]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * fs/f2fs/super.c | |
3 | * | |
4 | * Copyright (c) 2012 Samsung Electronics Co., Ltd. | |
5 | * http://www.samsung.com/ | |
6 | * | |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License version 2 as | |
9 | * published by the Free Software Foundation. | |
10 | */ | |
11 | #include <linux/module.h> | |
12 | #include <linux/init.h> | |
13 | #include <linux/fs.h> | |
14 | #include <linux/statfs.h> | |
15 | #include <linux/buffer_head.h> | |
16 | #include <linux/backing-dev.h> | |
17 | #include <linux/kthread.h> | |
18 | #include <linux/parser.h> | |
19 | #include <linux/mount.h> | |
20 | #include <linux/seq_file.h> | |
21 | #include <linux/proc_fs.h> | |
22 | #include <linux/random.h> | |
23 | #include <linux/exportfs.h> | |
24 | #include <linux/blkdev.h> | |
25 | #include <linux/f2fs_fs.h> | |
26 | #include <linux/sysfs.h> | |
27 | ||
28 | #include "f2fs.h" | |
29 | #include "node.h" | |
30 | #include "segment.h" | |
31 | #include "xattr.h" | |
32 | #include "gc.h" | |
33 | #include "trace.h" | |
34 | ||
35 | #define CREATE_TRACE_POINTS | |
36 | #include <trace/events/f2fs.h> | |
37 | ||
38 | static struct proc_dir_entry *f2fs_proc_root; | |
39 | static struct kmem_cache *f2fs_inode_cachep; | |
40 | static struct kset *f2fs_kset; | |
41 | ||
42 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
43 | ||
44 | char *fault_name[FAULT_MAX] = { | |
45 | [FAULT_KMALLOC] = "kmalloc", | |
46 | [FAULT_PAGE_ALLOC] = "page alloc", | |
47 | [FAULT_ALLOC_NID] = "alloc nid", | |
48 | [FAULT_ORPHAN] = "orphan", | |
49 | [FAULT_BLOCK] = "no more block", | |
50 | [FAULT_DIR_DEPTH] = "too big dir depth", | |
51 | [FAULT_EVICT_INODE] = "evict_inode fail", | |
52 | [FAULT_TRUNCATE] = "truncate fail", | |
53 | [FAULT_IO] = "IO error", | |
54 | [FAULT_CHECKPOINT] = "checkpoint error", | |
55 | }; | |
56 | ||
57 | static void f2fs_build_fault_attr(struct f2fs_sb_info *sbi, | |
58 | unsigned int rate) | |
59 | { | |
60 | struct f2fs_fault_info *ffi = &sbi->fault_info; | |
61 | ||
62 | if (rate) { | |
63 | atomic_set(&ffi->inject_ops, 0); | |
64 | ffi->inject_rate = rate; | |
65 | ffi->inject_type = (1 << FAULT_MAX) - 1; | |
66 | } else { | |
67 | memset(ffi, 0, sizeof(struct f2fs_fault_info)); | |
68 | } | |
69 | } | |
70 | #endif | |
71 | ||
72 | /* f2fs-wide shrinker description */ | |
73 | static struct shrinker f2fs_shrinker_info = { | |
74 | .scan_objects = f2fs_shrink_scan, | |
75 | .count_objects = f2fs_shrink_count, | |
76 | .seeks = DEFAULT_SEEKS, | |
77 | }; | |
78 | ||
79 | enum { | |
80 | Opt_gc_background, | |
81 | Opt_disable_roll_forward, | |
82 | Opt_norecovery, | |
83 | Opt_discard, | |
84 | Opt_nodiscard, | |
85 | Opt_noheap, | |
86 | Opt_heap, | |
87 | Opt_user_xattr, | |
88 | Opt_nouser_xattr, | |
89 | Opt_acl, | |
90 | Opt_noacl, | |
91 | Opt_active_logs, | |
92 | Opt_disable_ext_identify, | |
93 | Opt_inline_xattr, | |
94 | Opt_noinline_xattr, | |
95 | Opt_inline_data, | |
96 | Opt_inline_dentry, | |
97 | Opt_noinline_dentry, | |
98 | Opt_flush_merge, | |
99 | Opt_noflush_merge, | |
100 | Opt_nobarrier, | |
101 | Opt_fastboot, | |
102 | Opt_extent_cache, | |
103 | Opt_noextent_cache, | |
104 | Opt_noinline_data, | |
105 | Opt_data_flush, | |
106 | Opt_mode, | |
107 | Opt_io_size_bits, | |
108 | Opt_fault_injection, | |
109 | Opt_lazytime, | |
110 | Opt_nolazytime, | |
111 | Opt_err, | |
112 | }; | |
113 | ||
114 | static match_table_t f2fs_tokens = { | |
115 | {Opt_gc_background, "background_gc=%s"}, | |
116 | {Opt_disable_roll_forward, "disable_roll_forward"}, | |
117 | {Opt_norecovery, "norecovery"}, | |
118 | {Opt_discard, "discard"}, | |
119 | {Opt_nodiscard, "nodiscard"}, | |
120 | {Opt_noheap, "no_heap"}, | |
121 | {Opt_heap, "heap"}, | |
122 | {Opt_user_xattr, "user_xattr"}, | |
123 | {Opt_nouser_xattr, "nouser_xattr"}, | |
124 | {Opt_acl, "acl"}, | |
125 | {Opt_noacl, "noacl"}, | |
126 | {Opt_active_logs, "active_logs=%u"}, | |
127 | {Opt_disable_ext_identify, "disable_ext_identify"}, | |
128 | {Opt_inline_xattr, "inline_xattr"}, | |
129 | {Opt_noinline_xattr, "noinline_xattr"}, | |
130 | {Opt_inline_data, "inline_data"}, | |
131 | {Opt_inline_dentry, "inline_dentry"}, | |
132 | {Opt_noinline_dentry, "noinline_dentry"}, | |
133 | {Opt_flush_merge, "flush_merge"}, | |
134 | {Opt_noflush_merge, "noflush_merge"}, | |
135 | {Opt_nobarrier, "nobarrier"}, | |
136 | {Opt_fastboot, "fastboot"}, | |
137 | {Opt_extent_cache, "extent_cache"}, | |
138 | {Opt_noextent_cache, "noextent_cache"}, | |
139 | {Opt_noinline_data, "noinline_data"}, | |
140 | {Opt_data_flush, "data_flush"}, | |
141 | {Opt_mode, "mode=%s"}, | |
142 | {Opt_io_size_bits, "io_bits=%u"}, | |
143 | {Opt_fault_injection, "fault_injection=%u"}, | |
144 | {Opt_lazytime, "lazytime"}, | |
145 | {Opt_nolazytime, "nolazytime"}, | |
146 | {Opt_err, NULL}, | |
147 | }; | |
148 | ||
149 | /* Sysfs support for f2fs */ | |
150 | enum { | |
151 | GC_THREAD, /* struct f2fs_gc_thread */ | |
152 | SM_INFO, /* struct f2fs_sm_info */ | |
153 | DCC_INFO, /* struct discard_cmd_control */ | |
154 | NM_INFO, /* struct f2fs_nm_info */ | |
155 | F2FS_SBI, /* struct f2fs_sb_info */ | |
156 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
157 | FAULT_INFO_RATE, /* struct f2fs_fault_info */ | |
158 | FAULT_INFO_TYPE, /* struct f2fs_fault_info */ | |
159 | #endif | |
160 | }; | |
161 | ||
162 | struct f2fs_attr { | |
163 | struct attribute attr; | |
164 | ssize_t (*show)(struct f2fs_attr *, struct f2fs_sb_info *, char *); | |
165 | ssize_t (*store)(struct f2fs_attr *, struct f2fs_sb_info *, | |
166 | const char *, size_t); | |
167 | int struct_type; | |
168 | int offset; | |
169 | }; | |
170 | ||
171 | static unsigned char *__struct_ptr(struct f2fs_sb_info *sbi, int struct_type) | |
172 | { | |
173 | if (struct_type == GC_THREAD) | |
174 | return (unsigned char *)sbi->gc_thread; | |
175 | else if (struct_type == SM_INFO) | |
176 | return (unsigned char *)SM_I(sbi); | |
177 | else if (struct_type == DCC_INFO) | |
178 | return (unsigned char *)SM_I(sbi)->dcc_info; | |
179 | else if (struct_type == NM_INFO) | |
180 | return (unsigned char *)NM_I(sbi); | |
181 | else if (struct_type == F2FS_SBI) | |
182 | return (unsigned char *)sbi; | |
183 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
184 | else if (struct_type == FAULT_INFO_RATE || | |
185 | struct_type == FAULT_INFO_TYPE) | |
186 | return (unsigned char *)&sbi->fault_info; | |
187 | #endif | |
188 | return NULL; | |
189 | } | |
190 | ||
191 | static ssize_t lifetime_write_kbytes_show(struct f2fs_attr *a, | |
192 | struct f2fs_sb_info *sbi, char *buf) | |
193 | { | |
194 | struct super_block *sb = sbi->sb; | |
195 | ||
196 | if (!sb->s_bdev->bd_part) | |
197 | return snprintf(buf, PAGE_SIZE, "0\n"); | |
198 | ||
199 | return snprintf(buf, PAGE_SIZE, "%llu\n", | |
200 | (unsigned long long)(sbi->kbytes_written + | |
201 | BD_PART_WRITTEN(sbi))); | |
202 | } | |
203 | ||
204 | static ssize_t f2fs_sbi_show(struct f2fs_attr *a, | |
205 | struct f2fs_sb_info *sbi, char *buf) | |
206 | { | |
207 | unsigned char *ptr = NULL; | |
208 | unsigned int *ui; | |
209 | ||
210 | ptr = __struct_ptr(sbi, a->struct_type); | |
211 | if (!ptr) | |
212 | return -EINVAL; | |
213 | ||
214 | ui = (unsigned int *)(ptr + a->offset); | |
215 | ||
216 | return snprintf(buf, PAGE_SIZE, "%u\n", *ui); | |
217 | } | |
218 | ||
219 | static ssize_t f2fs_sbi_store(struct f2fs_attr *a, | |
220 | struct f2fs_sb_info *sbi, | |
221 | const char *buf, size_t count) | |
222 | { | |
223 | unsigned char *ptr; | |
224 | unsigned long t; | |
225 | unsigned int *ui; | |
226 | ssize_t ret; | |
227 | ||
228 | ptr = __struct_ptr(sbi, a->struct_type); | |
229 | if (!ptr) | |
230 | return -EINVAL; | |
231 | ||
232 | ui = (unsigned int *)(ptr + a->offset); | |
233 | ||
234 | ret = kstrtoul(skip_spaces(buf), 0, &t); | |
235 | if (ret < 0) | |
236 | return ret; | |
237 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
238 | if (a->struct_type == FAULT_INFO_TYPE && t >= (1 << FAULT_MAX)) | |
239 | return -EINVAL; | |
240 | #endif | |
241 | *ui = t; | |
242 | return count; | |
243 | } | |
244 | ||
245 | static ssize_t f2fs_attr_show(struct kobject *kobj, | |
246 | struct attribute *attr, char *buf) | |
247 | { | |
248 | struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, | |
249 | s_kobj); | |
250 | struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); | |
251 | ||
252 | return a->show ? a->show(a, sbi, buf) : 0; | |
253 | } | |
254 | ||
255 | static ssize_t f2fs_attr_store(struct kobject *kobj, struct attribute *attr, | |
256 | const char *buf, size_t len) | |
257 | { | |
258 | struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, | |
259 | s_kobj); | |
260 | struct f2fs_attr *a = container_of(attr, struct f2fs_attr, attr); | |
261 | ||
262 | return a->store ? a->store(a, sbi, buf, len) : 0; | |
263 | } | |
264 | ||
265 | static void f2fs_sb_release(struct kobject *kobj) | |
266 | { | |
267 | struct f2fs_sb_info *sbi = container_of(kobj, struct f2fs_sb_info, | |
268 | s_kobj); | |
269 | complete(&sbi->s_kobj_unregister); | |
270 | } | |
271 | ||
272 | #define F2FS_ATTR_OFFSET(_struct_type, _name, _mode, _show, _store, _offset) \ | |
273 | static struct f2fs_attr f2fs_attr_##_name = { \ | |
274 | .attr = {.name = __stringify(_name), .mode = _mode }, \ | |
275 | .show = _show, \ | |
276 | .store = _store, \ | |
277 | .struct_type = _struct_type, \ | |
278 | .offset = _offset \ | |
279 | } | |
280 | ||
281 | #define F2FS_RW_ATTR(struct_type, struct_name, name, elname) \ | |
282 | F2FS_ATTR_OFFSET(struct_type, name, 0644, \ | |
283 | f2fs_sbi_show, f2fs_sbi_store, \ | |
284 | offsetof(struct struct_name, elname)) | |
285 | ||
286 | #define F2FS_GENERAL_RO_ATTR(name) \ | |
287 | static struct f2fs_attr f2fs_attr_##name = __ATTR(name, 0444, name##_show, NULL) | |
288 | ||
289 | F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_min_sleep_time, min_sleep_time); | |
290 | F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_max_sleep_time, max_sleep_time); | |
291 | F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time, no_gc_sleep_time); | |
292 | F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle); | |
293 | F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments); | |
294 | F2FS_RW_ATTR(DCC_INFO, discard_cmd_control, max_small_discards, max_discards); | |
295 | F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections); | |
296 | F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy); | |
