xref: /kernel/linux/linux-5.10/fs/f2fs/super.c (revision 8c2ecf20)
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * fs/f2fs/super.c
4 *
5 * Copyright (c) 2012 Samsung Electronics Co., Ltd.
6 *             http://www.samsung.com/
7 */
8#include <linux/module.h>
9#include <linux/init.h>
10#include <linux/fs.h>
11#include <linux/statfs.h>
12#include <linux/buffer_head.h>
13#include <linux/backing-dev.h>
14#include <linux/kthread.h>
15#include <linux/parser.h>
16#include <linux/mount.h>
17#include <linux/seq_file.h>
18#include <linux/proc_fs.h>
19#include <linux/random.h>
20#include <linux/exportfs.h>
21#include <linux/blkdev.h>
22#include <linux/quotaops.h>
23#include <linux/f2fs_fs.h>
24#include <linux/sysfs.h>
25#include <linux/quota.h>
26#include <linux/unicode.h>
27#include <linux/part_stat.h>
28
29#include "f2fs.h"
30#include "node.h"
31#include "segment.h"
32#include "xattr.h"
33#include "gc.h"
34#include "trace.h"
35
36#define CREATE_TRACE_POINTS
37#include <trace/events/f2fs.h>
38
39#ifdef CONFIG_F2FS_GRADING_SSR
40#define SSR_DEFALT_SPACE_LIMIT      (5<<20) /* 5G default space limit */
41#define SSR_DEFALT_WATERLINE        80      /* 80% default waterline */
42#define SSR_HN_SAPCE_LIMIT_128G     (8<<20) /* 8G default sapce limit for 128G devices */
43#define SSR_HN_WATERLINE_128G       80      /* 80% default hot node waterline for 128G devices */
44#define SSR_WN_SAPCE_LIMIT_128G     (5<<20) /* 5G default warm node sapce limit for 128G devices */
45#define SSR_WN_WATERLINE_128G       70      /* 70% default warm node waterline for 128G devices */
46#define SSR_HD_SAPCE_LIMIT_128G     (8<<20) /* 8G default hot data sapce limit for 128G devices */
47#define SSR_HD_WATERLINE_128G       65      /* 65% default hot data waterline for 128G devices */
48#define SSR_WD_SAPCE_LIMIT_128G     (5<<20) /* 5G default warm data sapce limit for 128G devices */
49#define SSR_WD_WATERLINE_128G       60      /* 60% default warm data waterline for 128G devices */
50#endif
51
52static struct kmem_cache *f2fs_inode_cachep;
53
54#ifdef CONFIG_F2FS_FAULT_INJECTION
55
56const char *f2fs_fault_name[FAULT_MAX] = {
57	[FAULT_KMALLOC]		= "kmalloc",
58	[FAULT_KVMALLOC]	= "kvmalloc",
59	[FAULT_PAGE_ALLOC]	= "page alloc",
60	[FAULT_PAGE_GET]	= "page get",
61	[FAULT_ALLOC_BIO]	= "alloc bio",
62	[FAULT_ALLOC_NID]	= "alloc nid",
63	[FAULT_ORPHAN]		= "orphan",
64	[FAULT_BLOCK]		= "no more block",
65	[FAULT_DIR_DEPTH]	= "too big dir depth",
66	[FAULT_EVICT_INODE]	= "evict_inode fail",
67	[FAULT_TRUNCATE]	= "truncate fail",
68	[FAULT_READ_IO]		= "read IO error",
69	[FAULT_CHECKPOINT]	= "checkpoint error",
70	[FAULT_DISCARD]		= "discard error",
71	[FAULT_WRITE_IO]	= "write IO error",
72};
73
74int f2fs_build_fault_attr(struct f2fs_sb_info *sbi, unsigned long rate,
75							unsigned long type)
76{
77	struct f2fs_fault_info *ffi = &F2FS_OPTION(sbi).fault_info;
78
79	if (rate) {
80		if (rate > INT_MAX)
81			return -EINVAL;
82		atomic_set(&ffi->inject_ops, 0);
83		ffi->inject_rate = (int)rate;
84	}
85
86	if (type) {
87		if (type >= BIT(FAULT_MAX))
88			return -EINVAL;
89		ffi->inject_type = (unsigned int)type;
90	}
91
92	if (!rate && !type)
93		memset(ffi, 0, sizeof(struct f2fs_fault_info));
94	else
95		f2fs_info(sbi,
96			"build fault injection attr: rate: %lu, type: 0x%lx",
97								rate, type);
98	return 0;
99}
100#endif
101
102/* f2fs-wide shrinker description */
103static struct shrinker f2fs_shrinker_info = {
104	.scan_objects = f2fs_shrink_scan,
105	.count_objects = f2fs_shrink_count,
106	.seeks = DEFAULT_SEEKS,
107};
108
109enum {
110	Opt_gc_background,
111	Opt_disable_roll_forward,
112	Opt_norecovery,
113	Opt_discard,
114	Opt_nodiscard,
115	Opt_noheap,
116	Opt_heap,
117	Opt_user_xattr,
118	Opt_nouser_xattr,
119	Opt_acl,
120	Opt_noacl,
121	Opt_active_logs,
122	Opt_disable_ext_identify,
123	Opt_inline_xattr,
124	Opt_noinline_xattr,
125	Opt_inline_xattr_size,
126	Opt_inline_data,
127	Opt_inline_dentry,
128	Opt_noinline_dentry,
129	Opt_flush_merge,
130	Opt_noflush_merge,
131	Opt_nobarrier,
132	Opt_fastboot,
133	Opt_extent_cache,
134	Opt_noextent_cache,
135	Opt_noinline_data,
136	Opt_data_flush,
137	Opt_reserve_root,
138	Opt_resgid,
139	Opt_resuid,
140	Opt_mode,
141	Opt_io_size_bits,
142	Opt_fault_injection,
143	Opt_fault_type,
144	Opt_lazytime,
145	Opt_nolazytime,
146	Opt_quota,
147	Opt_noquota,
148	Opt_usrquota,
149	Opt_grpquota,
150	Opt_prjquota,
151	Opt_usrjquota,
152	Opt_grpjquota,
153	Opt_prjjquota,
154	Opt_offusrjquota,
155	Opt_offgrpjquota,
156	Opt_offprjjquota,
157	Opt_jqfmt_vfsold,
158	Opt_jqfmt_vfsv0,
159	Opt_jqfmt_vfsv1,
160	Opt_whint,
161	Opt_alloc,
162	Opt_fsync,
163	Opt_test_dummy_encryption,
164	Opt_inlinecrypt,
165	Opt_checkpoint_disable,
166	Opt_checkpoint_disable_cap,
167	Opt_checkpoint_disable_cap_perc,
168	Opt_checkpoint_enable,
169	Opt_compress_algorithm,
170	Opt_compress_log_size,
171	Opt_compress_extension,
172	Opt_atgc,
173	Opt_gc_merge,
174	Opt_nogc_merge,
175	Opt_err,
176};
177
178static match_table_t f2fs_tokens = {
179	{Opt_gc_background, "background_gc=%s"},
180	{Opt_disable_roll_forward, "disable_roll_forward"},
181	{Opt_norecovery, "norecovery"},
182	{Opt_discard, "discard"},
183	{Opt_nodiscard, "nodiscard"},
184	{Opt_noheap, "no_heap"},
185	{Opt_heap, "heap"},
186	{Opt_user_xattr, "user_xattr"},
187	{Opt_nouser_xattr, "nouser_xattr"},
188	{Opt_acl, "acl"},
189	{Opt_noacl, "noacl"},
190	{Opt_active_logs, "active_logs=%u"},
191	{Opt_disable_ext_identify, "disable_ext_identify"},
192	{Opt_inline_xattr, "inline_xattr"},
193	{Opt_noinline_xattr, "noinline_xattr"},
194	{Opt_inline_xattr_size, "inline_xattr_size=%u"},
195	{Opt_inline_data, "inline_data"},
196	{Opt_inline_dentry, "inline_dentry"},
197	{Opt_noinline_dentry, "noinline_dentry"},
198	{Opt_flush_merge, "flush_merge"},
199	{Opt_noflush_merge, "noflush_merge"},
200	{Opt_nobarrier, "nobarrier"},
201	{Opt_fastboot, "fastboot"},
202	{Opt_extent_cache, "extent_cache"},
203	{Opt_noextent_cache, "noextent_cache"},
204	{Opt_noinline_data, "noinline_data"},
205	{Opt_data_flush, "data_flush"},
206	{Opt_reserve_root, "reserve_root=%u"},
207	{Opt_resgid, "resgid=%u"},
208	{Opt_resuid, "resuid=%u"},
209	{Opt_mode, "mode=%s"},
210	{Opt_io_size_bits, "io_bits=%u"},
211	{Opt_fault_injection, "fault_injection=%u"},
212	{Opt_fault_type, "fault_type=%u"},
213	{Opt_lazytime, "lazytime"},
214	{Opt_nolazytime, "nolazytime"},
215	{Opt_quota, "quota"},
216	{Opt_noquota, "noquota"},
217	{Opt_usrquota, "usrquota"},
218	{Opt_grpquota, "grpquota"},
219	{Opt_prjquota, "prjquota"},
220	{Opt_usrjquota, "usrjquota=%s"},
221	{Opt_grpjquota, "grpjquota=%s"},
222	{Opt_prjjquota, "prjjquota=%s"},
223	{Opt_offusrjquota, "usrjquota="},
224	{Opt_offgrpjquota, "grpjquota="},
225	{Opt_offprjjquota, "prjjquota="},
226	{Opt_jqfmt_vfsold, "jqfmt=vfsold"},
227	{Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
228	{Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
229	{Opt_whint, "whint_mode=%s"},
230	{Opt_alloc, "alloc_mode=%s"},
231	{Opt_fsync, "fsync_mode=%s"},
232	{Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
233	{Opt_test_dummy_encryption, "test_dummy_encryption"},
234	{Opt_inlinecrypt, "inlinecrypt"},
235	{Opt_checkpoint_disable, "checkpoint=disable"},
236	{Opt_checkpoint_disable_cap, "checkpoint=disable:%u"},
237	{Opt_checkpoint_disable_cap_perc, "checkpoint=disable:%u%%"},
238	{Opt_checkpoint_enable, "checkpoint=enable"},
239	{Opt_compress_algorithm, "compress_algorithm=%s"},
240	{Opt_compress_log_size, "compress_log_size=%u"},
241	{Opt_compress_extension, "compress_extension=%s"},
242	{Opt_atgc, "atgc"},
243	{Opt_gc_merge, "gc_merge"},
244	{Opt_nogc_merge, "nogc_merge"},
245	{Opt_err, NULL},
246};
247
248void f2fs_printk(struct f2fs_sb_info *sbi, const char *fmt, ...)
249{
250	struct va_format vaf;
251	va_list args;
252	int level;
253
254	va_start(args, fmt);
255
256	level = printk_get_level(fmt);
257	vaf.fmt = printk_skip_level(fmt);
258	vaf.va = &args;
259	printk("%c%cF2FS-fs (%s): %pV\n",
260	       KERN_SOH_ASCII, level, sbi->sb->s_id, &vaf);
261
262	va_end(args);
263}
264
265#ifdef CONFIG_UNICODE
266static const struct f2fs_sb_encodings {
267	__u16 magic;
268	char *name;
269	char *version;
270} f2fs_sb_encoding_map[] = {
271	{F2FS_ENC_UTF8_12_1, "utf8", "12.1.0"},
272};
273
274static int f2fs_sb_read_encoding(const struct f2fs_super_block *sb,
275				 const struct f2fs_sb_encodings **encoding,
276				 __u16 *flags)
277{
278	__u16 magic = le16_to_cpu(sb->s_encoding);
279	int i;
280
281	for (i = 0; i < ARRAY_SIZE(f2fs_sb_encoding_map); i++)
282		if (magic == f2fs_sb_encoding_map[i].magic)
283			break;
284
285	if (i >= ARRAY_SIZE(f2fs_sb_encoding_map))
286		return -EINVAL;
287
288	*encoding = &f2fs_sb_encoding_map[i];
289	*flags = le16_to_cpu(sb->s_encoding_flags);
290
291	return 0;
292}
293#endif
294
295static inline void limit_reserve_root(struct f2fs_sb_info *sbi)
296{
297	block_t limit = min((sbi->user_block_count >> 3),
298			sbi->user_block_count - sbi->reserved_blocks);
299
300	/* limit is 12.5% */
301	if (test_opt(sbi, RESERVE_ROOT) &&
302			F2FS_OPTION(sbi).root_reserved_blocks > limit) {
303		F2FS_OPTION(sbi).root_reserved_blocks = limit;
304		f2fs_info(sbi, "Reduce reserved blocks for root = %u",
305			  F2FS_OPTION(sbi).root_reserved_blocks);
306	}
307	if (!test_opt(sbi, RESERVE_ROOT) &&
308		(!uid_eq(F2FS_OPTION(sbi).s_resuid,
309				make_kuid(&init_user_ns, F2FS_DEF_RESUID)) ||
310		!gid_eq(F2FS_OPTION(sbi).s_resgid,
311				make_kgid(&init_user_ns, F2FS_DEF_RESGID))))
312		f2fs_info(sbi, "Ignore s_resuid=%u, s_resgid=%u w/o reserve_root",
313			  from_kuid_munged(&init_user_ns,
314					   F2FS_OPTION(sbi).s_resuid),
315			  from_kgid_munged(&init_user_ns,
316					   F2FS_OPTION(sbi).s_resgid));
317}
318
319static inline int adjust_reserved_segment(struct f2fs_sb_info *sbi)
320{
321	unsigned int sec_blks = sbi->blocks_per_seg * sbi->segs_per_sec;
322	unsigned int avg_vblocks;
323	unsigned int wanted_reserved_segments;
324	block_t avail_user_block_count;
325
326	if (!F2FS_IO_ALIGNED(sbi))
327		return 0;
328
329	/* average valid block count in section in worst case */
330	avg_vblocks = sec_blks / F2FS_IO_SIZE(sbi);
331
332	/*
333	 * we need enough free space when migrating one section in worst case
334	 */
335	wanted_reserved_segments = (F2FS_IO_SIZE(sbi) / avg_vblocks) *
336						reserved_segments(sbi);
337	wanted_reserved_segments -= reserved_segments(sbi);
338
339	avail_user_block_count = sbi->user_block_count -
340				sbi->current_reserved_blocks -
341				F2FS_OPTION(sbi).root_reserved_blocks;
342
343	if (wanted_reserved_segments * sbi->blocks_per_seg >
344					avail_user_block_count) {
345		f2fs_err(sbi, "IO align feature can't grab additional reserved segment: %u, available segments: %u",
346			wanted_reserved_segments,
347			avail_user_block_count >> sbi->log_blocks_per_seg);
348		return -ENOSPC;
349	}
350
351	SM_I(sbi)->additional_reserved_segments = wanted_reserved_segments;
352
353	f2fs_info(sbi, "IO align feature needs additional reserved segment: %u",
354			 wanted_reserved_segments);
355
356	return 0;
357}
358
359static inline void adjust_unusable_cap_perc(struct f2fs_sb_info *sbi)
360{
361	if (!F2FS_OPTION(sbi).unusable_cap_perc)
362		return;
363
364	if (F2FS_OPTION(sbi).unusable_cap_perc == 100)
365		F2FS_OPTION(sbi).unusable_cap = sbi->user_block_count;
366	else
367		F2FS_OPTION(sbi).unusable_cap = (sbi->user_block_count / 100) *
368					F2FS_OPTION(sbi).unusable_cap_perc;
369
370	f2fs_info(sbi, "Adjust unusable cap for checkpoint=disable = %u / %u%%",
371			F2FS_OPTION(sbi).unusable_cap,
372			F2FS_OPTION(sbi).unusable_cap_perc);
373}
374
375static void init_once(void *foo)
376{
377	struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo;
378
379	inode_init_once(&fi->vfs_inode);
380}
381
382#ifdef CONFIG_QUOTA
383static const char * const quotatypes[] = INITQFNAMES;
384#define QTYPE2NAME(t) (quotatypes[t])
385static int f2fs_set_qf_name(struct super_block *sb, int qtype,
386							substring_t *args)
387{
388	struct f2fs_sb_info *sbi = F2FS_SB(sb);
389	char *qname;
390	int ret = -EINVAL;
391
392	if (sb_any_quota_loaded(sb) && !F2FS_OPTION(sbi).s_qf_names[qtype]) {
393		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
394		return -EINVAL;
395	}
396	if (f2fs_sb_has_quota_ino(sbi)) {
397		f2fs_info(sbi, "QUOTA feature is enabled, so ignore qf_name");
398		return 0;
399	}
400
401	qname = match_strdup(args);
402	if (!qname) {
403		f2fs_err(sbi, "Not enough memory for storing quotafile name");
404		return -ENOMEM;
405	}
406	if (F2FS_OPTION(sbi).s_qf_names[qtype]) {
407		if (strcmp(F2FS_OPTION(sbi).s_qf_names[qtype], qname) == 0)
408			ret = 0;
409		else
410			f2fs_err(sbi, "%s quota file already specified",
411				 QTYPE2NAME(qtype));
412		goto errout;
413	}
414	if (strchr(qname, '/')) {
415		f2fs_err(sbi, "quotafile must be on filesystem root");
416		goto errout;
417	}
418	F2FS_OPTION(sbi).s_qf_names[qtype] = qname;
419	set_opt(sbi, QUOTA);
420	return 0;
421errout:
422	kfree(qname);
423	return ret;
424}
425
426static int f2fs_clear_qf_name(struct super_block *sb, int qtype)
427{
428	struct f2fs_sb_info *sbi = F2FS_SB(sb);
429
430	if (sb_any_quota_loaded(sb) && F2FS_OPTION(sbi).s_qf_names[qtype]) {
431		f2fs_err(sbi, "Cannot change journaled quota options when quota turned on");
432		return -EINVAL;
433	}
434	kfree(F2FS_OPTION(sbi).s_qf_names[qtype]);
435	F2FS_OPTION(sbi).s_qf_names[qtype] = NULL;
436	return 0;
437}
438
439static int f2fs_check_quota_options(struct f2fs_sb_info *sbi)
440{
441	/*
442	 * We do the test below only for project quotas. 'usrquota' and
443	 * 'grpquota' mount options are allowed even without quota feature
444	 * to support legacy quotas in quota files.
445	 */
446	if (test_opt(sbi, PRJQUOTA) && !f2fs_sb_has_project_quota(sbi)) {
447		f2fs_err(sbi, "Project quota feature not enabled. Cannot enable project quota enforcement.");
448		return -1;
449	}
450	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA] ||
451			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA] ||
452			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]) {
453		if (test_opt(sbi, USRQUOTA) &&
454				F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
455			clear_opt(sbi, USRQUOTA);
456
457		if (test_opt(sbi, GRPQUOTA) &&
458				F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
459			clear_opt(sbi, GRPQUOTA);
460
461		if (test_opt(sbi, PRJQUOTA) &&
462				F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
463			clear_opt(sbi, PRJQUOTA);
464
465		if (test_opt(sbi, GRPQUOTA) || test_opt(sbi, USRQUOTA) ||
466				test_opt(sbi, PRJQUOTA)) {
467			f2fs_err(sbi, "old and new quota format mixing");
468			return -1;
469		}
470
471		if (!F2FS_OPTION(sbi).s_jquota_fmt) {
472			f2fs_err(sbi, "journaled quota format not specified");
473			return -1;
474		}
475	}
476
477	if (f2fs_sb_has_quota_ino(sbi) && F2FS_OPTION(sbi).s_jquota_fmt) {
478		f2fs_info(sbi, "QUOTA feature is enabled, so ignore jquota_fmt");
479		F2FS_OPTION(sbi).s_jquota_fmt = 0;
480	}
481	return 0;
482}
483#endif
484
485static int f2fs_set_test_dummy_encryption(struct super_block *sb,
486					  const char *opt,
487					  const substring_t *arg,
488					  bool is_remount)
489{
490	struct f2fs_sb_info *sbi = F2FS_SB(sb);
491#ifdef CONFIG_FS_ENCRYPTION
492	int err;
493
494	if (!f2fs_sb_has_encrypt(sbi)) {
495		f2fs_err(sbi, "Encrypt feature is off");
496		return -EINVAL;
497	}
498
499	/*
500	 * This mount option is just for testing, and it's not worthwhile to
501	 * implement the extra complexity (e.g. RCU protection) that would be
502	 * needed to allow it to be set or changed during remount.  We do allow
503	 * it to be specified during remount, but only if there is no change.
