]>
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> |
cc632e7f | 20 | #include <linux/quotaops.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 | /** |
42 | * gfs2_tune_init - Fill a gfs2_tune structure with default values | |
43 | * @gt: tune | |
44 | * | |
45 | */ | |
46 | ||
47 | static void gfs2_tune_init(struct gfs2_tune *gt) | |
48 | { | |
49 | spin_lock_init(>->gt_spin); | |
50 | ||
9b8df98f SW |
51 | gt->gt_quota_simul_sync = 64; |
52 | gt->gt_quota_warn_period = 10; | |
53 | gt->gt_quota_scale_num = 1; | |
54 | gt->gt_quota_scale_den = 1; | |
9b8df98f SW |
55 | gt->gt_new_files_jdata = 0; |
56 | gt->gt_max_readahead = 1 << 18; | |
9b8df98f | 57 | gt->gt_complain_secs = 10; |
9b8df98f SW |
58 | } |
59 | ||
b3b94faa DT |
60 | static struct gfs2_sbd *init_sbd(struct super_block *sb) |
61 | { | |
62 | struct gfs2_sbd *sdp; | |
b3b94faa | 63 | |
85d1da67 | 64 | sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL); |
b3b94faa DT |
65 | if (!sdp) |
66 | return NULL; | |
67 | ||
5c676f6d | 68 | sb->s_fs_info = sdp; |
b3b94faa | 69 | sdp->sd_vfs = sb; |
ba6e9364 | 70 | set_bit(SDF_NOJOURNALID, &sdp->sd_flags); |
b3b94faa DT |
71 | gfs2_tune_init(&sdp->sd_tune); |
72 | ||
e402746a SW |
73 | init_waitqueue_head(&sdp->sd_glock_wait); |
74 | atomic_set(&sdp->sd_glock_disposal, 0); | |
b3b94faa | 75 | spin_lock_init(&sdp->sd_statfs_spin); |
b3b94faa DT |
76 | |
77 | spin_lock_init(&sdp->sd_rindex_spin); | |
f55ab26a | 78 | mutex_init(&sdp->sd_rindex_mutex); |
b3b94faa DT |
79 | INIT_LIST_HEAD(&sdp->sd_rindex_list); |
80 | INIT_LIST_HEAD(&sdp->sd_rindex_mru_list); | |
b3b94faa DT |
81 | |
82 | INIT_LIST_HEAD(&sdp->sd_jindex_list); | |
83 | spin_lock_init(&sdp->sd_jindex_spin); | |
f55ab26a | 84 | mutex_init(&sdp->sd_jindex_mutex); |
b3b94faa | 85 | |
b3b94faa | 86 | INIT_LIST_HEAD(&sdp->sd_quota_list); |
f55ab26a | 87 | mutex_init(&sdp->sd_quota_mutex); |
37b2c837 | 88 | init_waitqueue_head(&sdp->sd_quota_wait); |
813e0c46 SW |
89 | INIT_LIST_HEAD(&sdp->sd_trunc_list); |
90 | spin_lock_init(&sdp->sd_trunc_lock); | |
b3b94faa DT |
91 | |
92 | spin_lock_init(&sdp->sd_log_lock); | |
5e687eac | 93 | atomic_set(&sdp->sd_log_pinned, 0); |
b3b94faa DT |
94 | INIT_LIST_HEAD(&sdp->sd_log_le_buf); |
95 | INIT_LIST_HEAD(&sdp->sd_log_le_revoke); | |
96 | INIT_LIST_HEAD(&sdp->sd_log_le_rg); | |
97 | INIT_LIST_HEAD(&sdp->sd_log_le_databuf); | |
d7b616e2 | 98 | INIT_LIST_HEAD(&sdp->sd_log_le_ordered); |
b3b94faa | 99 | |
5e687eac BM |
100 | init_waitqueue_head(&sdp->sd_log_waitq); |
101 | init_waitqueue_head(&sdp->sd_logd_waitq); | |
b3b94faa DT |
102 | INIT_LIST_HEAD(&sdp->sd_ail1_list); |
103 | INIT_LIST_HEAD(&sdp->sd_ail2_list); | |
104 | ||
484adff8 | 105 | init_rwsem(&sdp->sd_log_flush_lock); |
16615be1 SW |
106 | atomic_set(&sdp->sd_log_in_flight, 0); |
107 | init_waitqueue_head(&sdp->sd_log_flush_wait); | |
b3b94faa DT |
108 | |
109 | INIT_LIST_HEAD(&sdp->sd_revoke_list); | |
110 | ||
f55ab26a | 111 | mutex_init(&sdp->sd_freeze_lock); |
b3b94faa DT |
112 | |
113 | return sdp; | |
114 | } | |
115 | ||
b3b94faa | 116 | |
9b8df98f SW |
117 | /** |
118 | * gfs2_check_sb - Check superblock | |
119 | * @sdp: the filesystem | |
120 | * @sb: The superblock | |
121 | * @silent: Don't print a message if the check fails | |
122 | * | |
123 | * Checks the version code of the FS is one that we understand how to | |
124 | * read and that the sizes of the various on-disk structures have not | |
125 | * changed. | |
126 | */ | |
127 | ||
128 | static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent) | |
129 | { | |
9b8df98f SW |
130 | if (sb->sb_magic != GFS2_MAGIC || |
131 | sb->sb_type != GFS2_METATYPE_SB) { | |
132 | if (!silent) | |
133 | printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n"); | |
134 | return -EINVAL; | |
135 | } | |
136 | ||
137 | /* If format numbers match exactly, we're done. */ | |
138 | ||
139 | if (sb->sb_fs_format == GFS2_FORMAT_FS && | |
140 | sb->sb_multihost_format == GFS2_FORMAT_MULTI) | |
141 | return 0; | |
142 | ||
c80dbb58 | 143 | fs_warn(sdp, "Unknown on-disk format, unable to mount\n"); |
9b8df98f | 144 | |
c80dbb58 | 145 | return -EINVAL; |
9b8df98f SW |
146 | } |
147 | ||
148 | static void end_bio_io_page(struct bio *bio, int error) | |
149 | { | |
150 | struct page *page = bio->bi_private; | |
151 | ||
152 | if (!error) | |
153 | SetPageUptodate(page); | |
154 | else | |
155 | printk(KERN_WARNING "gfs2: error %d reading superblock\n", error); | |
156 | unlock_page(page); | |
157 | } | |
158 | ||
159 | static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf) | |
160 | { | |
161 | const struct gfs2_sb *str = buf; | |
162 | ||
163 | sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic); | |
164 | sb->sb_type = be32_to_cpu(str->sb_header.mh_type); | |
165 | sb->sb_format = be32_to_cpu(str->sb_header.mh_format); | |
166 | sb->sb_fs_format = be32_to_cpu(str->sb_fs_format); | |
167 | sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format); | |
168 | sb->sb_bsize = be32_to_cpu(str->sb_bsize); | |
169 | sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift); | |
170 | sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr); | |
171 | sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino); | |
172 | sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr); | |
173 | sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino); | |
174 | ||
175 | memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN); | |
176 | memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN); | |
02e3cc70 | 177 | memcpy(sb->sb_uuid, str->sb_uuid, 16); |
9b8df98f SW |
178 | } |
179 | ||
180 | /** | |
181 | * gfs2_read_super - Read the gfs2 super block from disk | |
182 | * @sdp: The GFS2 super block | |
183 | * @sector: The location of the super block | |
184 | * @error: The error code to return | |
185 | * | |
186 | * This uses the bio functions to read the super block from disk | |
187 | * because we want to be 100% sure that we never read cached data. | |
188 | * A super block is read twice only during each GFS2 mount and is | |
189 | * never written to by the filesystem. The first time its read no | |
190 | * locks are held, and the only details which are looked at are those | |
191 | * relating to the locking protocol. Once locking is up and working, | |
192 | * the sb is read again under the lock to establish the location of | |
