]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - fs/f2fs/super.c
f2fs: reduce region of cp_rwsem covered in f2fs_do_collapse
[mirror_ubuntu-hirsute-kernel.git] / fs / f2fs / super.c
CommitLineData
0a8165d7 1/*
aff063e2
JK
2 * fs/f2fs/super.c
3 *
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License version 2 as
9 * published by the Free Software Foundation.
10 */
11#include <linux/module.h>
12#include <linux/init.h>
13#include <linux/fs.h>
14#include <linux/statfs.h>
aff063e2
JK
15#include <linux/buffer_head.h>
16#include <linux/backing-dev.h>
17#include <linux/kthread.h>
18#include <linux/parser.h>
19#include <linux/mount.h>
20#include <linux/seq_file.h>
5e176d54 21#include <linux/proc_fs.h>
aff063e2
JK
22#include <linux/random.h>
23#include <linux/exportfs.h>
d3ee456d 24#include <linux/blkdev.h>
aff063e2 25#include <linux/f2fs_fs.h>
b59d0bae 26#include <linux/sysfs.h>
aff063e2
JK
27
28#include "f2fs.h"
29#include "node.h"
5ec4e49f 30#include "segment.h"
aff063e2 31#include "xattr.h"
b59d0bae 32#include "gc.h"
db9f7c1a 33#include "trace.h"
aff063e2 34
a2a4a7e4
NJ
35#define CREATE_TRACE_POINTS
36#include <trace/events/f2fs.h>
37
5e176d54 38static struct proc_dir_entry *f2fs_proc_root;
aff063e2 39static struct kmem_cache *f2fs_inode_cachep;
b59d0bae 40static struct kset *f2fs_kset;
aff063e2 41
2658e50d
JK
42/* f2fs-wide shrinker description */
43static struct shrinker f2fs_shrinker_info = {
44 .scan_objects = f2fs_shrink_scan,
45 .count_objects = f2fs_shrink_count,
46 .seeks = DEFAULT_SEEKS,
47};
48
aff063e2 49enum {
696c018c 50 Opt_gc_background,
aff063e2 51 Opt_disable_roll_forward,
2d834bf9 52 Opt_norecovery,
aff063e2
JK
53 Opt_discard,
54 Opt_noheap,
4058c511 55 Opt_user_xattr,
aff063e2 56 Opt_nouser_xattr,
4058c511 57 Opt_acl,
aff063e2
JK
58 Opt_noacl,
59 Opt_active_logs,
60 Opt_disable_ext_identify,
444c580f 61 Opt_inline_xattr,
8274de77 62 Opt_inline_data,
5efd3c6f 63 Opt_inline_dentry,
6b4afdd7 64 Opt_flush_merge,
0f7b2abd 65 Opt_nobarrier,
d5053a34 66 Opt_fastboot,
89672159 67 Opt_extent_cache,
7daaea25 68 Opt_noextent_cache,
75342797 69 Opt_noinline_data,
aff063e2
JK
70 Opt_err,
71};
72
73static match_table_t f2fs_tokens = {
696c018c 74 {Opt_gc_background, "background_gc=%s"},
aff063e2 75 {Opt_disable_roll_forward, "disable_roll_forward"},
2d834bf9 76 {Opt_norecovery, "norecovery"},
aff063e2
JK
77 {Opt_discard, "discard"},
78 {Opt_noheap, "no_heap"},
4058c511 79 {Opt_user_xattr, "user_xattr"},
aff063e2 80 {Opt_nouser_xattr, "nouser_xattr"},
4058c511 81 {Opt_acl, "acl"},
aff063e2
JK
82 {Opt_noacl, "noacl"},
83 {Opt_active_logs, "active_logs=%u"},
84 {Opt_disable_ext_identify, "disable_ext_identify"},
444c580f 85 {Opt_inline_xattr, "inline_xattr"},
8274de77 86 {Opt_inline_data, "inline_data"},
5efd3c6f 87 {Opt_inline_dentry, "inline_dentry"},
6b4afdd7 88 {Opt_flush_merge, "flush_merge"},
0f7b2abd 89 {Opt_nobarrier, "nobarrier"},
d5053a34 90 {Opt_fastboot, "fastboot"},
89672159 91 {Opt_extent_cache, "extent_cache"},
7daaea25 92 {Opt_noextent_cache, "noextent_cache"},
75342797 93 {Opt_noinline_data, "noinline_data"},
aff063e2
JK
94 {Opt_err, NULL},
95};
96
b59d0bae 97/* Sysfs support for f2fs */
ea91e9b0
JK
98enum {
99 GC_THREAD, /* struct f2fs_gc_thread */
100 SM_INFO, /* struct f2fs_sm_info */
cdfc41c1 101 NM_INFO, /* struct f2fs_nm_info */
b1c57c1c 102 F2FS_SBI, /* struct f2fs_sb_info */
ea91e9b0
JK
103};
104
b59d0bae
NJ
105struct f2fs_attr {
106 struct attribute attr;
107 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
108 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
109 const char *, size_t);
ea91e9b0 110 int struct_type;
b59d0bae
NJ
111 int offset;
112};
113
ea91e9b0
JK
114static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
115{
116 if (struct_type == GC_THREAD)
117 return (unsigned char *)sbi->gc_thread;
118 else if (struct_type == SM_INFO)
119 return (unsigned char *)SM_I(sbi);
cdfc41c1
JK
120 else if (struct_type == NM_INFO)
121 return (unsigned char *)NM_I(sbi);
b1c57c1c
JK
122 else if (struct_type == F2FS_SBI)
123 return (unsigned char *)sbi;
ea91e9b0
JK
124 return NULL;
125}
126
b59d0bae
NJ
127static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
128 struct f2fs_sb_info *sbi, char *buf)
129{
ea91e9b0 130 unsigned char *ptr = NULL;
b59d0bae
NJ
131 unsigned int *ui;
132
ea91e9b0
JK
133 ptr = __struct_ptr(sbi, a->struct_type);
134 if (!ptr)
b59d0bae
NJ
135 return -EINVAL;
136
ea91e9b0 137 ui = (unsigned int *)(ptr + a->offset);
b59d0bae
NJ
138
139 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
140}
141
142static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
143 struct f2fs_sb_info *sbi,
144 const char *buf, size_t count)
145{
ea91e9b0 146 unsigned char *ptr;
b59d0bae
NJ
147 unsigned long t;
148 unsigned int *ui;
149 ssize_t ret;
150
ea91e9b0
JK
151 ptr = __struct_ptr(sbi, a->struct_type);
152 if (!ptr)
b59d0bae
NJ
153 return -EINVAL;
154
ea91e9b0 155 ui = (unsigned int *)(ptr + a->offset);
b59d0bae
NJ
156
157 ret = kstrtoul(skip_spaces(buf), 0, &t);
158 if (ret < 0)
159 return ret;
160 *ui = t;
161 return count;
162}
163
164static ssize_t f2fs_attr_show(struct kobject *kobj,
165 struct attribute *attr, char *buf)
166{
167 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
168 s_kobj);
169 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
170
171 return a->show ? a->show(a, sbi, buf) : 0;
172}
173
174static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
175 const char *buf, size_t len)
176{
177 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
