• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /**
2  * fsck.c
3  *
4  * Copyright (c) 2013 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 "fsck.h"
12 #include "xattr.h"
13 #include "quotaio.h"
14 #include <time.h>
15 
16 char *tree_mark;
17 uint32_t tree_mark_size = 256;
18 
f2fs_set_main_bitmap(struct f2fs_sb_info * sbi,u32 blk,int type)19 int f2fs_set_main_bitmap(struct f2fs_sb_info *sbi, u32 blk, int type)
20 {
21 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
22 	struct seg_entry *se;
23 	int fix = 0;
24 
25 	se = get_seg_entry(sbi, GET_SEGNO(sbi, blk));
26 	if (se->type >= NO_CHECK_TYPE)
27 		fix = 1;
28 	else if (IS_DATASEG(se->type) != IS_DATASEG(type))
29 		fix = 1;
30 
31 	/* just check data and node types */
32 	if (fix) {
33 		DBG(1, "Wrong segment type [0x%x] %x -> %x",
34 				GET_SEGNO(sbi, blk), se->type, type);
35 		se->type = type;
36 	}
37 	return f2fs_set_bit(BLKOFF_FROM_MAIN(sbi, blk), fsck->main_area_bitmap);
38 }
39 
f2fs_test_main_bitmap(struct f2fs_sb_info * sbi,u32 blk)40 static inline int f2fs_test_main_bitmap(struct f2fs_sb_info *sbi, u32 blk)
41 {
42 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
43 
44 	return f2fs_test_bit(BLKOFF_FROM_MAIN(sbi, blk),
45 						fsck->main_area_bitmap);
46 }
47 
f2fs_clear_main_bitmap(struct f2fs_sb_info * sbi,u32 blk)48 static inline int f2fs_clear_main_bitmap(struct f2fs_sb_info *sbi, u32 blk)
49 {
50 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
51 
52 	return f2fs_clear_bit(BLKOFF_FROM_MAIN(sbi, blk),
53 						fsck->main_area_bitmap);
54 }
55 
f2fs_test_sit_bitmap(struct f2fs_sb_info * sbi,u32 blk)56 static inline int f2fs_test_sit_bitmap(struct f2fs_sb_info *sbi, u32 blk)
57 {
58 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
59 
60 	return f2fs_test_bit(BLKOFF_FROM_MAIN(sbi, blk), fsck->sit_area_bitmap);
61 }
62 
f2fs_set_sit_bitmap(struct f2fs_sb_info * sbi,u32 blk)63 int f2fs_set_sit_bitmap(struct f2fs_sb_info *sbi, u32 blk)
64 {
65 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
66 
67 	return f2fs_set_bit(BLKOFF_FROM_MAIN(sbi, blk), fsck->sit_area_bitmap);
68 }
69 
add_into_hard_link_list(struct f2fs_sb_info * sbi,u32 nid,u32 link_cnt)70 static int add_into_hard_link_list(struct f2fs_sb_info *sbi,
71 						u32 nid, u32 link_cnt)
72 {
73 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
74 	struct hard_link_node *node = NULL, *tmp = NULL, *prev = NULL;
75 
76 	node = calloc(sizeof(struct hard_link_node), 1);
77 	ASSERT(node != NULL);
78 
79 	node->nid = nid;
80 	node->links = link_cnt;
81 	node->actual_links = 1;
82 	node->next = NULL;
83 
84 	if (fsck->hard_link_list_head == NULL) {
85 		fsck->hard_link_list_head = node;
86 		goto out;
87 	}
88 
89 	tmp = fsck->hard_link_list_head;
90 
91 	/* Find insertion position */
92 	while (tmp && (nid < tmp->nid)) {
93 		ASSERT(tmp->nid != nid);
94 		prev = tmp;
95 		tmp = tmp->next;
96 	}
97 
98 	if (tmp == fsck->hard_link_list_head) {
99 		node->next = tmp;
100 		fsck->hard_link_list_head = node;
101 	} else {
102 		prev->next = node;
103 		node->next = tmp;
104 	}
105 
106 out:
107 	DBG(2, "ino[0x%x] has hard links [0x%x]\n", nid, link_cnt);
108 	return 0;
109 }
110 
find_and_dec_hard_link_list(struct f2fs_sb_info * sbi,u32 nid)111 static int find_and_dec_hard_link_list(struct f2fs_sb_info *sbi, u32 nid)
112 {
113 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
114 	struct hard_link_node *node = NULL, *prev = NULL;
115 
116 	if (fsck->hard_link_list_head == NULL)
117 		return -EINVAL;
118 
119 	node = fsck->hard_link_list_head;
120 
121 	while (node && (nid < node->nid)) {
122 		prev = node;
123 		node = node->next;
124 	}
125 
126 	if (node == NULL || (nid != node->nid))
127 		return -EINVAL;
128 
129 	/* Decrease link count */
130 	node->links = node->links - 1;
131 	node->actual_links++;
132 
133 	/* if link count becomes one, remove the node */
134 	if (node->links == 1) {
135 		if (fsck->hard_link_list_head == node)
136 			fsck->hard_link_list_head = node->next;
137 		else
138 			prev->next = node->next;
139 		free(node);
140 	}
141 	return 0;
142 }
143 
is_valid_ssa_node_blk(struct f2fs_sb_info * sbi,u32 nid,u32 blk_addr)144 static int is_valid_ssa_node_blk(struct f2fs_sb_info *sbi, u32 nid,
145 							u32 blk_addr)
146 {
147 	struct f2fs_summary_block *sum_blk;
148 	struct f2fs_summary *sum_entry;
149 	struct seg_entry * se;
150 	u32 segno, offset;
151 	int need_fix = 0, ret = 0;
152 	int type;
153 
154 	segno = GET_SEGNO(sbi, blk_addr);
155 	offset = OFFSET_IN_SEG(sbi, blk_addr);
156 
157 	sum_blk = get_sum_block(sbi, segno, &type);
158 
159 	if (type != SEG_TYPE_NODE && type != SEG_TYPE_CUR_NODE) {
160 		/* can't fix current summary, then drop the block */
161 		if (!c.fix_on || type < 0) {
162 			ASSERT_MSG("Summary footer is not for node segment");
163 			ret = -EINVAL;
164 			goto out;
165 		}
166 
167 		need_fix = 1;
168 		se = get_seg_entry(sbi, segno);
169 		if(IS_NODESEG(se->type)) {
170 			FIX_MSG("Summary footer indicates a node segment: 0x%x", segno);
171 			sum_blk->footer.entry_type = SUM_TYPE_NODE;
172 		} else {
173 			ret = -EINVAL;
174 			goto out;
175 		}
176 	}
177 
178 	sum_entry = &(sum_blk->entries[offset]);
179 
180 	if (le32_to_cpu(sum_entry->nid) != nid) {
181 		if (!c.fix_on || type < 0) {
182 			DBG(0, "nid                       [0x%x]\n", nid);
183 			DBG(0, "target blk_addr           [0x%x]\n", blk_addr);
184 			DBG(0, "summary blk_addr          [0x%x]\n",
185 						GET_SUM_BLKADDR(sbi,
186 						GET_SEGNO(sbi, blk_addr)));
187 			DBG(0, "seg no / offset           [0x%x / 0x%x]\n",
188 						GET_SEGNO(sbi, blk_addr),
189 						OFFSET_IN_SEG(sbi, blk_addr));
190 			DBG(0, "summary_entry.nid         [0x%x]\n",
191 						le32_to_cpu(sum_entry->nid));
192 			DBG(0, "--> node block's nid      [0x%x]\n", nid);
193 			ASSERT_MSG("Invalid node seg summary\n");
194 			ret = -EINVAL;
195 		} else {
196 			FIX_MSG("Set node summary 0x%x -> [0x%x] [0x%x]",
197 						segno, nid, blk_addr);
198 			sum_entry->nid = cpu_to_le32(nid);
199 			need_fix = 1;
200 		}
201 	}
202 	if (need_fix && f2fs_dev_is_writable()) {
203 		u64 ssa_blk;
204 		int ret2;
205 
206 		ssa_blk = GET_SUM_BLKADDR(sbi, segno);
207 		ret2 = dev_write_block(sum_blk, ssa_blk);
208 		ASSERT(ret2 >= 0);
209 	}
210 out:
211 	if (type == SEG_TYPE_NODE || type == SEG_TYPE_DATA ||
212 					type == SEG_TYPE_MAX)
213 		free(sum_blk);
214 	return ret;
215 }
216 
is_valid_summary(struct f2fs_sb_info * sbi,struct f2fs_summary * sum,u32 blk_addr)217 static int is_valid_summary(struct f2fs_sb_info *sbi, struct f2fs_summary *sum,
218 							u32 blk_addr)
219 {
220 	u16 ofs_in_node = le16_to_cpu(sum->ofs_in_node);
221 	u32 nid = le32_to_cpu(sum->nid);
222 	struct f2fs_node *node_blk = NULL;
223 	__le32 target_blk_addr;
224 	struct node_info ni;
225 	int ret = 0;
226 
227 	node_blk = (struct f2fs_node *)calloc(BLOCK_SZ, 1);
228 	ASSERT(node_blk != NULL);
229 
230 	if (!IS_VALID_NID(sbi, nid))
231 		goto out;
232 
233 	get_node_info(sbi, nid, &ni);
234 
235 	if (!IS_VALID_BLK_ADDR(sbi, ni.blk_addr))
236 		goto out;
237 
238 	/* read node_block */
239 	ret = dev_read_block(node_blk, ni.blk_addr);
240 	ASSERT(ret >= 0);
241 
242 	if (le32_to_cpu(node_blk->footer.nid) != nid)
243 		goto out;
244 
245 	/* check its block address */
246 	if (node_blk->footer.nid == node_blk->footer.ino) {
247 		int ofs = get_extra_isize(node_blk);
248 
249 		target_blk_addr = node_blk->i.i_addr[ofs + ofs_in_node];
250 	} else {
251 		target_blk_addr = node_blk->dn.addr[ofs_in_node];
252 	}
253 
254 	if (blk_addr == le32_to_cpu(target_blk_addr))
255 		ret = 1;
256 out:
257 	free(node_blk);
258 	return ret;
259 }
260 
is_valid_ssa_data_blk(struct f2fs_sb_info * sbi,u32 blk_addr,u32 parent_nid,u16 idx_in_node,u8 version)261 static int is_valid_ssa_data_blk(struct f2fs_sb_info *sbi, u32 blk_addr,
262 		u32 parent_nid, u16 idx_in_node, u8 version)
263 {
264 	struct f2fs_summary_block *sum_blk;
265 	struct f2fs_summary *sum_entry;
266 	struct seg_entry * se;
267 	u32 segno, offset;
268 	int need_fix = 0, ret = 0;
269 	int type;
270 
271 	segno = GET_SEGNO(sbi, blk_addr);
272 	offset = OFFSET_IN_SEG(sbi, blk_addr);
273 
274 	sum_blk = get_sum_block(sbi, segno, &type);
275 
276 	if (type != SEG_TYPE_DATA && type != SEG_TYPE_CUR_DATA) {
277 		/* can't fix current summary, then drop the block */
278 		if (!c.fix_on || type < 0) {
279 			ASSERT_MSG("Summary footer is not for data segment");
280 			ret = -EINVAL;
281 			goto out;
282 		}
283 
284 		need_fix = 1;
285 		se = get_seg_entry(sbi, segno);
286 		if (IS_DATASEG(se->type)) {
287 			FIX_MSG("Summary footer indicates a data segment: 0x%x", segno);
288 			sum_blk->footer.entry_type = SUM_TYPE_DATA;
289 		} else {
290 			ret = -EINVAL;
291 			goto out;
292 		}
293 	}
294 
295 	sum_entry = &(sum_blk->entries[offset]);
296 
297 	if (le32_to_cpu(sum_entry->nid) != parent_nid ||
298 			sum_entry->version != version ||
299 			le16_to_cpu(sum_entry->ofs_in_node) != idx_in_node) {
300 		if (!c.fix_on || type < 0) {
301 			DBG(0, "summary_entry.nid         [0x%x]\n",
302 					le32_to_cpu(sum_entry->nid));
303 			DBG(0, "summary_entry.version     [0x%x]\n",
304 					sum_entry->version);
305 			DBG(0, "summary_entry.ofs_in_node [0x%x]\n",
306 					le16_to_cpu(sum_entry->ofs_in_node));
307 			DBG(0, "parent nid                [0x%x]\n",
308 					parent_nid);
309 			DBG(0, "version from nat          [0x%x]\n", version);
310 			DBG(0, "idx in parent node        [0x%x]\n",
311 					idx_in_node);
312 
313 			DBG(0, "Target data block addr    [0x%x]\n", blk_addr);
314 			ASSERT_MSG("Invalid data seg summary\n");
315 			ret = -EINVAL;
316 		} else if (is_valid_summary(sbi, sum_entry, blk_addr)) {
317 			/* delete wrong index */
318 			ret = -EINVAL;
319 		} else {
320 			FIX_MSG("Set data summary 0x%x -> [0x%x] [0x%x] [0x%x]",
321 					segno, parent_nid, version, idx_in_node);
322 			sum_entry->nid = cpu_to_le32(parent_nid);
323 			sum_entry->version = version;
324 			sum_entry->ofs_in_node = cpu_to_le16(idx_in_node);
325 			need_fix = 1;
326 		}
327 	}
328 	if (need_fix && f2fs_dev_is_writable()) {
329 		u64 ssa_blk;
330 		int ret2;
331 
332 		ssa_blk = GET_SUM_BLKADDR(sbi, segno);
333 		ret2 = dev_write_block(sum_blk, ssa_blk);
334 		ASSERT(ret2 >= 0);
335 	}
336 out:
337 	if (type == SEG_TYPE_NODE || type == SEG_TYPE_DATA ||
338 					type == SEG_TYPE_MAX)
339 		free(sum_blk);
340 	return ret;
341 }
342 
__check_inode_mode(u32 nid,enum FILE_TYPE ftype,u16 mode)343 static int __check_inode_mode(u32 nid, enum FILE_TYPE ftype, u16 mode)
344 {
345 	if (ftype >= F2FS_FT_MAX)
346 		return 0;
347 	/* f2fs_iget will return -EIO if mode is not valid file type */
348 	if (!S_ISLNK(mode) && !S_ISREG(mode) && !S_ISDIR(mode) &&
349 	    !S_ISCHR(mode) && !S_ISBLK(mode) && !S_ISFIFO(mode) &&
350 	    !S_ISSOCK(mode)) {
351 		ASSERT_MSG("inode [0x%x] unknown file type i_mode [0x%x]",
352 			   nid, mode);
353 		return -1;
354 	}
355 
356 	if (S_ISLNK(mode) && ftype != F2FS_FT_SYMLINK)
357 		goto err;
358 	if (S_ISREG(mode) && ftype != F2FS_FT_REG_FILE)
359 		goto err;
360 	if (S_ISDIR(mode) && ftype != F2FS_FT_DIR)
361 		goto err;
362 	if (S_ISCHR(mode) && ftype != F2FS_FT_CHRDEV)
363 		goto err;
364 	if (S_ISBLK(mode) && ftype != F2FS_FT_BLKDEV)
365 		goto err;
366 	if (S_ISFIFO(mode) && ftype != F2FS_FT_FIFO)
367 		goto err;
368 	if (S_ISSOCK(mode) && ftype != F2FS_FT_SOCK)
369 		goto err;
370 	return 0;
371 err:
372 	ASSERT_MSG("inode [0x%x] mismatch i_mode [0x%x vs. 0x%x]",
373 		   nid, ftype, mode);
374 	return -1;
375 }
376 
sanity_check_nid(struct f2fs_sb_info * sbi,u32 nid,struct f2fs_node * node_blk,enum FILE_TYPE ftype,enum NODE_TYPE ntype,struct node_info * ni)377 static int sanity_check_nid(struct f2fs_sb_info *sbi, u32 nid,
378 			struct f2fs_node *node_blk,
379 			enum FILE_TYPE ftype, enum NODE_TYPE ntype,
380 			struct node_info *ni)
381 {
382 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
383 	int ret;
384 
385 	if (!IS_VALID_NID(sbi, nid)) {
386 		ASSERT_MSG("nid is not valid. [0x%x]", nid);
387 		return -EINVAL;
388 	}
389 
390 	get_node_info(sbi, nid, ni);
391 	if (ni->ino == 0) {
392 		ASSERT_MSG("nid[0x%x] ino is 0", nid);
393 		return -EINVAL;
394 	}
395 
396 	if (ni->blk_addr == NEW_ADDR) {
397 		ASSERT_MSG("nid is NEW_ADDR. [0x%x]", nid);
398 		return -EINVAL;
399 	}
400 
401 	if (!IS_VALID_BLK_ADDR(sbi, ni->blk_addr)) {
402 		ASSERT_MSG("blkaddress is not valid. [0x%x]", ni->blk_addr);
403 		return -EINVAL;
404 	}
405 
406 	ret = dev_read_block(node_blk, ni->blk_addr);
407 	ASSERT(ret >= 0);
408 
409 	if (ntype == TYPE_INODE &&
410 			node_blk->footer.nid != node_blk->footer.ino) {
411 		ASSERT_MSG("nid[0x%x] footer.nid[0x%x] footer.ino[0x%x]",
412 				nid, le32_to_cpu(node_blk->footer.nid),
413 				le32_to_cpu(node_blk->footer.ino));
414 		return -EINVAL;
415 	}
416 	if (ni->ino != le32_to_cpu(node_blk->footer.ino)) {
417 		ASSERT_MSG("nid[0x%x] nat_entry->ino[0x%x] footer.ino[0x%x]",
418 				nid, ni->ino, le32_to_cpu(node_blk->footer.ino));
419 		return -EINVAL;
420 	}
421 	if (ntype != TYPE_INODE &&
422 			node_blk->footer.nid == node_blk->footer.ino) {
423 		ASSERT_MSG("nid[0x%x] footer.nid[0x%x] footer.ino[0x%x]",
424 				nid, le32_to_cpu(node_blk->footer.nid),
425 				le32_to_cpu(node_blk->footer.ino));