193 | * the master directory (contains pointers to journals etc) and the | |
194 | * root directory. | |
195 | * | |
196 | * Returns: 0 on success or error | |
197 | */ | |
198 | ||
199 | static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector) | |
200 | { | |
201 | struct super_block *sb = sdp->sd_vfs; | |
202 | struct gfs2_sb *p; | |
203 | struct page *page; | |
204 | struct bio *bio; | |
205 | ||
206 | page = alloc_page(GFP_NOFS); | |
207 | if (unlikely(!page)) | |
208 | return -ENOBUFS; | |
209 | ||
210 | ClearPageUptodate(page); | |
211 | ClearPageDirty(page); | |
212 | lock_page(page); | |
213 | ||
214 | bio = bio_alloc(GFP_NOFS, 1); | |
9b8df98f SW |
215 | bio->bi_sector = sector * (sb->s_blocksize >> 9); |
216 | bio->bi_bdev = sb->s_bdev; | |
217 | bio_add_page(bio, page, PAGE_SIZE, 0); | |
218 | ||
219 | bio->bi_end_io = end_bio_io_page; | |
220 | bio->bi_private = page; | |
7b6d91da | 221 | submit_bio(READ_SYNC | REQ_META, bio); |
9b8df98f SW |
222 | wait_on_page_locked(page); |
223 | bio_put(bio); | |
224 | if (!PageUptodate(page)) { | |
225 | __free_page(page); | |
226 | return -EIO; | |
227 | } | |
228 | p = kmap(page); | |
229 | gfs2_sb_in(&sdp->sd_sb, p); | |
230 | kunmap(page); | |
231 | __free_page(page); | |
232 | return 0; | |
233 | } | |
6bac243f | 234 | |
9b8df98f SW |
235 | /** |
236 | * gfs2_read_sb - Read super block | |
237 | * @sdp: The GFS2 superblock | |
9b8df98f SW |
238 | * @silent: Don't print message if mount fails |
239 | * | |
240 | */ | |
241 | ||
6bac243f | 242 | static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent) |
9b8df98f SW |
243 | { |
244 | u32 hash_blocks, ind_blocks, leaf_blocks; | |
245 | u32 tmp_blocks; | |
246 | unsigned int x; | |
247 | int error; | |
248 | ||
249 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); | |
250 | if (error) { | |
251 | if (!silent) | |
252 | fs_err(sdp, "can't read superblock\n"); | |
253 | return error; | |
254 | } | |
255 | ||
256 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
257 | if (error) | |
258 | return error; | |
259 | ||
260 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
261 | GFS2_BASIC_BLOCK_SHIFT; | |
262 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
263 | sdp->sd_diptrs = (sdp->sd_sb.sb_bsize - | |
264 | sizeof(struct gfs2_dinode)) / sizeof(u64); | |
265 | sdp->sd_inptrs = (sdp->sd_sb.sb_bsize - | |
266 | sizeof(struct gfs2_meta_header)) / sizeof(u64); | |
267 | sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header); | |
268 | sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2; | |
269 | sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1; | |
270 | sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64); | |
271 | sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize - | |
272 | sizeof(struct gfs2_meta_header)) / | |
273 | sizeof(struct gfs2_quota_change); | |
274 | ||
275 | /* Compute maximum reservation required to add a entry to a directory */ | |
276 | ||
277 | hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH), | |
278 | sdp->sd_jbsize); | |
279 | ||
280 | ind_blocks = 0; | |
281 | for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) { | |
282 | tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs); | |
283 | ind_blocks += tmp_blocks; | |
284 | } | |
285 | ||
286 | leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH; | |
287 | ||
288 | sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks; | |
289 | ||
290 | sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize - | |
291 | sizeof(struct gfs2_dinode); | |
292 | sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs; | |
293 | for (x = 2;; x++) { | |
294 | u64 space, d; | |
295 | u32 m; | |
296 | ||
297 | space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs; | |
298 | d = space; | |
299 | m = do_div(d, sdp->sd_inptrs); | |
300 | ||
301 | if (d != sdp->sd_heightsize[x - 1] || m) | |
302 | break; | |
303 | sdp->sd_heightsize[x] = space; | |
304 | } | |
305 | sdp->sd_max_height = x; | |
306 | sdp->sd_heightsize[x] = ~0; | |
307 | gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT); | |
308 | ||
309 | sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize - | |
310 | sizeof(struct gfs2_dinode); | |
311 | sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs; | |
312 | for (x = 2;; x++) { | |
313 | u64 space, d; | |
314 | u32 m; | |
315 | ||
316 | space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs; | |
317 | d = space; | |
318 | m = do_div(d, sdp->sd_inptrs); | |
319 | ||
320 | if (d != sdp->sd_jheightsize[x - 1] || m) | |
321 | break; | |
322 | sdp->sd_jheightsize[x] = space; | |
323 | } | |
324 | sdp->sd_max_jheight = x; | |
325 | sdp->sd_jheightsize[x] = ~0; | |
326 | gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT); | |
327 | ||
328 | return 0; | |
329 | } | |
330 | ||
b3b94faa DT |
331 | static int init_names(struct gfs2_sbd *sdp, int silent) |
332 | { | |
b3b94faa DT |
333 | char *proto, *table; |
334 | int error = 0; | |
335 | ||
336 | proto = sdp->sd_args.ar_lockproto; | |
337 | table = sdp->sd_args.ar_locktable; | |
338 | ||
339 | /* Try to autodetect */ | |
340 | ||
341 | if (!proto[0] || !table[0]) { | |
bb8d8a6f SW |
342 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift); |
343 | if (error) | |
344 | return error; | |
3cf1e7be SW |
345 | |
346 | error = gfs2_check_sb(sdp, &sdp->sd_sb, silent); | |
b3b94faa DT |
347 | if (error) |
348 | goto out; | |
349 | ||
350 | if (!proto[0]) | |
3cf1e7be | 351 | proto = sdp->sd_sb.sb_lockproto; |
b3b94faa | 352 | if (!table[0]) |
3cf1e7be | 353 | table = sdp->sd_sb.sb_locktable; |
b3b94faa DT |
354 | } |
355 | ||
356 | if (!table[0]) | |
357 | table = sdp->sd_vfs->s_id; | |
358 | ||
00377d8e JD |
359 | strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN); |
360 | strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN); | |
b3b94faa | 361 | |
5d35e31f DC |
362 | table = sdp->sd_table_name; |
363 | while ((table = strchr(table, '/'))) | |
b35997d4 RP |
364 | *table = '_'; |
365 | ||
a91ea69f | 366 | out: |
b3b94faa DT |
367 | return error; |
368 | } | |
369 | ||
370 | static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh, | |
371 | int undo) | |
372 | { | |
b3b94faa DT |
373 | int error = 0; |
374 | ||
375 | if (undo) | |
376 | goto fail_trans; | |
377 | ||
b3b94faa DT |
