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); | |
d6a079e8 | 102 | spin_lock_init(&sdp->sd_ail_lock); |
b3b94faa DT |
103 | INIT_LIST_HEAD(&sdp->sd_ail1_list); |
104 | INIT_LIST_HEAD(&sdp->sd_ail2_list); | |
105 | ||
484adff8 | 106 | init_rwsem(&sdp->sd_log_flush_lock); |
16615be1 SW |
107 | atomic_set(&sdp->sd_log_in_flight, 0); |
108 | init_waitqueue_head(&sdp->sd_log_flush_wait); | |
b3b94faa DT |
109 | |
110 | INIT_LIST_HEAD(&sdp->sd_revoke_list); | |
111 | ||
f55ab26a | 112 | mutex_init(&sdp->sd_freeze_lock); |
b3b94faa DT |
113 | |
114 | return sdp; | |
115 | } | |
116 | ||
b3b94faa | 117 | |
9b8df98f SW |
118 | /** |
119 | * gfs2_check_sb - Check superblock | |
120 | * @sdp: the filesystem | |
121 | * @sb: The superblock | |
122 | * @silent: Don't print a message if the check fails | |
123 | * | |
124 | * Checks the version code of the FS is one that we understand how to | |
125 | * read and that the sizes of the various on-disk structures have not | |
126 | * changed. | |
127 | */ | |
128 | ||
32e471ef | 129 | static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent) |
9b8df98f | 130 | { |
32e471ef SW |
131 | struct gfs2_sb_host *sb = &sdp->sd_sb; |
132 | ||
9b8df98f SW |
133 | if (sb->sb_magic != GFS2_MAGIC || |
134 | sb->sb_type != GFS2_METATYPE_SB) { | |
135 | if (!silent) | |
136 | printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n"); | |
137 | return -EINVAL; | |
138 | } | |
139 | ||
140 | /* If format numbers match exactly, we're done. */ | |
141 | ||
142 | if (sb->sb_fs_format == GFS2_FORMAT_FS && | |
143 | sb->sb_multihost_format == GFS2_FORMAT_MULTI) | |
144 | return 0; | |
145 | ||
c80dbb58 | 146 | fs_warn(sdp, "Unknown on-disk format, unable to mount\n"); |
9b8df98f | 147 | |
c80dbb58 | 148 | return -EINVAL; |
9b8df98f SW |
149 | } |
150 | ||
151 | static void end_bio_io_page(struct bio *bio, int error) | |
152 | { | |
153 | struct page *page = bio->bi_private; | |
154 | ||
155 | if (!error) | |
156 | SetPageUptodate(page); | |
157 | else | |
158 | printk(KERN_WARNING "gfs2: error %d reading superblock\n", error); | |
159 | unlock_page(page); | |
160 | } | |
161 | ||
32e471ef | 162 | static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf) |
9b8df98f | 163 | { |
32e471ef SW |
164 | struct gfs2_sb_host *sb = &sdp->sd_sb; |
165 | struct super_block *s = sdp->sd_vfs; | |
9b8df98f SW |
166 | const struct gfs2_sb *str = buf; |
167 | ||
168 | sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic); | |
169 | sb->sb_type = be32_to_cpu(str->sb_header.mh_type); | |
170 | sb->sb_format = be32_to_cpu(str->sb_header.mh_format); | |
171 | sb->sb_fs_format = be32_to_cpu(str->sb_fs_format); | |
172 | sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format); | |
173 | sb->sb_bsize = be32_to_cpu(str->sb_bsize); | |
174 | sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift); | |
175 | sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr); | |
176 | sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino); | |
177 | sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr); | |
178 | sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino); | |
179 | ||
180 | memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN); | |
181 | memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN); | |
32e471ef | 182 | memcpy(s->s_uuid, str->sb_uuid, 16); |
9b8df98f SW |
183 | } |
184 | ||
185 | /** | |
186 | * gfs2_read_super - Read the gfs2 super block from disk | |
187 | * @sdp: The GFS2 super block | |
188 | * @sector: The location of the super block | |
189 | * @error: The error code to return | |
190 | * | |
191 | * This uses the bio functions to read the super block from disk | |
192 | * because we want to be 100% sure that we never read cached data. | |
193 | * A super block is read twice only during each GFS2 mount and is | |
194 | * never written to by the filesystem. The first time its read no | |
195 | * locks are held, and the only details which are looked at are those | |
196 | * relating to the locking protocol. Once locking is up and working, | |
197 | * the sb is read again under the lock to establish the location of | |
198 | * the master directory (contains pointers to journals etc) and the | |
199 | * root directory. | |
200 | * | |
201 | * Returns: 0 on success or error | |
202 | */ | |
203 | ||
32e471ef | 204 | static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent) |
9b8df98f SW |
205 | { |
206 | struct super_block *sb = sdp->sd_vfs; | |
207 | struct gfs2_sb *p; | |
208 | struct page *page; | |
209 | struct bio *bio; | |
210 | ||
211 | page = alloc_page(GFP_NOFS); | |
212 | if (unlikely(!page)) | |
213 | return -ENOBUFS; | |
214 | ||
215 | ClearPageUptodate(page); | |
216 | ClearPageDirty(page); | |
217 | lock_page(page); | |
218 | ||
219 | bio = bio_alloc(GFP_NOFS, 1); | |
9b8df98f SW |
220 | bio->bi_sector = sector * (sb->s_blocksize >> 9); |
221 | bio->bi_bdev = sb->s_bdev; | |
222 | bio_add_page(bio, page, PAGE_SIZE, 0); | |
223 | ||
224 | bio->bi_end_io = end_bio_io_page; | |
225 | bio->bi_private = page; | |
7b6d91da | 226 | submit_bio(READ_SYNC | REQ_META, bio); |
9b8df98f SW |
227 | wait_on_page_locked(page); |
228 | bio_put(bio); | |
229 | if (!PageUptodate(page)) { | |
230 | __free_page(page); | |
231 | return -EIO; | |
232 | } | |
233 | p = kmap(page); | |
32e471ef | 234 | gfs2_sb_in(sdp, p); |
9b8df98f SW |
235 | kunmap(page); |
236 | __free_page(page); | |
32e471ef | 237 | return gfs2_check_sb(sdp, silent); |
9b8df98f | 238 | } |
6bac243f | 239 | |
9b8df98f SW |
240 | /** |
241 | * gfs2_read_sb - Read super block | |
242 | * @sdp: The GFS2 superblock | |
9b8df98f SW |
243 | * @silent: Don't print message if mount fails |
244 | * | |
245 | */ | |
246 | ||
6bac243f | 247 | static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent) |
9b8df98f SW |
248 | { |
249 | u32 hash_blocks, ind_blocks, leaf_blocks; | |
250 | u32 tmp_blocks; | |
251 | unsigned int x; | |
252 | int error; | |
253 | ||
32e471ef | 254 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent); |