178 s_kobj);
179 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
180
181 return a->store ? a->store(a, sbi, buf, len) : 0;
182}
183
184static void f2fs_sb_release(struct kobject *kobj)
185{
186 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
187 s_kobj);
188 complete(&sbi->s_kobj_unregister);
189}
190
ea91e9b0 191#define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
b59d0bae
NJ
192static struct f2fs_attr f2fs_attr_##_name = { \
193 .attr = {.name = __stringify(_name), .mode = _mode }, \
194 .show = _show, \
195 .store = _store, \
ea91e9b0
JK
196 .struct_type = _struct_type, \
197 .offset = _offset \
b59d0bae
NJ
198}
199
ea91e9b0
JK
200#define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
201 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
202 f2fs_sbi_show, f2fs_sbi_store, \
203 offsetof(struct struct_name, elname))
b59d0bae 204
ea91e9b0
JK
205F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
206F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
207F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
208F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
209F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
7ac8c3b0 210F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
bba681cb 211F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
216fbd64
JK
212F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
213F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
c1ce1b02 214F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
cdfc41c1 215F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
b1c57c1c 216F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
ab9fa662 217F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
b59d0bae
NJ
218
219#define ATTR_LIST(name) (&f2fs_attr_##name.attr)
220static struct attribute *f2fs_attrs[] = {
221 ATTR_LIST(gc_min_sleep_time),
222 ATTR_LIST(gc_max_sleep_time),
223 ATTR_LIST(gc_no_gc_sleep_time),
d2dc095f 224 ATTR_LIST(gc_idle),
ea91e9b0 225 ATTR_LIST(reclaim_segments),
7ac8c3b0 226 ATTR_LIST(max_small_discards),
bba681cb 227 ATTR_LIST(batched_trim_sections),
216fbd64
JK
228 ATTR_LIST(ipu_policy),
229 ATTR_LIST(min_ipu_util),
c1ce1b02 230 ATTR_LIST(min_fsync_blocks),
b1c57c1c 231 ATTR_LIST(max_victim_search),
ab9fa662 232 ATTR_LIST(dir_level),
cdfc41c1 233 ATTR_LIST(ram_thresh),
b59d0bae
NJ
234 NULL,
235};
236
237static const struct sysfs_ops f2fs_attr_ops = {
238 .show = f2fs_attr_show,
239 .store = f2fs_attr_store,
240};
241
242static struct kobj_type f2fs_ktype = {
243 .default_attrs = f2fs_attrs,
244 .sysfs_ops = &f2fs_attr_ops,
245 .release = f2fs_sb_release,
246};
247
a07ef784
NJ
248void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
249{
250 struct va_format vaf;
251 va_list args;
252
253 va_start(args, fmt);
254 vaf.fmt = fmt;
255 vaf.va = &args;
256 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
257 va_end(args);
258}
259
aff063e2
JK
260static void init_once(void *foo)
261{
262 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
263
aff063e2
JK
264 inode_init_once(&fi->vfs_inode);
265}
266
696c018c
NJ
267static int parse_options(struct super_block *sb, char *options)
268{
269 struct f2fs_sb_info *sbi = F2FS_SB(sb);
09d54cdd 270 struct request_queue *q;
696c018c
NJ
271 substring_t args[MAX_OPT_ARGS];
272 char *p, *name;
273 int arg = 0;
274
275 if (!options)
276 return 0;
277
278 while ((p = strsep(&options, ",")) != NULL) {
279 int token;
280 if (!*p)
281 continue;
282 /*
283 * Initialize args struct so we know whether arg was
284 * found; some options take optional arguments.
285 */
286 args[0].to = args[0].from = NULL;
287 token = match_token(p, f2fs_tokens, args);
288
289 switch (token) {
290 case Opt_gc_background:
291 name = match_strdup(&args[0]);
292
293 if (!name)
294 return -ENOMEM;
04c09388 295 if (strlen(name) == 2 && !strncmp(name, "on", 2))
696c018c 296 set_opt(sbi, BG_GC);
04c09388 297 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
696c018c
NJ
298 clear_opt(sbi, BG_GC);
299 else {
300 kfree(name);
301 return -EINVAL;
302 }
303 kfree(name);
304 break;
305 case Opt_disable_roll_forward:
306 set_opt(sbi, DISABLE_ROLL_FORWARD);
307 break;
2d834bf9
JK
308 case Opt_norecovery:
309 /* this option mounts f2fs with ro */
310 set_opt(sbi, DISABLE_ROLL_FORWARD);
311 if (!f2fs_readonly(sb))
312 return -EINVAL;
313 break;
696c018c 314 case Opt_discard:
09d54cdd
CY
315 q = bdev_get_queue(sb->s_bdev);
316 if (blk_queue_discard(q)) {
317 set_opt(sbi, DISCARD);
318 } else {
319 f2fs_msg(sb, KERN_WARNING,
320 "mounting with \"discard\" option, but "
321 "the device does not support discard");
322 }
696c018c
NJ
323 break;
324 case Opt_noheap:
325 set_opt(sbi, NOHEAP);
326 break;
327#ifdef CONFIG_F2FS_FS_XATTR
4058c511
KA
328 case Opt_user_xattr:
329 set_opt(sbi, XATTR_USER);
330 break;
696c018c
NJ
331 case Opt_nouser_xattr:
332 clear_opt(sbi, XATTR_USER);
333 break;
444c580f
JK
334 case Opt_inline_xattr:
335 set_opt(sbi, INLINE_XATTR);
336 break;
696c018c 337#else
4058c511
KA
338 case Opt_user_xattr:
339 f2fs_msg(sb, KERN_INFO,
340 "user_xattr options not supported");
341 break;
696c018c
NJ
342 case Opt_nouser_xattr:
343 f2fs_msg(sb, KERN_INFO,
344 "nouser_xattr options not supported");
345 break;
444c580f
JK
346 case Opt_inline_xattr:
347 f2fs_msg(sb, KERN_INFO,
348 "inline_xattr options not supported");
349 break;
696c018c
NJ
350#endif
351#ifdef CONFIG_F2FS_FS_POSIX_ACL
4058c511
KA
352 case Opt_acl:
353 set_opt(sbi, POSIX_ACL);
354 break;
696c018c
NJ
355 case Opt_noacl:
356 clear_opt(sbi, POSIX_ACL);
357 break;
