1 /**
2  * mount.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 "node.h"
13 #include "xattr.h"
14 #include <locale.h>
15 #include <stdbool.h>
16 #ifdef HAVE_LINUX_POSIX_ACL_H
17 #include <linux/posix_acl.h>
18 #endif
19 #ifdef HAVE_SYS_ACL_H
20 #include <sys/acl.h>
21 #endif
22 
23 #ifndef ACL_UNDEFINED_TAG
24 #define ACL_UNDEFINED_TAG	(0x00)
25 #define ACL_USER_OBJ		(0x01)
26 #define ACL_USER		(0x02)
27 #define ACL_GROUP_OBJ		(0x04)
28 #define ACL_GROUP		(0x08)
29 #define ACL_MASK		(0x10)
30 #define ACL_OTHER		(0x20)
31 #endif
32 
get_device_idx(struct f2fs_sb_info * sbi,u_int32_t segno)33 static int get_device_idx(struct f2fs_sb_info *sbi, u_int32_t segno)
34 {
35 	block_t seg_start_blkaddr;
36 	int i;
37 
38 	seg_start_blkaddr = SM_I(sbi)->main_blkaddr +
39 				segno * DEFAULT_BLOCKS_PER_SEGMENT;
40 	for (i = 0; i < c.ndevs; i++)
41 		if (c.devices[i].start_blkaddr <= seg_start_blkaddr &&
42 			c.devices[i].end_blkaddr > seg_start_blkaddr)
43 			return i;
44 	return 0;
45 }
46 
47 #ifdef HAVE_LINUX_BLKZONED_H
48 
get_zone_idx_from_dev(struct f2fs_sb_info * sbi,u_int32_t segno,u_int32_t dev_idx)49 static int get_zone_idx_from_dev(struct f2fs_sb_info *sbi,
50 					u_int32_t segno, u_int32_t dev_idx)
51 {
52 	block_t seg_start_blkaddr = START_BLOCK(sbi, segno);
53 
54 	return (seg_start_blkaddr - c.devices[dev_idx].start_blkaddr) >>
55 			log_base_2(sbi->segs_per_sec * sbi->blocks_per_seg);
56 }
57 
is_usable_seg(struct f2fs_sb_info * sbi,unsigned int segno)58 bool is_usable_seg(struct f2fs_sb_info *sbi, unsigned int segno)
59 {
60 	unsigned int secno = segno / sbi->segs_per_sec;
61 	block_t seg_start = START_BLOCK(sbi, segno);
62 	block_t blocks_per_sec = sbi->blocks_per_seg * sbi->segs_per_sec;
63 	unsigned int dev_idx = get_device_idx(sbi, segno);
64 	unsigned int zone_idx = get_zone_idx_from_dev(sbi, segno, dev_idx);
65 	unsigned int sec_off = SM_I(sbi)->main_blkaddr >>
66 						log_base_2(blocks_per_sec);
67 
68 	if (zone_idx < c.devices[dev_idx].nr_rnd_zones)
69 		return true;
70 
71 	if (c.devices[dev_idx].zoned_model != F2FS_ZONED_HM)
72 		return true;
73 
74 	return seg_start < ((sec_off + secno) * blocks_per_sec) +
75 				c.devices[dev_idx].zone_cap_blocks[zone_idx];
76 }
77 
get_usable_seg_count(struct f2fs_sb_info * sbi)78 unsigned int get_usable_seg_count(struct f2fs_sb_info *sbi)
79 {
80 	unsigned int i, usable_seg_count = 0;
81 
82 	for (i = 0; i < TOTAL_SEGS(sbi); i++)
83 		if (is_usable_seg(sbi, i))
84 			usable_seg_count++;
85 
86 	return usable_seg_count;
87 }
88 
89 #else
90 
is_usable_seg(struct f2fs_sb_info * UNUSED (sbi),unsigned int UNUSED (segno))91 bool is_usable_seg(struct f2fs_sb_info *UNUSED(sbi), unsigned int UNUSED(segno))
92 {
93 	return true;
94 }
95 
get_usable_seg_count(struct f2fs_sb_info * sbi)96 unsigned int get_usable_seg_count(struct f2fs_sb_info *sbi)
97 {
98 	return TOTAL_SEGS(sbi);
99 }
100 
101 #endif
102 
get_free_segments(struct f2fs_sb_info * sbi)103 u32 get_free_segments(struct f2fs_sb_info *sbi)
104 {
105 	u32 i, free_segs = 0;
106 
107 	for (i = 0; i < TOTAL_SEGS(sbi); i++) {
108 		struct seg_entry *se = get_seg_entry(sbi, i);
109 
110 		if (se->valid_blocks == 0x0 && !IS_CUR_SEGNO(sbi, i) &&
111 							is_usable_seg(sbi, i))
112 			free_segs++;
113 	}
114 	return free_segs;
115 }
116 
update_free_segments(struct f2fs_sb_info * sbi)117 void update_free_segments(struct f2fs_sb_info *sbi)
118 {
119 	char *progress = "-*|*-";
120 	static int i = 0;
121 
122 	if (c.dbg_lv)
123 		return;
124 
125 	MSG(0, "\r [ %c ] Free segments: 0x%x", progress[i % 5], get_free_segments(sbi));
126 	fflush(stdout);
127 	i++;
128 }
129 
130 #if defined(HAVE_LINUX_POSIX_ACL_H) || defined(HAVE_SYS_ACL_H)
print_acl(const u8 * value,int size)131 static void print_acl(const u8 *value, int size)
132 {
133 	const struct f2fs_acl_header *hdr = (struct f2fs_acl_header *)value;
134 	const struct f2fs_acl_entry *entry = (struct f2fs_acl_entry *)(hdr + 1);
135 	const u8 *end = value + size;
136 	int i, count;
137 
138 	if (hdr->a_version != cpu_to_le32(F2FS_ACL_VERSION)) {
139 		MSG(0, "Invalid ACL version [0x%x : 0x%x]\n",
140 				le32_to_cpu(hdr->a_version), F2FS_ACL_VERSION);
141 		return;
142 	}
143 
144 	count = f2fs_acl_count(size);
145 	if (count <= 0) {
146 		MSG(0, "Invalid ACL value size %d\n", size);
147 		return;
148 	}
149 
150 	for (i = 0; i < count; i++) {
151 		if ((u8 *)entry > end) {
152 			MSG(0, "Invalid ACL entries count %d\n", count);
153 			return;
154 		}
155 
156 		switch (le16_to_cpu(entry->e_tag)) {
157 		case ACL_USER_OBJ:
158 		case ACL_GROUP_OBJ:
159 		case ACL_MASK:
160 		case ACL_OTHER:
161 			MSG(0, "tag:0x%x perm:0x%x\n",
162 					le16_to_cpu(entry->e_tag),
163 					le16_to_cpu(entry->e_perm));
164 			entry = (struct f2fs_acl_entry *)((char *)entry +
165 					sizeof(struct f2fs_acl_entry_short));
166 			break;
167 		case ACL_USER:
168 			MSG(0, "tag:0x%x perm:0x%x uid:%u\n",
169 					le16_to_cpu(entry->e_tag),
170 					le16_to_cpu(entry->e_perm),
171 					le32_to_cpu(entry->e_id));
172 			entry = (struct f2fs_acl_entry *)((char *)entry +
173 					sizeof(struct f2fs_acl_entry));
174 			break;
175 		case ACL_GROUP:
176 			MSG(0, "tag:0x%x perm:0x%x gid:%u\n",
177 					le16_to_cpu(entry->e_tag),
178 					le16_to_cpu(entry->e_perm),
179 					le32_to_cpu(entry->e_id));
180 			entry = (struct f2fs_acl_entry *)((char *)entry +
181 					sizeof(struct f2fs_acl_entry));
182 			break;
183 		default:
184 			MSG(0, "Unknown ACL tag 0x%x\n",
185 					le16_to_cpu(entry->e_tag));
186 			return;
187 		}
188 	}
189 }
190 #endif /* HAVE_LINUX_POSIX_ACL_H || HAVE_SYS_ACL_H */
191 
print_xattr_entry(const struct f2fs_xattr_entry * ent)192 static void print_xattr_entry(const struct f2fs_xattr_entry *ent)
193 {
194 	const u8 *value = (const u8 *)&ent->e_name[ent->e_name_len];
195 	const int size = le16_to_cpu(ent->e_value_size);
196 	const struct fscrypt_context *ctx;
197 	int i;
198 
199 	MSG(0, "\nxattr: e_name_index:%d e_name:", ent->e_name_index);
200 	for (i = 0; i < ent->e_name_len; i++)
201 		MSG(0, "%c", ent->e_name[i]);
202 	MSG(0, " e_name_len:%d e_value_size:%d e_value:\n",
203 			ent->e_name_len, size);
204 
205 	switch (ent->e_name_index) {
206 #if defined(HAVE_LINUX_POSIX_ACL_H) || defined(HAVE_SYS_ACL_H)
207 	case F2FS_XATTR_INDEX_POSIX_ACL_ACCESS:
208 	case F2FS_XATTR_INDEX_POSIX_ACL_DEFAULT:
209 		print_acl(value, size);
210 		return;
211 #endif
212 	case F2FS_XATTR_INDEX_ENCRYPTION:
213 		ctx = (const struct fscrypt_context *)value;
214 		if (size != sizeof(*ctx) ||
215 		    ctx->format != FS_ENCRYPTION_CONTEXT_FORMAT_V1)
216 			break;
217 		MSG(0, "format: %d\n", ctx->format);
218 		MSG(0, "contents_encryption_mode: 0x%x\n", ctx->contents_encryption_mode);
219 		MSG(0, "filenames_encryption_mode: 0x%x\n", ctx->filenames_encryption_mode);
220 		MSG(0, "flags: 0x%x\n", ctx->flags);
221 		MSG(0, "master_key_descriptor: ");
222 		for (i = 0; i < FS_KEY_DESCRIPTOR_SIZE; i++)
223 			MSG(0, "%02X", ctx->master_key_descriptor[i]);
224 		MSG(0, "\nnonce: ");
225 		for (i = 0; i < FS_KEY_DERIVATION_NONCE_SIZE; i++)
226 			MSG(0, "%02X", ctx->nonce[i]);
227 		MSG(0, "\n");
228 		return;
229 	}
230 	for (i = 0; i < size; i++)
231 		MSG(0, "%02X", value[i]);
232 	MSG(0, "\n");
233 }
234 
print_inode_info(struct f2fs_sb_info * sbi,struct f2fs_node * node,int name)235 void print_inode_info(struct f2fs_sb_info *sbi,
236 			struct f2fs_node *node, int name)
237 {
238 	struct f2fs_inode *inode = &node->i;
239 	void *xattr_addr;
240 	struct f2fs_xattr_entry *ent;
241 	char en[F2FS_PRINT_NAMELEN];
242 	unsigned int i = 0;
243 	u32 namelen = le32_to_cpu(inode->i_namelen);
244 	int enc_name = file_enc_name(inode);
245 	int ofs = get_extra_isize(node);
246 
247 	pretty_print_filename(inode->i_name, namelen, en, enc_name);
248 	if (name && en[0]) {
249 		MSG(0, " - File name         : %s%s\n", en,
250 				enc_name ? " <encrypted>" : "");
251 		setlocale(LC_ALL, "");
252 		MSG(0, " - File size         : %'llu (bytes)\n",
253 				le64_to_cpu(inode->i_size));
254 		return;
255 	}
256 
257 	DISP_u32(inode, i_mode);
258 	DISP_u32(inode, i_advise);
259 	DISP_u32(inode, i_uid);
260 	DISP_u32(inode, i_gid);
261 	DISP_u32(inode, i_links);
262 	DISP_u64(inode, i_size);
263 	DISP_u64(inode, i_blocks);
264 
265 	DISP_u64(inode, i_atime);
266 	DISP_u32(inode, i_atime_nsec);
267 	DISP_u64(inode, i_ctime);
268 	DISP_u32(inode, i_ctime_nsec);
269 	DISP_u64(inode, i_mtime);
270 	DISP_u32(inode, i_mtime_nsec);
271 
272 	DISP_u32(inode, i_generation);
273 	DISP_u32(inode, i_current_depth);
274 	DISP_u32(inode, i_xattr_nid);
275 	DISP_u32(inode, i_flags);
276 	DISP_u32(inode, i_inline);
277 	DISP_u32(inode, i_pino);
278 	DISP_u32(inode, i_dir_level);
279 
280 	if (en[0]) {
281 		DISP_u32(inode, i_namelen);
282 		printf("%-30s\t\t[%s]\n", "i_name", en);
283 	}
284 
285 	printf("i_ext: fofs:%x blkaddr:%x len:%x\n",
286 			le32_to_cpu(inode->i_ext.fofs),
287 			le32_to_cpu(inode->i_ext.blk_addr),
288 			le32_to_cpu(inode->i_ext.len));
289 
290 	if (c.feature & cpu_to_le32(F2FS_FEATURE_EXTRA_ATTR)) {
291 		DISP_u16(inode, i_extra_isize);
292 		if (c.feature & cpu_to_le32(F2FS_FEATURE_FLEXIBLE_INLINE_XATTR))
293 			DISP_u16(inode, i_inline_xattr_size);
294 		if (c.feature & cpu_to_le32(F2FS_FEATURE_PRJQUOTA))
295 			DISP_u32(inode, i_projid);
296 		if (c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM))
297 			DISP_u32(inode, i_inode_checksum);
298 		if (c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CRTIME)) {
299 			DISP_u64(inode, i_crtime);
300 			DISP_u32(inode, i_crtime_nsec);
301 		}
302 		if (c.feature & cpu_to_le32(F2FS_FEATURE_COMPRESSION)) {
303 			DISP_u64(inode, i_compr_blocks);
304 			DISP_u32(inode, i_compress_algrithm);
305 			DISP_u32(inode, i_log_cluster_size);
306 			DISP_u32(inode, i_padding);
307 		}
308 	}
309 
310 	for (i = 0; i < ADDRS_PER_INODE(inode); i++) {
311 		block_t blkaddr;
312 		char *flag = "";
313 
314 		if (i + ofs >= DEF_ADDRS_PER_INODE)
315 			break;
316 
317 		blkaddr = le32_to_cpu(inode->i_addr[i + ofs]);
318 
319 		if (blkaddr == 0x0)
320 			continue;
321 		if (blkaddr == COMPRESS_ADDR)
322 			flag = "cluster flag";
323 		else if (blkaddr == NEW_ADDR)
324 			flag = "reserved flag";
325 		printf("i_addr[0x%x] %-16s\t\t[0x%8x : %u]\n", i + ofs, flag,
326 				blkaddr, blkaddr);
327 	}
328 
329 	DISP_u32(inode, i_nid[0]);	/* direct */
330 	DISP_u32(inode, i_nid[1]);	/* direct */
331 	DISP_u32(inode, i_nid[2]);	/* indirect */
332 	DISP_u32(inode, i_nid[3]);	/* indirect */
333 	DISP_u32(inode, i_nid[4]);	/* double indirect */
334 
335 	xattr_addr = read_all_xattrs(sbi, node);
336 	if (xattr_addr) {
337 		list_for_each_xattr(ent, xattr_addr) {
338 			print_xattr_entry(ent);
339 		}
340 		free(xattr_addr);
341 	}
342 
343 	printf("\n");
344 }
345 
print_node_info(struct f2fs_sb_info * sbi,struct f2fs_node * node_block,int verbose)346 void print_node_info(struct f2fs_sb_info *sbi,
347 			struct f2fs_node *node_block, int verbose)
348 {
349 	nid_t ino = le32_to_cpu(node_block->footer.ino);
350 	nid_t nid = le32_to_cpu(node_block->footer.nid);
351 	/* Is this inode? */
352 	if (ino == nid) {
353 		DBG(verbose, "Node ID [0x%x:%u] is inode\n", nid, nid);
354 		print_inode_info(sbi, node_block, verbose);
355 	} else {
356 		int i;
357 		u32 *dump_blk = (u32 *)node_block;
358 		DBG(verbose,
359 			"Node ID [0x%x:%u] is direct node or indirect node.\n",
360 								nid, nid);
361 		for (i = 0; i < DEF_ADDRS_PER_BLOCK; i++)
362 			MSG(verbose, "[%d]\t\t\t[0x%8x : %d]\n",
363 						i, dump_blk[i], dump_blk[i]);
364 	}
365 }
366 
DISP_label(u_int16_t * name)367 static void DISP_label(u_int16_t *name)
368 {
369 	char buffer[MAX_VOLUME_NAME];
370 
371 	utf16_to_utf8(buffer, name, MAX_VOLUME_NAME, MAX_VOLUME_NAME);
372 	if (c.layout)
373 		printf("%-30s %s\n", "Filesystem volume name:", buffer);
374 	else
375 		printf("%-30s" "\t\t[%s]\n", "volum_name", buffer);
376 }
377 
print_raw_sb_info(struct f2fs_super_block * sb)378 void print_raw_sb_info(struct f2fs_super_block *sb)
379 {
380 	if (c.layout)
381 		goto printout;
382 	if (!c.dbg_lv)
383 		return;
384 
385 	printf("\n");
386 	printf("+--------------------------------------------------------+\n");
387 	printf("| Super block                                            |\n");
388 	printf("+--------------------------------------------------------+\n");
389 printout:
390 	DISP_u32(sb, magic);
391 	DISP_u32(sb, major_ver);
392 
393 	DISP_label(sb->volume_name);
394 
395 	DISP_u32(sb, minor_ver);
396 	DISP_u32(sb, log_sectorsize);
397 	DISP_u32(sb, log_sectors_per_block);
398 
399 	DISP_u32(sb, log_blocksize);
400 	DISP_u32(sb, log_blocks_per_seg);
401 	DISP_u32(sb, segs_per_sec);
402 	DISP_u32(sb, secs_per_zone);
403 	DISP_u32(sb, checksum_offset);
404 	DISP_u64(sb, block_count);
405 
406 	DISP_u32(sb, section_count);
407 	DISP_u32(sb, segment_count);
408 	DISP_u32(sb, segment_count_ckpt);
409 	DISP_u32(sb, segment_count_sit);