9b8df98f SW |
255 | if (error) { |
256 | if (!silent) | |
257 | fs_err(sdp, "can't read superblock\n"); | |
258 | return error; | |
259 | } | |
260 | ||
9b8df98f SW |
261 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - |
262 | GFS2_BASIC_BLOCK_SHIFT; | |
263 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
264 | sdp->sd_diptrs = (sdp->sd_sb.sb_bsize - | |
265 | sizeof(struct gfs2_dinode)) / sizeof(u64); | |
266 | sdp->sd_inptrs = (sdp->sd_sb.sb_bsize - | |
267 | sizeof(struct gfs2_meta_header)) / sizeof(u64); | |
268 | sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header); | |
269 | sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2; | |
270 | sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1; | |
271 | sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64); | |
272 | sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize - | |
273 | sizeof(struct gfs2_meta_header)) / | |
274 | sizeof(struct gfs2_quota_change); | |
275 | ||
276 | /* Compute maximum reservation required to add a entry to a directory */ | |
277 | ||
278 | hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH), | |
279 | sdp->sd_jbsize); | |
280 | ||
281 | ind_blocks = 0; | |
282 | for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) { | |
283 | tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs); | |
284 | ind_blocks += tmp_blocks; | |
285 | } | |
286 | ||
287 | leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH; | |
288 | ||
289 | sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks; | |
290 | ||
291 | sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize - | |
292 | sizeof(struct gfs2_dinode); | |
293 | sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs; | |
294 | for (x = 2;; x++) { | |
295 | u64 space, d; | |
296 | u32 m; | |
297 | ||
298 | space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs; | |
299 | d = space; | |
300 | m = do_div(d, sdp->sd_inptrs); | |
301 | ||
302 | if (d != sdp->sd_heightsize[x - 1] || m) | |
303 | break; | |
304 | sdp->sd_heightsize[x] = space; | |
305 | } | |
306 | sdp->sd_max_height = x; | |
307 | sdp->sd_heightsize[x] = ~0; | |
308 | gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT); | |
309 | ||
310 | sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize - | |
311 | sizeof(struct gfs2_dinode); | |
312 | sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs; | |
313 | for (x = 2;; x++) { | |
314 | u64 space, d; | |
315 | u32 m; | |
316 | ||
317 | space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs; | |
318 | d = space; | |
319 | m = do_div(d, sdp->sd_inptrs); | |
320 | ||
321 | if (d != sdp->sd_jheightsize[x - 1] || m) | |
322 | break; | |
323 | sdp->sd_jheightsize[x] = space; | |
324 | } | |
325 | sdp->sd_max_jheight = x; | |
326 | sdp->sd_jheightsize[x] = ~0; | |
327 | gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT); | |
328 | ||
329 | return 0; | |
330 | } | |
331 | ||
b3b94faa DT |
332 | static int init_names(struct gfs2_sbd *sdp, int silent) |
333 | { | |
b3b94faa DT |
334 | char *proto, *table; |
335 | int error = 0; | |
336 | ||
337 | proto = sdp->sd_args.ar_lockproto; | |
338 | table = sdp->sd_args.ar_locktable; | |
339 | ||
340 | /* Try to autodetect */ | |
341 | ||
342 | if (!proto[0] || !table[0]) { | |
32e471ef | 343 | error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent); |
bb8d8a6f SW |
344 | if (error) |
345 | return error; | |
3cf1e7be | 346 | |
b3b94faa | 347 | if (!proto[0]) |
3cf1e7be | 348 | proto = sdp->sd_sb.sb_lockproto; |
b3b94faa | 349 | if (!table[0]) |
3cf1e7be | 350 | table = sdp->sd_sb.sb_locktable; |
b3b94faa DT |
351 | } |
352 | ||
353 | if (!table[0]) | |
354 | table = sdp->sd_vfs->s_id; | |
355 | ||
00377d8e JD |
356 | strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN); |
357 | strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN); | |
b3b94faa | 358 | |
5d35e31f DC |
359 | table = sdp->sd_table_name; |
360 | while ((table = strchr(table, '/'))) | |
b35997d4 RP |
361 | *table = '_'; |
362 | ||
b3b94faa DT |
363 | return error; |
364 | } | |
365 | ||
366 | static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh, | |
367 | int undo) | |
368 | { | |
b3b94faa DT |
369 | int error = 0; |
370 | ||
371 | if (undo) | |
372 | goto fail_trans; | |
373 | ||
b3b94faa DT |
374 | error = gfs2_glock_nq_num(sdp, |
375 | GFS2_MOUNT_LOCK, &gfs2_nondisk_glops, | |
376 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE, | |
377 | mount_gh); | |
378 | if (error) { | |
379 | fs_err(sdp, "can't acquire mount glock: %d\n", error); | |
380 | goto fail; | |
381 | } | |
382 | ||
383 | error = gfs2_glock_nq_num(sdp, | |
384 | GFS2_LIVE_LOCK, &gfs2_nondisk_glops, | |
385 | LM_ST_SHARED, | |
579b78a4 | 386 | LM_FLAG_NOEXP | GL_EXACT, |
b3b94faa DT |
387 | &sdp->sd_live_gh); |
388 | if (error) { | |
389 | fs_err(sdp, "can't acquire live glock: %d\n", error); | |
390 | goto fail_mount; | |
391 | } | |
392 | ||
393 | error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops, | |
394 | CREATE, &sdp->sd_rename_gl); | |
395 | if (error) { | |
396 | fs_err(sdp, "can't create rename glock: %d\n", error); | |
397 | goto fail_live; | |
398 | } | |
399 | ||
400 | error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops, | |
401 | CREATE, &sdp->sd_trans_gl); | |
402 | if (error) { | |
403 | fs_err(sdp, "can't create transaction glock: %d\n", error); | |
404 | goto fail_rename; | |
405 | } | |
b3b94faa DT |
406 | |
407 | return 0; | |
408 | ||
feaa7bba | 409 | fail_trans: |
b3b94faa | 410 | gfs2_glock_put(sdp->sd_trans_gl); |
feaa7bba | 411 | fail_rename: |
b3b94faa | 412 | gfs2_glock_put(sdp->sd_rename_gl); |
feaa7bba | 413 | fail_live: |
b3b94faa | 414 | gfs2_glock_dq_uninit(&sdp->sd_live_gh); |
feaa7bba | 415 | fail_mount: |
b3b94faa | 416 | gfs2_glock_dq_uninit(mount_gh); |
feaa7bba | 417 | fail: |
b3b94faa DT |
418 | return error; |
419 | } | |
420 | ||
9b8df98f SW |
421 | static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr, |
422 | u64 no_addr, const char *name) | |
f42faf4f | 423 | { |