358#else
4058c511
KA
359 case Opt_acl:
360 f2fs_msg(sb, KERN_INFO, "acl options not supported");
361 break;
696c018c
NJ
362 case Opt_noacl:
363 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
364 break;
365#endif
366 case Opt_active_logs:
367 if (args->from && match_int(args, &arg))
368 return -EINVAL;
369 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
370 return -EINVAL;
371 sbi->active_logs = arg;
372 break;
373 case Opt_disable_ext_identify:
374 set_opt(sbi, DISABLE_EXT_IDENTIFY);
375 break;
8274de77
HL
376 case Opt_inline_data:
377 set_opt(sbi, INLINE_DATA);
378 break;
5efd3c6f
CY
379 case Opt_inline_dentry:
380 set_opt(sbi, INLINE_DENTRY);
381 break;
6b4afdd7
JK
382 case Opt_flush_merge:
383 set_opt(sbi, FLUSH_MERGE);
384 break;
0f7b2abd
JK
385 case Opt_nobarrier:
386 set_opt(sbi, NOBARRIER);
387 break;
d5053a34
JK
388 case Opt_fastboot:
389 set_opt(sbi, FASTBOOT);
390 break;
89672159
CY
391 case Opt_extent_cache:
392 set_opt(sbi, EXTENT_CACHE);
393 break;
7daaea25
JK
394 case Opt_noextent_cache:
395 clear_opt(sbi, EXTENT_CACHE);
396 break;
75342797
WL
397 case Opt_noinline_data:
398 clear_opt(sbi, INLINE_DATA);
399 break;
696c018c
NJ
400 default:
401 f2fs_msg(sb, KERN_ERR,
402 "Unrecognized mount option \"%s\" or missing value",
403 p);
404 return -EINVAL;
405 }
406 }
407 return 0;
408}
409
aff063e2
JK
410static struct inode *f2fs_alloc_inode(struct super_block *sb)
411{
412 struct f2fs_inode_info *fi;
413
a0acdfe0 414 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
aff063e2
JK
415 if (!fi)
416 return NULL;
417
418 init_once((void *) fi);
419
434720fa 420 /* Initialize f2fs-specific inode info */
aff063e2 421 fi->vfs_inode.i_version = 1;
a7ffdbe2 422 atomic_set(&fi->dirty_pages, 0);
aff063e2
JK
423 fi->i_current_depth = 1;
424 fi->i_advise = 0;
d928bfbf 425 init_rwsem(&fi->i_sem);
34ba94ba 426 INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
88b88a66
JK
427 INIT_LIST_HEAD(&fi->inmem_pages);
428 mutex_init(&fi->inmem_lock);
aff063e2
JK
429
430 set_inode_flag(fi, FI_NEW_INODE);
431
444c580f
JK
432 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
433 set_inode_flag(fi, FI_INLINE_XATTR);
434
ab9fa662
JK
435 /* Will be used by directory only */
436 fi->i_dir_level = F2FS_SB(sb)->dir_level;
437
57e5055b
JK
438#ifdef CONFIG_F2FS_FS_ENCRYPTION
439 fi->i_crypt_info = NULL;
440#endif
aff063e2
JK
441 return &fi->vfs_inode;
442}
443
531ad7d5
JK
444static int f2fs_drop_inode(struct inode *inode)
445{
446 /*
447 * This is to avoid a deadlock condition like below.
448 * writeback_single_inode(inode)
449 * - f2fs_write_data_page
450 * - f2fs_gc -> iput -> evict
451 * - inode_wait_for_writeback(inode)
452 */
06e1bc05
JK
453 if (!inode_unhashed(inode) && inode->i_state & I_SYNC) {
454 if (!inode->i_nlink && !is_bad_inode(inode)) {
3e72f721
JK
455 /* to avoid evict_inode call simultaneously */
456 atomic_inc(&inode->i_count);
06e1bc05
JK
457 spin_unlock(&inode->i_lock);
458
459 /* some remained atomic pages should discarded */
460 if (f2fs_is_atomic_file(inode))
461 commit_inmem_pages(inode, true);
462
3e72f721
JK
463 /* should remain fi->extent_tree for writepage */
464 f2fs_destroy_extent_node(inode);
465
06e1bc05
JK
466 sb_start_intwrite(inode->i_sb);
467 i_size_write(inode, 0);
468
469 if (F2FS_HAS_BLOCKS(inode))
470 f2fs_truncate(inode);
471
472 sb_end_intwrite(inode->i_sb);
473
474#ifdef CONFIG_F2FS_FS_ENCRYPTION
475 if (F2FS_I(inode)->i_crypt_info)
26bf3dc7
JK
476 f2fs_free_encryption_info(inode,
477 F2FS_I(inode)->i_crypt_info);
06e1bc05
JK
478#endif
479 spin_lock(&inode->i_lock);
3e72f721 480 atomic_dec(&inode->i_count);
06e1bc05 481 }
531ad7d5 482 return 0;
06e1bc05 483 }
531ad7d5
JK
484 return generic_drop_inode(inode);
485}
486
b3783873
JK
487/*
488 * f2fs_dirty_inode() is called from __mark_inode_dirty()
489 *
490 * We should call set_dirty_inode to write the dirty inode through write_inode.
491 */
492static void f2fs_dirty_inode(struct inode *inode, int flags)
493{
494 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
b3783873
JK
495}
496
aff063e2
JK
497static void f2fs_i_callback(struct rcu_head *head)
498{
499 struct inode *inode = container_of(head, struct inode, i_rcu);
500 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
501}
502
25ca923b 503static void f2fs_destroy_inode(struct inode *inode)
aff063e2
JK
504{
505 call_rcu(&inode->i_rcu, f2fs_i_callback);
506}
507
508static void f2fs_put_super(struct super_block *sb)
509{
510 struct f2fs_sb_info *sbi = F2FS_SB(sb);
511
5e176d54
JK
512 if (sbi->s_proc) {
513 remove_proc_entry("segment_info", sbi->s_proc);
514 remove_proc_entry(sb->s_id, f2fs_proc_root);
515 }
b59d0bae 516 kobject_del(&sbi->s_kobj);
5e176d54 517
aff063e2
JK
518 stop_gc_thread(sbi);
519
2658e50d
JK
520 /* prevent remaining shrinker jobs */
521 mutex_lock(&sbi->umount_mutex);
522
85dc2f2c
JK
523 /*
524 * We don't need to do checkpoint when superblock is clean.
525 * But, the previous checkpoint was not done by umount, it needs to do
526 * clean checkpoint again.
527 */
caf0047e 528 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
85dc2f2c 529 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
75ab4cb8
JK
530 struct cp_control cpc = {
531 .reason = CP_UMOUNT,
532 };
533 write_checkpoint(sbi, &cpc);
534 }
aff063e2 535
eca616f8
JK
536 /* write_checkpoint can update stat informaion */
537 f2fs_destroy_stats(sbi);
538
cf779cab
JK
539 /*
540 * normally superblock is clean, so we need to release this.
541 * In addition, EIO will skip do checkpoint, we need this as well.