297 | F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util); | |
298 | F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks); | |
299 | F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_hot_blocks, min_hot_blocks); | |
300 | F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ram_thresh, ram_thresh); | |
301 | F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, ra_nid_pages, ra_nid_pages); | |
302 | F2FS_RW_ATTR(NM_INFO, f2fs_nm_info, dirty_nats_ratio, dirty_nats_ratio); | |
303 | F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_victim_search, max_victim_search); | |
304 | F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, dir_level, dir_level); | |
305 | F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, cp_interval, interval_time[CP_TIME]); | |
306 | F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, idle_interval, interval_time[REQ_TIME]); | |
307 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
308 | F2FS_RW_ATTR(FAULT_INFO_RATE, f2fs_fault_info, inject_rate, inject_rate); | |
309 | F2FS_RW_ATTR(FAULT_INFO_TYPE, f2fs_fault_info, inject_type, inject_type); | |
310 | #endif | |
311 | F2FS_GENERAL_RO_ATTR(lifetime_write_kbytes); | |
312 | ||
313 | #define ATTR_LIST(name) (&f2fs_attr_##name.attr) | |
314 | static struct attribute *f2fs_attrs[] = { | |
315 | ATTR_LIST(gc_min_sleep_time), | |
316 | ATTR_LIST(gc_max_sleep_time), | |
317 | ATTR_LIST(gc_no_gc_sleep_time), | |
318 | ATTR_LIST(gc_idle), | |
319 | ATTR_LIST(reclaim_segments), | |
320 | ATTR_LIST(max_small_discards), | |
321 | ATTR_LIST(batched_trim_sections), | |
322 | ATTR_LIST(ipu_policy), | |
323 | ATTR_LIST(min_ipu_util), | |
324 | ATTR_LIST(min_fsync_blocks), | |
325 | ATTR_LIST(min_hot_blocks), | |
326 | ATTR_LIST(max_victim_search), | |
327 | ATTR_LIST(dir_level), | |
328 | ATTR_LIST(ram_thresh), | |
329 | ATTR_LIST(ra_nid_pages), | |
330 | ATTR_LIST(dirty_nats_ratio), | |
331 | ATTR_LIST(cp_interval), | |
332 | ATTR_LIST(idle_interval), | |
333 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
334 | ATTR_LIST(inject_rate), | |
335 | ATTR_LIST(inject_type), | |
336 | #endif | |
337 | ATTR_LIST(lifetime_write_kbytes), | |
338 | NULL, | |
339 | }; | |
340 | ||
341 | static const struct sysfs_ops f2fs_attr_ops = { | |
342 | .show = f2fs_attr_show, | |
343 | .store = f2fs_attr_store, | |
344 | }; | |
345 | ||
346 | static struct kobj_type f2fs_ktype = { | |
347 | .default_attrs = f2fs_attrs, | |
348 | .sysfs_ops = &f2fs_attr_ops, | |
349 | .release = f2fs_sb_release, | |
350 | }; | |
351 | ||
352 | void f2fs_msg(struct super_block *sb, const char *level, const char *fmt, ...) | |
353 | { | |
354 | struct va_format vaf; | |
355 | va_list args; | |
356 | ||
357 | va_start(args, fmt); | |
358 | vaf.fmt = fmt; | |
359 | vaf.va = &args; | |
360 | printk("%sF2FS-fs (%s): %pV\n", level, sb->s_id, &vaf); | |
361 | va_end(args); | |
362 | } | |
363 | ||
364 | static void init_once(void *foo) | |
365 | { | |
366 | struct f2fs_inode_info *fi = (struct f2fs_inode_info *) foo; | |
367 | ||
368 | inode_init_once(&fi->vfs_inode); | |
369 | } | |
370 | ||
371 | static int parse_options(struct super_block *sb, char *options) | |
372 | { | |
373 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
374 | struct request_queue *q; | |
375 | substring_t args[MAX_OPT_ARGS]; | |
376 | char *p, *name; | |
377 | int arg = 0; | |
378 | ||
379 | if (!options) | |
380 | return 0; | |
381 | ||
382 | while ((p = strsep(&options, ",")) != NULL) { | |
383 | int token; | |
384 | if (!*p) | |
385 | continue; | |
386 | /* | |
387 | * Initialize args struct so we know whether arg was | |
388 | * found; some options take optional arguments. | |
389 | */ | |
390 | args[0].to = args[0].from = NULL; | |
391 | token = match_token(p, f2fs_tokens, args); | |
392 | ||
393 | switch (token) { | |
394 | case Opt_gc_background: | |
395 | name = match_strdup(&args[0]); | |
396 | ||
397 | if (!name) | |
398 | return -ENOMEM; | |
399 | if (strlen(name) == 2 && !strncmp(name, "on", 2)) { | |
400 | set_opt(sbi, BG_GC); | |
401 | clear_opt(sbi, FORCE_FG_GC); | |
402 | } else if (strlen(name) == 3 && !strncmp(name, "off", 3)) { | |
403 | clear_opt(sbi, BG_GC); | |
404 | clear_opt(sbi, FORCE_FG_GC); | |
405 | } else if (strlen(name) == 4 && !strncmp(name, "sync", 4)) { | |
406 | set_opt(sbi, BG_GC); | |
407 | set_opt(sbi, FORCE_FG_GC); | |
408 | } else { | |
409 | kfree(name); | |
410 | return -EINVAL; | |
411 | } | |
412 | kfree(name); | |
413 | break; | |
414 | case Opt_disable_roll_forward: | |
415 | set_opt(sbi, DISABLE_ROLL_FORWARD); | |
416 | break; | |
417 | case Opt_norecovery: | |
418 | /* this option mounts f2fs with ro */ | |
419 | set_opt(sbi, DISABLE_ROLL_FORWARD); | |
420 | if (!f2fs_readonly(sb)) | |
421 | return -EINVAL; | |
422 | break; | |
423 | case Opt_discard: | |
424 | q = bdev_get_queue(sb->s_bdev); | |
425 | if (blk_queue_discard(q)) { | |
426 | set_opt(sbi, DISCARD); | |
427 | } else if (!f2fs_sb_mounted_blkzoned(sb)) { | |
428 | f2fs_msg(sb, KERN_WARNING, | |
429 | "mounting with \"discard\" option, but " | |
430 | "the device does not support discard"); | |
431 | } | |
432 | break; | |
433 | case Opt_nodiscard: | |
434 | if (f2fs_sb_mounted_blkzoned(sb)) { | |
435 | f2fs_msg(sb, KERN_WARNING, | |
436 | "discard is required for zoned block devices"); | |
437 | return -EINVAL; | |
438 | } | |
439 | clear_opt(sbi, DISCARD); | |
440 | break; | |
441 | case Opt_noheap: | |
442 | set_opt(sbi, NOHEAP); | |
443 | break; | |
444 | case Opt_heap: | |
445 | clear_opt(sbi, NOHEAP); | |
446 | break; | |
447 | #ifdef CONFIG_F2FS_FS_XATTR | |
448 | case Opt_user_xattr: | |
449 | set_opt(sbi, XATTR_USER); | |
450 | break; | |
451 | case Opt_nouser_xattr: | |
452 | clear_opt(sbi, XATTR_USER); | |
453 | break; | |
454 | case Opt_inline_xattr: | |
455 | set_opt(sbi, INLINE_XATTR); | |
456 | break; | |
457 | case Opt_noinline_xattr: | |
458 | clear_opt(sbi, INLINE_XATTR); | |
459 | break; | |
460 | #else | |
461 | case Opt_user_xattr: | |
462 | f2fs_msg(sb, KERN_INFO, | |
463 | "user_xattr options not supported"); | |
464 | break; | |
465 | case Opt_nouser_xattr: | |
466 | f2fs_msg(sb, KERN_INFO, | |
467 | "nouser_xattr options not supported"); | |
468 | break; | |
469 | case Opt_inline_xattr: | |
470 | f2fs_msg(sb, KERN_INFO, | |
471 | "inline_xattr options not supported"); | |
472 | break; | |
473 | case Opt_noinline_xattr: | |
474 | f2fs_msg(sb, KERN_INFO, | |
475 | "noinline_xattr options not supported"); | |
476 | break; | |
477 | #endif | |
478 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
479 | case Opt_acl: | |
480 | set_opt(sbi, POSIX_ACL); | |
481 | break; | |
482 | case Opt_noacl: | |
483 | clear_opt(sbi, POSIX_ACL); | |
484 | break; | |
485 | #else | |
486 | case Opt_acl: | |
487 | f2fs_msg(sb, KERN_INFO, "acl options not supported"); | |
488 | break; | |
489 | case Opt_noacl: | |
490 | f2fs_msg(sb, KERN_INFO, "noacl options not supported"); | |
491 | break; | |
492 | #endif | |
493 | case Opt_active_logs: | |
494 | if (args->from && match_int(args, &arg)) | |
495 | return -EINVAL; | |
496 | if (arg != 2 && arg != 4 && arg != NR_CURSEG_TYPE) | |
497 | return -EINVAL; | |
498 | sbi->active_logs = arg; | |
499 | break; | |
500 | case Opt_disable_ext_identify: | |
501 | set_opt(sbi, DISABLE_EXT_IDENTIFY); | |
502 | break; | |
503 | case Opt_inline_data: | |
504 | set_opt(sbi, INLINE_DATA); | |
505 | break; | |
506 | case Opt_inline_dentry: | |
507 | set_opt(sbi, INLINE_DENTRY); | |
508 | break; | |
509 | case Opt_noinline_dentry: | |
510 | clear_opt(sbi, INLINE_DENTRY); | |
511 | break; | |
512 | case Opt_flush_merge: | |
513 | set_opt(sbi, FLUSH_MERGE); | |
514 | break; | |
515 | case Opt_noflush_merge: | |
516 | clear_opt(sbi, FLUSH_MERGE); | |
517 | break; | |
518 | case Opt_nobarrier: | |
519 | set_opt(sbi, NOBARRIER); | |
520 | break; | |
521 | case Opt_fastboot: | |
522 | set_opt(sbi, FASTBOOT); | |
523 | break; | |
524 | case Opt_extent_cache: | |
525 | set_opt(sbi, EXTENT_CACHE); | |
526 | break; | |
527 | case Opt_noextent_cache: | |
528 | clear_opt(sbi, EXTENT_CACHE); | |
529 | break; | |
530 | case Opt_noinline_data: | |
531 | clear_opt(sbi, INLINE_DATA); | |
532 | break; | |
533 | case Opt_data_flush: | |
534 | set_opt(sbi, DATA_FLUSH); | |
535 | break; | |
536 | case Opt_mode: | |
537 | name = match_strdup(&args[0]); | |
538 | ||
539 | if (!name) | |
540 | return -ENOMEM; | |
541 | if (strlen(name) == 8 && | |
542 | !strncmp(name, "adaptive", 8)) { | |
543 | if (f2fs_sb_mounted_blkzoned(sb)) { | |
544 | f2fs_msg(sb, KERN_WARNING, | |
545 | "adaptive mode is not allowed with " | |
546 | "zoned block device feature"); | |
547 | kfree(name); | |
548 | return -EINVAL; | |
549 | } | |
550 | set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); | |
551 | } else if (strlen(name) == 3 && | |
552 | !strncmp(name, "lfs", 3)) { | |
553 | set_opt_mode(sbi, F2FS_MOUNT_LFS); | |
554 | } else { | |
555 | kfree(name); | |
556 | return -EINVAL; | |
557 | } | |
558 | kfree(name); | |
559 | break; | |
560 | case Opt_io_size_bits: | |
561 | if (args->from && match_int(args, &arg)) | |
562 | return -EINVAL; | |
563 | if (arg > __ilog2_u32(BIO_MAX_PAGES)) { | |
564 | f2fs_msg(sb, KERN_WARNING, | |
565 | "Not support %d, larger than %d", | |
566 | 1 << arg, BIO_MAX_PAGES); | |
567 | return -EINVAL; | |
568 | } | |
569 | sbi->write_io_size_bits = arg; | |
570 | break; | |
571 | case Opt_fault_injection: | |
572 | if (args->from && match_int(args, &arg)) | |
573 | return -EINVAL; | |
574 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
575 | f2fs_build_fault_attr(sbi, arg); | |
576 | set_opt(sbi, FAULT_INJECTION); | |
577 | #else | |
578 | f2fs_msg(sb, KERN_INFO, | |
579 | "FAULT_INJECTION was not selected"); | |
580 | #endif | |
581 | break; | |
582 | case Opt_lazytime: | |
583 | sb->s_flags |= MS_LAZYTIME; | |
584 | break; | |
585 | case Opt_nolazytime: | |
586 | sb->s_flags &= ~MS_LAZYTIME; | |
587 | break; | |
588 | default: | |
589 | f2fs_msg(sb, KERN_ERR, | |
590 | "Unrecognized mount option \"%s\" or missing value", | |
591 | p); | |
592 | return -EINVAL; | |
593 | } | |
594 | } | |
595 | ||
596 | if (F2FS_IO_SIZE_BITS(sbi) && !test_opt(sbi, LFS)) { | |
597 | f2fs_msg(sb, KERN_ERR, | |
598 | "Should set mode=lfs with %uKB-sized IO", | |
599 | F2FS_IO_SIZE_KB(sbi)); | |
600 | return -EINVAL; | |
601 | } | |
602 | return 0; | |
603 | } | |
604 | ||
605 | static struct inode *f2fs_alloc_inode(struct super_block *sb) | |