426 		return -EINVAL;
427 	}
428 
429 	if (le32_to_cpu(node_blk->footer.nid) != nid) {
430 		ASSERT_MSG("nid[0x%x] blk_addr[0x%x] footer.nid[0x%x]",
431 				nid, ni->blk_addr,
432 				le32_to_cpu(node_blk->footer.nid));
433 		return -EINVAL;
434 	}
435 
436 	if (ntype == TYPE_XATTR) {
437 		u32 flag = le32_to_cpu(node_blk->footer.flag);
438 
439 		if ((flag >> OFFSET_BIT_SHIFT) != XATTR_NODE_OFFSET) {
440 			ASSERT_MSG("xnid[0x%x] has wrong ofs:[0x%x]",
441 					nid, flag);
442 			return -EINVAL;
443 		}
444 	}
445 
446 	if ((ntype == TYPE_INODE && ftype == F2FS_FT_DIR) ||
447 			(ntype == TYPE_XATTR && ftype == F2FS_FT_XATTR)) {
448 		/* not included '.' & '..' */
449 		if (f2fs_test_main_bitmap(sbi, ni->blk_addr) != 0) {
450 			ASSERT_MSG("Duplicated node blk. nid[0x%x][0x%x]\n",
451 					nid, ni->blk_addr);
452 			return -EINVAL;
453 		}
454 	}
455 
456 	/* this if only from fix_hard_links */
457 	if (ftype == F2FS_FT_MAX)
458 		return 0;
459 
460 	if (ntype == TYPE_INODE &&
461 		__check_inode_mode(nid, ftype, le16_to_cpu(node_blk->i.i_mode)))
462 		return -EINVAL;
463 
464 	/* workaround to fix later */
465 	if (ftype != F2FS_FT_ORPHAN ||
466 			f2fs_test_bit(nid, fsck->nat_area_bitmap) != 0) {
467 		f2fs_clear_bit(nid, fsck->nat_area_bitmap);
468 		/* avoid reusing nid when reconnecting files */
469 		f2fs_set_bit(nid, NM_I(sbi)->nid_bitmap);
470 	} else
471 		ASSERT_MSG("orphan or xattr nid is duplicated [0x%x]\n",
472 				nid);
473 
474 	if (is_valid_ssa_node_blk(sbi, nid, ni->blk_addr)) {
475 		ASSERT_MSG("summary node block is not valid. [0x%x]", nid);
476 		return -EINVAL;
477 	}
478 
479 	if (f2fs_test_sit_bitmap(sbi, ni->blk_addr) == 0)
480 		ASSERT_MSG("SIT bitmap is 0x0. blk_addr[0x%x]",
481 				ni->blk_addr);
482 
483 	if (f2fs_test_main_bitmap(sbi, ni->blk_addr) == 0) {
484 		fsck->chk.valid_blk_cnt++;
485 		fsck->chk.valid_node_cnt++;
486 	}
487 	return 0;
488 }
489 
fsck_sanity_check_nid(struct f2fs_sb_info * sbi,u32 nid,struct f2fs_node * node_blk,enum FILE_TYPE ftype,enum NODE_TYPE ntype,struct node_info * ni)490 int fsck_sanity_check_nid(struct f2fs_sb_info *sbi, u32 nid,
491 			struct f2fs_node *node_blk,
492 			enum FILE_TYPE ftype, enum NODE_TYPE ntype,
493 			struct node_info *ni)
494 {
495 	return sanity_check_nid(sbi, nid, node_blk, ftype, ntype, ni);
496 }
497 
fsck_chk_xattr_blk(struct f2fs_sb_info * sbi,u32 ino,u32 x_nid,u32 * blk_cnt)498 static int fsck_chk_xattr_blk(struct f2fs_sb_info *sbi, u32 ino,
499 					u32 x_nid, u32 *blk_cnt)
500 {
501 	struct f2fs_node *node_blk = NULL;
502 	struct node_info ni;
503 	int ret = 0;
504 
505 	if (x_nid == 0x0)
506 		return 0;
507 
508 	node_blk = (struct f2fs_node *)calloc(BLOCK_SZ, 1);
509 	ASSERT(node_blk != NULL);
510 
511 	/* Sanity check */
512 	if (sanity_check_nid(sbi, x_nid, node_blk,
513 				F2FS_FT_XATTR, TYPE_XATTR, &ni)) {
514 		ret = -EINVAL;
515 		goto out;
516 	}
517 
518 	*blk_cnt = *blk_cnt + 1;
519 	f2fs_set_main_bitmap(sbi, ni.blk_addr, CURSEG_COLD_NODE);
520 	DBG(2, "ino[0x%x] x_nid[0x%x]\n", ino, x_nid);
521 out:
522 	free(node_blk);
523 	return ret;
524 }
525 
fsck_chk_node_blk(struct f2fs_sb_info * sbi,struct f2fs_inode * inode,u32 nid,enum FILE_TYPE ftype,enum NODE_TYPE ntype,u32 * blk_cnt,struct child_info * child)526 int fsck_chk_node_blk(struct f2fs_sb_info *sbi, struct f2fs_inode *inode,
527 		u32 nid, enum FILE_TYPE ftype, enum NODE_TYPE ntype,
528 		u32 *blk_cnt, struct child_info *child)
529 {
530 	struct node_info ni;
531 	struct f2fs_node *node_blk = NULL;
532 
533 	node_blk = (struct f2fs_node *)calloc(BLOCK_SZ, 1);
534 	ASSERT(node_blk != NULL);
535 
536 	if (sanity_check_nid(sbi, nid, node_blk, ftype, ntype, &ni))
537 		goto err;
538 
539 	if (ntype == TYPE_INODE) {
540 		struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
541 
542 		fsck_chk_inode_blk(sbi, nid, ftype, node_blk, blk_cnt, &ni, child);
543 		quota_add_inode_usage(fsck->qctx, nid, &node_blk->i);
544 	} else {
545 		switch (ntype) {
546 		case TYPE_DIRECT_NODE:
547 			f2fs_set_main_bitmap(sbi, ni.blk_addr,
548 							CURSEG_WARM_NODE);
549 			fsck_chk_dnode_blk(sbi, inode, nid, ftype, node_blk,
550 					blk_cnt, child, &ni);
551 			break;
552 		case TYPE_INDIRECT_NODE:
553 			f2fs_set_main_bitmap(sbi, ni.blk_addr,
554 							CURSEG_COLD_NODE);
555 			fsck_chk_idnode_blk(sbi, inode, ftype, node_blk,
556 					blk_cnt, child);
557 			break;
558 		case TYPE_DOUBLE_INDIRECT_NODE:
559 			f2fs_set_main_bitmap(sbi, ni.blk_addr,
560 							CURSEG_COLD_NODE);
561 			fsck_chk_didnode_blk(sbi, inode, ftype, node_blk,
562 					blk_cnt, child);
563 			break;
564 		default:
565 			ASSERT(0);
566 		}
567 	}
568 	free(node_blk);
569 	return 0;
570 err:
571 	free(node_blk);
572 	return -EINVAL;
573 }
574 
get_extent_info(struct extent_info * ext,struct f2fs_extent * i_ext)575 static inline void get_extent_info(struct extent_info *ext,
576 					struct f2fs_extent *i_ext)
577 {
578 	ext->fofs = le32_to_cpu(i_ext->fofs);
579 	ext->blk = le32_to_cpu(i_ext->blk_addr);
580 	ext->len = le32_to_cpu(i_ext->len);
581 }
582 
check_extent_info(struct child_info * child,block_t blkaddr,int last)583 static void check_extent_info(struct child_info *child,
584 						block_t blkaddr, int last)
585 {
586 	struct extent_info *ei = &child->ei;
587 	u32 pgofs = child->pgofs;
588 	int is_hole = 0;
589 
590 	if (!ei->len)
591 		return;
592 
593 	if (child->state & FSCK_UNMATCHED_EXTENT)
594 		return;
595 
596 	if ((child->state & FSCK_INLINE_INODE) && ei->len)
597 		goto unmatched;
598 
599 	if (last) {
600 		/* hole exist in the back of extent */
601 		if (child->last_blk != ei->blk + ei->len - 1)
602 			child->state |= FSCK_UNMATCHED_EXTENT;
603 		return;
604 	}
605 
606 	if (blkaddr == NULL_ADDR || blkaddr == NEW_ADDR)
607 		is_hole = 1;
608 
609 	if (pgofs >= ei->fofs && pgofs < ei->fofs + ei->len) {
610 		/* unmatched blkaddr */
611 		if (is_hole || (blkaddr != pgofs - ei->fofs + ei->blk))
612 			goto unmatched;
613 
614 		if (!child->last_blk) {
615 			/* hole exists in the front of extent */
616 			if (pgofs != ei->fofs)
617 				goto unmatched;
618 		} else if (child->last_blk + 1 != blkaddr) {
619 			/* hole exists in the middle of extent */
620 			goto unmatched;
621 		}
622 		child->last_blk = blkaddr;
623 		return;
624 	}
625 
626 	if (is_hole)
627 		return;
628 
629 	if (blkaddr < ei->blk || blkaddr >= ei->blk + ei->len)
630 		return;
631 	/* unmatched file offset */
632 unmatched:
633 	child->state |= FSCK_UNMATCHED_EXTENT;
634 }
635 
fsck_reada_node_block(struct f2fs_sb_info * sbi,u32 nid)636 void fsck_reada_node_block(struct f2fs_sb_info *sbi, u32 nid)
637 {
638 	struct node_info ni;
639 
640 	if (nid != 0 && IS_VALID_NID(sbi, nid)) {
641 		get_node_info(sbi, nid, &ni);
642 		if (IS_VALID_BLK_ADDR(sbi, ni.blk_addr))
643 			dev_reada_block(ni.blk_addr);
644 	}
645 }
646 
fsck_reada_all_direct_node_blocks(struct f2fs_sb_info * sbi,struct f2fs_node * node_blk)647 void fsck_reada_all_direct_node_blocks(struct f2fs_sb_info *sbi,
648 						struct f2fs_node *node_blk)
649 {
650 	int i;
651 
652 	for (i = 0; i < NIDS_PER_BLOCK; i++) {
653 		u32 nid = le32_to_cpu(node_blk->in.nid[i]);
654 
655 		fsck_reada_node_block(sbi, nid);
656 	}
657 }
658 
659 /* start with valid nid and blkaddr */
fsck_chk_inode_blk(struct f2fs_sb_info * sbi,u32 nid,enum FILE_TYPE ftype,struct f2fs_node * node_blk,u32 * blk_cnt,struct node_info * ni,struct child_info * child_d)660 void fsck_chk_inode_blk(struct f2fs_sb_info *sbi, u32 nid,
661 		enum FILE_TYPE ftype, struct f2fs_node *node_blk,
662 		u32 *blk_cnt, struct node_info *ni, struct child_info *child_d)
663 {
664 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
665 	struct child_info child;
666 	enum NODE_TYPE ntype;
667 	u32 i_links = le32_to_cpu(node_blk->i.i_links);
668 	u64 i_size = le64_to_cpu(node_blk->i.i_size);
669 	u64 i_blocks = le64_to_cpu(node_blk->i.i_blocks);
670 	nid_t i_xattr_nid = le32_to_cpu(node_blk->i.i_xattr_nid);
671 	int ofs;
672 	char *en;
673 	u32 namelen;
674 	unsigned int idx = 0;
675 	unsigned short i_gc_failures;
676 	int need_fix = 0;
677 	int ret;
678 
679 	memset(&child, 0, sizeof(child));
680 	child.links = 2;
681 	child.p_ino = nid;
682 	child.pp_ino = le32_to_cpu(node_blk->i.i_pino);
683 	child.dir_level = node_blk->i.i_dir_level;
684 
685 	if (f2fs_test_main_bitmap(sbi, ni->blk_addr) == 0)
686 		fsck->chk.valid_inode_cnt++;
687 
688 	if (ftype == F2FS_FT_DIR) {
689 		f2fs_set_main_bitmap(sbi, ni->blk_addr, CURSEG_HOT_NODE);
690 	} else {
691 		if (f2fs_test_main_bitmap(sbi, ni->blk_addr) == 0) {
692 			f2fs_set_main_bitmap(sbi, ni->blk_addr,
693 							CURSEG_WARM_NODE);
694 			if (i_links > 1 && ftype != F2FS_FT_ORPHAN &&
695 					!is_qf_ino(F2FS_RAW_SUPER(sbi), nid)) {
696 				/* First time. Create new hard link node */
697 				add_into_hard_link_list(sbi, nid, i_links);
698 				fsck->chk.multi_hard_link_files++;
699 			}
700 		} else {
701 			DBG(3, "[0x%x] has hard links [0x%x]\n", nid, i_links);
702 			if (find_and_dec_hard_link_list(sbi, nid)) {
703 				ASSERT_MSG("[0x%x] needs more i_links=0x%x",
704 						nid, i_links);
705 				if (c.fix_on) {
706 					node_blk->i.i_links =
707 						cpu_to_le32(i_links + 1);
708 					need_fix = 1;
709 					FIX_MSG("File: 0x%x "
710 						"i_links= 0x%x -> 0x%x",
711 						nid, i_links, i_links + 1);
712 				}
713 				goto skip_blkcnt_fix;
714 			}
715 			/* No need to go deep into the node */
716 			return;
717 		}
718 	}
719 
720 	/* readahead xattr node block */
721 	fsck_reada_node_block(sbi, i_xattr_nid);
722 
723 	if (fsck_chk_xattr_blk(sbi, nid, i_xattr_nid, blk_cnt)) {
724 		if (c.fix_on) {
725 			node_blk->i.i_xattr_nid = 0;
726 			need_fix = 1;
727 			FIX_MSG("Remove xattr block: 0x%x, x_nid = 0x%x",
728 							nid, i_xattr_nid);
729 		}
730 	}
731 
732 	if (ftype == F2FS_FT_CHRDEV || ftype == F2FS_FT_BLKDEV ||
733 			ftype == F2FS_FT_FIFO || ftype == F2FS_FT_SOCK)
734 		goto check;
735 
736 	/* init extent info */
737 	get_extent_info(&child.ei, &node_blk->i.i_ext);
738 	child.last_blk = 0;
739 
740 	if (f2fs_has_extra_isize(&node_blk->i)) {
741 		if (c.feature & cpu_to_le32(F2FS_FEATURE_EXTRA_ATTR)) {
742 			unsigned int isize =
743 				le16_to_cpu(node_blk->i.i_extra_isize);
744 			if (isize > 4 * DEF_ADDRS_PER_INODE) {
745 				ASSERT_MSG("[0x%x] wrong i_extra_isize=0x%x",
746 						nid, isize);
747 				if (c.fix_on) {
748 					FIX_MSG("ino[0x%x] recover i_extra_isize "
749 						"from %u to %u",
750 						nid, isize,
751 						calc_extra_isize());
752 					node_blk->i.i_extra_isize =
753 						cpu_to_le16(calc_extra_isize());
754 					need_fix = 1;
755 				}
756 			}
757 		} else {
758 			ASSERT_MSG("[0x%x] wrong extra_attr flag", nid);
759 			if (c.fix_on) {
760 				FIX_MSG("ino[0x%x] remove F2FS_EXTRA_ATTR "
761 					"flag in i_inline:%u",
762 					nid, node_blk->i.i_inline);
763 				/* we don't support tuning F2FS_FEATURE_EXTRA_ATTR now */
764 				node_blk->i.i_inline &= ~F2FS_EXTRA_ATTR;
765 				need_fix = 1;
766 			}
767 		}
768 
769 		if ((c.feature &
770 			cpu_to_le32(F2FS_FEATURE_FLEXIBLE_INLINE_XATTR)) &&
771 			(node_blk->i.i_inline & F2FS_INLINE_XATTR)) {
772 			unsigned int inline_size =
773 				le16_to_cpu(node_blk->i.i_inline_xattr_size);
774 
775 			if (!inline_size ||
776 					inline_size > MAX_INLINE_XATTR_SIZE) {
777 				ASSERT_MSG("[0x%x] wrong inline_xattr_size:%u",
778 						nid, inline_size);
779 				if (c.fix_on) {
780 					FIX_MSG("ino[0x%x] recover inline xattr size "
781 						"from %u to %u",
782 						nid, inline_size,
783 						DEFAULT_INLINE_XATTR_ADDRS);
784 					node_blk->i.i_inline_xattr_size =
785 						cpu_to_le16(DEFAULT_INLINE_XATTR_ADDRS);
786 					need_fix = 1;
787 				}
788 			}
789 		}
790 	}
791 	ofs = get_extra_isize(node_blk);
792 
793 	if ((node_blk->i.i_inline & F2FS_INLINE_DATA)) {
794 		unsigned int inline_size = MAX_INLINE_DATA(node_blk);
795 		if (cur_qtype != -1)
796 			qf_szchk_type[cur_qtype] = QF_SZCHK_INLINE;
797 		block_t blkaddr = le32_to_cpu(node_blk->i.i_addr[ofs]);
798 
799 		if (blkaddr != 0) {
800 			ASSERT_MSG("[0x%x] wrong inline reserve blkaddr:%u",
801 					nid, blkaddr);
802 			if (c.fix_on) {
803 				FIX_MSG("inline_data has wrong 0'th block = %x",
804 								blkaddr);
805 				node_blk->i.i_addr[ofs] = 0;
806 				node_blk->i.i_blocks = cpu_to_le64(*blk_cnt);
807 				need_fix = 1;
808 			}
809 		}
810 		if (i_size > inline_size) {
811 			ASSERT_MSG("[0x%x] wrong inline size:%lu",
812 					nid, (unsigned long)i_size);
813 			if (c.fix_on) {
814 				node_blk->i.i_size = cpu_to_le64(inline_size);
815 				FIX_MSG("inline_data has wrong i_size %lu",
816 							(unsigned long)i_size);
817 				need_fix = 1;
818 			}
819 		}
820 		if (!(node_blk->i.i_inline & F2FS_DATA_EXIST)) {
821 			char buf[MAX_INLINE_DATA(node_blk)];
822 			memset(buf, 0, MAX_INLINE_DATA(node_blk));
823 
824 			if (memcmp(buf, inline_data_addr(node_blk),
825 						MAX_INLINE_DATA(node_blk))) {
826 				ASSERT_MSG("[0x%x] junk inline data", nid);
827 				if (c.fix_on) {
828 					FIX_MSG("inline_data has DATA_EXIST");
829 					node_blk->i.i_inline |= F2FS_DATA_EXIST;
830 					need_fix = 1;
831 				}
832 			}
833 		}
834 		DBG(3, "ino[0x%x] has inline data!\n", nid);
835 		child.state |= FSCK_INLINE_INODE;
836 		goto check;
837 	}
838 