542 */
6f12ac25 543 release_dirty_inode(sbi);
4b2fecc8 544 release_discard_addrs(sbi);
6f12ac25 545
2658e50d
JK
546 f2fs_leave_shrinker(sbi);
547 mutex_unlock(&sbi->umount_mutex);
548
aff063e2
JK
549 iput(sbi->node_inode);
550 iput(sbi->meta_inode);
551
552 /* destroy f2fs internal modules */
553 destroy_node_manager(sbi);
554 destroy_segment_manager(sbi);
555
556 kfree(sbi->ckpt);
b59d0bae
NJ
557 kobject_put(&sbi->s_kobj);
558 wait_for_completion(&sbi->s_kobj_unregister);
aff063e2
JK
559
560 sb->s_fs_info = NULL;
561 brelse(sbi->raw_super_buf);
562 kfree(sbi);
563}
564
565int f2fs_sync_fs(struct super_block *sb, int sync)
566{
567 struct f2fs_sb_info *sbi = F2FS_SB(sb);
aff063e2 568
a2a4a7e4
NJ
569 trace_f2fs_sync_fs(sb, sync);
570
b7473754 571 if (sync) {
d5053a34
JK
572 struct cp_control cpc;
573
119ee914
JK
574 cpc.reason = __get_cp_reason(sbi);
575
b7473754 576 mutex_lock(&sbi->gc_mutex);
75ab4cb8 577 write_checkpoint(sbi, &cpc);
b7473754
JK
578 mutex_unlock(&sbi->gc_mutex);
579 } else {
7d82db83 580 f2fs_balance_fs(sbi);
b7473754 581 }
05ca3632 582 f2fs_trace_ios(NULL, 1);
aff063e2 583
64c576fe 584 return 0;
aff063e2
JK
585}
586
d6212a5f
CL
587static int f2fs_freeze(struct super_block *sb)
588{
589 int err;
590
77888c1e 591 if (f2fs_readonly(sb))
d6212a5f
CL
592 return 0;
593
594 err = f2fs_sync_fs(sb, 1);
595 return err;
596}
597
598static int f2fs_unfreeze(struct super_block *sb)
599{
600 return 0;
601}
602
aff063e2
JK
603static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
604{
605 struct super_block *sb = dentry->d_sb;
606 struct f2fs_sb_info *sbi = F2FS_SB(sb);
607 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
608 block_t total_count, user_block_count, start_count, ovp_count;
609
610 total_count = le64_to_cpu(sbi->raw_super->block_count);
611 user_block_count = sbi->user_block_count;
612 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
613 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
614 buf->f_type = F2FS_SUPER_MAGIC;
615 buf->f_bsize = sbi->blocksize;
616
617 buf->f_blocks = total_count - start_count;
618 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
619 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
620
c200b1aa
CY
621 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
622 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
aff063e2 623
5a20d339 624 buf->f_namelen = F2FS_NAME_LEN;
aff063e2
JK
625 buf->f_fsid.val[0] = (u32)id;
626 buf->f_fsid.val[1] = (u32)(id >> 32);
627
628 return 0;
629}
630
631static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
632{
633 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
634
b270ad6f 635 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
696c018c 636 seq_printf(seq, ",background_gc=%s", "on");
aff063e2 637 else
696c018c 638 seq_printf(seq, ",background_gc=%s", "off");
aff063e2
JK
639 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
640 seq_puts(seq, ",disable_roll_forward");
641 if (test_opt(sbi, DISCARD))
642 seq_puts(seq, ",discard");
643 if (test_opt(sbi, NOHEAP))
644 seq_puts(seq, ",no_heap_alloc");
645#ifdef CONFIG_F2FS_FS_XATTR
646 if (test_opt(sbi, XATTR_USER))
647 seq_puts(seq, ",user_xattr");
648 else
649 seq_puts(seq, ",nouser_xattr");
444c580f
JK
650 if (test_opt(sbi, INLINE_XATTR))
651 seq_puts(seq, ",inline_xattr");
aff063e2
JK
652#endif
653#ifdef CONFIG_F2FS_FS_POSIX_ACL
654 if (test_opt(sbi, POSIX_ACL))
655 seq_puts(seq, ",acl");
656 else
657 seq_puts(seq, ",noacl");
658#endif
659 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
aa43507f 660 seq_puts(seq, ",disable_ext_identify");
8274de77
HL
661 if (test_opt(sbi, INLINE_DATA))
662 seq_puts(seq, ",inline_data");
75342797
WL
663 else
664 seq_puts(seq, ",noinline_data");
5efd3c6f
CY
665 if (test_opt(sbi, INLINE_DENTRY))
666 seq_puts(seq, ",inline_dentry");
b270ad6f 667 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
6b4afdd7 668 seq_puts(seq, ",flush_merge");
0f7b2abd
JK
669 if (test_opt(sbi, NOBARRIER))
670 seq_puts(seq, ",nobarrier");
d5053a34
JK
671 if (test_opt(sbi, FASTBOOT))
672 seq_puts(seq, ",fastboot");
89672159
CY
673 if (test_opt(sbi, EXTENT_CACHE))
674 seq_puts(seq, ",extent_cache");
7daaea25
JK
675 else
676 seq_puts(seq, ",noextent_cache");
aff063e2
JK
677 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
678
679 return 0;
680}
681
5e176d54
JK
682static int segment_info_seq_show(struct seq_file *seq, void *offset)
683{
684 struct super_block *sb = seq->private;
685 struct f2fs_sb_info *sbi = F2FS_SB(sb);
6c311ec6
CF
686 unsigned int total_segs =
687 le32_to_cpu(sbi->raw_super->segment_count_main);
5e176d54
JK
688 int i;
689
90aa6dc9
CY
690 seq_puts(seq, "format: segment_type|valid_blocks\n"
691 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
692
5e176d54 693 for (i = 0; i < total_segs; i++) {
90aa6dc9
CY
694 struct seg_entry *se = get_seg_entry(sbi, i);
695
696 if ((i % 10) == 0)
697 seq_printf(seq, "%-5d", i);
698 seq_printf(seq, "%d|%-3u", se->type,
699 get_valid_blocks(sbi, i, 1));
46c04366
GZ
700 if ((i % 10) == 9 || i == (total_segs - 1))
701 seq_putc(seq, '\n');
5e176d54 702 else
46c04366 703 seq_putc(seq, ' ');
5e176d54 704 }
46c04366 705
5e176d54
JK
706 return 0;
707}
708
709static int segment_info_open_fs(struct inode *inode, struct file *file)
710{
711 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
712}
713
714static const struct file_operations f2fs_seq_segment_info_fops = {
715 .owner = THIS_MODULE,
716 .open = segment_info_open_fs,
717 .read = seq_read,
718 .llseek = seq_lseek,
719 .release = single_release,
720};
721
498c5e9f
YH
722static void default_options(struct f2fs_sb_info *sbi)
723{
724 /* init some FS parameters */
725 sbi->active_logs = NR_CURSEG_TYPE;
726
727 set_opt(sbi, BG_GC);
728 set_opt(sbi, INLINE_DATA);
3e72f721 729 set_opt(sbi, EXTENT_CACHE);
498c5e9f
YH
730
731#ifdef CONFIG_F2FS_FS_XATTR
732 set_opt(sbi, XATTR_USER);
733#endif
734#ifdef CONFIG_F2FS_FS_POSIX_ACL
735 set_opt(sbi, POSIX_ACL);
736#endif
737}
738
696c018c
NJ
739static int f2fs_remount(struct super_block *sb, int *flags, char *data)
740{
741 struct f2fs_sb_info *sbi = F2FS_SB(sb);
742 struct f2fs_mount_info org_mount_opt;
743 int err, active_logs;
876dc59e
GZ
744 bool need_restart_gc = false;
745 bool need_stop_gc = false;
696c018c 746
02b9984d
TT
747 sync_filesystem(sb);
748
696c018c
NJ
749 /*
750 * Save the old mount options in case we
751 * need to restore them.
