sdp->sd_sc_inode)) { | |
936 | error = PTR_ERR(sdp->sd_sc_inode); | |
b3b94faa DT |
937 | fs_err(sdp, "can't find local \"sc\" file: %d\n", error); |
938 | goto fail_ir_i; | |
939 | } | |
940 | ||
b3b94faa | 941 | sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
942 | sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf); |
943 | if (IS_ERR(sdp->sd_qc_inode)) { | |
944 | error = PTR_ERR(sdp->sd_qc_inode); | |
b3b94faa DT |
945 | fs_err(sdp, "can't find local \"qc\" file: %d\n", error); |
946 | goto fail_ut_i; | |
947 | } | |
948 | ||
f42faf4f | 949 | iput(pn); |
b3b94faa DT |
950 | pn = NULL; |
951 | ||
feaa7bba | 952 | ip = GFS2_I(sdp->sd_ir_inode); |
5c676f6d | 953 | error = gfs2_glock_nq_init(ip->i_gl, |
579b78a4 | 954 | LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
955 | &sdp->sd_ir_gh); |
956 | if (error) { | |
957 | fs_err(sdp, "can't lock local \"ir\" file: %d\n", error); | |
958 | goto fail_qc_i; | |
959 | } | |
960 | ||
feaa7bba | 961 | ip = GFS2_I(sdp->sd_sc_inode); |
5c676f6d | 962 | error = gfs2_glock_nq_init(ip->i_gl, |
579b78a4 | 963 | LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
964 | &sdp->sd_sc_gh); |
965 | if (error) { | |
966 | fs_err(sdp, "can't lock local \"sc\" file: %d\n", error); | |
967 | goto fail_ir_gh; | |
968 | } | |
969 | ||
feaa7bba | 970 | ip = GFS2_I(sdp->sd_qc_inode); |
5c676f6d | 971 | error = gfs2_glock_nq_init(ip->i_gl, |
579b78a4 | 972 | LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
973 | &sdp->sd_qc_gh); |
974 | if (error) { | |
975 | fs_err(sdp, "can't lock local \"qc\" file: %d\n", error); | |
976 | goto fail_ut_gh; | |
977 | } | |
978 | ||
979 | return 0; | |
980 | ||
a91ea69f | 981 | fail_qc_gh: |
b3b94faa | 982 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); |
a91ea69f | 983 | fail_ut_gh: |
b3b94faa | 984 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); |
a91ea69f | 985 | fail_ir_gh: |
b3b94faa | 986 | gfs2_glock_dq_uninit(&sdp->sd_ir_gh); |
a91ea69f | 987 | fail_qc_i: |
f42faf4f | 988 | iput(sdp->sd_qc_inode); |
a91ea69f | 989 | fail_ut_i: |
f42faf4f | 990 | iput(sdp->sd_sc_inode); |
a91ea69f | 991 | fail_ir_i: |
f42faf4f | 992 | iput(sdp->sd_ir_inode); |
a91ea69f | 993 | fail: |
b3b94faa | 994 | if (pn) |
f42faf4f | 995 | iput(pn); |
b3b94faa DT |
996 | return error; |
997 | } | |
998 | ||
999 | static int init_threads(struct gfs2_sbd *sdp, int undo) | |
1000 | { | |
1001 | struct task_struct *p; | |
1002 | int error = 0; | |
1003 | ||
1004 | if (undo) | |
feaa7bba | 1005 | goto fail_quotad; |
b3b94faa DT |
1006 | |
1007 | sdp->sd_log_flush_time = jiffies; | |
1008 | sdp->sd_jindex_refresh_time = jiffies; | |
1009 | ||
1010 | p = kthread_run(gfs2_logd, sdp, "gfs2_logd"); | |
1011 | error = IS_ERR(p); | |
1012 | if (error) { | |
1013 | fs_err(sdp, "can't start logd thread: %d\n", error); | |
1014 | return error; | |
1015 | } | |
1016 | sdp->sd_logd_process = p; | |
1017 | ||
b3b94faa DT |
1018 | p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad"); |
1019 | error = IS_ERR(p); | |
1020 | if (error) { | |
1021 | fs_err(sdp, "can't start quotad thread: %d\n", error); | |
1022 | goto fail; | |
1023 | } | |