504	 */
505	if (is_remount && !F2FS_OPTION(sbi).dummy_enc_policy.policy) {
506		f2fs_warn(sbi, "Can't set test_dummy_encryption on remount");
507		return -EINVAL;
508	}
509	err = fscrypt_set_test_dummy_encryption(
510		sb, arg->from, &F2FS_OPTION(sbi).dummy_enc_policy);
511	if (err) {
512		if (err == -EEXIST)
513			f2fs_warn(sbi,
514				  "Can't change test_dummy_encryption on remount");
515		else if (err == -EINVAL)
516			f2fs_warn(sbi, "Value of option \"%s\" is unrecognized",
517				  opt);
518		else
519			f2fs_warn(sbi, "Error processing option \"%s\" [%d]",
520				  opt, err);
521		return -EINVAL;
522	}
523	f2fs_warn(sbi, "Test dummy encryption mode enabled");
524#else
525	f2fs_warn(sbi, "Test dummy encryption mount option ignored");
526#endif
527	return 0;
528}
529
530static int parse_options(struct super_block *sb, char *options, bool is_remount)
531{
532	struct f2fs_sb_info *sbi = F2FS_SB(sb);
533	substring_t args[MAX_OPT_ARGS];
534#ifdef CONFIG_F2FS_FS_COMPRESSION
535	unsigned char (*ext)[F2FS_EXTENSION_LEN];
536	int ext_cnt;
537#endif
538	char *p, *name;
539	int arg = 0;
540	kuid_t uid;
541	kgid_t gid;
542	int ret;
543
544	if (!options)
545		return 0;
546
547	while ((p = strsep(&options, ",")) != NULL) {
548		int token;
549		if (!*p)
550			continue;
551		/*
552		 * Initialize args struct so we know whether arg was
553		 * found; some options take optional arguments.
554		 */
555		args[0].to = args[0].from = NULL;
556		token = match_token(p, f2fs_tokens, args);
557
558		switch (token) {
559		case Opt_gc_background:
560			name = match_strdup(&args[0]);
561
562			if (!name)
563				return -ENOMEM;
564			if (!strcmp(name, "on")) {
565				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
566			} else if (!strcmp(name, "off")) {
567				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_OFF;
568			} else if (!strcmp(name, "sync")) {
569				F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_SYNC;
570			} else {
571				kfree(name);
572				return -EINVAL;
573			}
574			kfree(name);
575			break;
576		case Opt_disable_roll_forward:
577			set_opt(sbi, DISABLE_ROLL_FORWARD);
578			break;
579		case Opt_norecovery:
580			/* this option mounts f2fs with ro */
581			set_opt(sbi, NORECOVERY);
582			if (!f2fs_readonly(sb))
583				return -EINVAL;
584			break;
585		case Opt_discard:
586			set_opt(sbi, DISCARD);
587			break;
588		case Opt_nodiscard:
589			if (f2fs_sb_has_blkzoned(sbi)) {
590				f2fs_warn(sbi, "discard is required for zoned block devices");
591				return -EINVAL;
592			}
593			clear_opt(sbi, DISCARD);
594			break;
595		case Opt_noheap:
596			set_opt(sbi, NOHEAP);
597			break;
598		case Opt_heap:
599			clear_opt(sbi, NOHEAP);
600			break;
601#ifdef CONFIG_F2FS_FS_XATTR
602		case Opt_user_xattr:
603			set_opt(sbi, XATTR_USER);
604			break;
605		case Opt_nouser_xattr:
606			clear_opt(sbi, XATTR_USER);
607			break;
608		case Opt_inline_xattr:
609			set_opt(sbi, INLINE_XATTR);
610			break;
611		case Opt_noinline_xattr:
612			clear_opt(sbi, INLINE_XATTR);
613			break;
614		case Opt_inline_xattr_size:
615			if (args->from && match_int(args, &arg))
616				return -EINVAL;
617			set_opt(sbi, INLINE_XATTR_SIZE);
618			F2FS_OPTION(sbi).inline_xattr_size = arg;
619			break;
620#else
621		case Opt_user_xattr:
622			f2fs_info(sbi, "user_xattr options not supported");
623			break;
624		case Opt_nouser_xattr:
625			f2fs_info(sbi, "nouser_xattr options not supported");
626			break;
627		case Opt_inline_xattr:
628			f2fs_info(sbi, "inline_xattr options not supported");
629			break;
630		case Opt_noinline_xattr:
631			f2fs_info(sbi, "noinline_xattr options not supported");
632			break;
633#endif
634#ifdef CONFIG_F2FS_FS_POSIX_ACL
635		case Opt_acl:
636			set_opt(sbi, POSIX_ACL);
637			break;
638		case Opt_noacl:
639			clear_opt(sbi, POSIX_ACL);
640			break;
641#else
642		case Opt_acl:
643			f2fs_info(sbi, "acl options not supported");
644			break;
645		case Opt_noacl:
646			f2fs_info(sbi, "noacl options not supported");
647			break;
648#endif
649		case Opt_active_logs:
650			if (args->from && match_int(args, &arg))
651				return -EINVAL;
652			if (arg != 2 && arg != 4 &&
653				arg != NR_CURSEG_PERSIST_TYPE)
654				return -EINVAL;
655			F2FS_OPTION(sbi).active_logs = arg;
656			break;
657		case Opt_disable_ext_identify:
658			set_opt(sbi, DISABLE_EXT_IDENTIFY);
659			break;
660		case Opt_inline_data:
661			set_opt(sbi, INLINE_DATA);
662			break;
663		case Opt_inline_dentry:
664			set_opt(sbi, INLINE_DENTRY);
665			break;
666		case Opt_noinline_dentry:
667			clear_opt(sbi, INLINE_DENTRY);
668			break;
669		case Opt_flush_merge:
670			set_opt(sbi, FLUSH_MERGE);
671			break;
672		case Opt_noflush_merge:
673			clear_opt(sbi, FLUSH_MERGE);
674			break;
675		case Opt_nobarrier:
676			set_opt(sbi, NOBARRIER);
677			break;
678		case Opt_fastboot:
679			set_opt(sbi, FASTBOOT);
680			break;
681		case Opt_extent_cache:
682			set_opt(sbi, EXTENT_CACHE);
683			break;
684		case Opt_noextent_cache:
685			clear_opt(sbi, EXTENT_CACHE);
686			break;
687		case Opt_noinline_data:
688			clear_opt(sbi, INLINE_DATA);
689			break;
690		case Opt_data_flush:
691			set_opt(sbi, DATA_FLUSH);
692			break;
693		case Opt_reserve_root:
694			if (args->from && match_int(args, &arg))
695				return -EINVAL;
696			if (test_opt(sbi, RESERVE_ROOT)) {
697				f2fs_info(sbi, "Preserve previous reserve_root=%u",
698					  F2FS_OPTION(sbi).root_reserved_blocks);
699			} else {
700				F2FS_OPTION(sbi).root_reserved_blocks = arg;
701				set_opt(sbi, RESERVE_ROOT);
702			}
703			break;
704		case Opt_resuid:
705			if (args->from && match_int(args, &arg))
706				return -EINVAL;
707			uid = make_kuid(current_user_ns(), arg);
708			if (!uid_valid(uid)) {
709				f2fs_err(sbi, "Invalid uid value %d", arg);
710				return -EINVAL;
711			}
712			F2FS_OPTION(sbi).s_resuid = uid;
713			break;
714		case Opt_resgid:
715			if (args->from && match_int(args, &arg))
716				return -EINVAL;
717			gid = make_kgid(current_user_ns(), arg);
718			if (!gid_valid(gid)) {
719				f2fs_err(sbi, "Invalid gid value %d", arg);
720				return -EINVAL;
721			}
722			F2FS_OPTION(sbi).s_resgid = gid;
723			break;
724		case Opt_mode:
725			name = match_strdup(&args[0]);
726
727			if (!name)
728				return -ENOMEM;
729			if (!strcmp(name, "adaptive")) {
730				if (f2fs_sb_has_blkzoned(sbi)) {
731					f2fs_warn(sbi, "adaptive mode is not allowed with zoned block device feature");
732					kfree(name);
733					return -EINVAL;
734				}
735				F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
736			} else if (!strcmp(name, "lfs")) {
737				F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
738			} else {
739				kfree(name);
740				return -EINVAL;
741			}
742			kfree(name);
743			break;
744		case Opt_io_size_bits:
745			if (args->from && match_int(args, &arg))
746				return -EINVAL;
747			if (arg <= 0 || arg > __ilog2_u32(BIO_MAX_PAGES)) {
748				f2fs_warn(sbi, "Not support %d, larger than %d",
749					  1 << arg, BIO_MAX_PAGES);
750				return -EINVAL;
751			}
752			F2FS_OPTION(sbi).write_io_size_bits = arg;
753			break;
754#ifdef CONFIG_F2FS_FAULT_INJECTION
755		case Opt_fault_injection:
756			if (args->from && match_int(args, &arg))
757				return -EINVAL;
758			if (f2fs_build_fault_attr(sbi, arg,
759					F2FS_ALL_FAULT_TYPE))
760				return -EINVAL;
761			set_opt(sbi, FAULT_INJECTION);
762			break;
763
764		case Opt_fault_type:
765			if (args->from && match_int(args, &arg))
766				return -EINVAL;
767			if (f2fs_build_fault_attr(sbi, 0, arg))
768				return -EINVAL;
769			set_opt(sbi, FAULT_INJECTION);
770			break;
771#else
772		case Opt_fault_injection:
773			f2fs_info(sbi, "fault_injection options not supported");
774			break;
775
776		case Opt_fault_type:
777			f2fs_info(sbi, "fault_type options not supported");
778			break;
779#endif
780		case Opt_lazytime:
781			sb->s_flags |= SB_LAZYTIME;
782			break;
783		case Opt_nolazytime:
784			sb->s_flags &= ~SB_LAZYTIME;
785			break;
786#ifdef CONFIG_QUOTA
787		case Opt_quota:
788		case Opt_usrquota:
789			set_opt(sbi, USRQUOTA);
790			break;
791		case Opt_grpquota:
792			set_opt(sbi, GRPQUOTA);
793			break;
794		case Opt_prjquota:
795			set_opt(sbi, PRJQUOTA);
796			break;
797		case Opt_usrjquota:
798			ret = f2fs_set_qf_name(sb, USRQUOTA, &args[0]);
799			if (ret)
800				return ret;
801			break;
802		case Opt_grpjquota:
803			ret = f2fs_set_qf_name(sb, GRPQUOTA, &args[0]);
804			if (ret)
805				return ret;
806			break;
807		case Opt_prjjquota:
808			ret = f2fs_set_qf_name(sb, PRJQUOTA, &args[0]);
809			if (ret)
810				return ret;
811			break;
812		case Opt_offusrjquota:
813			ret = f2fs_clear_qf_name(sb, USRQUOTA);
814			if (ret)
815				return ret;
816			break;
817		case Opt_offgrpjquota:
818			ret = f2fs_clear_qf_name(sb, GRPQUOTA);
819			if (ret)
820				return ret;
821			break;
822		case Opt_offprjjquota:
823			ret = f2fs_clear_qf_name(sb, PRJQUOTA);
824			if (ret)
825				return ret;
826			break;
827		case Opt_jqfmt_vfsold:
828			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_OLD;
829			break;
830		case Opt_jqfmt_vfsv0:
831			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V0;
832			break;
833		case Opt_jqfmt_vfsv1:
834			F2FS_OPTION(sbi).s_jquota_fmt = QFMT_VFS_V1;
835			break;
836		case Opt_noquota:
837			clear_opt(sbi, QUOTA);
838			clear_opt(sbi, USRQUOTA);
839			clear_opt(sbi, GRPQUOTA);
840			clear_opt(sbi, PRJQUOTA);
841			break;
842#else
843		case Opt_quota:
844		case Opt_usrquota:
845		case Opt_grpquota:
846		case Opt_prjquota:
847		case Opt_usrjquota:
848		case Opt_grpjquota:
849		case Opt_prjjquota:
850		case Opt_offusrjquota:
851		case Opt_offgrpjquota:
852		case Opt_offprjjquota:
853		case Opt_jqfmt_vfsold:
854		case Opt_jqfmt_vfsv0:
855		case Opt_jqfmt_vfsv1:
856		case Opt_noquota:
857			f2fs_info(sbi, "quota operations not supported");
858			break;
859#endif
860		case Opt_whint:
861			name = match_strdup(&args[0]);
862			if (!name)
863				return -ENOMEM;
864			if (!strcmp(name, "user-based")) {
865				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_USER;
866			} else if (!strcmp(name, "off")) {
867				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
868			} else if (!strcmp(name, "fs-based")) {
869				F2FS_OPTION(sbi).whint_mode = WHINT_MODE_FS;
870			} else {
871				kfree(name);
872				return -EINVAL;
873			}
874			kfree(name);
875			break;
876		case Opt_alloc:
877			name = match_strdup(&args[0]);
878			if (!name)
879				return -ENOMEM;
880
881			if (!strcmp(name, "default")) {
882				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
883			} else if (!strcmp(name, "reuse")) {
884				F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
885			} else {
886				kfree(name);
887				return -EINVAL;
888			}
889			kfree(name);
890			break;
891		case Opt_fsync:
892			name = match_strdup(&args[0]);
893			if (!name)
894				return -ENOMEM;
895			if (!strcmp(name, "posix")) {
896				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
897			} else if (!strcmp(name, "strict")) {
898				F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_STRICT;
899			} else if (!strcmp(name, "nobarrier")) {
900				F2FS_OPTION(sbi).fsync_mode =
901							FSYNC_MODE_NOBARRIER;
902			} else {
903				kfree(name);
904				return -EINVAL;
905			}
906			kfree(name);
907			break;
908		case Opt_test_dummy_encryption:
909			ret = f2fs_set_test_dummy_encryption(sb, p, &args[0],
910							     is_remount);
911			if (ret)
912				return ret;
913			break;
914		case Opt_inlinecrypt:
915#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
916			sb->s_flags |= SB_INLINECRYPT;
917#else
918			f2fs_info(sbi, "inline encryption not supported");
919#endif
920			break;
921		case Opt_checkpoint_disable_cap_perc:
922			if (args->from && match_int(args, &arg))
923				return -EINVAL;
924			if (arg < 0 || arg > 100)
925				return -EINVAL;
926			F2FS_OPTION(sbi).unusable_cap_perc = arg;
927			set_opt(sbi, DISABLE_CHECKPOINT);
928			break;
929		case Opt_checkpoint_disable_cap:
930			if (args->from && match_int(args, &arg))
931				return -EINVAL;
932			F2FS_OPTION(sbi).unusable_cap = arg;
933			set_opt(sbi, DISABLE_CHECKPOINT);
934			break;
935		case Opt_checkpoint_disable:
936			set_opt(sbi, DISABLE_CHECKPOINT);
937			break;
938		case Opt_checkpoint_enable:
939			clear_opt(sbi, DISABLE_CHECKPOINT);
940			break;
941#ifdef CONFIG_F2FS_FS_COMPRESSION
942		case Opt_compress_algorithm:
943			if (!f2fs_sb_has_compression(sbi)) {
944				f2fs_info(sbi, "Image doesn't support compression");
945				break;
946			}
947			name = match_strdup(&args[0]);
948			if (!name)
949				return -ENOMEM;
950			if (!strcmp(name, "lzo")) {
951				F2FS_OPTION(sbi).compress_algorithm =
952								COMPRESS_LZO;
953			} else if (!strcmp(name, "lz4")) {
954				F2FS_OPTION(sbi).compress_algorithm =
955								COMPRESS_LZ4;
956			} else if (!strcmp(name, "zstd")) {
957				F2FS_OPTION(sbi).compress_algorithm =
958								COMPRESS_ZSTD;
959			} else if (!strcmp(name, "lzo-rle")) {
960				F2FS_OPTION(sbi).compress_algorithm =
961								COMPRESS_LZORLE;
962			} else {
963				kfree(name);
964				return -EINVAL;
965			}
966			kfree(name);
967			break;
968		case Opt_compress_log_size:
969			if (!f2fs_sb_has_compression(sbi)) {
970				f2fs_info(sbi, "Image doesn't support compression");
971				break;
972			}
973			if (args->from && match_int(args, &arg))
974				return -EINVAL;
975			if (arg < MIN_COMPRESS_LOG_SIZE ||
976				arg > MAX_COMPRESS_LOG_SIZE) {
977				f2fs_err(sbi,
978					"Compress cluster log size is out of range");
979				return -EINVAL;
980			}
981			F2FS_OPTION(sbi).compress_log_size = arg;
982			break;
983		case Opt_compress_extension:
984			if (!f2fs_sb_has_compression(sbi)) {
985				f2fs_info(sbi, "Image doesn't support compression");
986				break;
987			}
988			name = match_strdup(&args[0]);
989			if (!name)
990				return -ENOMEM;
991
992			ext = F2FS_OPTION(sbi).extensions;
993			ext_cnt = F2FS_OPTION(sbi).compress_ext_cnt;
994
995			if (strlen(name) >= F2FS_EXTENSION_LEN ||
996				ext_cnt >= COMPRESS_EXT_NUM) {
997				f2fs_err(sbi,
998					"invalid extension length/number");
999				kfree(name);
1000				return -EINVAL;
1001			}
1002
1003			strcpy(ext[ext_cnt], name);
1004			F2FS_OPTION(sbi).compress_ext_cnt++;
1005			kfree(name);
1006			break;
1007#else
1008		case Opt_compress_algorithm:
1009		case Opt_compress_log_size:
1010		case Opt_compress_extension:
1011			f2fs_info(sbi, "compression options not supported");
1012			break;
1013#endif
1014		case Opt_atgc:
1015			set_opt(sbi, ATGC);
1016			break;
1017		case Opt_gc_merge:
1018			set_opt(sbi, GC_MERGE);
1019			break;
1020		case Opt_nogc_merge:
1021			clear_opt(sbi, GC_MERGE);
1022			break;
1023		default:
1024			f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
1025				 p);
1026			return -EINVAL;
1027		}
1028	}
1029#ifdef CONFIG_QUOTA
1030	if (f2fs_check_quota_options(sbi))
1031		return -EINVAL;
1032#else
1033	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sbi->sb)) {
1034		f2fs_info(sbi, "Filesystem with quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1035		return -EINVAL;
1036	}
1037	if (f2fs_sb_has_project_quota(sbi) && !f2fs_readonly(sbi->sb)) {
1038		f2fs_err(sbi, "Filesystem with project quota feature cannot be mounted RDWR without CONFIG_QUOTA");
1039		return -EINVAL;
1040	}
1041#endif
1042#ifndef CONFIG_UNICODE
1043	if (f2fs_sb_has_casefold(sbi)) {
1044		f2fs_err(sbi,
1045			"Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
1046		return -EINVAL;
1047	}
1048#endif
1049	/*
1050	 * The BLKZONED feature indicates that the drive was formatted with
1051	 * zone alignment optimization. This is optional for host-aware
1052	 * devices, but mandatory for host-managed zoned block devices.
1053	 */
1054#ifndef CONFIG_BLK_DEV_ZONED
1055	if (f2fs_sb_has_blkzoned(sbi)) {
1056		f2fs_err(sbi, "Zoned block device support is not enabled");
1057		return -EINVAL;
1058	}
1059#endif
1060
1061	if (F2FS_IO_SIZE_BITS(sbi) && !f2fs_lfs_mode(sbi)) {
1062		f2fs_err(sbi, "Should set mode=lfs with %uKB-sized IO",
1063			 F2FS_IO_SIZE_KB(sbi));
1064		return -EINVAL;
1065	}
1066
1067	if (test_opt(sbi, INLINE_XATTR_SIZE)) {
1068		int min_size, max_size;
1069
1070		if (!f2fs_sb_has_extra_attr(sbi) ||
1071			!f2fs_sb_has_flexible_inline_xattr(sbi)) {
1072			f2fs_err(sbi, "extra_attr or flexible_inline_xattr feature is off");
1073			return -EINVAL;
1074		}
1075		if (!test_opt(sbi, INLINE_XATTR)) {
1076			f2fs_err(sbi, "inline_xattr_size option should be set with inline_xattr option");
1077			return -EINVAL;
1078		}
1079
1080		min_size = sizeof(struct f2fs_xattr_header) / sizeof(__le32);
1081		max_size = MAX_INLINE_XATTR_SIZE;
1082
1083		if (F2FS_OPTION(sbi).inline_xattr_size < min_size ||
1084				F2FS_OPTION(sbi).inline_xattr_size > max_size) {
1085			f2fs_err(sbi, "inline xattr size is out of range: %d ~ %d",
1086				 min_size, max_size);
1087			return -EINVAL;
1088		}
1089	}
1090
1091	if (test_opt(sbi, DISABLE_CHECKPOINT) && f2fs_lfs_mode(sbi)) {
1092		f2fs_err(sbi, "LFS not compatible with checkpoint=disable\n");
1093		return -EINVAL;
1094	}
1095
1096	/* Not pass down write hints if the number of active logs is lesser
1097	 * than NR_CURSEG_PERSIST_TYPE.