378 | error = gfs2_glock_nq_num(sdp, |
379 | GFS2_MOUNT_LOCK, &gfs2_nondisk_glops, | |
380 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE, | |
381 | mount_gh); | |
382 | if (error) { | |
383 | fs_err(sdp, "can't acquire mount glock: %d\n", error); | |
384 | goto fail; | |
385 | } | |
386 | ||
387 | error = gfs2_glock_nq_num(sdp, | |
388 | GFS2_LIVE_LOCK, &gfs2_nondisk_glops, | |
389 | LM_ST_SHARED, | |
579b78a4 | 390 | LM_FLAG_NOEXP | GL_EXACT, |
b3b94faa DT |
391 | &sdp->sd_live_gh); |
392 | if (error) { | |
393 | fs_err(sdp, "can't acquire live glock: %d\n", error); | |
394 | goto fail_mount; | |
395 | } | |
396 | ||
397 | error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops, | |
398 | CREATE, &sdp->sd_rename_gl); | |
399 | if (error) { | |
400 | fs_err(sdp, "can't create rename glock: %d\n", error); | |
401 | goto fail_live; | |
402 | } | |
403 | ||
404 | error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops, | |
405 | CREATE, &sdp->sd_trans_gl); | |
406 | if (error) { | |
407 | fs_err(sdp, "can't create transaction glock: %d\n", error); | |
408 | goto fail_rename; | |
409 | } | |
b3b94faa DT |
410 | |
411 | return 0; | |
412 | ||
feaa7bba | 413 | fail_trans: |
b3b94faa | 414 | gfs2_glock_put(sdp->sd_trans_gl); |
feaa7bba | 415 | fail_rename: |
b3b94faa | 416 | gfs2_glock_put(sdp->sd_rename_gl); |
feaa7bba | 417 | fail_live: |
b3b94faa | 418 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); |
feaa7bba | 419 | fail_mount: |
b3b94faa | 420 | gfs2_glock_dq_uninit(mount_gh); |
feaa7bba | 421 | fail: |
b3b94faa DT |
422 | return error; |
423 | } | |
424 | ||
9b8df98f SW |
425 | static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr, |
426 | u64 no_addr, const char *name) | |
f42faf4f | 427 | { |
9b8df98f SW |
428 | struct gfs2_sbd *sdp = sb->s_fs_info; |
429 | struct dentry *dentry; | |
430 | struct inode *inode; | |
431 | ||
1a0eae88 | 432 | inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0); |
9b8df98f SW |
433 | if (IS_ERR(inode)) { |
434 | fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode)); | |
435 | return PTR_ERR(inode); | |
436 | } | |
437 | dentry = d_alloc_root(inode); | |
438 | if (!dentry) { | |
439 | fs_err(sdp, "can't alloc %s dentry\n", name); | |
440 | iput(inode); | |
441 | return -ENOMEM; | |
442 | } | |
443 | dentry->d_op = &gfs2_dops; | |
444 | *dptr = dentry; | |
445 | return 0; | |
f42faf4f SW |
446 | } |
447 | ||
9b8df98f | 448 | static int init_sb(struct gfs2_sbd *sdp, int silent) |
b3b94faa DT |
449 | { |
450 | struct super_block *sb = sdp->sd_vfs; | |
451 | struct gfs2_holder sb_gh; | |
dbb7cae2 | 452 | u64 no_addr; |
9b8df98f | 453 | int ret; |
b3b94faa | 454 | |
9b8df98f SW |
455 | ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops, |
456 | LM_ST_SHARED, 0, &sb_gh); | |
457 | if (ret) { | |
458 | fs_err(sdp, "can't acquire superblock glock: %d\n", ret); | |
459 | return ret; | |
b3b94faa | 460 | } |
907b9bce | 461 | |
6bac243f | 462 | ret = gfs2_read_sb(sdp, silent); |
9b8df98f SW |
463 | if (ret) { |
464 | fs_err(sdp, "can't read superblock: %d\n", ret); | |
b3b94faa DT |
465 | goto out; |
466 | } | |
467 | ||
468 | /* Set up the buffer cache and SB for real */ | |
e1defc4f | 469 | if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) { |
9b8df98f | 470 | ret = -EINVAL; |
b3b94faa DT |
471 | fs_err(sdp, "FS block size (%u) is too small for device " |
472 | "block size (%u)\n", | |
e1defc4f | 473 | sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev)); |
b3b94faa DT |
474 | goto out; |
475 | } | |
476 | if (sdp->sd_sb.sb_bsize > PAGE_SIZE) { | |
9b8df98f | 477 | ret = -EINVAL; |
b3b94faa DT |
478 | fs_err(sdp, "FS block size (%u) is too big for machine " |
479 | "page size (%u)\n", | |
480 | sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE); | |
481 | goto out; | |
482 | } | |
b3b94faa DT |
483 | sb_set_blocksize(sb, sdp->sd_sb.sb_bsize); |
484 | ||
f42faf4f | 485 | /* Get the root inode */ |
dbb7cae2 | 486 | no_addr = sdp->sd_sb.sb_root_dir.no_addr; |
9b8df98f SW |
487 | ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root"); |
488 | if (ret) | |
f42faf4f | 489 | goto out; |
b3b94faa | 490 | |
9b8df98f SW |
491 | /* Get the master inode */ |
492 | no_addr = sdp->sd_sb.sb_master_dir.no_addr; | |
493 | ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master"); | |
494 | if (ret) { | |
495 | dput(sdp->sd_root_dir); | |
496 | goto out; | |
497 | } | |
498 | sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir); | |
f42faf4f | 499 | out: |
b3b94faa | 500 | gfs2_glock_dq_uninit(&sb_gh); |
9b8df98f | 501 | return ret; |
b3b94faa DT |
502 | } |
503 | ||
da6dd40d BP |
504 | /** |
505 | * map_journal_extents - create a reusable "extent" mapping from all logical | |
506 | * blocks to all physical blocks for the given journal. This will save | |
507 | * us time when writing journal blocks. Most journals will have only one | |
508 | * extent that maps all their logical blocks. That's because gfs2.mkfs | |
509 | * arranges the journal blocks sequentially to maximize performance. | |
510 | * So the extent would map the first block for the entire file length. | |
511 | * However, gfs2_jadd can happen while file activity is happening, so | |
512 | * those journals may not be sequential. Less likely is the case where | |
513 | * the users created their own journals by mounting the metafs and | |
514 | * laying it out. But it's still possible. These journals might have | |
515 | * several extents. | |
516 | * | |
517 | * TODO: This should be done in bigger chunks rather than one block at a time, | |
518 | * but since it's only done at mount time, I'm not worried about the | |
519 | * time it takes. | |
520 | */ | |
521 | static int map_journal_extents(struct gfs2_sbd *sdp) | |
522 | { | |
523 | struct gfs2_jdesc *jd = sdp->sd_jdesc; | |
524 | unsigned int lb; | |
525 | u64 db, prev_db; /* logical block, disk block, prev disk block */ | |
526 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); | |
527 | struct gfs2_journal_extent *jext = NULL; | |
528 | struct buffer_head bh; | |
529 | int rc = 0; | |
530 | ||
da6dd40d BP |
531 | prev_db = 0; |
532 | ||
a2e0f799 | 533 | for (lb = 0; lb < i_size_read(jd->jd_inode) >> sdp->sd_sb.sb_bsize_shift; lb++) { |
da6dd40d BP |
534 | bh.b_state = 0; |
535 | bh.b_blocknr = 0; | |
536 | bh.b_size = 1 << ip->i_inode.i_blkbits; | |