606 | { | |
607 | struct f2fs_inode_info *fi; | |
608 | ||
609 | fi = kmem_cache_alloc(f2fs_inode_cachep, GFP_F2FS_ZERO); | |
610 | if (!fi) | |
611 | return NULL; | |
612 | ||
613 | init_once((void *) fi); | |
614 | ||
615 | /* Initialize f2fs-specific inode info */ | |
616 | fi->vfs_inode.i_version = 1; | |
617 | atomic_set(&fi->dirty_pages, 0); | |
618 | fi->i_current_depth = 1; | |
619 | fi->i_advise = 0; | |
620 | init_rwsem(&fi->i_sem); | |
621 | INIT_LIST_HEAD(&fi->dirty_list); | |
622 | INIT_LIST_HEAD(&fi->gdirty_list); | |
623 | INIT_LIST_HEAD(&fi->inmem_pages); | |
624 | mutex_init(&fi->inmem_lock); | |
625 | init_rwsem(&fi->dio_rwsem[READ]); | |
626 | init_rwsem(&fi->dio_rwsem[WRITE]); | |
627 | ||
628 | /* Will be used by directory only */ | |
629 | fi->i_dir_level = F2FS_SB(sb)->dir_level; | |
630 | return &fi->vfs_inode; | |
631 | } | |
632 | ||
633 | static int f2fs_drop_inode(struct inode *inode) | |
634 | { | |
635 | int ret; | |
636 | /* | |
637 | * This is to avoid a deadlock condition like below. | |
638 | * writeback_single_inode(inode) | |
639 | * - f2fs_write_data_page | |
640 | * - f2fs_gc -> iput -> evict | |
641 | * - inode_wait_for_writeback(inode) | |
642 | */ | |
643 | if ((!inode_unhashed(inode) && inode->i_state & I_SYNC)) { | |
644 | if (!inode->i_nlink && !is_bad_inode(inode)) { | |
645 | /* to avoid evict_inode call simultaneously */ | |
646 | atomic_inc(&inode->i_count); | |
647 | spin_unlock(&inode->i_lock); | |
648 | ||
649 | /* some remained atomic pages should discarded */ | |
650 | if (f2fs_is_atomic_file(inode)) | |
651 | drop_inmem_pages(inode); | |
652 | ||
653 | /* should remain fi->extent_tree for writepage */ | |
654 | f2fs_destroy_extent_node(inode); | |
655 | ||
656 | sb_start_intwrite(inode->i_sb); | |
657 | f2fs_i_size_write(inode, 0); | |
658 | ||
659 | if (F2FS_HAS_BLOCKS(inode)) | |
660 | f2fs_truncate(inode); | |
661 | ||
662 | sb_end_intwrite(inode->i_sb); | |
663 | ||
664 | fscrypt_put_encryption_info(inode, NULL); | |
665 | spin_lock(&inode->i_lock); | |
666 | atomic_dec(&inode->i_count); | |
667 | } | |
668 | trace_f2fs_drop_inode(inode, 0); | |
669 | return 0; | |
670 | } | |
671 | ret = generic_drop_inode(inode); | |
672 | trace_f2fs_drop_inode(inode, ret); | |
673 | return ret; | |
674 | } | |
675 | ||
676 | int f2fs_inode_dirtied(struct inode *inode, bool sync) | |
677 | { | |
678 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
679 | int ret = 0; | |
680 | ||
681 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
682 | if (is_inode_flag_set(inode, FI_DIRTY_INODE)) { | |
683 | ret = 1; | |
684 | } else { | |
685 | set_inode_flag(inode, FI_DIRTY_INODE); | |
686 | stat_inc_dirty_inode(sbi, DIRTY_META); | |
687 | } | |
688 | if (sync && list_empty(&F2FS_I(inode)->gdirty_list)) { | |
689 | list_add_tail(&F2FS_I(inode)->gdirty_list, | |
690 | &sbi->inode_list[DIRTY_META]); | |
691 | inc_page_count(sbi, F2FS_DIRTY_IMETA); | |
692 | } | |
693 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
694 | return ret; | |
695 | } | |
696 | ||
697 | void f2fs_inode_synced(struct inode *inode) | |
698 | { | |
699 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
700 | ||
701 | spin_lock(&sbi->inode_lock[DIRTY_META]); | |
702 | if (!is_inode_flag_set(inode, FI_DIRTY_INODE)) { | |
703 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
704 | return; | |
705 | } | |
706 | if (!list_empty(&F2FS_I(inode)->gdirty_list)) { | |
707 | list_del_init(&F2FS_I(inode)->gdirty_list); | |
708 | dec_page_count(sbi, F2FS_DIRTY_IMETA); | |
709 | } | |
710 | clear_inode_flag(inode, FI_DIRTY_INODE); | |
711 | clear_inode_flag(inode, FI_AUTO_RECOVER); | |
712 | stat_dec_dirty_inode(F2FS_I_SB(inode), DIRTY_META); | |
713 | spin_unlock(&sbi->inode_lock[DIRTY_META]); | |
714 | } | |
715 | ||
716 | /* | |
717 | * f2fs_dirty_inode() is called from __mark_inode_dirty() | |
718 | * | |
719 | * We should call set_dirty_inode to write the dirty inode through write_inode. | |
720 | */ | |
721 | static void f2fs_dirty_inode(struct inode *inode, int flags) | |
722 | { | |
723 | struct f2fs_sb_info *sbi = F2FS_I_SB(inode); | |
724 | ||
725 | if (inode->i_ino == F2FS_NODE_INO(sbi) || | |
726 | inode->i_ino == F2FS_META_INO(sbi)) | |
727 | return; | |
728 | ||
729 | if (flags == I_DIRTY_TIME) | |
730 | return; | |
731 | ||
732 | if (is_inode_flag_set(inode, FI_AUTO_RECOVER)) | |
733 | clear_inode_flag(inode, FI_AUTO_RECOVER); | |
734 | ||
735 | f2fs_inode_dirtied(inode, false); | |
736 | } | |
737 | ||
738 | static void f2fs_i_callback(struct rcu_head *head) | |
739 | { | |
740 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
741 | kmem_cache_free(f2fs_inode_cachep, F2FS_I(inode)); | |
742 | } | |
743 | ||
744 | static void f2fs_destroy_inode(struct inode *inode) | |
745 | { | |
746 | call_rcu(&inode->i_rcu, f2fs_i_callback); | |
747 | } | |
748 | ||
749 | static void destroy_percpu_info(struct f2fs_sb_info *sbi) | |
750 | { | |
751 | percpu_counter_destroy(&sbi->alloc_valid_block_count); | |
752 | percpu_counter_destroy(&sbi->total_valid_inode_count); | |
753 | } | |
754 | ||
755 | static void destroy_device_list(struct f2fs_sb_info *sbi) | |
756 | { | |
757 | int i; | |
758 | ||
759 | for (i = 0; i < sbi->s_ndevs; i++) { | |
760 | blkdev_put(FDEV(i).bdev, FMODE_EXCL); | |
761 | #ifdef CONFIG_BLK_DEV_ZONED | |
762 | kfree(FDEV(i).blkz_type); | |
763 | #endif | |
764 | } | |
765 | kfree(sbi->devs); | |
766 | } | |
767 | ||
768 | static void f2fs_put_super(struct super_block *sb) | |
769 | { | |
770 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
771 | int i; | |
772 | ||
773 | if (sbi->s_proc) { | |
774 | remove_proc_entry("segment_info", sbi->s_proc); | |
775 | remove_proc_entry("segment_bits", sbi->s_proc); | |
776 | remove_proc_entry(sb->s_id, f2fs_proc_root); | |
777 | } | |
778 | kobject_del(&sbi->s_kobj); | |
779 | ||
780 | stop_gc_thread(sbi); | |
781 | ||
782 | /* prevent remaining shrinker jobs */ | |
783 | mutex_lock(&sbi->umount_mutex); | |
784 | ||
785 | /* | |
786 | * We don't need to do checkpoint when superblock is clean. | |
787 | * But, the previous checkpoint was not done by umount, it needs to do | |
788 | * clean checkpoint again. | |
789 | */ | |
790 | if (is_sbi_flag_set(sbi, SBI_IS_DIRTY) || | |
791 | !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { | |
792 | struct cp_control cpc = { | |
793 | .reason = CP_UMOUNT, | |
794 | }; | |
795 | write_checkpoint(sbi, &cpc); | |
796 | } | |
797 | ||
798 | /* be sure to wait for any on-going discard commands */ | |
799 | f2fs_wait_discard_bios(sbi); | |
800 | ||
801 | if (!sbi->discard_blks) { | |
802 | struct cp_control cpc = { | |
803 | .reason = CP_UMOUNT | CP_TRIMMED, | |
804 | }; | |
805 | write_checkpoint(sbi, &cpc); | |
806 | } | |
807 | ||
808 | /* write_checkpoint can update stat informaion */ | |
809 | f2fs_destroy_stats(sbi); | |
810 | ||
811 | /* | |
812 | * normally superblock is clean, so we need to release this. | |
813 | * In addition, EIO will skip do checkpoint, we need this as well. | |
814 | */ | |
815 | release_ino_entry(sbi, true); | |
816 | ||
817 | f2fs_leave_shrinker(sbi); | |
818 | mutex_unlock(&sbi->umount_mutex); | |
819 | ||
820 | /* our cp_error case, we can wait for any writeback page */ | |
821 | f2fs_flush_merged_writes(sbi); | |
822 | ||
823 | iput(sbi->node_inode); | |
824 | iput(sbi->meta_inode); | |
825 | ||
826 | /* destroy f2fs internal modules */ | |
827 | destroy_node_manager(sbi); | |
828 | destroy_segment_manager(sbi); | |
829 | ||
830 | kfree(sbi->ckpt); | |
831 | kobject_put(&sbi->s_kobj); | |
832 | wait_for_completion(&sbi->s_kobj_unregister); | |
833 | ||
834 | sb->s_fs_info = NULL; | |
835 | if (sbi->s_chksum_driver) | |
836 | crypto_free_shash(sbi->s_chksum_driver); | |
837 | kfree(sbi->raw_super); | |
838 | ||
839 | destroy_device_list(sbi); | |
840 | mempool_destroy(sbi->write_io_dummy); | |
841 | destroy_percpu_info(sbi); | |
842 | for (i = 0; i < NR_PAGE_TYPE; i++) | |
843 | kfree(sbi->write_io[i]); | |
844 | kfree(sbi); | |
845 | } | |
846 | ||
847 | int f2fs_sync_fs(struct super_block *sb, int sync) | |
848 | { | |
849 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
850 | int err = 0; | |
851 | ||
852 | trace_f2fs_sync_fs(sb, sync); | |
853 | ||
854 | if (sync) { | |
855 | struct cp_control cpc; | |
856 | ||
857 | cpc.reason = __get_cp_reason(sbi); | |
858 | ||
859 | mutex_lock(&sbi->gc_mutex); | |
860 | err = write_checkpoint(sbi, &cpc); | |
861 | mutex_unlock(&sbi->gc_mutex); | |
862 | } | |
863 | f2fs_trace_ios(NULL, 1); | |
864 | ||
865 | return err; | |
866 | } | |
867 | ||
868 | static int f2fs_freeze(struct super_block *sb) | |
869 | { | |
870 | if (f2fs_readonly(sb)) | |
871 | return 0; | |
872 | ||
873 | /* IO error happened before */ | |
874 | if (unlikely(f2fs_cp_error(F2FS_SB(sb)))) | |
875 | return -EIO; | |
876 | ||
877 | /* must be clean, since sync_filesystem() was already called */ | |
878 | if (is_sbi_flag_set(F2FS_SB(sb), SBI_IS_DIRTY)) | |
879 | return -EINVAL; | |
880 | return 0; | |
881 | } | |
882 | ||
883 | static int f2fs_unfreeze(struct super_block *sb) | |
884 | { | |
885 | return 0; | |
886 | } | |
887 | ||
888 | static int f2fs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
889 | { | |
890 | struct super_block *sb = dentry->d_sb; | |
891 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
892 | u64 id = huge_encode_dev(sb->s_bdev->bd_dev); | |
893 | block_t total_count, user_block_count, start_count, ovp_count; | |
894 | ||
895 | total_count = le64_to_cpu(sbi->raw_super->block_count); | |
896 | user_block_count = sbi->user_block_count; | |
897 | start_count = le32_to_cpu(sbi->raw_super->segment0_blkaddr); | |
898 | ovp_count = SM_I(sbi)->ovp_segments << sbi->log_blocks_per_seg; | |
899 | buf->f_type = F2FS_SUPER_MAGIC; | |
900 | buf->f_bsize = sbi->blocksize; | |
901 | ||
902 | buf->f_blocks = total_count - start_count; | |
903 | buf->f_bfree = user_block_count - valid_user_blocks(sbi) + ovp_count; | |
904 | buf->f_bavail = user_block_count - valid_user_blocks(sbi); | |
905 | ||
906 | buf->f_files = sbi->total_node_count - F2FS_RESERVED_NODE_NUM; | |
907 | buf->f_ffree = min(buf->f_files - valid_node_count(sbi), | |
908 | buf->f_bavail); | |
909 | ||
910 | buf->f_namelen = F2FS_NAME_LEN; | |
911 | buf->f_fsid.val[0] = (u32)id; | |
912 | buf->f_fsid.val[1] = (u32)(id >> 32); | |
913 | ||
914 | return 0; | |
915 | } | |
916 | ||
917 | static int f2fs_show_options(struct seq_file *seq, struct dentry *root) | |