410 	DISP_u32(sb, segment_count_nat);
411 
412 	DISP_u32(sb, segment_count_ssa);
413 	DISP_u32(sb, segment_count_main);
414 	DISP_u32(sb, segment0_blkaddr);
415 
416 	DISP_u32(sb, cp_blkaddr);
417 	DISP_u32(sb, sit_blkaddr);
418 	DISP_u32(sb, nat_blkaddr);
419 	DISP_u32(sb, ssa_blkaddr);
420 	DISP_u32(sb, main_blkaddr);
421 
422 	DISP_u32(sb, root_ino);
423 	DISP_u32(sb, node_ino);
424 	DISP_u32(sb, meta_ino);
425 	DISP_u32(sb, cp_payload);
426 	DISP_u32(sb, crc);
427 	DISP("%-.256s", sb, version);
428 	printf("\n");
429 }
430 
print_ckpt_info(struct f2fs_sb_info * sbi)431 void print_ckpt_info(struct f2fs_sb_info *sbi)
432 {
433 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
434 
435 	if (c.layout)
436 		goto printout;
437 	if (!c.dbg_lv)
438 		return;
439 
440 	printf("\n");
441 	printf("+--------------------------------------------------------+\n");
442 	printf("| Checkpoint                                             |\n");
443 	printf("+--------------------------------------------------------+\n");
444 printout:
445 	DISP_u64(cp, checkpoint_ver);
446 	DISP_u64(cp, user_block_count);
447 	DISP_u64(cp, valid_block_count);
448 	DISP_u32(cp, rsvd_segment_count);
449 	DISP_u32(cp, overprov_segment_count);
450 	DISP_u32(cp, free_segment_count);
451 
452 	DISP_u32(cp, alloc_type[CURSEG_HOT_NODE]);
453 	DISP_u32(cp, alloc_type[CURSEG_WARM_NODE]);
454 	DISP_u32(cp, alloc_type[CURSEG_COLD_NODE]);
455 	DISP_u32(cp, cur_node_segno[0]);
456 	DISP_u32(cp, cur_node_segno[1]);
457 	DISP_u32(cp, cur_node_segno[2]);
458 
459 	DISP_u32(cp, cur_node_blkoff[0]);
460 	DISP_u32(cp, cur_node_blkoff[1]);
461 	DISP_u32(cp, cur_node_blkoff[2]);
462 
463 
464 	DISP_u32(cp, alloc_type[CURSEG_HOT_DATA]);
465 	DISP_u32(cp, alloc_type[CURSEG_WARM_DATA]);
466 	DISP_u32(cp, alloc_type[CURSEG_COLD_DATA]);
467 	DISP_u32(cp, cur_data_segno[0]);
468 	DISP_u32(cp, cur_data_segno[1]);
469 	DISP_u32(cp, cur_data_segno[2]);
470 
471 	DISP_u32(cp, cur_data_blkoff[0]);
472 	DISP_u32(cp, cur_data_blkoff[1]);
473 	DISP_u32(cp, cur_data_blkoff[2]);
474 
475 	DISP_u32(cp, ckpt_flags);
476 	DISP_u32(cp, cp_pack_total_block_count);
477 	DISP_u32(cp, cp_pack_start_sum);
478 	DISP_u32(cp, valid_node_count);
479 	DISP_u32(cp, valid_inode_count);
480 	DISP_u32(cp, next_free_nid);
481 	DISP_u32(cp, sit_ver_bitmap_bytesize);
482 	DISP_u32(cp, nat_ver_bitmap_bytesize);
483 	DISP_u32(cp, checksum_offset);
484 	DISP_u64(cp, elapsed_time);
485 
486 	DISP_u32(cp, sit_nat_version_bitmap[0]);
487 	printf("\n\n");
488 }
489 
print_cp_state(u32 flag)490 void print_cp_state(u32 flag)
491 {
492 	MSG(0, "Info: checkpoint state = %x : ", flag);
493 	if (flag & CP_QUOTA_NEED_FSCK_FLAG)
494 		MSG(0, "%s", " quota_need_fsck");
495 	if (flag & CP_LARGE_NAT_BITMAP_FLAG)
496 		MSG(0, "%s", " large_nat_bitmap");
497 	if (flag & CP_NOCRC_RECOVERY_FLAG)
498 		MSG(0, "%s", " allow_nocrc");
499 	if (flag & CP_TRIMMED_FLAG)
500 		MSG(0, "%s", " trimmed");
501 	if (flag & CP_NAT_BITS_FLAG)
502 		MSG(0, "%s", " nat_bits");
503 	if (flag & CP_CRC_RECOVERY_FLAG)
504 		MSG(0, "%s", " crc");
505 	if (flag & CP_FASTBOOT_FLAG)
506 		MSG(0, "%s", " fastboot");
507 	if (flag & CP_FSCK_FLAG)
508 		MSG(0, "%s", " fsck");
509 	if (flag & CP_ERROR_FLAG)
510 		MSG(0, "%s", " error");
511 	if (flag & CP_COMPACT_SUM_FLAG)
512 		MSG(0, "%s", " compacted_summary");
513 	if (flag & CP_ORPHAN_PRESENT_FLAG)
514 		MSG(0, "%s", " orphan_inodes");
515 	if (flag & CP_DISABLED_FLAG)
516 		MSG(0, "%s", " disabled");
517 	if (flag & CP_RESIZEFS_FLAG)
518 		MSG(0, "%s", " resizefs");
519 	if (flag & CP_UMOUNT_FLAG)
520 		MSG(0, "%s", " unmount");
521 	else
522 		MSG(0, "%s", " sudden-power-off");
523 	MSG(0, "\n");
524 }
525 
print_sb_state(struct f2fs_super_block * sb)526 void print_sb_state(struct f2fs_super_block *sb)
527 {
528 	__le32 f = sb->feature;
529 	int i;
530 
531 	MSG(0, "Info: superblock features = %x : ", f);
532 	if (f & cpu_to_le32(F2FS_FEATURE_ENCRYPT)) {
533 		MSG(0, "%s", " encrypt");
534 	}
535 	if (f & cpu_to_le32(F2FS_FEATURE_VERITY)) {
536 		MSG(0, "%s", " verity");
537 	}
538 	if (f & cpu_to_le32(F2FS_FEATURE_BLKZONED)) {
539 		MSG(0, "%s", " blkzoned");
540 	}
541 	if (f & cpu_to_le32(F2FS_FEATURE_EXTRA_ATTR)) {
542 		MSG(0, "%s", " extra_attr");
543 	}
544 	if (f & cpu_to_le32(F2FS_FEATURE_PRJQUOTA)) {
545 		MSG(0, "%s", " project_quota");
546 	}
547 	if (f & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM)) {
548 		MSG(0, "%s", " inode_checksum");
549 	}
550 	if (f & cpu_to_le32(F2FS_FEATURE_FLEXIBLE_INLINE_XATTR)) {
551 		MSG(0, "%s", " flexible_inline_xattr");
552 	}
553 	if (f & cpu_to_le32(F2FS_FEATURE_QUOTA_INO)) {
554 		MSG(0, "%s", " quota_ino");
555 	}
556 	if (f & cpu_to_le32(F2FS_FEATURE_INODE_CRTIME)) {
557 		MSG(0, "%s", " inode_crtime");
558 	}
559 	if (f & cpu_to_le32(F2FS_FEATURE_LOST_FOUND)) {
560 		MSG(0, "%s", " lost_found");
561 	}
562 	if (f & cpu_to_le32(F2FS_FEATURE_SB_CHKSUM)) {
563 		MSG(0, "%s", " sb_checksum");
564 	}
565 	if (f & cpu_to_le32(F2FS_FEATURE_CASEFOLD)) {
566 		MSG(0, "%s", " casefold");
567 	}
568 	if (f & cpu_to_le32(F2FS_FEATURE_COMPRESSION)) {
569 		MSG(0, "%s", " compression");
570 	}
571 	if (f & cpu_to_le32(F2FS_FEATURE_RO)) {
572 		MSG(0, "%s", " ro");
573 	}
574 	MSG(0, "\n");
575 	MSG(0, "Info: superblock encrypt level = %d, salt = ",
576 					sb->encryption_level);
577 	for (i = 0; i < 16; i++)
578 		MSG(0, "%02x", sb->encrypt_pw_salt[i]);
579 	MSG(0, "\n");
580 }
581 
is_valid_data_blkaddr(block_t blkaddr)582 static inline bool is_valid_data_blkaddr(block_t blkaddr)
583 {
584 	if (blkaddr == NEW_ADDR || blkaddr == NULL_ADDR ||
585 				blkaddr == COMPRESS_ADDR)
586 		return 0;
587 	return 1;
588 }
589 
f2fs_is_valid_blkaddr(struct f2fs_sb_info * sbi,block_t blkaddr,int type)590 bool f2fs_is_valid_blkaddr(struct f2fs_sb_info *sbi,
591 					block_t blkaddr, int type)
592 {
593 	switch (type) {
594 	case META_NAT:
595 		break;
596 	case META_SIT:
597 		if (blkaddr >= SIT_BLK_CNT(sbi))
598 			return 0;
599 		break;
600 	case META_SSA:
601 		if (blkaddr >= MAIN_BLKADDR(sbi) ||
602 			blkaddr < SM_I(sbi)->ssa_blkaddr)
603 			return 0;
604 		break;
605 	case META_CP:
606 		if (blkaddr >= SIT_I(sbi)->sit_base_addr ||
607 			blkaddr < __start_cp_addr(sbi))
608 			return 0;
609 		break;
610 	case META_POR:
611 		if (blkaddr >= MAX_BLKADDR(sbi) ||
612 			blkaddr < MAIN_BLKADDR(sbi))
613 			return 0;
614 		break;
615 	default:
616 		ASSERT(0);
617 	}
618 
619 	return 1;
620 }
621 
622 static inline block_t current_sit_addr(struct f2fs_sb_info *sbi,
623 						unsigned int start);
624 
625 /*
626  * Readahead CP/NAT/SIT/SSA pages
627  */
f2fs_ra_meta_pages(struct f2fs_sb_info * sbi,block_t start,int nrpages,int type)628 int f2fs_ra_meta_pages(struct f2fs_sb_info *sbi, block_t start, int nrpages,
629 							int type)
630 {
631 	block_t blkno = start;
632 	block_t blkaddr, start_blk = 0, len = 0;
633 
634 	for (; nrpages-- > 0; blkno++) {
635 
636 		if (!f2fs_is_valid_blkaddr(sbi, blkno, type))
637 			goto out;
638 
639 		switch (type) {
640 		case META_NAT:
641 			if (blkno >= NAT_BLOCK_OFFSET(NM_I(sbi)->max_nid))
642 				blkno = 0;
643 			/* get nat block addr */
644 			blkaddr = current_nat_addr(sbi,
645 					blkno * NAT_ENTRY_PER_BLOCK, NULL);
646 			break;
647 		case META_SIT:
648 			/* get sit block addr */
649 			blkaddr = current_sit_addr(sbi,
650 					blkno * SIT_ENTRY_PER_BLOCK);
651 			break;
652 		case META_SSA:
653 		case META_CP:
654 		case META_POR:
655 			blkaddr = blkno;
656 			break;
657 		default:
658 			ASSERT(0);
659 		}
660 
661 		if (!len) {
662 			start_blk = blkaddr;
663 			len = 1;
664 		} else if (start_blk + len == blkaddr) {
665 			len++;
666 		} else {
667 			dev_readahead(start_blk << F2FS_BLKSIZE_BITS,
668 						len << F2FS_BLKSIZE_BITS);
669 		}
670 	}
671 out:
672 	if (len)
673 		dev_readahead(start_blk << F2FS_BLKSIZE_BITS,
674 					len << F2FS_BLKSIZE_BITS);
675 	return blkno - start;
676 }
677 
update_superblock(struct f2fs_super_block * sb,int sb_mask)678 void update_superblock(struct f2fs_super_block *sb, int sb_mask)
679 {
680 	int addr, ret;
681 	u_int8_t *buf;
682 	u32 old_crc, new_crc;
683 
684 	buf = calloc(BLOCK_SZ, 1);
685 	ASSERT(buf);
686 
687 	if (get_sb(feature) & F2FS_FEATURE_SB_CHKSUM) {
688 		old_crc = get_sb(crc);
689 		new_crc = f2fs_cal_crc32(F2FS_SUPER_MAGIC, sb,
690 						SB_CHKSUM_OFFSET);
691 		set_sb(crc, new_crc);
692 		MSG(1, "Info: SB CRC is updated (0x%x -> 0x%x)\n",
693 							old_crc, new_crc);
694 	}
695 
696 	memcpy(buf + F2FS_SUPER_OFFSET, sb, sizeof(*sb));
697 	for (addr = SB0_ADDR; addr < SB_MAX_ADDR; addr++) {
698 		if (SB_MASK(addr) & sb_mask) {
699 			ret = dev_write_block(buf, addr);
700 			ASSERT(ret >= 0);
701 		}
702 	}
703 
704 	free(buf);
705 	DBG(0, "Info: Done to update superblock\n");
706 }
707 
sanity_check_area_boundary(struct f2fs_super_block * sb,enum SB_ADDR sb_addr)708 static inline int sanity_check_area_boundary(struct f2fs_super_block *sb,
709 							enum SB_ADDR sb_addr)
710 {
711 	u32 segment0_blkaddr = get_sb(segment0_blkaddr);
712 	u32 cp_blkaddr = get_sb(cp_blkaddr);
713 	u32 sit_blkaddr = get_sb(sit_blkaddr);
714 	u32 nat_blkaddr = get_sb(nat_blkaddr);
715 	u32 ssa_blkaddr = get_sb(ssa_blkaddr);
716 	u32 main_blkaddr = get_sb(main_blkaddr);
717 	u32 segment_count_ckpt = get_sb(segment_count_ckpt);
718 	u32 segment_count_sit = get_sb(segment_count_sit);
719 	u32 segment_count_nat = get_sb(segment_count_nat);
720 	u32 segment_count_ssa = get_sb(segment_count_ssa);
721 	u32 segment_count_main = get_sb(segment_count_main);
722 	u32 segment_count = get_sb(segment_count);
723 	u32 log_blocks_per_seg = get_sb(log_blocks_per_seg);
724 	u64 main_end_blkaddr = main_blkaddr +
725 				(segment_count_main << log_blocks_per_seg);
726 	u64 seg_end_blkaddr = segment0_blkaddr +
727 				(segment_count << log_blocks_per_seg);
728 
729 	if (segment0_blkaddr != cp_blkaddr) {
730 		MSG(0, "\tMismatch segment0(%u) cp_blkaddr(%u)\n",
731 				segment0_blkaddr, cp_blkaddr);
732 		return -1;
733 	}
734 
735 	if (cp_blkaddr + (segment_count_ckpt << log_blocks_per_seg) !=
736 							sit_blkaddr) {
737 		MSG(0, "\tWrong CP boundary, start(%u) end(%u) blocks(%u)\n",
738 			cp_blkaddr, sit_blkaddr,
739 			segment_count_ckpt << log_blocks_per_seg);
740 		return -1;
741 	}
742 
743 	if (sit_blkaddr + (segment_count_sit << log_blocks_per_seg) !=
744 							nat_blkaddr) {
745 		MSG(0, "\tWrong SIT boundary, start(%u) end(%u) blocks(%u)\n",
746 			sit_blkaddr, nat_blkaddr,
747 			segment_count_sit << log_blocks_per_seg);
748 		return -1;
749 	}
750 
751 	if (nat_blkaddr + (segment_count_nat << log_blocks_per_seg) !=
752 							ssa_blkaddr) {
753 		MSG(0, "\tWrong NAT boundary, start(%u) end(%u) blocks(%u)\n",
754 			nat_blkaddr, ssa_blkaddr,
755 			segment_count_nat << log_blocks_per_seg);
756 		return -1;
757 	}
758 
759 	if (ssa_blkaddr + (segment_count_ssa << log_blocks_per_seg) !=
760 							main_blkaddr) {
761 		MSG(0, "\tWrong SSA boundary, start(%u) end(%u) blocks(%u)\n",
762 			ssa_blkaddr, main_blkaddr,
763 			segment_count_ssa << log_blocks_per_seg);
764 		return -1;
765 	}
766 
767 	if (main_end_blkaddr > seg_end_blkaddr) {
768 		MSG(0, "\tWrong MAIN_AREA, start(%u) end(%u) block(%u)\n",
769 			main_blkaddr,
770 			segment0_blkaddr +
771 				(segment_count << log_blocks_per_seg),
772 			segment_count_main << log_blocks_per_seg);
773 		return -1;
774 	} else if (main_end_blkaddr < seg_end_blkaddr) {
775 		set_sb(segment_count, (main_end_blkaddr -
776 				segment0_blkaddr) >> log_blocks_per_seg);
777 
778 		update_superblock(sb, SB_MASK(sb_addr));
779 		MSG(0, "Info: Fix alignment: start(%u) end(%u) block(%u)\n",
780 			main_blkaddr,
781 			segment0_blkaddr +
782 				(segment_count << log_blocks_per_seg),
783 			segment_count_main << log_blocks_per_seg);
784 	}
785 	return 0;
786 }
787 
verify_sb_chksum(struct f2fs_super_block * sb)788 static int verify_sb_chksum(struct f2fs_super_block *sb)
789 {
790 	if (SB_CHKSUM_OFFSET != get_sb(checksum_offset)) {
791 		MSG(0, "\tInvalid SB CRC offset: %u\n",
792 					get_sb(checksum_offset));
793 		return -1;
794 	}
795 	if (f2fs_crc_valid(get_sb(crc), sb,
796 			get_sb(checksum_offset))) {
797 		MSG(0, "\tInvalid SB CRC: 0x%x\n", get_sb(crc));
798 		return -1;
799 	}
800 	return 0;
801 }
802 
sanity_check_raw_super(struct f2fs_super_block * sb,enum SB_ADDR sb_addr)803 int sanity_check_raw_super(struct f2fs_super_block *sb, enum SB_ADDR sb_addr)
804 {
805 	unsigned int blocksize;
806 	unsigned int segment_count, segs_per_sec, secs_per_zone, segs_per_zone;
807 	unsigned int total_sections, blocks_per_seg;
808 
809 	if ((get_sb(feature) & F2FS_FEATURE_SB_CHKSUM) &&
810 					verify_sb_chksum(sb))
811 		return -1;
812 
813 	if (F2FS_SUPER_MAGIC != get_sb(magic)) {
814 		MSG(0, "Magic Mismatch, valid(0x%x) - read(0x%x)\n",
815 			F2FS_SUPER_MAGIC, get_sb(magic));
816 		return -1;
817 	}
818 
819 	if (F2FS_BLKSIZE != PAGE_CACHE_SIZE) {
