gro: Allow tunnel stacking in the case of FOU/GUE
[linux/fpc-iii.git] / fs / f2fs / super.c
blobb2dd1b01f07634e27f42fc09e8fd627a70395e3e
1 /*
2 * fs/f2fs/super.c
4 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
5 * http://www.samsung.com/
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.
11 #include <linux/module.h>
12 #include <linux/init.h>
13 #include <linux/fs.h>
14 #include <linux/statfs.h>
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>
21 #include <linux/proc_fs.h>
22 #include <linux/random.h>
23 #include <linux/exportfs.h>
24 #include <linux/blkdev.h>
25 #include <linux/f2fs_fs.h>
26 #include <linux/sysfs.h>
28 #include "f2fs.h"
29 #include "node.h"
30 #include "segment.h"
31 #include "xattr.h"
32 #include "gc.h"
33 #include "trace.h"
35 #define CREATE_TRACE_POINTS
36 #include <trace/events/f2fs.h>
38 static struct proc_dir_entry *f2fs_proc_root;
39 static struct kmem_cache *f2fs_inode_cachep;
40 static struct kset *f2fs_kset;
42 enum {
43 Opt_gc_background,
44 Opt_disable_roll_forward,
45 Opt_norecovery,
46 Opt_discard,
47 Opt_noheap,
48 Opt_user_xattr,
49 Opt_nouser_xattr,
50 Opt_acl,
51 Opt_noacl,
52 Opt_active_logs,
53 Opt_disable_ext_identify,
54 Opt_inline_xattr,
55 Opt_inline_data,
56 Opt_inline_dentry,
57 Opt_flush_merge,
58 Opt_nobarrier,
59 Opt_fastboot,
60 Opt_extent_cache,
61 Opt_noinline_data,
62 Opt_err,
65 static match_table_t f2fs_tokens = {
66 {Opt_gc_background, "background_gc=%s"},
67 {Opt_disable_roll_forward, "disable_roll_forward"},
68 {Opt_norecovery, "norecovery"},
69 {Opt_discard, "discard"},
70 {Opt_noheap, "no_heap"},
71 {Opt_user_xattr, "user_xattr"},
72 {Opt_nouser_xattr, "nouser_xattr"},
73 {Opt_acl, "acl"},
74 {Opt_noacl, "noacl"},
75 {Opt_active_logs, "active_logs=%u"},
76 {Opt_disable_ext_identify, "disable_ext_identify"},
77 {Opt_inline_xattr, "inline_xattr"},
78 {Opt_inline_data, "inline_data"},
79 {Opt_inline_dentry, "inline_dentry"},
80 {Opt_flush_merge, "flush_merge"},
81 {Opt_nobarrier, "nobarrier"},
82 {Opt_fastboot, "fastboot"},
83 {Opt_extent_cache, "extent_cache"},
84 {Opt_noinline_data, "noinline_data"},
85 {Opt_err, NULL},
88 /* Sysfs support for f2fs */
89 enum {
90 GC_THREAD, /* struct f2fs_gc_thread */
91 SM_INFO, /* struct f2fs_sm_info */
92 NM_INFO, /* struct f2fs_nm_info */
93 F2FS_SBI, /* struct f2fs_sb_info */
96 struct f2fs_attr {
97 struct attribute attr;
98 ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *);
99 ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *,
100 const char *, size_t);
101 int struct_type;
102 int offset;
105 static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type)
107 if (struct_type == GC_THREAD)
108 return (unsigned char *)sbi->gc_thread;
109 else if (struct_type == SM_INFO)
110 return (unsigned char *)SM_I(sbi);
111 else if (struct_type == NM_INFO)
112 return (unsigned char *)NM_I(sbi);
113 else if (struct_type == F2FS_SBI)
114 return (unsigned char *)sbi;
115 return NULL;
118 static ssize_t f2fs_sbi_show(struct f2fs_attr *a,
119 struct f2fs_sb_info *sbi, char *buf)
121 unsigned char *ptr = NULL;
122 unsigned int *ui;
124 ptr = __struct_ptr(sbi, a->struct_type);
125 if (!ptr)
126 return -EINVAL;
128 ui = (unsigned int *)(ptr + a->offset);
130 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
133 static ssize_t f2fs_sbi_store(struct f2fs_attr *a,
134 struct f2fs_sb_info *sbi,
135 const char *buf, size_t count)
137 unsigned char *ptr;
138 unsigned long t;
139 unsigned int *ui;
140 ssize_t ret;
142 ptr = __struct_ptr(sbi, a->struct_type);
143 if (!ptr)
144 return -EINVAL;
146 ui = (unsigned int *)(ptr + a->offset);
148 ret = kstrtoul(skip_spaces(buf), 0, &t);
149 if (ret < 0)
150 return ret;
151 *ui = t;
152 return count;
155 static ssize_t f2fs_attr_show(struct kobject *kobj,
156 struct attribute *attr, char *buf)
158 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
159 s_kobj);
160 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
162 return a->show ? a->show(a, sbi, buf) : 0;
165 static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr,
166 const char *buf, size_t len)
168 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
169 s_kobj);