9b8df98f SW |
424 | struct gfs2_sbd *sdp = sb->s_fs_info; |
425 | struct dentry *dentry; | |
426 | struct inode *inode; | |
427 | ||
44ad37d6 | 428 | inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0); |
9b8df98f SW |
429 | if (IS_ERR(inode)) { |
430 | fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode)); | |
431 | return PTR_ERR(inode); | |
432 | } | |
433 | dentry = d_alloc_root(inode); | |
434 | if (!dentry) { | |
435 | fs_err(sdp, "can't alloc %s dentry\n", name); | |
436 | iput(inode); | |
437 | return -ENOMEM; | |
438 | } | |
9b8df98f SW |
439 | *dptr = dentry; |
440 | return 0; | |
f42faf4f SW |
441 | } |
442 | ||
9b8df98f | 443 | static int init_sb(struct gfs2_sbd *sdp, int silent) |
b3b94faa DT |
444 | { |
445 | struct super_block *sb = sdp->sd_vfs; | |
446 | struct gfs2_holder sb_gh; | |
dbb7cae2 | 447 | u64 no_addr; |
9b8df98f | 448 | int ret; |
b3b94faa | 449 | |
9b8df98f SW |
450 | ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops, |
451 | LM_ST_SHARED, 0, &sb_gh); | |
452 | if (ret) { | |
453 | fs_err(sdp, "can't acquire superblock glock: %d\n", ret); | |
454 | return ret; | |
b3b94faa | 455 | } |
907b9bce | 456 | |
6bac243f | 457 | ret = gfs2_read_sb(sdp, silent); |
9b8df98f SW |
458 | if (ret) { |
459 | fs_err(sdp, "can't read superblock: %d\n", ret); | |
b3b94faa DT |
460 | goto out; |
461 | } | |
462 | ||
463 | /* Set up the buffer cache and SB for real */ | |
e1defc4f | 464 | if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) { |
9b8df98f | 465 | ret = -EINVAL; |
b3b94faa DT |
466 | fs_err(sdp, "FS block size (%u) is too small for device " |
467 | "block size (%u)\n", | |
e1defc4f | 468 | sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev)); |
b3b94faa DT |
469 | goto out; |
470 | } | |
471 | if (sdp->sd_sb.sb_bsize > PAGE_SIZE) { | |
9b8df98f | 472 | ret = -EINVAL; |
b3b94faa DT |
473 | fs_err(sdp, "FS block size (%u) is too big for machine " |
474 | "page size (%u)\n", | |
475 | sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE); | |
476 | goto out; | |
477 | } | |
b3b94faa DT |
478 | sb_set_blocksize(sb, sdp->sd_sb.sb_bsize); |
479 | ||
f42faf4f | 480 | /* Get the root inode */ |
dbb7cae2 | 481 | no_addr = sdp->sd_sb.sb_root_dir.no_addr; |
9b8df98f SW |
482 | ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root"); |
483 | if (ret) | |
f42faf4f | 484 | goto out; |
b3b94faa | 485 | |
9b8df98f SW |
486 | /* Get the master inode */ |
487 | no_addr = sdp->sd_sb.sb_master_dir.no_addr; | |
488 | ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master"); | |
489 | if (ret) { | |
490 | dput(sdp->sd_root_dir); | |
491 | goto out; | |
492 | } | |
493 | sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir); | |
f42faf4f | 494 | out: |
b3b94faa | 495 | gfs2_glock_dq_uninit(&sb_gh); |
9b8df98f | 496 | return ret; |
b3b94faa DT |
497 | } |
498 | ||
da6dd40d BP |
499 | /** |
500 | * map_journal_extents - create a reusable "extent" mapping from all logical | |
501 | * blocks to all physical blocks for the given journal. This will save | |
502 | * us time when writing journal blocks. Most journals will have only one | |
503 | * extent that maps all their logical blocks. That's because gfs2.mkfs | |
504 | * arranges the journal blocks sequentially to maximize performance. | |
505 | * So the extent would map the first block for the entire file length. | |
506 | * However, gfs2_jadd can happen while file activity is happening, so | |
507 | * those journals may not be sequential. Less likely is the case where | |
508 | * the users created their own journals by mounting the metafs and | |
509 | * laying it out. But it's still possible. These journals might have | |
510 | * several extents. | |
511 | * | |
512 | * TODO: This should be done in bigger chunks rather than one block at a time, | |
513 | * but since it's only done at mount time, I'm not worried about the | |
514 | * time it takes. | |
515 | */ | |
516 | static int map_journal_extents(struct gfs2_sbd *sdp) | |
517 | { | |
518 | struct gfs2_jdesc *jd = sdp->sd_jdesc; | |
519 | unsigned int lb; | |
520 | u64 db, prev_db; /* logical block, disk block, prev disk block */ | |
521 | struct gfs2_inode *ip = GFS2_I(jd->jd_inode); | |
522 | struct gfs2_journal_extent *jext = NULL; | |
523 | struct buffer_head bh; | |
524 | int rc = 0; | |
525 | ||
da6dd40d BP |
526 | prev_db = 0; |
527 | ||
a2e0f799 | 528 | for (lb = 0; lb < i_size_read(jd->jd_inode) >> sdp->sd_sb.sb_bsize_shift; lb++) { |
da6dd40d BP |
529 | bh.b_state = 0; |
530 | bh.b_blocknr = 0; | |
531 | bh.b_size = 1 << ip->i_inode.i_blkbits; | |
532 | rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0); | |
533 | db = bh.b_blocknr; | |
534 | if (rc || !db) { | |
535 | printk(KERN_INFO "GFS2 journal mapping error %d: lb=" | |
536 | "%u db=%llu\n", rc, lb, (unsigned long long)db); | |
537 | break; | |
538 | } | |
539 | if (!prev_db || db != prev_db + 1) { | |
540 | jext = kzalloc(sizeof(struct gfs2_journal_extent), | |
541 | GFP_KERNEL); | |
542 | if (!jext) { | |
543 | printk(KERN_INFO "GFS2 error: out of memory " | |
544 | "mapping journal extents.\n"); | |
545 | rc = -ENOMEM; | |
546 | break; | |
547 | } | |
548 | jext->dblock = db; | |
549 | jext->lblock = lb; | |
550 | jext->blocks = 1; | |
551 | list_add_tail(&jext->extent_list, &jd->extent_list); | |
552 | } else { | |
553 | jext->blocks++; | |
554 | } | |
555 | prev_db = db; | |
556 | } | |
557 | return rc; | |
558 | } | |
559 | ||
f057f6cd | 560 | static void gfs2_others_may_mount(struct gfs2_sbd *sdp) |
da755fdb | 561 | { |
f057f6cd SW |
562 | char *message = "FIRSTMOUNT=Done"; |
563 | char *envp[] = { message, NULL }; | |
564 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
565 | ls->ls_first_done = 1; | |
566 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp); | |
da755fdb SW |
567 | } |
568 | ||
b5289681 SW |
569 | /** |
570 | * gfs2_jindex_hold - Grab a lock on the jindex | |
571 | * @sdp: The GFS2 superblock | |
572 | * @ji_gh: the holder for the jindex glock | |
573 | * | |
574 | * Returns: errno | |