820 		MSG(0, "Invalid page_cache_size (%d), supports only 4KB\n",
821 			PAGE_CACHE_SIZE);
822 		return -1;
823 	}
824 
825 	blocksize = 1 << get_sb(log_blocksize);
826 	if (F2FS_BLKSIZE != blocksize) {
827 		MSG(0, "Invalid blocksize (%u), supports only 4KB\n",
828 			blocksize);
829 		return -1;
830 	}
831 
832 	/* check log blocks per segment */
833 	if (get_sb(log_blocks_per_seg) != 9) {
834 		MSG(0, "Invalid log blocks per segment (%u)\n",
835 			get_sb(log_blocks_per_seg));
836 		return -1;
837 	}
838 
839 	/* Currently, support 512/1024/2048/4096 bytes sector size */
840 	if (get_sb(log_sectorsize) > F2FS_MAX_LOG_SECTOR_SIZE ||
841 			get_sb(log_sectorsize) < F2FS_MIN_LOG_SECTOR_SIZE) {
842 		MSG(0, "Invalid log sectorsize (%u)\n", get_sb(log_sectorsize));
843 		return -1;
844 	}
845 
846 	if (get_sb(log_sectors_per_block) + get_sb(log_sectorsize) !=
847 						F2FS_MAX_LOG_SECTOR_SIZE) {
848 		MSG(0, "Invalid log sectors per block(%u) log sectorsize(%u)\n",
849 			get_sb(log_sectors_per_block),
850 			get_sb(log_sectorsize));
851 		return -1;
852 	}
853 
854 	segment_count = get_sb(segment_count);
855 	segs_per_sec = get_sb(segs_per_sec);
856 	secs_per_zone = get_sb(secs_per_zone);
857 	total_sections = get_sb(section_count);
858 	segs_per_zone = segs_per_sec * secs_per_zone;
859 
860 	/* blocks_per_seg should be 512, given the above check */
861 	blocks_per_seg = 1 << get_sb(log_blocks_per_seg);
862 
863 	if (segment_count > F2FS_MAX_SEGMENT ||
864 			segment_count < F2FS_MIN_SEGMENTS) {
865 		MSG(0, "\tInvalid segment count (%u)\n", segment_count);
866 		return -1;
867 	}
868 
869 	if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO)) &&
870 			(total_sections > segment_count ||
871 			total_sections < F2FS_MIN_SEGMENTS ||
872 			segs_per_sec > segment_count || !segs_per_sec)) {
873 		MSG(0, "\tInvalid segment/section count (%u, %u x %u)\n",
874 			segment_count, total_sections, segs_per_sec);
875 		return 1;
876 	}
877 
878 	if ((segment_count / segs_per_sec) < total_sections) {
879 		MSG(0, "Small segment_count (%u < %u * %u)\n",
880 			segment_count, segs_per_sec, total_sections);
881 		return 1;
882 	}
883 
884 	if (segment_count > (get_sb(block_count) >> 9)) {
885 		MSG(0, "Wrong segment_count / block_count (%u > %llu)\n",
886 			segment_count, get_sb(block_count));
887 		return 1;
888 	}
889 
890 	if (sb->devs[0].path[0]) {
891 		unsigned int dev_segs = le32_to_cpu(sb->devs[0].total_segments);
892 		int i = 1;
893 
894 		while (i < MAX_DEVICES && sb->devs[i].path[0]) {
895 			dev_segs += le32_to_cpu(sb->devs[i].total_segments);
896 			i++;
897 		}
898 		if (segment_count != dev_segs / segs_per_zone * segs_per_zone) {
899 			MSG(0, "Segment count (%u) mismatch with total segments from devices (%u)",
900 				segment_count, dev_segs);
901 			return 1;
902 		}
903 	}
904 
905 	if (secs_per_zone > total_sections || !secs_per_zone) {
906 		MSG(0, "Wrong secs_per_zone / total_sections (%u, %u)\n",
907 			secs_per_zone, total_sections);
908 		return 1;
909 	}
910 	if (get_sb(extension_count) > F2FS_MAX_EXTENSION ||
911 			sb->hot_ext_count > F2FS_MAX_EXTENSION ||
912 			get_sb(extension_count) +
913 			sb->hot_ext_count > F2FS_MAX_EXTENSION) {
914 		MSG(0, "Corrupted extension count (%u + %u > %u)\n",
915 			get_sb(extension_count),
916 			sb->hot_ext_count,
917 			F2FS_MAX_EXTENSION);
918 		return 1;
919 	}
920 
921 	if (get_sb(cp_payload) > (blocks_per_seg - F2FS_CP_PACKS)) {
922 		MSG(0, "Insane cp_payload (%u > %u)\n",
923 			get_sb(cp_payload), blocks_per_seg - F2FS_CP_PACKS);
924 		return 1;
925 	}
926 
927 	/* check reserved ino info */
928 	if (get_sb(node_ino) != 1 || get_sb(meta_ino) != 2 ||
929 						get_sb(root_ino) != 3) {
930 		MSG(0, "Invalid Fs Meta Ino: node(%u) meta(%u) root(%u)\n",
931 			get_sb(node_ino), get_sb(meta_ino), get_sb(root_ino));
932 		return -1;
933 	}
934 
935 	/* Check zoned block device feature */
936 	if (c.devices[0].zoned_model == F2FS_ZONED_HM &&
937 			!(sb->feature & cpu_to_le32(F2FS_FEATURE_BLKZONED))) {
938 		MSG(0, "\tMissing zoned block device feature\n");
939 		return -1;
940 	}
941 
942 	if (sanity_check_area_boundary(sb, sb_addr))
943 		return -1;
944 	return 0;
945 }
946 
validate_super_block(struct f2fs_sb_info * sbi,enum SB_ADDR sb_addr)947 int validate_super_block(struct f2fs_sb_info *sbi, enum SB_ADDR sb_addr)
948 {
949 	char buf[F2FS_BLKSIZE];
950 
951 	sbi->raw_super = malloc(sizeof(struct f2fs_super_block));
952 	if (!sbi->raw_super)
953 		return -ENOMEM;
954 
955 	if (dev_read_block(buf, sb_addr))
956 		return -1;
957 
958 	memcpy(sbi->raw_super, buf + F2FS_SUPER_OFFSET,
959 					sizeof(struct f2fs_super_block));
960 
961 	if (!sanity_check_raw_super(sbi->raw_super, sb_addr)) {
962 		/* get kernel version */
963 		if (c.kd >= 0) {
964 			dev_read_version(c.version, 0, VERSION_LEN);
965 			get_kernel_version(c.version);
966 		} else {
967 			get_kernel_uname_version(c.version);
968 		}
969 
970 		/* build sb version */
971 		memcpy(c.sb_version, sbi->raw_super->version, VERSION_LEN);
972 		get_kernel_version(c.sb_version);
973 		memcpy(c.init_version, sbi->raw_super->init_version, VERSION_LEN);
974 		get_kernel_version(c.init_version);
975 
976 		MSG(0, "Info: MKFS version\n  \"%s\"\n", c.init_version);
977 		MSG(0, "Info: FSCK version\n  from \"%s\"\n    to \"%s\"\n",
978 					c.sb_version, c.version);
979 		if (!c.no_kernel_check &&
980 				memcmp(c.sb_version, c.version, VERSION_LEN)) {
981 			c.auto_fix = 0;
982 			c.fix_on = 1;
983 		}
984 		if (c.fix_on) {
985 			memcpy(sbi->raw_super->version,
986 						c.version, VERSION_LEN);
987 			update_superblock(sbi->raw_super, SB_MASK(sb_addr));
988 		}
989 		print_sb_state(sbi->raw_super);
990 		return 0;
991 	}
992 
993 	free(sbi->raw_super);
994 	sbi->raw_super = NULL;
995 	MSG(0, "\tCan't find a valid F2FS superblock at 0x%x\n", sb_addr);
996 
997 	return -EINVAL;
998 }
999 
init_sb_info(struct f2fs_sb_info * sbi)1000 int init_sb_info(struct f2fs_sb_info *sbi)
1001 {
1002 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1003 	u64 total_sectors;
1004 	int i;
1005 
1006 	sbi->log_sectors_per_block = get_sb(log_sectors_per_block);
1007 	sbi->log_blocksize = get_sb(log_blocksize);
1008 	sbi->blocksize = 1 << sbi->log_blocksize;
1009 	sbi->log_blocks_per_seg = get_sb(log_blocks_per_seg);
1010 	sbi->blocks_per_seg = 1 << sbi->log_blocks_per_seg;
1011 	sbi->segs_per_sec = get_sb(segs_per_sec);
1012 	sbi->secs_per_zone = get_sb(secs_per_zone);
1013 	sbi->total_sections = get_sb(section_count);
1014 	sbi->total_node_count = (get_sb(segment_count_nat) / 2) *
1015 				sbi->blocks_per_seg * NAT_ENTRY_PER_BLOCK;
1016 	sbi->root_ino_num = get_sb(root_ino);
1017 	sbi->node_ino_num = get_sb(node_ino);
1018 	sbi->meta_ino_num = get_sb(meta_ino);
1019 	sbi->cur_victim_sec = NULL_SEGNO;
1020 
1021 	for (i = 0; i < MAX_DEVICES; i++) {
1022 		if (!sb->devs[i].path[0])
1023 			break;
1024 
1025 		if (i) {
1026 			c.devices[i].path = strdup((char *)sb->devs[i].path);
1027 			if (get_device_info(i))
1028 				ASSERT(0);
1029 		} else {
1030 			ASSERT(!strcmp((char *)sb->devs[i].path,
1031 						(char *)c.devices[i].path));
1032 		}
1033 
1034 		c.devices[i].total_segments =
1035 			le32_to_cpu(sb->devs[i].total_segments);
1036 		if (i)
1037 			c.devices[i].start_blkaddr =
1038 				c.devices[i - 1].end_blkaddr + 1;
1039 		c.devices[i].end_blkaddr = c.devices[i].start_blkaddr +
1040 			c.devices[i].total_segments *
1041 			c.blks_per_seg - 1;
1042 		if (i == 0)
1043 			c.devices[i].end_blkaddr += get_sb(segment0_blkaddr);
1044 
1045 		c.ndevs = i + 1;
1046 		MSG(0, "Info: Device[%d] : %s blkaddr = %"PRIx64"--%"PRIx64"\n",
1047 				i, c.devices[i].path,
1048 				c.devices[i].start_blkaddr,
1049 				c.devices[i].end_blkaddr);
1050 	}
1051 
1052 	total_sectors = get_sb(block_count) << sbi->log_sectors_per_block;
1053 	MSG(0, "Info: total FS sectors = %"PRIu64" (%"PRIu64" MB)\n",
1054 				total_sectors, total_sectors >>
1055 						(20 - get_sb(log_sectorsize)));
1056 	return 0;
1057 }
1058 
verify_checksum_chksum(struct f2fs_checkpoint * cp)1059 static int verify_checksum_chksum(struct f2fs_checkpoint *cp)
1060 {
1061 	unsigned int chksum_offset = get_cp(checksum_offset);
1062 	unsigned int crc, cal_crc;
1063 
1064 	if (chksum_offset < CP_MIN_CHKSUM_OFFSET ||
1065 			chksum_offset > CP_CHKSUM_OFFSET) {
1066 		MSG(0, "\tInvalid CP CRC offset: %u\n", chksum_offset);
1067 		return -1;
1068 	}
1069 
1070 	crc = le32_to_cpu(*(__le32 *)((unsigned char *)cp + chksum_offset));
1071 	cal_crc = f2fs_checkpoint_chksum(cp);
1072 	if (cal_crc != crc) {
1073 		MSG(0, "\tInvalid CP CRC: offset:%u, crc:0x%x, calc:0x%x\n",
1074 			chksum_offset, crc, cal_crc);
1075 		return -1;
1076 	}
1077 	return 0;
1078 }
1079 
get_checkpoint_version(block_t cp_addr)1080 static void *get_checkpoint_version(block_t cp_addr)
1081 {
1082 	void *cp_page;
1083 
1084 	cp_page = malloc(PAGE_SIZE);
1085 	ASSERT(cp_page);
1086 
1087 	if (dev_read_block(cp_page, cp_addr) < 0)
1088 		ASSERT(0);
1089 
1090 	if (verify_checksum_chksum((struct f2fs_checkpoint *)cp_page))
1091 		goto out;
1092 	return cp_page;
1093 out:
1094 	free(cp_page);
1095 	return NULL;
1096 }
1097 
validate_checkpoint(struct f2fs_sb_info * sbi,block_t cp_addr,unsigned long long * version)1098 void *validate_checkpoint(struct f2fs_sb_info *sbi, block_t cp_addr,
1099 				unsigned long long *version)
1100 {
1101 	void *cp_page_1, *cp_page_2;
1102 	struct f2fs_checkpoint *cp;
1103 	unsigned long long cur_version = 0, pre_version = 0;
1104 
1105 	/* Read the 1st cp block in this CP pack */
1106 	cp_page_1 = get_checkpoint_version(cp_addr);
1107 	if (!cp_page_1)
1108 		return NULL;
1109 
1110 	cp = (struct f2fs_checkpoint *)cp_page_1;
1111 	if (get_cp(cp_pack_total_block_count) > sbi->blocks_per_seg)
1112 		goto invalid_cp1;
1113 
1114 	pre_version = get_cp(checkpoint_ver);
1115 
1116 	/* Read the 2nd cp block in this CP pack */
1117 	cp_addr += get_cp(cp_pack_total_block_count) - 1;
1118 	cp_page_2 = get_checkpoint_version(cp_addr);
1119 	if (!cp_page_2)
1120 		goto invalid_cp1;
1121 
1122 	cp = (struct f2fs_checkpoint *)cp_page_2;
1123 	cur_version = get_cp(checkpoint_ver);
1124 
1125 	if (cur_version == pre_version) {
1126 		*version = cur_version;
1127 		free(cp_page_2);
1128 		return cp_page_1;
1129 	}
1130 
1131 	free(cp_page_2);
1132 invalid_cp1:
1133 	free(cp_page_1);
1134 	return NULL;
1135 }
1136 
get_valid_checkpoint(struct f2fs_sb_info * sbi)1137 int get_valid_checkpoint(struct f2fs_sb_info *sbi)
1138 {
1139 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1140 	void *cp1, *cp2, *cur_page;
1141 	unsigned long blk_size = sbi->blocksize;
1142 	unsigned long long cp1_version = 0, cp2_version = 0, version;
1143 	unsigned long long cp_start_blk_no;
1144 	unsigned int cp_payload, cp_blks;
1145 	int ret;
1146 
1147 	cp_payload = get_sb(cp_payload);
1148 	if (cp_payload > F2FS_BLK_ALIGN(MAX_SIT_BITMAP_SIZE))
1149 		return -EINVAL;
1150 
1151 	cp_blks = 1 + cp_payload;
1152 	sbi->ckpt = malloc(cp_blks * blk_size);
1153 	if (!sbi->ckpt)
1154 		return -ENOMEM;
1155 	/*
1156 	 * Finding out valid cp block involves read both
1157 	 * sets( cp pack1 and cp pack 2)
1158 	 */
1159 	cp_start_blk_no = get_sb(cp_blkaddr);
1160 	cp1 = validate_checkpoint(sbi, cp_start_blk_no, &cp1_version);
1161 
1162 	/* The second checkpoint pack should start at the next segment */
1163 	cp_start_blk_no += 1 << get_sb(log_blocks_per_seg);
1164 	cp2 = validate_checkpoint(sbi, cp_start_blk_no, &cp2_version);
1165 
1166 	if (cp1 && cp2) {
1167 		if (ver_after(cp2_version, cp1_version)) {
1168 			cur_page = cp2;
1169 			sbi->cur_cp = 2;
1170 			version = cp2_version;
1171 		} else {
1172 			cur_page = cp1;
1173 			sbi->cur_cp = 1;
1174 			version = cp1_version;
1175 		}
1176 	} else if (cp1) {
1177 		cur_page = cp1;
1178 		sbi->cur_cp = 1;
1179 		version = cp1_version;
1180 	} else if (cp2) {
1181 		cur_page = cp2;
1182 		sbi->cur_cp = 2;
1183 		version = cp2_version;
1184 	} else
1185 		goto fail_no_cp;
1186 
1187 	MSG(0, "Info: CKPT version = %llx\n", version);
1188 
1189 	memcpy(sbi->ckpt, cur_page, blk_size);
1190 
1191 	if (cp_blks > 1) {
1192 		unsigned int i;
1193 		unsigned long long cp_blk_no;
1194 
1195 		cp_blk_no = get_sb(cp_blkaddr);
1196 		if (cur_page == cp2)
1197 			cp_blk_no += 1 << get_sb(log_blocks_per_seg);
1198 
1199 		/* copy sit bitmap */
1200 		for (i = 1; i < cp_blks; i++) {
1201 			unsigned char *ckpt = (unsigned char *)sbi->ckpt;
1202 			ret = dev_read_block(cur_page, cp_blk_no + i);
1203 			ASSERT(ret >= 0);
1204 			memcpy(ckpt + i * blk_size, cur_page, blk_size);
1205 		}
1206 	}
1207 	if (cp1)
1208 		free(cp1);
1209 	if (cp2)
1210 		free(cp2);
1211 	return 0;
1212 
1213 fail_no_cp:
1214 	free(sbi->ckpt);
1215 	sbi->ckpt = NULL;
1216 	return -EINVAL;
1217 }
1218 
1219 /*
1220  * For a return value of 1, caller should further check for c.fix_on state
1221  * and take appropriate action.