170 struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr);
172 return a->store ? a->store(a, sbi, buf, len) : 0;
175 static void f2fs_sb_release(struct kobject *kobj)
177 struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info,
178 s_kobj);
179 complete(&sbi->s_kobj_unregister);
182 #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \
183 static struct f2fs_attr f2fs_attr_##_name = { \
184 .attr = {.name = __stringify(_name), .mode = _mode }, \
185 .show = _show, \
186 .store = _store, \
187 .struct_type = _struct_type, \
188 .offset = _offset \
191 #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \
192 F2FS_ATTR_OFFSET(struct_type, name, 0644, \
193 f2fs_sbi_show, f2fs_sbi_store, \
194 offsetof(struct struct_name, elname))
196 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time);
197 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time);
198 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time);
199 F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
200 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
201 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
202 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
203 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
204 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
205 F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
206 F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh);
207 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search);
208 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level);
210 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
211 static struct attribute *f2fs_attrs[] = {
212 ATTR_LIST(gc_min_sleep_time),
213 ATTR_LIST(gc_max_sleep_time),
214 ATTR_LIST(gc_no_gc_sleep_time),
215 ATTR_LIST(gc_idle),
216 ATTR_LIST(reclaim_segments),
217 ATTR_LIST(max_small_discards),
218 ATTR_LIST(batched_trim_sections),
219 ATTR_LIST(ipu_policy),
220 ATTR_LIST(min_ipu_util),
221 ATTR_LIST(min_fsync_blocks),
222 ATTR_LIST(max_victim_search),
223 ATTR_LIST(dir_level),
224 ATTR_LIST(ram_thresh),
225 NULL,
228 static const struct sysfs_ops f2fs_attr_ops = {
229 .show = f2fs_attr_show,
230 .store = f2fs_attr_store,
233 static struct kobj_type f2fs_ktype = {
234 .default_attrs = f2fs_attrs,
235 .sysfs_ops = &f2fs_attr_ops,
236 .release = f2fs_sb_release,
239 void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...)
241 struct va_format vaf;
242 va_list args;
244 va_start(args, fmt);
245 vaf.fmt = fmt;
246 vaf.va = &args;
247 printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf);
248 va_end(args);
251 static void init_once(void *foo)
253 struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
255 inode_init_once(&fi->vfs_inode);
258 static int parse_options(struct super_block *sb, char *options)
260 struct f2fs_sb_info *sbi = F2FS_SB(sb);
261 substring_t args[MAX_OPT_ARGS];
262 char *p, *name;
263 int arg = 0;
265 if (!options)
266 return 0;
268 while ((p = strsep(&options, ",")) != NULL) {
269 int token;
270 if (!*p)
271 continue;
273 * Initialize args struct so we know whether arg was
274 * found; some options take optional arguments.
276 args[0].to = args[0].from = NULL;
277 token = match_token(p, f2fs_tokens, args);
279 switch (token) {
280 case Opt_gc_background:
281 name = match_strdup(&args[0]);
283 if (!name)
284 return -ENOMEM;
285 if (strlen(name) == 2 && !strncmp(name, "on", 2))
286 set_opt(sbi, BG_GC);
287 else if (strlen(name) == 3 && !strncmp(name, "off", 3))
288 clear_opt(sbi, BG_GC);
289 else {
290 kfree(name);
291 return -EINVAL;
293 kfree(name);
294 break;
295 case Opt_disable_roll_forward:
296 set_opt(sbi, DISABLE_ROLL_FORWARD);
297 break;
298 case Opt_norecovery:
299 /* this option mounts f2fs with ro */
300 set_opt(sbi, DISABLE_ROLL_FORWARD);
301 if (!f2fs_readonly(sb))
302 return -EINVAL;
303 break;
304 case Opt_discard:
305 set_opt(sbi, DISCARD);
306 break;
307 case Opt_noheap:
308 set_opt(sbi, NOHEAP);
309 break;
310 #ifdef CONFIG_F2FS_FS_XATTR
311 case Opt_user_xattr:
312 set_opt(sbi, XATTR_USER);
313 break;
314 case Opt_nouser_xattr:
315 clear_opt(sbi, XATTR_USER);
316 break;
317 case Opt_inline_xattr:
