]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/gfs2/ops_fstype.c
GFS2: Limit number of transaction blocks requested for truncates
[mirror_ubuntu-artful-kernel.git] / fs / gfs2 / ops_fstype.c
CommitLineData
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
d77d1b58
JP
10#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
11
b3b94faa
DT
12#include <linux/sched.h>
13#include <linux/slab.h>
14#include <linux/spinlock.h>
15#include <linux/completion.h>
16#include <linux/buffer_head.h>
b3b94faa
DT
17#include <linux/blkdev.h>
18#include <linux/kthread.h>
afeacc8c 19#include <linux/export.h>
86384605
AD
20#include <linux/namei.h>
21#include <linux/mount.h>
5c676f6d 22#include <linux/gfs2_ondisk.h>
cc632e7f 23#include <linux/quotaops.h>
56aa72d0 24#include <linux/lockdep.h>
7f78e035 25#include <linux/module.h>
b3b94faa
DT
26
27#include "gfs2.h"
5c676f6d 28#include "incore.h"
da6dd40d 29#include "bmap.h"
b3b94faa
DT
30#include "glock.h"
31#include "glops.h"
32#include "inode.h"
b3b94faa
DT
33#include "recovery.h"
34#include "rgrp.h"
35#include "super.h"
b3b94faa 36#include "sys.h"
5c676f6d 37#include "util.h"
bb3b0e3d 38#include "log.h"
9ac1b4d9 39#include "quota.h"
b5289681 40#include "dir.h"
70d4ee94 41#include "meta_io.h"
63997775 42#include "trace_gfs2.h"
b3b94faa
DT
43
44#define DO 0
45#define UNDO 1
46
9b8df98f
SW
47/**
48 * gfs2_tune_init - Fill a gfs2_tune structure with default values
49 * @gt: tune
50 *
51 */
52
53static void gfs2_tune_init(struct gfs2_tune *gt)
54{
55 spin_lock_init(&gt->gt_spin);
56
9b8df98f
SW
57 gt->gt_quota_warn_period = 10;
58 gt->gt_quota_scale_num = 1;
59 gt->gt_quota_scale_den = 1;
9b8df98f 60 gt->gt_new_files_jdata = 0;
47a9a527 61 gt->gt_max_readahead = BIT(18);
9b8df98f 62 gt->gt_complain_secs = 10;
9b8df98f
SW
63}
64
b3b94faa
DT
65static struct gfs2_sbd *init_sbd(struct super_block *sb)
66{
67 struct gfs2_sbd *sdp;
70d4ee94 68 struct address_space *mapping;
b3b94faa 69
85d1da67 70 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
b3b94faa
DT
71 if (!sdp)
72 return NULL;
73
5c676f6d 74 sb->s_fs_info = sdp;
b3b94faa 75 sdp->sd_vfs = sb;
a245769f
SW
76 sdp->sd_lkstats = alloc_percpu(struct gfs2_pcpu_lkstats);
77 if (!sdp->sd_lkstats) {
78 kfree(sdp);
79 return NULL;
80 }
81
ba6e9364 82 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
b3b94faa
DT
83 gfs2_tune_init(&sdp->sd_tune);
84
e402746a
SW
85 init_waitqueue_head(&sdp->sd_glock_wait);
86 atomic_set(&sdp->sd_glock_disposal, 0);
3942ae53 87 init_completion(&sdp->sd_locking_init);
fd95e81c 88 init_completion(&sdp->sd_wdack);
b3b94faa 89 spin_lock_init(&sdp->sd_statfs_spin);
b3b94faa
DT
90
91 spin_lock_init(&sdp->sd_rindex_spin);
7c9ca621 92 sdp->sd_rindex_tree.rb_node = NULL;
b3b94faa
DT
93
94 INIT_LIST_HEAD(&sdp->sd_jindex_list);
95 spin_lock_init(&sdp->sd_jindex_spin);
f55ab26a 96 mutex_init(&sdp->sd_jindex_mutex);
0e48e055 97 init_completion(&sdp->sd_journal_ready);
b3b94faa 98
b3b94faa 99 INIT_LIST_HEAD(&sdp->sd_quota_list);
f55ab26a 100 mutex_init(&sdp->sd_quota_mutex);
e46c772d 101 mutex_init(&sdp->sd_quota_sync_mutex);
37b2c837 102 init_waitqueue_head(&sdp->sd_quota_wait);
813e0c46
SW
103 INIT_LIST_HEAD(&sdp->sd_trunc_list);
104 spin_lock_init(&sdp->sd_trunc_lock);
2d9e7230 105 spin_lock_init(&sdp->sd_bitmap_lock);
b3b94faa 106
70d4ee94
SW
107 mapping = &sdp->sd_aspace;
108
39849d69 109 address_space_init_once(mapping);
1b2ad412 110 mapping->a_ops = &gfs2_rgrp_aops;
70d4ee94
SW
111 mapping->host = sb->s_bdev->bd_inode;
112 mapping->flags = 0;
113 mapping_set_gfp_mask(mapping, GFP_NOFS);
114 mapping->private_data = NULL;
70d4ee94
SW
115 mapping->writeback_index = 0;
116
b3b94faa 117 spin_lock_init(&sdp->sd_log_lock);
5e687eac 118 atomic_set(&sdp->sd_log_pinned, 0);
b3b94faa 119 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
d7b616e2 120 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
45138990 121 spin_lock_init(&sdp->sd_ordered_lock);
b3b94faa 122
5e687eac
BM
123 init_waitqueue_head(&sdp->sd_log_waitq);
124 init_waitqueue_head(&sdp->sd_logd_waitq);
d6a079e8 125 spin_lock_init(&sdp->sd_ail_lock);
b3b94faa
DT
126 INIT_LIST_HEAD(&sdp->sd_ail1_list);
127 INIT_LIST_HEAD(&sdp->sd_ail2_list);
128
484adff8 129 init_rwsem(&sdp->sd_log_flush_lock);
16615be1 130 atomic_set(&sdp->sd_log_in_flight, 0);
2e60d768
BM
131 atomic_set(&sdp->sd_reserving_log, 0);
132 init_waitqueue_head(&sdp->sd_reserving_log_wait);
16615be1 133 init_waitqueue_head(&sdp->sd_log_flush_wait);
2e60d768
BM
134 atomic_set(&sdp->sd_freeze_state, SFS_UNFROZEN);
135 mutex_init(&sdp->sd_freeze_mutex);
b3b94faa 136
b3b94faa
DT
137 return sdp;
138}
139
b3b94faa 140
9b8df98f
SW
141/**
142 * gfs2_check_sb - Check superblock
143 * @sdp: the filesystem
144 * @sb: The superblock
145 * @silent: Don't print a message if the check fails
146 *
147 * Checks the version code of the FS is one that we understand how to
148 * read and that the sizes of the various on-disk structures have not
149 * changed.