1222  */
f2fs_should_proceed(struct f2fs_super_block * sb,u32 flag)1223 static int f2fs_should_proceed(struct f2fs_super_block *sb, u32 flag)
1224 {
1225 	if (!c.fix_on && (c.auto_fix || c.preen_mode)) {
1226 		if (flag & CP_FSCK_FLAG ||
1227 			flag & CP_QUOTA_NEED_FSCK_FLAG ||
1228 			(exist_qf_ino(sb) && (flag & CP_ERROR_FLAG))) {
1229 			c.fix_on = 1;
1230 		} else if (!c.preen_mode) {
1231 			print_cp_state(flag);
1232 			return 0;
1233 		}
1234 	}
1235 	return 1;
1236 }
1237 
sanity_check_ckpt(struct f2fs_sb_info * sbi)1238 int sanity_check_ckpt(struct f2fs_sb_info *sbi)
1239 {
1240 	unsigned int total, fsmeta;
1241 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1242 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1243 	unsigned int flag = get_cp(ckpt_flags);
1244 	unsigned int ovp_segments, reserved_segments;
1245 	unsigned int main_segs, blocks_per_seg;
1246 	unsigned int sit_segs, nat_segs;
1247 	unsigned int sit_bitmap_size, nat_bitmap_size;
1248 	unsigned int log_blocks_per_seg;
1249 	unsigned int segment_count_main;
1250 	unsigned int cp_pack_start_sum, cp_payload;
1251 	block_t user_block_count;
1252 	int i;
1253 
1254 	total = get_sb(segment_count);
1255 	fsmeta = get_sb(segment_count_ckpt);
1256 	sit_segs = get_sb(segment_count_sit);
1257 	fsmeta += sit_segs;
1258 	nat_segs = get_sb(segment_count_nat);
1259 	fsmeta += nat_segs;
1260 	fsmeta += get_cp(rsvd_segment_count);
1261 	fsmeta += get_sb(segment_count_ssa);
1262 
1263 	if (fsmeta >= total)
1264 		return 1;
1265 
1266 	ovp_segments = get_cp(overprov_segment_count);
1267 	reserved_segments = get_cp(rsvd_segment_count);
1268 
1269 	if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO)) &&
1270 		(fsmeta < F2FS_MIN_SEGMENT || ovp_segments == 0 ||
1271 					reserved_segments == 0)) {
1272 		MSG(0, "\tWrong layout: check mkfs.f2fs version\n");
1273 		return 1;
1274 	}
1275 
1276 	user_block_count = get_cp(user_block_count);
1277 	segment_count_main = get_sb(segment_count_main) +
1278 				(cpu_to_le32(F2FS_FEATURE_RO) ? 1 : 0);
1279 	log_blocks_per_seg = get_sb(log_blocks_per_seg);
1280 	if (!user_block_count || user_block_count >=
1281 			segment_count_main << log_blocks_per_seg) {
1282 		ASSERT_MSG("\tWrong user_block_count(%u)\n", user_block_count);
1283 
1284 		if (!f2fs_should_proceed(sb, flag))
1285 			return 1;
1286 		if (!c.fix_on)
1287 			return 1;
1288 
1289 		if (flag & (CP_FSCK_FLAG | CP_RESIZEFS_FLAG)) {
1290 			u32 valid_user_block_cnt;
1291 			u32 seg_cnt_main = get_sb(segment_count) -
1292 					(get_sb(segment_count_ckpt) +
1293 					 get_sb(segment_count_sit) +
1294 					 get_sb(segment_count_nat) +
1295 					 get_sb(segment_count_ssa));
1296 
1297 			/* validate segment_count_main in sb first */
1298 			if (seg_cnt_main != get_sb(segment_count_main)) {
1299 				MSG(0, "Inconsistent segment_cnt_main %u in sb\n",
1300 						segment_count_main << log_blocks_per_seg);
1301 				return 1;
1302 			}
1303 			valid_user_block_cnt = ((get_sb(segment_count_main) -
1304 						get_cp(overprov_segment_count)) * c.blks_per_seg);
1305 			MSG(0, "Info: Fix wrong user_block_count in CP: (%u) -> (%u)\n",
1306 					user_block_count, valid_user_block_cnt);
1307 			set_cp(user_block_count, valid_user_block_cnt);
1308 			c.bug_on = 1;
1309 		}
1310 	}
1311 
1312 	main_segs = get_sb(segment_count_main);
1313 	blocks_per_seg = sbi->blocks_per_seg;
1314 
1315 	for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) {
1316 		if (get_cp(cur_node_segno[i]) >= main_segs ||
1317 			get_cp(cur_node_blkoff[i]) >= blocks_per_seg)
1318 			return 1;
1319 	}
1320 	for (i = 0; i < NR_CURSEG_DATA_TYPE; i++) {
1321 		if (get_cp(cur_data_segno[i]) >= main_segs ||
1322 			get_cp(cur_data_blkoff[i]) >= blocks_per_seg)
1323 			return 1;
1324 	}
1325 
1326 	sit_bitmap_size = get_cp(sit_ver_bitmap_bytesize);
1327 	nat_bitmap_size = get_cp(nat_ver_bitmap_bytesize);
1328 
1329 	if (sit_bitmap_size != ((sit_segs / 2) << log_blocks_per_seg) / 8 ||
1330 		nat_bitmap_size != ((nat_segs / 2) << log_blocks_per_seg) / 8) {
1331 		MSG(0, "\tWrong bitmap size: sit(%u), nat(%u)\n",
1332 				sit_bitmap_size, nat_bitmap_size);
1333 		return 1;
1334 	}
1335 
1336 	cp_pack_start_sum = __start_sum_addr(sbi);
1337 	cp_payload = __cp_payload(sbi);
1338 	if (cp_pack_start_sum < cp_payload + 1 ||
1339 		cp_pack_start_sum > blocks_per_seg - 1 -
1340 			NR_CURSEG_TYPE) {
1341 		MSG(0, "\tWrong cp_pack_start_sum(%u) or cp_payload(%u)\n",
1342 			cp_pack_start_sum, cp_payload);
1343 		if ((get_sb(feature) & F2FS_FEATURE_SB_CHKSUM))
1344 			return 1;
1345 		set_sb(cp_payload, cp_pack_start_sum - 1);
1346 		update_superblock(sb, SB_MASK_ALL);
1347 	}
1348 
1349 	return 0;
1350 }
1351 
current_nat_addr(struct f2fs_sb_info * sbi,nid_t start,int * pack)1352 pgoff_t current_nat_addr(struct f2fs_sb_info *sbi, nid_t start, int *pack)
1353 {
1354 	struct f2fs_nm_info *nm_i = NM_I(sbi);
1355 	pgoff_t block_off;
1356 	pgoff_t block_addr;
1357 	int seg_off;
1358 
1359 	block_off = NAT_BLOCK_OFFSET(start);
1360 	seg_off = block_off >> sbi->log_blocks_per_seg;
1361 
1362 	block_addr = (pgoff_t)(nm_i->nat_blkaddr +
1363 			(seg_off << sbi->log_blocks_per_seg << 1) +
1364 			(block_off & ((1 << sbi->log_blocks_per_seg) -1)));
1365 	if (pack)
1366 		*pack = 1;
1367 
1368 	if (f2fs_test_bit(block_off, nm_i->nat_bitmap)) {
1369 		block_addr += sbi->blocks_per_seg;
1370 		if (pack)
1371 			*pack = 2;
1372 	}
1373 
1374 	return block_addr;
1375 }
1376 
1377 /* will not init nid_bitmap from nat */
f2fs_early_init_nid_bitmap(struct f2fs_sb_info * sbi)1378 static int f2fs_early_init_nid_bitmap(struct f2fs_sb_info *sbi)
1379 {
1380 	struct f2fs_nm_info *nm_i = NM_I(sbi);
1381 	int nid_bitmap_size = (nm_i->max_nid + BITS_PER_BYTE - 1) / BITS_PER_BYTE;
1382 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1383 	struct f2fs_summary_block *sum = curseg->sum_blk;
1384 	struct f2fs_journal *journal = &sum->journal;
1385 	nid_t nid;
1386 	int i;
1387 
1388 	if (!(c.func == SLOAD || c.func == FSCK))
1389 		return 0;
1390 
1391 	nm_i->nid_bitmap = (char *)calloc(nid_bitmap_size, 1);
1392 	if (!nm_i->nid_bitmap)
1393 		return -ENOMEM;
1394 
1395 	/* arbitrarily set 0 bit */
1396 	f2fs_set_bit(0, nm_i->nid_bitmap);
1397 
1398 	if (nats_in_cursum(journal) > NAT_JOURNAL_ENTRIES) {
1399 		MSG(0, "\tError: f2fs_init_nid_bitmap truncate n_nats(%u) to "
1400 			"NAT_JOURNAL_ENTRIES(%lu)\n",
1401 			nats_in_cursum(journal), NAT_JOURNAL_ENTRIES);
1402 		journal->n_nats = cpu_to_le16(NAT_JOURNAL_ENTRIES);
1403 		c.fix_on = 1;
1404 	}
1405 
1406 	for (i = 0; i < nats_in_cursum(journal); i++) {
1407 		block_t addr;
1408 
1409 		addr = le32_to_cpu(nat_in_journal(journal, i).block_addr);
1410 		if (!IS_VALID_BLK_ADDR(sbi, addr)) {
1411 			MSG(0, "\tError: f2fs_init_nid_bitmap: addr(%u) is invalid!!!\n", addr);
1412 			journal->n_nats = cpu_to_le16(i);
1413 			c.fix_on = 1;
1414 			continue;
1415 		}
1416 
1417 		nid = le32_to_cpu(nid_in_journal(journal, i));
1418 		if (!IS_VALID_NID(sbi, nid)) {
1419 			MSG(0, "\tError: f2fs_init_nid_bitmap: nid(%u) is invalid!!!\n", nid);
1420 			journal->n_nats = cpu_to_le16(i);
1421 			c.fix_on = 1;
1422 			continue;
1423 		}
1424 		if (addr != NULL_ADDR)
1425 			f2fs_set_bit(nid, nm_i->nid_bitmap);
1426 	}
1427 	return 0;
1428 }
1429 
1430 /* will init nid_bitmap from nat */
f2fs_late_init_nid_bitmap(struct f2fs_sb_info * sbi)1431 static int f2fs_late_init_nid_bitmap(struct f2fs_sb_info *sbi)
1432 {
1433 	struct f2fs_nm_info *nm_i = NM_I(sbi);
1434 	struct f2fs_nat_block *nat_block;
1435 	block_t start_blk;
1436 	nid_t nid;
1437 
1438 	if (!(c.func == SLOAD || c.func == FSCK))
1439 		return 0;
1440 
1441 	nat_block = malloc(F2FS_BLKSIZE);
1442 	if (!nat_block) {
1443 		free(nm_i->nid_bitmap);
1444 		return -ENOMEM;
1445 	}
1446 
1447 	f2fs_ra_meta_pages(sbi, 0, NAT_BLOCK_OFFSET(nm_i->max_nid),
1448 							META_NAT);
1449 	for (nid = 0; nid < nm_i->max_nid; nid++) {
1450 		if (!(nid % NAT_ENTRY_PER_BLOCK)) {
1451 			int ret;
1452 
1453 			start_blk = current_nat_addr(sbi, nid, NULL);
1454 			ret = dev_read_block(nat_block, start_blk);
1455 			ASSERT(ret >= 0);
1456 		}
1457 
1458 		if (nat_block->entries[nid % NAT_ENTRY_PER_BLOCK].block_addr)
1459 			f2fs_set_bit(nid, nm_i->nid_bitmap);
1460 	}
1461 
1462 	free(nat_block);
1463 	return 0;
1464 }
1465 
update_nat_bits_flags(struct f2fs_super_block * sb,struct f2fs_checkpoint * cp,u32 flags)1466 u32 update_nat_bits_flags(struct f2fs_super_block *sb,
1467 				struct f2fs_checkpoint *cp, u32 flags)
1468 {
1469 	u_int32_t nat_bits_bytes, nat_bits_blocks;
1470 
1471 	nat_bits_bytes = get_sb(segment_count_nat) << 5;
1472 	nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) + 8 +
1473 						F2FS_BLKSIZE - 1);
1474 	if (get_cp(cp_pack_total_block_count) <=
1475 			(1 << get_sb(log_blocks_per_seg)) - nat_bits_blocks)
1476 		flags |= CP_NAT_BITS_FLAG;
1477 	else
1478 		flags &= (~CP_NAT_BITS_FLAG);
1479 
1480 	return flags;
1481 }
1482 
1483 /* should call flush_journal_entries() bfore this */
write_nat_bits(struct f2fs_sb_info * sbi,struct f2fs_super_block * sb,struct f2fs_checkpoint * cp,int set)1484 void write_nat_bits(struct f2fs_sb_info *sbi,
1485 	struct f2fs_super_block *sb, struct f2fs_checkpoint *cp, int set)
1486 {
1487 	struct f2fs_nm_info *nm_i = NM_I(sbi);
1488 	u_int32_t nat_blocks = get_sb(segment_count_nat) <<
1489 				(get_sb(log_blocks_per_seg) - 1);
1490 	u_int32_t nat_bits_bytes = nat_blocks >> 3;
1491 	u_int32_t nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) +
1492 					8 + F2FS_BLKSIZE - 1);
1493 	unsigned char *nat_bits, *full_nat_bits, *empty_nat_bits;
1494 	struct f2fs_nat_block *nat_block;
1495 	u_int32_t i, j;
1496 	block_t blkaddr;
1497 	int ret;
1498 
1499 	nat_bits = calloc(F2FS_BLKSIZE, nat_bits_blocks);
1500 	ASSERT(nat_bits);
1501 
1502 	nat_block = malloc(F2FS_BLKSIZE);
1503 	ASSERT(nat_block);
1504 
1505 	full_nat_bits = nat_bits + 8;
1506 	empty_nat_bits = full_nat_bits + nat_bits_bytes;
1507 
1508 	memset(full_nat_bits, 0, nat_bits_bytes);
1509 	memset(empty_nat_bits, 0, nat_bits_bytes);
1510 
1511 	for (i = 0; i < nat_blocks; i++) {
1512 		int seg_off = i >> get_sb(log_blocks_per_seg);
1513 		int valid = 0;
1514 
1515 		blkaddr = (pgoff_t)(get_sb(nat_blkaddr) +
1516 				(seg_off << get_sb(log_blocks_per_seg) << 1) +
1517 				(i & ((1 << get_sb(log_blocks_per_seg)) - 1)));
1518 
1519 		/*
1520 		 * Should consider new nat_blocks is larger than old
1521 		 * nm_i->nat_blocks, since nm_i->nat_bitmap is based on
1522 		 * old one.
1523 		 */
1524 		if (i < nm_i->nat_blocks && f2fs_test_bit(i, nm_i->nat_bitmap))
1525 			blkaddr += (1 << get_sb(log_blocks_per_seg));
1526 
1527 		ret = dev_read_block(nat_block, blkaddr);
1528 		ASSERT(ret >= 0);
1529 
1530 		for (j = 0; j < NAT_ENTRY_PER_BLOCK; j++) {
1531 			if ((i == 0 && j == 0) ||
1532 				nat_block->entries[j].block_addr != NULL_ADDR)
1533 				valid++;
1534 		}
1535 		if (valid == 0)
1536 			test_and_set_bit_le(i, empty_nat_bits);
1537 		else if (valid == NAT_ENTRY_PER_BLOCK)
1538 			test_and_set_bit_le(i, full_nat_bits);
1539 	}
1540 	*(__le64 *)nat_bits = get_cp_crc(cp);
1541 	free(nat_block);
1542 
1543 	blkaddr = get_sb(segment0_blkaddr) + (set <<
1544 				get_sb(log_blocks_per_seg)) - nat_bits_blocks;
1545 
1546 	DBG(1, "\tWriting NAT bits pages, at offset 0x%08x\n", blkaddr);
1547 
1548 	for (i = 0; i < nat_bits_blocks; i++) {
1549 		if (dev_write_block(nat_bits + i * F2FS_BLKSIZE, blkaddr + i))
1550 			ASSERT_MSG("\tError: write NAT bits to disk!!!\n");
1551 	}
1552 	MSG(0, "Info: Write valid nat_bits in checkpoint\n");
1553 
1554 	free(nat_bits);
1555 }
1556 
check_nat_bits(struct f2fs_sb_info * sbi,struct f2fs_super_block * sb,struct f2fs_checkpoint * cp)1557 static int check_nat_bits(struct f2fs_sb_info *sbi,
1558 	struct f2fs_super_block *sb, struct f2fs_checkpoint *cp)
1559 {
1560 	struct f2fs_nm_info *nm_i = NM_I(sbi);
1561 	u_int32_t nat_blocks = get_sb(segment_count_nat) <<
1562 				(get_sb(log_blocks_per_seg) - 1);
1563 	u_int32_t nat_bits_bytes = nat_blocks >> 3;
1564 	u_int32_t nat_bits_blocks = F2FS_BYTES_TO_BLK((nat_bits_bytes << 1) +
1565 					8 + F2FS_BLKSIZE - 1);
1566 	unsigned char *nat_bits, *full_nat_bits, *empty_nat_bits;
1567 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1568 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
1569 	u_int32_t i, j;
1570 	block_t blkaddr;
1571 	int err = 0;
1572 
1573 	nat_bits = calloc(F2FS_BLKSIZE, nat_bits_blocks);
1574 	ASSERT(nat_bits);
1575 
1576 	full_nat_bits = nat_bits + 8;
1577 	empty_nat_bits = full_nat_bits + nat_bits_bytes;
1578 
1579 	blkaddr = get_sb(segment0_blkaddr) + (sbi->cur_cp <<
1580 				get_sb(log_blocks_per_seg)) - nat_bits_blocks;
1581 
1582 	for (i = 0; i < nat_bits_blocks; i++) {
1583 		if (dev_read_block(nat_bits + i * F2FS_BLKSIZE, blkaddr + i))
1584 			ASSERT_MSG("\tError: read NAT bits to disk!!!\n");
1585 	}
1586 
1587 	if (*(__le64 *)nat_bits != get_cp_crc(cp) || nats_in_cursum(journal)) {
1588 		/*
1589 		 * if there is a journal, f2fs was not shutdown cleanly. Let's
1590 		 * flush them with nat_bits.