318 set_opt(sbi, INLINE_XATTR);
319 break;
320 #else
321 case Opt_user_xattr:
322 f2fs_msg(sb, KERN_INFO,
323 "user_xattr options not supported");
324 break;
325 case Opt_nouser_xattr:
326 f2fs_msg(sb, KERN_INFO,
327 "nouser_xattr options not supported");
328 break;
329 case Opt_inline_xattr:
330 f2fs_msg(sb, KERN_INFO,
331 "inline_xattr options not supported");
332 break;
333 #endif
334 #ifdef CONFIG_F2FS_FS_POSIX_ACL
335 case Opt_acl:
336 set_opt(sbi, POSIX_ACL);
337 break;
338 case Opt_noacl:
339 clear_opt(sbi, POSIX_ACL);
340 break;
341 #else
342 case Opt_acl:
343 f2fs_msg(sb, KERN_INFO, "acl options not supported");
344 break;
345 case Opt_noacl:
346 f2fs_msg(sb, KERN_INFO, "noacl options not supported");
347 break;
348 #endif
349 case Opt_active_logs:
350 if (args->from && match_int(args, &arg))
351 return -EINVAL;
352 if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE)
353 return -EINVAL;
354 sbi->active_logs = arg;
355 break;
356 case Opt_disable_ext_identify:
357 set_opt(sbi, DISABLE_EXT_IDENTIFY);
358 break;
359 case Opt_inline_data:
360 set_opt(sbi, INLINE_DATA);
361 break;
362 case Opt_inline_dentry:
363 set_opt(sbi, INLINE_DENTRY);
364 break;
365 case Opt_flush_merge:
366 set_opt(sbi, FLUSH_MERGE);
367 break;
368 case Opt_nobarrier:
369 set_opt(sbi, NOBARRIER);
370 break;
371 case Opt_fastboot:
372 set_opt(sbi, FASTBOOT);
373 break;
374 case Opt_extent_cache:
375 set_opt(sbi, EXTENT_CACHE);
376 break;
377 case Opt_noinline_data:
378 clear_opt(sbi, INLINE_DATA);
379 break;
380 default:
381 f2fs_msg(sb, KERN_ERR,
382 "Unrecognized mount option \"%s\" or missing value",
384 return -EINVAL;
387 return 0;
390 static struct inode *f2fs_alloc_inode(struct super_block *sb)
392 struct f2fs_inode_info *fi;
394 fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
395 if (!fi)
396 return NULL;
398 init_once((void *) fi);
400 /* Initialize f2fs-specific inode info */
401 fi->vfs_inode.i_version = 1;
402 atomic_set(&fi->dirty_pages, 0);
403 fi->i_current_depth = 1;
404 fi->i_advise = 0;
405 rwlock_init(&fi->ext_lock);
406 init_rwsem(&fi->i_sem);
407 INIT_RADIX_TREE(&fi->inmem_root, GFP_NOFS);
408 INIT_LIST_HEAD(&fi->inmem_pages);
409 mutex_init(&fi->inmem_lock);
411 set_inode_flag(fi, FI_NEW_INODE);
413 if (test_opt(F2FS_SB(sb), INLINE_XATTR))
414 set_inode_flag(fi, FI_INLINE_XATTR);
416 /* Will be used by directory only */
417 fi->i_dir_level = F2FS_SB(sb)->dir_level;
419 return &fi->vfs_inode;
422 static int f2fs_drop_inode(struct inode *inode)
425 * This is to avoid a deadlock condition like below.
426 * writeback_single_inode(inode)
427 * - f2fs_write_data_page
428 * - f2fs_gc -> iput -> evict
429 * - inode_wait_for_writeback(inode)
431 if (!inode_unhashed(inode) && inode->i_state & I_SYNC)
432 return 0;
433 return generic_drop_inode(inode);
437 * f2fs_dirty_inode() is called from __mark_inode_dirty()
439 * We should call set_dirty_inode to write the dirty inode through write_inode.
441 static void f2fs_dirty_inode(struct inode *inode, int flags)
443 set_inode_flag(F2FS_I(inode), FI_DIRTY_INODE);
446 static void f2fs_i_callback(struct rcu_head *head)
448 struct inode *inode = container_of(head, struct inode, i_rcu);
449 kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
452 static void f2fs_destroy_inode(struct inode *inode)
454 call_rcu(&inode->i_rcu, f2fs_i_callback);
457 static void f2fs_put_super(struct super_block *sb)
459 struct f2fs_sb_info *sbi = F2FS_SB(sb);
461 if (sbi->s_proc) {
462 remove_proc_entry("segment_info", sbi->s_proc);
463 remove_proc_entry(sb->s_id, f2fs_proc_root);
465 kobject_del(&sbi->s_kobj);
467 f2fs_destroy_stats(sbi);
468 stop_gc_thread(sbi);
471 * We don't need to do checkpoint when superblock is clean.
472 * But, the previous checkpoint was not done by umount, it needs to do
473 * clean checkpoint again.
475 if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
476 !is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
477 struct cp_control cpc = {
478 .reason = CP_UMOUNT,
480 write_checkpoint(sbi, &cpc);
484 * normally superblock is clean, so we need to release this.
485 * In addition, EIO will skip do checkpoint, we need this as well.