918 | { | |
919 | struct f2fs_sb_info *sbi = F2FS_SB(root->d_sb); | |
920 | ||
921 | if (!f2fs_readonly(sbi->sb) && test_opt(sbi, BG_GC)) { | |
922 | if (test_opt(sbi, FORCE_FG_GC)) | |
923 | seq_printf(seq, ",background_gc=%s", "sync"); | |
924 | else | |
925 | seq_printf(seq, ",background_gc=%s", "on"); | |
926 | } else { | |
927 | seq_printf(seq, ",background_gc=%s", "off"); | |
928 | } | |
929 | if (test_opt(sbi, DISABLE_ROLL_FORWARD)) | |
930 | seq_puts(seq, ",disable_roll_forward"); | |
931 | if (test_opt(sbi, DISCARD)) | |
932 | seq_puts(seq, ",discard"); | |
933 | if (test_opt(sbi, NOHEAP)) | |
934 | seq_puts(seq, ",no_heap"); | |
935 | else | |
936 | seq_puts(seq, ",heap"); | |
937 | #ifdef CONFIG_F2FS_FS_XATTR | |
938 | if (test_opt(sbi, XATTR_USER)) | |
939 | seq_puts(seq, ",user_xattr"); | |
940 | else | |
941 | seq_puts(seq, ",nouser_xattr"); | |
942 | if (test_opt(sbi, INLINE_XATTR)) | |
943 | seq_puts(seq, ",inline_xattr"); | |
944 | else | |
945 | seq_puts(seq, ",noinline_xattr"); | |
946 | #endif | |
947 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
948 | if (test_opt(sbi, POSIX_ACL)) | |
949 | seq_puts(seq, ",acl"); | |
950 | else | |
951 | seq_puts(seq, ",noacl"); | |
952 | #endif | |
953 | if (test_opt(sbi, DISABLE_EXT_IDENTIFY)) | |
954 | seq_puts(seq, ",disable_ext_identify"); | |
955 | if (test_opt(sbi, INLINE_DATA)) | |
956 | seq_puts(seq, ",inline_data"); | |
957 | else | |
958 | seq_puts(seq, ",noinline_data"); | |
959 | if (test_opt(sbi, INLINE_DENTRY)) | |
960 | seq_puts(seq, ",inline_dentry"); | |
961 | else | |
962 | seq_puts(seq, ",noinline_dentry"); | |
963 | if (!f2fs_readonly(sbi->sb) && test_opt(sbi, FLUSH_MERGE)) | |
964 | seq_puts(seq, ",flush_merge"); | |
965 | if (test_opt(sbi, NOBARRIER)) | |
966 | seq_puts(seq, ",nobarrier"); | |
967 | if (test_opt(sbi, FASTBOOT)) | |
968 | seq_puts(seq, ",fastboot"); | |
969 | if (test_opt(sbi, EXTENT_CACHE)) | |
970 | seq_puts(seq, ",extent_cache"); | |
971 | else | |
972 | seq_puts(seq, ",noextent_cache"); | |
973 | if (test_opt(sbi, DATA_FLUSH)) | |
974 | seq_puts(seq, ",data_flush"); | |
975 | ||
976 | seq_puts(seq, ",mode="); | |
977 | if (test_opt(sbi, ADAPTIVE)) | |
978 | seq_puts(seq, "adaptive"); | |
979 | else if (test_opt(sbi, LFS)) | |
980 | seq_puts(seq, "lfs"); | |
981 | seq_printf(seq, ",active_logs=%u", sbi->active_logs); | |
982 | if (F2FS_IO_SIZE_BITS(sbi)) | |
983 | seq_printf(seq, ",io_size=%uKB", F2FS_IO_SIZE_KB(sbi)); | |
984 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
985 | if (test_opt(sbi, FAULT_INJECTION)) | |
986 | seq_puts(seq, ",fault_injection"); | |
987 | #endif | |
988 | ||
989 | return 0; | |
990 | } | |
991 | ||
992 | static int segment_info_seq_show(struct seq_file *seq, void *offset) | |
993 | { | |
994 | struct super_block *sb = seq->private; | |
995 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
996 | unsigned int total_segs = | |
997 | le32_to_cpu(sbi->raw_super->segment_count_main); | |
998 | int i; | |
999 | ||
1000 | seq_puts(seq, "format: segment_type|valid_blocks\n" | |
1001 | "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n"); | |
1002 | ||
1003 | for (i = 0; i < total_segs; i++) { | |
1004 | struct seg_entry *se = get_seg_entry(sbi, i); | |
1005 | ||
1006 | if ((i % 10) == 0) | |
1007 | seq_printf(seq, "%-10d", i); | |
1008 | seq_printf(seq, "%d|%-3u", se->type, | |
1009 | get_valid_blocks(sbi, i, false)); | |
1010 | if ((i % 10) == 9 || i == (total_segs - 1)) | |
1011 | seq_putc(seq, '\n'); | |
1012 | else | |
1013 | seq_putc(seq, ' '); | |
1014 | } | |
1015 | ||
1016 | return 0; | |
1017 | } | |
1018 | ||
1019 | static int segment_bits_seq_show(struct seq_file *seq, void *offset) | |
1020 | { | |
1021 | struct super_block *sb = seq->private; | |
1022 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1023 | unsigned int total_segs = | |
1024 | le32_to_cpu(sbi->raw_super->segment_count_main); | |
1025 | int i, j; | |
1026 | ||
1027 | seq_puts(seq, "format: segment_type|valid_blocks|bitmaps\n" | |
1028 | "segment_type(0:HD, 1:WD, 2:CD, 3:HN, 4:WN, 5:CN)\n"); | |
1029 | ||
1030 | for (i = 0; i < total_segs; i++) { | |
1031 | struct seg_entry *se = get_seg_entry(sbi, i); | |
1032 | ||
1033 | seq_printf(seq, "%-10d", i); | |
1034 | seq_printf(seq, "%d|%-3u|", se->type, | |
1035 | get_valid_blocks(sbi, i, false)); | |
1036 | for (j = 0; j < SIT_VBLOCK_MAP_SIZE; j++) | |
1037 | seq_printf(seq, " %.2x", se->cur_valid_map[j]); | |
1038 | seq_putc(seq, '\n'); | |
1039 | } | |
1040 | return 0; | |
1041 | } | |
1042 | ||
1043 | #define F2FS_PROC_FILE_DEF(_name) \ | |
1044 | static int _name##_open_fs(struct inode *inode, struct file *file) \ | |
1045 | { \ | |
1046 | return single_open(file, _name##_seq_show, PDE_DATA(inode)); \ | |
1047 | } \ | |
1048 | \ | |
1049 | static const struct file_operations f2fs_seq_##_name##_fops = { \ | |
1050 | .open = _name##_open_fs, \ | |
1051 | .read = seq_read, \ | |
1052 | .llseek = seq_lseek, \ | |
1053 | .release = single_release, \ | |
1054 | }; | |
1055 | ||
1056 | F2FS_PROC_FILE_DEF(segment_info); | |
1057 | F2FS_PROC_FILE_DEF(segment_bits); | |
1058 | ||
1059 | static void default_options(struct f2fs_sb_info *sbi) | |
1060 | { | |
1061 | /* init some FS parameters */ | |
1062 | sbi->active_logs = NR_CURSEG_TYPE; | |
1063 | ||
1064 | set_opt(sbi, BG_GC); | |
1065 | set_opt(sbi, INLINE_XATTR); | |
1066 | set_opt(sbi, INLINE_DATA); | |
1067 | set_opt(sbi, INLINE_DENTRY); | |
1068 | set_opt(sbi, EXTENT_CACHE); | |
1069 | set_opt(sbi, NOHEAP); | |
1070 | sbi->sb->s_flags |= MS_LAZYTIME; | |
1071 | set_opt(sbi, FLUSH_MERGE); | |
1072 | if (f2fs_sb_mounted_blkzoned(sbi->sb)) { | |
1073 | set_opt_mode(sbi, F2FS_MOUNT_LFS); | |
1074 | set_opt(sbi, DISCARD); | |
1075 | } else { | |
1076 | set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE); | |
1077 | } | |
1078 | ||
1079 | #ifdef CONFIG_F2FS_FS_XATTR | |
1080 | set_opt(sbi, XATTR_USER); | |
1081 | #endif | |
1082 | #ifdef CONFIG_F2FS_FS_POSIX_ACL | |
1083 | set_opt(sbi, POSIX_ACL); | |
1084 | #endif | |
1085 | ||
1086 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
1087 | f2fs_build_fault_attr(sbi, 0); | |
1088 | #endif | |
1089 | } | |
1090 | ||
1091 | static int f2fs_remount(struct super_block *sb, int *flags, char *data) | |
1092 | { | |
1093 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1094 | struct f2fs_mount_info org_mount_opt; | |
1095 | int err, active_logs; | |
1096 | bool need_restart_gc = false; | |
1097 | bool need_stop_gc = false; | |
1098 | bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE); | |
1099 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
1100 | struct f2fs_fault_info ffi = sbi->fault_info; | |
1101 | #endif | |
1102 | ||
1103 | /* | |
1104 | * Save the old mount options in case we | |
1105 | * need to restore them. | |
1106 | */ | |
1107 | org_mount_opt = sbi->mount_opt; | |
1108 | active_logs = sbi->active_logs; | |
1109 | ||
1110 | /* recover superblocks we couldn't write due to previous RO mount */ | |
1111 | if (!(*flags & MS_RDONLY) && is_sbi_flag_set(sbi, SBI_NEED_SB_WRITE)) { | |
1112 | err = f2fs_commit_super(sbi, false); | |
1113 | f2fs_msg(sb, KERN_INFO, | |
1114 | "Try to recover all the superblocks, ret: %d", err); | |
1115 | if (!err) | |
1116 | clear_sbi_flag(sbi, SBI_NEED_SB_WRITE); | |
1117 | } | |
1118 | ||
1119 | sbi->mount_opt.opt = 0; | |
1120 | default_options(sbi); | |
1121 | ||
1122 | /* parse mount options */ | |
1123 | err = parse_options(sb, data); | |
1124 | if (err) | |
1125 | goto restore_opts; | |
1126 | ||
1127 | /* | |
1128 | * Previous and new state of filesystem is RO, | |
1129 | * so skip checking GC and FLUSH_MERGE conditions. | |
1130 | */ | |
1131 | if (f2fs_readonly(sb) && (*flags & MS_RDONLY)) | |
1132 | goto skip; | |
1133 | ||
1134 | /* disallow enable/disable extent_cache dynamically */ | |
1135 | if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) { | |
1136 | err = -EINVAL; | |
1137 | f2fs_msg(sbi->sb, KERN_WARNING, | |
1138 | "switch extent_cache option is not allowed"); | |
1139 | goto restore_opts; | |
1140 | } | |
1141 | ||
1142 | /* | |
1143 | * We stop the GC thread if FS is mounted as RO | |
1144 | * or if background_gc = off is passed in mount | |
1145 | * option. Also sync the filesystem. | |
1146 | */ | |
1147 | if ((*flags & MS_RDONLY) || !test_opt(sbi, BG_GC)) { | |
1148 | if (sbi->gc_thread) { | |
1149 | stop_gc_thread(sbi); | |
1150 | need_restart_gc = true; | |
1151 | } | |
1152 | } else if (!sbi->gc_thread) { | |
1153 | err = start_gc_thread(sbi); | |
1154 | if (err) | |
1155 | goto restore_opts; | |
1156 | need_stop_gc = true; | |
1157 | } | |
1158 | ||
1159 | if (*flags & MS_RDONLY) { | |
1160 | writeback_inodes_sb(sb, WB_REASON_SYNC); | |
1161 | sync_inodes_sb(sb); | |
1162 | ||
1163 | set_sbi_flag(sbi, SBI_IS_DIRTY); | |
1164 | set_sbi_flag(sbi, SBI_IS_CLOSE); | |
1165 | f2fs_sync_fs(sb, 1); | |
1166 | clear_sbi_flag(sbi, SBI_IS_CLOSE); | |
1167 | } | |
1168 | ||
1169 | /* | |
1170 | * We stop issue flush thread if FS is mounted as RO | |
1171 | * or if flush_merge is not passed in mount option. | |
1172 | */ | |
1173 | if ((*flags & MS_RDONLY) || !test_opt(sbi, FLUSH_MERGE)) { | |
1174 | clear_opt(sbi, FLUSH_MERGE); | |
1175 | destroy_flush_cmd_control(sbi, false); | |
1176 | } else { | |
1177 | err = create_flush_cmd_control(sbi); | |
1178 | if (err) | |
1179 | goto restore_gc; | |
1180 | } | |
1181 | skip: | |
1182 | /* Update the POSIXACL Flag */ | |
1183 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
1184 | (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0); | |
1185 | ||
1186 | return 0; | |
1187 | restore_gc: | |
1188 | if (need_restart_gc) { | |
1189 | if (start_gc_thread(sbi)) | |
1190 | f2fs_msg(sbi->sb, KERN_WARNING, | |
1191 | "background gc thread has stopped"); | |
1192 | } else if (need_stop_gc) { | |
1193 | stop_gc_thread(sbi); | |
1194 | } | |
1195 | restore_opts: | |
1196 | sbi->mount_opt = org_mount_opt; | |
1197 | sbi->active_logs = active_logs; | |
1198 | #ifdef CONFIG_F2FS_FAULT_INJECTION | |
1199 | sbi->fault_info = ffi; | |
1200 | #endif | |
1201 | return err; | |
1202 | } | |
1203 | ||
1204 | static struct super_operations f2fs_sops = { | |
1205 | .alloc_inode = f2fs_alloc_inode, | |
1206 | .drop_inode = f2fs_drop_inode, | |
1207 | .destroy_inode = f2fs_destroy_inode, | |
1208 | .write_inode = f2fs_write_inode, | |
1209 | .dirty_inode = f2fs_dirty_inode, | |
1210 | .show_options = f2fs_show_options, | |