1591 		 */
1592 		if (c.fix_on)
1593 			err = -1;
1594 		/* Otherwise, kernel will disable nat_bits */
1595 		goto out;
1596 	}
1597 
1598 	for (i = 0; i < nat_blocks; i++) {
1599 		u_int32_t start_nid = i * NAT_ENTRY_PER_BLOCK;
1600 		u_int32_t valid = 0;
1601 		int empty = test_bit_le(i, empty_nat_bits);
1602 		int full = test_bit_le(i, full_nat_bits);
1603 
1604 		for (j = 0; j < NAT_ENTRY_PER_BLOCK; j++) {
1605 			if (f2fs_test_bit(start_nid + j, nm_i->nid_bitmap))
1606 				valid++;
1607 		}
1608 		if (valid == 0) {
1609 			if (!empty || full) {
1610 				err = -1;
1611 				goto out;
1612 			}
1613 		} else if (valid == NAT_ENTRY_PER_BLOCK) {
1614 			if (empty || !full) {
1615 				err = -1;
1616 				goto out;
1617 			}
1618 		} else {
1619 			if (empty || full) {
1620 				err = -1;
1621 				goto out;
1622 			}
1623 		}
1624 	}
1625 out:
1626 	free(nat_bits);
1627 	if (!err) {
1628 		MSG(0, "Info: Checked valid nat_bits in checkpoint\n");
1629 	} else {
1630 		c.bug_nat_bits = 1;
1631 		MSG(0, "Info: Corrupted valid nat_bits in checkpoint\n");
1632 	}
1633 	return err;
1634 }
1635 
init_node_manager(struct f2fs_sb_info * sbi)1636 int init_node_manager(struct f2fs_sb_info *sbi)
1637 {
1638 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1639 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1640 	struct f2fs_nm_info *nm_i = NM_I(sbi);
1641 	unsigned char *version_bitmap;
1642 	unsigned int nat_segs;
1643 
1644 	nm_i->nat_blkaddr = get_sb(nat_blkaddr);
1645 
1646 	/* segment_count_nat includes pair segment so divide to 2. */
1647 	nat_segs = get_sb(segment_count_nat) >> 1;
1648 	nm_i->nat_blocks = nat_segs << get_sb(log_blocks_per_seg);
1649 	nm_i->max_nid = NAT_ENTRY_PER_BLOCK * nm_i->nat_blocks;
1650 	nm_i->fcnt = 0;
1651 	nm_i->nat_cnt = 0;
1652 	nm_i->init_scan_nid = get_cp(next_free_nid);
1653 	nm_i->next_scan_nid = get_cp(next_free_nid);
1654 
1655 	nm_i->bitmap_size = __bitmap_size(sbi, NAT_BITMAP);
1656 
1657 	nm_i->nat_bitmap = malloc(nm_i->bitmap_size);
1658 	if (!nm_i->nat_bitmap)
1659 		return -ENOMEM;
1660 	version_bitmap = __bitmap_ptr(sbi, NAT_BITMAP);
1661 	if (!version_bitmap)
1662 		return -EFAULT;
1663 
1664 	/* copy version bitmap */
1665 	memcpy(nm_i->nat_bitmap, version_bitmap, nm_i->bitmap_size);
1666 	return f2fs_early_init_nid_bitmap(sbi);
1667 }
1668 
build_node_manager(struct f2fs_sb_info * sbi)1669 int build_node_manager(struct f2fs_sb_info *sbi)
1670 {
1671 	int err;
1672 	sbi->nm_info = malloc(sizeof(struct f2fs_nm_info));
1673 	if (!sbi->nm_info)
1674 		return -ENOMEM;
1675 
1676 	err = init_node_manager(sbi);
1677 	if (err)
1678 		return err;
1679 
1680 	return 0;
1681 }
1682 
build_sit_info(struct f2fs_sb_info * sbi)1683 int build_sit_info(struct f2fs_sb_info *sbi)
1684 {
1685 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1686 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1687 	struct sit_info *sit_i;
1688 	unsigned int sit_segs;
1689 	int start;
1690 	char *src_bitmap, *dst_bitmap;
1691 	unsigned char *bitmap;
1692 	unsigned int bitmap_size;
1693 
1694 	sit_i = malloc(sizeof(struct sit_info));
1695 	if (!sit_i) {
1696 		MSG(1, "\tError: Malloc failed for build_sit_info!\n");
1697 		return -ENOMEM;
1698 	}
1699 
1700 	SM_I(sbi)->sit_info = sit_i;
1701 
1702 	sit_i->sentries = calloc(TOTAL_SEGS(sbi) * sizeof(struct seg_entry), 1);
1703 	if (!sit_i->sentries) {
1704 		MSG(1, "\tError: Calloc failed for build_sit_info!\n");
1705 		goto free_sit_info;
1706 	}
1707 
1708 	bitmap_size = TOTAL_SEGS(sbi) * SIT_VBLOCK_MAP_SIZE;
1709 
1710 	if (need_fsync_data_record(sbi))
1711 		bitmap_size += bitmap_size;
1712 
1713 	sit_i->bitmap = calloc(bitmap_size, 1);
1714 	if (!sit_i->bitmap) {
1715 		MSG(1, "\tError: Calloc failed for build_sit_info!!\n");
1716 		goto free_sentries;
1717 	}
1718 
1719 	bitmap = sit_i->bitmap;
1720 
1721 	for (start = 0; start < TOTAL_SEGS(sbi); start++) {
1722 		sit_i->sentries[start].cur_valid_map = bitmap;
1723 		bitmap += SIT_VBLOCK_MAP_SIZE;
1724 
1725 		if (need_fsync_data_record(sbi)) {
1726 			sit_i->sentries[start].ckpt_valid_map = bitmap;
1727 			bitmap += SIT_VBLOCK_MAP_SIZE;
1728 		}
1729 	}
1730 
1731 	sit_segs = get_sb(segment_count_sit) >> 1;
1732 	bitmap_size = __bitmap_size(sbi, SIT_BITMAP);
1733 	src_bitmap = __bitmap_ptr(sbi, SIT_BITMAP);
1734 
1735 	dst_bitmap = malloc(bitmap_size);
1736 	if (!dst_bitmap) {
1737 		MSG(1, "\tError: Malloc failed for build_sit_info!!\n");
1738 		goto free_validity_maps;
1739 	}
1740 
1741 	memcpy(dst_bitmap, src_bitmap, bitmap_size);
1742 
1743 	sit_i->sit_base_addr = get_sb(sit_blkaddr);
1744 	sit_i->sit_blocks = sit_segs << sbi->log_blocks_per_seg;
1745 	sit_i->written_valid_blocks = get_cp(valid_block_count);
1746 	sit_i->sit_bitmap = dst_bitmap;
1747 	sit_i->bitmap_size = bitmap_size;
1748 	sit_i->dirty_sentries = 0;
1749 	sit_i->sents_per_block = SIT_ENTRY_PER_BLOCK;
1750 	sit_i->elapsed_time = get_cp(elapsed_time);
1751 	return 0;
1752 
1753 free_validity_maps:
1754 	free(sit_i->bitmap);
1755 free_sentries:
1756 	free(sit_i->sentries);
1757 free_sit_info:
1758 	free(sit_i);
1759 
1760 	return -ENOMEM;
1761 }
1762 
reset_curseg(struct f2fs_sb_info * sbi,int type)1763 void reset_curseg(struct f2fs_sb_info *sbi, int type)
1764 {
1765 	struct curseg_info *curseg = CURSEG_I(sbi, type);
1766 	struct summary_footer *sum_footer;
1767 	struct seg_entry *se;
1768 
1769 	sum_footer = &(curseg->sum_blk->footer);
1770 	memset(sum_footer, 0, sizeof(struct summary_footer));
1771 	if (IS_DATASEG(type))
1772 		SET_SUM_TYPE(sum_footer, SUM_TYPE_DATA);
1773 	if (IS_NODESEG(type))
1774 		SET_SUM_TYPE(sum_footer, SUM_TYPE_NODE);
1775 	se = get_seg_entry(sbi, curseg->segno);
1776 	se->type = type;
1777 	se->dirty = 1;
1778 }
1779 
read_compacted_summaries(struct f2fs_sb_info * sbi)1780 static void read_compacted_summaries(struct f2fs_sb_info *sbi)
1781 {
1782 	struct curseg_info *curseg;
1783 	unsigned int i, j, offset;
1784 	block_t start;
1785 	char *kaddr;
1786 	int ret;
1787 
1788 	start = start_sum_block(sbi);
1789 
1790 	kaddr = (char *)malloc(PAGE_SIZE);
1791 	ASSERT(kaddr);
1792 
1793 	ret = dev_read_block(kaddr, start++);
1794 	ASSERT(ret >= 0);
1795 
1796 	curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
1797 	memcpy(&curseg->sum_blk->journal.n_nats, kaddr, SUM_JOURNAL_SIZE);
1798 
1799 	curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
1800 	memcpy(&curseg->sum_blk->journal.n_sits, kaddr + SUM_JOURNAL_SIZE,
1801 						SUM_JOURNAL_SIZE);
1802 
1803 	offset = 2 * SUM_JOURNAL_SIZE;
1804 	for (i = CURSEG_HOT_DATA; i <= CURSEG_COLD_DATA; i++) {
1805 		unsigned short blk_off;
1806 		struct curseg_info *curseg = CURSEG_I(sbi, i);
1807 
1808 		reset_curseg(sbi, i);
1809 
1810 		if (curseg->alloc_type == SSR)
1811 			blk_off = sbi->blocks_per_seg;
1812 		else
1813 			blk_off = curseg->next_blkoff;
1814 
1815 		ASSERT(blk_off <= ENTRIES_IN_SUM);
1816 
1817 		for (j = 0; j < blk_off; j++) {
1818 			struct f2fs_summary *s;
1819 			s = (struct f2fs_summary *)(kaddr + offset);
1820 			curseg->sum_blk->entries[j] = *s;
1821 			offset += SUMMARY_SIZE;
1822 			if (offset + SUMMARY_SIZE <=
1823 					PAGE_CACHE_SIZE - SUM_FOOTER_SIZE)
1824 				continue;
1825 			memset(kaddr, 0, PAGE_SIZE);
1826 			ret = dev_read_block(kaddr, start++);
1827 			ASSERT(ret >= 0);
1828 			offset = 0;
1829 		}
1830 	}
1831 	free(kaddr);
1832 }
1833 
restore_node_summary(struct f2fs_sb_info * sbi,unsigned int segno,struct f2fs_summary_block * sum_blk)1834 static void restore_node_summary(struct f2fs_sb_info *sbi,
1835 		unsigned int segno, struct f2fs_summary_block *sum_blk)
1836 {
1837 	struct f2fs_node *node_blk;
1838 	struct f2fs_summary *sum_entry;
1839 	block_t addr;
1840 	unsigned int i;
1841 	int ret;
1842 
1843 	node_blk = malloc(F2FS_BLKSIZE);
1844 	ASSERT(node_blk);
1845 
1846 	/* scan the node segment */
1847 	addr = START_BLOCK(sbi, segno);
1848 	sum_entry = &sum_blk->entries[0];
1849 
1850 	for (i = 0; i < sbi->blocks_per_seg; i++, sum_entry++) {
1851 		ret = dev_read_block(node_blk, addr);
1852 		ASSERT(ret >= 0);
1853 		sum_entry->nid = node_blk->footer.nid;
1854 		addr++;
1855 	}
1856 	free(node_blk);
1857 }
1858 
read_normal_summaries(struct f2fs_sb_info * sbi,int type)1859 static void read_normal_summaries(struct f2fs_sb_info *sbi, int type)
1860 {
1861 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1862 	struct f2fs_summary_block *sum_blk;
1863 	struct curseg_info *curseg;
1864 	unsigned int segno = 0;
1865 	block_t blk_addr = 0;
1866 	int ret;
1867 
1868 	if (IS_DATASEG(type)) {
1869 		segno = get_cp(cur_data_segno[type]);
1870 		if (is_set_ckpt_flags(cp, CP_UMOUNT_FLAG))
1871 			blk_addr = sum_blk_addr(sbi, NR_CURSEG_TYPE, type);
1872 		else
1873 			blk_addr = sum_blk_addr(sbi, NR_CURSEG_DATA_TYPE, type);
1874 	} else {
1875 		segno = get_cp(cur_node_segno[type - CURSEG_HOT_NODE]);
1876 		if (is_set_ckpt_flags(cp, CP_UMOUNT_FLAG))
1877 			blk_addr = sum_blk_addr(sbi, NR_CURSEG_NODE_TYPE,
1878 							type - CURSEG_HOT_NODE);
1879 		else
1880 			blk_addr = GET_SUM_BLKADDR(sbi, segno);
1881 	}
1882 
1883 	sum_blk = (struct f2fs_summary_block *)malloc(PAGE_SIZE);
1884 	ASSERT(sum_blk);
1885 
1886 	ret = dev_read_block(sum_blk, blk_addr);
1887 	ASSERT(ret >= 0);
1888 
1889 	if (IS_NODESEG(type) && !is_set_ckpt_flags(cp, CP_UMOUNT_FLAG))
1890 		restore_node_summary(sbi, segno, sum_blk);
1891 
1892 	curseg = CURSEG_I(sbi, type);
1893 	memcpy(curseg->sum_blk, sum_blk, PAGE_CACHE_SIZE);
1894 	reset_curseg(sbi, type);
1895 	free(sum_blk);
1896 }
1897 
update_sum_entry(struct f2fs_sb_info * sbi,block_t blk_addr,struct f2fs_summary * sum)1898 void update_sum_entry(struct f2fs_sb_info *sbi, block_t blk_addr,
1899 					struct f2fs_summary *sum)
1900 {
1901 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
1902 	struct f2fs_summary_block *sum_blk;
1903 	u32 segno, offset;
1904 	int type, ret;
1905 	struct seg_entry *se;
1906 
1907 	if (get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO))
1908 		return;
1909 
1910 	segno = GET_SEGNO(sbi, blk_addr);
1911 	offset = OFFSET_IN_SEG(sbi, blk_addr);
1912 
1913 	se = get_seg_entry(sbi, segno);
1914 
1915 	sum_blk = get_sum_block(sbi, segno, &type);
1916 	memcpy(&sum_blk->entries[offset], sum, sizeof(*sum));
1917 	sum_blk->footer.entry_type = IS_NODESEG(se->type) ? SUM_TYPE_NODE :
1918 							SUM_TYPE_DATA;
1919 
1920 	/* write SSA all the time */
1921 	ret = dev_write_block(sum_blk, GET_SUM_BLKADDR(sbi, segno));
1922 	ASSERT(ret >= 0);
1923 
1924 	if (type == SEG_TYPE_NODE || type == SEG_TYPE_DATA ||
1925 					type == SEG_TYPE_MAX)
1926 		free(sum_blk);
1927 }
1928 
restore_curseg_summaries(struct f2fs_sb_info * sbi)1929 static void restore_curseg_summaries(struct f2fs_sb_info *sbi)
1930 {
1931 	int type = CURSEG_HOT_DATA;
1932 
1933 	if (is_set_ckpt_flags(F2FS_CKPT(sbi), CP_COMPACT_SUM_FLAG)) {
1934 		read_compacted_summaries(sbi);
1935 		type = CURSEG_HOT_NODE;
1936 	}
1937 
1938 	for (; type <= CURSEG_COLD_NODE; type++)
1939 		read_normal_summaries(sbi, type);
1940 }
1941 
build_curseg(struct f2fs_sb_info * sbi)1942 static int build_curseg(struct f2fs_sb_info *sbi)
1943 {
1944 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
1945 	struct curseg_info *array;
1946 	unsigned short blk_off;
1947 	unsigned int segno;
1948 	int i;
1949 
1950 	array = malloc(sizeof(*array) * NR_CURSEG_TYPE);
1951 	if (!array) {
1952 		MSG(1, "\tError: Malloc failed for build_curseg!\n");
1953 		return -ENOMEM;
1954 	}
1955 
1956 	SM_I(sbi)->curseg_array = array;
1957 
1958 	for (i = 0; i < NR_CURSEG_TYPE; i++) {
1959 		array[i].sum_blk = calloc(PAGE_CACHE_SIZE, 1);
1960 		if (!array[i].sum_blk) {
1961 			MSG(1, "\tError: Calloc failed for build_curseg!!\n");
1962 			goto seg_cleanup;
1963 		}
1964 
1965 		if (i <= CURSEG_COLD_DATA) {
1966 			blk_off = get_cp(cur_data_blkoff[i]);
1967 			segno = get_cp(cur_data_segno[i]);
1968 		}
1969 		if (i > CURSEG_COLD_DATA) {
1970 			blk_off = get_cp(cur_node_blkoff[i - CURSEG_HOT_NODE]);
1971 			segno = get_cp(cur_node_segno[i - CURSEG_HOT_NODE]);
1972 		}
1973 		ASSERT(segno < TOTAL_SEGS(sbi));
1974 		ASSERT(blk_off < DEFAULT_BLOCKS_PER_SEGMENT);
1975 
1976 		array[i].segno = segno;
1977 		array[i].zone = GET_ZONENO_FROM_SEGNO(sbi, segno);
1978 		array[i].next_segno = NULL_SEGNO;
1979 		array[i].next_blkoff = blk_off;
1980 		array[i].alloc_type = cp->alloc_type[i];
1981 	}
1982 	restore_curseg_summaries(sbi);
1983 	return 0;
1984 
1985 seg_cleanup:
1986 	for(--i ; i >=0; --i)
1987 		free(array[i].sum_blk);
1988 	free(array);
1989 
1990 	return -ENOMEM;
1991 }
1992 
check_seg_range(struct f2fs_sb_info * sbi,unsigned int segno)1993 static inline void check_seg_range(struct f2fs_sb_info *sbi, unsigned int segno)
1994 {
1995 	unsigned int end_segno = SM_I(sbi)->segment_count - 1;
1996 	ASSERT(segno <= end_segno);
1997 }
1998 
current_sit_addr(struct f2fs_sb_info * sbi,unsigned int segno)1999 static inline block_t current_sit_addr(struct f2fs_sb_info *sbi,
2000 						unsigned int segno)
2001 {
2002 	struct sit_info *sit_i = SIT_I(sbi);
2003 	unsigned int offset = SIT_BLOCK_OFFSET(sit_i, segno);
2004 	block_t blk_addr = sit_i->sit_base_addr + offset;
2005 
2006 	check_seg_range(sbi, segno);
2007 
2008 	/* calculate sit block address */
2009 	if (f2fs_test_bit(offset, sit_i->sit_bitmap))
2010 		blk_addr += sit_i->sit_blocks;
2011 
2012 	return blk_addr;