839 	if ((node_blk->i.i_inline & F2FS_INLINE_DENTRY)) {
840 		block_t blkaddr = le32_to_cpu(node_blk->i.i_addr[ofs]);
841 
842 		DBG(3, "ino[0x%x] has inline dentry!\n", nid);
843 		if (blkaddr != 0) {
844 			ASSERT_MSG("[0x%x] wrong inline reserve blkaddr:%u",
845 								nid, blkaddr);
846 			if (c.fix_on) {
847 				FIX_MSG("inline_dentry has wrong 0'th block = %x",
848 								blkaddr);
849 				node_blk->i.i_addr[ofs] = 0;
850 				node_blk->i.i_blocks = cpu_to_le64(*blk_cnt);
851 				need_fix = 1;
852 			}
853 		}
854 
855 		ret = fsck_chk_inline_dentries(sbi, node_blk, &child);
856 		if (ret < 0) {
857 			if (c.fix_on)
858 				need_fix = 1;
859 		}
860 		child.state |= FSCK_INLINE_INODE;
861 		goto check;
862 	}
863 
864 	/* check data blocks in inode */
865 	if (cur_qtype != -1) {
866 		qf_szchk_type[cur_qtype] = QF_SZCHK_REGFILE;
867 		qf_maxsize[cur_qtype] = (ADDRS_PER_INODE(&node_blk->i) +
868 				2 * ADDRS_PER_BLOCK(&node_blk->i) +
869 				2 * ADDRS_PER_BLOCK(&node_blk->i) *
870 				NIDS_PER_BLOCK +
871 				(u64) ADDRS_PER_BLOCK(&node_blk->i) *
872 				NIDS_PER_BLOCK * NIDS_PER_BLOCK) * F2FS_BLKSIZE;
873 	}
874 	for (idx = 0; idx < ADDRS_PER_INODE(&node_blk->i);
875 						idx++, child.pgofs++) {
876 		block_t blkaddr = le32_to_cpu(node_blk->i.i_addr[ofs + idx]);
877 
878 		/* check extent info */
879 		check_extent_info(&child, blkaddr, 0);
880 
881 		if (blkaddr == COMPRESS_ADDR) {
882 			if (node_blk->i.i_compr_blocks) {
883 				fsck->chk.valid_blk_cnt++;
884 				*blk_cnt = *blk_cnt + 1;
885 			}
886 			continue;
887 		}
888 
889 		if (blkaddr != 0) {
890 			ret = fsck_chk_data_blk(sbi,
891 					IS_CASEFOLDED(&node_blk->i),
892 					blkaddr,
893 					&child, (i_blocks == *blk_cnt),
894 					ftype, nid, idx, ni->version,
895 					file_is_encrypt(&node_blk->i));
896 			if (!ret) {
897 				*blk_cnt = *blk_cnt + 1;
898 				if (cur_qtype != -1 && blkaddr != NEW_ADDR)
899 					qf_last_blkofs[cur_qtype] = child.pgofs;
900 			} else if (c.fix_on) {
901 				node_blk->i.i_addr[ofs + idx] = 0;
902 				need_fix = 1;
903 				FIX_MSG("[0x%x] i_addr[%d] = 0",
904 							nid, ofs + idx);
905 			}
906 		}
907 	}
908 
909 	/* readahead node blocks */
910 	for (idx = 0; idx < 5; idx++) {
911 		u32 nid = le32_to_cpu(node_blk->i.i_nid[idx]);
912 		fsck_reada_node_block(sbi, nid);
913 	}
914 
915 	/* check node blocks in inode */
916 	for (idx = 0; idx < 5; idx++) {
917 		nid_t i_nid = le32_to_cpu(node_blk->i.i_nid[idx]);
918 
919 		if (idx == 0 || idx == 1)
920 			ntype = TYPE_DIRECT_NODE;
921 		else if (idx == 2 || idx == 3)
922 			ntype = TYPE_INDIRECT_NODE;
923 		else if (idx == 4)
924 			ntype = TYPE_DOUBLE_INDIRECT_NODE;
925 		else
926 			ASSERT(0);
927 
928 		if (i_nid == 0x0)
929 			goto skip;
930 
931 		ret = fsck_chk_node_blk(sbi, &node_blk->i, i_nid,
932 					ftype, ntype, blk_cnt, &child);
933 		if (!ret) {
934 			*blk_cnt = *blk_cnt + 1;
935 		} else if (ret == -EINVAL) {
936 			if (c.fix_on) {
937 				node_blk->i.i_nid[idx] = 0;
938 				need_fix = 1;
939 				FIX_MSG("[0x%x] i_nid[%d] = 0", nid, idx);
940 			}
941 skip:
942 			if (ntype == TYPE_DIRECT_NODE)
943 				child.pgofs += ADDRS_PER_BLOCK(&node_blk->i);
944 			else if (ntype == TYPE_INDIRECT_NODE)
945 				child.pgofs += ADDRS_PER_BLOCK(&node_blk->i) *
946 								NIDS_PER_BLOCK;
947 			else
948 				child.pgofs += ADDRS_PER_BLOCK(&node_blk->i) *
949 						NIDS_PER_BLOCK * NIDS_PER_BLOCK;
950 		}
951 
952 	}
953 
954 check:
955 	/* check uncovered range in the back of extent */
956 	check_extent_info(&child, 0, 1);
957 
958 	if (child.state & FSCK_UNMATCHED_EXTENT) {
959 		ASSERT_MSG("ino: 0x%x has wrong ext: [pgofs:%u, blk:%u, len:%u]",
960 				nid, child.ei.fofs, child.ei.blk, child.ei.len);
961 		if (c.fix_on)
962 			need_fix = 1;
963 	}
964 
965 	if (i_blocks != *blk_cnt) {
966 		ASSERT_MSG("ino: 0x%x has i_blocks: %08"PRIx64", "
967 				"but has %u blocks",
968 				nid, i_blocks, *blk_cnt);
969 		if (c.fix_on) {
970 			node_blk->i.i_blocks = cpu_to_le64(*blk_cnt);
971 			need_fix = 1;
972 			FIX_MSG("[0x%x] i_blocks=0x%08"PRIx64" -> 0x%x",
973 					nid, i_blocks, *blk_cnt);
974 		}
975 	}
976 skip_blkcnt_fix:
977 	en = malloc(F2FS_PRINT_NAMELEN);
978 	ASSERT(en);
979 
980 	namelen = le32_to_cpu(node_blk->i.i_namelen);
981 	if (namelen > F2FS_NAME_LEN) {
982 		if (child_d && child_d->i_namelen <= F2FS_NAME_LEN) {
983 			ASSERT_MSG("ino: 0x%x has i_namelen: 0x%x, "
984 					"but has %d characters for name",
985 					nid, namelen, child_d->i_namelen);
986 			if (c.fix_on) {
987 				FIX_MSG("[0x%x] i_namelen=0x%x -> 0x%x", nid, namelen,
988 					child_d->i_namelen);
989 				node_blk->i.i_namelen = cpu_to_le32(child_d->i_namelen);
990 				need_fix = 1;
991 			}
992 			namelen = child_d->i_namelen;
993 		} else
994 			namelen = F2FS_NAME_LEN;
995 	}
996 	pretty_print_filename(node_blk->i.i_name, namelen, en,
997 			      file_enc_name(&node_blk->i));
998 	if (ftype == F2FS_FT_ORPHAN)
999 		DBG(1, "Orphan Inode: 0x%x [%s] i_blocks: %u\n\n",
1000 				le32_to_cpu(node_blk->footer.ino),
1001 				en, (u32)i_blocks);
1002 
1003 	if (is_qf_ino(F2FS_RAW_SUPER(sbi), nid))
1004 		DBG(1, "Quota Inode: 0x%x [%s] i_blocks: %u\n\n",
1005 				le32_to_cpu(node_blk->footer.ino),
1006 				en, (u32)i_blocks);
1007 
1008 	if (ftype == F2FS_FT_DIR) {
1009 		DBG(1, "Directory Inode: 0x%x [%s] depth: %d has %d files\n\n",
1010 				le32_to_cpu(node_blk->footer.ino), en,
1011 				le32_to_cpu(node_blk->i.i_current_depth),
1012 				child.files);
1013 
1014 		if (i_links != child.links) {
1015 			ASSERT_MSG("ino: 0x%x i_links: %u, real links: %u",
1016 					nid, i_links, child.links);
1017 			if (c.fix_on) {
1018 				node_blk->i.i_links = cpu_to_le32(child.links);
1019 				need_fix = 1;
1020 				FIX_MSG("Dir: 0x%x i_links= 0x%x -> 0x%x",
1021 						nid, i_links, child.links);
1022 			}
1023 		}
1024 		if (child.dots < 2 &&
1025 				!(node_blk->i.i_inline & F2FS_INLINE_DOTS)) {
1026 			ASSERT_MSG("ino: 0x%x dots: %u",
1027 					nid, child.dots);
1028 			if (c.fix_on) {
1029 				node_blk->i.i_inline |= F2FS_INLINE_DOTS;
1030 				need_fix = 1;
1031 				FIX_MSG("Dir: 0x%x set inline_dots", nid);
1032 			}
1033 		}
1034 	}
1035 
1036 	i_gc_failures = le16_to_cpu(node_blk->i.i_gc_failures);
1037 
1038 	/*
1039 	 * old kernel initialized i_gc_failures as 0x01, in preen mode 2,
1040 	 * let's skip repairing.
1041 	 */
1042 	if (ftype == F2FS_FT_REG_FILE && i_gc_failures &&
1043 		(c.preen_mode != PREEN_MODE_2 || i_gc_failures != 0x01)) {
1044 
1045 		DBG(1, "Regular Inode: 0x%x [%s] depth: %d\n\n",
1046 				le32_to_cpu(node_blk->footer.ino), en,
1047 				i_gc_failures);
1048 
1049 		if (c.fix_on) {
1050 			node_blk->i.i_gc_failures = cpu_to_le16(0);
1051 			need_fix = 1;
1052 			FIX_MSG("Regular: 0x%x reset i_gc_failures from 0x%x to 0x00",
1053 					nid, i_gc_failures);
1054 		}
1055 	}
1056 
1057 	free(en);
1058 
1059 	if (ftype == F2FS_FT_SYMLINK && i_size == 0 &&
1060 			i_blocks == (i_xattr_nid ? 3 : 2)) {
1061 		ASSERT_MSG("ino: 0x%x i_blocks: %lu with zero i_size\n",
1062 						nid, (unsigned long)i_blocks);
1063 		if (c.fix_on) {
1064 			node_blk->i.i_size = cpu_to_le64(F2FS_BLKSIZE);
1065 			need_fix = 1;
1066 			FIX_MSG("Symlink: recover 0x%x with i_size=%lu",
1067 					nid, (unsigned long)F2FS_BLKSIZE);
1068 		}
1069 	}
1070 
1071 	if (ftype == F2FS_FT_ORPHAN && i_links) {
1072 		ASSERT_MSG("ino: 0x%x is orphan inode, but has i_links: %u",
1073 				nid, i_links);
1074 		if (c.fix_on) {
1075 			node_blk->i.i_links = 0;
1076 			need_fix = 1;
1077 			FIX_MSG("ino: 0x%x orphan_inode, i_links= 0x%x -> 0",
1078 					nid, i_links);
1079 		}
1080 	}
1081 
1082 	/* drop extent information to avoid potential wrong access */
1083 	if (need_fix && f2fs_dev_is_writable())
1084 		node_blk->i.i_ext.len = 0;
1085 
1086 	if ((c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM)) &&
1087 				f2fs_has_extra_isize(&node_blk->i)) {
1088 		__u32 provided, calculated;
1089 
1090 		provided = le32_to_cpu(node_blk->i.i_inode_checksum);
1091 		calculated = f2fs_inode_chksum(node_blk);
1092 
1093 		if (provided != calculated) {
1094 			ASSERT_MSG("ino: 0x%x chksum:0x%x, but calculated one is: 0x%x",
1095 				nid, provided, calculated);
1096 			if (c.fix_on) {
1097 				node_blk->i.i_inode_checksum =
1098 							cpu_to_le32(calculated);
1099 				need_fix = 1;
1100 				FIX_MSG("ino: 0x%x recover, i_inode_checksum= 0x%x -> 0x%x",
1101 						nid, provided, calculated);
1102 			}
1103 		}
1104 	}
1105 
1106 	if (need_fix && f2fs_dev_is_writable()) {
1107 		ret = dev_write_block(node_blk, ni->blk_addr);
1108 		ASSERT(ret >= 0);
1109 	}
1110 }
1111 
fsck_chk_dnode_blk(struct f2fs_sb_info * sbi,struct f2fs_inode * inode,u32 nid,enum FILE_TYPE ftype,struct f2fs_node * node_blk,u32 * blk_cnt,struct child_info * child,struct node_info * ni)1112 int fsck_chk_dnode_blk(struct f2fs_sb_info *sbi, struct f2fs_inode *inode,
1113 		u32 nid, enum FILE_TYPE ftype, struct f2fs_node *node_blk,
1114 		u32 *blk_cnt, struct child_info *child, struct node_info *ni)
1115 {
1116 	int idx, ret;
1117 	int need_fix = 0;
1118 	child->p_ino = nid;
1119 	child->pp_ino = le32_to_cpu(inode->i_pino);
1120 
1121 	for (idx = 0; idx < ADDRS_PER_BLOCK(inode); idx++, child->pgofs++) {
1122 		block_t blkaddr = le32_to_cpu(node_blk->dn.addr[idx]);
1123 
1124 		check_extent_info(child, blkaddr, 0);
1125 
1126 		if (blkaddr == 0x0)
1127 			continue;
1128 		if (blkaddr == COMPRESS_ADDR) {
1129 			if (inode->i_compr_blocks) {
1130 				F2FS_FSCK(sbi)->chk.valid_blk_cnt++;
1131 				*blk_cnt = *blk_cnt + 1;
1132 			}
1133 			continue;
1134 		}
1135 		ret = fsck_chk_data_blk(sbi, IS_CASEFOLDED(inode),
1136 			blkaddr, child,
1137 			le64_to_cpu(inode->i_blocks) == *blk_cnt, ftype,
1138 			nid, idx, ni->version,
1139 			file_is_encrypt(inode));
1140 		if (!ret) {
1141 			*blk_cnt = *blk_cnt + 1;
1142 			if (cur_qtype != -1 && blkaddr != NEW_ADDR)
1143 				qf_last_blkofs[cur_qtype] = child->pgofs;
1144 		} else if (c.fix_on) {
1145 			node_blk->dn.addr[idx] = 0;
1146 			need_fix = 1;
1147 			FIX_MSG("[0x%x] dn.addr[%d] = 0", nid, idx);
1148 		}
1149 	}
1150 	if (need_fix && f2fs_dev_is_writable()) {
1151 		ret = dev_write_block(node_blk, ni->blk_addr);
1152 		ASSERT(ret >= 0);
1153 	}
1154 	return 0;
1155 }
1156 
fsck_chk_idnode_blk(struct f2fs_sb_info * sbi,struct f2fs_inode * inode,enum FILE_TYPE ftype,struct f2fs_node * node_blk,u32 * blk_cnt,struct child_info * child)1157 int fsck_chk_idnode_blk(struct f2fs_sb_info *sbi, struct f2fs_inode *inode,
1158 		enum FILE_TYPE ftype, struct f2fs_node *node_blk, u32 *blk_cnt,
1159 		struct child_info *child)
1160 {
1161 	int need_fix = 0, ret;
1162 	int i = 0;
1163 
1164 	fsck_reada_all_direct_node_blocks(sbi, node_blk);
1165 
1166 	for (i = 0; i < NIDS_PER_BLOCK; i++) {
1167 		if (le32_to_cpu(node_blk->in.nid[i]) == 0x0)
1168 			goto skip;
1169 		ret = fsck_chk_node_blk(sbi, inode,
1170 				le32_to_cpu(node_blk->in.nid[i]),
1171 				ftype, TYPE_DIRECT_NODE, blk_cnt, child);
1172 		if (!ret)
1173 			*blk_cnt = *blk_cnt + 1;
1174 		else if (ret == -EINVAL) {
1175 			if (!c.fix_on)
1176 				printf("should delete in.nid[i] = 0;\n");
1177 			else {
1178 				node_blk->in.nid[i] = 0;
1179 				need_fix = 1;
1180 				FIX_MSG("Set indirect node 0x%x -> 0", i);
1181 			}
1182 skip:
1183 			child->pgofs += ADDRS_PER_BLOCK(&node_blk->i);
1184 		}
1185 	}
1186 
1187 	if (need_fix && f2fs_dev_is_writable()) {
1188 		struct node_info ni;
1189 		nid_t nid = le32_to_cpu(node_blk->footer.nid);
1190 
1191 		get_node_info(sbi, nid, &ni);
1192 		ret = dev_write_block(node_blk, ni.blk_addr);
1193 		ASSERT(ret >= 0);
1194 	}
1195 
1196 	return 0;
1197 }
1198 
fsck_chk_didnode_blk(struct f2fs_sb_info * sbi,struct f2fs_inode * inode,enum FILE_TYPE ftype,struct f2fs_node * node_blk,u32 * blk_cnt,struct child_info * child)1199 int fsck_chk_didnode_blk(struct f2fs_sb_info *sbi, struct f2fs_inode *inode,
1200 		enum FILE_TYPE ftype, struct f2fs_node *node_blk, u32 *blk_cnt,
1201 		struct child_info *child)
1202 {
1203 	int i = 0;
1204 	int need_fix = 0, ret = 0;
1205 
1206 	fsck_reada_all_direct_node_blocks(sbi, node_blk);
1207 
1208 	for (i = 0; i < NIDS_PER_BLOCK; i++) {
1209 		if (le32_to_cpu(node_blk->in.nid[i]) == 0x0)
1210 			goto skip;
1211 		ret = fsck_chk_node_blk(sbi, inode,
1212 				le32_to_cpu(node_blk->in.nid[i]),
1213 				ftype, TYPE_INDIRECT_NODE, blk_cnt, child);
1214 		if (!ret)
1215 			*blk_cnt = *blk_cnt + 1;
1216 		else if (ret == -EINVAL) {
1217 			if (!c.fix_on)
1218 				printf("should delete in.nid[i] = 0;\n");
1219 			else {
1220 				node_blk->in.nid[i] = 0;
1221 				need_fix = 1;
1222 				FIX_MSG("Set double indirect node 0x%x -> 0", i);
1223 			}
1224 skip:
1225 			child->pgofs += ADDRS_PER_BLOCK(&node_blk->i) *
1226 							NIDS_PER_BLOCK;
1227 		}
1228 	}
1229 
1230 	if (need_fix && f2fs_dev_is_writable()) {
1231 		struct node_info ni;
1232 		nid_t nid = le32_to_cpu(node_blk->footer.nid);
1233 
1234 		get_node_info(sbi, nid, &ni);
1235 		ret = dev_write_block(node_blk, ni.blk_addr);
1236 		ASSERT(ret >= 0);
1237 	}
1238 
1239 	return 0;
1240 }
1241 
1242 static const char *lookup_table =
1243         "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+,";
1244 
1245 /**
1246  * base64_encode() -
1247  *
1248  * Encodes the input string using characters from the set [A-Za-z0-9+,].