1098	 */
1099	if (F2FS_OPTION(sbi).active_logs != NR_CURSEG_PERSIST_TYPE)
1100		F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1101	return 0;
1102}
1103
1104static struct inode *f2fs_alloc_inode(struct super_block *sb)
1105{
1106	struct f2fs_inode_info *fi;
1107
1108	fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO);
1109	if (!fi)
1110		return NULL;
1111
1112	init_once((void *) fi);
1113
1114	/* Initialize f2fs-specific inode info */
1115	atomic_set(&fi->dirty_pages, 0);
1116	atomic_set(&fi->i_compr_blocks, 0);
1117	init_rwsem(&fi->i_sem);
1118	spin_lock_init(&fi->i_size_lock);
1119	INIT_LIST_HEAD(&fi->dirty_list);
1120	INIT_LIST_HEAD(&fi->gdirty_list);
1121	INIT_LIST_HEAD(&fi->inmem_ilist);
1122	INIT_LIST_HEAD(&fi->inmem_pages);
1123	mutex_init(&fi->inmem_lock);
1124	init_rwsem(&fi->i_gc_rwsem[READ]);
1125	init_rwsem(&fi->i_gc_rwsem[WRITE]);
1126	init_rwsem(&fi->i_mmap_sem);
1127	init_rwsem(&fi->i_xattr_sem);
1128
1129	/* Will be used by directory only */
1130	fi->i_dir_level = F2FS_SB(sb)->dir_level;
1131
1132	fi->ra_offset = -1;
1133
1134	return &fi->vfs_inode;
1135}
1136
1137static int f2fs_drop_inode(struct inode *inode)
1138{
1139	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1140	int ret;
1141
1142	/*
1143	 * during filesystem shutdown, if checkpoint is disabled,
1144	 * drop useless meta/node dirty pages.
1145	 */
1146	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED))) {
1147		if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1148			inode->i_ino == F2FS_META_INO(sbi)) {
1149			trace_f2fs_drop_inode(inode, 1);
1150			return 1;
1151		}
1152	}
1153
1154	/*
1155	 * This is to avoid a deadlock condition like below.
1156	 * writeback_single_inode(inode)
1157	 *  - f2fs_write_data_page
1158	 *    - f2fs_gc -> iput -> evict
1159	 *       - inode_wait_for_writeback(inode)
1160	 */
1161	if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) {
1162		if (!inode->i_nlink && !is_bad_inode(inode)) {
1163			/* to avoid evict_inode call simultaneously */
1164			atomic_inc(&inode->i_count);
1165			spin_unlock(&inode->i_lock);
1166
1167			/* some remained atomic pages should discarded */
1168			if (f2fs_is_atomic_file(inode))
1169				f2fs_drop_inmem_pages(inode);
1170
1171			/* should remain fi->extent_tree for writepage */
1172			f2fs_destroy_extent_node(inode);
1173
1174			sb_start_intwrite(inode->i_sb);
1175			f2fs_i_size_write(inode, 0);
1176
1177			f2fs_submit_merged_write_cond(F2FS_I_SB(inode),
1178					inode, NULL, 0, DATA);
1179			truncate_inode_pages_final(inode->i_mapping);
1180
1181			if (F2FS_HAS_BLOCKS(inode))
1182				f2fs_truncate(inode);
1183
1184			sb_end_intwrite(inode->i_sb);
1185
1186			spin_lock(&inode->i_lock);
1187			atomic_dec(&inode->i_count);
1188		}
1189		trace_f2fs_drop_inode(inode, 0);
1190		return 0;
1191	}
1192	ret = generic_drop_inode(inode);
1193	if (!ret)
1194		ret = fscrypt_drop_inode(inode);
1195	trace_f2fs_drop_inode(inode, ret);
1196	return ret;
1197}
1198
1199int f2fs_inode_dirtied(struct inode *inode, bool sync)
1200{
1201	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1202	int ret = 0;
1203
1204	spin_lock(&sbi->inode_lock[DIRTY_META]);
1205	if (is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1206		ret = 1;
1207	} else {
1208		set_inode_flag(inode, FI_DIRTY_INODE);
1209		stat_inc_dirty_inode(sbi, DIRTY_META);
1210	}
1211	if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) {
1212		list_add_tail(&F2FS_I(inode)->gdirty_list,
1213				&sbi->inode_list[DIRTY_META]);
1214		inc_page_count(sbi, F2FS_DIRTY_IMETA);
1215	}
1216	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1217	return ret;
1218}
1219
1220void f2fs_inode_synced(struct inode *inode)
1221{
1222	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1223
1224	spin_lock(&sbi->inode_lock[DIRTY_META]);
1225	if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) {
1226		spin_unlock(&sbi->inode_lock[DIRTY_META]);
1227		return;
1228	}
1229	if (!list_empty(&F2FS_I(inode)->gdirty_list)) {
1230		list_del_init(&F2FS_I(inode)->gdirty_list);
1231		dec_page_count(sbi, F2FS_DIRTY_IMETA);
1232	}
1233	clear_inode_flag(inode, FI_DIRTY_INODE);
1234	clear_inode_flag(inode, FI_AUTO_RECOVER);
1235	stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META);
1236	spin_unlock(&sbi->inode_lock[DIRTY_META]);
1237}
1238
1239/*
1240 * f2fs_dirty_inode() is called from __mark_inode_dirty()
1241 *
1242 * We should call set_dirty_inode to write the dirty inode through write_inode.
1243 */
1244static void f2fs_dirty_inode(struct inode *inode, int flags)
1245{
1246	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
1247
1248	if (inode->i_ino == F2FS_NODE_INO(sbi) ||
1249			inode->i_ino == F2FS_META_INO(sbi))
1250		return;
1251
1252	if (flags == I_DIRTY_TIME)
1253		return;
1254
1255	if (is_inode_flag_set(inode, FI_AUTO_RECOVER))
1256		clear_inode_flag(inode, FI_AUTO_RECOVER);
1257
1258	f2fs_inode_dirtied(inode, false);
1259}
1260
1261static void f2fs_free_inode(struct inode *inode)
1262{
1263	fscrypt_free_inode(inode);
1264	kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode));
1265}
1266
1267static void destroy_percpu_info(struct f2fs_sb_info *sbi)
1268{
1269	percpu_counter_destroy(&sbi->alloc_valid_block_count);
1270	percpu_counter_destroy(&sbi->total_valid_inode_count);
1271}
1272
1273static void destroy_device_list(struct f2fs_sb_info *sbi)
1274{
1275	int i;
1276
1277	for (i = 0; i < sbi->s_ndevs; i++) {
1278		blkdev_put(FDEV(i).bdev, FMODE_EXCL);
1279#ifdef CONFIG_BLK_DEV_ZONED
1280		kvfree(FDEV(i).blkz_seq);
1281#endif
1282	}
1283	kvfree(sbi->devs);
1284}
1285
1286static void f2fs_put_super(struct super_block *sb)
1287{
1288	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1289	int i;
1290	bool dropped;
1291
1292	/* unregister procfs/sysfs entries in advance to avoid race case */
1293	f2fs_unregister_sysfs(sbi);
1294
1295	f2fs_quota_off_umount(sb);
1296
1297	/* prevent remaining shrinker jobs */
1298	mutex_lock(&sbi->umount_mutex);
1299
1300	/*
1301	 * We don't need to do checkpoint when superblock is clean.
1302	 * But, the previous checkpoint was not done by umount, it needs to do
1303	 * clean checkpoint again.
1304	 */
1305	if ((is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
1306			!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG))) {
1307		struct cp_control cpc = {
1308			.reason = CP_UMOUNT,
1309		};
1310		f2fs_write_checkpoint(sbi, &cpc);
1311	}
1312
1313	/* be sure to wait for any on-going discard commands */
1314	dropped = f2fs_issue_discard_timeout(sbi);
1315
1316	if ((f2fs_hw_support_discard(sbi) || f2fs_hw_should_discard(sbi)) &&
1317					!sbi->discard_blks && !dropped) {
1318		struct cp_control cpc = {
1319			.reason = CP_UMOUNT | CP_TRIMMED,
1320		};
1321		f2fs_write_checkpoint(sbi, &cpc);
1322	}
1323
1324	/*
1325	 * normally superblock is clean, so we need to release this.
1326	 * In addition, EIO will skip do checkpoint, we need this as well.
1327	 */
1328	f2fs_release_ino_entry(sbi, true);
1329
1330	f2fs_leave_shrinker(sbi);
1331	mutex_unlock(&sbi->umount_mutex);
1332
1333	/* our cp_error case, we can wait for any writeback page */
1334	f2fs_flush_merged_writes(sbi);
1335
1336	f2fs_wait_on_all_pages(sbi, F2FS_WB_CP_DATA);
1337
1338	f2fs_bug_on(sbi, sbi->fsync_node_num);
1339
1340	iput(sbi->node_inode);
1341	sbi->node_inode = NULL;
1342
1343	iput(sbi->meta_inode);
1344	sbi->meta_inode = NULL;
1345
1346	/*
1347	 * iput() can update stat information, if f2fs_write_checkpoint()
1348	 * above failed with error.
1349	 */
1350	f2fs_destroy_stats(sbi);
1351
1352	/* destroy f2fs internal modules */
1353	f2fs_destroy_node_manager(sbi);
1354	f2fs_destroy_segment_manager(sbi);
1355
1356	f2fs_destroy_post_read_wq(sbi);
1357
1358	kvfree(sbi->ckpt);
1359
1360	sb->s_fs_info = NULL;
1361	if (sbi->s_chksum_driver)
1362		crypto_free_shash(sbi->s_chksum_driver);
1363	kfree(sbi->raw_super);
1364
1365	destroy_device_list(sbi);
1366	f2fs_destroy_page_array_cache(sbi);
1367	f2fs_destroy_xattr_caches(sbi);
1368	mempool_destroy(sbi->write_io_dummy);
1369#ifdef CONFIG_QUOTA
1370	for (i = 0; i < MAXQUOTAS; i++)
1371		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
1372#endif
1373	fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
1374	destroy_percpu_info(sbi);
1375	for (i = 0; i < NR_PAGE_TYPE; i++)
1376		kvfree(sbi->write_io[i]);
1377#ifdef CONFIG_UNICODE
1378	utf8_unload(sb->s_encoding);
1379#endif
1380	kfree(sbi);
1381}
1382
1383int f2fs_sync_fs(struct super_block *sb, int sync)
1384{
1385	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1386	int err = 0;
1387
1388	if (unlikely(f2fs_cp_error(sbi)))
1389		return 0;
1390	if (unlikely(is_sbi_flag_set(sbi, SBI_CP_DISABLED)))
1391		return 0;
1392
1393	trace_f2fs_sync_fs(sb, sync);
1394
1395	if (unlikely(is_sbi_flag_set(sbi, SBI_POR_DOING)))
1396		return -EAGAIN;
1397
1398	if (sync) {
1399		struct cp_control cpc;
1400
1401		cpc.reason = __get_cp_reason(sbi);
1402
1403		down_write(&sbi->gc_lock);
1404		err = f2fs_write_checkpoint(sbi, &cpc);
1405		up_write(&sbi->gc_lock);
1406	}
1407	f2fs_trace_ios(NULL, 1);
1408
1409	return err;
1410}
1411
1412static int f2fs_freeze(struct super_block *sb)
1413{
1414	if (f2fs_readonly(sb))
1415		return 0;
1416
1417	/* IO error happened before */
1418	if (unlikely(f2fs_cp_error(F2FS_SB(sb))))
1419		return -EIO;
1420
1421	/* must be clean, since sync_filesystem() was already called */
1422	if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY))
1423		return -EINVAL;
1424	return 0;
1425}
1426
1427static int f2fs_unfreeze(struct super_block *sb)
1428{
1429	return 0;
1430}
1431
1432#ifdef CONFIG_QUOTA
1433static int f2fs_statfs_project(struct super_block *sb,
1434				kprojid_t projid, struct kstatfs *buf)
1435{
1436	struct kqid qid;
1437	struct dquot *dquot;
1438	u64 limit;
1439	u64 curblock;
1440
1441	qid = make_kqid_projid(projid);
1442	dquot = dqget(sb, qid);
1443	if (IS_ERR(dquot))
1444		return PTR_ERR(dquot);
1445	spin_lock(&dquot->dq_dqb_lock);
1446
1447	limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
1448					dquot->dq_dqb.dqb_bhardlimit);
1449	if (limit)
1450		limit >>= sb->s_blocksize_bits;
1451
1452	if (limit && buf->f_blocks > limit) {
1453		curblock = (dquot->dq_dqb.dqb_curspace +
1454			    dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
1455		buf->f_blocks = limit;
1456		buf->f_bfree = buf->f_bavail =
1457			(buf->f_blocks > curblock) ?
1458			 (buf->f_blocks - curblock) : 0;
1459	}
1460
1461	limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
1462					dquot->dq_dqb.dqb_ihardlimit);
1463
1464	if (limit && buf->f_files > limit) {
1465		buf->f_files = limit;
1466		buf->f_ffree =
1467			(buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
1468			 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
1469	}
1470
1471	spin_unlock(&dquot->dq_dqb_lock);
1472	dqput(dquot);
1473	return 0;
1474}
1475#endif
1476
1477static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf)
1478{
1479	struct super_block *sb = dentry->d_sb;
1480	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1481	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
1482	block_t total_count, user_block_count, start_count;
1483	u64 avail_node_count;
1484
1485	total_count = le64_to_cpu(sbi->raw_super->block_count);
1486	user_block_count = sbi->user_block_count;
1487	start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr);
1488	buf->f_type = F2FS_SUPER_MAGIC;
1489	buf->f_bsize = sbi->blocksize;
1490
1491	buf->f_blocks = total_count - start_count;
1492	buf->f_bfree = user_block_count - valid_user_blocks(sbi) -
1493						sbi->current_reserved_blocks;
1494
1495	spin_lock(&sbi->stat_lock);
1496	if (unlikely(buf->f_bfree <= sbi->unusable_block_count))
1497		buf->f_bfree = 0;
1498	else
1499		buf->f_bfree -= sbi->unusable_block_count;
1500	spin_unlock(&sbi->stat_lock);
1501
1502	if (buf->f_bfree > F2FS_OPTION(sbi).root_reserved_blocks)
1503		buf->f_bavail = buf->f_bfree -
1504				F2FS_OPTION(sbi).root_reserved_blocks;
1505	else
1506		buf->f_bavail = 0;
1507
1508	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
1509
1510	if (avail_node_count > user_block_count) {
1511		buf->f_files = user_block_count;
1512		buf->f_ffree = buf->f_bavail;
1513	} else {
1514		buf->f_files = avail_node_count;
1515		buf->f_ffree = min(avail_node_count - valid_node_count(sbi),
1516					buf->f_bavail);
1517	}
1518
1519	buf->f_namelen = F2FS_NAME_LEN;
1520	buf->f_fsid    = u64_to_fsid(id);
1521
1522#ifdef CONFIG_QUOTA
1523	if (is_inode_flag_set(dentry->d_inode, FI_PROJ_INHERIT) &&
1524			sb_has_quota_limits_enabled(sb, PRJQUOTA)) {
1525		f2fs_statfs_project(sb, F2FS_I(dentry->d_inode)->i_projid, buf);
1526	}
1527#endif
1528	return 0;
1529}
1530
1531static inline void f2fs_show_quota_options(struct seq_file *seq,
1532					   struct super_block *sb)
1533{
1534#ifdef CONFIG_QUOTA
1535	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1536
1537	if (F2FS_OPTION(sbi).s_jquota_fmt) {
1538		char *fmtname = "";
1539
1540		switch (F2FS_OPTION(sbi).s_jquota_fmt) {
1541		case QFMT_VFS_OLD:
1542			fmtname = "vfsold";
1543			break;
1544		case QFMT_VFS_V0:
1545			fmtname = "vfsv0";
1546			break;
1547		case QFMT_VFS_V1:
1548			fmtname = "vfsv1";
1549			break;
1550		}
1551		seq_printf(seq, ",jqfmt=%s", fmtname);
1552	}
1553
1554	if (F2FS_OPTION(sbi).s_qf_names[USRQUOTA])
1555		seq_show_option(seq, "usrjquota",
1556			F2FS_OPTION(sbi).s_qf_names[USRQUOTA]);
1557
1558	if (F2FS_OPTION(sbi).s_qf_names[GRPQUOTA])
1559		seq_show_option(seq, "grpjquota",
1560			F2FS_OPTION(sbi).s_qf_names[GRPQUOTA]);
1561
1562	if (F2FS_OPTION(sbi).s_qf_names[PRJQUOTA])
1563		seq_show_option(seq, "prjjquota",
1564			F2FS_OPTION(sbi).s_qf_names[PRJQUOTA]);
1565#endif
1566}
1567
1568static inline void f2fs_show_compress_options(struct seq_file *seq,
1569							struct super_block *sb)
1570{
1571	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1572	char *algtype = "";
1573	int i;
1574
1575	if (!f2fs_sb_has_compression(sbi))