150 */
151
32e471ef 152static int gfs2_check_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f 153{
32e471ef
SW
154 struct gfs2_sb_host *sb = &sdp->sd_sb;
155
9b8df98f
SW
156 if (sb->sb_magic != GFS2_MAGIC ||
157 sb->sb_type != GFS2_METATYPE_SB) {
158 if (!silent)
d77d1b58 159 pr_warn("not a GFS2 filesystem\n");
9b8df98f
SW
160 return -EINVAL;
161 }
162
163 /* If format numbers match exactly, we're done. */
164
165 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
166 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
167 return 0;
168
c80dbb58 169 fs_warn(sdp, "Unknown on-disk format, unable to mount\n");
9b8df98f 170
c80dbb58 171 return -EINVAL;
9b8df98f
SW
172}
173
4246a0b6 174static void end_bio_io_page(struct bio *bio)
9b8df98f
SW
175{
176 struct page *page = bio->bi_private;
177
4246a0b6 178 if (!bio->bi_error)
9b8df98f
SW
179 SetPageUptodate(page);
180 else
4246a0b6 181 pr_warn("error %d reading superblock\n", bio->bi_error);
9b8df98f
SW
182 unlock_page(page);
183}
184
32e471ef 185static void gfs2_sb_in(struct gfs2_sbd *sdp, const void *buf)
9b8df98f 186{
32e471ef
SW
187 struct gfs2_sb_host *sb = &sdp->sd_sb;
188 struct super_block *s = sdp->sd_vfs;
9b8df98f
SW
189 const struct gfs2_sb *str = buf;
190
191 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
192 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
193 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
194 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
195 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
196 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
197 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
198 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
199 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
200 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
201 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
202
203 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
204 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
32e471ef 205 memcpy(s->s_uuid, str->sb_uuid, 16);
9b8df98f
SW
206}
207
208/**
209 * gfs2_read_super - Read the gfs2 super block from disk
210 * @sdp: The GFS2 super block
211 * @sector: The location of the super block
212 * @error: The error code to return
213 *
214 * This uses the bio functions to read the super block from disk
215 * because we want to be 100% sure that we never read cached data.
216 * A super block is read twice only during each GFS2 mount and is
217 * never written to by the filesystem. The first time its read no
218 * locks are held, and the only details which are looked at are those
219 * relating to the locking protocol. Once locking is up and working,
220 * the sb is read again under the lock to establish the location of
221 * the master directory (contains pointers to journals etc) and the
222 * root directory.
223 *
224 * Returns: 0 on success or error
225 */
226
32e471ef 227static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector, int silent)
9b8df98f
SW
228{
229 struct super_block *sb = sdp->sd_vfs;
230 struct gfs2_sb *p;
231 struct page *page;
232 struct bio *bio;
233
234 page = alloc_page(GFP_NOFS);
235 if (unlikely(!page))
ac3beb6a 236 return -ENOMEM;
9b8df98f
SW
237
238 ClearPageUptodate(page);
239 ClearPageDirty(page);
240 lock_page(page);
241
242 bio = bio_alloc(GFP_NOFS, 1);
4f024f37 243 bio->bi_iter.bi_sector = sector * (sb->s_blocksize >> 9);
9b8df98f
SW
244 bio->bi_bdev = sb->s_bdev;
245 bio_add_page(bio, page, PAGE_SIZE, 0);
246
247 bio->bi_end_io = end_bio_io_page;
248 bio->bi_private = page;
e1b1afa6 249 bio_set_op_attrs(bio, REQ_OP_READ, READ_SYNC | REQ_META);
4e49ea4a 250 submit_bio(bio);
9b8df98f
SW
251 wait_on_page_locked(page);
252 bio_put(bio);
253 if (!PageUptodate(page)) {
254 __free_page(page);
255 return -EIO;
256 }
257 p = kmap(page);
32e471ef 258 gfs2_sb_in(sdp, p);
9b8df98f
SW
259 kunmap(page);
260 __free_page(page);
32e471ef 261 return gfs2_check_sb(sdp, silent);
9b8df98f 262}
6bac243f 263
9b8df98f
SW
264/**
265 * gfs2_read_sb - Read super block
266 * @sdp: The GFS2 superblock
9b8df98f
SW
267 * @silent: Don't print message if mount fails
268 *
269 */
270
6bac243f 271static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
9b8df98f
SW
272{
273 u32 hash_blocks, ind_blocks, leaf_blocks;
274 u32 tmp_blocks;
275 unsigned int x;
276 int error;
277
32e471ef 278 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
9b8df98f
SW
279 if (error) {
280 if (!silent)
281 fs_err(sdp, "can't read superblock\n");
282 return error;
283 }
284
9b8df98f
SW
285 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
286 GFS2_BASIC_BLOCK_SHIFT;
47a9a527 287 sdp->sd_fsb2bb = BIT(sdp->sd_fsb2bb_shift);
9b8df98f
SW
288 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
289 sizeof(struct gfs2_dinode)) / sizeof(u64);
290 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
291 sizeof(struct gfs2_meta_header)) / sizeof(u64);
292 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
293 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
294 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
295 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
296 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
297 sizeof(struct gfs2_meta_header)) /
298 sizeof(struct gfs2_quota_change);
a68a0a35
BP
299 sdp->sd_blocks_per_bitmap = (sdp->sd_sb.sb_bsize -
300 sizeof(struct gfs2_meta_header))