575 | */ | |
576 | ||
577 | static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh) | |
578 | { | |
579 | struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex); | |
580 | struct qstr name; | |
581 | char buf[20]; | |
582 | struct gfs2_jdesc *jd; | |
583 | int error; | |
584 | ||
585 | name.name = buf; | |
586 | ||
587 | mutex_lock(&sdp->sd_jindex_mutex); | |
588 | ||
589 | for (;;) { | |
590 | error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh); | |
591 | if (error) | |
592 | break; | |
593 | ||
594 | name.len = sprintf(buf, "journal%u", sdp->sd_journals); | |
595 | name.hash = gfs2_disk_hash(name.name, name.len); | |
596 | ||
597 | error = gfs2_dir_check(sdp->sd_jindex, &name, NULL); | |
598 | if (error == -ENOENT) { | |
599 | error = 0; | |
600 | break; | |
601 | } | |
602 | ||
603 | gfs2_glock_dq_uninit(ji_gh); | |
604 | ||
605 | if (error) | |
606 | break; | |
607 | ||
608 | error = -ENOMEM; | |
609 | jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL); | |
610 | if (!jd) | |
611 | break; | |
612 | ||
613 | INIT_LIST_HEAD(&jd->extent_list); | |
6ecd7c2d | 614 | INIT_WORK(&jd->jd_work, gfs2_recover_func); |
b5289681 SW |
615 | jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1); |
616 | if (!jd->jd_inode || IS_ERR(jd->jd_inode)) { | |
617 | if (!jd->jd_inode) | |
618 | error = -ENOENT; | |
619 | else | |
620 | error = PTR_ERR(jd->jd_inode); | |
621 | kfree(jd); | |
622 | break; | |
623 | } | |
624 | ||
625 | spin_lock(&sdp->sd_jindex_spin); | |
626 | jd->jd_jid = sdp->sd_journals++; | |
627 | list_add_tail(&jd->jd_list, &sdp->sd_jindex_list); | |
628 | spin_unlock(&sdp->sd_jindex_spin); | |
629 | } | |
630 | ||
631 | mutex_unlock(&sdp->sd_jindex_mutex); | |
632 | ||
633 | return error; | |
634 | } | |
635 | ||
b3b94faa DT |
636 | static int init_journal(struct gfs2_sbd *sdp, int undo) |
637 | { | |
9b8df98f | 638 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa | 639 | struct gfs2_holder ji_gh; |
5c676f6d | 640 | struct gfs2_inode *ip; |
b3b94faa DT |
641 | int jindex = 1; |
642 | int error = 0; | |
643 | ||
644 | if (undo) { | |
645 | jindex = 0; | |
fe64d517 | 646 | goto fail_jinode_gh; |
b3b94faa DT |
647 | } |
648 | ||
9b8df98f | 649 | sdp->sd_jindex = gfs2_lookup_simple(master, "jindex"); |
c752666c | 650 | if (IS_ERR(sdp->sd_jindex)) { |
b3b94faa | 651 | fs_err(sdp, "can't lookup journal index: %d\n", error); |
c752666c | 652 | return PTR_ERR(sdp->sd_jindex); |
b3b94faa | 653 | } |
feaa7bba | 654 | ip = GFS2_I(sdp->sd_jindex); |
b3b94faa DT |
655 | |
656 | /* Load in the journal index special file */ | |
657 | ||
658 | error = gfs2_jindex_hold(sdp, &ji_gh); | |
659 | if (error) { | |
660 | fs_err(sdp, "can't read journal index: %d\n", error); | |
661 | goto fail; | |
662 | } | |
663 | ||
0e5a9fb0 | 664 | error = -EUSERS; |
b3b94faa DT |
665 | if (!gfs2_jindex_size(sdp)) { |
666 | fs_err(sdp, "no journals!\n"); | |
907b9bce | 667 | goto fail_jindex; |
b3b94faa DT |
668 | } |
669 | ||
670 | if (sdp->sd_args.ar_spectator) { | |
671 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0); | |
fd041f0b | 672 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
5e687eac BM |
673 | atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5); |
674 | atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5); | |
b3b94faa DT |
675 | } else { |
676 | if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) { | |
677 | fs_err(sdp, "can't mount journal #%u\n", | |
678 | sdp->sd_lockstruct.ls_jid); | |
679 | fs_err(sdp, "there are only %u journals (0 - %u)\n", | |
680 | gfs2_jindex_size(sdp), | |
681 | gfs2_jindex_size(sdp) - 1); | |
682 | goto fail_jindex; | |
683 | } | |
684 | sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid); | |
685 | ||
feaa7bba | 686 | error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid, |
b3b94faa DT |
687 | &gfs2_journal_glops, |
688 | LM_ST_EXCLUSIVE, LM_FLAG_NOEXP, | |
689 | &sdp->sd_journal_gh); | |
690 | if (error) { | |
691 | fs_err(sdp, "can't acquire journal glock: %d\n", error); | |
692 | goto fail_jindex; | |
693 | } | |
694 | ||
feaa7bba SW |
695 | ip = GFS2_I(sdp->sd_jdesc->jd_inode); |
696 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, | |
75be73a8 | 697 | LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE, |
b3b94faa DT |
698 | &sdp->sd_jinode_gh); |
699 | if (error) { | |
700 | fs_err(sdp, "can't acquire journal inode glock: %d\n", | |
701 | error); | |
702 | goto fail_journal_gh; | |
703 | } | |
704 | ||
705 | error = gfs2_jdesc_check(sdp->sd_jdesc); | |
706 | if (error) { | |
707 | fs_err(sdp, "my journal (%u) is bad: %d\n", | |
708 | sdp->sd_jdesc->jd_jid, error); | |
709 | goto fail_jinode_gh; | |
710 | } | |
fd041f0b | 711 | atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks); |
5e687eac BM |
712 | atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5); |
713 | atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5); | |
da6dd40d BP |
714 | |
715 | /* Map the extents for this journal's blocks */ | |
716 | map_journal_extents(sdp); | |
b3b94faa | 717 | } |
63997775 | 718 | trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free)); |
b3b94faa DT |
719 | |
720 | if (sdp->sd_lockstruct.ls_first) { | |
721 | unsigned int x; | |
722 | for (x = 0; x < sdp->sd_journals; x++) { | |
6ecd7c2d TH |
723 | error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x), |
724 | true); | |
b3b94faa DT |
725 | if (error) { |
726 | fs_err(sdp, "error recovering journal %u: %d\n", | |
727 | x, error); | |
728 | goto fail_jinode_gh; | |
729 | } | |
730 | } | |
731 | ||
f057f6cd | 732 | gfs2_others_may_mount(sdp); |
b3b94faa | 733 | } else if (!sdp->sd_args.ar_spectator) { |
6ecd7c2d | 734 | error = gfs2_recover_journal(sdp->sd_jdesc, true); |
b3b94faa DT |
735 | if (error) { |
736 | fs_err(sdp, "error recovering my journal: %d\n", error); | |
737 | goto fail_jinode_gh; | |
738 | } | |
739 | } | |
740 | ||
741 | set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags); | |
742 | gfs2_glock_dq_uninit(&ji_gh); | |