752 */
753 org_mount_opt = sbi->mount_opt;
754 active_logs = sbi->active_logs;
755
26666c8a 756 sbi->mount_opt.opt = 0;
498c5e9f 757 default_options(sbi);
26666c8a 758
696c018c
NJ
759 /* parse mount options */
760 err = parse_options(sb, data);
761 if (err)
762 goto restore_opts;
763
764 /*
765 * Previous and new state of filesystem is RO,
876dc59e 766 * so skip checking GC and FLUSH_MERGE conditions.
696c018c 767 */
6b2920a5 768 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
696c018c
NJ
769 goto skip;
770
771 /*
772 * We stop the GC thread if FS is mounted as RO
773 * or if background_gc = off is passed in mount
774 * option. Also sync the filesystem.
775 */
776 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
777 if (sbi->gc_thread) {
778 stop_gc_thread(sbi);
779 f2fs_sync_fs(sb, 1);
876dc59e 780 need_restart_gc = true;
696c018c 781 }
aba291b3 782 } else if (!sbi->gc_thread) {
696c018c
NJ
783 err = start_gc_thread(sbi);
784 if (err)
785 goto restore_opts;
876dc59e
GZ
786 need_stop_gc = true;
787 }
788
789 /*
790 * We stop issue flush thread if FS is mounted as RO
791 * or if flush_merge is not passed in mount option.
792 */
793 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2163d198 794 destroy_flush_cmd_control(sbi);
aba291b3 795 } else if (!SM_I(sbi)->cmd_control_info) {
2163d198
GZ
796 err = create_flush_cmd_control(sbi);
797 if (err)
a688b9d9 798 goto restore_gc;
696c018c
NJ
799 }
800skip:
801 /* Update the POSIXACL Flag */
802 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
803 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
804 return 0;
876dc59e
GZ
805restore_gc:
806 if (need_restart_gc) {
807 if (start_gc_thread(sbi))
808 f2fs_msg(sbi->sb, KERN_WARNING,
e1c42045 809 "background gc thread has stopped");
876dc59e
GZ
810 } else if (need_stop_gc) {
811 stop_gc_thread(sbi);
812 }
696c018c
NJ
813restore_opts:
814 sbi->mount_opt = org_mount_opt;
815 sbi->active_logs = active_logs;
816 return err;
817}
818
aff063e2
JK
819static struct super_operations f2fs_sops = {
820 .alloc_inode = f2fs_alloc_inode,
531ad7d5 821 .drop_inode = f2fs_drop_inode,
aff063e2
JK
822 .destroy_inode = f2fs_destroy_inode,
823 .write_inode = f2fs_write_inode,
b3783873 824 .dirty_inode = f2fs_dirty_inode,
aff063e2
JK
825 .show_options = f2fs_show_options,
826 .evict_inode = f2fs_evict_inode,
827 .put_super = f2fs_put_super,
828 .sync_fs = f2fs_sync_fs,
d6212a5f
CL
829 .freeze_fs = f2fs_freeze,
830 .unfreeze_fs = f2fs_unfreeze,
aff063e2 831 .statfs = f2fs_statfs,
696c018c 832 .remount_fs = f2fs_remount,
aff063e2
JK
833};
834
835static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
836 u64 ino, u32 generation)
837{
838 struct f2fs_sb_info *sbi = F2FS_SB(sb);
839 struct inode *inode;
840
d6b7d4b3 841 if (check_nid_range(sbi, ino))
910bb12d 842 return ERR_PTR(-ESTALE);
aff063e2
JK
843
844 /*
845 * f2fs_iget isn't quite right if the inode is currently unallocated!
846 * However f2fs_iget currently does appropriate checks to handle stale
847 * inodes so everything is OK.
848 */
849 inode = f2fs_iget(sb, ino);
850 if (IS_ERR(inode))
851 return ERR_CAST(inode);
6bacf52f 852 if (unlikely(generation && inode->i_generation != generation)) {
aff063e2
JK
853 /* we didn't find the right inode.. */
854 iput(inode);
855 return ERR_PTR(-ESTALE);
856 }
857 return inode;
858}
859
860static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
861 int fh_len, int fh_type)
862{
863 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
864 f2fs_nfs_get_inode);
865}
866
867static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
868 int fh_len, int fh_type)
869{
870 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
871 f2fs_nfs_get_inode);
872}
873
874static const struct export_operations f2fs_export_ops = {
875 .fh_to_dentry = f2fs_fh_to_dentry,
876 .fh_to_parent = f2fs_fh_to_parent,
877 .get_parent = f2fs_get_parent,
878};
879
aff063e2
JK
880static loff_t max_file_size(unsigned bits)
881{
de93653f 882 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
aff063e2
JK
883 loff_t leaf_count = ADDRS_PER_BLOCK;
884
885 /* two direct node blocks */
886 result += (leaf_count * 2);
887
888 /* two indirect node blocks */
889 leaf_count *= NIDS_PER_BLOCK;
890 result += (leaf_count * 2);
891
892 /* one double indirect node block */
893 leaf_count *= NIDS_PER_BLOCK;
894 result += leaf_count;
895
896 result <<= bits;
897 return result;
898}
899
a07ef784
NJ
900static int sanity_check_raw_super(struct super_block *sb,
901 struct f2fs_super_block *raw_super)
aff063e2
JK
902{
903 unsigned int blocksize;
904
a07ef784
NJ
905 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
906 f2fs_msg(sb, KERN_INFO,
907 "Magic Mismatch, valid(0x%x) - read(0x%x)",
908 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
aff063e2 909 return 1;
a07ef784 910 }
aff063e2 911
5c9b4692 912 /* Currently, support only 4KB page cache size */
913 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
914 f2fs_msg(sb, KERN_INFO,
14d7e9de 915 "Invalid page_cache_size (%lu), supports only 4KB\n",
5c9b4692 916 PAGE_CACHE_SIZE);
917 return 1;
918 }
919
aff063e2
JK
920 /* Currently, support only 4KB block size */
921 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
5c9b4692 922 if (blocksize != F2FS_BLKSIZE) {
a07ef784
NJ
923 f2fs_msg(sb, KERN_INFO,
924 "Invalid blocksize (%u), supports only 4KB\n",
925 blocksize);
aff063e2 926 return 1;
a07ef784 927 }
5c9b4692 928
55cf9cb6
CY
929 /* Currently, support 512/1024/2048/4096 bytes sector size */
930 if (le32_to_cpu(raw_super->log_sectorsize) >
931 F2FS_MAX_LOG_SECTOR_SIZE ||
932 le32_to_cpu(raw_super->log_sectorsize) <
933 F2FS_MIN_LOG_SECTOR_SIZE) {
934 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
935 le32_to_cpu(raw_super->log_sectorsize));
aff063e2 936 return 1;
a07ef784 937 }
55cf9cb6
CY
938 if (le32_to_cpu(raw_super->log_sectors_per_block) +
939 le32_to_cpu(raw_super->log_sectorsize) !=
940 F2FS_MAX_LOG_SECTOR_SIZE) {
941 f2fs_msg(sb, KERN_INFO,
942 "Invalid log sectors per block(%u) log sectorsize(%u)",