301 * GFS2_NBBY; /* not the rgrp bitmap, subsequent bitmaps only */
9b8df98f
SW
302
303 /* Compute maximum reservation required to add a entry to a directory */
304
47a9a527 305 hash_blocks = DIV_ROUND_UP(sizeof(u64) * BIT(GFS2_DIR_MAX_DEPTH),
9b8df98f
SW
306 sdp->sd_jbsize);
307
308 ind_blocks = 0;
309 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
310 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
311 ind_blocks += tmp_blocks;
312 }
313
314 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
315
316 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
317
318 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
319 sizeof(struct gfs2_dinode);
320 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
321 for (x = 2;; x++) {
322 u64 space, d;
323 u32 m;
324
325 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
326 d = space;
327 m = do_div(d, sdp->sd_inptrs);
328
329 if (d != sdp->sd_heightsize[x - 1] || m)
330 break;
331 sdp->sd_heightsize[x] = space;
332 }
333 sdp->sd_max_height = x;
334 sdp->sd_heightsize[x] = ~0;
335 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
336
337 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
338 sizeof(struct gfs2_dinode);
339 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
340 for (x = 2;; x++) {
341 u64 space, d;
342 u32 m;
343
344 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
345 d = space;
346 m = do_div(d, sdp->sd_inptrs);
347
348 if (d != sdp->sd_jheightsize[x - 1] || m)
349 break;
350 sdp->sd_jheightsize[x] = space;
351 }
352 sdp->sd_max_jheight = x;
353 sdp->sd_jheightsize[x] = ~0;
354 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
355
471f3db2
BM
356 sdp->sd_max_dents_per_leaf = (sdp->sd_sb.sb_bsize -
357 sizeof(struct gfs2_leaf)) /
358 GFS2_MIN_DIRENT_SIZE;
9b8df98f
SW
359 return 0;
360}
361
b3b94faa
DT
362static int init_names(struct gfs2_sbd *sdp, int silent)
363{
b3b94faa
DT
364 char *proto, *table;
365 int error = 0;
366
367 proto = sdp->sd_args.ar_lockproto;
368 table = sdp->sd_args.ar_locktable;
369
370 /* Try to autodetect */
371
372 if (!proto[0] || !table[0]) {
32e471ef 373 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift, silent);
bb8d8a6f
SW
374 if (error)
375 return error;
3cf1e7be 376
b3b94faa 377 if (!proto[0])
3cf1e7be 378 proto = sdp->sd_sb.sb_lockproto;
b3b94faa 379 if (!table[0])
3cf1e7be 380 table = sdp->sd_sb.sb_locktable;
b3b94faa
DT
381 }
382
383 if (!table[0])
384 table = sdp->sd_vfs->s_id;
385
00377d8e
JD
386 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
387 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
b3b94faa 388
5d35e31f
DC
389 table = sdp->sd_table_name;
390 while ((table = strchr(table, '/')))
b35997d4
RP
391 *table = '_';
392
b3b94faa
DT
393 return error;
394}
395
396static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
397 int undo)
398{
b3b94faa
DT
399 int error = 0;
400
401 if (undo)
402 goto fail_trans;
403
b3b94faa
DT
404 error = gfs2_glock_nq_num(sdp,
405 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
406 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
407 mount_gh);
408 if (error) {
409 fs_err(sdp, "can't acquire mount glock: %d\n", error);
410 goto fail;
411 }
412
413 error = gfs2_glock_nq_num(sdp,
414 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
415 LM_ST_SHARED,
579b78a4 416 LM_FLAG_NOEXP | GL_EXACT,
b3b94faa
DT
417 &sdp->sd_live_gh);
418 if (error) {
419 fs_err(sdp, "can't acquire live glock: %d\n", error);
420 goto fail_mount;
421 }
422
423 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
424 CREATE, &sdp->sd_rename_gl);
425 if (error) {
426 fs_err(sdp, "can't create rename glock: %d\n", error);
427 goto fail_live;
428 }
429
24972557
BM
430 error = gfs2_glock_get(sdp, GFS2_FREEZE_LOCK, &gfs2_freeze_glops,
431 CREATE, &sdp->sd_freeze_gl);
b3b94faa
DT
432 if (error) {
433 fs_err(sdp, "can't create transaction glock: %d\n", error);
434 goto fail_rename;
435 }
b3b94faa
DT
436
437 return 0;
438
feaa7bba 439fail_trans:
24972557 440 gfs2_glock_put(sdp->sd_freeze_gl);
feaa7bba 441fail_rename:
b3b94faa 442 gfs2_glock_put(sdp->sd_rename_gl);
feaa7bba 443fail_live:
b3b94faa 444 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
feaa7bba 445fail_mount:
b3b94faa 446 gfs2_glock_dq_uninit(mount_gh);
feaa7bba 447fail:
b3b94faa
DT
448 return error;
449}
450
9b8df98f
SW
451static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
452 u64 no_addr, const char *name)
f42faf4f 453{
9b8df98f
SW
454 struct gfs2_sbd *sdp = sb->s_fs_info;
455 struct dentry *dentry;
456 struct inode *inode;
457
3ce37b2c
AG
458 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0,
459 GFS2_BLKST_FREE /* ignore */);
9b8df98f
SW
460 if (IS_ERR(inode)) {
461 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
462 return PTR_ERR(inode);
463 }
48fde701 464 dentry = d_make_root(inode);
9b8df98f
SW
465 if (!dentry) {
466 fs_err(sdp, "can't alloc %s dentry\n", name);
9b8df98f
SW
467 return -ENOMEM;
468 }
9b8df98f
SW
469 *dptr = dentry;
470 return 0;
f42faf4f
SW
471}
472
9b8df98f 473static int init_sb(struct gfs2_sbd *sdp, int silent)
b3b94faa
DT
474{
475 struct super_block *sb = sdp->sd_vfs;
476 struct gfs2_holder sb_gh;
dbb7cae2 477 u64 no_addr;
9b8df98f 478 int ret;
b3b94faa 479
9b8df98f
SW
480 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
481 LM_ST_SHARED, 0, &sb_gh);
482 if (ret) {
483 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