743 | jindex = 0; | |
744 | ||
b3b94faa DT |
745 | return 0; |
746 | ||
a91ea69f | 747 | fail_jinode_gh: |
b3b94faa DT |
748 | if (!sdp->sd_args.ar_spectator) |
749 | gfs2_glock_dq_uninit(&sdp->sd_jinode_gh); | |
a91ea69f | 750 | fail_journal_gh: |
b3b94faa DT |
751 | if (!sdp->sd_args.ar_spectator) |
752 | gfs2_glock_dq_uninit(&sdp->sd_journal_gh); | |
a91ea69f | 753 | fail_jindex: |
b3b94faa DT |
754 | gfs2_jindex_free(sdp); |
755 | if (jindex) | |
756 | gfs2_glock_dq_uninit(&ji_gh); | |
a91ea69f | 757 | fail: |
f42faf4f | 758 | iput(sdp->sd_jindex); |
b3b94faa DT |
759 | return error; |
760 | } | |
761 | ||
b3b94faa DT |
762 | |
763 | static int init_inodes(struct gfs2_sbd *sdp, int undo) | |
764 | { | |
b3b94faa | 765 | int error = 0; |
5c676f6d | 766 | struct gfs2_inode *ip; |
9b8df98f | 767 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
768 | |
769 | if (undo) | |
f42faf4f SW |
770 | goto fail_qinode; |
771 | ||
f42faf4f SW |
772 | error = init_journal(sdp, undo); |
773 | if (error) | |
9b8df98f | 774 | goto fail; |
b3b94faa | 775 | |
b3b94faa | 776 | /* Read in the master statfs inode */ |
9b8df98f | 777 | sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs"); |
c752666c SW |
778 | if (IS_ERR(sdp->sd_statfs_inode)) { |
779 | error = PTR_ERR(sdp->sd_statfs_inode); | |
b3b94faa | 780 | fs_err(sdp, "can't read in statfs inode: %d\n", error); |
8d8291ae | 781 | goto fail_journal; |
b3b94faa DT |
782 | } |
783 | ||
784 | /* Read in the resource index inode */ | |
9b8df98f | 785 | sdp->sd_rindex = gfs2_lookup_simple(master, "rindex"); |
c752666c SW |
786 | if (IS_ERR(sdp->sd_rindex)) { |
787 | error = PTR_ERR(sdp->sd_rindex); | |
b3b94faa DT |
788 | fs_err(sdp, "can't get resource index inode: %d\n", error); |
789 | goto fail_statfs; | |
790 | } | |
feaa7bba | 791 | ip = GFS2_I(sdp->sd_rindex); |
cf45b752 | 792 | sdp->sd_rindex_uptodate = 0; |
b3b94faa DT |
793 | |
794 | /* Read in the quota inode */ | |
9b8df98f | 795 | sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota"); |
c752666c SW |
796 | if (IS_ERR(sdp->sd_quota_inode)) { |
797 | error = PTR_ERR(sdp->sd_quota_inode); | |
b3b94faa DT |
798 | fs_err(sdp, "can't get quota file inode: %d\n", error); |
799 | goto fail_rindex; | |
800 | } | |
b3b94faa DT |
801 | return 0; |
802 | ||
f42faf4f SW |
803 | fail_qinode: |
804 | iput(sdp->sd_quota_inode); | |
f42faf4f | 805 | fail_rindex: |
b3b94faa | 806 | gfs2_clear_rgrpd(sdp); |
f42faf4f | 807 | iput(sdp->sd_rindex); |
f42faf4f SW |
808 | fail_statfs: |
809 | iput(sdp->sd_statfs_inode); | |
f42faf4f SW |
810 | fail_journal: |
811 | init_journal(sdp, UNDO); | |
f42faf4f | 812 | fail: |
b3b94faa DT |
813 | return error; |
814 | } | |
815 | ||
816 | static int init_per_node(struct gfs2_sbd *sdp, int undo) | |
817 | { | |
f42faf4f | 818 | struct inode *pn = NULL; |
b3b94faa DT |
819 | char buf[30]; |
820 | int error = 0; | |
5c676f6d | 821 | struct gfs2_inode *ip; |
9b8df98f | 822 | struct inode *master = sdp->sd_master_dir->d_inode; |
b3b94faa DT |
823 | |
824 | if (sdp->sd_args.ar_spectator) | |
825 | return 0; | |
826 | ||
827 | if (undo) | |
828 | goto fail_qc_gh; | |
829 | ||
9b8df98f | 830 | pn = gfs2_lookup_simple(master, "per_node"); |
c752666c SW |
831 | if (IS_ERR(pn)) { |
832 | error = PTR_ERR(pn); | |
b3b94faa DT |
833 | fs_err(sdp, "can't find per_node directory: %d\n", error); |
834 | return error; | |
835 | } | |
836 | ||
b3b94faa | 837 | sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
838 | sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf); |
839 | if (IS_ERR(sdp->sd_sc_inode)) { | |
840 | error = PTR_ERR(sdp->sd_sc_inode); | |
b3b94faa | 841 | fs_err(sdp, "can't find local \"sc\" file: %d\n", error); |
8d8291ae | 842 | goto fail; |
b3b94faa DT |
843 | } |
844 | ||
b3b94faa | 845 | sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid); |
c752666c SW |
846 | sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf); |
847 | if (IS_ERR(sdp->sd_qc_inode)) { | |
848 | error = PTR_ERR(sdp->sd_qc_inode); | |
b3b94faa DT |
849 | fs_err(sdp, "can't find local \"qc\" file: %d\n", error); |
850 | goto fail_ut_i; | |
851 | } | |
852 | ||
f42faf4f | 853 | iput(pn); |
b3b94faa DT |
854 | pn = NULL; |
855 | ||
feaa7bba | 856 | ip = GFS2_I(sdp->sd_sc_inode); |
8d8291ae | 857 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
858 | &sdp->sd_sc_gh); |
859 | if (error) { | |
860 | fs_err(sdp, "can't lock local \"sc\" file: %d\n", error); | |
8d8291ae | 861 | goto fail_qc_i; |
b3b94faa DT |
862 | } |
863 | ||
feaa7bba | 864 | ip = GFS2_I(sdp->sd_qc_inode); |
8d8291ae | 865 | error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, |
b3b94faa DT |
866 | &sdp->sd_qc_gh); |
867 | if (error) { | |
868 | fs_err(sdp, "can't lock local \"qc\" file: %d\n", error); | |
869 | goto fail_ut_gh; | |
870 | } | |
871 | ||
872 | return 0; | |
873 | ||
a91ea69f | 874 | fail_qc_gh: |
b3b94faa | 875 | gfs2_glock_dq_uninit(&sdp->sd_qc_gh); |
a91ea69f | 876 | fail_ut_gh: |
b3b94faa | 877 | gfs2_glock_dq_uninit(&sdp->sd_sc_gh); |
a91ea69f | 878 | fail_qc_i: |
f42faf4f | 879 | iput(sdp->sd_qc_inode); |
a91ea69f | 880 | fail_ut_i: |
f42faf4f | 881 | iput(sdp->sd_sc_inode); |
a91ea69f | 882 | fail: |
b3b94faa | 883 | if (pn) |
f42faf4f | 884 | iput(pn); |
b3b94faa DT |
885 | return error; |
886 | } | |
887 | ||
888 | static int init_threads(struct gfs2_sbd *sdp, int undo) | |
889 | { | |
890 | struct task_struct *p; | |
891 | int error = 0; | |
892 | ||
893 | if (undo) | |
feaa7bba | 894 | goto fail_quotad; |
b3b94faa | 895 | |
b3b94faa DT |
896 | p = kthread_run(gfs2_logd, sdp, "gfs2_logd"); |
897 | error = IS_ERR(p); | |
898 | if (error) { | |
899 | fs_err(sdp, "can't start logd thread: %d\n", error); | |
900 | return error; | |
901 | } | |
902 | sdp->sd_logd_process = p; | |
903 | ||
b3b94faa DT |
904 | p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad"); |
905 | error = IS_ERR(p); | |
906 | if (error) { | |
907 | fs_err(sdp, "can't start quotad thread: %d\n", error); | |
908 | goto fail; | |
909 | } | |
910 | sdp->sd_quotad_process = p; | |