1024 | sdp->sd_quotad_process = p; | |
1025 | ||
b3b94faa DT |
1026 | return 0; |
1027 | ||
b3b94faa | 1028 | |
feaa7bba | 1029 | fail_quotad: |
b3b94faa | 1030 | kthread_stop(sdp->sd_quotad_process); |
feaa7bba | 1031 | fail: |
b3b94faa | 1032 | kthread_stop(sdp->sd_logd_process); |
b3b94faa DT |
1033 | return error; |
1034 | } | |
da755fdb SW |
1035 | |
1036 | /** | |
1037 | * gfs2_lm_mount - mount a locking protocol | |
1038 | * @sdp: the filesystem | |
1039 | * @args: mount arguements | |
1040 | * @silent: if 1, don't complain if the FS isn't a GFS2 fs | |
1041 | * | |
1042 | * Returns: errno | |
1043 | */ | |
1044 | ||
1045 | static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent) | |
1046 | { | |
1047 | char *proto = sdp->sd_proto_name; | |
1048 | char *table = sdp->sd_table_name; | |
58e9fee1 | 1049 | int flags = LM_MFLAG_CONV_NODROP; |
da755fdb SW |
1050 | int error; |
1051 | ||
1052 | if (sdp->sd_args.ar_spectator) | |
1053 | flags |= LM_MFLAG_SPECTATOR; | |
1054 | ||
1055 | fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table); | |
1056 | ||
1057 | error = gfs2_mount_lockproto(proto, table, sdp->sd_args.ar_hostdata, | |
1058 | gfs2_glock_cb, sdp, | |
1059 | GFS2_MIN_LVB_SIZE, flags, | |
1060 | &sdp->sd_lockstruct, &sdp->sd_kobj); | |
1061 | if (error) { | |
1062 | fs_info(sdp, "can't mount proto=%s, table=%s, hostdata=%s\n", | |
1063 | proto, table, sdp->sd_args.ar_hostdata); | |
1064 | goto out; | |
1065 | } | |
1066 | ||
048bca22 | 1067 | if (gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_ops) || |
da755fdb SW |
1068 | gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lvb_size >= |
1069 | GFS2_MIN_LVB_SIZE)) { | |
1070 | gfs2_unmount_lockproto(&sdp->sd_lockstruct); | |
1071 | goto out; | |
1072 | } | |
1073 | ||
1074 | if (sdp->sd_args.ar_spectator) | |
1075 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table); | |
1076 | else | |
1077 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table, | |
1078 | sdp->sd_lockstruct.ls_jid); | |
1079 | ||
1080 | fs_info(sdp, "Joined cluster. Now mounting FS...\n"); | |
1081 | ||
1082 | if ((sdp->sd_lockstruct.ls_flags & LM_LSFLAG_LOCAL) && | |
1083 | !sdp->sd_args.ar_ignore_local_fs) { | |
1084 | sdp->sd_args.ar_localflocks = 1; | |
1085 | sdp->sd_args.ar_localcaching = 1; | |
1086 | } | |
1087 | ||
1088 | out: | |
1089 | return error; | |
1090 | } | |
1091 | ||
1092 | void gfs2_lm_unmount(struct gfs2_sbd *sdp) | |
1093 | { | |
1094 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) | |
1095 | gfs2_unmount_lockproto(&sdp->sd_lockstruct); | |
1096 | } | |
b3b94faa DT |
1097 | |
1098 | /** | |
1099 | * fill_super - Read in superblock | |
1100 | * @sb: The VFS superblock | |
1101 | * @data: Mount options | |
1102 | * @silent: Don't complain if it's not a GFS2 filesystem | |
1103 | * | |
1104 | * Returns: errno | |
1105 | */ | |
1106 | ||
1107 | static int fill_super(struct super_block *sb, void *data, int silent) | |
1108 | { | |
1109 | struct gfs2_sbd *sdp; | |
1110 | struct gfs2_holder mount_gh; | |
1111 | int error; | |
1112 | ||
1113 | sdp = init_sbd(sb); | |
1114 | if (!sdp) { | |