2013 }
2014 
get_current_sit_page(struct f2fs_sb_info * sbi,unsigned int segno,struct f2fs_sit_block * sit_blk)2015 void get_current_sit_page(struct f2fs_sb_info *sbi,
2016 			unsigned int segno, struct f2fs_sit_block *sit_blk)
2017 {
2018 	block_t blk_addr = current_sit_addr(sbi, segno);
2019 
2020 	ASSERT(dev_read_block(sit_blk, blk_addr) >= 0);
2021 }
2022 
rewrite_current_sit_page(struct f2fs_sb_info * sbi,unsigned int segno,struct f2fs_sit_block * sit_blk)2023 void rewrite_current_sit_page(struct f2fs_sb_info *sbi,
2024 			unsigned int segno, struct f2fs_sit_block *sit_blk)
2025 {
2026 	block_t blk_addr = current_sit_addr(sbi, segno);
2027 
2028 	ASSERT(dev_write_block(sit_blk, blk_addr) >= 0);
2029 }
2030 
check_block_count(struct f2fs_sb_info * sbi,unsigned int segno,struct f2fs_sit_entry * raw_sit)2031 void check_block_count(struct f2fs_sb_info *sbi,
2032 		unsigned int segno, struct f2fs_sit_entry *raw_sit)
2033 {
2034 	struct f2fs_sm_info *sm_info = SM_I(sbi);
2035 	unsigned int end_segno = sm_info->segment_count - 1;
2036 	int valid_blocks = 0;
2037 	unsigned int i;
2038 
2039 	/* check segment usage */
2040 	if (GET_SIT_VBLOCKS(raw_sit) > sbi->blocks_per_seg)
2041 		ASSERT_MSG("Invalid SIT vblocks: segno=0x%x, %u",
2042 				segno, GET_SIT_VBLOCKS(raw_sit));
2043 
2044 	/* check boundary of a given segment number */
2045 	if (segno > end_segno)
2046 		ASSERT_MSG("Invalid SEGNO: 0x%x", segno);
2047 
2048 	/* check bitmap with valid block count */
2049 	for (i = 0; i < SIT_VBLOCK_MAP_SIZE; i++)
2050 		valid_blocks += get_bits_in_byte(raw_sit->valid_map[i]);
2051 
2052 	if (GET_SIT_VBLOCKS(raw_sit) != valid_blocks)
2053 		ASSERT_MSG("Wrong SIT valid blocks: segno=0x%x, %u vs. %u",
2054 				segno, GET_SIT_VBLOCKS(raw_sit), valid_blocks);
2055 
2056 	if (GET_SIT_TYPE(raw_sit) >= NO_CHECK_TYPE)
2057 		ASSERT_MSG("Wrong SIT type: segno=0x%x, %u",
2058 				segno, GET_SIT_TYPE(raw_sit));
2059 }
2060 
__seg_info_from_raw_sit(struct seg_entry * se,struct f2fs_sit_entry * raw_sit)2061 void __seg_info_from_raw_sit(struct seg_entry *se,
2062 		struct f2fs_sit_entry *raw_sit)
2063 {
2064 	se->valid_blocks = GET_SIT_VBLOCKS(raw_sit);
2065 	memcpy(se->cur_valid_map, raw_sit->valid_map, SIT_VBLOCK_MAP_SIZE);
2066 	se->type = GET_SIT_TYPE(raw_sit);
2067 	se->orig_type = GET_SIT_TYPE(raw_sit);
2068 	se->mtime = le64_to_cpu(raw_sit->mtime);
2069 }
2070 
seg_info_from_raw_sit(struct f2fs_sb_info * sbi,struct seg_entry * se,struct f2fs_sit_entry * raw_sit)2071 void seg_info_from_raw_sit(struct f2fs_sb_info *sbi, struct seg_entry *se,
2072 						struct f2fs_sit_entry *raw_sit)
2073 {
2074 	__seg_info_from_raw_sit(se, raw_sit);
2075 
2076 	if (!need_fsync_data_record(sbi))
2077 		return;
2078 	se->ckpt_valid_blocks = se->valid_blocks;
2079 	memcpy(se->ckpt_valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2080 	se->ckpt_type = se->type;
2081 }
2082 
get_seg_entry(struct f2fs_sb_info * sbi,unsigned int segno)2083 struct seg_entry *get_seg_entry(struct f2fs_sb_info *sbi,
2084 		unsigned int segno)
2085 {
2086 	struct sit_info *sit_i = SIT_I(sbi);
2087 	return &sit_i->sentries[segno];
2088 }
2089 
get_seg_vblocks(struct f2fs_sb_info * sbi,struct seg_entry * se)2090 unsigned short get_seg_vblocks(struct f2fs_sb_info *sbi, struct seg_entry *se)
2091 {
2092 	if (!need_fsync_data_record(sbi))
2093 		return se->valid_blocks;
2094 	else
2095 		return se->ckpt_valid_blocks;
2096 }
2097 
get_seg_bitmap(struct f2fs_sb_info * sbi,struct seg_entry * se)2098 unsigned char *get_seg_bitmap(struct f2fs_sb_info *sbi, struct seg_entry *se)
2099 {
2100 	if (!need_fsync_data_record(sbi))
2101 		return se->cur_valid_map;
2102 	else
2103 		return se->ckpt_valid_map;
2104 }
2105 
get_seg_type(struct f2fs_sb_info * sbi,struct seg_entry * se)2106 unsigned char get_seg_type(struct f2fs_sb_info *sbi, struct seg_entry *se)
2107 {
2108 	if (!need_fsync_data_record(sbi))
2109 		return se->type;
2110 	else
2111 		return se->ckpt_type;
2112 }
2113 
get_sum_block(struct f2fs_sb_info * sbi,unsigned int segno,int * ret_type)2114 struct f2fs_summary_block *get_sum_block(struct f2fs_sb_info *sbi,
2115 				unsigned int segno, int *ret_type)
2116 {
2117 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2118 	struct f2fs_summary_block *sum_blk;
2119 	struct curseg_info *curseg;
2120 	int type, ret;
2121 	u64 ssa_blk;
2122 
2123 	*ret_type= SEG_TYPE_MAX;
2124 
2125 	ssa_blk = GET_SUM_BLKADDR(sbi, segno);
2126 	for (type = 0; type < NR_CURSEG_NODE_TYPE; type++) {
2127 		if (segno == get_cp(cur_node_segno[type])) {
2128 			curseg = CURSEG_I(sbi, CURSEG_HOT_NODE + type);
2129 			if (!IS_SUM_NODE_SEG(curseg->sum_blk->footer)) {
2130 				ASSERT_MSG("segno [0x%x] indicates a data "
2131 						"segment, but should be node",
2132 						segno);
2133 				*ret_type = -SEG_TYPE_CUR_NODE;
2134 			} else {
2135 				*ret_type = SEG_TYPE_CUR_NODE;
2136 			}
2137 			return curseg->sum_blk;
2138 		}
2139 	}
2140 
2141 	for (type = 0; type < NR_CURSEG_DATA_TYPE; type++) {
2142 		if (segno == get_cp(cur_data_segno[type])) {
2143 			curseg = CURSEG_I(sbi, type);
2144 			if (IS_SUM_NODE_SEG(curseg->sum_blk->footer)) {
2145 				ASSERT_MSG("segno [0x%x] indicates a node "
2146 						"segment, but should be data",
2147 						segno);
2148 				*ret_type = -SEG_TYPE_CUR_DATA;
2149 			} else {
2150 				*ret_type = SEG_TYPE_CUR_DATA;
2151 			}
2152 			return curseg->sum_blk;
2153 		}
2154 	}
2155 
2156 	sum_blk = calloc(BLOCK_SZ, 1);
2157 	ASSERT(sum_blk);
2158 
2159 	ret = dev_read_block(sum_blk, ssa_blk);
2160 	ASSERT(ret >= 0);
2161 
2162 	if (IS_SUM_NODE_SEG(sum_blk->footer))
2163 		*ret_type = SEG_TYPE_NODE;
2164 	else if (IS_SUM_DATA_SEG(sum_blk->footer))
2165 		*ret_type = SEG_TYPE_DATA;
2166 
2167 	return sum_blk;
2168 }
2169 
get_sum_entry(struct f2fs_sb_info * sbi,u32 blk_addr,struct f2fs_summary * sum_entry)2170 int get_sum_entry(struct f2fs_sb_info *sbi, u32 blk_addr,
2171 				struct f2fs_summary *sum_entry)
2172 {
2173 	struct f2fs_summary_block *sum_blk;
2174 	u32 segno, offset;
2175 	int type;
2176 
2177 	segno = GET_SEGNO(sbi, blk_addr);
2178 	offset = OFFSET_IN_SEG(sbi, blk_addr);
2179 
2180 	sum_blk = get_sum_block(sbi, segno, &type);
2181 	memcpy(sum_entry, &(sum_blk->entries[offset]),
2182 				sizeof(struct f2fs_summary));
2183 	if (type == SEG_TYPE_NODE || type == SEG_TYPE_DATA ||
2184 					type == SEG_TYPE_MAX)
2185 		free(sum_blk);
2186 	return type;
2187 }
2188 
get_nat_entry(struct f2fs_sb_info * sbi,nid_t nid,struct f2fs_nat_entry * raw_nat)2189 static void get_nat_entry(struct f2fs_sb_info *sbi, nid_t nid,
2190 				struct f2fs_nat_entry *raw_nat)
2191 {
2192 	struct f2fs_nat_block *nat_block;
2193 	pgoff_t block_addr;
2194 	int entry_off;
2195 	int ret;
2196 
2197 	if (lookup_nat_in_journal(sbi, nid, raw_nat) >= 0)
2198 		return;
2199 
2200 	nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2201 	ASSERT(nat_block);
2202 
2203 	entry_off = nid % NAT_ENTRY_PER_BLOCK;
2204 	block_addr = current_nat_addr(sbi, nid, NULL);
2205 
2206 	ret = dev_read_block(nat_block, block_addr);
2207 	ASSERT(ret >= 0);
2208 
2209 	memcpy(raw_nat, &nat_block->entries[entry_off],
2210 					sizeof(struct f2fs_nat_entry));
2211 	free(nat_block);
2212 }
2213 
update_data_blkaddr(struct f2fs_sb_info * sbi,nid_t nid,u16 ofs_in_node,block_t newaddr)2214 void update_data_blkaddr(struct f2fs_sb_info *sbi, nid_t nid,
2215 				u16 ofs_in_node, block_t newaddr)
2216 {
2217 	struct f2fs_node *node_blk = NULL;
2218 	struct node_info ni;
2219 	block_t oldaddr, startaddr, endaddr;
2220 	int ret;
2221 
2222 	node_blk = (struct f2fs_node *)calloc(BLOCK_SZ, 1);
2223 	ASSERT(node_blk);
2224 
2225 	get_node_info(sbi, nid, &ni);
2226 
2227 	/* read node_block */
2228 	ret = dev_read_block(node_blk, ni.blk_addr);
2229 	ASSERT(ret >= 0);
2230 
2231 	/* check its block address */
2232 	if (node_blk->footer.nid == node_blk->footer.ino) {
2233 		int ofs = get_extra_isize(node_blk);
2234 
2235 		oldaddr = le32_to_cpu(node_blk->i.i_addr[ofs + ofs_in_node]);
2236 		node_blk->i.i_addr[ofs + ofs_in_node] = cpu_to_le32(newaddr);
2237 		ret = write_inode(node_blk, ni.blk_addr);
2238 		ASSERT(ret >= 0);
2239 	} else {
2240 		oldaddr = le32_to_cpu(node_blk->dn.addr[ofs_in_node]);
2241 		node_blk->dn.addr[ofs_in_node] = cpu_to_le32(newaddr);
2242 		ret = dev_write_block(node_blk, ni.blk_addr);
2243 		ASSERT(ret >= 0);
2244 	}
2245 
2246 	/* check extent cache entry */
2247 	if (node_blk->footer.nid != node_blk->footer.ino) {
2248 		get_node_info(sbi, le32_to_cpu(node_blk->footer.ino), &ni);
2249 
2250 		/* read inode block */
2251 		ret = dev_read_block(node_blk, ni.blk_addr);
2252 		ASSERT(ret >= 0);
2253 	}
2254 
2255 	startaddr = le32_to_cpu(node_blk->i.i_ext.blk_addr);
2256 	endaddr = startaddr + le32_to_cpu(node_blk->i.i_ext.len);
2257 	if (oldaddr >= startaddr && oldaddr < endaddr) {
2258 		node_blk->i.i_ext.len = 0;
2259 
2260 		/* update inode block */
2261 		ASSERT(write_inode(node_blk, ni.blk_addr) >= 0);
2262 	}
2263 	free(node_blk);
2264 }
2265 
update_nat_blkaddr(struct f2fs_sb_info * sbi,nid_t ino,nid_t nid,block_t newaddr)2266 void update_nat_blkaddr(struct f2fs_sb_info *sbi, nid_t ino,
2267 					nid_t nid, block_t newaddr)
2268 {
2269 	struct f2fs_nat_block *nat_block;
2270 	pgoff_t block_addr;
2271 	int entry_off;
2272 	int ret;
2273 
2274 	nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2275 	ASSERT(nat_block);
2276 
2277 	entry_off = nid % NAT_ENTRY_PER_BLOCK;
2278 	block_addr = current_nat_addr(sbi, nid, NULL);
2279 
2280 	ret = dev_read_block(nat_block, block_addr);
2281 	ASSERT(ret >= 0);
2282 
2283 	if (ino)
2284 		nat_block->entries[entry_off].ino = cpu_to_le32(ino);
2285 	nat_block->entries[entry_off].block_addr = cpu_to_le32(newaddr);
2286 	if (c.func == FSCK)
2287 		F2FS_FSCK(sbi)->entries[nid] = nat_block->entries[entry_off];
2288 
2289 	ret = dev_write_block(nat_block, block_addr);
2290 	ASSERT(ret >= 0);
2291 	free(nat_block);
2292 }
2293 
get_node_info(struct f2fs_sb_info * sbi,nid_t nid,struct node_info * ni)2294 void get_node_info(struct f2fs_sb_info *sbi, nid_t nid, struct node_info *ni)
2295 {
2296 	struct f2fs_nat_entry raw_nat;
2297 
2298 	ni->nid = nid;
2299 	if (c.func == FSCK && F2FS_FSCK(sbi)->nr_nat_entries) {
2300 		node_info_from_raw_nat(ni, &(F2FS_FSCK(sbi)->entries[nid]));
2301 		if (ni->blk_addr)
2302 			return;
2303 		/* nat entry is not cached, read it */
2304 	}
2305 
2306 	get_nat_entry(sbi, nid, &raw_nat);
2307 	node_info_from_raw_nat(ni, &raw_nat);
2308 }
2309 
build_sit_entries(struct f2fs_sb_info * sbi)2310 static int build_sit_entries(struct f2fs_sb_info *sbi)
2311 {
2312 	struct sit_info *sit_i = SIT_I(sbi);
2313 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
2314 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
2315 	struct f2fs_sit_block *sit_blk;
2316 	struct seg_entry *se;
2317 	struct f2fs_sit_entry sit;
2318 	int sit_blk_cnt = SIT_BLK_CNT(sbi);
2319 	unsigned int i, segno, end;
2320 	unsigned int readed, start_blk = 0;
2321 
2322 	sit_blk = calloc(BLOCK_SZ, 1);
2323 	if (!sit_blk) {
2324 		MSG(1, "\tError: Calloc failed for build_sit_entries!\n");
2325 		return -ENOMEM;
2326 	}
2327 
2328 	do {
2329 		readed = f2fs_ra_meta_pages(sbi, start_blk, MAX_RA_BLOCKS,
2330 								META_SIT);
2331 
2332 		segno = start_blk * sit_i->sents_per_block;
2333 		end = (start_blk + readed) * sit_i->sents_per_block;
2334 
2335 		for (; segno < end && segno < TOTAL_SEGS(sbi); segno++) {
2336 			se = &sit_i->sentries[segno];
2337 
2338 			get_current_sit_page(sbi, segno, sit_blk);
2339 			sit = sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2340 
2341 			check_block_count(sbi, segno, &sit);
2342 			seg_info_from_raw_sit(sbi, se, &sit);
2343 		}
2344 		start_blk += readed;
2345 	} while (start_blk < sit_blk_cnt);
2346 
2347 
2348 	free(sit_blk);
2349 
2350 	if (sits_in_cursum(journal) > SIT_JOURNAL_ENTRIES) {
2351 		MSG(0, "\tError: build_sit_entries truncate n_sits(%u) to "
2352 			"SIT_JOURNAL_ENTRIES(%lu)\n",
2353 			sits_in_cursum(journal), SIT_JOURNAL_ENTRIES);
2354 		journal->n_sits = cpu_to_le16(SIT_JOURNAL_ENTRIES);
2355 		c.fix_on = 1;
2356 	}
2357 
2358 	for (i = 0; i < sits_in_cursum(journal); i++) {
2359 		segno = le32_to_cpu(segno_in_journal(journal, i));
2360 
2361 		if (segno >= TOTAL_SEGS(sbi)) {
2362 			MSG(0, "\tError: build_sit_entries: segno(%u) is invalid!!!\n", segno);
2363 			journal->n_sits = cpu_to_le16(i);
2364 			c.fix_on = 1;
2365 			continue;
2366 		}
2367 
2368 		se = &sit_i->sentries[segno];
2369 		sit = sit_in_journal(journal, i);
2370 
2371 		check_block_count(sbi, segno, &sit);
2372 		seg_info_from_raw_sit(sbi, se, &sit);
2373 	}
2374 	return 0;
2375 }
2376 
early_build_segment_manager(struct f2fs_sb_info * sbi)2377 static int early_build_segment_manager(struct f2fs_sb_info *sbi)
2378 {
2379 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2380 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2381 	struct f2fs_sm_info *sm_info;
2382 
2383 	sm_info = malloc(sizeof(struct f2fs_sm_info));
2384 	if (!sm_info) {
2385 		MSG(1, "\tError: Malloc failed for build_segment_manager!\n");
2386 		return -ENOMEM;
2387 	}
2388 
2389 	/* init sm info */
2390 	sbi->sm_info = sm_info;
2391 	sm_info->seg0_blkaddr = get_sb(segment0_blkaddr);
2392 	sm_info->main_blkaddr = get_sb(main_blkaddr);
2393 	sm_info->segment_count = get_sb(segment_count);