1576		return;
1577
1578	switch (F2FS_OPTION(sbi).compress_algorithm) {
1579	case COMPRESS_LZO:
1580		algtype = "lzo";
1581		break;
1582	case COMPRESS_LZ4:
1583		algtype = "lz4";
1584		break;
1585	case COMPRESS_ZSTD:
1586		algtype = "zstd";
1587		break;
1588	case COMPRESS_LZORLE:
1589		algtype = "lzo-rle";
1590		break;
1591	}
1592	seq_printf(seq, ",compress_algorithm=%s", algtype);
1593
1594	seq_printf(seq, ",compress_log_size=%u",
1595			F2FS_OPTION(sbi).compress_log_size);
1596
1597	for (i = 0; i < F2FS_OPTION(sbi).compress_ext_cnt; i++) {
1598		seq_printf(seq, ",compress_extension=%s",
1599			F2FS_OPTION(sbi).extensions[i]);
1600	}
1601}
1602
1603static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
1604{
1605	struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb);
1606
1607	if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_SYNC)
1608		seq_printf(seq, ",background_gc=%s", "sync");
1609	else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_ON)
1610		seq_printf(seq, ",background_gc=%s", "on");
1611	else if (F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF)
1612		seq_printf(seq, ",background_gc=%s", "off");
1613
1614	if (test_opt(sbi, GC_MERGE))
1615		seq_puts(seq, ",gc_merge");
1616
1617	if (test_opt(sbi, DISABLE_ROLL_FORWARD))
1618		seq_puts(seq, ",disable_roll_forward");
1619	if (test_opt(sbi, NORECOVERY))
1620		seq_puts(seq, ",norecovery");
1621	if (test_opt(sbi, DISCARD))
1622		seq_puts(seq, ",discard");
1623	else
1624		seq_puts(seq, ",nodiscard");
1625	if (test_opt(sbi, NOHEAP))
1626		seq_puts(seq, ",no_heap");
1627	else
1628		seq_puts(seq, ",heap");
1629#ifdef CONFIG_F2FS_FS_XATTR
1630	if (test_opt(sbi, XATTR_USER))
1631		seq_puts(seq, ",user_xattr");
1632	else
1633		seq_puts(seq, ",nouser_xattr");
1634	if (test_opt(sbi, INLINE_XATTR))
1635		seq_puts(seq, ",inline_xattr");
1636	else
1637		seq_puts(seq, ",noinline_xattr");
1638	if (test_opt(sbi, INLINE_XATTR_SIZE))
1639		seq_printf(seq, ",inline_xattr_size=%u",
1640					F2FS_OPTION(sbi).inline_xattr_size);
1641#endif
1642#ifdef CONFIG_F2FS_FS_POSIX_ACL
1643	if (test_opt(sbi, POSIX_ACL))
1644		seq_puts(seq, ",acl");
1645	else
1646		seq_puts(seq, ",noacl");
1647#endif
1648	if (test_opt(sbi, DISABLE_EXT_IDENTIFY))
1649		seq_puts(seq, ",disable_ext_identify");
1650	if (test_opt(sbi, INLINE_DATA))
1651		seq_puts(seq, ",inline_data");
1652	else
1653		seq_puts(seq, ",noinline_data");
1654	if (test_opt(sbi, INLINE_DENTRY))
1655		seq_puts(seq, ",inline_dentry");
1656	else
1657		seq_puts(seq, ",noinline_dentry");
1658	if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE))
1659		seq_puts(seq, ",flush_merge");
1660	if (test_opt(sbi, NOBARRIER))
1661		seq_puts(seq, ",nobarrier");
1662	if (test_opt(sbi, FASTBOOT))
1663		seq_puts(seq, ",fastboot");
1664	if (test_opt(sbi, EXTENT_CACHE))
1665		seq_puts(seq, ",extent_cache");
1666	else
1667		seq_puts(seq, ",noextent_cache");
1668	if (test_opt(sbi, DATA_FLUSH))
1669		seq_puts(seq, ",data_flush");
1670
1671	seq_puts(seq, ",mode=");
1672	if (F2FS_OPTION(sbi).fs_mode == FS_MODE_ADAPTIVE)
1673		seq_puts(seq, "adaptive");
1674	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
1675		seq_puts(seq, "lfs");
1676	seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
1677	if (test_opt(sbi, RESERVE_ROOT))
1678		seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
1679				F2FS_OPTION(sbi).root_reserved_blocks,
1680				from_kuid_munged(&init_user_ns,
1681					F2FS_OPTION(sbi).s_resuid),
1682				from_kgid_munged(&init_user_ns,
1683					F2FS_OPTION(sbi).s_resgid));
1684	if (F2FS_IO_SIZE_BITS(sbi))
1685		seq_printf(seq, ",io_bits=%u",
1686				F2FS_OPTION(sbi).write_io_size_bits);
1687#ifdef CONFIG_F2FS_FAULT_INJECTION
1688	if (test_opt(sbi, FAULT_INJECTION)) {
1689		seq_printf(seq, ",fault_injection=%u",
1690				F2FS_OPTION(sbi).fault_info.inject_rate);
1691		seq_printf(seq, ",fault_type=%u",
1692				F2FS_OPTION(sbi).fault_info.inject_type);
1693	}
1694#endif
1695#ifdef CONFIG_QUOTA
1696	if (test_opt(sbi, QUOTA))
1697		seq_puts(seq, ",quota");
1698	if (test_opt(sbi, USRQUOTA))
1699		seq_puts(seq, ",usrquota");
1700	if (test_opt(sbi, GRPQUOTA))
1701		seq_puts(seq, ",grpquota");
1702	if (test_opt(sbi, PRJQUOTA))
1703		seq_puts(seq, ",prjquota");
1704#endif
1705	f2fs_show_quota_options(seq, sbi->sb);
1706	if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_USER)
1707		seq_printf(seq, ",whint_mode=%s", "user-based");
1708	else if (F2FS_OPTION(sbi).whint_mode == WHINT_MODE_FS)
1709		seq_printf(seq, ",whint_mode=%s", "fs-based");
1710
1711	fscrypt_show_test_dummy_encryption(seq, ',', sbi->sb);
1712
1713	if (sbi->sb->s_flags & SB_INLINECRYPT)
1714		seq_puts(seq, ",inlinecrypt");
1715
1716	if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_DEFAULT)
1717		seq_printf(seq, ",alloc_mode=%s", "default");
1718	else if (F2FS_OPTION(sbi).alloc_mode == ALLOC_MODE_REUSE)
1719		seq_printf(seq, ",alloc_mode=%s", "reuse");
1720
1721	if (test_opt(sbi, DISABLE_CHECKPOINT))
1722		seq_printf(seq, ",checkpoint=disable:%u",
1723				F2FS_OPTION(sbi).unusable_cap);
1724	if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_POSIX)
1725		seq_printf(seq, ",fsync_mode=%s", "posix");
1726	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_STRICT)
1727		seq_printf(seq, ",fsync_mode=%s", "strict");
1728	else if (F2FS_OPTION(sbi).fsync_mode == FSYNC_MODE_NOBARRIER)
1729		seq_printf(seq, ",fsync_mode=%s", "nobarrier");
1730
1731#ifdef CONFIG_F2FS_FS_COMPRESSION
1732	f2fs_show_compress_options(seq, sbi->sb);
1733#endif
1734
1735	if (test_opt(sbi, ATGC))
1736		seq_puts(seq, ",atgc");
1737	return 0;
1738}
1739
1740static void default_options(struct f2fs_sb_info *sbi)
1741{
1742	/* init some FS parameters */
1743	F2FS_OPTION(sbi).active_logs = NR_CURSEG_PERSIST_TYPE;
1744	F2FS_OPTION(sbi).inline_xattr_size = DEFAULT_INLINE_XATTR_ADDRS;
1745	F2FS_OPTION(sbi).whint_mode = WHINT_MODE_OFF;
1746	F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_DEFAULT;
1747	F2FS_OPTION(sbi).fsync_mode = FSYNC_MODE_POSIX;
1748	F2FS_OPTION(sbi).s_resuid = make_kuid(&init_user_ns, F2FS_DEF_RESUID);
1749	F2FS_OPTION(sbi).s_resgid = make_kgid(&init_user_ns, F2FS_DEF_RESGID);
1750	F2FS_OPTION(sbi).compress_algorithm = COMPRESS_LZ4;
1751	F2FS_OPTION(sbi).compress_log_size = MIN_COMPRESS_LOG_SIZE;
1752	F2FS_OPTION(sbi).compress_ext_cnt = 0;
1753	F2FS_OPTION(sbi).bggc_mode = BGGC_MODE_ON;
1754
1755	set_opt(sbi, INLINE_XATTR);
1756	set_opt(sbi, INLINE_DATA);
1757	set_opt(sbi, INLINE_DENTRY);
1758	set_opt(sbi, EXTENT_CACHE);
1759	set_opt(sbi, NOHEAP);
1760	clear_opt(sbi, DISABLE_CHECKPOINT);
1761	F2FS_OPTION(sbi).unusable_cap = 0;
1762	sbi->sb->s_flags |= SB_LAZYTIME;
1763	set_opt(sbi, FLUSH_MERGE);
1764	set_opt(sbi, DISCARD);
1765	if (f2fs_sb_has_blkzoned(sbi))
1766		F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
1767	else
1768		F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
1769
1770#ifdef CONFIG_F2FS_FS_XATTR
1771	set_opt(sbi, XATTR_USER);
1772#endif
1773#ifdef CONFIG_F2FS_FS_POSIX_ACL
1774	set_opt(sbi, POSIX_ACL);
1775#endif
1776
1777	f2fs_build_fault_attr(sbi, 0, 0);
1778}
1779
1780#ifdef CONFIG_QUOTA
1781static int f2fs_enable_quotas(struct super_block *sb);
1782#endif
1783
1784static int f2fs_disable_checkpoint(struct f2fs_sb_info *sbi)
1785{
1786	unsigned int s_flags = sbi->sb->s_flags;
1787	struct cp_control cpc;
1788	int err = 0;
1789	int ret;
1790	block_t unusable;
1791
1792	if (s_flags & SB_RDONLY) {
1793		f2fs_err(sbi, "checkpoint=disable on readonly fs");
1794		return -EINVAL;
1795	}
1796	sbi->sb->s_flags |= SB_ACTIVE;
1797
1798	f2fs_update_time(sbi, DISABLE_TIME);
1799
1800	while (!f2fs_time_over(sbi, DISABLE_TIME)) {
1801		down_write(&sbi->gc_lock);
1802		err = f2fs_gc(sbi, true, false, false, NULL_SEGNO);
1803		if (err == -ENODATA) {
1804			err = 0;
1805			break;
1806		}
1807		if (err && err != -EAGAIN)
1808			break;
1809	}
1810
1811	ret = sync_filesystem(sbi->sb);
1812	if (ret || err) {
1813		err = ret ? ret: err;
1814		goto restore_flag;
1815	}
1816
1817	unusable = f2fs_get_unusable_blocks(sbi);
1818	if (f2fs_disable_cp_again(sbi, unusable)) {
1819		err = -EAGAIN;
1820		goto restore_flag;
1821	}
1822
1823	down_write(&sbi->gc_lock);
1824	cpc.reason = CP_PAUSE;
1825	set_sbi_flag(sbi, SBI_CP_DISABLED);
1826	err = f2fs_write_checkpoint(sbi, &cpc);
1827	if (err)
1828		goto out_unlock;
1829
1830	spin_lock(&sbi->stat_lock);
1831	sbi->unusable_block_count = unusable;
1832	spin_unlock(&sbi->stat_lock);
1833
1834out_unlock:
1835	up_write(&sbi->gc_lock);
1836restore_flag:
1837	sbi->sb->s_flags = s_flags;	/* Restore SB_RDONLY status */
1838	return err;
1839}
1840
1841static void f2fs_enable_checkpoint(struct f2fs_sb_info *sbi)
1842{
1843	int retry = DEFAULT_RETRY_IO_COUNT;
1844
1845	/* we should flush all the data to keep data consistency */
1846	do {
1847		sync_inodes_sb(sbi->sb);
1848		cond_resched();
1849		congestion_wait(BLK_RW_ASYNC, DEFAULT_IO_TIMEOUT);
1850	} while (get_pages(sbi, F2FS_DIRTY_DATA) && retry--);
1851
1852	if (unlikely(retry < 0))
1853		f2fs_warn(sbi, "checkpoint=enable has some unwritten data.");
1854
1855	down_write(&sbi->gc_lock);
1856	f2fs_dirty_to_prefree(sbi);
1857
1858	clear_sbi_flag(sbi, SBI_CP_DISABLED);
1859	set_sbi_flag(sbi, SBI_IS_DIRTY);
1860	up_write(&sbi->gc_lock);
1861
1862	f2fs_sync_fs(sbi->sb, 1);
1863}
1864
1865static int f2fs_remount(struct super_block *sb, int *flags, char *data)
1866{
1867	struct f2fs_sb_info *sbi = F2FS_SB(sb);
1868	struct f2fs_mount_info org_mount_opt;
1869	unsigned long old_sb_flags;
1870	int err;
1871	bool need_restart_gc = false;
1872	bool need_stop_gc = false;
1873	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
1874	bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
1875	bool no_io_align = !F2FS_IO_ALIGNED(sbi);
1876	bool no_atgc = !test_opt(sbi, ATGC);
1877	bool checkpoint_changed;
1878#ifdef CONFIG_QUOTA
1879	int i, j;
1880#endif
1881
1882	/*
1883	 * Save the old mount options in case we
1884	 * need to restore them.
1885	 */
1886	org_mount_opt = sbi->mount_opt;
1887	old_sb_flags = sb->s_flags;
1888
1889#ifdef CONFIG_QUOTA
1890	org_mount_opt.s_jquota_fmt = F2FS_OPTION(sbi).s_jquota_fmt;
1891	for (i = 0; i < MAXQUOTAS; i++) {
1892		if (F2FS_OPTION(sbi).s_qf_names[i]) {
1893			org_mount_opt.s_qf_names[i] =
1894				kstrdup(F2FS_OPTION(sbi).s_qf_names[i],
1895				GFP_KERNEL);
1896			if (!org_mount_opt.s_qf_names[i]) {
1897				for (j = 0; j < i; j++)
1898					kfree(org_mount_opt.s_qf_names[j]);
1899				return -ENOMEM;
1900			}
1901		} else {
1902			org_mount_opt.s_qf_names[i] = NULL;
1903		}
1904	}
1905#endif
1906
1907	/* recover superblocks we couldn't write due to previous RO mount */
1908	if (!(*flags & SB_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) {
1909		err = f2fs_commit_super(sbi, false);
1910		f2fs_info(sbi, "Try to recover all the superblocks, ret: %d",
1911			  err);
1912		if (!err)
1913			clear_sbi_flag(sbi, SBI_NEED_SB_WRITE);
1914	}
1915
1916	default_options(sbi);
1917
1918	/* parse mount options */
1919	err = parse_options(sb, data, true);
1920	if (err)
1921		goto restore_opts;
1922	checkpoint_changed =
1923			disable_checkpoint != test_opt(sbi, DISABLE_CHECKPOINT);
1924
1925	/*
1926	 * Previous and new state of filesystem is RO,
1927	 * so skip checking GC and FLUSH_MERGE conditions.
1928	 */
1929	if (f2fs_readonly(sb) && (*flags & SB_RDONLY))
1930		goto skip;
1931
1932#ifdef CONFIG_QUOTA
1933	if (!f2fs_readonly(sb) && (*flags & SB_RDONLY)) {
1934		err = dquot_suspend(sb, -1);
1935		if (err < 0)
1936			goto restore_opts;
1937	} else if (f2fs_readonly(sb) && !(*flags & SB_RDONLY)) {
1938		/* dquot_resume needs RW */
1939		sb->s_flags &= ~SB_RDONLY;
1940		if (sb_any_quota_suspended(sb)) {
1941			dquot_resume(sb, -1);
1942		} else if (f2fs_sb_has_quota_ino(sbi)) {
1943			err = f2fs_enable_quotas(sb);
1944			if (err)
1945				goto restore_opts;
1946		}
1947	}
1948#endif
1949	/* disallow enable atgc dynamically */
1950	if (no_atgc == !!test_opt(sbi, ATGC)) {
1951		err = -EINVAL;
1952		f2fs_warn(sbi, "switch atgc option is not allowed");
1953		goto restore_opts;
1954	}
1955
1956	/* disallow enable/disable extent_cache dynamically */
1957	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
1958		err = -EINVAL;
1959		f2fs_warn(sbi, "switch extent_cache option is not allowed");
1960		goto restore_opts;
1961	}
1962
1963	if (no_io_align == !!F2FS_IO_ALIGNED(sbi)) {
1964		err = -EINVAL;
1965		f2fs_warn(sbi, "switch io_bits option is not allowed");
1966		goto restore_opts;
1967	}
1968
1969	if ((*flags & SB_RDONLY) && test_opt(sbi, DISABLE_CHECKPOINT)) {
1970		err = -EINVAL;
1971		f2fs_warn(sbi, "disabling checkpoint not compatible with read-only");
1972		goto restore_opts;
1973	}
1974
1975	/*
1976	 * We stop the GC thread if FS is mounted as RO
1977	 * or if background_gc = off is passed in mount
1978	 * option. Also sync the filesystem.
1979	 */
1980	if ((*flags & SB_RDONLY) ||
1981			(F2FS_OPTION(sbi).bggc_mode == BGGC_MODE_OFF &&
1982			!test_opt(sbi, GC_MERGE))) {
1983		if (sbi->gc_thread) {
1984			f2fs_stop_gc_thread(sbi);
1985			need_restart_gc = true;
1986		}
1987	} else if (!sbi->gc_thread) {
1988		err = f2fs_start_gc_thread(sbi);
1989		if (err)
1990			goto restore_opts;
1991		need_stop_gc = true;
1992	}
1993
1994	if (*flags & SB_RDONLY ||
1995		F2FS_OPTION(sbi).whint_mode != org_mount_opt.whint_mode) {
1996		writeback_inodes_sb(sb, WB_REASON_SYNC);
1997		sync_inodes_sb(sb);
1998
1999		set_sbi_flag(sbi, SBI_IS_DIRTY);
2000		set_sbi_flag(sbi, SBI_IS_CLOSE);
2001		f2fs_sync_fs(sb, 1);
2002		clear_sbi_flag(sbi, SBI_IS_CLOSE);
2003	}
2004
2005	if (checkpoint_changed) {
2006		if (test_opt(sbi, DISABLE_CHECKPOINT)) {
2007			err = f2fs_disable_checkpoint(sbi);
2008			if (err)
2009				goto restore_gc;
2010		} else {
2011			f2fs_enable_checkpoint(sbi);
2012		}
2013	}
2014
2015	/*
2016	 * We stop issue flush thread if FS is mounted as RO
2017	 * or if flush_merge is not passed in mount option.