484 return ret;
b3b94faa 485 }
907b9bce 486
6bac243f 487 ret = gfs2_read_sb(sdp, silent);
9b8df98f
SW
488 if (ret) {
489 fs_err(sdp, "can't read superblock: %d\n", ret);
b3b94faa
DT
490 goto out;
491 }
492
493 /* Set up the buffer cache and SB for real */
e1defc4f 494 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
9b8df98f 495 ret = -EINVAL;
b3b94faa
DT
496 fs_err(sdp, "FS block size (%u) is too small for device "
497 "block size (%u)\n",
e1defc4f 498 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
b3b94faa
DT
499 goto out;
500 }
501 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
9b8df98f 502 ret = -EINVAL;
b3b94faa
DT
503 fs_err(sdp, "FS block size (%u) is too big for machine "
504 "page size (%u)\n",
505 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
506 goto out;
507 }
b3b94faa
DT
508 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
509
f42faf4f 510 /* Get the root inode */
dbb7cae2 511 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
9b8df98f
SW
512 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
513 if (ret)
f42faf4f 514 goto out;
b3b94faa 515
9b8df98f
SW
516 /* Get the master inode */
517 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
518 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
519 if (ret) {
520 dput(sdp->sd_root_dir);
521 goto out;
522 }
523 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
f42faf4f 524out:
b3b94faa 525 gfs2_glock_dq_uninit(&sb_gh);
9b8df98f 526 return ret;
b3b94faa
DT
527}
528
f057f6cd 529static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
da755fdb 530{
f057f6cd
SW
531 char *message = "FIRSTMOUNT=Done";
532 char *envp[] = { message, NULL };
e0c2a9aa
DT
533
534 fs_info(sdp, "first mount done, others may mount\n");
535
536 if (sdp->sd_lockstruct.ls_ops->lm_first_done)
537 sdp->sd_lockstruct.ls_ops->lm_first_done(sdp);
538
f057f6cd 539 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
da755fdb
SW
540}
541
b5289681
SW
542/**
543 * gfs2_jindex_hold - Grab a lock on the jindex
544 * @sdp: The GFS2 superblock
545 * @ji_gh: the holder for the jindex glock
546 *
547 * Returns: errno
548 */
549
550static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
551{
552 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
553 struct qstr name;
554 char buf[20];
555 struct gfs2_jdesc *jd;
556 int error;
557
558 name.name = buf;
559
560 mutex_lock(&sdp->sd_jindex_mutex);
561
562 for (;;) {
563 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
564 if (error)
565 break;
566
567 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
568 name.hash = gfs2_disk_hash(name.name, name.len);
569
570 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
571 if (error == -ENOENT) {
572 error = 0;
573 break;
574 }
575
576 gfs2_glock_dq_uninit(ji_gh);
577
578 if (error)
579 break;
580
581 error = -ENOMEM;
582 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
583 if (!jd)
584 break;
585
586 INIT_LIST_HEAD(&jd->extent_list);
a17d758b
BP
587 INIT_LIST_HEAD(&jd->jd_revoke_list);
588
6ecd7c2d 589 INIT_WORK(&jd->jd_work, gfs2_recover_func);
b5289681
SW
590 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
591 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
592 if (!jd->jd_inode)
593 error = -ENOENT;
594 else
595 error = PTR_ERR(jd->jd_inode);
596 kfree(jd);
597 break;
598 }
599
600 spin_lock(&sdp->sd_jindex_spin);
601 jd->jd_jid = sdp->sd_journals++;
602 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
603 spin_unlock(&sdp->sd_jindex_spin);
604 }
605
606 mutex_unlock(&sdp->sd_jindex_mutex);
607
608 return error;
609}
610
1d12d175
BP
611/**
612 * check_journal_clean - Make sure a journal is clean for a spectator mount
613 * @sdp: The GFS2 superblock
614 * @jd: The journal descriptor
615 *
616 * Returns: 0 if the journal is clean or locked, else an error
617 */
618static int check_journal_clean(struct gfs2_sbd *sdp, struct gfs2_jdesc *jd)
619{
620 int error;
621 struct gfs2_holder j_gh;
622 struct gfs2_log_header_host head;
623 struct gfs2_inode *ip;
624
625 ip = GFS2_I(jd->jd_inode);
626 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_NOEXP |
627 GL_EXACT | GL_NOCACHE, &j_gh);
628 if (error) {
629 fs_err(sdp, "Error locking journal for spectator mount.\n");
630 return -EPERM;
631 }
632 error = gfs2_jdesc_check(jd);
633 if (error) {
634 fs_err(sdp, "Error checking journal for spectator mount.\n");
635 goto out_unlock;
636 }
637 error = gfs2_find_jhead(jd, &head);
638 if (error) {
639 fs_err(sdp, "Error parsing journal for spectator mount.\n");
640 goto out_unlock;
641 }
642 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
643 error = -EPERM;
644 fs_err(sdp, "jid=%u: Journal is dirty, so the first mounter "
645 "must not be a spectator.\n", jd->jd_jid);
646 }
647
648out_unlock:
649 gfs2_glock_dq_uninit(&j_gh);
650 return error;
651}
652
b3b94faa
DT
653static int init_journal(struct gfs2_sbd *sdp, int undo)
654{
2b0143b5 655 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa 656 struct gfs2_holder ji_gh;
5c676f6d 657 struct gfs2_inode *ip;
b3b94faa
DT
658 int jindex = 1;
659 int error = 0;
660
661 if (undo) {
662 jindex = 0;
fe64d517 663 goto fail_jinode_gh;
b3b94faa
DT
664 }
665
9b8df98f 666 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
c752666c 667 if (IS_ERR(sdp->sd_jindex)) {
b3b94faa 668 fs_err(sdp, "can't lookup journal index: %d\n", error);
c752666c 669 return PTR_ERR(sdp->sd_jindex);