943 le32_to_cpu(raw_super->log_sectors_per_block),
944 le32_to_cpu(raw_super->log_sectorsize));
aff063e2 945 return 1;
a07ef784 946 }
aff063e2
JK
947 return 0;
948}
949
577e3495 950static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
aff063e2
JK
951{
952 unsigned int total, fsmeta;
577e3495
JK
953 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
954 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
aff063e2
JK
955
956 total = le32_to_cpu(raw_super->segment_count);
957 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
958 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
959 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
960 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
961 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
962
6bacf52f 963 if (unlikely(fsmeta >= total))
aff063e2 964 return 1;
577e3495 965
1e968fdf 966 if (unlikely(f2fs_cp_error(sbi))) {
577e3495
JK
967 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
968 return 1;
969 }
aff063e2
JK
970 return 0;
971}
972
973static void init_sb_info(struct f2fs_sb_info *sbi)
974{
975 struct f2fs_super_block *raw_super = sbi->raw_super;
976 int i;
977
978 sbi->log_sectors_per_block =
979 le32_to_cpu(raw_super->log_sectors_per_block);
980 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
981 sbi->blocksize = 1 << sbi->log_blocksize;
982 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
983 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
984 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
985 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
986 sbi->total_sections = le32_to_cpu(raw_super->section_count);
987 sbi->total_node_count =
988 (le32_to_cpu(raw_super->segment_count_nat) / 2)
989 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
990 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
991 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
992 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
5ec4e49f 993 sbi->cur_victim_sec = NULL_SECNO;
b1c57c1c 994 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
aff063e2
JK
995
996 for (i = 0; i < NR_COUNT_TYPE; i++)
997 atomic_set(&sbi->nr_pages[i], 0);
ab9fa662
JK
998
999 sbi->dir_level = DEF_DIR_LEVEL;
caf0047e 1000 clear_sbi_flag(sbi, SBI_NEED_FSCK);
2658e50d
JK
1001
1002 INIT_LIST_HEAD(&sbi->s_list);
1003 mutex_init(&sbi->umount_mutex);
aff063e2
JK
1004}
1005
9076a75f
GZ
1006/*
1007 * Read f2fs raw super block.
1008 * Because we have two copies of super block, so read the first one at first,
1009 * if the first one is invalid, move to read the second one.
1010 */
1011static int read_raw_super_block(struct super_block *sb,
1012 struct f2fs_super_block **raw_super,
da554e48 1013 struct buffer_head **raw_super_buf,
1014 int *recovery)
14d7e9de 1015{
9076a75f 1016 int block = 0;
da554e48 1017 struct buffer_head *buffer;
1018 struct f2fs_super_block *super;
1019 int err = 0;
14d7e9de 1020
9076a75f 1021retry:
da554e48 1022 buffer = sb_bread(sb, block);
1023 if (!buffer) {
1024 *recovery = 1;
9076a75f
GZ
1025 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
1026 block + 1);
1027 if (block == 0) {
1028 block++;
1029 goto retry;
1030 } else {
da554e48 1031 err = -EIO;
1032 goto out;
9076a75f 1033 }
14d7e9de 1034 }
1035
da554e48 1036 super = (struct f2fs_super_block *)
1037 ((char *)(buffer)->b_data + F2FS_SUPER_OFFSET);
14d7e9de 1038
1039 /* sanity checking of raw super */
da554e48 1040 if (sanity_check_raw_super(sb, super)) {
1041 brelse(buffer);
1042 *recovery = 1;
6c311ec6
CF
1043 f2fs_msg(sb, KERN_ERR,
1044 "Can't find valid F2FS filesystem in %dth superblock",
1045 block + 1);
6bacf52f 1046 if (block == 0) {
9076a75f
GZ
1047 block++;
1048 goto retry;
1049 } else {
da554e48 1050 err = -EINVAL;
1051 goto out;
9076a75f
GZ
1052 }
1053 }
14d7e9de 1054
da554e48 1055 if (!*raw_super) {
1056 *raw_super_buf = buffer;
1057 *raw_super = super;
1058 } else {
1059 /* already have a valid superblock */
1060 brelse(buffer);
1061 }
1062
1063 /* check the validity of the second superblock */
1064 if (block == 0) {
1065 block++;
1066 goto retry;
1067 }
1068
1069out:
1070 /* No valid superblock */
1071 if (!*raw_super)
1072 return err;
1073
9076a75f 1074 return 0;
14d7e9de 1075}
1076
c5bda1c8 1077int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
26d815ad
JK
1078{
1079 struct buffer_head *sbh = sbi->raw_super_buf;
1080 sector_t block = sbh->b_blocknr;
1081 int err;
1082
1083 /* write back-up superblock first */
1084 sbh->b_blocknr = block ? 0 : 1;
1085 mark_buffer_dirty(sbh);
1086 err = sync_dirty_buffer(sbh);
1087
1088 sbh->b_blocknr = block;
c5bda1c8
CY
1089
1090 /* if we are in recovery path, skip writing valid superblock */
1091 if (recover || err)
26d815ad
JK
1092 goto out;
1093
1094 /* write current valid superblock */
1095 mark_buffer_dirty(sbh);
1096 err = sync_dirty_buffer(sbh);
1097out:
1098 clear_buffer_write_io_error(sbh);
1099 set_buffer_uptodate(sbh);
1100 return err;
1101}
1102
aff063e2
JK
1103static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
1104{
1105 struct f2fs_sb_info *sbi;
da554e48 1106 struct f2fs_super_block *raw_super;
aff063e2
JK
1107 struct buffer_head *raw_super_buf;
1108 struct inode *root;
da554e48 1109 long err;
2adc3505 1110 bool retry = true, need_fsck = false;
dabc4a5c 1111 char *options = NULL;
da554e48 1112 int recovery, i;
aff063e2 1113
ed2e621a 1114try_onemore:
da554e48 1115 err = -EINVAL;
1116 raw_super = NULL;
1117 raw_super_buf = NULL;
1118 recovery = 0;
1119
aff063e2
JK
1120 /* allocate memory for f2fs-specific super block info */
1121 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
1122 if (!sbi)
1123 return -ENOMEM;
1124
ff9234ad 1125 /* set a block size */
6bacf52f 1126 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
a07ef784 1127 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
aff063e2 1128 goto free_sbi;
a07ef784 1129 }
aff063e2 1130
da554e48 1131 err = read_raw_super_block(sb, &raw_super, &raw_super_buf, &recovery);
9076a75f
GZ
1132 if (err)
1133 goto free_sbi;
1134
5fb08372 1135 sb->s_fs_info = sbi;