1211 | .evict_inode = f2fs_evict_inode, | |
1212 | .put_super = f2fs_put_super, | |
1213 | .sync_fs = f2fs_sync_fs, | |
1214 | .freeze_fs = f2fs_freeze, | |
1215 | .unfreeze_fs = f2fs_unfreeze, | |
1216 | .statfs = f2fs_statfs, | |
1217 | .remount_fs = f2fs_remount, | |
1218 | }; | |
1219 | ||
1220 | #ifdef CONFIG_F2FS_FS_ENCRYPTION | |
1221 | static int f2fs_get_context(struct inode *inode, void *ctx, size_t len) | |
1222 | { | |
1223 | return f2fs_getxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, | |
1224 | F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, | |
1225 | ctx, len, NULL); | |
1226 | } | |
1227 | ||
1228 | static int f2fs_set_context(struct inode *inode, const void *ctx, size_t len, | |
1229 | void *fs_data) | |
1230 | { | |
1231 | return f2fs_setxattr(inode, F2FS_XATTR_INDEX_ENCRYPTION, | |
1232 | F2FS_XATTR_NAME_ENCRYPTION_CONTEXT, | |
1233 | ctx, len, fs_data, XATTR_CREATE); | |
1234 | } | |
1235 | ||
1236 | static unsigned f2fs_max_namelen(struct inode *inode) | |
1237 | { | |
1238 | return S_ISLNK(inode->i_mode) ? | |
1239 | inode->i_sb->s_blocksize : F2FS_NAME_LEN; | |
1240 | } | |
1241 | ||
1242 | static const struct fscrypt_operations f2fs_cryptops = { | |
1243 | .key_prefix = "f2fs:", | |
1244 | .get_context = f2fs_get_context, | |
1245 | .set_context = f2fs_set_context, | |
1246 | .is_encrypted = f2fs_encrypted_inode, | |
1247 | .empty_dir = f2fs_empty_dir, | |
1248 | .max_namelen = f2fs_max_namelen, | |
1249 | }; | |
1250 | #else | |
1251 | static const struct fscrypt_operations f2fs_cryptops = { | |
1252 | .is_encrypted = f2fs_encrypted_inode, | |
1253 | }; | |
1254 | #endif | |
1255 | ||
1256 | static struct inode *f2fs_nfs_get_inode(struct super_block *sb, | |
1257 | u64 ino, u32 generation) | |
1258 | { | |
1259 | struct f2fs_sb_info *sbi = F2FS_SB(sb); | |
1260 | struct inode *inode; | |
1261 | ||
1262 | if (check_nid_range(sbi, ino)) | |
1263 | return ERR_PTR(-ESTALE); | |
1264 | ||
1265 | /* | |
1266 | * f2fs_iget isn't quite right if the inode is currently unallocated! | |
1267 | * However f2fs_iget currently does appropriate checks to handle stale | |
1268 | * inodes so everything is OK. | |
1269 | */ | |
1270 | inode = f2fs_iget(sb, ino); | |
1271 | if (IS_ERR(inode)) | |
1272 | return ERR_CAST(inode); | |
1273 | if (unlikely(generation && inode->i_generation != generation)) { | |
1274 | /* we didn't find the right inode.. */ | |
1275 | iput(inode); | |
1276 | return ERR_PTR(-ESTALE); | |
1277 | } | |
1278 | return inode; | |
1279 | } | |
1280 | ||
1281 | static struct dentry *f2fs_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
1282 | int fh_len, int fh_type) | |
1283 | { | |
1284 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
1285 | f2fs_nfs_get_inode); | |
1286 | } | |
1287 | ||
1288 | static struct dentry *f2fs_fh_to_parent(struct super_block *sb, struct fid *fid, | |
1289 | int fh_len, int fh_type) | |
1290 | { | |
1291 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
1292 | f2fs_nfs_get_inode); | |
1293 | } | |
1294 | ||
1295 | static const struct export_operations f2fs_export_ops = { | |
1296 | .fh_to_dentry = f2fs_fh_to_dentry, | |
1297 | .fh_to_parent = f2fs_fh_to_parent, | |
1298 | .get_parent = f2fs_get_parent, | |
1299 | }; | |
1300 | ||
1301 | static loff_t max_file_blocks(void) | |
1302 | { | |
1303 | loff_t result = (DEF_ADDRS_PER_INODE - F2FS_INLINE_XATTR_ADDRS); | |
1304 | loff_t leaf_count = ADDRS_PER_BLOCK; | |
1305 | ||
1306 | /* two direct node blocks */ | |
1307 | result += (leaf_count * 2); | |
1308 | ||
1309 | /* two indirect node blocks */ | |
1310 | leaf_count *= NIDS_PER_BLOCK; | |
1311 | result += (leaf_count * 2); | |
1312 | ||
1313 | /* one double indirect node block */ | |
1314 | leaf_count *= NIDS_PER_BLOCK; | |
1315 | result += leaf_count; | |
1316 | ||
1317 | return result; | |
1318 | } | |
1319 | ||
1320 | static int __f2fs_commit_super(struct buffer_head *bh, | |
1321 | struct f2fs_super_block *super) | |
1322 | { | |
1323 | lock_buffer(bh); | |
1324 | if (super) | |
1325 | memcpy(bh->b_data + F2FS_SUPER_OFFSET, super, sizeof(*super)); | |
1326 | set_buffer_uptodate(bh); | |
1327 | set_buffer_dirty(bh); | |
1328 | unlock_buffer(bh); | |
1329 | ||
1330 | /* it's rare case, we can do fua all the time */ | |
1331 | return __sync_dirty_buffer(bh, REQ_SYNC | REQ_PREFLUSH | REQ_FUA); | |
1332 | } | |
1333 | ||
1334 | static inline bool sanity_check_area_boundary(struct f2fs_sb_info *sbi, | |
1335 | struct buffer_head *bh) | |
1336 | { | |
1337 | struct f2fs_super_block *raw_super = (struct f2fs_super_block *) | |
1338 | (bh->b_data + F2FS_SUPER_OFFSET); | |
1339 | struct super_block *sb = sbi->sb; | |
1340 | u32 segment0_blkaddr = le32_to_cpu(raw_super->segment0_blkaddr); | |
1341 | u32 cp_blkaddr = le32_to_cpu(raw_super->cp_blkaddr); | |
1342 | u32 sit_blkaddr = le32_to_cpu(raw_super->sit_blkaddr); | |
1343 | u32 nat_blkaddr = le32_to_cpu(raw_super->nat_blkaddr); | |
1344 | u32 ssa_blkaddr = le32_to_cpu(raw_super->ssa_blkaddr); | |
1345 | u32 main_blkaddr = le32_to_cpu(raw_super->main_blkaddr); | |
1346 | u32 segment_count_ckpt = le32_to_cpu(raw_super->segment_count_ckpt); | |
1347 | u32 segment_count_sit = le32_to_cpu(raw_super->segment_count_sit); | |
1348 | u32 segment_count_nat = le32_to_cpu(raw_super->segment_count_nat); | |
1349 | u32 segment_count_ssa = le32_to_cpu(raw_super->segment_count_ssa); | |
1350 | u32 segment_count_main = le32_to_cpu(raw_super->segment_count_main); | |
1351 | u32 segment_count = le32_to_cpu(raw_super->segment_count); | |
1352 | u32 log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); | |
1353 | u64 main_end_blkaddr = main_blkaddr + | |
1354 | (segment_count_main << log_blocks_per_seg); | |
1355 | u64 seg_end_blkaddr = segment0_blkaddr + | |
1356 | (segment_count << log_blocks_per_seg); | |
1357 | ||
1358 | if (segment0_blkaddr != cp_blkaddr) { | |
1359 | f2fs_msg(sb, KERN_INFO, | |
1360 | "Mismatch start address, segment0(%u) cp_blkaddr(%u)", | |
1361 | segment0_blkaddr, cp_blkaddr); | |
1362 | return true; | |
1363 | } | |
1364 | ||
1365 | if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) != | |
1366 | sit_blkaddr) { | |
1367 | f2fs_msg(sb, KERN_INFO, | |
1368 | "Wrong CP boundary, start(%u) end(%u) blocks(%u)", | |
1369 | cp_blkaddr, sit_blkaddr, | |
1370 | segment_count_ckpt << log_blocks_per_seg); | |
1371 | return true; | |
1372 | } | |
1373 | ||
1374 | if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) != | |
1375 | nat_blkaddr) { | |
1376 | f2fs_msg(sb, KERN_INFO, | |
1377 | "Wrong SIT boundary, start(%u) end(%u) blocks(%u)", | |
1378 | sit_blkaddr, nat_blkaddr, | |
1379 | segment_count_sit << log_blocks_per_seg); | |
1380 | return true; | |
1381 | } | |
1382 | ||
1383 | if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) != | |
1384 | ssa_blkaddr) { | |
1385 | f2fs_msg(sb, KERN_INFO, | |
1386 | "Wrong NAT boundary, start(%u) end(%u) blocks(%u)", | |
1387 | nat_blkaddr, ssa_blkaddr, | |
1388 | segment_count_nat << log_blocks_per_seg); | |
1389 | return true; | |
1390 | } | |
1391 | ||
1392 | if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) != | |
1393 | main_blkaddr) { | |
1394 | f2fs_msg(sb, KERN_INFO, | |
1395 | "Wrong SSA boundary, start(%u) end(%u) blocks(%u)", | |
1396 | ssa_blkaddr, main_blkaddr, | |
1397 | segment_count_ssa << log_blocks_per_seg); | |
1398 | return true; | |
1399 | } | |
1400 | ||
1401 | if (main_end_blkaddr > seg_end_blkaddr) { | |
1402 | f2fs_msg(sb, KERN_INFO, | |
1403 | "Wrong MAIN_AREA boundary, start(%u) end(%u) block(%u)", | |
1404 | main_blkaddr, | |
1405 | segment0_blkaddr + | |
1406 | (segment_count << log_blocks_per_seg), | |
1407 | segment_count_main << log_blocks_per_seg); | |
1408 | return true; | |
1409 | } else if (main_end_blkaddr < seg_end_blkaddr) { | |
1410 | int err = 0; | |
1411 | char *res; | |
1412 | ||
1413 | /* fix in-memory information all the time */ | |
1414 | raw_super->segment_count = cpu_to_le32((main_end_blkaddr - | |
1415 | segment0_blkaddr) >> log_blocks_per_seg); | |
1416 | ||
1417 | if (f2fs_readonly(sb) || bdev_read_only(sb->s_bdev)) { | |
1418 | set_sbi_flag(sbi, SBI_NEED_SB_WRITE); | |
1419 | res = "internally"; | |
1420 | } else { | |
1421 | err = __f2fs_commit_super(bh, NULL); | |
1422 | res = err ? "failed" : "done"; | |
1423 | } | |
1424 | f2fs_msg(sb, KERN_INFO, | |
1425 | "Fix alignment : %s, start(%u) end(%u) block(%u)", | |
1426 | res, main_blkaddr, | |
1427 | segment0_blkaddr + | |
1428 | (segment_count << log_blocks_per_seg), | |
1429 | segment_count_main << log_blocks_per_seg); | |
1430 | if (err) | |
1431 | return true; | |
1432 | } | |
1433 | return false; | |
1434 | } | |
1435 | ||
1436 | static int sanity_check_raw_super(struct f2fs_sb_info *sbi, | |
1437 | struct buffer_head *bh) | |
1438 | { | |
1439 | struct f2fs_super_block *raw_super = (struct f2fs_super_block *) | |
1440 | (bh->b_data + F2FS_SUPER_OFFSET); | |
1441 | struct super_block *sb = sbi->sb; | |
1442 | unsigned int blocksize; | |
1443 | ||
1444 | if (F2FS_SUPER_MAGIC != le32_to_cpu(raw_super->magic)) { | |
1445 | f2fs_msg(sb, KERN_INFO, | |
1446 | "Magic Mismatch, valid(0x%x) - read(0x%x)", | |
1447 | F2FS_SUPER_MAGIC, le32_to_cpu(raw_super->magic)); | |
1448 | return 1; | |
1449 | } | |
1450 | ||
1451 | /* Currently, support only 4KB page cache size */ | |
1452 | if (F2FS_BLKSIZE != PAGE_SIZE) { | |
1453 | f2fs_msg(sb, KERN_INFO, | |
1454 | "Invalid page_cache_size (%lu), supports only 4KB\n", | |
1455 | PAGE_SIZE); | |
1456 | return 1; | |
1457 | } | |
1458 | ||
1459 | /* Currently, support only 4KB block size */ | |
1460 | blocksize = 1 << le32_to_cpu(raw_super->log_blocksize); | |
1461 | if (blocksize != F2FS_BLKSIZE) { | |
1462 | f2fs_msg(sb, KERN_INFO, | |
1463 | "Invalid blocksize (%u), supports only 4KB\n", | |
1464 | blocksize); | |
1465 | return 1; | |
1466 | } | |
1467 | ||
1468 | /* check log blocks per segment */ | |
1469 | if (le32_to_cpu(raw_super->log_blocks_per_seg) != 9) { | |
1470 | f2fs_msg(sb, KERN_INFO, | |
1471 | "Invalid log blocks per segment (%u)\n", | |
1472 | le32_to_cpu(raw_super->log_blocks_per_seg)); | |
1473 | return 1; | |
1474 | } | |
1475 | ||
1476 | /* Currently, support 512/1024/2048/4096 bytes sector size */ | |
1477 | if (le32_to_cpu(raw_super->log_sectorsize) > | |
1478 | F2FS_MAX_LOG_SECTOR_SIZE || | |
1479 | le32_to_cpu(raw_super->log_sectorsize) < | |
1480 | F2FS_MIN_LOG_SECTOR_SIZE) { | |
1481 | f2fs_msg(sb, KERN_INFO, "Invalid log sectorsize (%u)", | |