1249  * The encoded string is roughly 4/3 times the size of the input string.
1250  */
base64_encode(const u8 * src,int len,char * dst)1251 static int base64_encode(const u8 *src, int len, char *dst)
1252 {
1253 	int i, bits = 0, ac = 0;
1254 	char *cp = dst;
1255 
1256 	for (i = 0; i < len; i++) {
1257 		ac += src[i] << bits;
1258 		bits += 8;
1259 		do {
1260 			*cp++ = lookup_table[ac & 0x3f];
1261 			ac >>= 6;
1262 			bits -= 6;
1263 		} while (bits >= 6);
1264 	}
1265 	if (bits)
1266 		*cp++ = lookup_table[ac & 0x3f];
1267 	return cp - dst;
1268 }
1269 
pretty_print_filename(const u8 * raw_name,u32 len,char out[F2FS_PRINT_NAMELEN],int enc_name)1270 void pretty_print_filename(const u8 *raw_name, u32 len,
1271 			   char out[F2FS_PRINT_NAMELEN], int enc_name)
1272 {
1273 	len = min(len, (u32)F2FS_NAME_LEN);
1274 
1275 	if (enc_name)
1276 		len = base64_encode(raw_name, len, out);
1277 	else
1278 		memcpy(out, raw_name, len);
1279 	out[len] = 0;
1280 }
1281 
print_dentry(__u32 depth,__u8 * name,u8 * bitmap,struct f2fs_dir_entry * dentry,int max,int idx,int last_blk,int enc_name)1282 static void print_dentry(__u32 depth, __u8 *name,
1283 		u8 *bitmap, struct f2fs_dir_entry *dentry,
1284 		int max, int idx, int last_blk, int enc_name)
1285 {
1286 	int last_de = 0;
1287 	int next_idx = 0;
1288 	u32 name_len;
1289 	unsigned int i;
1290 	int bit_offset;
1291 	char new[F2FS_PRINT_NAMELEN];
1292 
1293 	if (!c.show_dentry)
1294 		return;
1295 
1296 	name_len = le16_to_cpu(dentry[idx].name_len);
1297 	next_idx = idx + (name_len + F2FS_SLOT_LEN - 1) / F2FS_SLOT_LEN;
1298 
1299 	bit_offset = find_next_bit_le(bitmap, max, next_idx);
1300 	if (bit_offset >= max && last_blk)
1301 		last_de = 1;
1302 
1303 	if (tree_mark_size <= depth) {
1304 		tree_mark_size *= 2;
1305 		ASSERT(tree_mark_size != 0);
1306 		tree_mark = realloc(tree_mark, tree_mark_size);
1307 		ASSERT(tree_mark != NULL);
1308 	}
1309 
1310 	if (last_de)
1311 		tree_mark[depth] = '`';
1312 	else
1313 		tree_mark[depth] = '|';
1314 
1315 	if (tree_mark[depth - 1] == '`')
1316 		tree_mark[depth - 1] = ' ';
1317 
1318 	for (i = 1; i < depth; i++)
1319 		printf("%c   ", tree_mark[i]);
1320 
1321 	pretty_print_filename(name, name_len, new, enc_name);
1322 
1323 	printf("%c-- %s <ino = 0x%x>, <encrypted (%d)>\n",
1324 			last_de ? '`' : '|',
1325 			new, le32_to_cpu(dentry[idx].ino),
1326 			enc_name);
1327 }
1328 
f2fs_check_hash_code(int encoding,int casefolded,struct f2fs_dir_entry * dentry,const unsigned char * name,u32 len,int enc_name)1329 static int f2fs_check_hash_code(int encoding, int casefolded,
1330 			struct f2fs_dir_entry *dentry,
1331 			const unsigned char *name, u32 len, int enc_name)
1332 {
1333 	/* Casefolded Encrypted names require a key to compute siphash */
1334 	if (enc_name && casefolded)
1335 		return 0;
1336 
1337 	f2fs_hash_t hash_code = f2fs_dentry_hash(encoding, casefolded, name, len);
1338 	/* fix hash_code made by old buggy code */
1339 	if (dentry->hash_code != hash_code) {
1340 		char new[F2FS_PRINT_NAMELEN];
1341 
1342 		pretty_print_filename(name, len, new, enc_name);
1343 		FIX_MSG("Mismatch hash_code for \"%s\" [%x:%x]",
1344 				new, le32_to_cpu(dentry->hash_code),
1345 				hash_code);
1346 		dentry->hash_code = cpu_to_le32(hash_code);
1347 		return 1;
1348 	}
1349 	return 0;
1350 }
1351 
1352 
__get_current_level(int dir_level,u32 pgofs)1353 static int __get_current_level(int dir_level, u32 pgofs)
1354 {
1355 	unsigned int bidx = 0;
1356 	int i;
1357 
1358 	for (i = 0; i < MAX_DIR_HASH_DEPTH; i++) {
1359 		bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
1360 		if (bidx > pgofs)
1361 			break;
1362 	}
1363 	return i;
1364 }
1365 
f2fs_check_dirent_position(const struct f2fs_dir_entry * dentry,const char * printable_name,u32 pgofs,u8 dir_level,u32 pino)1366 static int f2fs_check_dirent_position(const struct f2fs_dir_entry *dentry,
1367 				      const char *printable_name,
1368 				      u32 pgofs, u8 dir_level, u32 pino)
1369 {
1370 	unsigned int nbucket, nblock;
1371 	unsigned int bidx, end_block;
1372 	int level;
1373 
1374 	level = __get_current_level(dir_level, pgofs);
1375 
1376 	nbucket = dir_buckets(level, dir_level);
1377 	nblock = bucket_blocks(level);
1378 
1379 	bidx = dir_block_index(level, dir_level,
1380 			       le32_to_cpu(dentry->hash_code) % nbucket);
1381 	end_block = bidx + nblock;
1382 
1383 	if (pgofs >= bidx && pgofs < end_block)
1384 		return 0;
1385 
1386 	ASSERT_MSG("Wrong position of dirent pino:%u, name:%s, level:%d, "
1387 		"dir_level:%d, pgofs:%u, correct range:[%u, %u]\n",
1388 		pino, printable_name, level, dir_level, pgofs, bidx,
1389 		end_block - 1);
1390 	return 1;
1391 }
1392 
__chk_dots_dentries(struct f2fs_sb_info * sbi,int casefolded,struct f2fs_dir_entry * dentry,struct child_info * child,u8 * name,int len,__u8 (* filename)[F2FS_SLOT_LEN],int enc_name)1393 static int __chk_dots_dentries(struct f2fs_sb_info *sbi,
1394 			       int casefolded,
1395 			       struct f2fs_dir_entry *dentry,
1396 			       struct child_info *child,
1397 			       u8 *name, int len,
1398 			       __u8 (*filename)[F2FS_SLOT_LEN],
1399 			       int enc_name)
1400 {
1401 	int fixed = 0;
1402 
1403 	if ((name[0] == '.' && len == 1)) {
1404 		if (le32_to_cpu(dentry->ino) != child->p_ino) {
1405 			ASSERT_MSG("Bad inode number[0x%x] for '.', parent_ino is [0x%x]\n",
1406 				le32_to_cpu(dentry->ino), child->p_ino);
1407 			dentry->ino = cpu_to_le32(child->p_ino);
1408 			fixed = 1;
1409 		}
1410 	}
1411 
1412 	if (name[0] == '.' && name[1] == '.' && len == 2) {
1413 		if (child->p_ino == F2FS_ROOT_INO(sbi)) {
1414 			if (le32_to_cpu(dentry->ino) != F2FS_ROOT_INO(sbi)) {
1415 				ASSERT_MSG("Bad inode number[0x%x] for '..'\n",
1416 					le32_to_cpu(dentry->ino));
1417 				dentry->ino = cpu_to_le32(F2FS_ROOT_INO(sbi));
1418 				fixed = 1;
1419 			}
1420 		} else if (le32_to_cpu(dentry->ino) != child->pp_ino) {
1421 			ASSERT_MSG("Bad inode number[0x%x] for '..', parent parent ino is [0x%x]\n",
1422 				le32_to_cpu(dentry->ino), child->pp_ino);
1423 			dentry->ino = cpu_to_le32(child->pp_ino);
1424 			fixed = 1;
1425 		}
1426 	}
1427 
1428 	if (f2fs_check_hash_code(get_encoding(sbi), casefolded, dentry, name, len, enc_name))
1429 		fixed = 1;
1430 
1431 	if (name[len] != '\0') {
1432 		ASSERT_MSG("'.' is not NULL terminated\n");
1433 		name[len] = '\0';
1434 		memcpy(*filename, name, len);
1435 		fixed = 1;
1436 	}
1437 	return fixed;
1438 }
1439 
nullify_dentry(struct f2fs_dir_entry * dentry,int offs,__u8 (* filename)[F2FS_SLOT_LEN],u8 ** bitmap)1440 static void nullify_dentry(struct f2fs_dir_entry *dentry, int offs,
1441 			   __u8 (*filename)[F2FS_SLOT_LEN], u8 **bitmap)
1442 {
1443 	memset(dentry, 0, sizeof(struct f2fs_dir_entry));
1444 	test_and_clear_bit_le(offs, *bitmap);
1445 	memset(*filename, 0, F2FS_SLOT_LEN);
1446 }
1447 
__chk_dentries(struct f2fs_sb_info * sbi,int casefolded,struct child_info * child,u8 * bitmap,struct f2fs_dir_entry * dentry,__u8 (* filenames)[F2FS_SLOT_LEN],int max,int last_blk,int enc_name)1448 static int __chk_dentries(struct f2fs_sb_info *sbi, int casefolded,
1449 			struct child_info *child,
1450 			u8 *bitmap, struct f2fs_dir_entry *dentry,
1451 			__u8 (*filenames)[F2FS_SLOT_LEN],
1452 			int max, int last_blk, int enc_name)
1453 {
1454 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
1455 	enum FILE_TYPE ftype;
1456 	int dentries = 0;
1457 	u32 blk_cnt;
1458 	u8 *name;
1459 	char en[F2FS_PRINT_NAMELEN];
1460 	u16 name_len;
1461 	int ret = 0;
1462 	int fixed = 0;
1463 	int i, slots;
1464 
1465 	/* readahead inode blocks */
1466 	for (i = 0; i < max; i++) {
1467 		u32 ino;
1468 
1469 		if (test_bit_le(i, bitmap) == 0)
1470 			continue;
1471 
1472 		ino = le32_to_cpu(dentry[i].ino);
1473 
1474 		if (IS_VALID_NID(sbi, ino)) {
1475 			struct node_info ni;
1476 
1477 			get_node_info(sbi, ino, &ni);
1478 			if (IS_VALID_BLK_ADDR(sbi, ni.blk_addr)) {
1479 				dev_reada_block(ni.blk_addr);
1480 				name_len = le16_to_cpu(dentry[i].name_len);
1481 				if (name_len > 0)
1482 					i += (name_len + F2FS_SLOT_LEN - 1) / F2FS_SLOT_LEN - 1;
1483 			}
1484 		}
1485 	}
1486 
1487 	for (i = 0; i < max;) {
1488 		if (test_bit_le(i, bitmap) == 0) {
1489 			i++;
1490 			continue;
1491 		}
1492 		if (!IS_VALID_NID(sbi, le32_to_cpu(dentry[i].ino))) {
1493 			ASSERT_MSG("Bad dentry 0x%x with invalid NID/ino 0x%x",
1494 				    i, le32_to_cpu(dentry[i].ino));
1495 			if (c.fix_on) {
1496 				FIX_MSG("Clear bad dentry 0x%x with bad ino 0x%x",
1497 					i, le32_to_cpu(dentry[i].ino));
1498 				test_and_clear_bit_le(i, bitmap);
1499 				fixed = 1;
1500 			}
1501 			i++;
1502 			continue;
1503 		}
1504 
1505 		ftype = dentry[i].file_type;
1506 		if ((ftype <= F2FS_FT_UNKNOWN || ftype > F2FS_FT_LAST_FILE_TYPE)) {
1507 			ASSERT_MSG("Bad dentry 0x%x with unexpected ftype 0x%x",
1508 						le32_to_cpu(dentry[i].ino), ftype);
1509 			if (c.fix_on) {
1510 				FIX_MSG("Clear bad dentry 0x%x with bad ftype 0x%x",
1511 					i, ftype);
1512 				test_and_clear_bit_le(i, bitmap);
1513 				fixed = 1;
1514 			}
1515 			i++;
1516 			continue;
1517 		}
1518 
1519 		name_len = le16_to_cpu(dentry[i].name_len);
1520 
1521 		if (name_len == 0 || name_len > F2FS_NAME_LEN) {
1522 			ASSERT_MSG("Bad dentry 0x%x with invalid name_len", i);
1523 			if (c.fix_on) {
1524 				FIX_MSG("Clear bad dentry 0x%x", i);
1525 				test_and_clear_bit_le(i, bitmap);
1526 				fixed = 1;
1527 			}
1528 			i++;
1529 			continue;
1530 		}
1531 		name = calloc(name_len + 1, 1);
1532 		ASSERT(name);
1533 
1534 		memcpy(name, filenames[i], name_len);
1535 		slots = (name_len + F2FS_SLOT_LEN - 1) / F2FS_SLOT_LEN;
1536 
1537 		/* Becareful. 'dentry.file_type' is not imode. */
1538 		if (ftype == F2FS_FT_DIR) {
1539 			if ((name[0] == '.' && name_len == 1) ||
1540 				(name[0] == '.' && name[1] == '.' &&
1541 							name_len == 2)) {
1542 				ret = __chk_dots_dentries(sbi, casefolded, &dentry[i],
1543 					child, name, name_len, &filenames[i],
1544 					enc_name);
1545 				switch (ret) {
1546 				case 1:
1547 					fixed = 1;
1548 				case 0:
1549 					child->dots++;
1550 					break;
1551 				}
1552 
1553 				if (child->dots > 2) {
1554 					ASSERT_MSG("More than one '.' or '..', should delete the extra one\n");
1555 					nullify_dentry(&dentry[i], i,
1556 						       &filenames[i], &bitmap);
1557 					child->dots--;
1558 					fixed = 1;
1559 				}
1560 
1561 				i++;
1562 				free(name);
1563 				continue;
1564 			}
1565 		}
1566 
1567 		if (f2fs_check_hash_code(get_encoding(sbi), casefolded, dentry + i, name, name_len, enc_name))
1568 			fixed = 1;
1569 
1570 		pretty_print_filename(name, name_len, en, enc_name);
1571 
1572 		if (max == NR_DENTRY_IN_BLOCK) {
1573 			ret = f2fs_check_dirent_position(dentry + i, en,
1574 					child->pgofs, child->dir_level,
1575 					child->p_ino);
1576 			if (ret) {
1577 				if (c.fix_on) {
1578 					FIX_MSG("Clear bad dentry 0x%x", i);
1579 					test_and_clear_bit_le(i, bitmap);
1580 					fixed = 1;
1581 				}
1582 				i++;
1583 				free(name);
1584 				continue;
1585 			}
1586 		}
1587 
1588 		DBG(1, "[%3u]-[0x%x] name[%s] len[0x%x] ino[0x%x] type[0x%x]\n",
1589 				fsck->dentry_depth, i, en, name_len,
1590 				le32_to_cpu(dentry[i].ino),
1591 				dentry[i].file_type);
1592 
1593 		print_dentry(fsck->dentry_depth, name, bitmap,
1594 				dentry, max, i, last_blk, enc_name);
1595 
1596 		blk_cnt = 1;
1597 		child->i_namelen = name_len;
1598 		ret = fsck_chk_node_blk(sbi,
1599 				NULL, le32_to_cpu(dentry[i].ino),
1600 				ftype, TYPE_INODE, &blk_cnt, child);
1601 
1602 		if (ret && c.fix_on) {
1603 			int j;
1604 
1605 			for (j = 0; j < slots; j++)
1606 				test_and_clear_bit_le(i + j, bitmap);
1607 			FIX_MSG("Unlink [0x%x] - %s len[0x%x], type[0x%x]",
1608 					le32_to_cpu(dentry[i].ino),
1609 					en, name_len,
1610 					dentry[i].file_type);
1611 			fixed = 1;
1612 		} else if (ret == 0) {
1613 			if (ftype == F2FS_FT_DIR)
1614 				child->links++;
1615 			dentries++;
1616 			child->files++;
1617 		}
1618 
1619 		i += slots;
1620 		free(name);
1621 	}
1622 	return fixed ? -1 : dentries;
1623 }
1624 
fsck_chk_inline_dentries(struct f2fs_sb_info * sbi,struct f2fs_node * node_blk,struct child_info * child)1625 int fsck_chk_inline_dentries(struct f2fs_sb_info *sbi,
1626 		struct f2fs_node *node_blk, struct child_info *child)
1627 {
1628 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
1629 	struct f2fs_dentry_ptr d;
1630 	void *inline_dentry;
1631 	int dentries;
1632 
1633 	inline_dentry = inline_data_addr(node_blk);
1634 	ASSERT(inline_dentry != NULL);
1635 
1636 	make_dentry_ptr(&d, node_blk, inline_dentry, 2);
1637 
1638 	fsck->dentry_depth++;
1639 	dentries = __chk_dentries(sbi, IS_CASEFOLDED(&node_blk->i), child,