487 release_dirty_inode(sbi);
488 release_discard_addrs(sbi);
490 iput(sbi->node_inode);
491 iput(sbi->meta_inode);
493 /* destroy f2fs internal modules */
494 destroy_node_manager(sbi);
495 destroy_segment_manager(sbi);
497 kfree(sbi->ckpt);
498 kobject_put(&sbi->s_kobj);
499 wait_for_completion(&sbi->s_kobj_unregister);
501 sb->s_fs_info = NULL;
502 brelse(sbi->raw_super_buf);
503 kfree(sbi);
506 int f2fs_sync_fs(struct super_block *sb, int sync)
508 struct f2fs_sb_info *sbi = F2FS_SB(sb);
510 trace_f2fs_sync_fs(sb, sync);
512 if (sync) {
513 struct cp_control cpc;
515 cpc.reason = __get_cp_reason(sbi);
517 mutex_lock(&sbi->gc_mutex);
518 write_checkpoint(sbi, &cpc);
519 mutex_unlock(&sbi->gc_mutex);
520 } else {
521 f2fs_balance_fs(sbi);
523 f2fs_trace_ios(NULL, NULL, 1);
525 return 0;
528 static int f2fs_freeze(struct super_block *sb)
530 int err;
532 if (f2fs_readonly(sb))
533 return 0;
535 err = f2fs_sync_fs(sb, 1);
536 return err;
539 static int f2fs_unfreeze(struct super_block *sb)
541 return 0;
544 static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
546 struct super_block *sb = dentry->d_sb;
547 struct f2fs_sb_info *sbi = F2FS_SB(sb);
548 u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
549 block_t total_count, user_block_count, start_count, ovp_count;
551 total_count = le64_to_cpu(sbi->raw_super->block_count);
552 user_block_count = sbi->user_block_count;
553 start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
554 ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg;
555 buf->f_type = F2FS_SUPER_MAGIC;
556 buf->f_bsize = sbi->blocksize;
558 buf->f_blocks = total_count - start_count;
559 buf->f_bfree = buf->f_blocks - valid_user_blocks(sbi) - ovp_count;
560 buf->f_bavail = user_block_count - valid_user_blocks(sbi);
562 buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
563 buf->f_ffree = buf->f_files - valid_inode_count(sbi);
565 buf->f_namelen = F2FS_NAME_LEN;
566 buf->f_fsid.val[0] = (u32)id;
567 buf->f_fsid.val[1] = (u32)(id >> 32);
569 return 0;
572 static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
574 struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
576 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC))
577 seq_printf(seq, ",background_gc=%s", "on");
578 else
579 seq_printf(seq, ",background_gc=%s", "off");
580 if (test_opt(sbi, DISABLE_ROLL_FORWARD))
581 seq_puts(seq, ",disable_roll_forward");
582 if (test_opt(sbi, DISCARD))
583 seq_puts(seq, ",discard");
584 if (test_opt(sbi, NOHEAP))
585 seq_puts(seq, ",no_heap_alloc");
586 #ifdef CONFIG_F2FS_FS_XATTR
587 if (test_opt(sbi, XATTR_USER))
588 seq_puts(seq, ",user_xattr");
589 else
590 seq_puts(seq, ",nouser_xattr");
591 if (test_opt(sbi, INLINE_XATTR))
592 seq_puts(seq, ",inline_xattr");
593 #endif
594 #ifdef CONFIG_F2FS_FS_POSIX_ACL
595 if (test_opt(sbi, POSIX_ACL))
596 seq_puts(seq, ",acl");
597 else
598 seq_puts(seq, ",noacl");
599 #endif
600 if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
601 seq_puts(seq, ",disable_ext_identify");
602 if (test_opt(sbi, INLINE_DATA))
603 seq_puts(seq, ",inline_data");
604 else
605 seq_puts(seq, ",noinline_data");
606 if (test_opt(sbi, INLINE_DENTRY))
607 seq_puts(seq, ",inline_dentry");
608 if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
609 seq_puts(seq, ",flush_merge");
610 if (test_opt(sbi, NOBARRIER))
611 seq_puts(seq, ",nobarrier");
612 if (test_opt(sbi, FASTBOOT))
613 seq_puts(seq, ",fastboot");
614 if (test_opt(sbi, EXTENT_CACHE))
615 seq_puts(seq, ",extent_cache");
616 seq_printf(seq, ",active_logs=%u", sbi->active_logs);
618 return 0;
621 static int segment_info_seq_show(struct seq_file *seq, void *offset)
623 struct super_block *sb = seq->private;
624 struct f2fs_sb_info *sbi = F2FS_SB(sb);
625 unsigned int total_segs =
626 le32_to_cpu(sbi->raw_super->segment_count_main);
627 int i;
629 seq_puts(seq, "format: segment_type|valid_blocks\n"
630 "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n");
632 for (i = 0; i < total_segs; i++) {
633 struct seg_entry *se = get_seg_entry(sbi, i);
635 if ((i % 10) == 0)
636 seq_printf(seq, "%-5d", i);
637 seq_printf(seq, "%d|%-3u", se->type,
638 get_valid_blocks(sbi, i, 1));
639 if ((i % 10) == 9 || i == (total_segs - 1))
640 seq_putc(seq, '\n');
641 else
642 seq_putc(seq, ' ');
645 return 0;
648 static int segment_info_open_fs(struct inode *inode, struct file *file)
650 return single_open(file, segment_info_seq_show, PDE_DATA(inode));
653 static const struct file_operations f2fs_seq_segment_info_fops = {
654 .owner = THIS_MODULE,
655 .open = segment_info_open_fs,
656 .read = seq_read,
657 .llseek = seq_lseek,
658 .release = single_release,
661 static int f2fs_remount(struct super_block *sb, int *flags, char *data)
663 struct f2fs_sb_info *sbi = F2FS_SB(sb);
664 struct f2fs_mount_info org_mount_opt;
665 int err, active_logs;
666 bool need_restart_gc = false;
667 bool need_stop_gc = false;
669 sync_filesystem(sb);
672 * Save the old mount options in case we
673 * need to restore them.
675 org_mount_opt = sbi->mount_opt;
676 active_logs = sbi->active_logs;
678 sbi->mount_opt.opt = 0;
679 sbi->active_logs = NR_CURSEG_TYPE;
681 /* parse mount options */
682 err = parse_options(sb, data);
683 if (err)
684 goto restore_opts;
687 * Previous and new state of filesystem is RO,
688 * so skip checking GC and FLUSH_MERGE conditions.