1482 | le32_to_cpu(raw_super->log_sectorsize)); | |
1483 | return 1; | |
1484 | } | |
1485 | if (le32_to_cpu(raw_super->log_sectors_per_block) + | |
1486 | le32_to_cpu(raw_super->log_sectorsize) != | |
1487 | F2FS_MAX_LOG_SECTOR_SIZE) { | |
1488 | f2fs_msg(sb, KERN_INFO, | |
1489 | "Invalid log sectors per block(%u) log sectorsize(%u)", | |
1490 | le32_to_cpu(raw_super->log_sectors_per_block), | |
1491 | le32_to_cpu(raw_super->log_sectorsize)); | |
1492 | return 1; | |
1493 | } | |
1494 | ||
1495 | /* check reserved ino info */ | |
1496 | if (le32_to_cpu(raw_super->node_ino) != 1 || | |
1497 | le32_to_cpu(raw_super->meta_ino) != 2 || | |
1498 | le32_to_cpu(raw_super->root_ino) != 3) { | |
1499 | f2fs_msg(sb, KERN_INFO, | |
1500 | "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)", | |
1501 | le32_to_cpu(raw_super->node_ino), | |
1502 | le32_to_cpu(raw_super->meta_ino), | |
1503 | le32_to_cpu(raw_super->root_ino)); | |
1504 | return 1; | |
1505 | } | |
1506 | ||
1507 | if (le32_to_cpu(raw_super->segment_count) > F2FS_MAX_SEGMENT) { | |
1508 | f2fs_msg(sb, KERN_INFO, | |
1509 | "Invalid segment count (%u)", | |
1510 | le32_to_cpu(raw_super->segment_count)); | |
1511 | return 1; | |
1512 | } | |
1513 | ||
1514 | /* check CP/SIT/NAT/SSA/MAIN_AREA area boundary */ | |
1515 | if (sanity_check_area_boundary(sbi, bh)) | |
1516 | return 1; | |
1517 | ||
1518 | return 0; | |
1519 | } | |
1520 | ||
1521 | int sanity_check_ckpt(struct f2fs_sb_info *sbi) | |
1522 | { | |
1523 | unsigned int total, fsmeta; | |
1524 | struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); | |
1525 | struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); | |
1526 | unsigned int ovp_segments, reserved_segments; | |
1527 | unsigned int main_segs, blocks_per_seg; | |
1528 | int i; | |
1529 | ||
1530 | total = le32_to_cpu(raw_super->segment_count); | |
1531 | fsmeta = le32_to_cpu(raw_super->segment_count_ckpt); | |
1532 | fsmeta += le32_to_cpu(raw_super->segment_count_sit); | |
1533 | fsmeta += le32_to_cpu(raw_super->segment_count_nat); | |
1534 | fsmeta += le32_to_cpu(ckpt->rsvd_segment_count); | |
1535 | fsmeta += le32_to_cpu(raw_super->segment_count_ssa); | |
1536 | ||
1537 | if (unlikely(fsmeta >= total)) | |
1538 | return 1; | |
1539 | ||
1540 | ovp_segments = le32_to_cpu(ckpt->overprov_segment_count); | |
1541 | reserved_segments = le32_to_cpu(ckpt->rsvd_segment_count); | |
1542 | ||
1543 | if (unlikely(fsmeta < F2FS_MIN_SEGMENTS || | |
1544 | ovp_segments == 0 || reserved_segments == 0)) { | |
1545 | f2fs_msg(sbi->sb, KERN_ERR, | |
1546 | "Wrong layout: check mkfs.f2fs version"); | |
1547 | return 1; | |
1548 | } | |
1549 | ||
1550 | main_segs = le32_to_cpu(raw_super->segment_count_main); | |
1551 | blocks_per_seg = sbi->blocks_per_seg; | |
1552 | ||
1553 | for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { | |
1554 | if (le32_to_cpu(ckpt->cur_node_segno[i]) >= main_segs || | |
1555 | le16_to_cpu(ckpt->cur_node_blkoff[i]) >= blocks_per_seg) | |
1556 | return 1; | |
1557 | } | |
1558 | for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) { | |
1559 | if (le32_to_cpu(ckpt->cur_data_segno[i]) >= main_segs || | |
1560 | le16_to_cpu(ckpt->cur_data_blkoff[i]) >= blocks_per_seg) | |
1561 | return 1; | |
1562 | } | |
1563 | ||
1564 | if (unlikely(f2fs_cp_error(sbi))) { | |
1565 | f2fs_msg(sbi->sb, KERN_ERR, "A bug case: need to run fsck"); | |
1566 | return 1; | |
1567 | } | |
1568 | return 0; | |
1569 | } | |
1570 | ||
1571 | static void init_sb_info(struct f2fs_sb_info *sbi) | |
1572 | { | |
1573 | struct f2fs_super_block *raw_super = sbi->raw_super; | |
1574 | int i, j; | |
1575 | ||
1576 | sbi->log_sectors_per_block = | |
1577 | le32_to_cpu(raw_super->log_sectors_per_block); | |
1578 | sbi->log_blocksize = le32_to_cpu(raw_super->log_blocksize); | |
1579 | sbi->blocksize = 1 << sbi->log_blocksize; | |
1580 | sbi->log_blocks_per_seg = le32_to_cpu(raw_super->log_blocks_per_seg); | |
1581 | sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg; | |
1582 | sbi->segs_per_sec = le32_to_cpu(raw_super->segs_per_sec); | |
1583 | sbi->secs_per_zone = le32_to_cpu(raw_super->secs_per_zone); | |
1584 | sbi->total_sections = le32_to_cpu(raw_super->section_count); | |
1585 | sbi->total_node_count = | |
1586 | (le32_to_cpu(raw_super->segment_count_nat) / 2) | |
1587 | * sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK; | |
1588 | sbi->root_ino_num = le32_to_cpu(raw_super->root_ino); | |
1589 | sbi->node_ino_num = le32_to_cpu(raw_super->node_ino); | |
1590 | sbi->meta_ino_num = le32_to_cpu(raw_super->meta_ino); | |
1591 | sbi->cur_victim_sec = NULL_SECNO; | |
1592 | sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH; | |
1593 | ||
1594 | sbi->dir_level = DEF_DIR_LEVEL; | |
1595 | sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL; | |
1596 | sbi->interval_time[REQ_TIME] = DEF_IDLE_INTERVAL; | |
1597 | clear_sbi_flag(sbi, SBI_NEED_FSCK); | |
1598 | ||
1599 | for (i = 0; i < NR_COUNT_TYPE; i++) | |
1600 | atomic_set(&sbi->nr_pages[i], 0); | |
1601 | ||
1602 | atomic_set(&sbi->wb_sync_req, 0); | |
1603 | ||
1604 | INIT_LIST_HEAD(&sbi->s_list); | |
1605 | mutex_init(&sbi->umount_mutex); | |
1606 | for (i = 0; i < NR_PAGE_TYPE - 1; i++) | |
1607 | for (j = HOT; j < NR_TEMP_TYPE; j++) | |
1608 | mutex_init(&sbi->wio_mutex[i][j]); | |
1609 | spin_lock_init(&sbi->cp_lock); | |
1610 | } | |
1611 | ||
1612 | static int init_percpu_info(struct f2fs_sb_info *sbi) | |
1613 | { | |
1614 | int err; | |
1615 | ||
1616 | err = percpu_counter_init(&sbi->alloc_valid_block_count, 0, GFP_KERNEL); | |
1617 | if (err) | |
1618 | return err; | |
1619 | ||
1620 | return percpu_counter_init(&sbi->total_valid_inode_count, 0, | |
1621 | GFP_KERNEL); | |
1622 | } | |
1623 | ||
1624 | #ifdef CONFIG_BLK_DEV_ZONED | |
1625 | static int init_blkz_info(struct f2fs_sb_info *sbi, int devi) | |
1626 | { | |
1627 | struct block_device *bdev = FDEV(devi).bdev; | |
1628 | sector_t nr_sectors = bdev->bd_part->nr_sects; | |
1629 | sector_t sector = 0; | |
1630 | struct blk_zone *zones; | |
1631 | unsigned int i, nr_zones; | |
1632 | unsigned int n = 0; | |
1633 | int err = -EIO; | |
1634 | ||
1635 | if (!f2fs_sb_mounted_blkzoned(sbi->sb)) | |
1636 | return 0; | |
1637 | ||
1638 | if (sbi->blocks_per_blkz && sbi->blocks_per_blkz != | |
1639 | SECTOR_TO_BLOCK(bdev_zone_sectors(bdev))) | |
1640 | return -EINVAL; | |
1641 | sbi->blocks_per_blkz = SECTOR_TO_BLOCK(bdev_zone_sectors(bdev)); | |
1642 | if (sbi->log_blocks_per_blkz && sbi->log_blocks_per_blkz != | |
1643 | __ilog2_u32(sbi->blocks_per_blkz)) | |
1644 | return -EINVAL; | |
1645 | sbi->log_blocks_per_blkz = __ilog2_u32(sbi->blocks_per_blkz); | |
1646 | FDEV(devi).nr_blkz = SECTOR_TO_BLOCK(nr_sectors) >> | |
1647 | sbi->log_blocks_per_blkz; | |
1648 | if (nr_sectors & (bdev_zone_sectors(bdev) - 1)) | |
1649 | FDEV(devi).nr_blkz++; | |
1650 | ||
1651 | FDEV(devi).blkz_type = kmalloc(FDEV(devi).nr_blkz, GFP_KERNEL); | |
1652 | if (!FDEV(devi).blkz_type) | |
1653 | return -ENOMEM; | |
1654 | ||
1655 | #define F2FS_REPORT_NR_ZONES 4096 | |
1656 | ||
1657 | zones = kcalloc(F2FS_REPORT_NR_ZONES, sizeof(struct blk_zone), | |
1658 | GFP_KERNEL); | |
1659 | if (!zones) | |
1660 | return -ENOMEM; | |
1661 | ||
1662 | /* Get block zones type */ | |
1663 | while (zones && sector < nr_sectors) { | |
1664 | ||
1665 | nr_zones = F2FS_REPORT_NR_ZONES; | |
1666 | err = blkdev_report_zones(bdev, sector, | |
1667 | zones, &nr_zones, | |
1668 | GFP_KERNEL); | |
1669 | if (err) | |
1670 | break; | |
1671 | if (!nr_zones) { | |
1672 | err = -EIO; | |
1673 | break; | |
1674 | } | |
1675 | ||
1676 | for (i = 0; i < nr_zones; i++) { | |
1677 | FDEV(devi).blkz_type[n] = zones[i].type; | |
1678 | sector += zones[i].len; | |
1679 | n++; | |
1680 | } | |
1681 | } | |
1682 | ||
1683 | kfree(zones); | |
1684 | ||
1685 | return err; | |
1686 | } | |
1687 | #endif | |
1688 | ||
1689 | /* | |
1690 | * Read f2fs raw super block. | |
1691 | * Because we have two copies of super block, so read both of them | |
1692 | * to get the first valid one. If any one of them is broken, we pass | |
1693 | * them recovery flag back to the caller. | |
1694 | */ | |
1695 | static int read_raw_super_block(struct f2fs_sb_info *sbi, | |
1696 | struct f2fs_super_block **raw_super, | |
1697 | int *valid_super_block, int *recovery) | |
1698 | { | |
1699 | struct super_block *sb = sbi->sb; | |
1700 | int block; | |
1701 | struct buffer_head *bh; | |
1702 | struct f2fs_super_block *super; | |
1703 | int err = 0; | |
1704 | ||
1705 | super = kzalloc(sizeof(struct f2fs_super_block), GFP_KERNEL); | |
1706 | if (!super) | |
1707 | return -ENOMEM; | |
1708 | ||
1709 | for (block = 0; block < 2; block++) { | |
1710 | bh = sb_bread(sb, block); | |
1711 | if (!bh) { | |
1712 | f2fs_msg(sb, KERN_ERR, "Unable to read %dth superblock", | |
1713 | block + 1); | |
1714 | err = -EIO; | |
1715 | continue; | |
1716 | } | |
1717 | ||
1718 | /* sanity checking of raw super */ | |
1719 | if (sanity_check_raw_super(sbi, bh)) { | |
1720 | f2fs_msg(sb, KERN_ERR, | |
1721 | "Can't find valid F2FS filesystem in %dth superblock", | |
1722 | block + 1); | |
1723 | err = -EINVAL; | |
1724 | brelse(bh); | |
1725 | continue; | |
1726 | } | |
1727 | ||
1728 | if (!*raw_super) { | |
1729 | memcpy(super, bh->b_data + F2FS_SUPER_OFFSET, | |
1730 | sizeof(*super)); | |
1731 | *valid_super_block = block; | |
1732 | *raw_super = super; | |
1733 | } | |
1734 | brelse(bh); | |
1735 | } | |
1736 | ||
1737 | /* Fail to read any one of the superblocks*/ | |
1738 | if (err < 0) | |
1739 | *recovery = 1; | |
1740 | ||
1741 | /* No valid superblock */ | |
1742 | if (!*raw_super) | |
1743 | kfree(super); | |
1744 | else | |
1745 | err = 0; | |
1746 | ||
1747 | return err; | |
1748 | } | |
1749 | ||
1750 | int f2fs_commit_super(struct f2fs_sb_info *sbi, bool recover) | |
1751 | { | |
1752 | struct buffer_head *bh; | |
1753 | int err; | |
1754 | ||
1755 | if ((recover && f2fs_readonly(sbi->sb)) || | |
1756 | bdev_read_only(sbi->sb->s_bdev)) { | |
1757 | set_sbi_flag(sbi, SBI_NEED_SB_WRITE); | |
1758 | return -EROFS; | |
1759 | } | |
1760 | ||
1761 | /* write back-up superblock first */ | |
1762 | bh = sb_getblk(sbi->sb, sbi->valid_super_block ? 0: 1); | |
1763 | if (!bh) | |
1764 | return -EIO; | |
1765 | err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); | |
1766 | brelse(bh); | |
1767 | ||
1768 | /* if we are in recovery path, skip writing valid superblock */ | |