2018	 */
2019	if ((*flags & SB_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) {
2020		clear_opt(sbi, FLUSH_MERGE);
2021		f2fs_destroy_flush_cmd_control(sbi, false);
2022	} else {
2023		err = f2fs_create_flush_cmd_control(sbi);
2024		if (err)
2025			goto restore_gc;
2026	}
2027skip:
2028#ifdef CONFIG_QUOTA
2029	/* Release old quota file names */
2030	for (i = 0; i < MAXQUOTAS; i++)
2031		kfree(org_mount_opt.s_qf_names[i]);
2032#endif
2033	/* Update the POSIXACL Flag */
2034	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
2035		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
2036
2037	limit_reserve_root(sbi);
2038	adjust_unusable_cap_perc(sbi);
2039	*flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
2040	return 0;
2041restore_gc:
2042	if (need_restart_gc) {
2043		if (f2fs_start_gc_thread(sbi))
2044			f2fs_warn(sbi, "background gc thread has stopped");
2045	} else if (need_stop_gc) {
2046		f2fs_stop_gc_thread(sbi);
2047	}
2048restore_opts:
2049#ifdef CONFIG_QUOTA
2050	F2FS_OPTION(sbi).s_jquota_fmt = org_mount_opt.s_jquota_fmt;
2051	for (i = 0; i < MAXQUOTAS; i++) {
2052		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
2053		F2FS_OPTION(sbi).s_qf_names[i] = org_mount_opt.s_qf_names[i];
2054	}
2055#endif
2056	sbi->mount_opt = org_mount_opt;
2057	sb->s_flags = old_sb_flags;
2058	return err;
2059}
2060
2061#ifdef CONFIG_QUOTA
2062/* Read data from quotafile */
2063static ssize_t f2fs_quota_read(struct super_block *sb, int type, char *data,
2064			       size_t len, loff_t off)
2065{
2066	struct inode *inode = sb_dqopt(sb)->files[type];
2067	struct address_space *mapping = inode->i_mapping;
2068	block_t blkidx = F2FS_BYTES_TO_BLK(off);
2069	int offset = off & (sb->s_blocksize - 1);
2070	int tocopy;
2071	size_t toread;
2072	loff_t i_size = i_size_read(inode);
2073	struct page *page;
2074
2075	if (off > i_size)
2076		return 0;
2077
2078	if (off + len > i_size)
2079		len = i_size - off;
2080	toread = len;
2081	while (toread > 0) {
2082		tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
2083repeat:
2084		page = read_cache_page_gfp(mapping, blkidx, GFP_NOFS);
2085		if (IS_ERR(page)) {
2086			if (PTR_ERR(page) == -ENOMEM) {
2087				congestion_wait(BLK_RW_ASYNC,
2088						DEFAULT_IO_TIMEOUT);
2089				goto repeat;
2090			}
2091			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2092			return PTR_ERR(page);
2093		}
2094
2095		lock_page(page);
2096
2097		if (unlikely(page->mapping != mapping)) {
2098			f2fs_put_page(page, 1);
2099			goto repeat;
2100		}
2101		if (unlikely(!PageUptodate(page))) {
2102			f2fs_put_page(page, 1);
2103			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2104			return -EIO;
2105		}
2106
2107		memcpy_from_page(data, page, offset, tocopy);
2108		f2fs_put_page(page, 1);
2109
2110		offset = 0;
2111		toread -= tocopy;
2112		data += tocopy;
2113		blkidx++;
2114	}
2115	return len;
2116}
2117
2118/* Write to quotafile */
2119static ssize_t f2fs_quota_write(struct super_block *sb, int type,
2120				const char *data, size_t len, loff_t off)
2121{
2122	struct inode *inode = sb_dqopt(sb)->files[type];
2123	struct address_space *mapping = inode->i_mapping;
2124	const struct address_space_operations *a_ops = mapping->a_ops;
2125	int offset = off & (sb->s_blocksize - 1);
2126	size_t towrite = len;
2127	struct page *page;
2128	void *fsdata = NULL;
2129	int err = 0;
2130	int tocopy;
2131
2132	while (towrite > 0) {
2133		tocopy = min_t(unsigned long, sb->s_blocksize - offset,
2134								towrite);
2135retry:
2136		err = a_ops->write_begin(NULL, mapping, off, tocopy, 0,
2137							&page, &fsdata);
2138		if (unlikely(err)) {
2139			if (err == -ENOMEM) {
2140				congestion_wait(BLK_RW_ASYNC,
2141						DEFAULT_IO_TIMEOUT);
2142				goto retry;
2143			}
2144			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2145			break;
2146		}
2147
2148		memcpy_to_page(page, offset, data, tocopy);
2149
2150		a_ops->write_end(NULL, mapping, off, tocopy, tocopy,
2151						page, fsdata);
2152		offset = 0;
2153		towrite -= tocopy;
2154		off += tocopy;
2155		data += tocopy;
2156		cond_resched();
2157	}
2158
2159	if (len == towrite)
2160		return err;
2161	inode->i_mtime = inode->i_ctime = current_time(inode);
2162	f2fs_mark_inode_dirty_sync(inode, false);
2163	return len - towrite;
2164}
2165
2166static struct dquot **f2fs_get_dquots(struct inode *inode)
2167{
2168	return F2FS_I(inode)->i_dquot;
2169}
2170
2171static qsize_t *f2fs_get_reserved_space(struct inode *inode)
2172{
2173	return &F2FS_I(inode)->i_reserved_quota;
2174}
2175
2176static int f2fs_quota_on_mount(struct f2fs_sb_info *sbi, int type)
2177{
2178	if (is_set_ckpt_flags(sbi, CP_QUOTA_NEED_FSCK_FLAG)) {
2179		f2fs_err(sbi, "quota sysfile may be corrupted, skip loading it");
2180		return 0;
2181	}
2182
2183	return dquot_quota_on_mount(sbi->sb, F2FS_OPTION(sbi).s_qf_names[type],
2184					F2FS_OPTION(sbi).s_jquota_fmt, type);
2185}
2186
2187int f2fs_enable_quota_files(struct f2fs_sb_info *sbi, bool rdonly)
2188{
2189	int enabled = 0;
2190	int i, err;
2191
2192	if (f2fs_sb_has_quota_ino(sbi) && rdonly) {
2193		err = f2fs_enable_quotas(sbi->sb);
2194		if (err) {
2195			f2fs_err(sbi, "Cannot turn on quota_ino: %d", err);
2196			return 0;
2197		}
2198		return 1;
2199	}
2200
2201	for (i = 0; i < MAXQUOTAS; i++) {
2202		if (F2FS_OPTION(sbi).s_qf_names[i]) {
2203			err = f2fs_quota_on_mount(sbi, i);
2204			if (!err) {
2205				enabled = 1;
2206				continue;
2207			}
2208			f2fs_err(sbi, "Cannot turn on quotas: %d on %d",
2209				 err, i);
2210		}
2211	}
2212	return enabled;
2213}
2214
2215static int f2fs_quota_enable(struct super_block *sb, int type, int format_id,
2216			     unsigned int flags)
2217{
2218	struct inode *qf_inode;
2219	unsigned long qf_inum;
2220	int err;
2221
2222	BUG_ON(!f2fs_sb_has_quota_ino(F2FS_SB(sb)));
2223
2224	qf_inum = f2fs_qf_ino(sb, type);
2225	if (!qf_inum)
2226		return -EPERM;
2227
2228	qf_inode = f2fs_iget(sb, qf_inum);
2229	if (IS_ERR(qf_inode)) {
2230		f2fs_err(F2FS_SB(sb), "Bad quota inode %u:%lu", type, qf_inum);
2231		return PTR_ERR(qf_inode);
2232	}
2233
2234	/* Don't account quota for quota files to avoid recursion */
2235	qf_inode->i_flags |= S_NOQUOTA;
2236	err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
2237	iput(qf_inode);
2238	return err;
2239}
2240
2241static int f2fs_enable_quotas(struct super_block *sb)
2242{
2243	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2244	int type, err = 0;
2245	unsigned long qf_inum;
2246	bool quota_mopt[MAXQUOTAS] = {
2247		test_opt(sbi, USRQUOTA),
2248		test_opt(sbi, GRPQUOTA),
2249		test_opt(sbi, PRJQUOTA),
2250	};
2251
2252	if (is_set_ckpt_flags(F2FS_SB(sb), CP_QUOTA_NEED_FSCK_FLAG)) {
2253		f2fs_err(sbi, "quota file may be corrupted, skip loading it");
2254		return 0;
2255	}
2256
2257	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
2258
2259	for (type = 0; type < MAXQUOTAS; type++) {
2260		qf_inum = f2fs_qf_ino(sb, type);
2261		if (qf_inum) {
2262			err = f2fs_quota_enable(sb, type, QFMT_VFS_V1,
2263				DQUOT_USAGE_ENABLED |
2264				(quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
2265			if (err) {
2266				f2fs_err(sbi, "Failed to enable quota tracking (type=%d, err=%d). Please run fsck to fix.",
2267					 type, err);
2268				for (type--; type >= 0; type--)
2269					dquot_quota_off(sb, type);
2270				set_sbi_flag(F2FS_SB(sb),
2271						SBI_QUOTA_NEED_REPAIR);
2272				return err;
2273			}
2274		}
2275	}
2276	return 0;
2277}
2278
2279static int f2fs_quota_sync_file(struct f2fs_sb_info *sbi, int type)
2280{
2281	struct quota_info *dqopt = sb_dqopt(sbi->sb);
2282	struct address_space *mapping = dqopt->files[type]->i_mapping;
2283	int ret = 0;
2284
2285	ret = dquot_writeback_dquots(sbi->sb, type);
2286	if (ret)
2287		goto out;
2288
2289	ret = filemap_fdatawrite(mapping);
2290	if (ret)
2291		goto out;
2292
2293	/* if we are using journalled quota */
2294	if (is_journalled_quota(sbi))
2295		goto out;
2296
2297	ret = filemap_fdatawait(mapping);
2298
2299	truncate_inode_pages(&dqopt->files[type]->i_data, 0);
2300out:
2301	if (ret)
2302		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2303	return ret;
2304}
2305
2306int f2fs_quota_sync(struct super_block *sb, int type)
2307{
2308	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2309	struct quota_info *dqopt = sb_dqopt(sb);
2310	int cnt;
2311	int ret = 0;
2312
2313	/*
2314	 * Now when everything is written we can discard the pagecache so
2315	 * that userspace sees the changes.
2316	 */
2317	for (cnt = 0; cnt < MAXQUOTAS; cnt++) {
2318
2319		if (type != -1 && cnt != type)
2320			continue;
2321
2322		if (!sb_has_quota_active(sb, cnt))
2323			continue;
2324
2325		if (!f2fs_sb_has_quota_ino(sbi))
2326			inode_lock(dqopt->files[cnt]);
2327
2328		/*
2329		 * do_quotactl
2330		 *  f2fs_quota_sync
2331		 *  down_read(quota_sem)
2332		 *  dquot_writeback_dquots()
2333		 *  f2fs_dquot_commit
2334		 *			      block_operation
2335		 *			      down_read(quota_sem)
2336		 */
2337		f2fs_lock_op(sbi);
2338		down_read(&sbi->quota_sem);
2339
2340		ret = f2fs_quota_sync_file(sbi, cnt);
2341
2342		up_read(&sbi->quota_sem);
2343		f2fs_unlock_op(sbi);
2344
2345		if (!f2fs_sb_has_quota_ino(sbi))
2346			inode_unlock(dqopt->files[cnt]);
2347
2348		if (ret)
2349			break;
2350	}
2351	return ret;
2352}
2353
2354static int f2fs_quota_on(struct super_block *sb, int type, int format_id,
2355							const struct path *path)
2356{
2357	struct inode *inode;
2358	int err;
2359
2360	/* if quota sysfile exists, deny enabling quota with specific file */
2361	if (f2fs_sb_has_quota_ino(F2FS_SB(sb))) {
2362		f2fs_err(F2FS_SB(sb), "quota sysfile already exists");
2363		return -EBUSY;
2364	}
2365
2366	err = f2fs_quota_sync(sb, type);
2367	if (err)
2368		return err;
2369
2370	err = dquot_quota_on(sb, type, format_id, path);
2371	if (err)
2372		return err;
2373
2374	inode = d_inode(path->dentry);
2375
2376	inode_lock(inode);
2377	F2FS_I(inode)->i_flags |= F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL;
2378	f2fs_set_inode_flags(inode);
2379	inode_unlock(inode);
2380	f2fs_mark_inode_dirty_sync(inode, false);
2381
2382	return 0;
2383}
2384
2385static int __f2fs_quota_off(struct super_block *sb, int type)
2386{
2387	struct inode *inode = sb_dqopt(sb)->files[type];
2388	int err;
2389
2390	if (!inode || !igrab(inode))
2391		return dquot_quota_off(sb, type);
2392
2393	err = f2fs_quota_sync(sb, type);
2394	if (err)
2395		goto out_put;
2396
2397	err = dquot_quota_off(sb, type);
2398	if (err || f2fs_sb_has_quota_ino(F2FS_SB(sb)))
2399		goto out_put;
2400
2401	inode_lock(inode);
2402	F2FS_I(inode)->i_flags &= ~(F2FS_NOATIME_FL | F2FS_IMMUTABLE_FL);
2403	f2fs_set_inode_flags(inode);
2404	inode_unlock(inode);
2405	f2fs_mark_inode_dirty_sync(inode, false);
2406out_put:
2407	iput(inode);
2408	return err;
2409}
2410
2411static int f2fs_quota_off(struct super_block *sb, int type)
2412{
2413	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2414	int err;
2415
2416	err = __f2fs_quota_off(sb, type);
2417
2418	/*
2419	 * quotactl can shutdown journalled quota, result in inconsistence
2420	 * between quota record and fs data by following updates, tag the
2421	 * flag to let fsck be aware of it.
2422	 */
2423	if (is_journalled_quota(sbi))
2424		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2425	return err;
2426}
2427
2428void f2fs_quota_off_umount(struct super_block *sb)
2429{
2430	int type;
2431	int err;
2432
2433	for (type = 0; type < MAXQUOTAS; type++) {
2434		err = __f2fs_quota_off(sb, type);
2435		if (err) {
2436			int ret = dquot_quota_off(sb, type);
2437
2438			f2fs_err(F2FS_SB(sb), "Fail to turn off disk quota (type: %d, err: %d, ret:%d), Please run fsck to fix it.",
2439				 type, err, ret);
2440			set_sbi_flag(F2FS_SB(sb), SBI_QUOTA_NEED_REPAIR);
2441		}
2442	}
2443	/*
2444	 * In case of checkpoint=disable, we must flush quota blocks.
2445	 * This can cause NULL exception for node_inode in end_io, since
2446	 * put_super already dropped it.
2447	 */
2448	sync_filesystem(sb);
2449}
2450
2451static void f2fs_truncate_quota_inode_pages(struct super_block *sb)
2452{
2453	struct quota_info *dqopt = sb_dqopt(sb);
2454	int type;
2455
2456	for (type = 0; type < MAXQUOTAS; type++) {
2457		if (!dqopt->files[type])
2458			continue;
2459		f2fs_inode_synced(dqopt->files[type]);
2460	}
2461}
2462
2463static int f2fs_dquot_commit(struct dquot *dquot)
2464{
2465	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2466	int ret;
2467
2468	down_read_nested(&sbi->quota_sem, SINGLE_DEPTH_NESTING);
2469	ret = dquot_commit(dquot);
2470	if (ret < 0)
2471		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2472	up_read(&sbi->quota_sem);
2473	return ret;
2474}
2475
2476static int f2fs_dquot_acquire(struct dquot *dquot)
2477{
2478	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2479	int ret;
2480
2481	down_read(&sbi->quota_sem);
2482	ret = dquot_acquire(dquot);
2483	if (ret < 0)
2484		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2485	up_read(&sbi->quota_sem);
2486	return ret;
2487}
2488
2489static int f2fs_dquot_release(struct dquot *dquot)
2490{
2491	struct f2fs_sb_info *sbi = F2FS_SB(dquot->dq_sb);
2492	int ret = dquot_release(dquot);
2493
2494	if (ret < 0)
2495		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2496	return ret;
2497}
2498
2499static int f2fs_dquot_mark_dquot_dirty(struct dquot *dquot)
2500{
2501	struct super_block *sb = dquot->dq_sb;
2502	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2503	int ret = dquot_mark_dquot_dirty(dquot);
2504
2505	/* if we are using journalled quota */
2506	if (is_journalled_quota(sbi))
2507		set_sbi_flag(sbi, SBI_QUOTA_NEED_FLUSH);
2508
2509	return ret;
2510}
2511
2512static int f2fs_dquot_commit_info(struct super_block *sb, int type)
2513{
2514	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2515	int ret = dquot_commit_info(sb, type);
2516
2517	if (ret < 0)
2518		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
2519	return ret;
2520}
2521
2522static int f2fs_get_projid(struct inode *inode, kprojid_t *projid)
2523{
2524	*projid = F2FS_I(inode)->i_projid;
2525	return 0;
2526}
2527
2528static const struct dquot_operations f2fs_quota_operations = {
2529	.get_reserved_space = f2fs_get_reserved_space,
2530	.write_dquot	= f2fs_dquot_commit,
2531	.acquire_dquot	= f2fs_dquot_acquire,
2532	.release_dquot	= f2fs_dquot_release,
2533	.mark_dirty	= f2fs_dquot_mark_dquot_dirty,
2534	.write_info	= f2fs_dquot_commit_info,
2535	.alloc_dquot	= dquot_alloc,
2536	.destroy_dquot	= dquot_destroy,
2537	.get_projid	= f2fs_get_projid,
2538	.get_next_id	= dquot_get_next_id,
2539};
2540
2541static const struct quotactl_ops f2fs_quotactl_ops = {
2542	.quota_on	= f2fs_quota_on,
2543	.quota_off	= f2fs_quota_off,
2544	.quota_sync	= f2fs_quota_sync,
2545	.get_state	= dquot_get_state,
2546	.set_info	= dquot_set_dqinfo,
2547	.get_dqblk	= dquot_get_dqblk,
2548	.set_dqblk	= dquot_set_dqblk,
2549	.get_nextdqblk	= dquot_get_next_dqblk,
2550};
2551#else
2552int f2fs_quota_sync(struct super_block *sb, int type)
2553{
2554	return 0;
2555}
2556
2557void f2fs_quota_off_umount(struct super_block *sb)
2558{
2559}
2560#endif
2561
2562static const struct super_operations f2fs_sops = {
2563	.alloc_inode	= f2fs_alloc_inode,
2564	.free_inode	= f2fs_free_inode,
2565	.drop_inode	= f2fs_drop_inode,
2566	.write_inode	= f2fs_write_inode,
2567	.dirty_inode	= f2fs_dirty_inode,
2568	.show_options	= f2fs_show_options,
2569#ifdef CONFIG_QUOTA
2570	.quota_read	= f2fs_quota_read,
2571	.quota_write	= f2fs_quota_write,
2572	.get_dquots	= f2fs_get_dquots,
2573#endif
2574	.evict_inode	= f2fs_evict_inode,
2575	.put_super	= f2fs_put_super,
2576	.sync_fs	= f2fs_sync_fs,
2577	.freeze_fs	= f2fs_freeze,
2578	.unfreeze_fs	= f2fs_unfreeze,
2579	.statfs		= f2fs_statfs,
2580	.remount_fs	= f2fs_remount,
2581};
2582
2583#ifdef CONFIG_FS_ENCRYPTION
2584static int f2fs_get_context(struct inode *inode, void *ctx, size_t len)
2585{
2586	return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2587				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2588				ctx, len, NULL);
2589}
2590
2591static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len,
2592							void *fs_data)
2593{
2594	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
2595
2596	/*
2597	 * Encrypting the root directory is not allowed because fsck
2598	 * expects lost+found directory to exist and remain unencrypted
2599	 * if LOST_FOUND feature is enabled.
2600	 *
2601	 */
2602	if (f2fs_sb_has_lost_found(sbi) &&
2603			inode->i_ino == F2FS_ROOT_INO(sbi))
2604		return -EPERM;
2605
2606	return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION,
2607				F2FS_XATTR_NAME_ENCRYPTION_CONTEXT,
2608				ctx, len, fs_data, XATTR_CREATE);
2609}
2610
2611static const union fscrypt_policy *f2fs_get_dummy_policy(struct super_block *sb)
2612{
2613	return F2FS_OPTION(F2FS_SB(sb)).dummy_enc_policy.policy;
2614}
2615
2616static bool f2fs_has_stable_inodes(struct super_block *sb)
2617{
2618	return true;
2619}
2620
2621static void f2fs_get_ino_and_lblk_bits(struct super_block *sb,
2622				       int *ino_bits_ret, int *lblk_bits_ret)
2623{
2624	*ino_bits_ret = 8 * sizeof(nid_t);
2625	*lblk_bits_ret = 8 * sizeof(block_t);
2626}
2627
2628static int f2fs_get_num_devices(struct super_block *sb)
2629{
2630	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2631
2632	if (f2fs_is_multi_device(sbi))
2633		return sbi->s_ndevs;
2634	return 1;
2635}
2636
2637static void f2fs_get_devices(struct super_block *sb,
2638			     struct request_queue **devs)
2639{
2640	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2641	int i;
2642
2643	for (i = 0; i < sbi->s_ndevs; i++)
2644		devs[i] = bdev_get_queue(FDEV(i).bdev);
2645}
2646
2647static const struct fscrypt_operations f2fs_cryptops = {
2648	.key_prefix		= "f2fs:",
2649	.get_context		= f2fs_get_context,
2650	.set_context		= f2fs_set_context,
2651	.get_dummy_policy	= f2fs_get_dummy_policy,
2652	.empty_dir		= f2fs_empty_dir,
2653	.max_namelen		= F2FS_NAME_LEN,
2654	.has_stable_inodes	= f2fs_has_stable_inodes,
2655	.get_ino_and_lblk_bits	= f2fs_get_ino_and_lblk_bits,
2656	.get_num_devices	= f2fs_get_num_devices,
2657	.get_devices		= f2fs_get_devices,
2658};
2659#endif
2660
2661static struct inode *f2fs_nfs_get_inode(struct super_block *sb,
2662		u64 ino, u32 generation)
2663{
2664	struct f2fs_sb_info *sbi = F2FS_SB(sb);
2665	struct inode *inode;
2666
2667	if (f2fs_check_nid_range(sbi, ino))
2668		return ERR_PTR(-ESTALE);
2669
2670	/*
2671	 * f2fs_iget isn't quite right if the inode is currently unallocated!
2672	 * However f2fs_iget currently does appropriate checks to handle stale
2673	 * inodes so everything is OK.
2674	 */
2675	inode = f2fs_iget(sb, ino);
2676	if (IS_ERR(inode))
2677		return ERR_CAST(inode);
2678	if (unlikely(generation && inode->i_generation != generation)) {
2679		/* we didn't find the right inode.. */
2680		iput(inode);
2681		return ERR_PTR(-ESTALE);
2682	}
2683	return inode;
2684}
2685
2686static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid,
2687		int fh_len, int fh_type)
2688{
2689	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
2690				    f2fs_nfs_get_inode);
2691}
2692
2693static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid,
2694		int fh_len, int fh_type)
2695{
2696	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
2697				    f2fs_nfs_get_inode);
2698}
2699
2700static const struct export_operations f2fs_export_ops = {
2701	.fh_to_dentry = f2fs_fh_to_dentry,
2702	.fh_to_parent = f2fs_fh_to_parent,
2703	.get_parent = f2fs_get_parent,
2704};
2705
2706static loff_t max_file_blocks(void)
2707{
2708	loff_t result = 0;
2709	loff_t leaf_count = DEF_ADDRS_PER_BLOCK;
2710
2711	/*
2712	 * note: previously, result is equal to (DEF_ADDRS_PER_INODE -
2713	 * DEFAULT_INLINE_XATTR_ADDRS), but now f2fs try to reserve more
2714	 * space in inode.i_addr, it will be more safe to reassign
2715	 * result as zero.