690 if (f2fs_readonly(sb) && (*flags & MS_RDONLY))
691 goto skip;
694 * We stop the GC thread if FS is mounted as RO
695 * or if background_gc = off is passed in mount
696 * option. Also sync the filesystem.
698 if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) {
699 if (sbi->gc_thread) {
700 stop_gc_thread(sbi);
701 f2fs_sync_fs(sb, 1);
702 need_restart_gc = true;
704 } else if (!sbi->gc_thread) {
705 err = start_gc_thread(sbi);
706 if (err)
707 goto restore_opts;
708 need_stop_gc = true;
712 * We stop issue flush thread if FS is mounted as RO
713 * or if flush_merge is not passed in mount option.
715 if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
716 destroy_flush_cmd_control(sbi);
717 } else if (!SM_I(sbi)->cmd_control_info) {
718 err = create_flush_cmd_control(sbi);
719 if (err)
720 goto restore_gc;
722 skip:
723 /* Update the POSIXACL Flag */
724 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
725 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
726 return 0;
727 restore_gc:
728 if (need_restart_gc) {
729 if (start_gc_thread(sbi))
730 f2fs_msg(sbi->sb, KERN_WARNING,
731 "background gc thread has stopped");
732 } else if (need_stop_gc) {
733 stop_gc_thread(sbi);
735 restore_opts:
736 sbi->mount_opt = org_mount_opt;
737 sbi->active_logs = active_logs;
738 return err;
741 static struct super_operations f2fs_sops = {
742 .alloc_inode = f2fs_alloc_inode,
743 .drop_inode = f2fs_drop_inode,
744 .destroy_inode = f2fs_destroy_inode,
745 .write_inode = f2fs_write_inode,
746 .dirty_inode = f2fs_dirty_inode,
747 .show_options = f2fs_show_options,
748 .evict_inode = f2fs_evict_inode,
749 .put_super = f2fs_put_super,
750 .sync_fs = f2fs_sync_fs,
751 .freeze_fs = f2fs_freeze,
752 .unfreeze_fs = f2fs_unfreeze,
753 .statfs = f2fs_statfs,
754 .remount_fs = f2fs_remount,
757 static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
758 u64 ino, u32 generation)
760 struct f2fs_sb_info *sbi = F2FS_SB(sb);
761 struct inode *inode;
763 if (check_nid_range(sbi, ino))
764 return ERR_PTR(-ESTALE);
767 * f2fs_iget isn't quite right if the inode is currently unallocated!
768 * However f2fs_iget currently does appropriate checks to handle stale
769 * inodes so everything is OK.
771 inode = f2fs_iget(sb, ino);
772 if (IS_ERR(inode))
773 return ERR_CAST(inode);
774 if (unlikely(generation && inode->i_generation != generation)) {
775 /* we didn't find the right inode.. */
776 iput(inode);
777 return ERR_PTR(-ESTALE);
779 return inode;
782 static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
783 int fh_len, int fh_type)
785 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
786 f2fs_nfs_get_inode);
789 static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
790 int fh_len, int fh_type)
792 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
793 f2fs_nfs_get_inode);
796 static const struct export_operations f2fs_export_ops = {
797 .fh_to_dentry = f2fs_fh_to_dentry,
798 .fh_to_parent = f2fs_fh_to_parent,
799 .get_parent = f2fs_get_parent,
802 static loff_t max_file_size(unsigned bits)
804 loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS);
805 loff_t leaf_count = ADDRS_PER_BLOCK;
807 /* two direct node blocks */
808 result += (leaf_count * 2);
810 /* two indirect node blocks */
811 leaf_count *= NIDS_PER_BLOCK;
812 result += (leaf_count * 2);
814 /* one double indirect node block */
815 leaf_count *= NIDS_PER_BLOCK;
816 result += leaf_count;
818 result <<= bits;
819 return result;
822 static int sanity_check_raw_super(struct super_block *sb,
823 struct f2fs_super_block *raw_super)
825 unsigned int blocksize;
827 if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) {
828 f2fs_msg(sb, KERN_INFO,
829 "Magic Mismatch, valid(0x%x) - read(0x%x)",
830 F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
831 return 1;
834 /* Currently, support only 4KB page cache size */
835 if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
836 f2fs_msg(sb, KERN_INFO,
837 "Invalid page_cache_size (%lu), supports only 4KB\n",
838 PAGE_CACHE_SIZE);
839 return 1;
842 /* Currently, support only 4KB block size */
843 blocksize = 1 << le32_to_cpu(raw_super->log_blocksize);
844 if (blocksize != F2FS_BLKSIZE) {
845 f2fs_msg(sb, KERN_INFO,
846 "Invalid blocksize (%u), supports only 4KB\n",
847 blocksize);
848 return 1;
851 /* Currently, support 512/1024/2048/4096 bytes sector size */
852 if (le32_to_cpu(raw_super->log_sectorsize) >
853 F2FS_MAX_LOG_SECTOR_SIZE ||
854 le32_to_cpu(raw_super->log_sectorsize) <
855 F2FS_MIN_LOG_SECTOR_SIZE) {
856 f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)",