1769 | if (recover || err) | |
1770 | return err; | |
1771 | ||
1772 | /* write current valid superblock */ | |
1773 | bh = sb_getblk(sbi->sb, sbi->valid_super_block); | |
1774 | if (!bh) | |
1775 | return -EIO; | |
1776 | err = __f2fs_commit_super(bh, F2FS_RAW_SUPER(sbi)); | |
1777 | brelse(bh); | |
1778 | return err; | |
1779 | } | |
1780 | ||
1781 | static int f2fs_scan_devices(struct f2fs_sb_info *sbi) | |
1782 | { | |
1783 | struct f2fs_super_block *raw_super = F2FS_RAW_SUPER(sbi); | |
1784 | unsigned int max_devices = MAX_DEVICES; | |
1785 | int i; | |
1786 | ||
1787 | /* Initialize single device information */ | |
1788 | if (!RDEV(0).path[0]) { | |
1789 | if (!bdev_is_zoned(sbi->sb->s_bdev)) | |
1790 | return 0; | |
1791 | max_devices = 1; | |
1792 | } | |
1793 | ||
1794 | /* | |
1795 | * Initialize multiple devices information, or single | |
1796 | * zoned block device information. | |
1797 | */ | |
1798 | sbi->devs = kcalloc(max_devices, sizeof(struct f2fs_dev_info), | |
1799 | GFP_KERNEL); | |
1800 | if (!sbi->devs) | |
1801 | return -ENOMEM; | |
1802 | ||
1803 | for (i = 0; i < max_devices; i++) { | |
1804 | ||
1805 | if (i > 0 && !RDEV(i).path[0]) | |
1806 | break; | |
1807 | ||
1808 | if (max_devices == 1) { | |
1809 | /* Single zoned block device mount */ | |
1810 | FDEV(0).bdev = | |
1811 | blkdev_get_by_dev(sbi->sb->s_bdev->bd_dev, | |
1812 | sbi->sb->s_mode, sbi->sb->s_type); | |
1813 | } else { | |
1814 | /* Multi-device mount */ | |
1815 | memcpy(FDEV(i).path, RDEV(i).path, MAX_PATH_LEN); | |
1816 | FDEV(i).total_segments = | |
1817 | le32_to_cpu(RDEV(i).total_segments); | |
1818 | if (i == 0) { | |
1819 | FDEV(i).start_blk = 0; | |
1820 | FDEV(i).end_blk = FDEV(i).start_blk + | |
1821 | (FDEV(i).total_segments << | |
1822 | sbi->log_blocks_per_seg) - 1 + | |
1823 | le32_to_cpu(raw_super->segment0_blkaddr); | |
1824 | } else { | |
1825 | FDEV(i).start_blk = FDEV(i - 1).end_blk + 1; | |
1826 | FDEV(i).end_blk = FDEV(i).start_blk + | |
1827 | (FDEV(i).total_segments << | |
1828 | sbi->log_blocks_per_seg) - 1; | |
1829 | } | |
1830 | FDEV(i).bdev = blkdev_get_by_path(FDEV(i).path, | |
1831 | sbi->sb->s_mode, sbi->sb->s_type); | |
1832 | } | |
1833 | if (IS_ERR(FDEV(i).bdev)) | |
1834 | return PTR_ERR(FDEV(i).bdev); | |
1835 | ||
1836 | /* to release errored devices */ | |
1837 | sbi->s_ndevs = i + 1; | |
1838 | ||
1839 | #ifdef CONFIG_BLK_DEV_ZONED | |
1840 | if (bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HM && | |
1841 | !f2fs_sb_mounted_blkzoned(sbi->sb)) { | |
1842 | f2fs_msg(sbi->sb, KERN_ERR, | |
1843 | "Zoned block device feature not enabled\n"); | |
1844 | return -EINVAL; | |
1845 | } | |
1846 | if (bdev_zoned_model(FDEV(i).bdev) != BLK_ZONED_NONE) { | |
1847 | if (init_blkz_info(sbi, i)) { | |
1848 | f2fs_msg(sbi->sb, KERN_ERR, | |
1849 | "Failed to initialize F2FS blkzone information"); | |
1850 | return -EINVAL; | |
1851 | } | |
1852 | if (max_devices == 1) | |
1853 | break; | |
1854 | f2fs_msg(sbi->sb, KERN_INFO, | |
1855 | "Mount Device [%2d]: %20s, %8u, %8x - %8x (zone: %s)", | |
1856 | i, FDEV(i).path, | |
1857 | FDEV(i).total_segments, | |
1858 | FDEV(i).start_blk, FDEV(i).end_blk, | |
1859 | bdev_zoned_model(FDEV(i).bdev) == BLK_ZONED_HA ? | |
1860 | "Host-aware" : "Host-managed"); | |
1861 | continue; | |
1862 | } | |
1863 | #endif | |
1864 | f2fs_msg(sbi->sb, KERN_INFO, | |
1865 | "Mount Device [%2d]: %20s, %8u, %8x - %8x", | |
1866 | i, FDEV(i).path, | |
1867 | FDEV(i).total_segments, | |
1868 | FDEV(i).start_blk, FDEV(i).end_blk); | |
1869 | } | |
1870 | f2fs_msg(sbi->sb, KERN_INFO, | |
1871 | "IO Block Size: %8d KB", F2FS_IO_SIZE_KB(sbi)); | |
1872 | return 0; | |
1873 | } | |
1874 | ||
1875 | static int f2fs_fill_super(struct super_block *sb, void *data, int silent) | |
1876 | { | |
1877 | struct f2fs_sb_info *sbi; | |
1878 | struct f2fs_super_block *raw_super; | |
1879 | struct inode *root; | |
1880 | int err; | |
1881 | bool retry = true, need_fsck = false; | |
1882 | char *options = NULL; | |
1883 | int recovery, i, valid_super_block; | |
1884 | struct curseg_info *seg_i; | |
1885 | ||
1886 | try_onemore: | |
1887 | err = -EINVAL; | |
1888 | raw_super = NULL; | |
1889 | valid_super_block = -1; | |
1890 | recovery = 0; | |
1891 | ||
1892 | /* allocate memory for f2fs-specific super block info */ | |
1893 | sbi = kzalloc(sizeof(struct f2fs_sb_info), GFP_KERNEL); | |
1894 | if (!sbi) | |
1895 | return -ENOMEM; | |
1896 | ||
1897 | sbi->sb = sb; | |
1898 | ||
1899 | /* Load the checksum driver */ | |
1900 | sbi->s_chksum_driver = crypto_alloc_shash("crc32", 0, 0); | |
1901 | if (IS_ERR(sbi->s_chksum_driver)) { | |
1902 | f2fs_msg(sb, KERN_ERR, "Cannot load crc32 driver."); | |
1903 | err = PTR_ERR(sbi->s_chksum_driver); | |
1904 | sbi->s_chksum_driver = NULL; | |
1905 | goto free_sbi; | |
1906 | } | |
1907 | ||
1908 | /* set a block size */ | |
1909 | if (unlikely(!sb_set_blocksize(sb, F2FS_BLKSIZE))) { | |
1910 | f2fs_msg(sb, KERN_ERR, "unable to set blocksize"); | |
1911 | goto free_sbi; | |
1912 | } | |
1913 | ||
1914 | err = read_raw_super_block(sbi, &raw_super, &valid_super_block, | |
1915 | &recovery); | |
1916 | if (err) | |
1917 | goto free_sbi; | |
1918 | ||
1919 | sb->s_fs_info = sbi; | |
1920 | sbi->raw_super = raw_super; | |
1921 | ||
1922 | /* | |
1923 | * The BLKZONED feature indicates that the drive was formatted with | |
1924 | * zone alignment optimization. This is optional for host-aware | |
1925 | * devices, but mandatory for host-managed zoned block devices. | |
1926 | */ | |
1927 | #ifndef CONFIG_BLK_DEV_ZONED | |
1928 | if (f2fs_sb_mounted_blkzoned(sb)) { | |
1929 | f2fs_msg(sb, KERN_ERR, | |
1930 | "Zoned block device support is not enabled\n"); | |
1931 | goto free_sb_buf; | |
1932 | } | |
1933 | #endif | |
1934 | default_options(sbi); | |
1935 | /* parse mount options */ | |
1936 | options = kstrdup((const char *)data, GFP_KERNEL); | |
1937 | if (data && !options) { | |
1938 | err = -ENOMEM; | |
1939 | goto free_sb_buf; | |
1940 | } | |
1941 | ||
1942 | err = parse_options(sb, options); | |
1943 | if (err) | |
1944 | goto free_options; | |
1945 | ||
1946 | sbi->max_file_blocks = max_file_blocks(); | |
1947 | sb->s_maxbytes = sbi->max_file_blocks << | |
1948 | le32_to_cpu(raw_super->log_blocksize); | |
1949 | sb->s_max_links = F2FS_LINK_MAX; | |
1950 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
1951 | ||
1952 | sb->s_op = &f2fs_sops; | |
1953 | sb->s_cop = &f2fs_cryptops; | |
1954 | sb->s_xattr = f2fs_xattr_handlers; | |
1955 | sb->s_export_op = &f2fs_export_ops; | |
1956 | sb->s_magic = F2FS_SUPER_MAGIC; | |
1957 | sb->s_time_gran = 1; | |
1958 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | | |
1959 | (test_opt(sbi, POSIX_ACL) ? MS_POSIXACL : 0); | |
1960 | memcpy(sb->s_uuid, raw_super->uuid, sizeof(raw_super->uuid)); | |
1961 | ||
1962 | /* init f2fs-specific super block info */ | |
1963 | sbi->valid_super_block = valid_super_block; | |
1964 | mutex_init(&sbi->gc_mutex); | |
1965 | mutex_init(&sbi->cp_mutex); | |
1966 | init_rwsem(&sbi->node_write); | |
1967 | init_rwsem(&sbi->node_change); | |
1968 | ||
1969 | /* disallow all the data/node/meta page writes */ | |
1970 | set_sbi_flag(sbi, SBI_POR_DOING); | |
1971 | spin_lock_init(&sbi->stat_lock); | |
1972 | ||
1973 | for (i = 0; i < NR_PAGE_TYPE; i++) { | |
1974 | int n = (i == META) ? 1: NR_TEMP_TYPE; | |
1975 | int j; | |
1976 | ||
1977 | sbi->write_io[i] = kmalloc(n * sizeof(struct f2fs_bio_info), | |
1978 | GFP_KERNEL); | |
1979 | if (!sbi->write_io[i]) | |
1980 | goto free_options; | |
1981 | ||
1982 | for (j = HOT; j < n; j++) { | |
1983 | init_rwsem(&sbi->write_io[i][j].io_rwsem); | |
1984 | sbi->write_io[i][j].sbi = sbi; | |
1985 | sbi->write_io[i][j].bio = NULL; | |
1986 | spin_lock_init(&sbi->write_io[i][j].io_lock); | |
1987 | INIT_LIST_HEAD(&sbi->write_io[i][j].io_list); | |
1988 | } | |
1989 | } | |
1990 | ||
1991 | init_rwsem(&sbi->cp_rwsem); | |
1992 | init_waitqueue_head(&sbi->cp_wait); | |
1993 | init_sb_info(sbi); | |
1994 | ||
1995 | err = init_percpu_info(sbi); | |
1996 | if (err) | |
1997 | goto free_options; | |
1998 | ||
1999 | if (F2FS_IO_SIZE(sbi) > 1) { | |
2000 | sbi->write_io_dummy = | |
2001 | mempool_create_page_pool(2 * (F2FS_IO_SIZE(sbi) - 1), 0); | |
2002 | if (!sbi->write_io_dummy) | |
2003 | goto free_options; | |
2004 | } | |
2005 | ||
2006 | /* get an inode for meta space */ | |
2007 | sbi->meta_inode = f2fs_iget(sb, F2FS_META_INO(sbi)); | |
2008 | if (IS_ERR(sbi->meta_inode)) { | |
2009 | f2fs_msg(sb, KERN_ERR, "Failed to read F2FS meta data inode"); | |
2010 | err = PTR_ERR(sbi->meta_inode); | |
2011 | goto free_io_dummy; | |
2012 | } | |
2013 | ||
2014 | err = get_valid_checkpoint(sbi); | |
2015 | if (err) { | |
2016 | f2fs_msg(sb, KERN_ERR, "Failed to get valid F2FS checkpoint"); | |
2017 | goto free_meta_inode; | |
2018 | } | |
2019 | ||
2020 | /* Initialize device list */ | |
2021 | err = f2fs_scan_devices(sbi); | |
2022 | if (err) { | |
2023 | f2fs_msg(sb, KERN_ERR, "Failed to find devices"); | |
2024 | goto free_devices; | |
2025 | } | |
2026 | ||
2027 | sbi->total_valid_node_count = | |
2028 | le32_to_cpu(sbi->ckpt->valid_node_count); | |
2029 | percpu_counter_set(&sbi->total_valid_inode_count, | |
2030 | le32_to_cpu(sbi->ckpt->valid_inode_count)); | |
2031 | sbi->user_block_count = le64_to_cpu(sbi->ckpt->user_block_count); | |
2032 | sbi->total_valid_block_count = | |
2033 | le64_to_cpu(sbi->ckpt->valid_block_count); | |
2034 | sbi->last_valid_block_count = sbi->total_valid_block_count; | |
2035 | ||
2036 | for (i = 0; i < NR_INODE_TYPE; i++) { | |
2037 | INIT_LIST_HEAD(&sbi->inode_list[i]); | |
2038 | spin_lock_init(&sbi->inode_lock[i]); | |
2039 | } | |
2040 | ||
2041 | init_extent_cache_info(sbi); | |
2042 | ||
2043 | init_ino_entry_info(sbi); | |
2044 | ||
2045 | /* setup f2fs internal modules */ | |
2046 | err = build_segment_manager(sbi); | |
2047 | if (err) { | |
2048 | f2fs_msg(sb, KERN_ERR, | |
2049 | "Failed to initialize F2FS segment manager"); | |
2050 | goto free_sm; | |
2051 | } | |
2052 | err = build_node_manager(sbi); | |
2053 | if (err) { | |
2054 | f2fs_msg(sb, KERN_ERR, | |
2055 | "Failed to initialize F2FS node manager"); | |
2056 | goto free_nm; | |
2057 | } | |
2058 | ||
2059 | /* For write statistics */ | |
2060 | if (sb->s_bdev->bd_part) | |
2061 | sbi->sectors_written_start = | |
2062 | (u64)part_stat_read(sb->s_bdev->bd_part, sectors[1]); | |
2063 | ||
2064 | /* Read accumulated write IO statistics if exists */ | |