2716	 */
2717
2718	/* two direct node blocks */
2719	result += (leaf_count * 2);
2720
2721	/* two indirect node blocks */
2722	leaf_count *= NIDS_PER_BLOCK;
2723	result += (leaf_count * 2);
2724
2725	/* one double indirect node block */
2726	leaf_count *= NIDS_PER_BLOCK;
2727	result += leaf_count;
2728
2729	return result;
2730}
2731
2732static int __f2fs_commit_super(struct buffer_head *bh,
2733			struct f2fs_super_block *super)
2734{
2735	lock_buffer(bh);
2736	if (super)
2737		memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super));
2738	set_buffer_dirty(bh);
2739	unlock_buffer(bh);
2740
2741	/* it's rare case, we can do fua all the time */
2742	return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA);
2743}
2744
2745static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi,
2746					struct buffer_head *bh)
2747{
2748	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2749					(bh->b_data + F2FS_SUPER_OFFSET);
2750	struct super_block *sb = sbi->sb;
2751	u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr);
2752	u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr);
2753	u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr);
2754	u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr);
2755	u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr);
2756	u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr);
2757	u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt);
2758	u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit);
2759	u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat);
2760	u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa);
2761	u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2762	u32 segment_count = le32_to_cpu(raw_super->segment_count);
2763	u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
2764	u64 main_end_blkaddr = main_blkaddr +
2765				(segment_count_main << log_blocks_per_seg);
2766	u64 seg_end_blkaddr = segment0_blkaddr +
2767				(segment_count << log_blocks_per_seg);
2768
2769	if (segment0_blkaddr != cp_blkaddr) {
2770		f2fs_info(sbi, "Mismatch start address, segment0(%u) cp_blkaddr(%u)",
2771			  segment0_blkaddr, cp_blkaddr);
2772		return true;
2773	}
2774
2775	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
2776							sit_blkaddr) {
2777		f2fs_info(sbi, "Wrong CP boundary, start(%u) end(%u) blocks(%u)",
2778			  cp_blkaddr, sit_blkaddr,
2779			  segment_count_ckpt << log_blocks_per_seg);
2780		return true;
2781	}
2782
2783	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
2784							nat_blkaddr) {
2785		f2fs_info(sbi, "Wrong SIT boundary, start(%u) end(%u) blocks(%u)",
2786			  sit_blkaddr, nat_blkaddr,
2787			  segment_count_sit << log_blocks_per_seg);
2788		return true;
2789	}
2790
2791	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
2792							ssa_blkaddr) {
2793		f2fs_info(sbi, "Wrong NAT boundary, start(%u) end(%u) blocks(%u)",
2794			  nat_blkaddr, ssa_blkaddr,
2795			  segment_count_nat << log_blocks_per_seg);
2796		return true;
2797	}
2798
2799	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
2800							main_blkaddr) {
2801		f2fs_info(sbi, "Wrong SSA boundary, start(%u) end(%u) blocks(%u)",
2802			  ssa_blkaddr, main_blkaddr,
2803			  segment_count_ssa << log_blocks_per_seg);
2804		return true;
2805	}
2806
2807	if (main_end_blkaddr > seg_end_blkaddr) {
2808		f2fs_info(sbi, "Wrong MAIN_AREA boundary, start(%u) end(%llu) block(%u)",
2809			  main_blkaddr, seg_end_blkaddr,
2810			  segment_count_main << log_blocks_per_seg);
2811		return true;
2812	} else if (main_end_blkaddr < seg_end_blkaddr) {
2813		int err = 0;
2814		char *res;
2815
2816		/* fix in-memory information all the time */
2817		raw_super->segment_count = cpu_to_le32((main_end_blkaddr -
2818				segment0_blkaddr) >> log_blocks_per_seg);
2819
2820		if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) {
2821			set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
2822			res = "internally";
2823		} else {
2824			err = __f2fs_commit_super(bh, NULL);
2825			res = err ? "failed" : "done";
2826		}
2827		f2fs_info(sbi, "Fix alignment : %s, start(%u) end(%llu) block(%u)",
2828			  res, main_blkaddr, seg_end_blkaddr,
2829			  segment_count_main << log_blocks_per_seg);
2830		if (err)
2831			return true;
2832	}
2833	return false;
2834}
2835
2836static int sanity_check_raw_super(struct f2fs_sb_info *sbi,
2837				struct buffer_head *bh)
2838{
2839	block_t segment_count, segs_per_sec, secs_per_zone, segment_count_main;
2840	block_t total_sections, blocks_per_seg;
2841	struct f2fs_super_block *raw_super = (struct f2fs_super_block *)
2842					(bh->b_data + F2FS_SUPER_OFFSET);
2843	size_t crc_offset = 0;
2844	__u32 crc = 0;
2845
2846	if (le32_to_cpu(raw_super->magic) != F2FS_SUPER_MAGIC) {
2847		f2fs_info(sbi, "Magic Mismatch, valid(0x%x) - read(0x%x)",
2848			  F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic));
2849		return -EINVAL;
2850	}
2851
2852	/* Check checksum_offset and crc in superblock */
2853	if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_SB_CHKSUM)) {
2854		crc_offset = le32_to_cpu(raw_super->checksum_offset);
2855		if (crc_offset !=
2856			offsetof(struct f2fs_super_block, crc)) {
2857			f2fs_info(sbi, "Invalid SB checksum offset: %zu",
2858				  crc_offset);
2859			return -EFSCORRUPTED;
2860		}
2861		crc = le32_to_cpu(raw_super->crc);
2862		if (!f2fs_crc_valid(sbi, crc, raw_super, crc_offset)) {
2863			f2fs_info(sbi, "Invalid SB checksum value: %u", crc);
2864			return -EFSCORRUPTED;
2865		}
2866	}
2867
2868	/* Currently, support only 4KB page cache size */
2869	if (F2FS_BLKSIZE != PAGE_SIZE) {
2870		f2fs_info(sbi, "Invalid page_cache_size (%lu), supports only 4KB",
2871			  PAGE_SIZE);
2872		return -EFSCORRUPTED;
2873	}
2874
2875	/* Currently, support only 4KB block size */
2876	if (le32_to_cpu(raw_super->log_blocksize) != F2FS_BLKSIZE_BITS) {
2877		f2fs_info(sbi, "Invalid log_blocksize (%u), supports only %u",
2878			  le32_to_cpu(raw_super->log_blocksize),
2879			  F2FS_BLKSIZE_BITS);
2880		return -EFSCORRUPTED;
2881	}
2882
2883	/* check log blocks per segment */
2884	if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) {
2885		f2fs_info(sbi, "Invalid log blocks per segment (%u)",
2886			  le32_to_cpu(raw_super->log_blocks_per_seg));
2887		return -EFSCORRUPTED;
2888	}
2889
2890	/* Currently, support 512/1024/2048/4096 bytes sector size */
2891	if (le32_to_cpu(raw_super->log_sectorsize) >
2892				F2FS_MAX_LOG_SECTOR_SIZE ||
2893		le32_to_cpu(raw_super->log_sectorsize) <
2894				F2FS_MIN_LOG_SECTOR_SIZE) {
2895		f2fs_info(sbi, "Invalid log sectorsize (%u)",
2896			  le32_to_cpu(raw_super->log_sectorsize));
2897		return -EFSCORRUPTED;
2898	}
2899	if (le32_to_cpu(raw_super->log_sectors_per_block) +
2900		le32_to_cpu(raw_super->log_sectorsize) !=
2901			F2FS_MAX_LOG_SECTOR_SIZE) {
2902		f2fs_info(sbi, "Invalid log sectors per block(%u) log sectorsize(%u)",
2903			  le32_to_cpu(raw_super->log_sectors_per_block),
2904			  le32_to_cpu(raw_super->log_sectorsize));
2905		return -EFSCORRUPTED;
2906	}
2907
2908	segment_count = le32_to_cpu(raw_super->segment_count);
2909	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
2910	segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
2911	secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
2912	total_sections = le32_to_cpu(raw_super->section_count);
2913
2914	/* blocks_per_seg should be 512, given the above check */
2915	blocks_per_seg = 1 << le32_to_cpu(raw_super->log_blocks_per_seg);
2916
2917	if (segment_count > F2FS_MAX_SEGMENT ||
2918				segment_count < F2FS_MIN_SEGMENTS) {
2919		f2fs_info(sbi, "Invalid segment count (%u)", segment_count);
2920		return -EFSCORRUPTED;
2921	}
2922
2923	if (total_sections > segment_count_main || total_sections < 1 ||
2924			segs_per_sec > segment_count || !segs_per_sec) {
2925		f2fs_info(sbi, "Invalid segment/section count (%u, %u x %u)",
2926			  segment_count, total_sections, segs_per_sec);
2927		return -EFSCORRUPTED;
2928	}
2929
2930	if (segment_count_main != total_sections * segs_per_sec) {
2931		f2fs_info(sbi, "Invalid segment/section count (%u != %u * %u)",
2932			  segment_count_main, total_sections, segs_per_sec);
2933		return -EFSCORRUPTED;
2934	}
2935
2936	if ((segment_count / segs_per_sec) < total_sections) {
2937		f2fs_info(sbi, "Small segment_count (%u < %u * %u)",
2938			  segment_count, segs_per_sec, total_sections);
2939		return -EFSCORRUPTED;
2940	}
2941
2942	if (segment_count > (le64_to_cpu(raw_super->block_count) >> 9)) {
2943		f2fs_info(sbi, "Wrong segment_count / block_count (%u > %llu)",
2944			  segment_count, le64_to_cpu(raw_super->block_count));
2945		return -EFSCORRUPTED;
2946	}
2947
2948	if (RDEV(0).path[0]) {
2949		block_t dev_seg_count = le32_to_cpu(RDEV(0).total_segments);
2950		int i = 1;
2951
2952		while (i < MAX_DEVICES && RDEV(i).path[0]) {
2953			dev_seg_count += le32_to_cpu(RDEV(i).total_segments);
2954			i++;
2955		}
2956		if (segment_count != dev_seg_count) {
2957			f2fs_info(sbi, "Segment count (%u) mismatch with total segments from devices (%u)",
2958					segment_count, dev_seg_count);
2959			return -EFSCORRUPTED;
2960		}
2961	} else {
2962		if (__F2FS_HAS_FEATURE(raw_super, F2FS_FEATURE_BLKZONED) &&
2963					!bdev_is_zoned(sbi->sb->s_bdev)) {
2964			f2fs_info(sbi, "Zoned block device path is missing");
2965			return -EFSCORRUPTED;
2966		}
2967	}
2968
2969	if (secs_per_zone > total_sections || !secs_per_zone) {
2970		f2fs_info(sbi, "Wrong secs_per_zone / total_sections (%u, %u)",
2971			  secs_per_zone, total_sections);
2972		return -EFSCORRUPTED;
2973	}
2974	if (le32_to_cpu(raw_super->extension_count) > F2FS_MAX_EXTENSION ||
2975			raw_super->hot_ext_count > F2FS_MAX_EXTENSION ||
2976			(le32_to_cpu(raw_super->extension_count) +
2977			raw_super->hot_ext_count) > F2FS_MAX_EXTENSION) {
2978		f2fs_info(sbi, "Corrupted extension count (%u + %u > %u)",
2979			  le32_to_cpu(raw_super->extension_count),
2980			  raw_super->hot_ext_count,
2981			  F2FS_MAX_EXTENSION);
2982		return -EFSCORRUPTED;
2983	}
2984
2985	if (le32_to_cpu(raw_super->cp_payload) >=
2986				(blocks_per_seg - F2FS_CP_PACKS -
2987				NR_CURSEG_PERSIST_TYPE)) {
2988		f2fs_info(sbi, "Insane cp_payload (%u >= %u)",
2989			  le32_to_cpu(raw_super->cp_payload),
2990			  blocks_per_seg - F2FS_CP_PACKS -
2991			  NR_CURSEG_PERSIST_TYPE);
2992		return -EFSCORRUPTED;
2993	}
2994
2995	/* check reserved ino info */
2996	if (le32_to_cpu(raw_super->node_ino) != 1 ||
2997		le32_to_cpu(raw_super->meta_ino) != 2 ||
2998		le32_to_cpu(raw_super->root_ino) != 3) {
2999		f2fs_info(sbi, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)",
3000			  le32_to_cpu(raw_super->node_ino),
3001			  le32_to_cpu(raw_super->meta_ino),
3002			  le32_to_cpu(raw_super->root_ino));
3003		return -EFSCORRUPTED;
3004	}
3005
3006	/* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */
3007	if (sanity_check_area_boundary(sbi, bh))
3008		return -EFSCORRUPTED;
3009
3010	return 0;
3011}
3012
3013int f2fs_sanity_check_ckpt(struct f2fs_sb_info *sbi)
3014{
3015	unsigned int total, fsmeta;
3016	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3017	struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi);
3018	unsigned int ovp_segments, reserved_segments;
3019	unsigned int main_segs, blocks_per_seg;
3020	unsigned int sit_segs, nat_segs;
3021	unsigned int sit_bitmap_size, nat_bitmap_size;
3022	unsigned int log_blocks_per_seg;
3023	unsigned int segment_count_main;
3024	unsigned int cp_pack_start_sum, cp_payload;
3025	block_t user_block_count, valid_user_blocks;
3026	block_t avail_node_count, valid_node_count;
3027	unsigned int nat_blocks, nat_bits_bytes, nat_bits_blocks;
3028	int i, j;
3029
3030	total = le32_to_cpu(raw_super->segment_count);
3031	fsmeta = le32_to_cpu(raw_super->segment_count_ckpt);
3032	sit_segs = le32_to_cpu(raw_super->segment_count_sit);
3033	fsmeta += sit_segs;
3034	nat_segs = le32_to_cpu(raw_super->segment_count_nat);
3035	fsmeta += nat_segs;
3036	fsmeta += le32_to_cpu(ckpt->rsvd_segment_count);
3037	fsmeta += le32_to_cpu(raw_super->segment_count_ssa);
3038
3039	if (unlikely(fsmeta >= total))
3040		return 1;
3041
3042	ovp_segments = le32_to_cpu(ckpt->overprov_segment_count);
3043	reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count);
3044
3045	if (unlikely(fsmeta < F2FS_MIN_META_SEGMENTS ||
3046			ovp_segments == 0 || reserved_segments == 0)) {
3047		f2fs_err(sbi, "Wrong layout: check mkfs.f2fs version");
3048		return 1;
3049	}
3050
3051	user_block_count = le64_to_cpu(ckpt->user_block_count);
3052	segment_count_main = le32_to_cpu(raw_super->segment_count_main);
3053	log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3054	if (!user_block_count || user_block_count >=
3055			segment_count_main << log_blocks_per_seg) {
3056		f2fs_err(sbi, "Wrong user_block_count: %u",
3057			 user_block_count);
3058		return 1;
3059	}
3060
3061	valid_user_blocks = le64_to_cpu(ckpt->valid_block_count);
3062	if (valid_user_blocks > user_block_count) {
3063		f2fs_err(sbi, "Wrong valid_user_blocks: %u, user_block_count: %u",
3064			 valid_user_blocks, user_block_count);
3065		return 1;
3066	}
3067
3068	valid_node_count = le32_to_cpu(ckpt->valid_node_count);
3069	avail_node_count = sbi->total_node_count - F2FS_RESERVED_NODE_NUM;
3070	if (valid_node_count > avail_node_count) {
3071		f2fs_err(sbi, "Wrong valid_node_count: %u, avail_node_count: %u",
3072			 valid_node_count, avail_node_count);
3073		return 1;
3074	}
3075
3076	main_segs = le32_to_cpu(raw_super->segment_count_main);
3077	blocks_per_seg = sbi->blocks_per_seg;
3078
3079	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3080		if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs ||
3081			le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg)
3082			return 1;
3083		for (j = i + 1; j < NR_CURSEG_NODE_TYPE; j++) {
3084			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3085				le32_to_cpu(ckpt->cur_node_segno[j])) {
3086				f2fs_err(sbi, "Node segment (%u, %u) has the same segno: %u",
3087					 i, j,
3088					 le32_to_cpu(ckpt->cur_node_segno[i]));
3089				return 1;
3090			}
3091		}
3092	}
3093	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
3094		if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs ||
3095			le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg)
3096			return 1;
3097		for (j = i + 1; j < NR_CURSEG_DATA_TYPE; j++) {
3098			if (le32_to_cpu(ckpt->cur_data_segno[i]) ==
3099				le32_to_cpu(ckpt->cur_data_segno[j])) {
3100				f2fs_err(sbi, "Data segment (%u, %u) has the same segno: %u",
3101					 i, j,
3102					 le32_to_cpu(ckpt->cur_data_segno[i]));
3103				return 1;
3104			}
3105		}
3106	}
3107	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
3108		for (j = 0; j < NR_CURSEG_DATA_TYPE; j++) {
3109			if (le32_to_cpu(ckpt->cur_node_segno[i]) ==
3110				le32_to_cpu(ckpt->cur_data_segno[j])) {
3111				f2fs_err(sbi, "Node segment (%u) and Data segment (%u) has the same segno: %u",
3112					 i, j,
3113					 le32_to_cpu(ckpt->cur_node_segno[i]));
3114				return 1;
3115			}
3116		}
3117	}
3118
3119	sit_bitmap_size = le32_to_cpu(ckpt->sit_ver_bitmap_bytesize);
3120	nat_bitmap_size = le32_to_cpu(ckpt->nat_ver_bitmap_bytesize);
3121
3122	if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
3123		nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
3124		f2fs_err(sbi, "Wrong bitmap size: sit: %u, nat:%u",
3125			 sit_bitmap_size, nat_bitmap_size);
3126		return 1;
3127	}
3128
3129	cp_pack_start_sum = __start_sum_addr(sbi);
3130	cp_payload = __cp_payload(sbi);
3131	if (cp_pack_start_sum < cp_payload + 1 ||
3132		cp_pack_start_sum > blocks_per_seg - 1 -
3133			NR_CURSEG_PERSIST_TYPE) {
3134		f2fs_err(sbi, "Wrong cp_pack_start_sum: %u",
3135			 cp_pack_start_sum);
3136		return 1;
3137	}
3138
3139	if (__is_set_ckpt_flags(ckpt, CP_LARGE_NAT_BITMAP_FLAG) &&
3140		le32_to_cpu(ckpt->checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
3141		f2fs_warn(sbi, "using deprecated layout of large_nat_bitmap, "
3142			  "please run fsck v1.13.0 or higher to repair, chksum_offset: %u, "
3143			  "fixed with patch: \"f2fs-tools: relocate chksum_offset for large_nat_bitmap feature\"",
3144			  le32_to_cpu(ckpt->checksum_offset));
3145		return 1;
3146	}
3147
3148	nat_blocks = nat_segs << log_blocks_per_seg;
3149	nat_bits_bytes = nat_blocks / BITS_PER_BYTE;
3150	nat_bits_blocks = F2FS_BLK_ALIGN((nat_bits_bytes << 1) + 8);
3151	if (__is_set_ckpt_flags(ckpt, CP_NAT_BITS_FLAG) &&
3152		(cp_payload + F2FS_CP_PACKS +
3153		NR_CURSEG_PERSIST_TYPE + nat_bits_blocks >= blocks_per_seg)) {
3154		f2fs_warn(sbi, "Insane cp_payload: %u, nat_bits_blocks: %u)",
3155			  cp_payload, nat_bits_blocks);
3156		return 1;
3157	}
3158
3159	if (unlikely(f2fs_cp_error(sbi))) {
3160		f2fs_err(sbi, "A bug case: need to run fsck");
3161		return 1;
3162	}
3163	return 0;
3164}
3165
3166static void init_sb_info(struct f2fs_sb_info *sbi)
3167{
3168	struct f2fs_super_block *raw_super = sbi->raw_super;
3169	int i;
3170
3171	sbi->log_sectors_per_block =
3172		le32_to_cpu(raw_super->log_sectors_per_block);
3173	sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize);
3174	sbi->blocksize = 1 << sbi->log_blocksize;
3175	sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg);
3176	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
3177	sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec);
3178	sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone);
3179	sbi->total_sections = le32_to_cpu(raw_super->section_count);
3180	sbi->total_node_count =
3181		(le32_to_cpu(raw_super->segment_count_nat) / 2)
3182			* sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
3183	sbi->root_ino_num = le32_to_cpu(raw_super->root_ino);
3184	sbi->node_ino_num = le32_to_cpu(raw_super->node_ino);
3185	sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino);
3186	sbi->cur_victim_sec = NULL_SECNO;
3187	sbi->next_victim_seg[BG_GC] = NULL_SEGNO;
3188	sbi->next_victim_seg[FG_GC] = NULL_SEGNO;
3189	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
3190	sbi->migration_granularity = sbi->segs_per_sec;
3191
3192	sbi->dir_level = DEF_DIR_LEVEL;
3193	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
3194	sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL;
3195	sbi->interval_time[DISCARD_TIME] = DEF_IDLE_INTERVAL;
3196	sbi->interval_time[GC_TIME] = DEF_IDLE_INTERVAL;
3197	sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_INTERVAL;
3198	sbi->interval_time[UMOUNT_DISCARD_TIMEOUT] =
3199				DEF_UMOUNT_DISCARD_TIMEOUT;
3200	clear_sbi_flag(sbi, SBI_NEED_FSCK);
3201
3202	for (i = 0; i < NR_COUNT_TYPE; i++)
3203		atomic_set(&sbi->nr_pages[i], 0);
3204
3205	for (i = 0; i < META; i++)
3206		atomic_set(&sbi->wb_sync_req[i], 0);
3207
3208	INIT_LIST_HEAD(&sbi->s_list);
3209	mutex_init(&sbi->umount_mutex);
3210	init_rwsem(&sbi->io_order_lock);
3211	spin_lock_init(&sbi->cp_lock);
3212
3213	sbi->dirty_device = 0;
3214	spin_lock_init(&sbi->dev_lock);
3215
3216	init_rwsem(&sbi->sb_lock);
3217	init_rwsem(&sbi->pin_sem);
3218}
3219
3220static int init_percpu_info(struct f2fs_sb_info *sbi)
3221{
3222	int err;
3223
3224	err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL);
3225	if (err)
3226		return err;
3227
3228	err = percpu_counter_init(&sbi->total_valid_inode_count, 0,
3229								GFP_KERNEL);
3230	if (err)
3231		percpu_counter_destroy(&sbi->alloc_valid_block_count);
3232
3233	return err;
3234}
3235
3236#ifdef CONFIG_BLK_DEV_ZONED
3237
3238struct f2fs_report_zones_args {
3239	struct f2fs_sb_info *sbi;
3240	struct f2fs_dev_info *dev;
3241};
3242
3243static int f2fs_report_zone_cb(struct blk_zone *zone, unsigned int idx,
3244			      void *data)
3245{
3246	struct f2fs_report_zones_args *rz_args = data;
3247	block_t unusable_blocks = (zone->len - zone->capacity) >>
3248					F2FS_LOG_SECTORS_PER_BLOCK;
3249
3250	if (zone->type == BLK_ZONE_TYPE_CONVENTIONAL)
3251		return 0;
3252
3253	set_bit(idx, rz_args->dev->blkz_seq);
3254	if (!rz_args->sbi->unusable_blocks_per_sec) {
3255		rz_args->sbi->unusable_blocks_per_sec = unusable_blocks;
3256		return 0;
3257	}
3258	if (rz_args->sbi->unusable_blocks_per_sec != unusable_blocks) {
3259		f2fs_err(rz_args->sbi, "F2FS supports single zone capacity\n");
3260		return -EINVAL;
3261	}
3262	return 0;
3263}
3264
3265static int init_blkz_info(struct f2fs_sb_info *sbi, int devi)
3266{
3267	struct block_device *bdev = FDEV(devi).bdev;
3268	sector_t nr_sectors = bdev->bd_part->nr_sects;
3269	struct f2fs_report_zones_args rep_zone_arg;
3270	int ret;
3271
3272	if (!f2fs_sb_has_blkzoned(sbi))
3273		return 0;
3274
3275	if (sbi->blocks_per_blkz && sbi->blocks_per_blkz !=
3276				SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)))
3277		return -EINVAL;
3278	sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev));
3279	if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz !=
3280				__ilog2_u32(sbi->blocks_per_blkz))
3281		return -EINVAL;
3282	sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz);
3283	FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >>
3284					sbi->log_blocks_per_blkz;
3285	if (nr_sectors & (bdev_zone_sectors(bdev) - 1))
3286		FDEV(devi).nr_blkz++;
3287
3288	FDEV(devi).blkz_seq = f2fs_kvzalloc(sbi,
3289					BITS_TO_LONGS(FDEV(devi).nr_blkz)
3290					* sizeof(unsigned long),
3291					GFP_KERNEL);
3292	if (!FDEV(devi).blkz_seq)
3293		return -ENOMEM;
3294
3295	rep_zone_arg.sbi = sbi;
3296	rep_zone_arg.dev = &FDEV(devi);
3297
3298	ret = blkdev_report_zones(bdev, 0, BLK_ALL_ZONES, f2fs_report_zone_cb,
3299				  &rep_zone_arg);
3300	if (ret < 0)
3301		return ret;
3302	return 0;
3303}
3304#endif
3305
3306/*
3307 * Read f2fs raw super block.