537 | rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0); | |
538 | db = bh.b_blocknr; | |
539 | if (rc || !db) { | |
540 | printk(KERN_INFO "GFS2 journal mapping error %d: lb=" | |
541 | "%u db=%llu\n", rc, lb, (unsigned long long)db); | |
542 | break; | |
543 | } | |
544 | if (!prev_db || db != prev_db + 1) { | |
545 | jext = kzalloc(sizeof(struct gfs2_journal_extent), | |
546 | GFP_KERNEL); | |
547 | if (!jext) { | |
548 | printk(KERN_INFO "GFS2 error: out of memory " | |
549 | "mapping journal extents.\n"); | |
550 | rc = -ENOMEM; | |
551 | break; | |
552 | } | |
553 | jext->dblock = db; | |
554 | jext->lblock = lb; | |
555 | jext->blocks = 1; | |
556 | list_add_tail(&jext->extent_list, &jd->extent_list); | |
557 | } else { | |
558 | jext->blocks++; | |
559 | } | |
560 | prev_db = db; | |
561 | } | |
562 | return rc; | |
563 | } | |
564 | ||
f057f6cd | 565 | static void gfs2_others_may_mount(struct gfs2_sbd *sdp) |
da755fdb | 566 | { |
f057f6cd SW |
567 | char *message = "FIRSTMOUNT=Done"; |
568 | char *envp[] = { message, NULL }; | |
569 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
570 | ls->ls_first_done = 1; | |
571 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp); | |
da755fdb SW |
572 | } |
573 | ||
b5289681 SW |
574 | /** |
575 | * gfs2_jindex_hold - Grab a lock on the jindex | |
576 | * @sdp: The GFS2 superblock | |
577 | * @ji_gh: the holder for the jindex glock | |
578 | * | |
579 | * Returns: errno | |
580 | */ | |
581 | ||
582 | static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh) | |
583 | { | |
584 | struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex); | |
585 | struct qstr name; | |
586 | char buf[20]; | |
587 | struct gfs2_jdesc *jd; | |
588 | int error; | |
589 | ||
590 | name.name = buf; | |
591 | ||
592 | mutex_lock(&sdp->sd_jindex_mutex); | |
593 | ||
594 | for (;;) { | |
595 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh); | |
596 | if (error) | |
597 | break; | |
598 | ||
599 | name.len = sprintf(buf, "journal%u", sdp->sd_journals); | |
600 | name.hash = gfs2_disk_hash(name.name, name.len); | |
601 | ||
602 | error = gfs2_dir_check(sdp->sd_jindex, &name, NULL); | |
603 | if (error == -ENOENT) { | |
604 | error = 0; | |
605 | break; | |
606 | } | |
607 | ||
608 | gfs2_glock_dq_uninit(ji_gh); | |
609 | ||
610 | if (error) | |
611 | break; | |
612 | ||
613 | error = -ENOMEM; | |
614 | jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL); | |
615 | if (!jd) | |
616 | break; | |
617 | ||
618 | INIT_LIST_HEAD(&jd->extent_list); | |
6ecd7c2d | 619 | INIT_WORK(&jd->jd_work, gfs2_recover_func); |
b5289681 SW |
620 | jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1); |
621 | if (!jd->jd_inode || IS_ERR(jd->jd_inode)) { | |
622 | if (!jd->jd_inode) | |
623 | error = -ENOENT; | |
624 | else | |
625 | error = PTR_ERR(jd->jd_inode); | |
626 | kfree(jd); | |
627 | break; | |
628 | } | |
629 | ||
630 | spin_lock(&sdp->sd_jindex_spin); | |
631 | jd->jd_jid = sdp->sd_journals++; | |
632 | list_add_tail(&jd->jd_list, &sdp->sd_jindex_list); | |
633 | spin_unlock(&sdp->sd_jindex_spin); | |
634 | } | |
635 | ||
636 | mutex_unlock(&sdp->sd_jindex_mutex); | |
637 | ||
638 | return error; | |
639 | } | |
640 | ||
b3b94faa DT |
641 | static int init_journal(struct gfs2_sbd *sdp, int undo) |
642 | { | |
9b8df98f | 643 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa | 644 | struct gfs2_holder ji_gh; |
5c676f6d | 645 | struct gfs2_inode *ip; |
b3b94faa DT |
646 | int jindex = 1; |
647 | int error = 0; | |
648 | ||
649 | if (undo) { | |
650 | jindex = 0; | |
fe64d517 | 651 | goto fail_jinode_gh; |
b3b94faa DT |
652 | } |
653 | ||
9b8df98f | 654 | sdp->sd_jindex = gfs2_lookup_simple(master, "jindex"); |
c752666c | 655 | if (IS_ERR(sdp->sd_jindex)) { |
b3b94faa | 656 | fs_err(sdp, "can't lookup journal index: %d\n", error); |
c752666c | 657 | return PTR_ERR(sdp->sd_jindex); |
b3b94faa | 658 | } |
feaa7bba | 659 | ip = GFS2_I(sdp->sd_jindex); |
b3b94faa DT |
660 | |
661 | /* Load in the journal index special file */ | |
662 | ||
663 | error = gfs2_jindex_hold(sdp, &ji_gh); | |
664 | if (error) { | |
665 | fs_err(sdp, "can't read journal index: %d\n", error); | |
666 | goto fail; | |
667 | } | |
668 | ||
0e5a9fb0 | 669 | error = -EUSERS; |
b3b94faa DT |
670 | if (!gfs2_jindex_size(sdp)) { |
671 | fs_err(sdp, "no journals!\n"); | |
907b9bce | 672 | goto fail_jindex; |
b3b94faa DT |
673 | } |
674 | ||
675 | if (sdp->sd_args.ar_spectator) { | |
676 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0); | |
fd041f0b | 677 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
5e687eac BM |
678 | atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5); |
679 | atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5); | |
b3b94faa DT |
680 | } else { |
681 | if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) { | |
682 | fs_err(sdp, "can't mount journal #%u\n", | |
683 | sdp->sd_lockstruct.ls_jid); | |
684 | fs_err(sdp, "there are only %u journals (0 - %u)\n", | |
685 | gfs2_jindex_size(sdp), | |
686 | gfs2_jindex_size(sdp) - 1); | |
687 | goto fail_jindex; | |
688 | } | |
689 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid); | |
690 | ||
feaa7bba | 691 | error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid, |
b3b94faa DT |
692 | &gfs2_journal_glops, |
693 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP, | |
694 | &sdp->sd_journal_gh); | |
695 | if (error) { | |
696 | fs_err(sdp, "can't acquire journal glock: %d\n", error); | |
697 | goto fail_jindex; | |
698 | } | |
699 | ||
feaa7bba SW |
700 | ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
701 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, | |
75be73a8 | 702 | LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE, |
b3b94faa DT |
703 | &sdp->sd_jinode_gh); |
704 | if (error) { | |
705 | fs_err(sdp, "can't acquire journal inode glock: %d\n", | |
706 | error); | |
707 | goto fail_journal_gh; | |
708 | } | |
709 | ||
710 | error = gfs2_jdesc_check(sdp->sd_jdesc); | |
711 | if (error) { | |
712 | fs_err(sdp, "my journal (%u) is bad: %d\n", | |
713 | sdp->sd_jdesc->jd_jid, error); | |
714 | goto fail_jinode_gh; | |
715 | } | |
fd041f0b | 716 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
5e687eac BM |
717 | atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5); |