2394 	sm_info->reserved_segments = get_cp(rsvd_segment_count);
2395 	sm_info->ovp_segments = get_cp(overprov_segment_count);
2396 	sm_info->main_segments = get_sb(segment_count_main);
2397 	sm_info->ssa_blkaddr = get_sb(ssa_blkaddr);
2398 
2399 	if (build_sit_info(sbi) || build_curseg(sbi)) {
2400 		free(sm_info);
2401 		return -ENOMEM;
2402 	}
2403 
2404 	return 0;
2405 }
2406 
late_build_segment_manager(struct f2fs_sb_info * sbi)2407 static int late_build_segment_manager(struct f2fs_sb_info *sbi)
2408 {
2409 	if (sbi->seg_manager_done)
2410 		return 1; /* this function was already called */
2411 
2412 	sbi->seg_manager_done = true;
2413 	if (build_sit_entries(sbi)) {
2414 		free (sbi->sm_info);
2415 		return -ENOMEM;
2416 	}
2417 
2418 	return 0;
2419 }
2420 
build_sit_area_bitmap(struct f2fs_sb_info * sbi)2421 void build_sit_area_bitmap(struct f2fs_sb_info *sbi)
2422 {
2423 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2424 	struct f2fs_sm_info *sm_i = SM_I(sbi);
2425 	unsigned int segno = 0;
2426 	char *ptr = NULL;
2427 	u32 sum_vblocks = 0;
2428 	u32 free_segs = 0;
2429 	struct seg_entry *se;
2430 
2431 	fsck->sit_area_bitmap_sz = sm_i->main_segments * SIT_VBLOCK_MAP_SIZE;
2432 	fsck->sit_area_bitmap = calloc(1, fsck->sit_area_bitmap_sz);
2433 	ASSERT(fsck->sit_area_bitmap);
2434 	ptr = fsck->sit_area_bitmap;
2435 
2436 	ASSERT(fsck->sit_area_bitmap_sz == fsck->main_area_bitmap_sz);
2437 
2438 	for (segno = 0; segno < TOTAL_SEGS(sbi); segno++) {
2439 		se = get_seg_entry(sbi, segno);
2440 
2441 		memcpy(ptr, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2442 		ptr += SIT_VBLOCK_MAP_SIZE;
2443 
2444 		if (se->valid_blocks == 0x0 && is_usable_seg(sbi, segno)) {
2445 			if (le32_to_cpu(sbi->ckpt->cur_node_segno[0]) == segno ||
2446 				le32_to_cpu(sbi->ckpt->cur_data_segno[0]) == segno ||
2447 				le32_to_cpu(sbi->ckpt->cur_node_segno[1]) == segno ||
2448 				le32_to_cpu(sbi->ckpt->cur_data_segno[1]) == segno ||
2449 				le32_to_cpu(sbi->ckpt->cur_node_segno[2]) == segno ||
2450 				le32_to_cpu(sbi->ckpt->cur_data_segno[2]) == segno) {
2451 				continue;
2452 			} else {
2453 				free_segs++;
2454 			}
2455 		} else {
2456 			sum_vblocks += se->valid_blocks;
2457 		}
2458 	}
2459 	fsck->chk.sit_valid_blocks = sum_vblocks;
2460 	fsck->chk.sit_free_segs = free_segs;
2461 
2462 	DBG(1, "Blocks [0x%x : %d] Free Segs [0x%x : %d]\n\n",
2463 			sum_vblocks, sum_vblocks,
2464 			free_segs, free_segs);
2465 }
2466 
rewrite_sit_area_bitmap(struct f2fs_sb_info * sbi)2467 void rewrite_sit_area_bitmap(struct f2fs_sb_info *sbi)
2468 {
2469 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
2470 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
2471 	struct sit_info *sit_i = SIT_I(sbi);
2472 	struct f2fs_sit_block *sit_blk;
2473 	unsigned int segno = 0;
2474 	struct f2fs_summary_block *sum = curseg->sum_blk;
2475 	char *ptr = NULL;
2476 
2477 	sit_blk = calloc(BLOCK_SZ, 1);
2478 	ASSERT(sit_blk);
2479 	/* remove sit journal */
2480 	sum->journal.n_sits = 0;
2481 
2482 	ptr = fsck->main_area_bitmap;
2483 
2484 	for (segno = 0; segno < TOTAL_SEGS(sbi); segno++) {
2485 		struct f2fs_sit_entry *sit;
2486 		struct seg_entry *se;
2487 		u16 valid_blocks = 0;
2488 		u16 type;
2489 		int i;
2490 
2491 		get_current_sit_page(sbi, segno, sit_blk);
2492 		sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2493 		memcpy(sit->valid_map, ptr, SIT_VBLOCK_MAP_SIZE);
2494 
2495 		/* update valid block count */
2496 		for (i = 0; i < SIT_VBLOCK_MAP_SIZE; i++)
2497 			valid_blocks += get_bits_in_byte(sit->valid_map[i]);
2498 
2499 		se = get_seg_entry(sbi, segno);
2500 		memcpy(se->cur_valid_map, ptr, SIT_VBLOCK_MAP_SIZE);
2501 		se->valid_blocks = valid_blocks;
2502 		type = se->type;
2503 		if (type >= NO_CHECK_TYPE) {
2504 			ASSERT_MSG("Invalid type and valid blocks=%x,%x",
2505 					segno, valid_blocks);
2506 			type = 0;
2507 		}
2508 		sit->vblocks = cpu_to_le16((type << SIT_VBLOCKS_SHIFT) |
2509 								valid_blocks);
2510 		rewrite_current_sit_page(sbi, segno, sit_blk);
2511 
2512 		ptr += SIT_VBLOCK_MAP_SIZE;
2513 	}
2514 
2515 	free(sit_blk);
2516 }
2517 
flush_sit_journal_entries(struct f2fs_sb_info * sbi)2518 static int flush_sit_journal_entries(struct f2fs_sb_info *sbi)
2519 {
2520 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_COLD_DATA);
2521 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
2522 	struct sit_info *sit_i = SIT_I(sbi);
2523 	struct f2fs_sit_block *sit_blk;
2524 	unsigned int segno;
2525 	int i;
2526 
2527 	sit_blk = calloc(BLOCK_SZ, 1);
2528 	ASSERT(sit_blk);
2529 	for (i = 0; i < sits_in_cursum(journal); i++) {
2530 		struct f2fs_sit_entry *sit;
2531 		struct seg_entry *se;
2532 
2533 		segno = segno_in_journal(journal, i);
2534 		se = get_seg_entry(sbi, segno);
2535 
2536 		get_current_sit_page(sbi, segno, sit_blk);
2537 		sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2538 
2539 		memcpy(sit->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2540 		sit->vblocks = cpu_to_le16((se->type << SIT_VBLOCKS_SHIFT) |
2541 							se->valid_blocks);
2542 		sit->mtime = cpu_to_le64(se->mtime);
2543 
2544 		rewrite_current_sit_page(sbi, segno, sit_blk);
2545 	}
2546 
2547 	free(sit_blk);
2548 	journal->n_sits = 0;
2549 	return i;
2550 }
2551 
flush_nat_journal_entries(struct f2fs_sb_info * sbi)2552 static int flush_nat_journal_entries(struct f2fs_sb_info *sbi)
2553 {
2554 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2555 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
2556 	struct f2fs_nat_block *nat_block;
2557 	pgoff_t block_addr;
2558 	int entry_off;
2559 	nid_t nid;
2560 	int ret;
2561 	int i = 0;
2562 
2563 	nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2564 	ASSERT(nat_block);
2565 next:
2566 	if (i >= nats_in_cursum(journal)) {
2567 		free(nat_block);
2568 		journal->n_nats = 0;
2569 		return i;
2570 	}
2571 
2572 	nid = le32_to_cpu(nid_in_journal(journal, i));
2573 
2574 	entry_off = nid % NAT_ENTRY_PER_BLOCK;
2575 	block_addr = current_nat_addr(sbi, nid, NULL);
2576 
2577 	ret = dev_read_block(nat_block, block_addr);
2578 	ASSERT(ret >= 0);
2579 
2580 	memcpy(&nat_block->entries[entry_off], &nat_in_journal(journal, i),
2581 					sizeof(struct f2fs_nat_entry));
2582 
2583 	ret = dev_write_block(nat_block, block_addr);
2584 	ASSERT(ret >= 0);
2585 	i++;
2586 	goto next;
2587 }
2588 
flush_journal_entries(struct f2fs_sb_info * sbi)2589 void flush_journal_entries(struct f2fs_sb_info *sbi)
2590 {
2591 	int n_nats = flush_nat_journal_entries(sbi);
2592 	int n_sits = flush_sit_journal_entries(sbi);
2593 
2594 	if (n_nats || n_sits)
2595 		write_checkpoints(sbi);
2596 }
2597 
flush_sit_entries(struct f2fs_sb_info * sbi)2598 void flush_sit_entries(struct f2fs_sb_info *sbi)
2599 {
2600 	struct sit_info *sit_i = SIT_I(sbi);
2601 	struct f2fs_sit_block *sit_blk;
2602 	unsigned int segno = 0;
2603 
2604 	sit_blk = calloc(BLOCK_SZ, 1);
2605 	ASSERT(sit_blk);
2606 	/* update free segments */
2607 	for (segno = 0; segno < TOTAL_SEGS(sbi); segno++) {
2608 		struct f2fs_sit_entry *sit;
2609 		struct seg_entry *se;
2610 
2611 		se = get_seg_entry(sbi, segno);
2612 
2613 		if (!se->dirty)
2614 			continue;
2615 
2616 		get_current_sit_page(sbi, segno, sit_blk);
2617 		sit = &sit_blk->entries[SIT_ENTRY_OFFSET(sit_i, segno)];
2618 		memcpy(sit->valid_map, se->cur_valid_map, SIT_VBLOCK_MAP_SIZE);
2619 		sit->vblocks = cpu_to_le16((se->type << SIT_VBLOCKS_SHIFT) |
2620 							se->valid_blocks);
2621 		rewrite_current_sit_page(sbi, segno, sit_blk);
2622 	}
2623 
2624 	free(sit_blk);
2625 }
2626 
relocate_curseg_offset(struct f2fs_sb_info * sbi,int type)2627 int relocate_curseg_offset(struct f2fs_sb_info *sbi, int type)
2628 {
2629 	struct curseg_info *curseg = CURSEG_I(sbi, type);
2630 	struct seg_entry *se = get_seg_entry(sbi, curseg->segno);
2631 	unsigned int i;
2632 
2633 	if (c.zoned_model == F2FS_ZONED_HM)
2634 		return -EINVAL;
2635 
2636 	for (i = 0; i < sbi->blocks_per_seg; i++) {
2637 		if (!f2fs_test_bit(i, (const char *)se->cur_valid_map))
2638 			break;
2639 	}
2640 
2641 	if (i == sbi->blocks_per_seg)
2642 		return -EINVAL;
2643 
2644 	DBG(1, "Update curseg[%d].next_blkoff %u -> %u, alloc_type %s -> SSR\n",
2645 			type, curseg->next_blkoff, i,
2646 			curseg->alloc_type == LFS ? "LFS" : "SSR");
2647 
2648 	curseg->next_blkoff = i;
2649 	curseg->alloc_type = SSR;
2650 
2651 	return 0;
2652 }
2653 
set_section_type(struct f2fs_sb_info * sbi,unsigned int segno,int type)2654 void set_section_type(struct f2fs_sb_info *sbi, unsigned int segno, int type)
2655 {
2656 	unsigned int i;
2657 
2658 	if (sbi->segs_per_sec == 1)
2659 		return;
2660 
2661 	for (i = 0; i < sbi->segs_per_sec; i++) {
2662 		struct seg_entry *se = get_seg_entry(sbi, segno + i);
2663 
2664 		se->type = type;
2665 	}
2666 }
2667 
2668 #ifdef HAVE_LINUX_BLKZONED_H
2669 
write_pointer_at_zone_start(struct f2fs_sb_info * sbi,unsigned int zone_segno)2670 static bool write_pointer_at_zone_start(struct f2fs_sb_info *sbi,
2671 					unsigned int zone_segno)
2672 {
2673 	u_int64_t sector;
2674 	struct blk_zone blkz;
2675 	block_t block = START_BLOCK(sbi, zone_segno);
2676 	int log_sectors_per_block = sbi->log_blocksize - SECTOR_SHIFT;
2677 	int ret, j;
2678 
2679 	if (c.zoned_model != F2FS_ZONED_HM)
2680 		return true;
2681 
2682 	for (j = 0; j < MAX_DEVICES; j++) {
2683 		if (!c.devices[j].path)
2684 			break;
2685 		if (c.devices[j].start_blkaddr <= block &&
2686 		    block <= c.devices[j].end_blkaddr)
2687 			break;
2688 	}
2689 
2690 	if (j >= MAX_DEVICES)
2691 		return false;
2692 
2693 	sector = (block - c.devices[j].start_blkaddr) << log_sectors_per_block;
2694 	ret = f2fs_report_zone(j, sector, &blkz);
2695 	if (ret)
2696 		return false;
2697 
2698 	if (blk_zone_type(&blkz) != BLK_ZONE_TYPE_SEQWRITE_REQ)
2699 		return true;
2700 
2701 	return blk_zone_sector(&blkz) == blk_zone_wp_sector(&blkz);
2702 }
2703 
2704 #else
2705 
write_pointer_at_zone_start(struct f2fs_sb_info * UNUSED (sbi),unsigned int UNUSED (zone_segno))2706 static bool write_pointer_at_zone_start(struct f2fs_sb_info *UNUSED(sbi),
2707 					unsigned int UNUSED(zone_segno))
2708 {
2709 	return true;
2710 }
2711 
2712 #endif
2713 
find_next_free_block(struct f2fs_sb_info * sbi,u64 * to,int left,int want_type,bool new_sec)2714 int find_next_free_block(struct f2fs_sb_info *sbi, u64 *to, int left,
2715 						int want_type, bool new_sec)
2716 {
2717 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2718 	struct seg_entry *se;
2719 	u32 segno;
2720 	u32 offset;
2721 	int not_enough = 0;
2722 	u64 end_blkaddr = (get_sb(segment_count_main) <<
2723 			get_sb(log_blocks_per_seg)) + get_sb(main_blkaddr);
2724 
2725 	if (*to > 0)
2726 		*to -= left;
2727 	if (get_free_segments(sbi) <= SM_I(sbi)->reserved_segments + 1)
2728 		not_enough = 1;
2729 
2730 	while (*to >= SM_I(sbi)->main_blkaddr && *to < end_blkaddr) {
2731 		unsigned short vblocks;
2732 		unsigned char *bitmap;
2733 		unsigned char type;
2734 
2735 		segno = GET_SEGNO(sbi, *to);
2736 		offset = OFFSET_IN_SEG(sbi, *to);
2737 
2738 		se = get_seg_entry(sbi, segno);
2739 
2740 		vblocks = get_seg_vblocks(sbi, se);
2741 		bitmap = get_seg_bitmap(sbi, se);
2742 		type = get_seg_type(sbi, se);
2743 
2744 		if (vblocks == sbi->blocks_per_seg) {
2745 next_segment:
2746 			*to = left ? START_BLOCK(sbi, segno) - 1:
2747 						START_BLOCK(sbi, segno + 1);
2748 			continue;
2749 		}
2750 		if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO)) &&
2751 						IS_CUR_SEGNO(sbi, segno))
2752 			goto next_segment;
2753 		if (vblocks == 0 && not_enough)
2754 			goto next_segment;
2755 
2756 		if (vblocks == 0 && !(segno % sbi->segs_per_sec)) {
2757 			struct seg_entry *se2;
2758 			unsigned int i;
2759 
2760 			for (i = 1; i < sbi->segs_per_sec; i++) {
2761 				se2 = get_seg_entry(sbi, segno + i);
2762 				if (get_seg_vblocks(sbi, se2))
2763 					break;
2764 			}
2765 
2766 			if (i == sbi->segs_per_sec &&
2767 			    write_pointer_at_zone_start(sbi, segno)) {
2768 				set_section_type(sbi, segno, want_type);
2769 				return 0;
2770 			}
2771 		}
2772 
2773 		if (type == want_type && !new_sec &&
2774 			!f2fs_test_bit(offset, (const char *)bitmap))
2775 			return 0;
2776 
2777 		*to = left ? *to - 1: *to + 1;
2778 	}
2779 	return -1;
2780 }
2781 
move_one_curseg_info(struct f2fs_sb_info * sbi,u64 from,int left,int i)2782 static void move_one_curseg_info(struct f2fs_sb_info *sbi, u64 from, int left,
2783 				 int i)
2784 {
2785 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2786 	struct curseg_info *curseg = CURSEG_I(sbi, i);
2787 	struct f2fs_summary_block buf;
2788 	u32 old_segno;
2789 	u64 ssa_blk, to;
2790 	int ret;
2791 
2792 	if ((get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO))) {
2793 		if (i != CURSEG_HOT_DATA && i != CURSEG_HOT_NODE)
2794 			return;
2795 		goto bypass_ssa;
2796 	}
2797 
2798 	/* update original SSA too */
2799 	ssa_blk = GET_SUM_BLKADDR(sbi, curseg->segno);
2800 	ret = dev_write_block(curseg->sum_blk, ssa_blk);
2801 	ASSERT(ret >= 0);
2802 bypass_ssa:
2803 	to = from;
2804 	ret = find_next_free_block(sbi, &to, left, i,
2805 				   c.zoned_model == F2FS_ZONED_HM);
2806 	ASSERT(ret == 0);
2807 
2808 	old_segno = curseg->segno;
2809 	curseg->segno = GET_SEGNO(sbi, to);
2810 	curseg->next_blkoff = OFFSET_IN_SEG(sbi, to);
2811 	curseg->alloc_type = c.zoned_model == F2FS_ZONED_HM ? LFS : SSR;