b3b94faa 670 }
b3b94faa
DT
671
672 /* Load in the journal index special file */
673
674 error = gfs2_jindex_hold(sdp, &ji_gh);
675 if (error) {
676 fs_err(sdp, "can't read journal index: %d\n", error);
677 goto fail;
678 }
679
0e5a9fb0 680 error = -EUSERS;
b3b94faa
DT
681 if (!gfs2_jindex_size(sdp)) {
682 fs_err(sdp, "no journals!\n");
907b9bce 683 goto fail_jindex;
b3b94faa
DT
684 }
685
686 if (sdp->sd_args.ar_spectator) {
687 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
fd041f0b 688 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
689 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
690 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
b3b94faa
DT
691 } else {
692 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
693 fs_err(sdp, "can't mount journal #%u\n",
694 sdp->sd_lockstruct.ls_jid);
695 fs_err(sdp, "there are only %u journals (0 - %u)\n",
696 gfs2_jindex_size(sdp),
697 gfs2_jindex_size(sdp) - 1);
698 goto fail_jindex;
699 }
700 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
701
feaa7bba 702 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
b3b94faa
DT
703 &gfs2_journal_glops,
704 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
705 &sdp->sd_journal_gh);
706 if (error) {
707 fs_err(sdp, "can't acquire journal glock: %d\n", error);
708 goto fail_jindex;
709 }
710
feaa7bba
SW
711 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
712 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
75be73a8 713 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
b3b94faa
DT
714 &sdp->sd_jinode_gh);
715 if (error) {
716 fs_err(sdp, "can't acquire journal inode glock: %d\n",
717 error);
718 goto fail_journal_gh;
719 }
720
721 error = gfs2_jdesc_check(sdp->sd_jdesc);
722 if (error) {
723 fs_err(sdp, "my journal (%u) is bad: %d\n",
724 sdp->sd_jdesc->jd_jid, error);
725 goto fail_jinode_gh;
726 }
fd041f0b 727 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
5e687eac
BM
728 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
729 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
da6dd40d
BP
730
731 /* Map the extents for this journal's blocks */
b50f227b 732 gfs2_map_journal_extents(sdp, sdp->sd_jdesc);
b3b94faa 733 }
63997775 734 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
b3b94faa
DT
735
736 if (sdp->sd_lockstruct.ls_first) {
737 unsigned int x;
738 for (x = 0; x < sdp->sd_journals; x++) {
1d12d175
BP
739 struct gfs2_jdesc *jd = gfs2_jdesc_find(sdp, x);
740
741 if (sdp->sd_args.ar_spectator) {
742 error = check_journal_clean(sdp, jd);
743 if (error)
744 goto fail_jinode_gh;
745 continue;
746 }
747 error = gfs2_recover_journal(jd, true);
b3b94faa
DT
748 if (error) {
749 fs_err(sdp, "error recovering journal %u: %d\n",
750 x, error);
751 goto fail_jinode_gh;
752 }
753 }
754
f057f6cd 755 gfs2_others_may_mount(sdp);
b3b94faa 756 } else if (!sdp->sd_args.ar_spectator) {
6ecd7c2d 757 error = gfs2_recover_journal(sdp->sd_jdesc, true);
b3b94faa
DT
758 if (error) {
759 fs_err(sdp, "error recovering my journal: %d\n", error);
760 goto fail_jinode_gh;
761 }
762 }
763
086cc672 764 sdp->sd_log_idle = 1;
b3b94faa
DT
765 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
766 gfs2_glock_dq_uninit(&ji_gh);
767 jindex = 0;
2e60d768 768 INIT_WORK(&sdp->sd_freeze_work, gfs2_freeze_func);
b3b94faa
DT
769 return 0;
770
a91ea69f 771fail_jinode_gh:
b3b94faa
DT
772 if (!sdp->sd_args.ar_spectator)
773 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
a91ea69f 774fail_journal_gh:
b3b94faa
DT
775 if (!sdp->sd_args.ar_spectator)
776 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
a91ea69f 777fail_jindex:
b3b94faa
DT
778 gfs2_jindex_free(sdp);
779 if (jindex)
780 gfs2_glock_dq_uninit(&ji_gh);
a91ea69f 781fail:
f42faf4f 782 iput(sdp->sd_jindex);
b3b94faa
DT
783 return error;
784}
785
56aa72d0 786static struct lock_class_key gfs2_quota_imutex_key;
b3b94faa
DT
787
788static int init_inodes(struct gfs2_sbd *sdp, int undo)
789{
b3b94faa 790 int error = 0;
2b0143b5 791 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa
DT
792
793 if (undo)
f42faf4f
SW
794 goto fail_qinode;
795
f42faf4f 796 error = init_journal(sdp, undo);
0e48e055 797 complete_all(&sdp->sd_journal_ready);
f42faf4f 798 if (error)
9b8df98f 799 goto fail;
b3b94faa 800
b3b94faa 801 /* Read in the master statfs inode */
9b8df98f 802 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
c752666c
SW
803 if (IS_ERR(sdp->sd_statfs_inode)) {
804 error = PTR_ERR(sdp->sd_statfs_inode);
b3b94faa 805 fs_err(sdp, "can't read in statfs inode: %d\n", error);
8d8291ae 806 goto fail_journal;
b3b94faa
DT
807 }
808
809 /* Read in the resource index inode */
9b8df98f 810 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
c752666c
SW
811 if (IS_ERR(sdp->sd_rindex)) {
812 error = PTR_ERR(sdp->sd_rindex);
b3b94faa
DT
813 fs_err(sdp, "can't get resource index inode: %d\n", error);
814 goto fail_statfs;
815 }
cf45b752 816 sdp->sd_rindex_uptodate = 0;
b3b94faa
DT
817
818 /* Read in the quota inode */
9b8df98f 819 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
c752666c
SW
820 if (IS_ERR(sdp->sd_quota_inode)) {
821 error = PTR_ERR(sdp->sd_quota_inode);
b3b94faa
DT
822 fs_err(sdp, "can't get quota file inode: %d\n", error);
823 goto fail_rindex;
824 }
56aa72d0
JK
825 /*
826 * i_mutex on quota files is special. Since this inode is hidden system
827 * file, we are safe to define locking ourselves.