718 | atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5); | |
da6dd40d BP |
719 | |
720 | /* Map the extents for this journal's blocks */ | |
721 | map_journal_extents(sdp); | |
b3b94faa | 722 | } |
63997775 | 723 | trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free)); |
b3b94faa DT |
724 | |
725 | if (sdp->sd_lockstruct.ls_first) { | |
726 | unsigned int x; | |
727 | for (x = 0; x < sdp->sd_journals; x++) { | |
6ecd7c2d TH |
728 | error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x), |
729 | true); | |
b3b94faa DT |
730 | if (error) { |
731 | fs_err(sdp, "error recovering journal %u: %d\n", | |
732 | x, error); | |
733 | goto fail_jinode_gh; | |
734 | } | |
735 | } | |
736 | ||
f057f6cd | 737 | gfs2_others_may_mount(sdp); |
b3b94faa | 738 | } else if (!sdp->sd_args.ar_spectator) { |
6ecd7c2d | 739 | error = gfs2_recover_journal(sdp->sd_jdesc, true); |
b3b94faa DT |
740 | if (error) { |
741 | fs_err(sdp, "error recovering my journal: %d\n", error); | |
742 | goto fail_jinode_gh; | |
743 | } | |
744 | } | |
745 | ||
746 | set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags); | |
747 | gfs2_glock_dq_uninit(&ji_gh); | |
748 | jindex = 0; | |
749 | ||
b3b94faa DT |
750 | return 0; |
751 | ||
a91ea69f | 752 | fail_jinode_gh: |
b3b94faa DT |
753 | if (!sdp->sd_args.ar_spectator) |
754 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
a91ea69f | 755 | fail_journal_gh: |
b3b94faa DT |
756 | if (!sdp->sd_args.ar_spectator) |
757 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
a91ea69f | 758 | fail_jindex: |
b3b94faa DT |
759 | gfs2_jindex_free(sdp); |
760 | if (jindex) | |
761 | gfs2_glock_dq_uninit(&ji_gh); | |
a91ea69f | 762 | fail: |
f42faf4f | 763 | iput(sdp->sd_jindex); |
b3b94faa DT |
764 | return error; |
765 | } | |
766 | ||
b3b94faa DT |
767 | |
768 | static int init_inodes(struct gfs2_sbd *sdp, int undo) | |
769 | { | |
b3b94faa | 770 | int error = 0; |
5c676f6d | 771 | struct gfs2_inode *ip; |
9b8df98f | 772 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
773 | |
774 | if (undo) | |
f42faf4f SW |
775 | goto fail_qinode; |
776 | ||
f42faf4f SW |
777 | error = init_journal(sdp, undo); |
778 | if (error) | |
9b8df98f | 779 | goto fail; |
b3b94faa | 780 | |
b3b94faa | 781 | /* Read in the master statfs inode */ |
9b8df98f | 782 | sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs"); |
c752666c SW |
783 | if (IS_ERR(sdp->sd_statfs_inode)) { |
784 | error = PTR_ERR(sdp->sd_statfs_inode); | |
b3b94faa | 785 | fs_err(sdp, "can't read in statfs inode: %d\n", error); |
8d8291ae | 786 | goto fail_journal; |
b3b94faa DT |
787 | } |
788 | ||
789 | /* Read in the resource index inode */ | |
9b8df98f | 790 | sdp->sd_rindex = gfs2_lookup_simple(master, "rindex"); |
c752666c SW |
791 | if (IS_ERR(sdp->sd_rindex)) { |
792 | error = PTR_ERR(sdp->sd_rindex); | |
b3b94faa DT |
793 | fs_err(sdp, "can't get resource index inode: %d\n", error); |
794 | goto fail_statfs; | |
795 | } | |
feaa7bba | 796 | ip = GFS2_I(sdp->sd_rindex); |
cf45b752 | 797 | sdp->sd_rindex_uptodate = 0; |
b3b94faa DT |
798 | |
799 | /* Read in the quota inode */ | |
9b8df98f | 800 | sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota"); |
c752666c SW |
801 | if (IS_ERR(sdp->sd_quota_inode)) { |
802 | error = PTR_ERR(sdp->sd_quota_inode); | |
b3b94faa DT |
803 | fs_err(sdp, "can't get quota file inode: %d\n", error); |
804 | goto fail_rindex; | |
805 | } | |
b3b94faa DT |
806 | return 0; |
807 | ||
f42faf4f SW |
808 | fail_qinode: |
809 | iput(sdp->sd_quota_inode); | |
f42faf4f | 810 | fail_rindex: |
b3b94faa | 811 | gfs2_clear_rgrpd(sdp); |
f42faf4f | 812 | iput(sdp->sd_rindex); |
f42faf4f SW |
813 | fail_statfs: |
814 | iput(sdp->sd_statfs_inode); | |
f42faf4f SW |
815 | fail_journal: |
816 | init_journal(sdp, UNDO); | |
f42faf4f | 817 | fail: |
b3b94faa DT |
818 | return error; |
819 | } | |
820 | ||
821 | static int init_per_node(struct gfs2_sbd *sdp, int undo) | |
822 | { | |
f42faf4f | 823 | struct inode *pn = NULL; |
b3b94faa DT |
824 | char buf[30]; |
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 (sdp->sd_args.ar_spectator) | |
830 | return 0; | |
831 | ||
832 | if (undo) | |
833 | goto fail_qc_gh; | |
834 | ||
9b8df98f | 835 | pn = gfs2_lookup_simple(master, "per_node"); |
c752666c SW |
836 | if (IS_ERR(pn)) { |
837 | error = PTR_ERR(pn); | |
b3b94faa DT |
838 | fs_err(sdp, "can't find per_node directory: %d\n", error); |
839 | return error; | |
840 | } | |
841 | ||
b3b94faa | 842 | sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
843 | sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf); |
844 | if (IS_ERR(sdp->sd_sc_inode)) { | |
845 | error = PTR_ERR(sdp->sd_sc_inode); | |
b3b94faa | 846 | fs_err(sdp, "can't find local \"sc\" file: %d\n", error); |
8d8291ae | 847 | goto fail; |
b3b94faa DT |
848 | } |
849 | ||
b3b94faa | 850 | sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
851 | sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf); |
852 | if (IS_ERR(sdp->sd_qc_inode)) { | |
853 | error = PTR_ERR(sdp->sd_qc_inode); | |
b3b94faa DT |
854 | fs_err(sdp, "can't find local \"qc\" file: %d\n", error); |
855 | goto fail_ut_i; | |
856 | } | |
857 | ||
f42faf4f | 858 | iput(pn); |
b3b94faa DT |
859 | pn = NULL; |
860 | ||
feaa7bba | 861 | ip = GFS2_I(sdp->sd_sc_inode); |
8d8291ae | 862 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
863 | &sdp->sd_sc_gh); |
864 | if (error) { | |
865 | fs_err(sdp, "can't lock local \"sc\" file: %d\n", error); | |
8d8291ae | 866 | goto fail_qc_i; |
b3b94faa DT |
867 | } |
868 | ||
feaa7bba | 869 | ip = GFS2_I(sdp->sd_qc_inode); |
8d8291ae | 870 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
871 | &sdp->sd_qc_gh); |
872 | if (error) { | |
873 | fs_err(sdp, "can't lock local \"qc\" file: %d\n", error); | |
874 | goto fail_ut_gh; | |
875 | } | |
876 | ||
877 | return 0; | |
878 | ||
a91ea69f | 879 | fail_qc_gh: |
b3b94faa | 880 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); |
a91ea69f | 881 | fail_ut_gh: |
b3b94faa | 882 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); |
a91ea69f | 883 | fail_qc_i: |
f42faf4f | 884 | iput(sdp->sd_qc_inode); |
a91ea69f | 885 | fail_ut_i: |
f42faf4f | 886 | iput(sdp->sd_sc_inode); |
a91ea69f | 887 | fail: |
b3b94faa | 888 | if (pn) |
f42faf4f | 889 | iput(pn); |
b3b94faa DT |
890 | return error; |
891 | } | |