498c5e9f 1136 default_options(sbi);
aff063e2 1137 /* parse mount options */
dabc4a5c
JK
1138 options = kstrdup((const char *)data, GFP_KERNEL);
1139 if (data && !options) {
1140 err = -ENOMEM;
aff063e2 1141 goto free_sb_buf;
dabc4a5c
JK
1142 }
1143
1144 err = parse_options(sb, options);
1145 if (err)
1146 goto free_options;
aff063e2 1147
25ca923b 1148 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
aff063e2
JK
1149 sb->s_max_links = F2FS_LINK_MAX;
1150 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1151
1152 sb->s_op = &f2fs_sops;
1153 sb->s_xattr = f2fs_xattr_handlers;
1154 sb->s_export_op = &f2fs_export_ops;
1155 sb->s_magic = F2FS_SUPER_MAGIC;
aff063e2
JK
1156 sb->s_time_gran = 1;
1157 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1158 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1159 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1160
1161 /* init f2fs-specific super block info */
1162 sbi->sb = sb;
1163 sbi->raw_super = raw_super;
1164 sbi->raw_super_buf = raw_super_buf;
1165 mutex_init(&sbi->gc_mutex);
5463e7c1 1166 mutex_init(&sbi->writepages);
aff063e2 1167 mutex_init(&sbi->cp_mutex);
b3582c68 1168 init_rwsem(&sbi->node_write);
caf0047e 1169 clear_sbi_flag(sbi, SBI_POR_DOING);
aff063e2 1170 spin_lock_init(&sbi->stat_lock);
971767ca 1171
df0f8dc0 1172 init_rwsem(&sbi->read_io.io_rwsem);
458e6197
JK
1173 sbi->read_io.sbi = sbi;
1174 sbi->read_io.bio = NULL;
1175 for (i = 0; i < NR_PAGE_TYPE; i++) {
df0f8dc0 1176 init_rwsem(&sbi->write_io[i].io_rwsem);
458e6197
JK
1177 sbi->write_io[i].sbi = sbi;
1178 sbi->write_io[i].bio = NULL;
1179 }
971767ca 1180
e479556b 1181 init_rwsem(&sbi->cp_rwsem);
fb51b5ef 1182 init_waitqueue_head(&sbi->cp_wait);
aff063e2
JK
1183 init_sb_info(sbi);
1184
1185 /* get an inode for meta space */
1186 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1187 if (IS_ERR(sbi->meta_inode)) {
a07ef784 1188 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
aff063e2 1189 err = PTR_ERR(sbi->meta_inode);
dabc4a5c 1190 goto free_options;
aff063e2
JK
1191 }
1192
1193 err = get_valid_checkpoint(sbi);
a07ef784
NJ
1194 if (err) {
1195 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
aff063e2 1196 goto free_meta_inode;
a07ef784 1197 }
aff063e2
JK
1198
1199 /* sanity checking of checkpoint */
1200 err = -EINVAL;
577e3495 1201 if (sanity_check_ckpt(sbi)) {
a07ef784 1202 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
aff063e2 1203 goto free_cp;
a07ef784 1204 }
aff063e2
JK
1205
1206 sbi->total_valid_node_count =
1207 le32_to_cpu(sbi->ckpt->valid_node_count);
1208 sbi->total_valid_inode_count =
1209 le32_to_cpu(sbi->ckpt->valid_inode_count);
1210 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1211 sbi->total_valid_block_count =
1212 le64_to_cpu(sbi->ckpt->valid_block_count);
1213 sbi->last_valid_block_count = sbi->total_valid_block_count;
1214 sbi->alloc_valid_block_count = 0;
1215 INIT_LIST_HEAD(&sbi->dir_inode_list);
1216 spin_lock_init(&sbi->dir_inode_lock);
1217
1dcc336b
CY
1218 init_extent_cache_info(sbi);
1219
6451e041 1220 init_ino_entry_info(sbi);
aff063e2
JK
1221
1222 /* setup f2fs internal modules */
1223 err = build_segment_manager(sbi);
a07ef784
NJ
1224 if (err) {
1225 f2fs_msg(sb, KERN_ERR,
1226 "Failed to initialize F2FS segment manager");
aff063e2 1227 goto free_sm;
a07ef784 1228 }
aff063e2 1229 err = build_node_manager(sbi);
a07ef784
NJ
1230 if (err) {
1231 f2fs_msg(sb, KERN_ERR,
1232 "Failed to initialize F2FS node manager");
aff063e2 1233 goto free_nm;
a07ef784 1234 }
aff063e2
JK
1235
1236 build_gc_manager(sbi);
1237
1238 /* get an inode for node space */
1239 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1240 if (IS_ERR(sbi->node_inode)) {
a07ef784 1241 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
aff063e2
JK
1242 err = PTR_ERR(sbi->node_inode);
1243 goto free_nm;
1244 }
1245
2658e50d
JK
1246 f2fs_join_shrinker(sbi);
1247
aff063e2 1248 /* if there are nt orphan nodes free them */
8f99a946 1249 recover_orphan_inodes(sbi);
aff063e2
JK
1250
1251 /* read root inode and dentry */
1252 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1253 if (IS_ERR(root)) {
a07ef784 1254 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
aff063e2
JK
1255 err = PTR_ERR(root);
1256 goto free_node_inode;
1257 }
8f99a946 1258 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
9d847950 1259 iput(root);
8f99a946 1260 err = -EINVAL;
9d847950 1261 goto free_node_inode;
8f99a946 1262 }
aff063e2
JK
1263
1264 sb->s_root = d_make_root(root); /* allocate root dentry */
1265 if (!sb->s_root) {
1266 err = -ENOMEM;
1267 goto free_root_inode;
1268 }
1269
aff063e2
JK
1270 err = f2fs_build_stats(sbi);
1271 if (err)
6437d1b0 1272 goto free_root_inode;
aff063e2 1273
5e176d54
JK
1274 if (f2fs_proc_root)
1275 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1276
1277 if (sbi->s_proc)
1278 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1279 &f2fs_seq_segment_info_fops, sb);
1280
b59d0bae
NJ
1281 sbi->s_kobj.kset = f2fs_kset;
1282 init_completion(&sbi->s_kobj_unregister);
1283 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1284 "%s", sb->s_id);
1285 if (err)
6437d1b0 1286 goto free_proc;
b59d0bae 1287
6437d1b0
JK
1288 /* recover fsynced data */
1289 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
081d78c2
JK
1290 /*
1291 * mount should be failed, when device has readonly mode, and
1292 * previous checkpoint was not done by clean system shutdown.
1293 */
1294 if (bdev_read_only(sb->s_bdev) &&
1295 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1296 err = -EROFS;
1297 goto free_kobj;
1298 }
2adc3505
CY
1299
1300 if (need_fsck)
1301 set_sbi_flag(sbi, SBI_NEED_FSCK);
1302
6437d1b0 1303 err = recover_fsync_data(sbi);
ed2e621a 1304 if (err) {
2adc3505 1305 need_fsck = true;
6437d1b0
JK
1306 f2fs_msg(sb, KERN_ERR,
1307 "Cannot recover all fsync data errno=%ld", err);
ed2e621a
JK
1308 goto free_kobj;
1309 }
6437d1b0 1310 }
b59d0bae 1311
6437d1b0
JK
1312 /*
1313 * If filesystem is not mounted as read-only then
1314 * do start the gc_thread.