911 | ||
b3b94faa DT |
912 | return 0; |
913 | ||
b3b94faa | 914 | |
feaa7bba | 915 | fail_quotad: |
b3b94faa | 916 | kthread_stop(sdp->sd_quotad_process); |
feaa7bba | 917 | fail: |
b3b94faa | 918 | kthread_stop(sdp->sd_logd_process); |
b3b94faa DT |
919 | return error; |
920 | } | |
da755fdb | 921 | |
f057f6cd SW |
922 | static const match_table_t nolock_tokens = { |
923 | { Opt_jid, "jid=%d\n", }, | |
924 | { Opt_err, NULL }, | |
925 | }; | |
926 | ||
927 | static const struct lm_lockops nolock_ops = { | |
928 | .lm_proto_name = "lock_nolock", | |
fc0e38da | 929 | .lm_put_lock = gfs2_glock_free, |
f057f6cd SW |
930 | .lm_tokens = &nolock_tokens, |
931 | }; | |
932 | ||
da755fdb SW |
933 | /** |
934 | * gfs2_lm_mount - mount a locking protocol | |
935 | * @sdp: the filesystem | |
3ad2f3fb | 936 | * @args: mount arguments |
da755fdb SW |
937 | * @silent: if 1, don't complain if the FS isn't a GFS2 fs |
938 | * | |
939 | * Returns: errno | |
940 | */ | |
941 | ||
942 | static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent) | |
943 | { | |
f057f6cd SW |
944 | const struct lm_lockops *lm; |
945 | struct lm_lockstruct *ls = &sdp->sd_lockstruct; | |
946 | struct gfs2_args *args = &sdp->sd_args; | |
947 | const char *proto = sdp->sd_proto_name; | |
948 | const char *table = sdp->sd_table_name; | |
949 | const char *fsname; | |
950 | char *o, *options; | |
951 | int ret; | |
da755fdb | 952 | |
f057f6cd SW |
953 | if (!strcmp("lock_nolock", proto)) { |
954 | lm = &nolock_ops; | |
955 | sdp->sd_args.ar_localflocks = 1; | |
f057f6cd SW |
956 | #ifdef CONFIG_GFS2_FS_LOCKING_DLM |
957 | } else if (!strcmp("lock_dlm", proto)) { | |
958 | lm = &gfs2_dlm_ops; | |
959 | #endif | |
960 | } else { | |
961 | printk(KERN_INFO "GFS2: can't find protocol %s\n", proto); | |
962 | return -ENOENT; | |
963 | } | |
da755fdb SW |
964 | |
965 | fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table); | |
966 | ||
f057f6cd SW |
967 | ls->ls_ops = lm; |
968 | ls->ls_first = 1; | |
da755fdb | 969 | |
f057f6cd SW |
970 | for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) { |
971 | substring_t tmp[MAX_OPT_ARGS]; | |
972 | int token, option; | |
973 | ||
974 | if (!o || !*o) | |
975 | continue; | |
976 | ||
977 | token = match_token(o, *lm->lm_tokens, tmp); | |
978 | switch (token) { | |
979 | case Opt_jid: | |
980 | ret = match_int(&tmp[0], &option); | |
981 | if (ret || option < 0) | |
982 | goto hostdata_error; | |
ba6e9364 SW |
983 | if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags)) |
984 | ls->ls_jid = option; | |
f057f6cd SW |
985 | break; |
986 | case Opt_id: | |
2b88f7c5 | 987 | /* Obsolete, but left for backward compat purposes */ |
f057f6cd SW |
988 | break; |
989 | case Opt_first: | |
990 | ret = match_int(&tmp[0], &option); | |
991 | if (ret || (option != 0 && option != 1)) | |
992 | goto hostdata_error; | |
993 | ls->ls_first = option; | |
994 | break; | |
995 | case Opt_nodir: | |
996 | ret = match_int(&tmp[0], &option); | |
997 | if (ret || (option != 0 && option != 1)) | |
998 | goto hostdata_error; | |
999 | ls->ls_nodir = option; | |
1000 | break; | |
1001 | case Opt_err: | |
1002 | default: | |
1003 | hostdata_error: | |
1004 | fs_info(sdp, "unknown hostdata (%s)\n", o); | |
1005 | return -EINVAL; | |
1006 | } | |
da755fdb SW |
1007 | } |
1008 | ||
1009 | if (sdp->sd_args.ar_spectator) | |
1010 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table); | |
1011 | else | |
1012 | snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table, | |
1013 | sdp->sd_lockstruct.ls_jid); | |
1014 | ||
f057f6cd SW |
1015 | fsname = strchr(table, ':'); |
1016 | if (fsname) | |
1017 | fsname++; | |
1018 | if (lm->lm_mount == NULL) { | |
1019 | fs_info(sdp, "Now mounting FS...\n"); | |
1020 | return 0; | |
da755fdb | 1021 | } |
f057f6cd SW |
1022 | ret = lm->lm_mount(sdp, fsname); |
1023 | if (ret == 0) | |
1024 | fs_info(sdp, "Joined cluster. Now mounting FS...\n"); | |
1025 | return ret; | |
da755fdb SW |
1026 | } |
1027 | ||
1028 | void gfs2_lm_unmount(struct gfs2_sbd *sdp) | |
1029 | { | |
f057f6cd SW |
1030 | const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops; |
1031 | if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) && | |
1032 | lm->lm_unmount) | |
1033 | lm->lm_unmount(sdp); | |
da755fdb | 1034 | } |
b3b94faa | 1035 | |
ba6e9364 SW |
1036 | static int gfs2_journalid_wait(void *word) |
1037 | { | |
1038 | if (signal_pending(current)) | |
1039 | return -EINTR; | |
1040 | schedule(); | |
1041 | return 0; | |
1042 | } | |
1043 | ||
1044 | static int wait_on_journal(struct gfs2_sbd *sdp) | |
1045 | { | |
ba6e9364 SW |
1046 | if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL) |
1047 | return 0; | |
1048 | ||
1049 | return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE); | |
1050 | } | |
1051 | ||
8633ecfa SW |
1052 | void gfs2_online_uevent(struct gfs2_sbd *sdp) |
1053 | { | |
1054 | struct super_block *sb = sdp->sd_vfs; | |
1055 | char ro[20]; | |
1056 | char spectator[20]; | |
1057 | char *envp[] = { ro, spectator, NULL }; | |
1058 | sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0); | |
1059 | sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0); | |
1060 | kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp); | |
1061 | } | |
1062 | ||
b3b94faa DT |
1063 | /** |
1064 | * fill_super - Read in superblock | |
1065 | * @sb: The VFS superblock | |
1066 | * @data: Mount options | |
1067 | * @silent: Don't complain if it's not a GFS2 filesystem | |
1068 | * | |
1069 | * Returns: errno | |
1070 | */ | |
1071 | ||
f55073ff | 1072 | static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent) |
b3b94faa DT |
1073 | { |
1074 | struct gfs2_sbd *sdp; | |
1075 | struct gfs2_holder mount_gh; | |
1076 | int error; | |
1077 | ||
1078 | sdp = init_sbd(sb); | |
1079 | if (!sdp) { | |
d92a8d48 | 1080 | printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n"); |
b3b94faa DT |
1081 | return -ENOMEM; |
1082 | } | |
f55073ff | 1083 | sdp->sd_args = *args; |
b3b94faa | 1084 | |