1640 			d.bitmap, d.dentry, d.filename, d.max, 1,
1641 			file_is_encrypt(&node_blk->i));// pass through
1642 	if (dentries < 0) {
1643 		DBG(1, "[%3d] Inline Dentry Block Fixed hash_codes\n\n",
1644 			fsck->dentry_depth);
1645 	} else {
1646 		DBG(1, "[%3d] Inline Dentry Block Done : "
1647 				"dentries:%d in %d slots (len:%d)\n\n",
1648 			fsck->dentry_depth, dentries,
1649 			d.max, F2FS_NAME_LEN);
1650 	}
1651 	fsck->dentry_depth--;
1652 	return dentries;
1653 }
1654 
fsck_chk_dentry_blk(struct f2fs_sb_info * sbi,int casefolded,u32 blk_addr,struct child_info * child,int last_blk,int enc_name)1655 int fsck_chk_dentry_blk(struct f2fs_sb_info *sbi, int casefolded, u32 blk_addr,
1656 		struct child_info *child, int last_blk, int enc_name)
1657 {
1658 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
1659 	struct f2fs_dentry_block *de_blk;
1660 	int dentries, ret;
1661 
1662 	de_blk = (struct f2fs_dentry_block *)calloc(BLOCK_SZ, 1);
1663 	ASSERT(de_blk != NULL);
1664 
1665 	ret = dev_read_block(de_blk, blk_addr);
1666 	ASSERT(ret >= 0);
1667 
1668 	fsck->dentry_depth++;
1669 	dentries = __chk_dentries(sbi, casefolded, child,
1670 			de_blk->dentry_bitmap,
1671 			de_blk->dentry, de_blk->filename,
1672 			NR_DENTRY_IN_BLOCK, last_blk, enc_name);
1673 
1674 	if (dentries < 0 && f2fs_dev_is_writable()) {
1675 		ret = dev_write_block(de_blk, blk_addr);
1676 		ASSERT(ret >= 0);
1677 		DBG(1, "[%3d] Dentry Block [0x%x] Fixed hash_codes\n\n",
1678 			fsck->dentry_depth, blk_addr);
1679 	} else {
1680 		DBG(1, "[%3d] Dentry Block [0x%x] Done : "
1681 				"dentries:%d in %d slots (len:%d)\n\n",
1682 			fsck->dentry_depth, blk_addr, dentries,
1683 			NR_DENTRY_IN_BLOCK, F2FS_NAME_LEN);
1684 	}
1685 	fsck->dentry_depth--;
1686 	free(de_blk);
1687 	return 0;
1688 }
1689 
fsck_chk_data_blk(struct f2fs_sb_info * sbi,int casefolded,u32 blk_addr,struct child_info * child,int last_blk,enum FILE_TYPE ftype,u32 parent_nid,u16 idx_in_node,u8 ver,int enc_name)1690 int fsck_chk_data_blk(struct f2fs_sb_info *sbi, int casefolded,
1691 		u32 blk_addr, struct child_info *child, int last_blk,
1692 		enum FILE_TYPE ftype, u32 parent_nid, u16 idx_in_node, u8 ver,
1693 		int enc_name)
1694 {
1695 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
1696 
1697 	/* Is it reserved block? */
1698 	if (blk_addr == NEW_ADDR) {
1699 		fsck->chk.valid_blk_cnt++;
1700 		return 0;
1701 	}
1702 
1703 	if (!IS_VALID_BLK_ADDR(sbi, blk_addr)) {
1704 		ASSERT_MSG("blkaddress is not valid. [0x%x]", blk_addr);
1705 		return -EINVAL;
1706 	}
1707 
1708 	if (is_valid_ssa_data_blk(sbi, blk_addr, parent_nid,
1709 						idx_in_node, ver)) {
1710 		ASSERT_MSG("summary data block is not valid. [0x%x]",
1711 						parent_nid);
1712 		return -EINVAL;
1713 	}
1714 
1715 	if (f2fs_test_sit_bitmap(sbi, blk_addr) == 0)
1716 		ASSERT_MSG("SIT bitmap is 0x0. blk_addr[0x%x]", blk_addr);
1717 
1718 	if (f2fs_test_main_bitmap(sbi, blk_addr) != 0)
1719 		ASSERT_MSG("Duplicated data [0x%x]. pnid[0x%x] idx[0x%x]",
1720 				blk_addr, parent_nid, idx_in_node);
1721 
1722 	fsck->chk.valid_blk_cnt++;
1723 
1724 	if (ftype == F2FS_FT_DIR) {
1725 		f2fs_set_main_bitmap(sbi, blk_addr, CURSEG_HOT_DATA);
1726 		return fsck_chk_dentry_blk(sbi, casefolded, blk_addr, child,
1727 						last_blk, enc_name);
1728 	} else {
1729 		f2fs_set_main_bitmap(sbi, blk_addr, CURSEG_WARM_DATA);
1730 	}
1731 	return 0;
1732 }
1733 
fsck_chk_orphan_node(struct f2fs_sb_info * sbi)1734 int fsck_chk_orphan_node(struct f2fs_sb_info *sbi)
1735 {
1736 	u32 blk_cnt = 0;
1737 	block_t start_blk, orphan_blkaddr, i, j;
1738 	struct f2fs_orphan_block *orphan_blk, *new_blk;
1739 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1740 	u32 entry_count;
1741 
1742 	if (!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_ORPHAN_PRESENT_FLAG))
1743 		return 0;
1744 
1745 	start_blk = __start_cp_addr(sbi) + 1 + get_sb(cp_payload);
1746 	orphan_blkaddr = __start_sum_addr(sbi) - 1 - get_sb(cp_payload);
1747 
1748 	f2fs_ra_meta_pages(sbi, start_blk, orphan_blkaddr, META_CP);
1749 
1750 	orphan_blk = calloc(BLOCK_SZ, 1);
1751 	ASSERT(orphan_blk);
1752 
1753 	new_blk = calloc(BLOCK_SZ, 1);
1754 	ASSERT(new_blk);
1755 
1756 	for (i = 0; i < orphan_blkaddr; i++) {
1757 		int ret = dev_read_block(orphan_blk, start_blk + i);
1758 		u32 new_entry_count = 0;
1759 
1760 		ASSERT(ret >= 0);
1761 		entry_count = le32_to_cpu(orphan_blk->entry_count);
1762 
1763 		for (j = 0; j < entry_count; j++) {
1764 			nid_t ino = le32_to_cpu(orphan_blk->ino[j]);
1765 			DBG(1, "[%3d] ino [0x%x]\n", i, ino);
1766 			struct node_info ni;
1767 			blk_cnt = 1;
1768 
1769 			if (c.preen_mode == PREEN_MODE_1 && !c.fix_on) {
1770 				get_node_info(sbi, ino, &ni);
1771 				if (!IS_VALID_NID(sbi, ino) ||
1772 						!IS_VALID_BLK_ADDR(sbi, ni.blk_addr))
1773 					return -EINVAL;
1774 
1775 				continue;
1776 			}
1777 
1778 			ret = fsck_chk_node_blk(sbi, NULL, ino,
1779 					F2FS_FT_ORPHAN, TYPE_INODE, &blk_cnt,
1780 					NULL);
1781 			if (!ret)
1782 				new_blk->ino[new_entry_count++] =
1783 							orphan_blk->ino[j];
1784 			else if (ret && c.fix_on)
1785 				FIX_MSG("[0x%x] remove from orphan list", ino);
1786 			else if (ret)
1787 				ASSERT_MSG("[0x%x] wrong orphan inode", ino);
1788 		}
1789 		if (f2fs_dev_is_writable() && c.fix_on &&
1790 				entry_count != new_entry_count) {
1791 			new_blk->entry_count = cpu_to_le32(new_entry_count);
1792 			ret = dev_write_block(new_blk, start_blk + i);
1793 			ASSERT(ret >= 0);
1794 		}
1795 		memset(orphan_blk, 0, BLOCK_SZ);
1796 		memset(new_blk, 0, BLOCK_SZ);
1797 	}
1798 	free(orphan_blk);
1799 	free(new_blk);
1800 
1801 	return 0;
1802 }
1803 
fsck_chk_quota_node(struct f2fs_sb_info * sbi)1804 int fsck_chk_quota_node(struct f2fs_sb_info *sbi)
1805 {
1806 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1807 	enum quota_type qtype;
1808 	int ret = 0;
1809 	u32 blk_cnt = 0;
1810 
1811 	for (qtype = 0; qtype < F2FS_MAX_QUOTAS; qtype++) {
1812 		cur_qtype = qtype;
1813 		if (sb->qf_ino[qtype] == 0)
1814 			continue;
1815 		nid_t ino = QUOTA_INO(sb, qtype);
1816 		struct node_info ni;
1817 
1818 		DBG(1, "qtype [%d] ino [0x%x]\n", qtype, ino);
1819 		blk_cnt = 1;
1820 
1821 		if (c.preen_mode == PREEN_MODE_1 && !c.fix_on) {
1822 			get_node_info(sbi, ino, &ni);
1823 			if (!IS_VALID_NID(sbi, ino) ||
1824 					!IS_VALID_BLK_ADDR(sbi, ni.blk_addr))
1825 				return -EINVAL;
1826 			continue;
1827 		}
1828 		ret = fsck_chk_node_blk(sbi, NULL, ino,
1829 				F2FS_FT_REG_FILE, TYPE_INODE, &blk_cnt, NULL);
1830 		if (ret) {
1831 			ASSERT_MSG("wrong quota inode, qtype [%d] ino [0x%x]",
1832 								qtype, ino);
1833 			qf_szchk_type[qtype] = QF_SZCHK_ERR;
1834 		}
1835 	}
1836 	cur_qtype = -1;
1837 	return ret;
1838 }
1839 
fsck_chk_quota_files(struct f2fs_sb_info * sbi)1840 int fsck_chk_quota_files(struct f2fs_sb_info *sbi)
1841 {
1842 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
1843 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1844 	enum quota_type qtype;
1845 	f2fs_ino_t ino;
1846 	int ret = 0;
1847 	int needs_writeout;
1848 
1849 	/* Return if quota feature is disabled */
1850 	if (!fsck->qctx)
1851 		return 0;
1852 
1853 	for (qtype = 0; qtype < F2FS_MAX_QUOTAS; qtype++) {
1854 		ino = sb->qf_ino[qtype];
1855 		if (!ino)
1856 			continue;
1857 
1858 	        DBG(1, "Checking Quota file ([%3d] ino [0x%x])\n", qtype, ino);
1859 		needs_writeout = 0;
1860 		ret = quota_compare_and_update(sbi, qtype, &needs_writeout,
1861 						c.preserve_limits);
1862 		if (ret == 0 && needs_writeout == 0) {
1863 			DBG(1, "OK\n");
1864 			continue;
1865 		}
1866 
1867 		/* Something is wrong */
1868 		if (c.fix_on) {
1869 			DBG(0, "Fixing Quota file ([%3d] ino [0x%x])\n",
1870 							qtype, ino);
1871 			f2fs_filesize_update(sbi, ino, 0);
1872 			ret = quota_write_inode(sbi, qtype);
1873 			if (!ret) {
1874 				c.bug_on = 1;
1875 				DBG(1, "OK\n");
1876 			} else {
1877 				ASSERT_MSG("Unable to write quota file");
1878 			}
1879 		} else {
1880 			ASSERT_MSG("Quota file is missing or invalid"
1881 					" quota file content found.");
1882 		}
1883 	}
1884 	return ret;
1885 }
1886 
fsck_chk_meta(struct f2fs_sb_info * sbi)1887 int fsck_chk_meta(struct f2fs_sb_info *sbi)
1888 {
1889 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
1890 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1891 	struct seg_entry *se;
1892 	unsigned int sit_valid_segs = 0, sit_node_blks = 0;
1893 	unsigned int i;
1894 
1895 	/* 1. check sit usage with CP: curseg is lost? */
1896 	for (i = 0; i < TOTAL_SEGS(sbi); i++) {
1897 		se = get_seg_entry(sbi, i);
1898 		if (se->valid_blocks != 0)
1899 			sit_valid_segs++;
1900 		else if (IS_CUR_SEGNO(sbi, i)) {
1901 			/* curseg has not been written back to device */
1902 			MSG(1, "\tInfo: curseg %u is counted in valid segs\n", i);
1903 			sit_valid_segs++;
1904 		}
1905 		if (IS_NODESEG(se->type))
1906 			sit_node_blks += se->valid_blocks;
1907 	}
1908 	if (fsck->chk.sit_free_segs + sit_valid_segs !=
1909 				get_usable_seg_count(sbi)) {
1910 		ASSERT_MSG("SIT usage does not match: sit_free_segs %u, "
1911 				"sit_valid_segs %u, total_segs %u",
1912 			fsck->chk.sit_free_segs, sit_valid_segs,
1913 			get_usable_seg_count(sbi));
1914 		return -EINVAL;
1915 	}
1916 
1917 	/* 2. check node count */
1918 	if (fsck->chk.valid_nat_entry_cnt != sit_node_blks) {
1919 		ASSERT_MSG("node count does not match: valid_nat_entry_cnt %u,"
1920 			" sit_node_blks %u",
1921 			fsck->chk.valid_nat_entry_cnt, sit_node_blks);
1922 		return -EINVAL;
1923 	}
1924 
1925 	/* 3. check SIT with CP */
1926 	if (fsck->chk.sit_free_segs != le32_to_cpu(cp->free_segment_count)) {
1927 		ASSERT_MSG("free segs does not match: sit_free_segs %u, "
1928 				"free_segment_count %u",
1929 				fsck->chk.sit_free_segs,
1930 				le32_to_cpu(cp->free_segment_count));
1931 		return -EINVAL;
1932 	}
1933 
1934 	/* 4. check NAT with CP */
1935 	if (fsck->chk.valid_nat_entry_cnt !=
1936 					le32_to_cpu(cp->valid_node_count)) {
1937 		ASSERT_MSG("valid node does not match: valid_nat_entry_cnt %u,"
1938 				" valid_node_count %u",
1939 				fsck->chk.valid_nat_entry_cnt,
1940 				le32_to_cpu(cp->valid_node_count));
1941 		return -EINVAL;
1942 	}
1943 
1944 	/* 4. check orphan inode simply */
1945 	if (fsck_chk_orphan_node(sbi))
1946 		return -EINVAL;
1947 
1948 	/* 5. check nat entry -- must be done before quota check */
1949 	for (i = 0; i < fsck->nr_nat_entries; i++) {
1950 		u32 blk = le32_to_cpu(fsck->entries[i].block_addr);
1951 		nid_t ino = le32_to_cpu(fsck->entries[i].ino);
1952 
1953 		if (!blk)
1954 			/*
1955 			 * skip entry whose ino is 0, otherwise, we will
1956 			 * get a negative number by BLKOFF_FROM_MAIN(sbi, blk)
1957 			 */
1958 			continue;
1959 
1960 		if (!IS_VALID_BLK_ADDR(sbi, blk)) {
1961 			MSG(0, "\tError: nat entry[ino %u block_addr 0x%x]"
1962 				" is in valid\n",
1963 				ino, blk);
1964 			return -EINVAL;
1965 		}
1966 
1967 		if (!f2fs_test_sit_bitmap(sbi, blk)) {
1968 			MSG(0, "\tError: nat entry[ino %u block_addr 0x%x]"
1969 				" not find it in sit_area_bitmap\n",
1970 				ino, blk);
1971 			return -EINVAL;
1972 		}
1973 
1974 		if (!IS_VALID_NID(sbi, ino)) {
1975 			MSG(0, "\tError: nat_entry->ino %u exceeds the range"
1976 				" of nat entries %u\n",
1977 				ino, fsck->nr_nat_entries);
1978 			return -EINVAL;
1979 		}
1980 
1981 		if (!f2fs_test_bit(ino, fsck->nat_area_bitmap)) {
1982 			MSG(0, "\tError: nat_entry->ino %u is not set in"
1983 				" nat_area_bitmap\n", ino);
1984 			return -EINVAL;
1985 		}
1986 	}
1987 
1988 	/* 6. check quota inode simply */
1989 	if (fsck_chk_quota_node(sbi))
1990 		return -EINVAL;
1991 
1992 	if (fsck->nat_valid_inode_cnt != le32_to_cpu(cp->valid_inode_count)) {
1993 		ASSERT_MSG("valid inode does not match: nat_valid_inode_cnt %u,"
1994 				" valid_inode_count %u",
1995 				fsck->nat_valid_inode_cnt,
1996 				le32_to_cpu(cp->valid_inode_count));
1997 		return -EINVAL;
1998 	}
1999 
2000 	return 0;
2001 }
2002 
fsck_chk_checkpoint(struct f2fs_sb_info * sbi)2003 void fsck_chk_checkpoint(struct f2fs_sb_info *sbi)
2004 {
2005 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2006 
2007 	if (get_cp(ckpt_flags) & CP_LARGE_NAT_BITMAP_FLAG) {
2008 		if (get_cp(checksum_offset) != CP_MIN_CHKSUM_OFFSET) {
2009 			ASSERT_MSG("Deprecated layout of large_nat_bitmap, "
2010 				"chksum_offset:%u", get_cp(checksum_offset));
2011 			c.fix_chksum = 1;
2012 		}
2013 	}
2014 }
2015 
fsck_init(struct f2fs_sb_info * sbi)2016 void fsck_init(struct f2fs_sb_info *sbi)
2017 {
2018 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2019 	struct f2fs_sm_info *sm_i = SM_I(sbi);
2020 
2021 	/*
2022 	 * We build three bitmap for main/sit/nat so that may check consistency
2023 	 * of filesystem.