857 le32_to_cpu(raw_super->log_sectorsize));
858 return 1;
860 if (le32_to_cpu(raw_super->log_sectors_per_block) +
861 le32_to_cpu(raw_super->log_sectorsize) !=
862 F2FS_MAX_LOG_SECTOR_SIZE) {
863 f2fs_msg(sb, KERN_INFO,
864 "Invalid log sectors per block(%u) log sectorsize(%u)",
865 le32_to_cpu(raw_super->log_sectors_per_block),
866 le32_to_cpu(raw_super->log_sectorsize));
867 return 1;
869 return 0;
872 static int sanity_check_ckpt(struct f2fs_sb_info *sbi)
874 unsigned int total, fsmeta;
875 struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
876 struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
878 total = le32_to_cpu(raw_super->segment_count);
879 fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
880 fsmeta += le32_to_cpu(raw_super->segment_count_sit);
881 fsmeta += le32_to_cpu(raw_super->segment_count_nat);
882 fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
883 fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
885 if (unlikely(fsmeta >= total))
886 return 1;
888 if (unlikely(f2fs_cp_error(sbi))) {
889 f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck");
890 return 1;
892 return 0;
895 static void init_sb_info(struct f2fs_sb_info *sbi)
897 struct f2fs_super_block *raw_super = sbi->raw_super;
898 int i;
900 sbi->log_sectors_per_block =
901 le32_to_cpu(raw_super->log_sectors_per_block);
902 sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
903 sbi->blocksize = 1 << sbi->log_blocksize;
904 sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
905 sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
906 sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
907 sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
908 sbi->total_sections = le32_to_cpu(raw_super->section_count);
909 sbi->total_node_count =
910 (le32_to_cpu(raw_super->segment_count_nat) / 2)
911 * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
912 sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
913 sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
914 sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
915 sbi->cur_victim_sec = NULL_SECNO;
916 sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
918 for (i = 0; i < NR_COUNT_TYPE; i++)
919 atomic_set(&sbi->nr_pages[i], 0);
921 sbi->dir_level = DEF_DIR_LEVEL;
922 clear_sbi_flag(sbi, SBI_NEED_FSCK);
926 * Read f2fs raw super block.
927 * Because we have two copies of super block, so read the first one at first,
928 * if the first one is invalid, move to read the second one.
930 static int read_raw_super_block(struct super_block *sb,
931 struct f2fs_super_block **raw_super,
932 struct buffer_head **raw_super_buf)
934 int block = 0;
936 retry:
937 *raw_super_buf = sb_bread(sb, block);
938 if (!*raw_super_buf) {
939 f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock",
940 block + 1);
941 if (block == 0) {
942 block++;
943 goto retry;
944 } else {
945 return -EIO;
949 *raw_super = (struct f2fs_super_block *)
950 ((char *)(*raw_super_buf)->b_data + F2FS_SUPER_OFFSET);
952 /* sanity checking of raw super */
953 if (sanity_check_raw_super(sb, *raw_super)) {
954 brelse(*raw_super_buf);
955 f2fs_msg(sb, KERN_ERR,
956 "Can't find valid F2FS filesystem in %dth superblock",
957 block + 1);
958 if (block == 0) {
959 block++;
960 goto retry;
961 } else {
962 return -EINVAL;
966 return 0;
969 static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
971 struct f2fs_sb_info *sbi;
972 struct f2fs_super_block *raw_super = NULL;
973 struct buffer_head *raw_super_buf;
974 struct inode *root;
975 long err = -EINVAL;
976 bool retry = true, need_fsck = false;
977 char *options = NULL;
978 int i;
980 try_onemore:
981 /* allocate memory for f2fs-specific super block info */
982 sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
983 if (!sbi)
984 return -ENOMEM;
986 /* set a block size */
987 if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
988 f2fs_msg(sb, KERN_ERR, "unable to set blocksize");
989 goto free_sbi;
992 err = read_raw_super_block(sb, &raw_super, &raw_super_buf);
993 if (err)
994 goto free_sbi;
996 sb->s_fs_info = sbi;
997 /* init some FS parameters */
998 sbi->active_logs = NR_CURSEG_TYPE;
1000 set_opt(sbi, BG_GC);
1001 set_opt(sbi, INLINE_DATA);
1003 #ifdef CONFIG_F2FS_FS_XATTR
1004 set_opt(sbi, XATTR_USER);
1005 #endif
1006 #ifdef CONFIG_F2FS_FS_POSIX_ACL
1007 set_opt(sbi, POSIX_ACL);
1008 #endif
1009 /* parse mount options */
1010 options = kstrdup((const char *)data, GFP_KERNEL);
1011 if (data && !options) {
1012 err = -ENOMEM;
1013 goto free_sb_buf;
1016 err = parse_options(sb, options);