d92a8d48 | 1115 | printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n"); |
b3b94faa DT |
1116 | return -ENOMEM; |
1117 | } | |
1118 | ||
1119 | error = gfs2_mount_args(sdp, (char *)data, 0); | |
1120 | if (error) { | |
d92a8d48 | 1121 | printk(KERN_WARNING "GFS2: can't parse mount arguments\n"); |
b3b94faa DT |
1122 | goto fail; |
1123 | } | |
1124 | ||
719ee344 SW |
1125 | sb->s_magic = GFS2_MAGIC; |
1126 | sb->s_op = &gfs2_super_ops; | |
1127 | sb->s_export_op = &gfs2_export_ops; | |
1128 | sb->s_time_gran = 1; | |
1129 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
419c93e0 SW |
1130 | |
1131 | /* Set up the buffer cache and fill in some fake block size values | |
1132 | to allow us to read-in the on-disk superblock. */ | |
1133 | sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK); | |
1134 | sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits; | |
1135 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
1136 | GFS2_BASIC_BLOCK_SHIFT; | |
1137 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
b3b94faa DT |
1138 | |
1139 | error = init_names(sdp, silent); | |
1140 | if (error) | |
1141 | goto fail; | |
1142 | ||
7c52b166 RP |
1143 | gfs2_create_debugfs_file(sdp); |
1144 | ||
b3b94faa DT |
1145 | error = gfs2_sys_fs_add(sdp); |
1146 | if (error) | |
1147 | goto fail; | |
1148 | ||
1149 | error = gfs2_lm_mount(sdp, silent); | |
1150 | if (error) | |
1151 | goto fail_sys; | |
1152 | ||
1153 | error = init_locking(sdp, &mount_gh, DO); | |
1154 | if (error) | |
1155 | goto fail_lm; | |
1156 | ||
9b8df98f | 1157 | error = init_sb(sdp, silent); |
b3b94faa DT |
1158 | if (error) |
1159 | goto fail_locking; | |
b3b94faa DT |
1160 | |
1161 | error = init_inodes(sdp, DO); | |
1162 | if (error) | |
f42faf4f | 1163 | goto fail_sb; |
b3b94faa DT |
1164 | |
1165 | error = init_per_node(sdp, DO); | |
1166 | if (error) | |
1167 | goto fail_inodes; | |
1168 | ||
1169 | error = gfs2_statfs_init(sdp); | |
1170 | if (error) { | |
1171 | fs_err(sdp, "can't initialize statfs subsystem: %d\n", error); | |
1172 | goto fail_per_node; | |
1173 | } | |
1174 | ||
1175 | error = init_threads(sdp, DO); | |
1176 | if (error) | |
1177 | goto fail_per_node; | |
1178 | ||
1179 | if (!(sb->s_flags & MS_RDONLY)) { | |
1180 | error = gfs2_make_fs_rw(sdp); | |
1181 | if (error) { | |
1182 | fs_err(sdp, "can't make FS RW: %d\n", error); | |
1183 | goto fail_threads; | |
1184 | } | |
1185 | } | |
1186 | ||
1187 | gfs2_glock_dq_uninit(&mount_gh); | |
1188 | ||
1189 | return 0; | |
1190 | ||
a91ea69f | 1191 | fail_threads: |
b3b94faa | 1192 | init_threads(sdp, UNDO); |
a91ea69f | 1193 | fail_per_node: |
b3b94faa | 1194 | init_per_node(sdp, UNDO); |
a91ea69f | 1195 | fail_inodes: |
b3b94faa | 1196 | init_inodes(sdp, UNDO); |
a91ea69f | 1197 | fail_sb: |
9b8df98f SW |
1198 | if (sdp->sd_root_dir) |
1199 | dput(sdp->sd_root_dir); | |
1200 | if (sdp->sd_master_dir) | |
1201 | dput(sdp->sd_master_dir); | |
1202 | sb->s_root = NULL; | |
a91ea69f | 1203 | fail_locking: |
b3b94faa | 1204 | init_locking(sdp, &mount_gh, UNDO); |
a91ea69f | 1205 | fail_lm: |
1bdad606 | 1206 | gfs2_gl_hash_clear(sdp); |
b3b94faa DT |
1207 | gfs2_lm_unmount(sdp); |