2024 	 * 1. main_area_bitmap will be used to check whether all blocks of main
2025 	 *    area is used or not.
2026 	 * 2. nat_area_bitmap has bitmap information of used nid in NAT.
2027 	 * 3. sit_area_bitmap has bitmap information of used main block.
2028 	 * At Last sequence, we compare main_area_bitmap with sit_area_bitmap.
2029 	 */
2030 	fsck->nr_main_blks = sm_i->main_segments << sbi->log_blocks_per_seg;
2031 	fsck->main_area_bitmap_sz = (fsck->nr_main_blks + 7) / 8;
2032 	fsck->main_area_bitmap = calloc(fsck->main_area_bitmap_sz, 1);
2033 	ASSERT(fsck->main_area_bitmap != NULL);
2034 
2035 	build_nat_area_bitmap(sbi);
2036 
2037 	build_sit_area_bitmap(sbi);
2038 
2039 	ASSERT(tree_mark_size != 0);
2040 	tree_mark = calloc(tree_mark_size, 1);
2041 	ASSERT(tree_mark != NULL);
2042 }
2043 
fix_hard_links(struct f2fs_sb_info * sbi)2044 static void fix_hard_links(struct f2fs_sb_info *sbi)
2045 {
2046 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2047 	struct hard_link_node *tmp, *node;
2048 	struct f2fs_node *node_blk = NULL;
2049 	struct node_info ni;
2050 	int ret;
2051 
2052 	if (fsck->hard_link_list_head == NULL)
2053 		return;
2054 
2055 	node_blk = (struct f2fs_node *)calloc(BLOCK_SZ, 1);
2056 	ASSERT(node_blk != NULL);
2057 
2058 	node = fsck->hard_link_list_head;
2059 	while (node) {
2060 		/* Sanity check */
2061 		if (sanity_check_nid(sbi, node->nid, node_blk,
2062 					F2FS_FT_MAX, TYPE_INODE, &ni))
2063 			FIX_MSG("Failed to fix, rerun fsck.f2fs");
2064 
2065 		node_blk->i.i_links = cpu_to_le32(node->actual_links);
2066 
2067 		FIX_MSG("File: 0x%x i_links= 0x%x -> 0x%x",
2068 				node->nid, node->links, node->actual_links);
2069 
2070 		ret = dev_write_block(node_blk, ni.blk_addr);
2071 		ASSERT(ret >= 0);
2072 		tmp = node;
2073 		node = node->next;
2074 		free(tmp);
2075 	}
2076 	free(node_blk);
2077 }
2078 
fix_nat_entries(struct f2fs_sb_info * sbi)2079 static void fix_nat_entries(struct f2fs_sb_info *sbi)
2080 {
2081 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2082 	u32 i;
2083 
2084 	for (i = 0; i < fsck->nr_nat_entries; i++)
2085 		if (f2fs_test_bit(i, fsck->nat_area_bitmap) != 0)
2086 			nullify_nat_entry(sbi, i);
2087 }
2088 
flush_curseg_sit_entries(struct f2fs_sb_info * sbi)2089 static void flush_curseg_sit_entries(struct f2fs_sb_info *sbi)
2090 {
2091 	struct sit_info *sit_i = SIT_I(sbi);
2092 	struct f2fs_sit_block *sit_blk;
2093 	int i;
2094 
2095 	sit_blk = calloc(BLOCK_SZ, 1);
2096 	ASSERT(sit_blk);
2097 	/* update curseg sit entries, since we may change
2098 	 * a segment type in move_curseg_info
2099 	 */
2100 	for (i = 0; i < NO_CHECK_TYPE; i++) {
2101 		struct curseg_info *curseg = CURSEG_I(sbi, i);
2102 		struct f2fs_sit_entry *sit;
2103 		struct seg_entry *se;
2104 
2105 		se = get_seg_entry(sbi, curseg->segno);
2106 		get_current_sit_page(sbi, curseg->segno, sit_blk);
2107 		sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, curseg->segno)];
2108 		sit->vblocks = cpu_to_le16((se->type << SIT_VBLOCKS_SHIFT) |
2109 							se->valid_blocks);
2110 		rewrite_current_sit_page(sbi, curseg->segno, sit_blk);
2111 	}
2112 
2113 	free(sit_blk);
2114 }
2115 
fix_checksum(struct f2fs_sb_info * sbi)2116 static void fix_checksum(struct f2fs_sb_info *sbi)
2117 {
2118 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2119 	struct f2fs_nm_info *nm_i = NM_I(sbi);
2120 	struct sit_info *sit_i = SIT_I(sbi);
2121 	void *bitmap_offset;
2122 
2123 	if (!c.fix_chksum)
2124 		return;
2125 
2126 	bitmap_offset = cp->sit_nat_version_bitmap + sizeof(__le32);
2127 
2128 	memcpy(bitmap_offset, nm_i->nat_bitmap, nm_i->bitmap_size);
2129 	memcpy(bitmap_offset + nm_i->bitmap_size,
2130 			sit_i->sit_bitmap, sit_i->bitmap_size);
2131 }
2132 
fix_checkpoint(struct f2fs_sb_info * sbi)2133 static void fix_checkpoint(struct f2fs_sb_info *sbi)
2134 {
2135 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2136 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2137 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2138 	unsigned long long cp_blk_no;
2139 	u32 flags = c.alloc_failed ? CP_FSCK_FLAG: CP_UMOUNT_FLAG;
2140 	block_t orphan_blks = 0;
2141 	block_t cp_blocks;
2142 	u32 i;
2143 	int ret;
2144 	u_int32_t crc = 0;
2145 
2146 	/* should call from fsck */
2147 	ASSERT(c.func == FSCK);
2148 
2149 	if (is_set_ckpt_flags(cp, CP_ORPHAN_PRESENT_FLAG)) {
2150 		orphan_blks = __start_sum_addr(sbi) - 1;
2151 		flags |= CP_ORPHAN_PRESENT_FLAG;
2152 	}
2153 	if (is_set_ckpt_flags(cp, CP_TRIMMED_FLAG))
2154 		flags |= CP_TRIMMED_FLAG;
2155 	if (is_set_ckpt_flags(cp, CP_DISABLED_FLAG))
2156 		flags |= CP_DISABLED_FLAG;
2157 	if (is_set_ckpt_flags(cp, CP_LARGE_NAT_BITMAP_FLAG)) {
2158 		flags |= CP_LARGE_NAT_BITMAP_FLAG;
2159 		set_cp(checksum_offset, CP_MIN_CHKSUM_OFFSET);
2160 	} else {
2161 		set_cp(checksum_offset, CP_CHKSUM_OFFSET);
2162 	}
2163 
2164 	if (flags & CP_UMOUNT_FLAG)
2165 		cp_blocks = 8;
2166 	else
2167 		cp_blocks = 5;
2168 
2169 	set_cp(cp_pack_total_block_count, cp_blocks +
2170 				orphan_blks + get_sb(cp_payload));
2171 
2172 	flags = update_nat_bits_flags(sb, cp, flags);
2173 	flags |= CP_NOCRC_RECOVERY_FLAG;
2174 	set_cp(ckpt_flags, flags);
2175 
2176 	set_cp(free_segment_count, get_free_segments(sbi));
2177 	set_cp(valid_block_count, fsck->chk.valid_blk_cnt);
2178 	set_cp(valid_node_count, fsck->chk.valid_node_cnt);
2179 	set_cp(valid_inode_count, fsck->chk.valid_inode_cnt);
2180 
2181 	crc = f2fs_checkpoint_chksum(cp);
2182 	*((__le32 *)((unsigned char *)cp + get_cp(checksum_offset))) =
2183 							cpu_to_le32(crc);
2184 
2185 	cp_blk_no = get_sb(cp_blkaddr);
2186 	if (sbi->cur_cp == 2)
2187 		cp_blk_no += 1 << get_sb(log_blocks_per_seg);
2188 
2189 	ret = dev_write_block(cp, cp_blk_no++);
2190 	ASSERT(ret >= 0);
2191 
2192 	for (i = 0; i < get_sb(cp_payload); i++) {
2193 		ret = dev_write_block(((unsigned char *)cp) +
2194 					(i + 1) * F2FS_BLKSIZE, cp_blk_no++);
2195 		ASSERT(ret >= 0);
2196 	}
2197 
2198 	cp_blk_no += orphan_blks;
2199 
2200 	for (i = 0; i < NO_CHECK_TYPE; i++) {
2201 		struct curseg_info *curseg = CURSEG_I(sbi, i);
2202 
2203 		if (!(flags & CP_UMOUNT_FLAG) && IS_NODESEG(i))
2204 			continue;
2205 
2206 		ret = dev_write_block(curseg->sum_blk, cp_blk_no++);
2207 		ASSERT(ret >= 0);
2208 	}
2209 
2210 	/* Write nat bits */
2211 	if (flags & CP_NAT_BITS_FLAG)
2212 		write_nat_bits(sbi, sb, cp, sbi->cur_cp);
2213 
2214 	ret = f2fs_fsync_device();
2215 	ASSERT(ret >= 0);
2216 
2217 	ret = dev_write_block(cp, cp_blk_no++);
2218 	ASSERT(ret >= 0);
2219 
2220 	ret = f2fs_fsync_device();
2221 	ASSERT(ret >= 0);
2222 }
2223 
fix_checkpoints(struct f2fs_sb_info * sbi)2224 static void fix_checkpoints(struct f2fs_sb_info *sbi)
2225 {
2226 	/* copy valid checkpoint to its mirror position */
2227 	duplicate_checkpoint(sbi);
2228 
2229 	/* repair checkpoint at CP #0 position */
2230 	sbi->cur_cp = 1;
2231 	fix_checkpoint(sbi);
2232 }
2233 
2234 #ifdef HAVE_LINUX_BLKZONED_H
2235 
2236 /*
2237  * Refer valid block map and return offset of the last valid block in the zone.
2238  * Obtain valid block map from SIT and fsync data.
2239  * If there is no valid block in the zone, return -1.
2240  */
last_vblk_off_in_zone(struct f2fs_sb_info * sbi,unsigned int zone_segno)2241 static int last_vblk_off_in_zone(struct f2fs_sb_info *sbi,
2242 				 unsigned int zone_segno)
2243 {
2244 	int s, b;
2245 	unsigned int segs_per_zone = sbi->segs_per_sec * sbi->secs_per_zone;
2246 	struct seg_entry *se;
2247 
2248 	for (s = segs_per_zone - 1; s >= 0; s--) {
2249 		se = get_seg_entry(sbi, zone_segno + s);
2250 
2251 		/*
2252 		 * Refer not cur_valid_map but ckpt_valid_map which reflects
2253 		 * fsync data.
2254 		 */
2255 		ASSERT(se->ckpt_valid_map);
2256 		for (b = sbi->blocks_per_seg - 1; b >= 0; b--)
2257 			if (f2fs_test_bit(b, (const char*)se->ckpt_valid_map))
2258 				return b + (s << sbi->log_blocks_per_seg);
2259 	}
2260 
2261 	return -1;
2262 }
2263 
check_curseg_write_pointer(struct f2fs_sb_info * sbi,int type)2264 static int check_curseg_write_pointer(struct f2fs_sb_info *sbi, int type)
2265 {
2266 	struct curseg_info *curseg = CURSEG_I(sbi, type);
2267 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2268 	struct blk_zone blkz;
2269 	block_t cs_block, wp_block, zone_last_vblock;
2270 	u_int64_t cs_sector, wp_sector;
2271 	int i, ret;
2272 	unsigned int zone_segno;
2273 	int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
2274 
2275 	/* get the device the curseg points to */
2276 	cs_block = START_BLOCK(sbi, curseg->segno) + curseg->next_blkoff;
2277 	for (i = 0; i < MAX_DEVICES; i++) {
2278 		if (!c.devices[i].path)
2279 			break;
2280 		if (c.devices[i].start_blkaddr <= cs_block &&
2281 		    cs_block <= c.devices[i].end_blkaddr)
2282 			break;
2283 	}
2284 
2285 	if (i >= MAX_DEVICES)
2286 		return -EINVAL;
2287 
2288 	/* get write pointer position of the zone the curseg points to */
2289 	cs_sector = (cs_block - c.devices[i].start_blkaddr)
2290 		<< log_sectors_per_block;
2291 	ret = f2fs_report_zone(i, cs_sector, &blkz);
2292 	if (ret)
2293 		return ret;
2294 
2295 	if (blk_zone_type(&blkz) != BLK_ZONE_TYPE_SEQWRITE_REQ)
2296 		return 0;
2297 
2298 	/* check consistency between the curseg and the write pointer */
2299 	wp_block = c.devices[i].start_blkaddr +
2300 		(blk_zone_wp_sector(&blkz) >> log_sectors_per_block);
2301 	wp_sector = blk_zone_wp_sector(&blkz);
2302 
2303 	if (cs_sector == wp_sector)
2304 		return 0;
2305 
2306 	if (cs_sector > wp_sector) {
2307 		MSG(0, "Inconsistent write pointer with curseg %d: "
2308 		    "curseg %d[0x%x,0x%x] > wp[0x%x,0x%x]\n",
2309 		    type, type, curseg->segno, curseg->next_blkoff,
2310 		    GET_SEGNO(sbi, wp_block), OFFSET_IN_SEG(sbi, wp_block));
2311 		fsck->chk.wp_inconsistent_zones++;
2312 		return -EINVAL;
2313 	}
2314 
2315 	MSG(0, "Write pointer goes advance from curseg %d: "
2316 	    "curseg %d[0x%x,0x%x] wp[0x%x,0x%x]\n",
2317 	    type, type, curseg->segno, curseg->next_blkoff,
2318 	    GET_SEGNO(sbi, wp_block), OFFSET_IN_SEG(sbi, wp_block));
2319 
2320 	zone_segno = GET_SEG_FROM_SEC(sbi,
2321 				      GET_SEC_FROM_SEG(sbi, curseg->segno));
2322 	zone_last_vblock = START_BLOCK(sbi, zone_segno) +
2323 		last_vblk_off_in_zone(sbi, zone_segno);
2324 
2325 	/*
2326 	 * If valid blocks exist between the curseg position and the write
2327 	 * pointer, they are fsync data. This is not an error to fix. Leave it
2328 	 * for kernel to recover later.
2329 	 * If valid blocks exist between the curseg's zone start and the curseg
2330 	 * position, or if there is no valid block in the curseg's zone, fix
2331 	 * the inconsistency between the curseg and the writ pointer.
2332 	 * Of Note is that if there is no valid block in the curseg's zone,
2333 	 * last_vblk_off_in_zone() returns -1 and zone_last_vblock is always
2334 	 * smaller than cs_block.