1017 if (err)
1018 goto free_options;
1020 sb->s_maxbytes = max_file_size(le32_to_cpu(raw_super->log_blocksize));
1021 sb->s_max_links = F2FS_LINK_MAX;
1022 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1024 sb->s_op = &f2fs_sops;
1025 sb->s_xattr = f2fs_xattr_handlers;
1026 sb->s_export_op = &f2fs_export_ops;
1027 sb->s_magic = F2FS_SUPER_MAGIC;
1028 sb->s_time_gran = 1;
1029 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1030 (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0);
1031 memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
1033 /* init f2fs-specific super block info */
1034 sbi->sb = sb;
1035 sbi->raw_super = raw_super;
1036 sbi->raw_super_buf = raw_super_buf;
1037 mutex_init(&sbi->gc_mutex);
1038 mutex_init(&sbi->writepages);
1039 mutex_init(&sbi->cp_mutex);
1040 init_rwsem(&sbi->node_write);
1041 clear_sbi_flag(sbi, SBI_POR_DOING);
1042 spin_lock_init(&sbi->stat_lock);
1044 init_rwsem(&sbi->read_io.io_rwsem);
1045 sbi->read_io.sbi = sbi;
1046 sbi->read_io.bio = NULL;
1047 for (i = 0; i < NR_PAGE_TYPE; i++) {
1048 init_rwsem(&sbi->write_io[i].io_rwsem);
1049 sbi->write_io[i].sbi = sbi;
1050 sbi->write_io[i].bio = NULL;
1053 init_rwsem(&sbi->cp_rwsem);
1054 init_waitqueue_head(&sbi->cp_wait);
1055 init_sb_info(sbi);
1057 /* get an inode for meta space */
1058 sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
1059 if (IS_ERR(sbi->meta_inode)) {
1060 f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode");
1061 err = PTR_ERR(sbi->meta_inode);
1062 goto free_options;
1065 err = get_valid_checkpoint(sbi);
1066 if (err) {
1067 f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint");
1068 goto free_meta_inode;
1071 /* sanity checking of checkpoint */
1072 err = -EINVAL;
1073 if (sanity_check_ckpt(sbi)) {
1074 f2fs_msg(sb, KERN_ERR, "Invalid F2FS checkpoint");
1075 goto free_cp;
1078 sbi->total_valid_node_count =
1079 le32_to_cpu(sbi->ckpt->valid_node_count);
1080 sbi->total_valid_inode_count =
1081 le32_to_cpu(sbi->ckpt->valid_inode_count);
1082 sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
1083 sbi->total_valid_block_count =
1084 le64_to_cpu(sbi->ckpt->valid_block_count);
1085 sbi->last_valid_block_count = sbi->total_valid_block_count;
1086 sbi->alloc_valid_block_count = 0;
1087 INIT_LIST_HEAD(&sbi->dir_inode_list);
1088 spin_lock_init(&sbi->dir_inode_lock);
1090 init_extent_cache_info(sbi);
1092 init_ino_entry_info(sbi);
1094 /* setup f2fs internal modules */
1095 err = build_segment_manager(sbi);
1096 if (err) {
1097 f2fs_msg(sb, KERN_ERR,
1098 "Failed to initialize F2FS segment manager");
1099 goto free_sm;
1101 err = build_node_manager(sbi);
1102 if (err) {
1103 f2fs_msg(sb, KERN_ERR,
1104 "Failed to initialize F2FS node manager");
1105 goto free_nm;
1108 build_gc_manager(sbi);
1110 /* get an inode for node space */
1111 sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
1112 if (IS_ERR(sbi->node_inode)) {
1113 f2fs_msg(sb, KERN_ERR, "Failed to read node inode");
1114 err = PTR_ERR(sbi->node_inode);
1115 goto free_nm;
1118 /* if there are nt orphan nodes free them */
1119 recover_orphan_inodes(sbi);
1121 /* read root inode and dentry */
1122 root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
1123 if (IS_ERR(root)) {
1124 f2fs_msg(sb, KERN_ERR, "Failed to read root inode");
1125 err = PTR_ERR(root);
1126 goto free_node_inode;
1128 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1129 iput(root);
1130 err = -EINVAL;
1131 goto free_node_inode;
1134 sb->s_root = d_make_root(root); /* allocate root dentry */
1135 if (!sb->s_root) {
1136 err = -ENOMEM;
1137 goto free_root_inode;
1140 err = f2fs_build_stats(sbi);
1141 if (err)
1142 goto free_root_inode;
1144 if (f2fs_proc_root)
1145 sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
1147 if (sbi->s_proc)
1148 proc_create_data("segment_info", S_IRUGO, sbi->s_proc,
1149 &f2fs_seq_segment_info_fops, sb);
1151 if (test_opt(sbi, DISCARD)) {
1152 struct request_queue *q = bdev_get_queue(sb->s_bdev);
1153 if (!blk_queue_discard(q))
1154 f2fs_msg(sb, KERN_WARNING,
1155 "mounting with \"discard\" option, but "
1156 "the device does not support discard");
1159 sbi->s_kobj.kset = f2fs_kset;
1160 init_completion(&sbi->s_kobj_unregister);
1161 err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL,
1162 "%s", sb->s_id);
1163 if (err)
1164 goto free_proc;
1166 /* recover fsynced data */
1167 if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) {
1169 * mount should be failed, when device has readonly mode, and
1170 * previous checkpoint was not done by clean system shutdown.