892 | ||
893 | static int init_threads(struct gfs2_sbd *sdp, int undo) | |
894 | { | |
895 | struct task_struct *p; | |
896 | int error = 0; | |
897 | ||
898 | if (undo) | |
feaa7bba | 899 | goto fail_quotad; |
b3b94faa | 900 | |
b3b94faa DT |
901 | p = kthread_run(gfs2_logd, sdp, "gfs2_logd"); |
902 | error = IS_ERR(p); | |
903 | if (error) { | |
904 | fs_err(sdp, "can't start logd thread: %d\n", error); | |
905 | return error; | |
906 | } | |
907 | sdp->sd_logd_process = p; | |
908 | ||
b3b94faa DT |
909 | p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad"); |
910 | error = IS_ERR(p); | |
911 | if (error) { | |
912 | fs_err(sdp, "can't start quotad thread: %d\n", error); | |
913 | goto fail; | |
914 | } | |
915 | sdp->sd_quotad_process = p; | |
916 | ||
b3b94faa DT |
917 | return 0; |
918 | ||
b3b94faa | 919 | |
feaa7bba | 920 | fail_quotad: |
b3b94faa | 921 | kthread_stop(sdp->sd_quotad_process); |
feaa7bba | 922 | fail: |
b3b94faa | 923 | kthread_stop(sdp->sd_logd_process); |
b3b94faa DT |
924 | return error; |
925 | } | |
da755fdb | 926 | |
f057f6cd SW |
927 | static const match_table_t nolock_tokens = { |
928 | { Opt_jid, "jid=%d\n", }, | |
929 | { Opt_err, NULL }, | |
930 | }; | |
931 | ||
8f05228e SW |
932 | static void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl) |
933 | { | |
934 | struct gfs2_sbd *sdp = gl->gl_sbd; | |
935 | kmem_cache_free(cachep, gl); | |
936 | if (atomic_dec_and_test(&sdp->sd_glock_disposal)) | |
937 | wake_up(&sdp->sd_glock_wait); | |
938 | } | |
939 | ||
f057f6cd SW |
940 | static const struct lm_lockops nolock_ops = { |
941 | .lm_proto_name = "lock_nolock", | |
8f05228e | 942 | .lm_put_lock = nolock_put_lock, |
f057f6cd SW |
943 | .lm_tokens = &nolock_tokens, |
944 | }; | |
945 | ||
da755fdb SW |
946 | /** |
947 | * gfs2_lm_mount - mount a locking protocol | |
948 | * @sdp: the filesystem | |
3ad2f3fb | 949 | * @args: mount arguments |
da755fdb SW |
950 | * @silent: if 1, don't complain if the FS isn't a GFS2 fs |
951 | * | |
952 | * Returns: errno | |
953 | */ | |
954 | ||
955 | static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent) | |
956 | { | |
f057f6cd SW |
957 | const struct lm_lockops *lm; |
958 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
959 | struct gfs2_args *args = &sdp->sd_args; | |
960 | const char *proto = sdp->sd_proto_name; | |
961 | const char *table = sdp->sd_table_name; | |
962 | const char *fsname; | |
963 | char *o, *options; | |
964 | int ret; | |
da755fdb | 965 | |
f057f6cd SW |
966 | if (!strcmp("lock_nolock", proto)) { |
967 | lm = &nolock_ops; | |
968 | sdp->sd_args.ar_localflocks = 1; | |
f057f6cd SW |
969 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM |
970 | } else if (!strcmp("lock_dlm", proto)) { | |
971 | lm = &gfs2_dlm_ops; | |
972 | #endif | |
973 | } else { | |
974 | printk(KERN_INFO "GFS2: can't find protocol %s\n", proto); | |
975 | return -ENOENT; | |
976 | } | |
da755fdb SW |
977 | |
978 | fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table); | |
979 | ||
f057f6cd SW |
980 | ls->ls_ops = lm; |
981 | ls->ls_first = 1; | |
da755fdb | 982 | |
f057f6cd SW |
983 | for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) { |
984 | substring_t tmp[MAX_OPT_ARGS]; | |
985 | int token, option; | |
986 | ||
987 | if (!o || !*o) | |
988 | continue; | |
989 | ||
990 | token = match_token(o, *lm->lm_tokens, tmp); | |
991 | switch (token) { | |
992 | case Opt_jid: | |
993 | ret = match_int(&tmp[0], &option); | |
994 | if (ret || option < 0) | |
995 | goto hostdata_error; | |
ba6e9364 SW |
996 | if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags)) |
997 | ls->ls_jid = option; | |
f057f6cd SW |
998 | break; |
999 | case Opt_id: | |
2b88f7c5 | 1000 | /* Obsolete, but left for backward compat purposes */ |
f057f6cd SW |
1001 | break; |
1002 | case Opt_first: | |
1003 | ret = match_int(&tmp[0], &option); | |
1004 | if (ret || (option != 0 && option != 1)) | |
1005 | goto hostdata_error; | |
1006 | ls->ls_first = option; | |
1007 | break; | |
1008 | case Opt_nodir: | |
1009 | ret = match_int(&tmp[0], &option); | |
1010 | if (ret || (option != 0 && option != 1)) | |
1011 | goto hostdata_error; | |
1012 | ls->ls_nodir = option; | |
1013 | break; | |
1014 | case Opt_err: | |
1015 | default: | |
1016 | hostdata_error: | |
1017 | fs_info(sdp, "unknown hostdata (%s)\n", o); | |
1018 | return -EINVAL; | |
1019 | } | |
da755fdb SW |
1020 | } |
1021 | ||
1022 | if (sdp->sd_args.ar_spectator) | |
1023 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table); | |
1024 | else | |
1025 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table, | |
1026 | sdp->sd_lockstruct.ls_jid); | |
1027 | ||
f057f6cd SW |
1028 | fsname = strchr(table, ':'); |
1029 | if (fsname) | |
1030 | fsname++; | |
1031 | if (lm->lm_mount == NULL) { | |
1032 | fs_info(sdp, "Now mounting FS...\n"); | |
1033 | return 0; | |
da755fdb | 1034 | } |
f057f6cd SW |
1035 | ret = lm->lm_mount(sdp, fsname); |
1036 | if (ret == 0) | |
1037 | fs_info(sdp, "Joined cluster. Now mounting FS...\n"); | |
1038 | return ret; | |
da755fdb SW |
1039 | } |
1040 | ||
1041 | void gfs2_lm_unmount(struct gfs2_sbd *sdp) | |
1042 | { | |
f057f6cd SW |
1043 | const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops; |
1044 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) && | |
1045 | lm->lm_unmount) | |
1046 | lm->lm_unmount(sdp); | |
da755fdb | 1047 | } |
b3b94faa | 1048 | |
ba6e9364 SW |
1049 | static int gfs2_journalid_wait(void *word) |
1050 | { | |
1051 | if (signal_pending(current)) | |
1052 | return -EINTR; | |
1053 | schedule(); | |
1054 | return 0; | |
1055 | } | |
1056 | ||
1057 | static int wait_on_journal(struct gfs2_sbd *sdp) | |
1058 | { | |
ba6e9364 SW |
1059 | if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL) |
1060 | return 0; | |
1061 | ||
1062 | return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE); | |
1063 | } | |
1064 | ||
8633ecfa SW |
1065 | void gfs2_online_uevent(struct gfs2_sbd *sdp) |
1066 | { | |
1067 | struct super_block *sb = sdp->sd_vfs; | |
1068 | char ro[20]; | |
1069 | char spectator[20]; | |
1070 | char *envp[] = { ro, spectator, NULL }; | |
1071 | sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0); | |
1072 | sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0); | |
1073 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp); | |