3308 * Because we have two copies of super block, so read both of them
3309 * to get the first valid one. If any one of them is broken, we pass
3310 * them recovery flag back to the caller.
3311 */
3312static int read_raw_super_block(struct f2fs_sb_info *sbi,
3313			struct f2fs_super_block **raw_super,
3314			int *valid_super_block, int *recovery)
3315{
3316	struct super_block *sb = sbi->sb;
3317	int block;
3318	struct buffer_head *bh;
3319	struct f2fs_super_block *super;
3320	int err = 0;
3321
3322	super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL);
3323	if (!super)
3324		return -ENOMEM;
3325
3326	for (block = 0; block < 2; block++) {
3327		bh = sb_bread(sb, block);
3328		if (!bh) {
3329			f2fs_err(sbi, "Unable to read %dth superblock",
3330				 block + 1);
3331			err = -EIO;
3332			*recovery = 1;
3333			continue;
3334		}
3335
3336		/* sanity checking of raw super */
3337		err = sanity_check_raw_super(sbi, bh);
3338		if (err) {
3339			f2fs_err(sbi, "Can't find valid F2FS filesystem in %dth superblock",
3340				 block + 1);
3341			brelse(bh);
3342			*recovery = 1;
3343			continue;
3344		}
3345
3346		if (!*raw_super) {
3347			memcpy(super, bh->b_data + F2FS_SUPER_OFFSET,
3348							sizeof(*super));
3349			*valid_super_block = block;
3350			*raw_super = super;
3351		}
3352		brelse(bh);
3353	}
3354
3355	/* No valid superblock */
3356	if (!*raw_super)
3357		kfree(super);
3358	else
3359		err = 0;
3360
3361	return err;
3362}
3363
3364int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover)
3365{
3366	struct buffer_head *bh;
3367	__u32 crc = 0;
3368	int err;
3369
3370	if ((recover && f2fs_readonly(sbi->sb)) ||
3371				bdev_read_only(sbi->sb->s_bdev)) {
3372		set_sbi_flag(sbi, SBI_NEED_SB_WRITE);
3373		return -EROFS;
3374	}
3375
3376	/* we should update superblock crc here */
3377	if (!recover && f2fs_sb_has_sb_chksum(sbi)) {
3378		crc = f2fs_crc32(sbi, F2FS_RAW_SUPER(sbi),
3379				offsetof(struct f2fs_super_block, crc));
3380		F2FS_RAW_SUPER(sbi)->crc = cpu_to_le32(crc);
3381	}
3382
3383	/* write back-up superblock first */
3384	bh = sb_bread(sbi->sb, sbi->valid_super_block ? 0 : 1);
3385	if (!bh)
3386		return -EIO;
3387	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3388	brelse(bh);
3389
3390	/* if we are in recovery path, skip writing valid superblock */
3391	if (recover || err)
3392		return err;
3393
3394	/* write current valid superblock */
3395	bh = sb_bread(sbi->sb, sbi->valid_super_block);
3396	if (!bh)
3397		return -EIO;
3398	err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi));
3399	brelse(bh);
3400	return err;
3401}
3402
3403static int f2fs_scan_devices(struct f2fs_sb_info *sbi)
3404{
3405	struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi);
3406	unsigned int max_devices = MAX_DEVICES;
3407	int i;
3408
3409	/* Initialize single device information */
3410	if (!RDEV(0).path[0]) {
3411		if (!bdev_is_zoned(sbi->sb->s_bdev))
3412			return 0;
3413		max_devices = 1;
3414	}
3415
3416	/*
3417	 * Initialize multiple devices information, or single
3418	 * zoned block device information.
3419	 */
3420	sbi->devs = f2fs_kzalloc(sbi,
3421				 array_size(max_devices,
3422					    sizeof(struct f2fs_dev_info)),
3423				 GFP_KERNEL);
3424	if (!sbi->devs)
3425		return -ENOMEM;
3426
3427	for (i = 0; i < max_devices; i++) {
3428
3429		if (i > 0 && !RDEV(i).path[0])
3430			break;
3431
3432		if (max_devices == 1) {
3433			/* Single zoned block device mount */
3434			FDEV(0).bdev =
3435				blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev,
3436					sbi->sb->s_mode, sbi->sb->s_type);
3437		} else {
3438			/* Multi-device mount */
3439			memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN);
3440			FDEV(i).total_segments =
3441				le32_to_cpu(RDEV(i).total_segments);
3442			if (i == 0) {
3443				FDEV(i).start_blk = 0;
3444				FDEV(i).end_blk = FDEV(i).start_blk +
3445				    (FDEV(i).total_segments <<
3446				    sbi->log_blocks_per_seg) - 1 +
3447				    le32_to_cpu(raw_super->segment0_blkaddr);
3448			} else {
3449				FDEV(i).start_blk = FDEV(i - 1).end_blk + 1;
3450				FDEV(i).end_blk = FDEV(i).start_blk +
3451					(FDEV(i).total_segments <<
3452					sbi->log_blocks_per_seg) - 1;
3453			}
3454			FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path,
3455					sbi->sb->s_mode, sbi->sb->s_type);
3456		}
3457		if (IS_ERR(FDEV(i).bdev))
3458			return PTR_ERR(FDEV(i).bdev);
3459
3460		/* to release errored devices */
3461		sbi->s_ndevs = i + 1;
3462
3463#ifdef CONFIG_BLK_DEV_ZONED
3464		if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM &&
3465				!f2fs_sb_has_blkzoned(sbi)) {
3466			f2fs_err(sbi, "Zoned block device feature not enabled\n");
3467			return -EINVAL;
3468		}
3469		if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) {
3470			if (init_blkz_info(sbi, i)) {
3471				f2fs_err(sbi, "Failed to initialize F2FS blkzone information");
3472				return -EINVAL;
3473			}
3474			if (max_devices == 1)
3475				break;
3476			f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)",
3477				  i, FDEV(i).path,
3478				  FDEV(i).total_segments,
3479				  FDEV(i).start_blk, FDEV(i).end_blk,
3480				  bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ?
3481				  "Host-aware" : "Host-managed");
3482			continue;
3483		}
3484#endif
3485		f2fs_info(sbi, "Mount Device [%2d]: %20s, %8u, %8x - %8x",
3486			  i, FDEV(i).path,
3487			  FDEV(i).total_segments,
3488			  FDEV(i).start_blk, FDEV(i).end_blk);
3489	}
3490	f2fs_info(sbi,
3491		  "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi));
3492	return 0;
3493}
3494
3495static int f2fs_setup_casefold(struct f2fs_sb_info *sbi)
3496{
3497#ifdef CONFIG_UNICODE
3498	if (f2fs_sb_has_casefold(sbi) && !sbi->sb->s_encoding) {
3499		const struct f2fs_sb_encodings *encoding_info;
3500		struct unicode_map *encoding;
3501		__u16 encoding_flags;
3502
3503		if (f2fs_sb_has_encrypt(sbi)) {
3504			f2fs_err(sbi,
3505				"Can't mount with encoding and encryption");
3506			return -EINVAL;
3507		}
3508
3509		if (f2fs_sb_read_encoding(sbi->raw_super, &encoding_info,
3510					  &encoding_flags)) {
3511			f2fs_err(sbi,
3512				 "Encoding requested by superblock is unknown");
3513			return -EINVAL;
3514		}
3515
3516		encoding = utf8_load(encoding_info->version);
3517		if (IS_ERR(encoding)) {
3518			f2fs_err(sbi,
3519				 "can't mount with superblock charset: %s-%s "
3520				 "not supported by the kernel. flags: 0x%x.",
3521				 encoding_info->name, encoding_info->version,
3522				 encoding_flags);
3523			return PTR_ERR(encoding);
3524		}
3525		f2fs_info(sbi, "Using encoding defined by superblock: "
3526			 "%s-%s with flags 0x%hx", encoding_info->name,
3527			 encoding_info->version?:"\b", encoding_flags);
3528
3529		sbi->sb->s_encoding = encoding;
3530		sbi->sb->s_encoding_flags = encoding_flags;
3531		sbi->sb->s_d_op = &f2fs_dentry_ops;
3532	}
3533#else
3534	if (f2fs_sb_has_casefold(sbi)) {
3535		f2fs_err(sbi, "Filesystem with casefold feature cannot be mounted without CONFIG_UNICODE");
3536		return -EINVAL;
3537	}
3538#endif
3539	return 0;
3540}
3541
3542static void f2fs_tuning_parameters(struct f2fs_sb_info *sbi)
3543{
3544	struct f2fs_sm_info *sm_i = SM_I(sbi);
3545
3546	/* adjust parameters according to the volume size */
3547	if (sm_i->main_segments <= SMALL_VOLUME_SEGMENTS) {
3548		F2FS_OPTION(sbi).alloc_mode = ALLOC_MODE_REUSE;
3549		sm_i->dcc_info->discard_granularity = 1;
3550		sm_i->ipu_policy = 1 << F2FS_IPU_FORCE;
3551	}
3552
3553	sbi->readdir_ra = 1;
3554}
3555
3556#ifdef CONFIG_F2FS_GRADING_SSR
3557static void f2fs_init_grading_ssr(struct f2fs_sb_info *sbi)
3558{
3559	u32 total_blocks = le64_to_cpu(sbi->raw_super->block_count) >> 18;
3560
3561	if (total_blocks > 64) { /* 64G */
3562		sbi->hot_cold_params.hot_data_lower_limit = SSR_HD_SAPCE_LIMIT_128G;
3563		sbi->hot_cold_params.hot_data_waterline = SSR_HD_WATERLINE_128G;
3564		sbi->hot_cold_params.warm_data_lower_limit = SSR_WD_SAPCE_LIMIT_128G;
3565		sbi->hot_cold_params.warm_data_waterline = SSR_WD_WATERLINE_128G;
3566		sbi->hot_cold_params.hot_node_lower_limit = SSR_HD_SAPCE_LIMIT_128G;
3567		sbi->hot_cold_params.hot_node_waterline = SSR_HN_WATERLINE_128G;
3568		sbi->hot_cold_params.warm_node_lower_limit = SSR_WN_SAPCE_LIMIT_128G;
3569		sbi->hot_cold_params.warm_node_waterline = SSR_WN_WATERLINE_128G;
3570		sbi->hot_cold_params.enable = GRADING_SSR_OFF;
3571	} else {
3572		sbi->hot_cold_params.hot_data_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3573		sbi->hot_cold_params.hot_data_waterline = SSR_DEFALT_WATERLINE;
3574		sbi->hot_cold_params.warm_data_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3575		sbi->hot_cold_params.warm_data_waterline = SSR_DEFALT_WATERLINE;
3576		sbi->hot_cold_params.hot_node_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3577		sbi->hot_cold_params.hot_node_waterline = SSR_DEFALT_WATERLINE;
3578		sbi->hot_cold_params.warm_node_lower_limit = SSR_DEFALT_SPACE_LIMIT;
3579		sbi->hot_cold_params.warm_node_waterline = SSR_DEFALT_WATERLINE;
3580		sbi->hot_cold_params.enable = GRADING_SSR_OFF;
3581	}
3582}
3583#endif
3584
3585static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
3586{
3587	struct f2fs_sb_info *sbi;
3588	struct f2fs_super_block *raw_super;
3589	struct inode *root;
3590	int err;
3591	bool skip_recovery = false, need_fsck = false;
3592	char *options = NULL;
3593	int recovery, i, valid_super_block;
3594	struct curseg_info *seg_i;
3595	int retry_cnt = 1;
3596
3597try_onemore:
3598	err = -EINVAL;
3599	raw_super = NULL;
3600	valid_super_block = -1;
3601	recovery = 0;
3602
3603	/* allocate memory for f2fs-specific super block info */
3604	sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL);
3605	if (!sbi)
3606		return -ENOMEM;
3607
3608	sbi->sb = sb;
3609
3610	/* Load the checksum driver */
3611	sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0);
3612	if (IS_ERR(sbi->s_chksum_driver)) {
3613		f2fs_err(sbi, "Cannot load crc32 driver.");
3614		err = PTR_ERR(sbi->s_chksum_driver);
3615		sbi->s_chksum_driver = NULL;
3616		goto free_sbi;
3617	}
3618
3619	/* set a block size */
3620	if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) {
3621		f2fs_err(sbi, "unable to set blocksize");
3622		goto free_sbi;
3623	}
3624
3625	err = read_raw_super_block(sbi, &raw_super, &valid_super_block,
3626								&recovery);
3627	if (err)
3628		goto free_sbi;
3629
3630	sb->s_fs_info = sbi;
3631	sbi->raw_super = raw_super;
3632
3633	/* precompute checksum seed for metadata */
3634	if (f2fs_sb_has_inode_chksum(sbi))
3635		sbi->s_chksum_seed = f2fs_chksum(sbi, ~0, raw_super->uuid,
3636						sizeof(raw_super->uuid));
3637
3638	default_options(sbi);
3639	/* parse mount options */
3640	options = kstrdup((const char *)data, GFP_KERNEL);
3641	if (data && !options) {
3642		err = -ENOMEM;
3643		goto free_sb_buf;
3644	}
3645
3646	err = parse_options(sb, options, false);
3647	if (err)
3648		goto free_options;
3649
3650	sbi->max_file_blocks = max_file_blocks();
3651	sb->s_maxbytes = sbi->max_file_blocks <<
3652				le32_to_cpu(raw_super->log_blocksize);
3653	sb->s_max_links = F2FS_LINK_MAX;
3654
3655	err = f2fs_setup_casefold(sbi);
3656	if (err)
3657		goto free_options;
3658
3659#ifdef CONFIG_QUOTA
3660	sb->dq_op = &f2fs_quota_operations;
3661	sb->s_qcop = &f2fs_quotactl_ops;
3662	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3663
3664	if (f2fs_sb_has_quota_ino(sbi)) {
3665		for (i = 0; i < MAXQUOTAS; i++) {
3666			if (f2fs_qf_ino(sbi->sb, i))
3667				sbi->nquota_files++;
3668		}
3669	}
3670#endif
3671
3672	sb->s_op = &f2fs_sops;
3673#ifdef CONFIG_FS_ENCRYPTION
3674	sb->s_cop = &f2fs_cryptops;
3675#endif
3676#ifdef CONFIG_FS_VERITY
3677	sb->s_vop = &f2fs_verityops;
3678#endif
3679	sb->s_xattr = f2fs_xattr_handlers;
3680	sb->s_export_op = &f2fs_export_ops;
3681	sb->s_magic = F2FS_SUPER_MAGIC;
3682	sb->s_time_gran = 1;
3683	sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3684		(test_opt(sbi, POSIX_ACL) ? SB_POSIXACL : 0);
3685	memcpy(&sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid));
3686	sb->s_iflags |= SB_I_CGROUPWB;
3687
3688	/* init f2fs-specific super block info */
3689	sbi->valid_super_block = valid_super_block;
3690	init_rwsem(&sbi->gc_lock);
3691	mutex_init(&sbi->writepages);
3692	mutex_init(&sbi->cp_mutex);
3693	init_rwsem(&sbi->node_write);
3694	init_rwsem(&sbi->node_change);
3695
3696	/* disallow all the data/node/meta page writes */
3697	set_sbi_flag(sbi, SBI_POR_DOING);
3698	spin_lock_init(&sbi->stat_lock);
3699
3700	/* init iostat info */
3701	spin_lock_init(&sbi->iostat_lock);
3702	sbi->iostat_enable = false;
3703	sbi->iostat_period_ms = DEFAULT_IOSTAT_PERIOD_MS;
3704
3705	for (i = 0; i < NR_PAGE_TYPE; i++) {
3706		int n = (i == META) ? 1: NR_TEMP_TYPE;
3707		int j;
3708
3709		sbi->write_io[i] =
3710			f2fs_kmalloc(sbi,
3711				     array_size(n,
3712						sizeof(struct f2fs_bio_info)),
3713				     GFP_KERNEL);
3714		if (!sbi->write_io[i]) {
3715			err = -ENOMEM;
3716			goto free_bio_info;
3717		}
3718
3719		for (j = HOT; j < n; j++) {
3720			init_rwsem(&sbi->write_io[i][j].io_rwsem);
3721			sbi->write_io[i][j].sbi = sbi;
3722			sbi->write_io[i][j].bio = NULL;
3723			spin_lock_init(&sbi->write_io[i][j].io_lock);
3724			INIT_LIST_HEAD(&sbi->write_io[i][j].io_list);
3725			INIT_LIST_HEAD(&sbi->write_io[i][j].bio_list);
3726			init_rwsem(&sbi->write_io[i][j].bio_list_lock);
3727		}
3728	}
3729
3730	init_rwsem(&sbi->cp_rwsem);
3731	init_rwsem(&sbi->quota_sem);
3732	init_waitqueue_head(&sbi->cp_wait);
3733	init_sb_info(sbi);
3734
3735	err = init_percpu_info(sbi);
3736	if (err)
3737		goto free_bio_info;
3738
3739	if (F2FS_IO_ALIGNED(sbi)) {
3740		sbi->write_io_dummy =
3741			mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0);
3742		if (!sbi->write_io_dummy) {
3743			err = -ENOMEM;
3744			goto free_percpu;
3745		}
3746	}
3747
3748	/* init per sbi slab cache */
3749	err = f2fs_init_xattr_caches(sbi);
3750	if (err)
3751		goto free_io_dummy;
3752	err = f2fs_init_page_array_cache(sbi);
3753	if (err)
3754		goto free_xattr_cache;
3755
3756	/* get an inode for meta space */
3757	sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi));
3758	if (IS_ERR(sbi->meta_inode)) {
3759		f2fs_err(sbi, "Failed to read F2FS meta data inode");
3760		err = PTR_ERR(sbi->meta_inode);
3761		goto free_page_array_cache;
3762	}
3763
3764	err = f2fs_get_valid_checkpoint(sbi);
3765	if (err) {
3766		f2fs_err(sbi, "Failed to get valid F2FS checkpoint");
3767		goto free_meta_inode;
3768	}
3769