1172 if (bdev_read_only(sb->s_bdev) &&
1173 !is_set_ckpt_flags(sbi->ckpt, CP_UMOUNT_FLAG)) {
1174 err = -EROFS;
1175 goto free_kobj;
1178 if (need_fsck)
1179 set_sbi_flag(sbi, SBI_NEED_FSCK);
1181 err = recover_fsync_data(sbi);
1182 if (err) {
1183 need_fsck = true;
1184 f2fs_msg(sb, KERN_ERR,
1185 "Cannot recover all fsync data errno=%ld", err);
1186 goto free_kobj;
1191 * If filesystem is not mounted as read-only then
1192 * do start the gc_thread.
1194 if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) {
1195 /* After POR, we can run background GC thread.*/
1196 err = start_gc_thread(sbi);
1197 if (err)
1198 goto free_kobj;
1200 kfree(options);
1201 return 0;
1203 free_kobj:
1204 kobject_del(&sbi->s_kobj);
1205 free_proc:
1206 if (sbi->s_proc) {
1207 remove_proc_entry("segment_info", sbi->s_proc);
1208 remove_proc_entry(sb->s_id, f2fs_proc_root);
1210 f2fs_destroy_stats(sbi);
1211 free_root_inode:
1212 dput(sb->s_root);
1213 sb->s_root = NULL;
1214 free_node_inode:
1215 iput(sbi->node_inode);
1216 free_nm:
1217 destroy_node_manager(sbi);
1218 free_sm:
1219 destroy_segment_manager(sbi);
1220 free_cp:
1221 kfree(sbi->ckpt);
1222 free_meta_inode:
1223 make_bad_inode(sbi->meta_inode);
1224 iput(sbi->meta_inode);
1225 free_options:
1226 kfree(options);
1227 free_sb_buf:
1228 brelse(raw_super_buf);
1229 free_sbi:
1230 kfree(sbi);
1232 /* give only one another chance */
1233 if (retry) {
1234 retry = false;
1235 shrink_dcache_sb(sb);
1236 goto try_onemore;
1238 return err;
1241 static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
1242 const char *dev_name, void *data)
1244 return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
1247 static void kill_f2fs_super(struct super_block *sb)
1249 if (sb->s_root)
1250 set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE);
1251 kill_block_super(sb);
1254 static struct file_system_type f2fs_fs_type = {
1255 .owner = THIS_MODULE,
1256 .name = "f2fs",
1257 .mount = f2fs_mount,
1258 .kill_sb = kill_f2fs_super,
1259 .fs_flags = FS_REQUIRES_DEV,
1261 MODULE_ALIAS_FS("f2fs");
1263 static int __init init_inodecache(void)
1265 f2fs_inode_cachep = f2fs_kmem_cache_create("f2fs_inode_cache",
1266 sizeof(struct f2fs_inode_info));
1267 if (!f2fs_inode_cachep)
1268 return -ENOMEM;
1269 return 0;
1272 static void destroy_inodecache(void)
1275 * Make sure all delayed rcu free inodes are flushed before we
1276 * destroy cache.
1278 rcu_barrier();
1279 kmem_cache_destroy(f2fs_inode_cachep);
1282 static int __init init_f2fs_fs(void)
1284 int err;
1286 f2fs_build_trace_ios();
1288 err = init_inodecache();
1289 if (err)
1290 goto fail;
1291 err = create_node_manager_caches();
1292 if (err)
1293 goto free_inodecache;
1294 err = create_segment_manager_caches();
1295 if (err)
1296 goto free_node_manager_caches;
1297 err = create_checkpoint_caches();
1298 if (err)
1299 goto free_segment_manager_caches;
1300 err = create_extent_cache();
1301 if (err)
1302 goto free_checkpoint_caches;
1303 f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj);
1304 if (!f2fs_kset) {
1305 err = -ENOMEM;
1306 goto free_extent_cache;
1308 err = register_filesystem(&f2fs_fs_type);
1309 if (err)
1310 goto free_kset;
1311 f2fs_create_root_stats();
1312 f2fs_proc_root = proc_mkdir("fs/f2fs", NULL);
1313 return 0;
1315 free_kset:
1316 kset_unregister(f2fs_kset);
1317 free_extent_cache:
1318 destroy_extent_cache();
1319 free_checkpoint_caches:
1320 destroy_checkpoint_caches();
1321 free_segment_manager_caches:
1322 destroy_segment_manager_caches();
1323 free_node_manager_caches:
1324 destroy_node_manager_caches();
1325 free_inodecache:
1326 destroy_inodecache();
1327 fail:
1328 return err;
1331 static void __exit exit_f2fs_fs(void)
1333 remove_proc_entry("fs/f2fs", NULL);
1334 f2fs_destroy_root_stats();
1335 unregister_filesystem(&f2fs_fs_type);
1336 destroy_extent_cache();
1337 destroy_checkpoint_caches();
1338 destroy_segment_manager_caches();
1339 destroy_node_manager_caches();
1340 destroy_inodecache();
1341 kset_unregister(f2fs_kset);
1342 f2fs_destroy_trace_ios();
1345 module_init(init_f2fs_fs)
1346 module_exit(exit_f2fs_fs)
1348 MODULE_AUTHOR("Samsung Electronics's Praesto Team");
1349 MODULE_DESCRIPTION("Flash Friendly File System");
1350 MODULE_LICENSE("GPL");