2335 	 */
2336 	if (cs_block <= zone_last_vblock && zone_last_vblock < wp_block) {
2337 		MSG(0, "Curseg has fsync data: curseg %d[0x%x,0x%x] "
2338 		    "last valid block in zone[0x%x,0x%x]\n",
2339 		    type, curseg->segno, curseg->next_blkoff,
2340 		    GET_SEGNO(sbi, zone_last_vblock),
2341 		    OFFSET_IN_SEG(sbi, zone_last_vblock));
2342 		return 0;
2343 	}
2344 
2345 	fsck->chk.wp_inconsistent_zones++;
2346 	return -EINVAL;
2347 }
2348 
2349 #else
2350 
check_curseg_write_pointer(struct f2fs_sb_info * UNUSED (sbi),int UNUSED (type))2351 static int check_curseg_write_pointer(struct f2fs_sb_info *UNUSED(sbi),
2352 					int UNUSED(type))
2353 {
2354 	return 0;
2355 }
2356 
2357 #endif
2358 
check_curseg_offset(struct f2fs_sb_info * sbi,int type)2359 int check_curseg_offset(struct f2fs_sb_info *sbi, int type)
2360 {
2361 	struct curseg_info *curseg = CURSEG_I(sbi, type);
2362 	struct seg_entry *se;
2363 	int j, nblocks;
2364 
2365 	if ((curseg->next_blkoff >> 3) >= SIT_VBLOCK_MAP_SIZE) {
2366 		ASSERT_MSG("Next block offset:%u is invalid, type:%d",
2367 			curseg->next_blkoff, type);
2368 		return -EINVAL;
2369 	}
2370 	se = get_seg_entry(sbi, curseg->segno);
2371 	if (f2fs_test_bit(curseg->next_blkoff,
2372 				(const char *)se->cur_valid_map)) {
2373 		ASSERT_MSG("Next block offset is not free, type:%d", type);
2374 		return -EINVAL;
2375 	}
2376 	if (curseg->alloc_type == SSR)
2377 		return 0;
2378 
2379 	nblocks = sbi->blocks_per_seg;
2380 	for (j = curseg->next_blkoff + 1; j < nblocks; j++) {
2381 		if (f2fs_test_bit(j, (const char *)se->cur_valid_map)) {
2382 			ASSERT_MSG("For LFS curseg, space after .next_blkoff "
2383 				"should be unused, type:%d", type);
2384 			return -EINVAL;
2385 		}
2386 	}
2387 
2388 	if (c.zoned_model == F2FS_ZONED_HM)
2389 		return check_curseg_write_pointer(sbi, type);
2390 
2391 	return 0;
2392 }
2393 
check_curseg_offsets(struct f2fs_sb_info * sbi)2394 int check_curseg_offsets(struct f2fs_sb_info *sbi)
2395 {
2396 	int i, ret;
2397 
2398 	for (i = 0; i < NO_CHECK_TYPE; i++) {
2399 		ret = check_curseg_offset(sbi, i);
2400 		if (ret)
2401 			return ret;
2402 	}
2403 	return 0;
2404 }
2405 
fix_curseg_info(struct f2fs_sb_info * sbi)2406 static void fix_curseg_info(struct f2fs_sb_info *sbi)
2407 {
2408 	int i, need_update = 0;
2409 
2410 	for (i = 0; i < NO_CHECK_TYPE; i++) {
2411 		if (check_curseg_offset(sbi, i)) {
2412 			update_curseg_info(sbi, i);
2413 			need_update = 1;
2414 		}
2415 	}
2416 
2417 	if (need_update) {
2418 		write_curseg_info(sbi);
2419 		flush_curseg_sit_entries(sbi);
2420 	}
2421 }
2422 
check_sit_types(struct f2fs_sb_info * sbi)2423 int check_sit_types(struct f2fs_sb_info *sbi)
2424 {
2425 	unsigned int i;
2426 	int err = 0;
2427 
2428 	for (i = 0; i < TOTAL_SEGS(sbi); i++) {
2429 		struct seg_entry *se;
2430 
2431 		se = get_seg_entry(sbi, i);
2432 		if (se->orig_type != se->type) {
2433 			if (se->orig_type == CURSEG_COLD_DATA &&
2434 					se->type <= CURSEG_COLD_DATA) {
2435 				se->type = se->orig_type;
2436 			} else {
2437 				FIX_MSG("Wrong segment type [0x%x] %x -> %x",
2438 						i, se->orig_type, se->type);
2439 				err = -EINVAL;
2440 			}
2441 		}
2442 	}
2443 	return err;
2444 }
2445 
fsck_get_lpf(struct f2fs_sb_info * sbi)2446 static struct f2fs_node *fsck_get_lpf(struct f2fs_sb_info *sbi)
2447 {
2448 	struct f2fs_node *node;
2449 	struct node_info ni;
2450 	nid_t lpf_ino;
2451 	int err;
2452 
2453 	/* read root inode first */
2454 	node = calloc(F2FS_BLKSIZE, 1);
2455 	ASSERT(node);
2456 	get_node_info(sbi, F2FS_ROOT_INO(sbi), &ni);
2457 	err = dev_read_block(node, ni.blk_addr);
2458 	ASSERT(err >= 0);
2459 
2460 	/* lookup lost+found in root directory */
2461 	lpf_ino = f2fs_lookup(sbi, node, (u8 *)LPF, strlen(LPF));
2462 	if (lpf_ino) { /* found */
2463 		get_node_info(sbi, lpf_ino, &ni);
2464 		err = dev_read_block(node, ni.blk_addr);
2465 		ASSERT(err >= 0);
2466 		DBG(1, "Found lost+found 0x%x at blkaddr [0x%x]\n",
2467 		    lpf_ino, ni.blk_addr);
2468 		if (!S_ISDIR(le16_to_cpu(node->i.i_mode))) {
2469 			ASSERT_MSG("lost+found is not directory [0%o]\n",
2470 				   le16_to_cpu(node->i.i_mode));
2471 			/* FIXME: give up? */
2472 			goto out;
2473 		}
2474 	} else { /* not found, create it */
2475 		struct dentry de;
2476 
2477 		memset(&de, 0, sizeof(de));
2478 		de.name = (u8 *) LPF;
2479 		de.len = strlen(LPF);
2480 		de.mode = 0x41c0;
2481 		de.pino = F2FS_ROOT_INO(sbi),
2482 		de.file_type = F2FS_FT_DIR,
2483 		de.uid = getuid();
2484 		de.gid = getgid();
2485 		de.mtime = time(NULL);
2486 
2487 		err = f2fs_mkdir(sbi, &de);
2488 		if (err) {
2489 			ASSERT_MSG("Failed create lost+found");
2490 			goto out;
2491 		}
2492 
2493 		get_node_info(sbi, de.ino, &ni);
2494 		err = dev_read_block(node, ni.blk_addr);
2495 		ASSERT(err >= 0);
2496 		DBG(1, "Create lost+found 0x%x at blkaddr [0x%x]\n",
2497 		    de.ino, ni.blk_addr);
2498 	}
2499 
2500 	c.lpf_ino = le32_to_cpu(node->footer.ino);
2501 	return node;
2502 out:
2503 	free(node);
2504 	return NULL;
2505 }
2506 
fsck_do_reconnect_file(struct f2fs_sb_info * sbi,struct f2fs_node * lpf,struct f2fs_node * fnode)2507 static int fsck_do_reconnect_file(struct f2fs_sb_info *sbi,
2508 				  struct f2fs_node *lpf,
2509 				  struct f2fs_node *fnode)
2510 {
2511 	char name[80];
2512 	size_t namelen;
2513 	nid_t ino = le32_to_cpu(fnode->footer.ino);
2514 	struct node_info ni;
2515 	int ftype, ret;
2516 
2517 	namelen = snprintf(name, 80, "%u", ino);
2518 	if (namelen >= 80)
2519 		/* ignore terminating '\0', should never happen */
2520 		namelen = 79;
2521 
2522 	if (f2fs_lookup(sbi, lpf, (u8 *)name, namelen)) {
2523 		ASSERT_MSG("Name %s already exist in lost+found", name);
2524 		return -EEXIST;
2525 	}
2526 
2527 	get_node_info(sbi, le32_to_cpu(lpf->footer.ino), &ni);
2528 	ftype = map_de_type(le16_to_cpu(fnode->i.i_mode));
2529 	ret = f2fs_add_link(sbi, lpf, (unsigned char *)name, namelen,
2530 			    ino, ftype, ni.blk_addr, 0);
2531 	if (ret) {
2532 		ASSERT_MSG("Failed to add inode [0x%x] to lost+found", ino);
2533 		return -EINVAL;
2534 	}
2535 
2536 	/* update fnode */
2537 	memcpy(fnode->i.i_name, name, namelen);
2538 	fnode->i.i_namelen = cpu_to_le32(namelen);
2539 	fnode->i.i_pino = c.lpf_ino;
2540 	get_node_info(sbi, le32_to_cpu(fnode->footer.ino), &ni);
2541 	ret = dev_write_block(fnode, ni.blk_addr);
2542 	ASSERT(ret >= 0);
2543 
2544 	DBG(1, "Reconnect inode [0x%x] to lost+found\n", ino);
2545 	return 0;
2546 }
2547 
fsck_failed_reconnect_file_dnode(struct f2fs_sb_info * sbi,nid_t nid)2548 static void fsck_failed_reconnect_file_dnode(struct f2fs_sb_info *sbi,
2549 					     nid_t nid)
2550 {
2551 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2552 	struct f2fs_node *node;
2553 	struct node_info ni;
2554 	u32 addr;
2555 	int i, err;
2556 
2557 	node = calloc(F2FS_BLKSIZE, 1);
2558 	ASSERT(node);
2559 
2560 	get_node_info(sbi, nid, &ni);
2561 	err = dev_read_block(node, ni.blk_addr);
2562 	ASSERT(err >= 0);
2563 
2564 	fsck->chk.valid_node_cnt--;
2565 	fsck->chk.valid_blk_cnt--;
2566 	f2fs_clear_main_bitmap(sbi, ni.blk_addr);
2567 
2568 	for (i = 0; i < ADDRS_PER_BLOCK(&node->i); i++) {
2569 		addr = le32_to_cpu(node->dn.addr[i]);
2570 		if (!addr)
2571 			continue;
2572 		fsck->chk.valid_blk_cnt--;
2573 		if (addr == NEW_ADDR)
2574 			continue;
2575 		f2fs_clear_main_bitmap(sbi, addr);
2576 	}
2577 
2578 	free(node);
2579 }
2580 
fsck_failed_reconnect_file_idnode(struct f2fs_sb_info * sbi,nid_t nid)2581 static void fsck_failed_reconnect_file_idnode(struct f2fs_sb_info *sbi,
2582 					      nid_t nid)
2583 {
2584 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2585 	struct f2fs_node *node;
2586 	struct node_info ni;
2587 	nid_t tmp;
2588 	int i, err;
2589 
2590 	node = calloc(F2FS_BLKSIZE, 1);
2591 	ASSERT(node);
2592 
2593 	get_node_info(sbi, nid, &ni);
2594 	err = dev_read_block(node, ni.blk_addr);
2595 	ASSERT(err >= 0);
2596 
2597 	fsck->chk.valid_node_cnt--;
2598 	fsck->chk.valid_blk_cnt--;
2599 	f2fs_clear_main_bitmap(sbi, ni.blk_addr);
2600 
2601 	for (i = 0; i < NIDS_PER_BLOCK; i++) {
2602 		tmp = le32_to_cpu(node->in.nid[i]);
2603 		if (!tmp)
2604 			continue;
2605 		fsck_failed_reconnect_file_dnode(sbi, tmp);
2606 	}
2607 
2608 	free(node);
2609 }
2610 
fsck_failed_reconnect_file_didnode(struct f2fs_sb_info * sbi,nid_t nid)2611 static void fsck_failed_reconnect_file_didnode(struct f2fs_sb_info *sbi,
2612 					       nid_t nid)
2613 {
2614 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2615 	struct f2fs_node *node;
2616 	struct node_info ni;
2617 	nid_t tmp;
2618 	int i, err;
2619 
2620 	node = calloc(F2FS_BLKSIZE, 1);
2621 	ASSERT(node);
2622 
2623 	get_node_info(sbi, nid, &ni);
2624 	err = dev_read_block(node, ni.blk_addr);
2625 	ASSERT(err >= 0);
2626 
2627 	fsck->chk.valid_node_cnt--;
2628 	fsck->chk.valid_blk_cnt--;
2629 	f2fs_clear_main_bitmap(sbi, ni.blk_addr);
2630 
2631 	for (i = 0; i < NIDS_PER_BLOCK; i++) {
2632 		tmp = le32_to_cpu(node->in.nid[i]);
2633 		if (!tmp)
2634 			continue;
2635 		fsck_failed_reconnect_file_idnode(sbi, tmp);
2636 	}
2637 
2638 	free(node);
2639 }
2640 
2641 /*
2642  * Counters and main_area_bitmap are already changed during checking
2643  * inode block, so clear them. There is no need to clear new blocks
2644  * allocted to lost+found.
2645  */
fsck_failed_reconnect_file(struct f2fs_sb_info * sbi,nid_t ino)2646 static void fsck_failed_reconnect_file(struct f2fs_sb_info *sbi, nid_t ino)
2647 {
2648 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2649 	struct f2fs_node *node;
2650 	struct node_info ni;
2651 	nid_t nid;
2652 	int ofs, i, err;
2653 
2654 	node = calloc(F2FS_BLKSIZE, 1);
2655 	ASSERT(node);
2656 
2657 	get_node_info(sbi, ino, &ni);
2658 	err = dev_read_block(node, ni.blk_addr);
2659 	ASSERT(err >= 0);
2660 
2661 	/* clear inode counters */
2662 	fsck->chk.valid_inode_cnt--;
2663 	fsck->chk.valid_node_cnt--;
2664 	fsck->chk.valid_blk_cnt--;
2665 	f2fs_clear_main_bitmap(sbi, ni.blk_addr);
2666 
2667 	/* clear xnid counters */
2668 	if (node->i.i_xattr_nid) {
2669 		nid = le32_to_cpu(node->i.i_xattr_nid);
2670 		fsck->chk.valid_node_cnt--;
2671 		fsck->chk.valid_blk_cnt--;
2672 		get_node_info(sbi, nid, &ni);
2673 		f2fs_clear_main_bitmap(sbi, ni.blk_addr);
2674 	}
2675 
2676 	/* clear data counters */
2677 	if(!(node->i.i_inline & F2FS_INLINE_DATA)) {
2678 		ofs = get_extra_isize(node);
2679 		for (i = 0; i < ADDRS_PER_INODE(&node->i); i++) {
2680 			block_t addr = le32_to_cpu(node->i.i_addr[ofs + i]);
2681 			if (!addr)
2682 				continue;
2683 			fsck->chk.valid_blk_cnt--;
2684 			if (addr == NEW_ADDR)
2685 				continue;
2686 			f2fs_clear_main_bitmap(sbi, addr);
2687 		}
2688 	}
2689 
2690 	for (i = 0; i < 5; i++) {
2691 		nid = le32_to_cpu(node->i.i_nid[i]);
2692 		if (!nid)
2693 			continue;
2694 
2695 		switch (i) {
2696 		case 0: /* direct node */
2697 		case 1:
2698 			fsck_failed_reconnect_file_dnode(sbi, nid);
2699 			break;
2700 		case 2: /* indirect node */
2701 		case 3:
2702 			fsck_failed_reconnect_file_idnode(sbi, nid);
2703 			break;
2704 		case 4: /* double indirect node */
2705 			fsck_failed_reconnect_file_didnode(sbi, nid);
2706 			break;
2707 		}
2708 	}
2709 
2710 	free(node);
2711 }
2712 
2713 /*
2714  * Scan unreachable nids and find only regular file inodes. If these files
2715  * are not corrupted, reconnect them to lost+found.
2716  *
2717  * Since all unreachable nodes are already checked, we can allocate new
2718  * blocks safely.
2719  *
2720  * This function returns the number of files been reconnected.