2065 | seg_i = CURSEG_I(sbi, CURSEG_HOT_NODE); | |
2066 | if (__exist_node_summaries(sbi)) | |
2067 | sbi->kbytes_written = | |
2068 | le64_to_cpu(seg_i->journal->info.kbytes_written); | |
2069 | ||
2070 | build_gc_manager(sbi); | |
2071 | ||
2072 | /* get an inode for node space */ | |
2073 | sbi->node_inode = f2fs_iget(sb, F2FS_NODE_INO(sbi)); | |
2074 | if (IS_ERR(sbi->node_inode)) { | |
2075 | f2fs_msg(sb, KERN_ERR, "Failed to read node inode"); | |
2076 | err = PTR_ERR(sbi->node_inode); | |
2077 | goto free_nm; | |
2078 | } | |
2079 | ||
2080 | f2fs_join_shrinker(sbi); | |
2081 | ||
2082 | err = f2fs_build_stats(sbi); | |
2083 | if (err) | |
2084 | goto free_nm; | |
2085 | ||
2086 | /* if there are nt orphan nodes free them */ | |
2087 | err = recover_orphan_inodes(sbi); | |
2088 | if (err) | |
2089 | goto free_node_inode; | |
2090 | ||
2091 | /* read root inode and dentry */ | |
2092 | root = f2fs_iget(sb, F2FS_ROOT_INO(sbi)); | |
2093 | if (IS_ERR(root)) { | |
2094 | f2fs_msg(sb, KERN_ERR, "Failed to read root inode"); | |
2095 | err = PTR_ERR(root); | |
2096 | goto free_node_inode; | |
2097 | } | |
2098 | if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { | |
2099 | iput(root); | |
2100 | err = -EINVAL; | |
2101 | goto free_node_inode; | |
2102 | } | |
2103 | ||
2104 | sb->s_root = d_make_root(root); /* allocate root dentry */ | |
2105 | if (!sb->s_root) { | |
2106 | err = -ENOMEM; | |
2107 | goto free_root_inode; | |
2108 | } | |
2109 | ||
2110 | if (f2fs_proc_root) | |
2111 | sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root); | |
2112 | ||
2113 | if (sbi->s_proc) { | |
2114 | proc_create_data("segment_info", S_IRUGO, sbi->s_proc, | |
2115 | &f2fs_seq_segment_info_fops, sb); | |
2116 | proc_create_data("segment_bits", S_IRUGO, sbi->s_proc, | |
2117 | &f2fs_seq_segment_bits_fops, sb); | |
2118 | } | |
2119 | ||
2120 | sbi->s_kobj.kset = f2fs_kset; | |
2121 | init_completion(&sbi->s_kobj_unregister); | |
2122 | err = kobject_init_and_add(&sbi->s_kobj, &f2fs_ktype, NULL, | |
2123 | "%s", sb->s_id); | |
2124 | if (err) | |
2125 | goto free_proc; | |
2126 | ||
2127 | /* recover fsynced data */ | |
2128 | if (!test_opt(sbi, DISABLE_ROLL_FORWARD)) { | |
2129 | /* | |
2130 | * mount should be failed, when device has readonly mode, and | |
2131 | * previous checkpoint was not done by clean system shutdown. | |
2132 | */ | |
2133 | if (bdev_read_only(sb->s_bdev) && | |
2134 | !is_set_ckpt_flags(sbi, CP_UMOUNT_FLAG)) { | |
2135 | err = -EROFS; | |
2136 | goto free_kobj; | |
2137 | } | |
2138 | ||
2139 | if (need_fsck) | |
2140 | set_sbi_flag(sbi, SBI_NEED_FSCK); | |
2141 | ||
2142 | if (!retry) | |
2143 | goto skip_recovery; | |
2144 | ||
2145 | err = recover_fsync_data(sbi, false); | |
2146 | if (err < 0) { | |
2147 | need_fsck = true; | |
2148 | f2fs_msg(sb, KERN_ERR, | |
2149 | "Cannot recover all fsync data errno=%d", err); | |
2150 | goto free_kobj; | |
2151 | } | |
2152 | } else { | |
2153 | err = recover_fsync_data(sbi, true); | |
2154 | ||
2155 | if (!f2fs_readonly(sb) && err > 0) { | |
2156 | err = -EINVAL; | |
2157 | f2fs_msg(sb, KERN_ERR, | |
2158 | "Need to recover fsync data"); | |
2159 | goto free_kobj; | |
2160 | } | |
2161 | } | |
2162 | skip_recovery: | |
2163 | /* recover_fsync_data() cleared this already */ | |
2164 | clear_sbi_flag(sbi, SBI_POR_DOING); | |
2165 | ||
2166 | /* | |
2167 | * If filesystem is not mounted as read-only then | |
2168 | * do start the gc_thread. | |
2169 | */ | |
2170 | if (test_opt(sbi, BG_GC) && !f2fs_readonly(sb)) { | |
2171 | /* After POR, we can run background GC thread.*/ | |
2172 | err = start_gc_thread(sbi); | |
2173 | if (err) | |
2174 | goto free_kobj; | |
2175 | } | |
2176 | kfree(options); | |
2177 | ||
2178 | /* recover broken superblock */ | |
2179 | if (recovery) { | |
2180 | err = f2fs_commit_super(sbi, true); | |
2181 | f2fs_msg(sb, KERN_INFO, | |
2182 | "Try to recover %dth superblock, ret: %d", | |
2183 | sbi->valid_super_block ? 1 : 2, err); | |
2184 | } | |
2185 | ||
2186 | f2fs_msg(sbi->sb, KERN_NOTICE, "Mounted with checkpoint version = %llx", | |
2187 | cur_cp_version(F2FS_CKPT(sbi))); | |
2188 | f2fs_update_time(sbi, CP_TIME); | |
2189 | f2fs_update_time(sbi, REQ_TIME); | |
2190 | return 0; | |
2191 | ||
2192 | free_kobj: | |
2193 | f2fs_sync_inode_meta(sbi); | |
2194 | kobject_del(&sbi->s_kobj); | |
2195 | kobject_put(&sbi->s_kobj); | |
2196 | wait_for_completion(&sbi->s_kobj_unregister); | |
2197 | free_proc: | |
2198 | if (sbi->s_proc) { | |
2199 | remove_proc_entry("segment_info", sbi->s_proc); | |
2200 | remove_proc_entry("segment_bits", sbi->s_proc); | |
2201 | remove_proc_entry(sb->s_id, f2fs_proc_root); | |
2202 | } | |
2203 | free_root_inode: | |
2204 | dput(sb->s_root); | |
2205 | sb->s_root = NULL; | |
2206 | free_node_inode: | |
2207 | truncate_inode_pages_final(NODE_MAPPING(sbi)); | |
2208 | mutex_lock(&sbi->umount_mutex); | |
2209 | release_ino_entry(sbi, true); | |
2210 | f2fs_leave_shrinker(sbi); | |
2211 | /* | |
2212 | * Some dirty meta pages can be produced by recover_orphan_inodes() | |
2213 | * failed by EIO. Then, iput(node_inode) can trigger balance_fs_bg() | |
2214 | * followed by write_checkpoint() through f2fs_write_node_pages(), which | |
2215 | * falls into an infinite loop in sync_meta_pages(). | |
2216 | */ | |
2217 | truncate_inode_pages_final(META_MAPPING(sbi)); | |
2218 | iput(sbi->node_inode); | |
2219 | mutex_unlock(&sbi->umount_mutex); | |
2220 | f2fs_destroy_stats(sbi); | |
2221 | free_nm: | |
2222 | destroy_node_manager(sbi); | |
2223 | free_sm: | |
2224 | destroy_segment_manager(sbi); | |
2225 | free_devices: | |
2226 | destroy_device_list(sbi); | |
2227 | kfree(sbi->ckpt); | |
2228 | free_meta_inode: | |
2229 | make_bad_inode(sbi->meta_inode); | |
2230 | iput(sbi->meta_inode); | |
2231 | free_io_dummy: | |
2232 | mempool_destroy(sbi->write_io_dummy); | |
2233 | free_options: | |
2234 | for (i = 0; i < NR_PAGE_TYPE; i++) | |
2235 | kfree(sbi->write_io[i]); | |
2236 | destroy_percpu_info(sbi); | |
2237 | kfree(options); | |
2238 | free_sb_buf: | |
2239 | kfree(raw_super); | |
2240 | free_sbi: | |
2241 | if (sbi->s_chksum_driver) | |
2242 | crypto_free_shash(sbi->s_chksum_driver); | |
2243 | kfree(sbi); | |
2244 | ||
2245 | /* give only one another chance */ | |
2246 | if (retry) { | |
2247 | retry = false; | |
2248 | shrink_dcache_sb(sb); | |
2249 | goto try_onemore; | |
2250 | } | |
2251 | return err; | |
2252 | } | |
2253 | ||
2254 | static struct dentry *f2fs_mount(struct file_system_type *fs_type, int flags, | |
2255 | const char *dev_name, void *data) | |
2256 | { | |
2257 | return mount_bdev(fs_type, flags, dev_name, data, f2fs_fill_super); | |
2258 | } | |
2259 | ||
2260 | static void kill_f2fs_super(struct super_block *sb) | |
2261 | { | |
2262 | if (sb->s_root) | |
2263 | set_sbi_flag(F2FS_SB(sb), SBI_IS_CLOSE); | |
2264 | kill_block_super(sb); | |
2265 | } | |
2266 | ||
2267 | static struct file_system_type f2fs_fs_type = { | |
2268 | .owner = THIS_MODULE, | |
2269 | .name = "f2fs", | |
2270 | .mount = f2fs_mount, | |
2271 | .kill_sb = kill_f2fs_super, | |
2272 | .fs_flags = FS_REQUIRES_DEV, | |
2273 | }; | |
2274 | MODULE_ALIAS_FS("f2fs"); | |
2275 | ||
2276 | static int __init init_inodecache(void) | |
2277 | { | |
2278 | f2fs_inode_cachep = kmem_cache_create("f2fs_inode_cache", | |
2279 | sizeof(struct f2fs_inode_info), 0, | |
2280 | SLAB_RECLAIM_ACCOUNT|SLAB_ACCOUNT, NULL); | |
2281 | if (!f2fs_inode_cachep) | |
2282 | return -ENOMEM; | |
2283 | return 0; | |
2284 | } | |
2285 | ||
2286 | static void destroy_inodecache(void) | |
2287 | { | |
2288 | /* | |
2289 | * Make sure all delayed rcu free inodes are flushed before we | |
2290 | * destroy cache. | |
2291 | */ | |
2292 | rcu_barrier(); | |
2293 | kmem_cache_destroy(f2fs_inode_cachep); | |
2294 | } | |
2295 | ||
2296 | static int __init init_f2fs_fs(void) | |
2297 | { | |
2298 | int err; | |
2299 | ||
2300 | f2fs_build_trace_ios(); | |
2301 | ||
2302 | err = init_inodecache(); | |
2303 | if (err) | |
2304 | goto fail; | |
2305 | err = create_node_manager_caches(); | |
2306 | if (err) | |
2307 | goto free_inodecache; | |
2308 | err = create_segment_manager_caches(); | |
2309 | if (err) | |
2310 | goto free_node_manager_caches; | |
2311 | err = create_checkpoint_caches(); | |
2312 | if (err) | |
2313 | goto free_segment_manager_caches; | |
2314 | err = create_extent_cache(); | |
2315 | if (err) | |
2316 | goto free_checkpoint_caches; | |
2317 | f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj); | |
2318 | if (!f2fs_kset) { | |
2319 | err = -ENOMEM; | |
2320 | goto free_extent_cache; | |
2321 | } | |
2322 | err = register_shrinker(&f2fs_shrinker_info); | |
2323 | if (err) | |
2324 | goto free_kset; | |
2325 | ||
2326 | err = register_filesystem(&f2fs_fs_type); | |
2327 | if (err) | |
2328 | goto free_shrinker; | |
2329 | err = f2fs_create_root_stats(); | |
2330 | if (err) | |
2331 | goto free_filesystem; | |
2332 | f2fs_proc_root = proc_mkdir("fs/f2fs", NULL); | |
2333 | return 0; | |
2334 | ||
2335 | free_filesystem: | |
2336 | unregister_filesystem(&f2fs_fs_type); | |
2337 | free_shrinker: | |
2338 | unregister_shrinker(&f2fs_shrinker_info); | |
2339 | free_kset: | |
2340 | kset_unregister(f2fs_kset); | |
2341 | free_extent_cache: | |
2342 | destroy_extent_cache(); | |
2343 | free_checkpoint_caches: | |
2344 | destroy_checkpoint_caches(); | |
2345 | free_segment_manager_caches: | |
2346 | destroy_segment_manager_caches(); | |
2347 | free_node_manager_caches: | |
2348 | destroy_node_manager_caches(); | |
2349 | free_inodecache: | |
2350 | destroy_inodecache(); | |
2351 | fail: | |
2352 | return err; | |
2353 | } | |
2354 | ||
2355 | static void __exit exit_f2fs_fs(void) | |
2356 | { | |
2357 | remove_proc_entry("fs/f2fs", NULL); | |
2358 | f2fs_destroy_root_stats(); | |
2359 | unregister_filesystem(&f2fs_fs_type); | |
2360 | unregister_shrinker(&f2fs_shrinker_info); | |
2361 | kset_unregister(f2fs_kset); | |
2362 | destroy_extent_cache(); | |
2363 | destroy_checkpoint_caches(); | |
2364 | destroy_segment_manager_caches(); | |
2365 | destroy_node_manager_caches(); | |
2366 | destroy_inodecache(); | |
2367 | f2fs_destroy_trace_ios(); | |
2368 | } | |
2369 | ||
2370 | module_init(init_f2fs_fs) | |
2371 | module_exit(exit_f2fs_fs) | |
2372 | ||
2373 | MODULE_AUTHOR("Samsung Electronics's Praesto Team"); | |
2374 | MODULE_DESCRIPTION("Flash Friendly File System"); | |
2375 | MODULE_LICENSE("GPL"); | |
2376 |