fe64d517 | 1085 | if (sdp->sd_args.ar_spectator) { |
6f04c1c7 | 1086 | sb->s_flags |= MS_RDONLY; |
fe64d517 SW |
1087 | set_bit(SDF_NORECOVERY, &sdp->sd_flags); |
1088 | } | |
6f04c1c7 SW |
1089 | if (sdp->sd_args.ar_posix_acl) |
1090 | sb->s_flags |= MS_POSIXACL; | |
f25934c5 CH |
1091 | if (sdp->sd_args.ar_nobarrier) |
1092 | set_bit(SDF_NOBARRIERS, &sdp->sd_flags); | |
6f04c1c7 | 1093 | |
719ee344 SW |
1094 | sb->s_magic = GFS2_MAGIC; |
1095 | sb->s_op = &gfs2_super_ops; | |
41ced6dc | 1096 | sb->s_d_op = &gfs2_dops; |
719ee344 | 1097 | sb->s_export_op = &gfs2_export_ops; |
40b78a32 | 1098 | sb->s_xattr = gfs2_xattr_handlers; |
cc632e7f SW |
1099 | sb->s_qcop = &gfs2_quotactl_ops; |
1100 | sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE; | |
719ee344 SW |
1101 | sb->s_time_gran = 1; |
1102 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
419c93e0 SW |
1103 | |
1104 | /* Set up the buffer cache and fill in some fake block size values | |
1105 | to allow us to read-in the on-disk superblock. */ | |
1106 | sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK); | |
1107 | sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits; | |
1108 | sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift - | |
1109 | GFS2_BASIC_BLOCK_SHIFT; | |
1110 | sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift; | |
b3b94faa | 1111 | |
5e687eac | 1112 | sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit; |
3d3c10f2 BM |
1113 | sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum; |
1114 | if (sdp->sd_args.ar_statfs_quantum) { | |
1115 | sdp->sd_tune.gt_statfs_slow = 0; | |
1116 | sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum; | |
32e471ef | 1117 | } else { |
3d3c10f2 BM |
1118 | sdp->sd_tune.gt_statfs_slow = 1; |
1119 | sdp->sd_tune.gt_statfs_quantum = 30; | |
1120 | } | |
48c2b613 | 1121 | |
b3b94faa DT |
1122 | error = init_names(sdp, silent); |
1123 | if (error) | |
1124 | goto fail; | |
1125 | ||
7c52b166 RP |
1126 | gfs2_create_debugfs_file(sdp); |
1127 | ||
b3b94faa DT |
1128 | error = gfs2_sys_fs_add(sdp); |
1129 | if (error) | |
1130 | goto fail; | |
1131 | ||
1132 | error = gfs2_lm_mount(sdp, silent); | |
1133 | if (error) | |
1134 | goto fail_sys; | |
1135 | ||
1136 | error = init_locking(sdp, &mount_gh, DO); | |
1137 | if (error) | |
1138 | goto fail_lm; | |
1139 | ||
9b8df98f | 1140 | error = init_sb(sdp, silent); |
b3b94faa DT |
1141 | if (error) |
1142 | goto fail_locking; | |
b3b94faa | 1143 | |
ba6e9364 SW |
1144 | error = wait_on_journal(sdp); |
1145 | if (error) | |
1146 | goto fail_sb; | |
1147 | ||
feb47ca9 SW |
1148 | /* |
1149 | * If user space has failed to join the cluster or some similar | |
1150 | * failure has occurred, then the journal id will contain a | |
1151 | * negative (error) number. This will then be returned to the | |
1152 | * caller (of the mount syscall). We do this even for spectator | |
1153 | * mounts (which just write a jid of 0 to indicate "ok" even though | |
1154 | * the jid is unused in the spectator case) | |
1155 | */ | |
1156 | if (sdp->sd_lockstruct.ls_jid < 0) { | |
1157 | error = sdp->sd_lockstruct.ls_jid; | |
1158 | sdp->sd_lockstruct.ls_jid = 0; | |
1159 | goto fail_sb; | |
1160 | } | |
1161 | ||
b3b94faa DT |
1162 | error = init_inodes(sdp, DO); |
1163 | if (error) | |
f42faf4f | 1164 | goto fail_sb; |
b3b94faa DT |
1165 | |
1166 | error = init_per_node(sdp, DO); | |
1167 | if (error) | |
1168 | goto fail_inodes; | |
1169 | ||
1170 | error = gfs2_statfs_init(sdp); | |
1171 | if (error) { | |
1172 | fs_err(sdp, "can't initialize statfs subsystem: %d\n", error); | |
1173 | goto fail_per_node; | |
1174 | } | |
1175 | ||
1176 | error = init_threads(sdp, DO); | |
1177 | if (error) | |
1178 | goto fail_per_node; | |
1179 | ||
1180 | if (!(sb->s_flags & MS_RDONLY)) { | |
1181 | error = gfs2_make_fs_rw(sdp); | |
1182 | if (error) { | |
1183 | fs_err(sdp, "can't make FS RW: %d\n", error); | |
1184 | goto fail_threads; | |
1185 | } | |
1186 | } | |
1187 | ||
1188 | gfs2_glock_dq_uninit(&mount_gh); | |
8633ecfa | 1189 | gfs2_online_uevent(sdp); |
b3b94faa DT |
1190 | return 0; |
1191 | ||
a91ea69f | 1192 | fail_threads: |
b3b94faa | 1193 | init_threads(sdp, UNDO); |
a91ea69f | 1194 | fail_per_node: |
b3b94faa | 1195 | init_per_node(sdp, UNDO); |
a91ea69f | 1196 | fail_inodes: |
b3b94faa | 1197 | init_inodes(sdp, UNDO); |
a91ea69f | 1198 | fail_sb: |
9b8df98f SW |
1199 | if (sdp->sd_root_dir) |
1200 | dput(sdp->sd_root_dir); | |
1201 | if (sdp->sd_master_dir) | |
1202 | dput(sdp->sd_master_dir); | |
e7c8707e SW |
1203 | if (sb->s_root) |
1204 | dput(sb->s_root); | |
9b8df98f | 1205 | sb->s_root = NULL; |
a91ea69f | 1206 | fail_locking: |
b3b94faa | 1207 | init_locking(sdp, &mount_gh, UNDO); |
a91ea69f | 1208 | fail_lm: |
fefc03bf | 1209 | gfs2_gl_hash_clear(sdp); |
b3b94faa | 1210 | gfs2_lm_unmount(sdp); |
a91ea69f | 1211 | fail_sys: |
b3b94faa | 1212 | gfs2_sys_fs_del(sdp); |
a91ea69f | 1213 | fail: |
5f882096 | 1214 | gfs2_delete_debugfs_file(sdp); |
a2c45807 | 1215 | kfree(sdp); |
5c676f6d | 1216 | sb->s_fs_info = NULL; |
b3b94faa DT |
1217 | return error; |
1218 | } | |
1219 | ||
f55073ff | 1220 | static int set_gfs2_super(struct super_block *s, void *data) |
b3b94faa | 1221 | { |
f55073ff SW |
1222 | s->s_bdev = data; |
1223 | s->s_dev = s->s_bdev->bd_dev; | |
1224 | ||
1225 | /* | |
1226 | * We set the bdi here to the queue backing, file systems can | |
1227 | * overwrite this in ->fill_super() | |
1228 | */ | |
1229 | s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info; | |
1230 | return 0; | |
86384605 AD |
1231 | } |
1232 | ||
f55073ff | 1233 | static int test_gfs2_super(struct super_block *s, void *ptr) |
f6d03139 SW |
1234 | { |
1235 | struct block_device *bdev = ptr; | |
1236 | return (bdev == s->s_bdev); | |
1237 | } | |
1238 | ||
f55073ff | 1239 | /** |
8bcbbf00 | 1240 | * gfs2_mount - Get the GFS2 superblock |
f55073ff SW |
1241 | * @fs_type: The GFS2 filesystem type |
1242 | * @flags: Mount flags | |
1243 | * @dev_name: The name of the device | |