3770	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_QUOTA_NEED_FSCK_FLAG))
3771		set_sbi_flag(sbi, SBI_QUOTA_NEED_REPAIR);
3772	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_DISABLED_QUICK_FLAG)) {
3773		set_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3774		sbi->interval_time[DISABLE_TIME] = DEF_DISABLE_QUICK_INTERVAL;
3775	}
3776
3777	if (__is_set_ckpt_flags(F2FS_CKPT(sbi), CP_FSCK_FLAG))
3778		set_sbi_flag(sbi, SBI_NEED_FSCK);
3779
3780	/* Initialize device list */
3781	err = f2fs_scan_devices(sbi);
3782	if (err) {
3783		f2fs_err(sbi, "Failed to find devices");
3784		goto free_devices;
3785	}
3786
3787	err = f2fs_init_post_read_wq(sbi);
3788	if (err) {
3789		f2fs_err(sbi, "Failed to initialize post read workqueue");
3790		goto free_devices;
3791	}
3792
3793	sbi->total_valid_node_count =
3794				le32_to_cpu(sbi->ckpt->valid_node_count);
3795	percpu_counter_set(&sbi->total_valid_inode_count,
3796				le32_to_cpu(sbi->ckpt->valid_inode_count));
3797	sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count);
3798	sbi->total_valid_block_count =
3799				le64_to_cpu(sbi->ckpt->valid_block_count);
3800	sbi->last_valid_block_count = sbi->total_valid_block_count;
3801	sbi->reserved_blocks = 0;
3802	sbi->current_reserved_blocks = 0;
3803	limit_reserve_root(sbi);
3804	adjust_unusable_cap_perc(sbi);
3805
3806	for (i = 0; i < NR_INODE_TYPE; i++) {
3807		INIT_LIST_HEAD(&sbi->inode_list[i]);
3808		spin_lock_init(&sbi->inode_lock[i]);
3809	}
3810	mutex_init(&sbi->flush_lock);
3811
3812	f2fs_init_extent_cache_info(sbi);
3813
3814	f2fs_init_ino_entry_info(sbi);
3815
3816	f2fs_init_fsync_node_info(sbi);
3817
3818	/* setup f2fs internal modules */
3819	err = f2fs_build_segment_manager(sbi);
3820	if (err) {
3821		f2fs_err(sbi, "Failed to initialize F2FS segment manager (%d)",
3822			 err);
3823		goto free_sm;
3824	}
3825	err = f2fs_build_node_manager(sbi);
3826	if (err) {
3827		f2fs_err(sbi, "Failed to initialize F2FS node manager (%d)",
3828			 err);
3829		goto free_nm;
3830	}
3831
3832	err = adjust_reserved_segment(sbi);
3833	if (err)
3834		goto free_nm;
3835
3836	/* For write statistics */
3837	if (sb->s_bdev->bd_part)
3838		sbi->sectors_written_start =
3839			(u64)part_stat_read(sb->s_bdev->bd_part,
3840					    sectors[STAT_WRITE]);
3841
3842	/* Read accumulated write IO statistics if exists */
3843	seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE);
3844	if (__exist_node_summaries(sbi))
3845		sbi->kbytes_written =
3846			le64_to_cpu(seg_i->journal->info.kbytes_written);
3847
3848	f2fs_build_gc_manager(sbi);
3849
3850	err = f2fs_build_stats(sbi);
3851	if (err)
3852		goto free_nm;
3853
3854	/* get an inode for node space */
3855	sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi));
3856	if (IS_ERR(sbi->node_inode)) {
3857		f2fs_err(sbi, "Failed to read node inode");
3858		err = PTR_ERR(sbi->node_inode);
3859		goto free_stats;
3860	}
3861
3862	/* read root inode and dentry */
3863	root = f2fs_iget(sb, F2FS_ROOT_INO(sbi));
3864	if (IS_ERR(root)) {
3865		f2fs_err(sbi, "Failed to read root inode");
3866		err = PTR_ERR(root);
3867		goto free_node_inode;
3868	}
3869	if (!S_ISDIR(root->i_mode) || !root->i_blocks ||
3870			!root->i_size || !root->i_nlink) {
3871		iput(root);
3872		err = -EINVAL;
3873		goto free_node_inode;
3874	}
3875
3876	sb->s_root = d_make_root(root); /* allocate root dentry */
3877	if (!sb->s_root) {
3878		err = -ENOMEM;
3879		goto free_node_inode;
3880	}
3881#ifdef CONFIG_F2FS_GRADING_SSR
3882	f2fs_init_grading_ssr(sbi);
3883#endif
3884	err = f2fs_register_sysfs(sbi);
3885	if (err)
3886		goto free_root_inode;
3887
3888#ifdef CONFIG_QUOTA
3889	/* Enable quota usage during mount */
3890	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb)) {
3891		err = f2fs_enable_quotas(sb);
3892		if (err)
3893			f2fs_err(sbi, "Cannot turn on quotas: error %d", err);
3894	}
3895#endif
3896	/* if there are any orphan inodes, free them */
3897	err = f2fs_recover_orphan_inodes(sbi);
3898	if (err)
3899		goto free_meta;
3900
3901	if (unlikely(is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)))
3902		goto reset_checkpoint;
3903
3904	/* recover fsynced data */
3905	if (!test_opt(sbi, DISABLE_ROLL_FORWARD) &&
3906			!test_opt(sbi, NORECOVERY)) {
3907		/*
3908		 * mount should be failed, when device has readonly mode, and
3909		 * previous checkpoint was not done by clean system shutdown.
3910		 */
3911		if (f2fs_hw_is_readonly(sbi)) {
3912			if (!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
3913				err = -EROFS;
3914				f2fs_err(sbi, "Need to recover fsync data, but write access unavailable");
3915				goto free_meta;
3916			}
3917			f2fs_info(sbi, "write access unavailable, skipping recovery");
3918			goto reset_checkpoint;
3919		}
3920
3921		if (need_fsck)
3922			set_sbi_flag(sbi, SBI_NEED_FSCK);
3923
3924		if (skip_recovery)
3925			goto reset_checkpoint;
3926
3927		err = f2fs_recover_fsync_data(sbi, false);
3928		if (err < 0) {
3929			if (err != -ENOMEM)
3930				skip_recovery = true;
3931			need_fsck = true;
3932			f2fs_err(sbi, "Cannot recover all fsync data errno=%d",
3933				 err);
3934			goto free_meta;
3935		}
3936	} else {
3937		err = f2fs_recover_fsync_data(sbi, true);
3938
3939		if (!f2fs_readonly(sb) && err > 0) {
3940			err = -EINVAL;
3941			f2fs_err(sbi, "Need to recover fsync data");
3942			goto free_meta;
3943		}
3944	}
3945
3946	/*
3947	 * If the f2fs is not readonly and fsync data recovery succeeds,
3948	 * check zoned block devices' write pointer consistency.
3949	 */
3950	if (!err && !f2fs_readonly(sb) && f2fs_sb_has_blkzoned(sbi)) {
3951		err = f2fs_check_write_pointer(sbi);
3952		if (err)
3953			goto free_meta;
3954	}
3955
3956reset_checkpoint:
3957	f2fs_init_inmem_curseg(sbi);
3958
3959	/* f2fs_recover_fsync_data() cleared this already */
3960	clear_sbi_flag(sbi, SBI_POR_DOING);
3961
3962	if (test_opt(sbi, DISABLE_CHECKPOINT)) {
3963		err = f2fs_disable_checkpoint(sbi);
3964		if (err)
3965			goto sync_free_meta;
3966	} else if (is_set_ckpt_flags(sbi, CP_DISABLED_FLAG)) {
3967		f2fs_enable_checkpoint(sbi);
3968	}
3969
3970	/*
3971	 * If filesystem is not mounted as read-only then
3972	 * do start the gc_thread.
3973	 */
3974	if ((F2FS_OPTION(sbi).bggc_mode != BGGC_MODE_OFF ||
3975		test_opt(sbi, GC_MERGE)) && !f2fs_readonly(sb)) {
3976		/* After POR, we can run background GC thread.*/
3977		err = f2fs_start_gc_thread(sbi);
3978		if (err)
3979			goto sync_free_meta;
3980	}
3981	kvfree(options);
3982
3983	/* recover broken superblock */
3984	if (recovery) {
3985		err = f2fs_commit_super(sbi, true);
3986		f2fs_info(sbi, "Try to recover %dth superblock, ret: %d",
3987			  sbi->valid_super_block ? 1 : 2, err);
3988	}
3989
3990	f2fs_join_shrinker(sbi);
3991
3992	f2fs_tuning_parameters(sbi);
3993
3994	f2fs_notice(sbi, "Mounted with checkpoint version = %llx",
3995		    cur_cp_version(F2FS_CKPT(sbi)));
3996	f2fs_update_time(sbi, CP_TIME);
3997	f2fs_update_time(sbi, REQ_TIME);
3998	clear_sbi_flag(sbi, SBI_CP_DISABLED_QUICK);
3999	return 0;
4000
4001sync_free_meta:
4002	/* safe to flush all the data */
4003	sync_filesystem(sbi->sb);
4004	retry_cnt = 0;
4005
4006free_meta:
4007#ifdef CONFIG_QUOTA
4008	f2fs_truncate_quota_inode_pages(sb);
4009	if (f2fs_sb_has_quota_ino(sbi) && !f2fs_readonly(sb))
4010		f2fs_quota_off_umount(sbi->sb);
4011#endif
4012	/*
4013	 * Some dirty meta pages can be produced by f2fs_recover_orphan_inodes()
4014	 * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg()
4015	 * followed by f2fs_write_checkpoint() through f2fs_write_node_pages(), which
4016	 * falls into an infinite loop in f2fs_sync_meta_pages().
4017	 */
4018	truncate_inode_pages_final(META_MAPPING(sbi));
4019	/* evict some inodes being cached by GC */
4020	evict_inodes(sb);
4021	f2fs_unregister_sysfs(sbi);
4022free_root_inode:
4023	dput(sb->s_root);
4024	sb->s_root = NULL;
4025free_node_inode:
4026	f2fs_release_ino_entry(sbi, true);
4027	truncate_inode_pages_final(NODE_MAPPING(sbi));
4028	iput(sbi->node_inode);
4029	sbi->node_inode = NULL;
4030free_stats:
4031	f2fs_destroy_stats(sbi);
4032free_nm:
4033	f2fs_destroy_node_manager(sbi);
4034free_sm:
4035	f2fs_destroy_segment_manager(sbi);
4036	f2fs_destroy_post_read_wq(sbi);
4037free_devices:
4038	destroy_device_list(sbi);
4039	kvfree(sbi->ckpt);
4040free_meta_inode:
4041	make_bad_inode(sbi->meta_inode);
4042	iput(sbi->meta_inode);
4043	sbi->meta_inode = NULL;
4044free_page_array_cache:
4045	f2fs_destroy_page_array_cache(sbi);
4046free_xattr_cache:
4047	f2fs_destroy_xattr_caches(sbi);
4048free_io_dummy:
4049	mempool_destroy(sbi->write_io_dummy);
4050free_percpu:
4051	destroy_percpu_info(sbi);
4052free_bio_info:
4053	for (i = 0; i < NR_PAGE_TYPE; i++)
4054		kvfree(sbi->write_io[i]);
4055
4056#ifdef CONFIG_UNICODE
4057	utf8_unload(sb->s_encoding);
4058	sb->s_encoding = NULL;
4059#endif
4060free_options:
4061#ifdef CONFIG_QUOTA
4062	for (i = 0; i < MAXQUOTAS; i++)
4063		kfree(F2FS_OPTION(sbi).s_qf_names[i]);
4064#endif
4065	fscrypt_free_dummy_policy(&F2FS_OPTION(sbi).dummy_enc_policy);
4066	kvfree(options);
4067free_sb_buf:
4068	kfree(raw_super);
4069free_sbi:
4070	if (sbi->s_chksum_driver)
4071		crypto_free_shash(sbi->s_chksum_driver);
4072	kfree(sbi);
4073
4074	/* give only one another chance */
4075	if (retry_cnt > 0 && skip_recovery) {
4076		retry_cnt--;
4077		shrink_dcache_sb(sb);
4078		goto try_onemore;
4079	}
4080	return err;
4081}
4082
4083static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags,
4084			const char *dev_name, void *data)
4085{
4086	return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super);
4087}
4088
4089static void kill_f2fs_super(struct super_block *sb)
4090{
4091	if (sb->s_root) {
4092		struct f2fs_sb_info *sbi = F2FS_SB(sb);
4093
4094		set_sbi_flag(sbi, SBI_IS_CLOSE);
4095		f2fs_stop_gc_thread(sbi);
4096		f2fs_stop_discard_thread(sbi);
4097
4098		if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) ||
4099				!is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) {
4100			struct cp_control cpc = {
4101				.reason = CP_UMOUNT,
4102			};
4103			f2fs_write_checkpoint(sbi, &cpc);
4104		}
4105
4106		if (is_sbi_flag_set(sbi, SBI_IS_RECOVERED) && f2fs_readonly(sb))
4107			sb->s_flags &= ~SB_RDONLY;
4108	}
4109	kill_block_super(sb);
4110}
4111
4112static struct file_system_type f2fs_fs_type = {
4113	.owner		= THIS_MODULE,
4114	.name		= "f2fs",
4115	.mount		= f2fs_mount,
4116	.kill_sb	= kill_f2fs_super,
4117	.fs_flags	= FS_REQUIRES_DEV,
4118};
4119MODULE_ALIAS_FS("f2fs");
4120
4121static int __init init_inodecache(void)
4122{
4123	f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache",
4124			sizeof(struct f2fs_inode_info), 0,
4125			SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL);
4126	if (!f2fs_inode_cachep)
4127		return -ENOMEM;
4128	return 0;
4129}
4130
4131static void destroy_inodecache(void)
4132{
4133	/*
4134	 * Make sure all delayed rcu free inodes are flushed before we
4135	 * destroy cache.
4136	 */
4137	rcu_barrier();
4138	kmem_cache_destroy(f2fs_inode_cachep);
4139}
4140
4141static int __init init_f2fs_fs(void)
4142{
4143	int err;
4144
4145	if (PAGE_SIZE != F2FS_BLKSIZE) {
4146		printk("F2FS not supported on PAGE_SIZE(%lu) != %d\n",
4147				PAGE_SIZE, F2FS_BLKSIZE);
4148		return -EINVAL;
4149	}
4150
4151	f2fs_build_trace_ios();
4152
4153	err = init_inodecache();
4154	if (err)
4155		goto fail;
4156	err = f2fs_create_node_manager_caches();
4157	if (err)
4158		goto free_inodecache;
4159	err = f2fs_create_segment_manager_caches();
4160	if (err)
4161		goto free_node_manager_caches;
4162	err = f2fs_create_checkpoint_caches();
4163	if (err)
4164		goto free_segment_manager_caches;
4165	err = f2fs_create_recovery_cache();
4166	if (err)
4167		goto free_checkpoint_caches;
4168	err = f2fs_create_extent_cache();
4169	if (err)
4170		goto free_recovery_cache;
4171	err = f2fs_create_garbage_collection_cache();
4172	if (err)
4173		goto free_extent_cache;
4174	err = f2fs_init_sysfs();
4175	if (err)
4176		goto free_garbage_collection_cache;
4177	err = register_shrinker(&f2fs_shrinker_info);
4178	if (err)
4179		goto free_sysfs;
4180	err = register_filesystem(&f2fs_fs_type);
4181	if (err)
4182		goto free_shrinker;
4183	f2fs_create_root_stats();
4184	err = f2fs_init_post_read_processing();
4185	if (err)
4186		goto free_root_stats;
4187	err = f2fs_init_bio_entry_cache();
4188	if (err)
4189		goto free_post_read;
4190	err = f2fs_init_bioset();
4191	if (err)
4192		goto free_bio_enrty_cache;
4193	err = f2fs_init_compress_mempool();
4194	if (err)
4195		goto free_bioset;
4196	err = f2fs_init_compress_cache();
4197	if (err)
4198		goto free_compress_mempool;
4199	return 0;
4200free_compress_mempool:
4201	f2fs_destroy_compress_mempool();
4202free_bioset:
4203	f2fs_destroy_bioset();
4204free_bio_enrty_cache:
4205	f2fs_destroy_bio_entry_cache();
4206free_post_read:
4207	f2fs_destroy_post_read_processing();
4208free_root_stats:
4209	f2fs_destroy_root_stats();
4210	unregister_filesystem(&f2fs_fs_type);
4211free_shrinker:
4212	unregister_shrinker(&f2fs_shrinker_info);
4213free_sysfs:
4214	f2fs_exit_sysfs();
4215free_garbage_collection_cache:
4216	f2fs_destroy_garbage_collection_cache();
4217free_extent_cache:
4218	f2fs_destroy_extent_cache();
4219free_recovery_cache:
4220	f2fs_destroy_recovery_cache();
4221free_checkpoint_caches:
4222	f2fs_destroy_checkpoint_caches();
4223free_segment_manager_caches:
4224	f2fs_destroy_segment_manager_caches();
4225free_node_manager_caches:
4226	f2fs_destroy_node_manager_caches();
4227free_inodecache:
4228	destroy_inodecache();
4229fail:
4230	return err;
4231}
4232
4233static void __exit exit_f2fs_fs(void)
4234{
4235	f2fs_destroy_compress_cache();
4236	f2fs_destroy_compress_mempool();
4237	f2fs_destroy_bioset();
4238	f2fs_destroy_bio_entry_cache();
4239	f2fs_destroy_post_read_processing();
4240	f2fs_destroy_root_stats();
4241	unregister_filesystem(&f2fs_fs_type);
4242	unregister_shrinker(&f2fs_shrinker_info);
4243	f2fs_exit_sysfs();
4244	f2fs_destroy_garbage_collection_cache();
4245	f2fs_destroy_extent_cache();
4246	f2fs_destroy_recovery_cache();
4247	f2fs_destroy_checkpoint_caches();
4248	f2fs_destroy_segment_manager_caches();
4249	f2fs_destroy_node_manager_caches();
4250	destroy_inodecache();
4251	f2fs_destroy_trace_ios();
4252}
4253
4254module_init(init_f2fs_fs)
4255module_exit(exit_f2fs_fs)
4256
4257MODULE_AUTHOR("Samsung Electronics's Praesto Team");
4258MODULE_DESCRIPTION("Flash Friendly File System");
4259MODULE_LICENSE("GPL");
4260MODULE_SOFTDEP("pre: crc32");
4261
4262