1074 | } | |
1075 | ||
b3b94faa DT |
1076 | /** |
1077 | * fill_super - Read in superblock | |
1078 | * @sb: The VFS superblock | |
1079 | * @data: Mount options | |
1080 | * @silent: Don't complain if it's not a GFS2 filesystem | |
1081 | * | |
1082 | * Returns: errno | |
1083 | */ | |
1084 | ||
f55073ff | 1085 | static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent) |
b3b94faa DT |
1086 | { |
1087 | struct gfs2_sbd *sdp; | |
1088 | struct gfs2_holder mount_gh; | |
1089 | int error; | |
1090 | ||
1091 | sdp = init_sbd(sb); | |
1092 | if (!sdp) { | |
d92a8d48 | 1093 | printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n"); |
b3b94faa DT |
1094 | return -ENOMEM; |
1095 | } | |
f55073ff | 1096 | sdp->sd_args = *args; |
b3b94faa | 1097 | |
fe64d517 | 1098 | if (sdp->sd_args.ar_spectator) { |
6f04c1c7 | 1099 | sb->s_flags |= MS_RDONLY; |
fe64d517 SW |
1100 | set_bit(SDF_NORECOVERY, &sdp->sd_flags); |
1101 | } | |
6f04c1c7 SW |
1102 | if (sdp->sd_args.ar_posix_acl) |
1103 | sb->s_flags |= MS_POSIXACL; | |
f25934c5 CH |
1104 | if (sdp->sd_args.ar_nobarrier) |
1105 | set_bit(SDF_NOBARRIERS, &sdp->sd_flags); | |
6f04c1c7 | 1106 | |
719ee344 SW |
1107 | sb->s_magic = GFS2_MAGIC; |
1108 | sb->s_op = &gfs2_super_ops; | |
1109 | sb->s_export_op = &gfs2_export_ops; | |
40b78a32 | 1110 | sb->s_xattr = gfs2_xattr_handlers; |
cc632e7f SW |
1111 | sb->s_qcop = &gfs2_quotactl_ops; |
1112 | sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; | |
719ee344 SW |
1113 | sb->s_time_gran = 1; |
1114 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
419c93e0 SW |
1115 | |
1116 | /* Set up the buffer cache and fill in some fake block size values | |
1117 | to allow us to read-in the on-disk superblock. */ | |
1118 | sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK); | |
1119 | sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits; | |
1120 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
1121 | GFS2_BASIC_BLOCK_SHIFT; | |
1122 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
b3b94faa | 1123 | |
5e687eac | 1124 | sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit; |
3d3c10f2 BM |
1125 | sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum; |
1126 | if (sdp->sd_args.ar_statfs_quantum) { | |
1127 | sdp->sd_tune.gt_statfs_slow = 0; | |
1128 | sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum; | |
1129 | } | |
1130 | else { | |
1131 | sdp->sd_tune.gt_statfs_slow = 1; | |
1132 | sdp->sd_tune.gt_statfs_quantum = 30; | |
1133 | } | |
48c2b613 | 1134 | |
b3b94faa DT |
1135 | error = init_names(sdp, silent); |
1136 | if (error) | |
1137 | goto fail; | |
1138 | ||
7c52b166 RP |
1139 | gfs2_create_debugfs_file(sdp); |
1140 | ||
b3b94faa DT |
1141 | error = gfs2_sys_fs_add(sdp); |
1142 | if (error) | |
1143 | goto fail; | |
1144 | ||
1145 | error = gfs2_lm_mount(sdp, silent); | |
1146 | if (error) | |
1147 | goto fail_sys; | |
1148 | ||
1149 | error = init_locking(sdp, &mount_gh, DO); | |
1150 | if (error) | |
1151 | goto fail_lm; | |
1152 | ||
9b8df98f | 1153 | error = init_sb(sdp, silent); |
b3b94faa DT |
1154 | if (error) |
1155 | goto fail_locking; | |
b3b94faa | 1156 | |
ba6e9364 SW |
1157 | error = wait_on_journal(sdp); |
1158 | if (error) | |
1159 | goto fail_sb; | |
1160 | ||
feb47ca9 SW |
1161 | /* |
1162 | * If user space has failed to join the cluster or some similar | |
1163 | * failure has occurred, then the journal id will contain a | |
1164 | * negative (error) number. This will then be returned to the | |
1165 | * caller (of the mount syscall). We do this even for spectator | |
1166 | * mounts (which just write a jid of 0 to indicate "ok" even though | |
1167 | * the jid is unused in the spectator case) | |
1168 | */ | |
1169 | if (sdp->sd_lockstruct.ls_jid < 0) { | |
1170 | error = sdp->sd_lockstruct.ls_jid; | |
1171 | sdp->sd_lockstruct.ls_jid = 0; | |
1172 | goto fail_sb; | |
1173 | } | |
1174 | ||
b3b94faa DT |
1175 | error = init_inodes(sdp, DO); |
1176 | if (error) | |
f42faf4f | 1177 | goto fail_sb; |
b3b94faa DT |
1178 | |
1179 | error = init_per_node(sdp, DO); | |
1180 | if (error) | |
1181 | goto fail_inodes; | |
1182 | ||
1183 | error = gfs2_statfs_init(sdp); | |
1184 | if (error) { | |
1185 | fs_err(sdp, "can't initialize statfs subsystem: %d\n", error); | |
1186 | goto fail_per_node; | |
1187 | } | |
1188 | ||
1189 | error = init_threads(sdp, DO); | |
1190 | if (error) | |
1191 | goto fail_per_node; | |
1192 | ||
1193 | if (!(sb->s_flags & MS_RDONLY)) { | |
1194 | error = gfs2_make_fs_rw(sdp); | |
1195 | if (error) { | |
1196 | fs_err(sdp, "can't make FS RW: %d\n", error); | |
1197 | goto fail_threads; | |
1198 | } | |
1199 | } | |
1200 | ||
1201 | gfs2_glock_dq_uninit(&mount_gh); | |
8633ecfa | 1202 | gfs2_online_uevent(sdp); |
b3b94faa DT |
1203 | return 0; |
1204 | ||
a91ea69f | 1205 | fail_threads: |
b3b94faa | 1206 | init_threads(sdp, UNDO); |
a91ea69f | 1207 | fail_per_node: |
b3b94faa | 1208 | init_per_node(sdp, UNDO); |
a91ea69f | 1209 | fail_inodes: |
b3b94faa | 1210 | init_inodes(sdp, UNDO); |
a91ea69f | 1211 | fail_sb: |
9b8df98f SW |
1212 | if (sdp->sd_root_dir) |
1213 | dput(sdp->sd_root_dir); | |
1214 | if (sdp->sd_master_dir) | |
1215 | dput(sdp->sd_master_dir); | |
e7c8707e SW |
1216 | if (sb->s_root) |
1217 | dput(sb->s_root); | |
9b8df98f | 1218 | sb->s_root = NULL; |
a91ea69f | 1219 | fail_locking: |
b3b94faa | 1220 | init_locking(sdp, &mount_gh, UNDO); |
a91ea69f | 1221 | fail_lm: |
fefc03bf | 1222 | gfs2_gl_hash_clear(sdp); |
b3b94faa | 1223 | gfs2_lm_unmount(sdp); |
a91ea69f | 1224 | fail_sys: |
b3b94faa | 1225 | gfs2_sys_fs_del(sdp); |
a91ea69f | 1226 | fail: |
5f882096 | 1227 | gfs2_delete_debugfs_file(sdp); |
a2c45807 | 1228 | kfree(sdp); |
5c676f6d | 1229 | sb->s_fs_info = NULL; |
b3b94faa DT |
1230 | return error; |
1231 | } | |
1232 | ||
f55073ff | 1233 | static int set_gfs2_super(struct super_block *s, void *data) |
b3b94faa | 1234 | { |
f55073ff SW |
1235 | s->s_bdev = data; |
1236 | s->s_dev = s->s_bdev->bd_dev; | |
1237 | ||
1238 | /* | |
1239 | * We set the bdi here to the queue backing, file systems can | |
1240 | * overwrite this in ->fill_super() | |
1241 | */ | |
1242 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1243 | return 0; | |
86384605 AD |
1244 | } |
1245 | ||
f55073ff | 1246 | static int test_gfs2_super(struct super_block *s, void *ptr) |