1244 | * @data: The mount arguments | |
f55073ff SW |
1245 | * |
1246 | * Q. Why not use get_sb_bdev() ? | |
1247 | * A. We need to select one of two root directories to mount, independent | |
1248 | * of whether this is the initial, or subsequent, mount of this sb | |
1249 | * | |
1250 | * Returns: 0 or -ve on error | |
1251 | */ | |
1252 | ||
8bcbbf00 AV |
1253 | static struct dentry *gfs2_mount(struct file_system_type *fs_type, int flags, |
1254 | const char *dev_name, void *data) | |
f55073ff SW |
1255 | { |
1256 | struct block_device *bdev; | |
1257 | struct super_block *s; | |
d4d77629 | 1258 | fmode_t mode = FMODE_READ | FMODE_EXCL; |
f55073ff SW |
1259 | int error; |
1260 | struct gfs2_args args; | |
1261 | struct gfs2_sbd *sdp; | |
1262 | ||
1263 | if (!(flags & MS_RDONLY)) | |
1264 | mode |= FMODE_WRITE; | |
1265 | ||
d4d77629 | 1266 | bdev = blkdev_get_by_path(dev_name, mode, fs_type); |
f55073ff | 1267 | if (IS_ERR(bdev)) |
8bcbbf00 | 1268 | return ERR_CAST(bdev); |
f55073ff SW |
1269 | |
1270 | /* | |
1271 | * once the super is inserted into the list by sget, s_umount | |
1272 | * will protect the lockfs code from trying to start a snapshot | |
1273 | * while we are mounting | |
1274 | */ | |
1275 | mutex_lock(&bdev->bd_fsfreeze_mutex); | |
1276 | if (bdev->bd_fsfreeze_count > 0) { | |
1277 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1278 | error = -EBUSY; | |
1279 | goto error_bdev; | |
1280 | } | |
1281 | s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev); | |
1282 | mutex_unlock(&bdev->bd_fsfreeze_mutex); | |
1283 | error = PTR_ERR(s); | |
1284 | if (IS_ERR(s)) | |
1285 | goto error_bdev; | |
1286 | ||
8bcbbf00 | 1287 | if (s->s_root) |
d4d77629 | 1288 | blkdev_put(bdev, mode); |
8bcbbf00 | 1289 | |
f55073ff SW |
1290 | memset(&args, 0, sizeof(args)); |
1291 | args.ar_quota = GFS2_QUOTA_DEFAULT; | |
1292 | args.ar_data = GFS2_DATA_DEFAULT; | |
5e687eac | 1293 | args.ar_commit = 30; |
3d3c10f2 BM |
1294 | args.ar_statfs_quantum = 30; |
1295 | args.ar_quota_quantum = 60; | |
f55073ff SW |
1296 | args.ar_errors = GFS2_ERRORS_DEFAULT; |
1297 | ||
1298 | error = gfs2_mount_args(&args, data); | |
1299 | if (error) { | |
1300 | printk(KERN_WARNING "GFS2: can't parse mount arguments\n"); | |
8bcbbf00 | 1301 | goto error_super; |
f55073ff SW |
1302 | } |
1303 | ||
1304 | if (s->s_root) { | |
1305 | error = -EBUSY; | |
1306 | if ((flags ^ s->s_flags) & MS_RDONLY) | |
1307 | goto error_super; | |
f55073ff SW |
1308 | } else { |
1309 | char b[BDEVNAME_SIZE]; | |
1310 | ||
1311 | s->s_flags = flags; | |
1312 | s->s_mode = mode; | |
1313 | strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id)); | |
1314 | sb_set_blocksize(s, block_size(bdev)); | |
1315 | error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0); | |
8bcbbf00 AV |
1316 | if (error) |
1317 | goto error_super; | |
f55073ff SW |
1318 | s->s_flags |= MS_ACTIVE; |
1319 | bdev->bd_super = s; | |
1320 | } | |
1321 | ||
1322 | sdp = s->s_fs_info; | |
f55073ff | 1323 | if (args.ar_meta) |
8bcbbf00 | 1324 | return dget(sdp->sd_master_dir); |
f55073ff | 1325 | else |
8bcbbf00 | 1326 | return dget(sdp->sd_root_dir); |
f55073ff SW |
1327 | |
1328 | error_super: | |
1329 | deactivate_locked_super(s); | |
8bcbbf00 | 1330 | return ERR_PTR(error); |
f55073ff | 1331 | error_bdev: |
d4d77629 | 1332 | blkdev_put(bdev, mode); |
8bcbbf00 | 1333 | return ERR_PTR(error); |
f55073ff SW |
1334 | } |
1335 | ||
f6d03139 SW |
1336 | static int set_meta_super(struct super_block *s, void *ptr) |
1337 | { | |
1338 | return -EINVAL; | |
1339 | } | |
1340 | ||
8bcbbf00 AV |
1341 | static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type, |
1342 | int flags, const char *dev_name, void *data) | |
86384605 | 1343 | { |
f6d03139 | 1344 | struct super_block *s; |
003dec89 | 1345 | struct gfs2_sbd *sdp; |
e24977d4 | 1346 | struct path path; |
86384605 | 1347 | int error; |
907b9bce | 1348 | |
e24977d4 | 1349 | error = kern_path(dev_name, LOOKUP_FOLLOW, &path); |
86384605 | 1350 | if (error) { |
9b8df98f SW |
1351 | printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n", |
1352 | dev_name, error); | |
8bcbbf00 | 1353 | return ERR_PTR(error); |
86384605 | 1354 | } |
f55073ff | 1355 | s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, |
f6d03139 | 1356 | path.dentry->d_inode->i_sb->s_bdev); |
e24977d4 | 1357 | path_put(&path); |
f6d03139 | 1358 | if (IS_ERR(s)) { |
86384605 | 1359 | printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n"); |
8bcbbf00 | 1360 | return ERR_CAST(s); |
86384605 | 1361 | } |
f55073ff SW |
1362 | if ((flags ^ s->s_flags) & MS_RDONLY) { |
1363 | deactivate_locked_super(s); | |
8bcbbf00 | 1364 | return ERR_PTR(-EBUSY); |
f55073ff | 1365 | } |
f6d03139 | 1366 | sdp = s->s_fs_info; |
8bcbbf00 | 1367 | return dget(sdp->sd_master_dir); |
b3b94faa DT |
1368 | } |
1369 | ||
419c93e0 | 1370 | static void gfs2_kill_sb(struct super_block *sb) |
86384605 AD |
1371 | { |
1372 | struct gfs2_sbd *sdp = sb->s_fs_info; | |
88a19ad0 SW |
1373 | |
1374 | if (sdp == NULL) { | |
1375 | kill_block_super(sb); | |
1376 | return; | |
3af165ac | 1377 | } |
88a19ad0 SW |
1378 | |
1379 | gfs2_meta_syncfs(sdp); | |
1380 | dput(sdp->sd_root_dir); | |
1381 | dput(sdp->sd_master_dir); | |
1382 | sdp->sd_root_dir = NULL; | |
1383 | sdp->sd_master_dir = NULL; | |
fefc03bf | 1384 | shrink_dcache_sb(sb); |
9b8df98f | 1385 | kill_block_super(sb); |
88a19ad0 SW |
1386 | gfs2_delete_debugfs_file(sdp); |
1387 | kfree(sdp); | |
86384605 AD |
1388 | } |
1389 | ||
b3b94faa DT |
1390 | struct file_system_type gfs2_fs_type = { |
1391 | .name = "gfs2", | |
1392 | .fs_flags = FS_REQUIRES_DEV, | |
8bcbbf00 | 1393 | .mount = gfs2_mount, |
419c93e0 SW |
1394 | .kill_sb = gfs2_kill_sb, |
1395 | .owner = THIS_MODULE, | |
1396 | }; | |
1397 | ||
1398 | struct file_system_type gfs2meta_fs_type = { | |
1399 | .name = "gfs2meta", | |
1400 | .fs_flags = FS_REQUIRES_DEV, | |
8bcbbf00 | 1401 | .mount = gfs2_mount_meta, |
b3b94faa DT |
1402 | .owner = THIS_MODULE, |
1403 | }; | |
1404 |