1208 | while (invalidate_inodes(sb)) | |
1209 | yield(); | |
a91ea69f | 1210 | fail_sys: |
b3b94faa | 1211 | gfs2_sys_fs_del(sdp); |
a91ea69f | 1212 | fail: |
5f882096 | 1213 | gfs2_delete_debugfs_file(sdp); |
a2c45807 | 1214 | kfree(sdp); |
5c676f6d | 1215 | sb->s_fs_info = NULL; |
b3b94faa DT |
1216 | return error; |
1217 | } | |
1218 | ||
ccd6efd0 | 1219 | static int gfs2_get_sb(struct file_system_type *fs_type, int flags, |
9b8df98f | 1220 | const char *dev_name, void *data, struct vfsmount *mnt) |
b3b94faa | 1221 | { |
9b8df98f | 1222 | return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt); |
86384605 AD |
1223 | } |
1224 | ||
9b8df98f | 1225 | static struct super_block *get_gfs2_sb(const char *dev_name) |
86384605 | 1226 | { |
9b8df98f | 1227 | struct super_block *sb; |
86384605 | 1228 | struct nameidata nd; |
86384605 | 1229 | int error; |
907b9bce | 1230 | |
86384605 AD |
1231 | error = path_lookup(dev_name, LOOKUP_FOLLOW, &nd); |
1232 | if (error) { | |
9b8df98f SW |
1233 | printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n", |
1234 | dev_name, error); | |
1235 | return NULL; | |
86384605 | 1236 | } |
9b8df98f SW |
1237 | sb = nd.path.dentry->d_inode->i_sb; |
1238 | if (sb && (sb->s_type == &gfs2_fs_type)) | |
1239 | atomic_inc(&sb->s_active); | |
1240 | else | |
1241 | sb = NULL; | |
1d957f9b | 1242 | path_put(&nd.path); |
86384605 AD |
1243 | return sb; |
1244 | } | |
1245 | ||
1246 | static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags, | |
1247 | const char *dev_name, void *data, struct vfsmount *mnt) | |
1248 | { | |
9b8df98f | 1249 | struct super_block *sb = NULL; |
86384605 | 1250 | struct gfs2_sbd *sdp; |
86384605 AD |
1251 | |
1252 | sb = get_gfs2_sb(dev_name); | |
1253 | if (!sb) { | |
1254 | printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n"); | |
9b8df98f | 1255 | return -ENOENT; |
86384605 | 1256 | } |
2d3ba1ea | 1257 | sdp = sb->s_fs_info; |
9b8df98f SW |
1258 | mnt->mnt_sb = sb; |
1259 | mnt->mnt_root = dget(sdp->sd_master_dir); | |
1260 | return 0; | |
b3b94faa DT |
1261 | } |
1262 | ||
419c93e0 | 1263 | static void gfs2_kill_sb(struct super_block *sb) |
86384605 AD |
1264 | { |
1265 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
acd2c8aa AD |
1266 | if (sdp) { |
1267 | gfs2_meta_syncfs(sdp); | |
1268 | dput(sdp->sd_root_dir); | |
1269 | dput(sdp->sd_master_dir); | |
1270 | sdp->sd_root_dir = NULL; | |
1271 | sdp->sd_master_dir = NULL; | |
1272 | } | |
9b8df98f SW |
1273 | shrink_dcache_sb(sb); |
1274 | kill_block_super(sb); | |
acd2c8aa AD |
1275 | if (sdp) |
1276 | gfs2_delete_debugfs_file(sdp); | |
86384605 AD |
1277 | } |
1278 | ||
b3b94faa DT |
1279 | struct file_system_type gfs2_fs_type = { |
1280 | .name = "gfs2", | |
1281 | .fs_flags = FS_REQUIRES_DEV, | |
1282 | .get_sb = gfs2_get_sb, | |
419c93e0 SW |
1283 | .kill_sb = gfs2_kill_sb, |
1284 | .owner = THIS_MODULE, | |
1285 | }; | |
1286 | ||
1287 | struct file_system_type gfs2meta_fs_type = { | |
1288 | .name = "gfs2meta", | |
1289 | .fs_flags = FS_REQUIRES_DEV, | |
86384605 | 1290 | .get_sb = gfs2_get_sb_meta, |
b3b94faa DT |
1291 | .owner = THIS_MODULE, |
1292 | }; | |
1293 |