f6d03139 SW |
1247 | { |
1248 | struct block_device *bdev = ptr; | |
1249 | return (bdev == s->s_bdev); | |
1250 | } | |
1251 | ||
f55073ff SW |
1252 | /** |
1253 | * gfs2_get_sb - Get the GFS2 superblock | |
1254 | * @fs_type: The GFS2 filesystem type | |
1255 | * @flags: Mount flags | |
1256 | * @dev_name: The name of the device | |
1257 | * @data: The mount arguments | |
1258 | * @mnt: The vfsmnt for this mount | |
1259 | * | |
1260 | * Q. Why not use get_sb_bdev() ? | |
1261 | * A. We need to select one of two root directories to mount, independent | |
1262 | * of whether this is the initial, or subsequent, mount of this sb | |
1263 | * | |
1264 | * Returns: 0 or -ve on error | |
1265 | */ | |
1266 | ||
1267 | static int gfs2_get_sb(struct file_system_type *fs_type, int flags, | |
1268 | const char *dev_name, void *data, struct vfsmount *mnt) | |
1269 | { | |
1270 | struct block_device *bdev; | |
1271 | struct super_block *s; | |
1272 | fmode_t mode = FMODE_READ; | |
1273 | int error; | |
1274 | struct gfs2_args args; | |
1275 | struct gfs2_sbd *sdp; | |
1276 | ||
1277 | if (!(flags & MS_RDONLY)) | |
1278 | mode |= FMODE_WRITE; | |
1279 | ||
1280 | bdev = open_bdev_exclusive(dev_name, mode, fs_type); | |
1281 | if (IS_ERR(bdev)) | |
1282 | return PTR_ERR(bdev); | |
1283 | ||
1284 | /* | |
1285 | * once the super is inserted into the list by sget, s_umount | |
1286 | * will protect the lockfs code from trying to start a snapshot | |
1287 | * while we are mounting | |
1288 | */ | |
1289 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
1290 | if (bdev->bd_fsfreeze_count > 0) { | |
1291 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1292 | error = -EBUSY; | |
1293 | goto error_bdev; | |
1294 | } | |
1295 | s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev); | |
1296 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1297 | error = PTR_ERR(s); | |
1298 | if (IS_ERR(s)) | |
1299 | goto error_bdev; | |
1300 | ||
1301 | memset(&args, 0, sizeof(args)); | |
1302 | args.ar_quota = GFS2_QUOTA_DEFAULT; | |
1303 | args.ar_data = GFS2_DATA_DEFAULT; | |
5e687eac | 1304 | args.ar_commit = 30; |
3d3c10f2 BM |
1305 | args.ar_statfs_quantum = 30; |
1306 | args.ar_quota_quantum = 60; | |
f55073ff SW |
1307 | args.ar_errors = GFS2_ERRORS_DEFAULT; |
1308 | ||
1309 | error = gfs2_mount_args(&args, data); | |
1310 | if (error) { | |
1311 | printk(KERN_WARNING "GFS2: can't parse mount arguments\n"); | |
1312 | if (s->s_root) | |
1313 | goto error_super; | |
1314 | deactivate_locked_super(s); | |
1315 | return error; | |
1316 | } | |
1317 | ||
1318 | if (s->s_root) { | |
1319 | error = -EBUSY; | |
1320 | if ((flags ^ s->s_flags) & MS_RDONLY) | |
1321 | goto error_super; | |
1322 | close_bdev_exclusive(bdev, mode); | |
1323 | } else { | |
1324 | char b[BDEVNAME_SIZE]; | |
1325 | ||
1326 | s->s_flags = flags; | |
1327 | s->s_mode = mode; | |
1328 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | |
1329 | sb_set_blocksize(s, block_size(bdev)); | |
1330 | error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0); | |
1331 | if (error) { | |
1332 | deactivate_locked_super(s); | |
1333 | return error; | |
1334 | } | |
1335 | s->s_flags |= MS_ACTIVE; | |
1336 | bdev->bd_super = s; | |
1337 | } | |
1338 | ||
1339 | sdp = s->s_fs_info; | |
1340 | mnt->mnt_sb = s; | |
1341 | if (args.ar_meta) | |
1342 | mnt->mnt_root = dget(sdp->sd_master_dir); | |
1343 | else | |
1344 | mnt->mnt_root = dget(sdp->sd_root_dir); | |
1345 | return 0; | |
1346 | ||
1347 | error_super: | |
1348 | deactivate_locked_super(s); | |
1349 | error_bdev: | |
1350 | close_bdev_exclusive(bdev, mode); | |
1351 | return error; | |
1352 | } | |
1353 | ||
f6d03139 SW |
1354 | static int set_meta_super(struct super_block *s, void *ptr) |
1355 | { | |
1356 | return -EINVAL; | |
1357 | } | |
1358 | ||
003dec89 SW |
1359 | static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags, |
1360 | const char *dev_name, void *data, struct vfsmount *mnt) | |
86384605 | 1361 | { |
f6d03139 | 1362 | struct super_block *s; |
003dec89 | 1363 | struct gfs2_sbd *sdp; |
e24977d4 | 1364 | struct path path; |
86384605 | 1365 | int error; |
907b9bce | 1366 | |
e24977d4 | 1367 | error = kern_path(dev_name, LOOKUP_FOLLOW, &path); |
86384605 | 1368 | if (error) { |
9b8df98f SW |
1369 | printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n", |
1370 | dev_name, error); | |
003dec89 | 1371 | return error; |
86384605 | 1372 | } |
f55073ff | 1373 | s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, |
f6d03139 | 1374 | path.dentry->d_inode->i_sb->s_bdev); |
e24977d4 | 1375 | path_put(&path); |
f6d03139 | 1376 | if (IS_ERR(s)) { |
86384605 | 1377 | printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n"); |
f6d03139 | 1378 | return PTR_ERR(s); |
86384605 | 1379 | } |
f55073ff SW |
1380 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
1381 | deactivate_locked_super(s); | |
1382 | return -EBUSY; | |
1383 | } | |
f6d03139 SW |
1384 | sdp = s->s_fs_info; |
1385 | mnt->mnt_sb = s; | |
9b8df98f SW |
1386 | mnt->mnt_root = dget(sdp->sd_master_dir); |
1387 | return 0; | |
b3b94faa DT |
1388 | } |
1389 | ||
419c93e0 | 1390 | static void gfs2_kill_sb(struct super_block *sb) |
86384605 AD |
1391 | { |
1392 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
88a19ad0 SW |
1393 | |
1394 | if (sdp == NULL) { | |
1395 | kill_block_super(sb); | |
1396 | return; | |
3af165ac | 1397 | } |
88a19ad0 SW |
1398 | |
1399 | gfs2_meta_syncfs(sdp); | |
1400 | dput(sdp->sd_root_dir); | |
1401 | dput(sdp->sd_master_dir); | |
1402 | sdp->sd_root_dir = NULL; | |
1403 | sdp->sd_master_dir = NULL; | |
fefc03bf | 1404 | shrink_dcache_sb(sb); |
9b8df98f | 1405 | kill_block_super(sb); |
88a19ad0 SW |
1406 | gfs2_delete_debugfs_file(sdp); |
1407 | kfree(sdp); | |
86384605 AD |
1408 | } |
1409 | ||
b3b94faa DT |
1410 | struct file_system_type gfs2_fs_type = { |
1411 | .name = "gfs2", | |
1412 | .fs_flags = FS_REQUIRES_DEV, | |
1413 | .get_sb = gfs2_get_sb, | |
419c93e0 SW |
1414 | .kill_sb = gfs2_kill_sb, |
1415 | .owner = THIS_MODULE, | |
1416 | }; | |
1417 | ||
1418 | struct file_system_type gfs2meta_fs_type = { | |
1419 | .name = "gfs2meta", | |
1420 | .fs_flags = FS_REQUIRES_DEV, | |
86384605 | 1421 | .get_sb = gfs2_get_sb_meta, |
b3b94faa DT |
1422 | .owner = THIS_MODULE, |
1423 | }; | |
1424 |