828 */
9902af79 829 lockdep_set_class(&sdp->sd_quota_inode->i_rwsem,
56aa72d0 830 &gfs2_quota_imutex_key);
a365fbf3
SW
831
832 error = gfs2_rindex_update(sdp);
833 if (error)
834 goto fail_qinode;
835
b3b94faa
DT
836 return 0;
837
f42faf4f
SW
838fail_qinode:
839 iput(sdp->sd_quota_inode);
f42faf4f 840fail_rindex:
b3b94faa 841 gfs2_clear_rgrpd(sdp);
f42faf4f 842 iput(sdp->sd_rindex);
f42faf4f
SW
843fail_statfs:
844 iput(sdp->sd_statfs_inode);
f42faf4f
SW
845fail_journal:
846 init_journal(sdp, UNDO);
f42faf4f 847fail:
b3b94faa
DT
848 return error;
849}
850
851static int init_per_node(struct gfs2_sbd *sdp, int undo)
852{
f42faf4f 853 struct inode *pn = NULL;
b3b94faa
DT
854 char buf[30];
855 int error = 0;
5c676f6d 856 struct gfs2_inode *ip;
2b0143b5 857 struct inode *master = d_inode(sdp->sd_master_dir);
b3b94faa
DT
858
859 if (sdp->sd_args.ar_spectator)
860 return 0;
861
862 if (undo)
863 goto fail_qc_gh;
864
9b8df98f 865 pn = gfs2_lookup_simple(master, "per_node");
c752666c
SW
866 if (IS_ERR(pn)) {
867 error = PTR_ERR(pn);
b3b94faa
DT
868 fs_err(sdp, "can't find per_node directory: %d\n", error);
869 return error;
870 }
871
b3b94faa 872 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
873 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
874 if (IS_ERR(sdp->sd_sc_inode)) {
875 error = PTR_ERR(sdp->sd_sc_inode);
b3b94faa 876 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
8d8291ae 877 goto fail;
b3b94faa
DT
878 }
879
b3b94faa 880 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
c752666c
SW
881 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
882 if (IS_ERR(sdp->sd_qc_inode)) {
883 error = PTR_ERR(sdp->sd_qc_inode);
b3b94faa
DT
884 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
885 goto fail_ut_i;
886 }
887
f42faf4f 888 iput(pn);
b3b94faa
DT
889 pn = NULL;
890
feaa7bba 891 ip = GFS2_I(sdp->sd_sc_inode);
8d8291ae 892 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
893 &sdp->sd_sc_gh);
894 if (error) {
895 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
8d8291ae 896 goto fail_qc_i;
b3b94faa
DT
897 }
898
feaa7bba 899 ip = GFS2_I(sdp->sd_qc_inode);
8d8291ae 900 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
b3b94faa
DT
901 &sdp->sd_qc_gh);
902 if (error) {
903 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
904 goto fail_ut_gh;
905 }
906
907 return 0;
908
a91ea69f 909fail_qc_gh:
b3b94faa 910 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
a91ea69f 911fail_ut_gh:
b3b94faa 912 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
a91ea69f 913fail_qc_i:
f42faf4f 914 iput(sdp->sd_qc_inode);
a91ea69f 915fail_ut_i:
f42faf4f 916 iput(sdp->sd_sc_inode);
a91ea69f 917fail:
6fde2242 918 iput(pn);
b3b94faa
DT
919 return error;
920}
921
f057f6cd
SW
922static const match_table_t nolock_tokens = {
923 { Opt_jid, "jid=%d\n", },
924 { Opt_err, NULL },
925};
926
927static 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
942static 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;
f057f6cd
SW
949 char *o, *options;
950 int ret;
da755fdb 951
f057f6cd
SW
952 if (!strcmp("lock_nolock", proto)) {
953 lm = &nolock_ops;
954 sdp->sd_args.ar_localflocks = 1;
f057f6cd
SW
955#ifdef CONFIG_GFS2_FS_LOCKING_DLM
956 } else if (!strcmp("lock_dlm", proto)) {
957 lm = &gfs2_dlm_ops;
958#endif
959 } else {
d77d1b58 960 pr_info("can't find protocol %s\n", proto);
f057f6cd
SW
961 return -ENOENT;
962 }
da755fdb
SW
963
964 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
965
f057f6cd
SW
966 ls->ls_ops = lm;
967 ls->ls_first = 1;
da755fdb 968
f057f6cd
SW
969 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
970 substring_t tmp[MAX_OPT_ARGS];
971 int token, option;
972
973 if (!o || !*o)
974 continue;
975
976 token = match_token(o, *lm->lm_tokens, tmp);
977 switch (token) {
978 case Opt_jid:
979 ret = match_int(&tmp[0], &option);
980 if (ret || option < 0)
981 goto hostdata_error;
ba6e9364
SW
982 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
983 ls->ls_jid = option;
f057f6cd
SW
984 break;
985 case Opt_id:
4875647a 986 case Opt_nodir:
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;
f057f6cd
SW
995 case Opt_err:
996 default:
997hostdata_error:
998 fs_info(sdp, "unknown hostdata (%s)\n", o);
999 return -EINVAL;
1000 }
da755fdb
SW
1001 }
1002
f057f6cd
SW
1003 if (lm->lm_mount == NULL) {
1004 fs_info(sdp, "Now mounting FS...\n");
19237039 1005 complete_all(&sdp->sd_locking_init);
f057f6cd 1006 return 0;
da755fdb 1007 }
e0c2a9aa 1008 ret = lm->lm_mount(sdp, table);
f057f6cd
SW
1009 if (ret == 0)
1010 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
19237039 1011 complete_all(&sdp->sd_locking_init);
f057f6cd 1012 return ret;
da755fdb
SW
1013}
1014
1015void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1016{
f057f6cd
SW
1017 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1018 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1019 lm->lm_unmount)
1020 lm->lm_unmount(sdp);
da755fdb 1021}
b3b94faa 1022
ba6e9364
SW
1023static int wait_on_journal(struct gfs2_sbd *sdp)
1024{
ba6e9364
SW
1025 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1026 return 0;
1027
74316201
N
1028 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, TASK_INTERRUPTIBLE)
1029 ? -EINTR : 0;
ba6e9364
SW
1030}
1031
8633ecfa
SW
1032void gfs2_online_uevent(struct gfs2_sbd *sdp)
1033{
1034 struct super_block *sb = sdp->sd_vfs;
1035 char ro[20];
1036 char spectator[20];
1037 char *envp[] = { ro, spectator, NULL };
1038 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1039 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1040 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1041}
1042
b3b94faa
DT
1043/**
1044 * fill_super - Read in superblock
1045 * @sb: The VFS superblock
1046 * @data: Mount options
1047 * @silent: Don't complain if it's not a GFS2 filesystem
1048 *
1049 * Returns: errno
1050 */
1051
f55073ff 1052static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
b3b94faa
DT
1053{
1054 struct gfs2_sbd *sdp;
1055 struct gfs2_holder mount_gh;
1056 int error;
1057
1058 sdp = init_sbd(sb);
1059 if (!sdp) {
d77d1b58 1060 pr_warn("can't alloc struct gfs2_sbd\n");
b3b94faa
DT