1315 */
6c029932 1316 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
6437d1b0
JK
1317 /* After POR, we can run background GC thread.*/
1318 err = start_gc_thread(sbi);
1319 if (err)
1320 goto free_kobj;
1321 }
dabc4a5c 1322 kfree(options);
da554e48 1323
1324 /* recover broken superblock */
1325 if (recovery && !f2fs_readonly(sb) && !bdev_read_only(sb->s_bdev)) {
1326 f2fs_msg(sb, KERN_INFO, "Recover invalid superblock");
c5bda1c8 1327 f2fs_commit_super(sbi, true);
da554e48 1328 }
1329
aff063e2 1330 return 0;
6437d1b0
JK
1331
1332free_kobj:
1333 kobject_del(&sbi->s_kobj);
1334free_proc:
1d15bd20
CY
1335 if (sbi->s_proc) {
1336 remove_proc_entry("segment_info", sbi->s_proc);
1337 remove_proc_entry(sb->s_id, f2fs_proc_root);
1338 }
1339 f2fs_destroy_stats(sbi);
aff063e2
JK
1340free_root_inode:
1341 dput(sb->s_root);
1342 sb->s_root = NULL;
1343free_node_inode:
2658e50d
JK
1344 mutex_lock(&sbi->umount_mutex);
1345 f2fs_leave_shrinker(sbi);
aff063e2 1346 iput(sbi->node_inode);
2658e50d 1347 mutex_unlock(&sbi->umount_mutex);
aff063e2
JK
1348free_nm:
1349 destroy_node_manager(sbi);
1350free_sm:
1351 destroy_segment_manager(sbi);
1352free_cp:
1353 kfree(sbi->ckpt);
1354free_meta_inode:
1355 make_bad_inode(sbi->meta_inode);
1356 iput(sbi->meta_inode);
dabc4a5c
JK
1357free_options:
1358 kfree(options);
aff063e2
JK
1359free_sb_buf:
1360 brelse(raw_super_buf);
1361free_sbi:
1362 kfree(sbi);
ed2e621a
JK
1363
1364 /* give only one another chance */
1365 if (retry) {
9df47ba7 1366 retry = false;
ed2e621a
JK
1367 shrink_dcache_sb(sb);
1368 goto try_onemore;
1369 }
aff063e2
JK
1370 return err;
1371}
1372
1373static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1374 const char *dev_name, void *data)
1375{
1376 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1377}
1378
30a5537f
JK
1379static void kill_f2fs_super(struct super_block *sb)
1380{
1381 if (sb->s_root)
caf0047e 1382 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
30a5537f
JK
1383 kill_block_super(sb);
1384}
1385
aff063e2
JK
1386static struct file_system_type f2fs_fs_type = {
1387 .owner = THIS_MODULE,
1388 .name = "f2fs",
1389 .mount = f2fs_mount,
30a5537f 1390 .kill_sb = kill_f2fs_super,
aff063e2
JK
1391 .fs_flags = FS_REQUIRES_DEV,
1392};
7f78e035 1393MODULE_ALIAS_FS("f2fs");
aff063e2 1394
6e6093a8 1395static int __init init_inodecache(void)
aff063e2
JK
1396{
1397 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
e8512d2e 1398 sizeof(struct f2fs_inode_info));
6bacf52f 1399 if (!f2fs_inode_cachep)
aff063e2
JK
1400 return -ENOMEM;
1401 return 0;
1402}
1403
1404static void destroy_inodecache(void)
1405{
1406 /*
1407 * Make sure all delayed rcu free inodes are flushed before we
1408 * destroy cache.
1409 */
1410 rcu_barrier();
1411 kmem_cache_destroy(f2fs_inode_cachep);
1412}
1413
1414static int __init init_f2fs_fs(void)
1415{
1416 int err;
1417
c0508650
JK
1418 f2fs_build_trace_ios();
1419
aff063e2
JK
1420 err = init_inodecache();
1421 if (err)
1422 goto fail;
1423 err = create_node_manager_caches();
1424 if (err)
9890ff3f 1425 goto free_inodecache;
7fd9e544 1426 err = create_segment_manager_caches();
aff063e2 1427 if (err)
9890ff3f 1428 goto free_node_manager_caches;
aff063e2
JK
1429 err = create_checkpoint_caches();
1430 if (err)
06292073 1431 goto free_segment_manager_caches;
1dcc336b
CY
1432 err = create_extent_cache();
1433 if (err)
1434 goto free_checkpoint_caches;
b59d0bae 1435 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
6e6b978c
WY
1436 if (!f2fs_kset) {
1437 err = -ENOMEM;
1dcc336b 1438 goto free_extent_cache;
6e6b978c 1439 }
cfc4d971 1440 err = f2fs_init_crypto();
4589d25d 1441 if (err)
9890ff3f 1442 goto free_kset;
2658e50d
JK
1443
1444 err = register_shrinker(&f2fs_shrinker_info);
cfc4d971
JK
1445 if (err)
1446 goto free_crypto;
2658e50d
JK
1447
1448 err = register_filesystem(&f2fs_fs_type);
1449 if (err)
1450 goto free_shrinker;
4589d25d 1451 f2fs_create_root_stats();
5e176d54 1452 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
9890ff3f
ZH
1453 return 0;
1454
2658e50d
JK
1455free_shrinker:
1456 unregister_shrinker(&f2fs_shrinker_info);
cfc4d971
JK
1457free_crypto:
1458 f2fs_exit_crypto();
9890ff3f
ZH
1459free_kset:
1460 kset_unregister(f2fs_kset);
1dcc336b
CY
1461free_extent_cache:
1462 destroy_extent_cache();
9890ff3f
ZH
1463free_checkpoint_caches:
1464 destroy_checkpoint_caches();
7fd9e544
JK
1465free_segment_manager_caches:
1466 destroy_segment_manager_caches();
9890ff3f
ZH
1467free_node_manager_caches:
1468 destroy_node_manager_caches();
1469free_inodecache:
1470 destroy_inodecache();
aff063e2
JK
1471fail:
1472 return err;
1473}
1474
1475static void __exit exit_f2fs_fs(void)
1476{
5e176d54 1477 remove_proc_entry("fs/f2fs", NULL);
4589d25d 1478 f2fs_destroy_root_stats();
2658e50d 1479 unregister_shrinker(&f2fs_shrinker_info);
aff063e2 1480 unregister_filesystem(&f2fs_fs_type);
cfc4d971 1481 f2fs_exit_crypto();
fdf6c8be 1482 destroy_extent_cache();
aff063e2 1483 destroy_checkpoint_caches();
5dcd8a71 1484 destroy_segment_manager_caches();
aff063e2
JK
1485 destroy_node_manager_caches();
1486 destroy_inodecache();
b59d0bae 1487 kset_unregister(f2fs_kset);
351f4fba 1488 f2fs_destroy_trace_ios();
aff063e2
JK
1489}
1490
1491module_init(init_f2fs_fs)
1492module_exit(exit_f2fs_fs)
1493
1494MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1495MODULE_DESCRIPTION("Flash Friendly File System");
1496MODULE_LICENSE("GPL");