2721  */
fsck_reconnect_file(struct f2fs_sb_info * sbi)2722 static int fsck_reconnect_file(struct f2fs_sb_info *sbi)
2723 {
2724 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2725 	struct f2fs_node *lpf_node, *node;
2726 	struct node_info ni;
2727 	char *reconnect_bitmap;
2728 	u32 blk_cnt;
2729 	nid_t nid;
2730 	int err, cnt = 0, ftype;
2731 
2732 	node = calloc(F2FS_BLKSIZE, 1);
2733 	ASSERT(node);
2734 
2735 	reconnect_bitmap = calloc(fsck->nat_area_bitmap_sz, 1);
2736 	ASSERT(reconnect_bitmap);
2737 
2738 	for (nid = 0; nid < fsck->nr_nat_entries; nid++) {
2739 		if (f2fs_test_bit(nid, fsck->nat_area_bitmap)) {
2740 			if (is_qf_ino(F2FS_RAW_SUPER(sbi), nid)) {
2741 				DBG(1, "Not support quota inode [0x%x]\n",
2742 				    nid);
2743 				continue;
2744 			}
2745 
2746 			get_node_info(sbi, nid, &ni);
2747 			err = dev_read_block(node, ni.blk_addr);
2748 			ASSERT(err >= 0);
2749 
2750 			/* reconnection will restore these nodes if needed */
2751 			if (node->footer.ino != node->footer.nid) {
2752 				DBG(1, "Not support non-inode node [0x%x]\n",
2753 				    nid);
2754 				continue;
2755 			}
2756 
2757 			if (S_ISDIR(le16_to_cpu(node->i.i_mode))) {
2758 				DBG(1, "Not support directory inode [0x%x]\n",
2759 				    nid);
2760 				continue;
2761 			}
2762 
2763 			ftype = map_de_type(le16_to_cpu(node->i.i_mode));
2764 			if (sanity_check_nid(sbi, nid, node, ftype,
2765 					     TYPE_INODE, &ni)) {
2766 				ASSERT_MSG("Invalid nid [0x%x]\n", nid);
2767 				continue;
2768 			}
2769 
2770 			DBG(1, "Check inode 0x%x\n", nid);
2771 			blk_cnt = 1;
2772 			fsck_chk_inode_blk(sbi, nid, ftype, node,
2773 					   &blk_cnt, &ni, NULL);
2774 
2775 			f2fs_set_bit(nid, reconnect_bitmap);
2776 		}
2777 	}
2778 
2779 	lpf_node = fsck_get_lpf(sbi);
2780 	if (!lpf_node)
2781 		goto out;
2782 
2783 	for (nid = 0; nid < fsck->nr_nat_entries; nid++) {
2784 		if (f2fs_test_bit(nid, reconnect_bitmap)) {
2785 			get_node_info(sbi, nid, &ni);
2786 			err = dev_read_block(node, ni.blk_addr);
2787 			ASSERT(err >= 0);
2788 
2789 			if (fsck_do_reconnect_file(sbi, lpf_node, node)) {
2790 				DBG(1, "Failed to reconnect inode [0x%x]\n",
2791 				    nid);
2792 				fsck_failed_reconnect_file(sbi, nid);
2793 				continue;
2794 			}
2795 
2796 			quota_add_inode_usage(fsck->qctx, nid, &node->i);
2797 
2798 			DBG(1, "Reconnected inode [0x%x] to lost+found\n", nid);
2799 			cnt++;
2800 		}
2801 	}
2802 
2803 out:
2804 	free(node);
2805 	free(lpf_node);
2806 	free(reconnect_bitmap);
2807 	return cnt;
2808 }
2809 
2810 #ifdef HAVE_LINUX_BLKZONED_H
2811 
2812 struct write_pointer_check_data {
2813 	struct f2fs_sb_info *sbi;
2814 	int dev_index;
2815 };
2816 
chk_and_fix_wp_with_sit(int i,void * blkzone,void * opaque)2817 static int chk_and_fix_wp_with_sit(int i, void *blkzone, void *opaque)
2818 {
2819 	struct blk_zone *blkz = (struct blk_zone *)blkzone;
2820 	struct write_pointer_check_data *wpd = opaque;
2821 	struct f2fs_sb_info *sbi = wpd->sbi;
2822 	struct device_info *dev = c.devices + wpd->dev_index;
2823 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2824 	block_t zone_block, wp_block, wp_blkoff;
2825 	unsigned int zone_segno, wp_segno;
2826 	struct curseg_info *cs;
2827 	int cs_index, ret, last_valid_blkoff;
2828 	int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
2829 	unsigned int segs_per_zone = sbi->segs_per_sec * sbi->secs_per_zone;
2830 
2831 	if (blk_zone_conv(blkz))
2832 		return 0;
2833 
2834 	zone_block = dev->start_blkaddr
2835 		+ (blk_zone_sector(blkz) >> log_sectors_per_block);
2836 	zone_segno = GET_SEGNO(sbi, zone_block);
2837 	if (zone_segno >= MAIN_SEGS(sbi))
2838 		return 0;
2839 
2840 	wp_block = dev->start_blkaddr
2841 		+ (blk_zone_wp_sector(blkz) >> log_sectors_per_block);
2842 	wp_segno = GET_SEGNO(sbi, wp_block);
2843 	wp_blkoff = wp_block - START_BLOCK(sbi, wp_segno);
2844 
2845 	/* if a curseg points to the zone, skip the check */
2846 	for (cs_index = 0; cs_index < NO_CHECK_TYPE; cs_index++) {
2847 		cs = &SM_I(sbi)->curseg_array[cs_index];
2848 		if (zone_segno <= cs->segno &&
2849 		    cs->segno < zone_segno + segs_per_zone)
2850 			return 0;
2851 	}
2852 
2853 	last_valid_blkoff = last_vblk_off_in_zone(sbi, zone_segno);
2854 
2855 	/*
2856 	 * When there is no valid block in the zone, check write pointer is
2857 	 * at zone start. If not, reset the write pointer.
2858 	 */
2859 	if (last_valid_blkoff < 0 &&
2860 	    blk_zone_wp_sector(blkz) != blk_zone_sector(blkz)) {
2861 		if (!c.fix_on) {
2862 			MSG(0, "Inconsistent write pointer: wp[0x%x,0x%x]\n",
2863 			    wp_segno, wp_blkoff);
2864 			fsck->chk.wp_inconsistent_zones++;
2865 			return 0;
2866 		}
2867 
2868 		FIX_MSG("Reset write pointer of zone at segment 0x%x",
2869 			zone_segno);
2870 		ret = f2fs_reset_zone(wpd->dev_index, blkz);
2871 		if (ret) {
2872 			printf("[FSCK] Write pointer reset failed: %s\n",
2873 			       dev->path);
2874 			return ret;
2875 		}
2876 		fsck->chk.wp_fixed = 1;
2877 		return 0;
2878 	}
2879 
2880 	/*
2881 	 * If valid blocks exist in the zone beyond the write pointer, it
2882 	 * is a bug. No need to fix because the zone is not selected for the
2883 	 * write. Just report it.
2884 	 */
2885 	if (last_valid_blkoff + zone_block > wp_block) {
2886 		MSG(0, "Unexpected invalid write pointer: wp[0x%x,0x%x]\n",
2887 		    wp_segno, wp_blkoff);
2888 		return 0;
2889 	}
2890 
2891 	return 0;
2892 }
2893 
fix_wp_sit_alignment(struct f2fs_sb_info * sbi)2894 static void fix_wp_sit_alignment(struct f2fs_sb_info *sbi)
2895 {
2896 	unsigned int i;
2897 	struct write_pointer_check_data wpd = {	sbi, 0 };
2898 
2899 	if (c.zoned_model != F2FS_ZONED_HM)
2900 		return;
2901 
2902 	for (i = 0; i < MAX_DEVICES; i++) {
2903 		if (!c.devices[i].path)
2904 			break;
2905 		if (c.devices[i].zoned_model != F2FS_ZONED_HM)
2906 			break;
2907 
2908 		wpd.dev_index = i;
2909 		if (f2fs_report_zones(i, chk_and_fix_wp_with_sit, &wpd)) {
2910 			printf("[FSCK] Write pointer check failed: %s\n",
2911 			       c.devices[i].path);
2912 			return;
2913 		}
2914 	}
2915 }
2916 
2917 #else
2918 
fix_wp_sit_alignment(struct f2fs_sb_info * UNUSED (sbi))2919 static void fix_wp_sit_alignment(struct f2fs_sb_info *UNUSED(sbi))
2920 {
2921 	return;
2922 }
2923 
2924 #endif
2925 
2926 /*
2927  * Check and fix consistency with write pointers at the beginning of
2928  * fsck so that following writes by fsck do not fail.
2929  */
fsck_chk_and_fix_write_pointers(struct f2fs_sb_info * sbi)2930 void fsck_chk_and_fix_write_pointers(struct f2fs_sb_info *sbi)
2931 {
2932 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2933 
2934 	if (c.zoned_model != F2FS_ZONED_HM)
2935 		return;
2936 
2937 	if (check_curseg_offsets(sbi) && c.fix_on) {
2938 		fix_curseg_info(sbi);
2939 		fsck->chk.wp_fixed = 1;
2940 	}
2941 
2942 	fix_wp_sit_alignment(sbi);
2943 }
2944 
fsck_chk_curseg_info(struct f2fs_sb_info * sbi)2945 int fsck_chk_curseg_info(struct f2fs_sb_info *sbi)
2946 {
2947 	struct curseg_info *curseg;
2948 	struct seg_entry *se;
2949 	struct f2fs_summary_block *sum_blk;
2950 	int i, ret = 0;
2951 
2952 	for (i = 0; i < NO_CHECK_TYPE; i++) {
2953 		curseg = CURSEG_I(sbi, i);
2954 		se = get_seg_entry(sbi, curseg->segno);
2955 		sum_blk = curseg->sum_blk;
2956 
2957 		if (se->type != i) {
2958 			ASSERT_MSG("Incorrect curseg [%d]: segno [0x%x] "
2959 				   "type(SIT) [%d]", i, curseg->segno,
2960 				   se->type);
2961 			if (c.fix_on || c.preen_mode)
2962 				se->type = i;
2963 			ret = -1;
2964 		}
2965 		if (i <= CURSEG_COLD_DATA && IS_SUM_DATA_SEG(sum_blk->footer)) {
2966 			continue;
2967 		} else if (i > CURSEG_COLD_DATA && IS_SUM_NODE_SEG(sum_blk->footer)) {
2968 			continue;
2969 		} else {
2970 			ASSERT_MSG("Incorrect curseg [%d]: segno [0x%x] "
2971 				   "type(SSA) [%d]", i, curseg->segno,
2972 				   sum_blk->footer.entry_type);
2973 			if (c.fix_on || c.preen_mode)
2974 				sum_blk->footer.entry_type =
2975 					i <= CURSEG_COLD_DATA ?
2976 					SUM_TYPE_DATA : SUM_TYPE_NODE;
2977 			ret = -1;
2978 		}
2979 	}
2980 
2981 	return ret;
2982 }
2983 
fsck_verify(struct f2fs_sb_info * sbi)2984 int fsck_verify(struct f2fs_sb_info *sbi)
2985 {
2986 	unsigned int i = 0;
2987 	int ret = 0;
2988 	int force = 0;
2989 	u32 nr_unref_nid = 0;
2990 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2991 	struct hard_link_node *node = NULL;
2992 
2993 	printf("\n");
2994 
2995 	if (c.zoned_model == F2FS_ZONED_HM) {
2996 		printf("[FSCK] Write pointers consistency                    ");
2997 		if (fsck->chk.wp_inconsistent_zones == 0x0) {
2998 			printf(" [Ok..]\n");
2999 		} else {
3000 			printf(" [Fail] [0x%x]\n",
3001 			       fsck->chk.wp_inconsistent_zones);
3002 			c.bug_on = 1;
3003 		}
3004 
3005 		if (fsck->chk.wp_fixed && c.fix_on)
3006 			force = 1;
3007 	}
3008 
3009 	if (c.feature & cpu_to_le32(F2FS_FEATURE_LOST_FOUND)) {
3010 		for (i = 0; i < fsck->nr_nat_entries; i++)
3011 			if (f2fs_test_bit(i, fsck->nat_area_bitmap) != 0)
3012 				break;
3013 		if (i < fsck->nr_nat_entries) {
3014 			i = fsck_reconnect_file(sbi);
3015 			printf("[FSCK] Reconnect %u files to lost+found\n", i);
3016 		}
3017 	}
3018 
3019 	for (i = 0; i < fsck->nr_nat_entries; i++) {
3020 		if (f2fs_test_bit(i, fsck->nat_area_bitmap) != 0) {
3021 			struct node_info ni;
3022 
3023 			get_node_info(sbi, i, &ni);
3024 			printf("NID[0x%x] is unreachable, blkaddr:0x%x\n",
3025 							i, ni.blk_addr);
3026 			nr_unref_nid++;
3027 		}
3028 	}
3029 
3030 	if (fsck->hard_link_list_head != NULL) {
3031 		node = fsck->hard_link_list_head;
3032 		while (node) {
3033 			printf("NID[0x%x] has [0x%x] more unreachable links\n",
3034 					node->nid, node->links);
3035 			node = node->next;
3036 		}
3037 		c.bug_on = 1;
3038 	}
3039 
3040 	printf("[FSCK] Unreachable nat entries                       ");
3041 	if (nr_unref_nid == 0x0) {
3042 		printf(" [Ok..] [0x%x]\n", nr_unref_nid);
3043 	} else {
3044 		printf(" [Fail] [0x%x]\n", nr_unref_nid);
3045 		ret = EXIT_ERR_CODE;
3046 		c.bug_on = 1;
3047 	}
3048 
3049 	printf("[FSCK] SIT valid block bitmap checking                ");
3050 	if (memcmp(fsck->sit_area_bitmap, fsck->main_area_bitmap,
3051 					fsck->sit_area_bitmap_sz) == 0x0) {
3052 		printf("[Ok..]\n");
3053 	} else {
3054 		printf("[Fail]\n");
3055 		ret = EXIT_ERR_CODE;
3056 		c.bug_on = 1;
3057 	}
3058 
3059 	printf("[FSCK] Hard link checking for regular file           ");
3060 	if (fsck->hard_link_list_head == NULL) {
3061 		printf(" [Ok..] [0x%x]\n", fsck->chk.multi_hard_link_files);
3062 	} else {
3063 		printf(" [Fail] [0x%x]\n", fsck->chk.multi_hard_link_files);
3064 		ret = EXIT_ERR_CODE;
3065 		c.bug_on = 1;
3066 	}
3067 
3068 	printf("[FSCK] valid_block_count matching with CP            ");
3069 	if (sbi->total_valid_block_count == fsck->chk.valid_blk_cnt) {
3070 		printf(" [Ok..] [0x%x]\n", (u32)fsck->chk.valid_blk_cnt);
3071 	} else {
3072 		printf(" [Fail] [0x%x]\n", (u32)fsck->chk.valid_blk_cnt);
3073 		ret = EXIT_ERR_CODE;
3074 		c.bug_on = 1;
3075 	}
3076 
3077 	printf("[FSCK] valid_node_count matching with CP (de lookup) ");
3078 	if (sbi->total_valid_node_count == fsck->chk.valid_node_cnt) {
3079 		printf(" [Ok..] [0x%x]\n", fsck->chk.valid_node_cnt);
3080 	} else {
3081 		printf(" [Fail] [0x%x]\n", fsck->chk.valid_node_cnt);
3082 		ret = EXIT_ERR_CODE;
3083 		c.bug_on = 1;
3084 	}
3085 
3086 	printf("[FSCK] valid_node_count matching with CP (nat lookup)");
3087 	if (sbi->total_valid_node_count == fsck->chk.valid_nat_entry_cnt) {
3088 		printf(" [Ok..] [0x%x]\n", fsck->chk.valid_nat_entry_cnt);
3089 	} else {
3090 		printf(" [Fail] [0x%x]\n", fsck->chk.valid_nat_entry_cnt);
3091 		ret = EXIT_ERR_CODE;
3092 		c.bug_on = 1;
3093 	}
3094 
3095 	printf("[FSCK] valid_inode_count matched with CP             ");
3096 	if (sbi->total_valid_inode_count == fsck->chk.valid_inode_cnt) {
3097 		printf(" [Ok..] [0x%x]\n", fsck->chk.valid_inode_cnt);
3098 	} else {
3099 		printf(" [Fail] [0x%x]\n", fsck->chk.valid_inode_cnt);
3100 		ret = EXIT_ERR_CODE;
3101 		c.bug_on = 1;
3102 	}
3103 
3104 	printf("[FSCK] free segment_count matched with CP            ");
3105 	if (le32_to_cpu(F2FS_CKPT(sbi)->free_segment_count) ==
3106 						fsck->chk.sit_free_segs) {
3107 		printf(" [Ok..] [0x%x]\n", fsck->chk.sit_free_segs);
3108 	} else {
3109 		printf(" [Fail] [0x%x]\n", fsck->chk.sit_free_segs);
3110 		ret = EXIT_ERR_CODE;
3111 		c.bug_on = 1;
3112 	}
3113 
3114 	printf("[FSCK] next block offset is free                     ");
3115 	if (check_curseg_offsets(sbi) == 0) {
3116 		printf(" [Ok..]\n");
3117 	} else {
3118 		printf(" [Fail]\n");
3119 		ret = EXIT_ERR_CODE;
3120 		c.bug_on = 1;
3121 	}
3122 
3123 	printf("[FSCK] fixing SIT types\n");
3124 	if (check_sit_types(sbi) != 0)
3125 		force = 1;
3126 
3127 	printf("[FSCK] other corrupted bugs                          ");
3128 	if (c.bug_on == 0) {
3129 		printf(" [Ok..]\n");
3130 	} else {
3131 		printf(" [Fail]\n");
3132 		ret = EXIT_ERR_CODE;
3133 	}
3134 
3135 #ifndef WITH_ANDROID
3136 	if (nr_unref_nid && !c.ro) {
3137 		char ans[255] = {0};
3138 
3139 		printf("\nDo you want to restore lost files into ./lost_found/? [Y/N] ");
3140 		ret = scanf("%s", ans);
3141 		ASSERT(ret >= 0);
3142 		if (!strcasecmp(ans, "y")) {
3143 			for (i = 0; i < fsck->nr_nat_entries; i++) {
3144 				if (f2fs_test_bit(i, fsck->nat_area_bitmap))
3145 					dump_node(sbi, i, 1);
3146 			}
3147 		}
3148 	}
3149 #endif
3150 
3151 	/* fix global metadata */
3152 	if (force || (c.fix_on && f2fs_dev_is_writable())) {
3153 		struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
3154 
3155 		if (force || c.bug_on || c.bug_nat_bits) {
3156 			/* flush nats to write_nit_bits below */
3157 			flush_journal_entries(sbi);
3158 			fix_hard_links(sbi);
3159 			fix_nat_entries(sbi);
3160 			rewrite_sit_area_bitmap(sbi);
3161 			fix_wp_sit_alignment(sbi);
3162 			fix_curseg_info(sbi);
3163 			fix_checksum(sbi);
3164 			fix_checkpoints(sbi);
3165 		} else if (is_set_ckpt_flags(cp, CP_FSCK_FLAG) ||
3166 			is_set_ckpt_flags(cp, CP_QUOTA_NEED_FSCK_FLAG)) {
3167 			write_checkpoints(sbi);
3168 		}
3169 		/* to return FSCK_ERROR_CORRECTED */
3170 		ret = 0;
3171 	}
3172 	return ret;
3173 }
3174 
fsck_free(struct f2fs_sb_info * sbi)3175 void fsck_free(struct f2fs_sb_info *sbi)
3176 {
3177 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
3178 
3179 	if (fsck->qctx)
3180 		quota_release_context(&fsck->qctx);
3181 
3182 	if (fsck->main_area_bitmap)
3183 		free(fsck->main_area_bitmap);
3184 
3185 	if (fsck->nat_area_bitmap)
3186 		free(fsck->nat_area_bitmap);
3187 
3188 	if (fsck->sit_area_bitmap)
3189 		free(fsck->sit_area_bitmap);
3190 
3191 	if (fsck->entries)
3192 		free(fsck->entries);
3193 
3194 	if (tree_mark)
3195 		free(tree_mark);
3196 }
3197