1061 return -ENOMEM;
1062 }
f55073ff 1063 sdp->sd_args = *args;
b3b94faa 1064
fe64d517 1065 if (sdp->sd_args.ar_spectator) {
6f04c1c7 1066 sb->s_flags |= MS_RDONLY;
e8ca5cc5 1067 set_bit(SDF_RORECOVERY, &sdp->sd_flags);
fe64d517 1068 }
6f04c1c7
SW
1069 if (sdp->sd_args.ar_posix_acl)
1070 sb->s_flags |= MS_POSIXACL;
f25934c5
CH
1071 if (sdp->sd_args.ar_nobarrier)
1072 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
6f04c1c7 1073
9964afbb 1074 sb->s_flags |= MS_NOSEC;
719ee344
SW
1075 sb->s_magic = GFS2_MAGIC;
1076 sb->s_op = &gfs2_super_ops;
41ced6dc 1077 sb->s_d_op = &gfs2_dops;
719ee344 1078 sb->s_export_op = &gfs2_export_ops;
40b78a32 1079 sb->s_xattr = gfs2_xattr_handlers;
cc632e7f 1080 sb->s_qcop = &gfs2_quotactl_ops;
de3b08d3 1081 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cc632e7f 1082 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
719ee344
SW
1083 sb->s_time_gran = 1;
1084 sb->s_maxbytes = MAX_LFS_FILESIZE;
419c93e0
SW
1085
1086 /* Set up the buffer cache and fill in some fake block size values
1087 to allow us to read-in the on-disk superblock. */
1088 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1089 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1090 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1091 GFS2_BASIC_BLOCK_SHIFT;
47a9a527 1092 sdp->sd_fsb2bb = BIT(sdp->sd_fsb2bb_shift);
b3b94faa 1093
5e687eac 1094 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
3d3c10f2
BM
1095 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1096 if (sdp->sd_args.ar_statfs_quantum) {
1097 sdp->sd_tune.gt_statfs_slow = 0;
1098 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
32e471ef 1099 } else {
3d3c10f2
BM
1100 sdp->sd_tune.gt_statfs_slow = 1;
1101 sdp->sd_tune.gt_statfs_quantum = 30;
1102 }
48c2b613 1103
b3b94faa 1104 error = init_names(sdp, silent);
0d515210
BP
1105 if (error) {
1106 /* In this case, we haven't initialized sysfs, so we have to
1107 manually free the sdp. */
1108 free_percpu(sdp->sd_lkstats);
1109 kfree(sdp);
1110 sb->s_fs_info = NULL;
1111 return error;
1112 }
b3b94faa 1113
e0c2a9aa
DT
1114 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
1115
b3b94faa 1116 error = gfs2_sys_fs_add(sdp);
0d515210
BP
1117 /*
1118 * If we hit an error here, gfs2_sys_fs_add will have called function
1119 * kobject_put which causes the sysfs usage count to go to zero, which
1120 * causes sysfs to call function gfs2_sbd_release, which frees sdp.
1121 * Subsequent error paths here will call gfs2_sys_fs_del, which also
1122 * kobject_put to free sdp.
1123 */
b3b94faa 1124 if (error)
0d515210
BP
1125 return error;
1126
1127 gfs2_create_debugfs_file(sdp);
b3b94faa
DT
1128
1129 error = gfs2_lm_mount(sdp, silent);
1130 if (error)
0d515210 1131 goto fail_debug;
b3b94faa
DT
1132
1133 error = init_locking(sdp, &mount_gh, DO);
1134 if (error)
1135 goto fail_lm;
1136
9b8df98f 1137 error = init_sb(sdp, silent);
b3b94faa
DT
1138 if (error)
1139 goto fail_locking;
b3b94faa 1140
ba6e9364
SW
1141 error = wait_on_journal(sdp);
1142 if (error)
1143 goto fail_sb;
1144
feb47ca9
SW
1145 /*
1146 * If user space has failed to join the cluster or some similar
1147 * failure has occurred, then the journal id will contain a
1148 * negative (error) number. This will then be returned to the
1149 * caller (of the mount syscall). We do this even for spectator
1150 * mounts (which just write a jid of 0 to indicate "ok" even though
1151 * the jid is unused in the spectator case)
1152 */
1153 if (sdp->sd_lockstruct.ls_jid < 0) {
1154 error = sdp->sd_lockstruct.ls_jid;
1155 sdp->sd_lockstruct.ls_jid = 0;
1156 goto fail_sb;
1157 }
1158
e0c2a9aa
DT
1159 if (sdp->sd_args.ar_spectator)
1160 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
1161 sdp->sd_table_name);
1162 else
1163 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
1164 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
1165
b3b94faa
DT
1166 error = init_inodes(sdp, DO);
1167 if (error)
f42faf4f 1168 goto fail_sb;
b3b94faa
DT
1169
1170 error = init_per_node(sdp, DO);
1171 if (error)
1172 goto fail_inodes;
1173
1174 error = gfs2_statfs_init(sdp);
1175 if (error) {
1176 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1177 goto fail_per_node;
1178 }
1179
b3b94faa
DT
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);
8ad151c2 1184 goto fail_per_node;
b3b94faa
DT
1185 }
1186 }
1187
1188 gfs2_glock_dq_uninit(&mount_gh);
8633ecfa 1189 gfs2_online_uevent(sdp);
b3b94faa
DT
1190 return 0;
1191
a91ea69f 1192fail_per_node:
b3b94faa 1193 init_per_node(sdp, UNDO);
a91ea69f 1194fail_inodes:
b3b94faa 1195 init_inodes(sdp, UNDO);
a91ea69f 1196fail_sb:
9b8df98f
SW
1197 if (sdp->sd_root_dir)
1198 dput(sdp->sd_root_dir);
1199 if (sdp->sd_master_dir)
1200 dput(sdp->sd_master_dir);
e7c8707e
SW
1201 if (sb->s_root)
1202 dput(sb->s_root);
9b8df98f 1203 sb->s_root = NULL;
a91ea69f 1204fail_locking:
b3b94faa 1205 init_locking(sdp, &mount_gh, UNDO);
a91ea69f 1206fail_lm:
0e48e055 1207 complete_all(&sdp->sd_journal_ready);
fefc03bf 1208 gfs2_gl_hash_clear(sdp);
b3b94faa 1209 gfs2_lm_unmount(sdp);
0d515210 1210fail_debug:
5f882096 1211 gfs2_delete_debugfs_file(sdp);
a245769f 1212 free_percpu(sdp->sd_lkstats);
0d515210
BP
1213 /* gfs2_sys_fs_del must be the last thing we do, since it causes
1214 * sysfs to call function gfs2_sbd_release, which frees sdp. */
1215 gfs2_sys_fs_del(sdp);
5c676f6d 1216 sb->s_fs_info = NULL;
b3b94faa
DT
1217 return error;
1218}
1219
f55073ff 1220static 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 1233static 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
1253static 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 }
9249e17f 1281 s = sget(fs_type, test_gfs2_super, set_gfs2_super, flags, bdev);
f55073ff
SW
1282 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1283 error = PTR_ERR(s);
1284 if (IS_ERR(s))
1285 goto error_bdev;
1286
dfe5b9ad
SW
1287 if (s->s_root) {
1288 /*
1289 * s_umount nests inside bd_mutex during
1290 * __invalidate_device(). blkdev_put() acquires
1291 * bd_mutex and can't be called under s_umount. Drop
1292 * s_umount temporarily. This is safe as we're
1293 * holding an active reference.
1294 */
1295 up_write(&s->s_umount);
d4d77629 1296 blkdev_put(bdev, mode);
dfe5b9ad 1297 down_write(&s->s_umount);
6de20eb0
AP
1298 } else {
1299 /* s_mode must be set before deactivate_locked_super calls */
1300 s->s_mode = mode;
dfe5b9ad 1301 }
8bcbbf00 1302
f55073ff
SW
1303 memset(&args, 0, sizeof(args));
1304 args.ar_quota = GFS2_QUOTA_DEFAULT;
1305 args.ar_data = GFS2_DATA_DEFAULT;
5e687eac 1306 args.ar_commit = 30;
3d3c10f2
BM
1307 args.ar_statfs_quantum = 30;
1308 args.ar_quota_quantum = 60;
f55073ff
SW
1309 args.ar_errors = GFS2_ERRORS_DEFAULT;
1310
1311 error = gfs2_mount_args(&args, data);
1312 if (error) {
d77d1b58 1313 pr_warn("can't parse mount arguments\n");
8bcbbf00 1314 goto error_super;
f55073ff
SW
1315 }
1316
1317 if (s->s_root) {
1318 error = -EBUSY;
1319 if ((flags ^ s->s_flags) & MS_RDONLY)
1320 goto error_super;
f55073ff 1321 } else {
a1c6f057 1322 snprintf(s->s_id, sizeof(s->s_id), "%pg", bdev);
f55073ff
SW
1323 sb_set_blocksize(s, block_size(bdev));
1324 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
8bcbbf00
AV
1325 if (error)
1326 goto error_super;
f55073ff
SW
1327 s->s_flags |= MS_ACTIVE;
1328 bdev->bd_super = s;
1329 }
1330
1331 sdp = s->s_fs_info;
f55073ff 1332 if (args.ar_meta)
8bcbbf00 1333 return dget(sdp->sd_master_dir);
f55073ff 1334 else
8bcbbf00 1335 return dget(sdp->sd_root_dir);
f55073ff
SW
1336
1337error_super:
1338 deactivate_locked_super(s);
8bcbbf00 1339 return ERR_PTR(error);
f55073ff 1340error_bdev:
d4d77629 1341 blkdev_put(bdev, mode);
8bcbbf00 1342 return ERR_PTR(error);
f55073ff
SW
1343}
1344
f6d03139
SW
1345static int set_meta_super(struct super_block *s, void *ptr)
1346{
1347 return -EINVAL;
1348}
1349
8bcbbf00
AV
1350static struct dentry *gfs2_mount_meta(struct file_system_type *fs_type,
1351 int flags, const char *dev_name, void *data)
86384605 1352{
f6d03139 1353 struct super_block *s;
003dec89 1354 struct gfs2_sbd *sdp;
e24977d4 1355 struct path path;
86384605 1356 int error;
907b9bce 1357
e24977d4 1358 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
86384605 1359 if (error) {
d77d1b58
JP
1360 pr_warn("path_lookup on %s returned error %d\n",
1361 dev_name, error);
8bcbbf00 1362 return ERR_PTR(error);
86384605 1363 }
9249e17f 1364 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super, flags,
fc64005c 1365 path.dentry->d_sb->s_bdev);
e24977d4 1366 path_put(&path);
f6d03139 1367 if (IS_ERR(s)) {
d77d1b58 1368 pr_warn("gfs2 mount does not exist\n");
8bcbbf00 1369 return ERR_CAST(s);
86384605 1370 }
f55073ff
SW
1371 if ((flags ^ s->s_flags) & MS_RDONLY) {
1372 deactivate_locked_super(s);
8bcbbf00 1373 return ERR_PTR(-EBUSY);
f55073ff 1374 }
f6d03139 1375 sdp = s->s_fs_info;
8bcbbf00 1376 return dget(sdp->sd_master_dir);
b3b94faa
DT
1377}
1378
419c93e0 1379static void gfs2_kill_sb(struct super_block *sb)
86384605
AD
1380{
1381 struct gfs2_sbd *sdp = sb->s_fs_info;
88a19ad0
SW
1382
1383 if (sdp == NULL) {
1384 kill_block_super(sb);
1385 return;
3af165ac 1386 }
88a19ad0 1387
24972557 1388 gfs2_log_flush(sdp, NULL, SYNC_FLUSH);
88a19ad0
SW
1389 dput(sdp->sd_root_dir);
1390 dput(sdp->sd_master_dir);
1391 sdp->sd_root_dir = NULL;
1392 sdp->sd_master_dir = NULL;
fefc03bf 1393 shrink_dcache_sb(sb);
88a19ad0 1394 gfs2_delete_debugfs_file(sdp);
a245769f 1395 free_percpu(sdp->sd_lkstats);
0d515210 1396 kill_block_super(sb);
86384605
AD
1397}
1398
b3b94faa
DT
1399struct file_system_type gfs2_fs_type = {
1400 .name = "gfs2",
1401 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1402 .mount = gfs2_mount,
419c93e0
SW
1403 .kill_sb = gfs2_kill_sb,
1404 .owner = THIS_MODULE,
1405};
7f78e035 1406MODULE_ALIAS_FS("gfs2");
419c93e0
SW
1407
1408struct file_system_type gfs2meta_fs_type = {
1409 .name = "gfs2meta",
1410 .fs_flags = FS_REQUIRES_DEV,
8bcbbf00 1411 .mount = gfs2_mount_meta,
b3b94faa
DT
1412 .owner = THIS_MODULE,
1413};
7f78e035 1414MODULE_ALIAS_FS("gfs2meta");