2812 
2813 	/* update new segno */
2814 	ssa_blk = GET_SUM_BLKADDR(sbi, curseg->segno);
2815 	ret = dev_read_block(&buf, ssa_blk);
2816 	ASSERT(ret >= 0);
2817 
2818 	memcpy(curseg->sum_blk, &buf, SUM_ENTRIES_SIZE);
2819 
2820 	/* update se->types */
2821 	reset_curseg(sbi, i);
2822 
2823 	FIX_MSG("Move curseg[%d] %x -> %x after %"PRIx64"\n",
2824 		i, old_segno, curseg->segno, from);
2825 }
2826 
move_curseg_info(struct f2fs_sb_info * sbi,u64 from,int left)2827 void move_curseg_info(struct f2fs_sb_info *sbi, u64 from, int left)
2828 {
2829 	int i;
2830 
2831 	/* update summary blocks having nullified journal entries */
2832 	for (i = 0; i < NO_CHECK_TYPE; i++)
2833 		move_one_curseg_info(sbi, from, left, i);
2834 }
2835 
update_curseg_info(struct f2fs_sb_info * sbi,int type)2836 void update_curseg_info(struct f2fs_sb_info *sbi, int type)
2837 {
2838 	if (!relocate_curseg_offset(sbi, type))
2839 		return;
2840 	move_one_curseg_info(sbi, SM_I(sbi)->main_blkaddr, 0, type);
2841 }
2842 
zero_journal_entries(struct f2fs_sb_info * sbi)2843 void zero_journal_entries(struct f2fs_sb_info *sbi)
2844 {
2845 	int i;
2846 
2847 	for (i = 0; i < NO_CHECK_TYPE; i++)
2848 		CURSEG_I(sbi, i)->sum_blk->journal.n_nats = 0;
2849 }
2850 
write_curseg_info(struct f2fs_sb_info * sbi)2851 void write_curseg_info(struct f2fs_sb_info *sbi)
2852 {
2853 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2854 	int i;
2855 
2856 	for (i = 0; i < NO_CHECK_TYPE; i++) {
2857 		cp->alloc_type[i] = CURSEG_I(sbi, i)->alloc_type;
2858 		if (i < CURSEG_HOT_NODE) {
2859 			set_cp(cur_data_segno[i], CURSEG_I(sbi, i)->segno);
2860 			set_cp(cur_data_blkoff[i],
2861 					CURSEG_I(sbi, i)->next_blkoff);
2862 		} else {
2863 			int n = i - CURSEG_HOT_NODE;
2864 
2865 			set_cp(cur_node_segno[n], CURSEG_I(sbi, i)->segno);
2866 			set_cp(cur_node_blkoff[n],
2867 					CURSEG_I(sbi, i)->next_blkoff);
2868 		}
2869 	}
2870 }
2871 
lookup_nat_in_journal(struct f2fs_sb_info * sbi,u32 nid,struct f2fs_nat_entry * raw_nat)2872 int lookup_nat_in_journal(struct f2fs_sb_info *sbi, u32 nid,
2873 					struct f2fs_nat_entry *raw_nat)
2874 {
2875 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2876 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
2877 	int i = 0;
2878 
2879 	for (i = 0; i < nats_in_cursum(journal); i++) {
2880 		if (le32_to_cpu(nid_in_journal(journal, i)) == nid) {
2881 			memcpy(raw_nat, &nat_in_journal(journal, i),
2882 						sizeof(struct f2fs_nat_entry));
2883 			DBG(3, "==> Found nid [0x%x] in nat cache\n", nid);
2884 			return i;
2885 		}
2886 	}
2887 	return -1;
2888 }
2889 
nullify_nat_entry(struct f2fs_sb_info * sbi,u32 nid)2890 void nullify_nat_entry(struct f2fs_sb_info *sbi, u32 nid)
2891 {
2892 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
2893 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
2894 	struct f2fs_nat_block *nat_block;
2895 	pgoff_t block_addr;
2896 	int entry_off;
2897 	int ret;
2898 	int i = 0;
2899 
2900 	/* check in journal */
2901 	for (i = 0; i < nats_in_cursum(journal); i++) {
2902 		if (le32_to_cpu(nid_in_journal(journal, i)) == nid) {
2903 			memset(&nat_in_journal(journal, i), 0,
2904 					sizeof(struct f2fs_nat_entry));
2905 			FIX_MSG("Remove nid [0x%x] in nat journal", nid);
2906 			return;
2907 		}
2908 	}
2909 	nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
2910 	ASSERT(nat_block);
2911 
2912 	entry_off = nid % NAT_ENTRY_PER_BLOCK;
2913 	block_addr = current_nat_addr(sbi, nid, NULL);
2914 
2915 	ret = dev_read_block(nat_block, block_addr);
2916 	ASSERT(ret >= 0);
2917 
2918 	if (nid == F2FS_NODE_INO(sbi) || nid == F2FS_META_INO(sbi)) {
2919 		FIX_MSG("nid [0x%x] block_addr= 0x%x -> 0x1", nid,
2920 			le32_to_cpu(nat_block->entries[entry_off].block_addr));
2921 		nat_block->entries[entry_off].block_addr = cpu_to_le32(0x1);
2922 	} else {
2923 		memset(&nat_block->entries[entry_off], 0,
2924 					sizeof(struct f2fs_nat_entry));
2925 		FIX_MSG("Remove nid [0x%x] in NAT", nid);
2926 	}
2927 
2928 	ret = dev_write_block(nat_block, block_addr);
2929 	ASSERT(ret >= 0);
2930 	free(nat_block);
2931 }
2932 
duplicate_checkpoint(struct f2fs_sb_info * sbi)2933 void duplicate_checkpoint(struct f2fs_sb_info *sbi)
2934 {
2935 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2936 	unsigned long long dst, src;
2937 	void *buf;
2938 	unsigned int seg_size = 1 << get_sb(log_blocks_per_seg);
2939 	int ret;
2940 
2941 	if (sbi->cp_backuped)
2942 		return;
2943 
2944 	buf = malloc(F2FS_BLKSIZE * seg_size);
2945 	ASSERT(buf);
2946 
2947 	if (sbi->cur_cp == 1) {
2948 		src = get_sb(cp_blkaddr);
2949 		dst = src + seg_size;
2950 	} else {
2951 		dst = get_sb(cp_blkaddr);
2952 		src = dst + seg_size;
2953 	}
2954 
2955 	ret = dev_read(buf, src << F2FS_BLKSIZE_BITS,
2956 				seg_size << F2FS_BLKSIZE_BITS);
2957 	ASSERT(ret >= 0);
2958 
2959 	ret = dev_write(buf, dst << F2FS_BLKSIZE_BITS,
2960 				seg_size << F2FS_BLKSIZE_BITS);
2961 	ASSERT(ret >= 0);
2962 
2963 	free(buf);
2964 
2965 	ret = f2fs_fsync_device();
2966 	ASSERT(ret >= 0);
2967 
2968 	sbi->cp_backuped = 1;
2969 
2970 	MSG(0, "Info: Duplicate valid checkpoint to mirror position "
2971 		"%llu -> %llu\n", src, dst);
2972 }
2973 
write_checkpoint(struct f2fs_sb_info * sbi)2974 void write_checkpoint(struct f2fs_sb_info *sbi)
2975 {
2976 	struct f2fs_checkpoint *cp = F2FS_CKPT(sbi);
2977 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
2978 	block_t orphan_blks = 0;
2979 	unsigned long long cp_blk_no;
2980 	u32 flags = CP_UMOUNT_FLAG;
2981 	int i, ret;
2982 	u_int32_t crc = 0;
2983 
2984 	if (is_set_ckpt_flags(cp, CP_ORPHAN_PRESENT_FLAG)) {
2985 		orphan_blks = __start_sum_addr(sbi) - 1;
2986 		flags |= CP_ORPHAN_PRESENT_FLAG;
2987 	}
2988 	if (is_set_ckpt_flags(cp, CP_TRIMMED_FLAG))
2989 		flags |= CP_TRIMMED_FLAG;
2990 	if (is_set_ckpt_flags(cp, CP_DISABLED_FLAG))
2991 		flags |= CP_DISABLED_FLAG;
2992 	if (is_set_ckpt_flags(cp, CP_LARGE_NAT_BITMAP_FLAG)) {
2993 		flags |= CP_LARGE_NAT_BITMAP_FLAG;
2994 		set_cp(checksum_offset, CP_MIN_CHKSUM_OFFSET);
2995 	} else {
2996 		set_cp(checksum_offset, CP_CHKSUM_OFFSET);
2997 	}
2998 
2999 	set_cp(free_segment_count, get_free_segments(sbi));
3000 	if (c.func == FSCK) {
3001 		struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
3002 
3003 		set_cp(valid_block_count, fsck->chk.valid_blk_cnt);
3004 		set_cp(valid_node_count, fsck->chk.valid_node_cnt);
3005 		set_cp(valid_inode_count, fsck->chk.valid_inode_cnt);
3006 	} else {
3007 		set_cp(valid_block_count, sbi->total_valid_block_count);
3008 		set_cp(valid_node_count, sbi->total_valid_node_count);
3009 		set_cp(valid_inode_count, sbi->total_valid_inode_count);
3010 	}
3011 	set_cp(cp_pack_total_block_count, 8 + orphan_blks + get_sb(cp_payload));
3012 
3013 	flags = update_nat_bits_flags(sb, cp, flags);
3014 	set_cp(ckpt_flags, flags);
3015 
3016 	crc = f2fs_checkpoint_chksum(cp);
3017 	*((__le32 *)((unsigned char *)cp + get_cp(checksum_offset))) =
3018 							cpu_to_le32(crc);
3019 
3020 	cp_blk_no = get_sb(cp_blkaddr);
3021 	if (sbi->cur_cp == 2)
3022 		cp_blk_no += 1 << get_sb(log_blocks_per_seg);
3023 
3024 	/* write the first cp */
3025 	ret = dev_write_block(cp, cp_blk_no++);
3026 	ASSERT(ret >= 0);
3027 
3028 	/* skip payload */
3029 	cp_blk_no += get_sb(cp_payload);
3030 	/* skip orphan blocks */
3031 	cp_blk_no += orphan_blks;
3032 
3033 	/* update summary blocks having nullified journal entries */
3034 	for (i = 0; i < NO_CHECK_TYPE; i++) {
3035 		struct curseg_info *curseg = CURSEG_I(sbi, i);
3036 		u64 ssa_blk;
3037 
3038 		ret = dev_write_block(curseg->sum_blk, cp_blk_no++);
3039 		ASSERT(ret >= 0);
3040 
3041 		if (!(get_sb(feature) & cpu_to_le32(F2FS_FEATURE_RO))) {
3042 			/* update original SSA too */
3043 			ssa_blk = GET_SUM_BLKADDR(sbi, curseg->segno);
3044 			ret = dev_write_block(curseg->sum_blk, ssa_blk);
3045 			ASSERT(ret >= 0);
3046 		}
3047 	}
3048 
3049 	/* Write nat bits */
3050 	if (flags & CP_NAT_BITS_FLAG)
3051 		write_nat_bits(sbi, sb, cp, sbi->cur_cp);
3052 
3053 	/* in case of sudden power off */
3054 	ret = f2fs_fsync_device();
3055 	ASSERT(ret >= 0);
3056 
3057 	/* write the last cp */
3058 	ret = dev_write_block(cp, cp_blk_no++);
3059 	ASSERT(ret >= 0);
3060 
3061 	ret = f2fs_fsync_device();
3062 	ASSERT(ret >= 0);
3063 }
3064 
write_checkpoints(struct f2fs_sb_info * sbi)3065 void write_checkpoints(struct f2fs_sb_info *sbi)
3066 {
3067 	/* copy valid checkpoint to its mirror position */
3068 	duplicate_checkpoint(sbi);
3069 
3070 	/* repair checkpoint at CP #0 position */
3071 	sbi->cur_cp = 1;
3072 	write_checkpoint(sbi);
3073 }
3074 
build_nat_area_bitmap(struct f2fs_sb_info * sbi)3075 void build_nat_area_bitmap(struct f2fs_sb_info *sbi)
3076 {
3077 	struct curseg_info *curseg = CURSEG_I(sbi, CURSEG_HOT_DATA);
3078 	struct f2fs_journal *journal = &curseg->sum_blk->journal;
3079 	struct f2fs_fsck *fsck = F2FS_FSCK(sbi);
3080 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
3081 	struct f2fs_nm_info *nm_i = NM_I(sbi);
3082 	struct f2fs_nat_block *nat_block;
3083 	struct node_info ni;
3084 	u32 nid, nr_nat_blks;
3085 	pgoff_t block_off;
3086 	pgoff_t block_addr;
3087 	int seg_off;
3088 	int ret;
3089 	unsigned int i;
3090 
3091 	nat_block = (struct f2fs_nat_block *)calloc(BLOCK_SZ, 1);
3092 	ASSERT(nat_block);
3093 
3094 	/* Alloc & build nat entry bitmap */
3095 	nr_nat_blks = (get_sb(segment_count_nat) / 2) <<
3096 					sbi->log_blocks_per_seg;
3097 
3098 	fsck->nr_nat_entries = nr_nat_blks * NAT_ENTRY_PER_BLOCK;
3099 	fsck->nat_area_bitmap_sz = (fsck->nr_nat_entries + 7) / 8;
3100 	fsck->nat_area_bitmap = calloc(fsck->nat_area_bitmap_sz, 1);
3101 	ASSERT(fsck->nat_area_bitmap);
3102 
3103 	fsck->entries = calloc(sizeof(struct f2fs_nat_entry),
3104 					fsck->nr_nat_entries);
3105 	ASSERT(fsck->entries);
3106 
3107 	for (block_off = 0; block_off < nr_nat_blks; block_off++) {
3108 
3109 		seg_off = block_off >> sbi->log_blocks_per_seg;
3110 		block_addr = (pgoff_t)(nm_i->nat_blkaddr +
3111 			(seg_off << sbi->log_blocks_per_seg << 1) +
3112 			(block_off & ((1 << sbi->log_blocks_per_seg) - 1)));
3113 
3114 		if (f2fs_test_bit(block_off, nm_i->nat_bitmap))
3115 			block_addr += sbi->blocks_per_seg;
3116 
3117 		ret = dev_read_block(nat_block, block_addr);
3118 		ASSERT(ret >= 0);
3119 
3120 		nid = block_off * NAT_ENTRY_PER_BLOCK;
3121 		for (i = 0; i < NAT_ENTRY_PER_BLOCK; i++) {
3122 			ni.nid = nid + i;
3123 
3124 			if ((nid + i) == F2FS_NODE_INO(sbi) ||
3125 					(nid + i) == F2FS_META_INO(sbi)) {
3126 				/*
3127 				 * block_addr of node/meta inode should be 0x1.
3128 				 * Set this bit, and fsck_verify will fix it.
3129 				 */
3130 				if (le32_to_cpu(nat_block->entries[i].block_addr) != 0x1) {
3131 					ASSERT_MSG("\tError: ino[0x%x] block_addr[0x%x] is invalid\n",
3132 							nid + i, le32_to_cpu(nat_block->entries[i].block_addr));
3133 					f2fs_set_bit(nid + i, fsck->nat_area_bitmap);
3134 				}
3135 				continue;
3136 			}
3137 
3138 			node_info_from_raw_nat(&ni, &nat_block->entries[i]);
3139 			if (ni.blk_addr == 0x0)
3140 				continue;
3141 			if (ni.ino == 0x0) {
3142 				ASSERT_MSG("\tError: ino[0x%8x] or blk_addr[0x%16x]"
3143 					" is invalid\n", ni.ino, ni.blk_addr);
3144 			}
3145 			if (ni.ino == (nid + i)) {
3146 				fsck->nat_valid_inode_cnt++;
3147 				DBG(3, "ino[0x%8x] maybe is inode\n", ni.ino);
3148 			}
3149 			if (nid + i == 0) {
3150 				/*
3151 				 * nat entry [0] must be null.  If
3152 				 * it is corrupted, set its bit in
3153 				 * nat_area_bitmap, fsck_verify will
3154 				 * nullify it
3155 				 */
3156 				ASSERT_MSG("Invalid nat entry[0]: "
3157 					"blk_addr[0x%x]\n", ni.blk_addr);
3158 				fsck->chk.valid_nat_entry_cnt--;
3159 			}
3160 
3161 			DBG(3, "nid[0x%8x] addr[0x%16x] ino[0x%8x]\n",
3162 				nid + i, ni.blk_addr, ni.ino);
3163 			f2fs_set_bit(nid + i, fsck->nat_area_bitmap);
3164 			fsck->chk.valid_nat_entry_cnt++;
3165 
3166 			fsck->entries[nid + i] = nat_block->entries[i];
3167 		}
3168 	}
3169 
3170 	/* Traverse nat journal, update the corresponding entries */
3171 	for (i = 0; i < nats_in_cursum(journal); i++) {
3172 		struct f2fs_nat_entry raw_nat;
3173 		nid = le32_to_cpu(nid_in_journal(journal, i));
3174 		ni.nid = nid;
3175 
3176 		DBG(3, "==> Found nid [0x%x] in nat cache, update it\n", nid);
3177 
3178 		/* Clear the original bit and count */
3179 		if (fsck->entries[nid].block_addr != 0x0) {
3180 			fsck->chk.valid_nat_entry_cnt--;
3181 			f2fs_clear_bit(nid, fsck->nat_area_bitmap);
3182 			if (fsck->entries[nid].ino == nid)
3183 				fsck->nat_valid_inode_cnt--;
3184 		}
3185 
3186 		/* Use nat entries in journal */
3187 		memcpy(&raw_nat, &nat_in_journal(journal, i),
3188 					sizeof(struct f2fs_nat_entry));
3189 		node_info_from_raw_nat(&ni, &raw_nat);
3190 		if (ni.blk_addr != 0x0) {
3191 			if (ni.ino == 0x0)
3192 				ASSERT_MSG("\tError: ino[0x%8x] or blk_addr[0x%16x]"
3193 					" is invalid\n", ni.ino, ni.blk_addr);
3194 			if (ni.ino == nid) {
3195 				fsck->nat_valid_inode_cnt++;
3196 				DBG(3, "ino[0x%8x] maybe is inode\n", ni.ino);
3197 			}
3198 			f2fs_set_bit(nid, fsck->nat_area_bitmap);
3199 			fsck->chk.valid_nat_entry_cnt++;
3200 			DBG(3, "nid[0x%x] in nat cache\n", nid);
3201 		}
3202 		fsck->entries[nid] = raw_nat;
3203 	}
3204 	free(nat_block);
3205 
3206 	DBG(1, "valid nat entries (block_addr != 0x0) [0x%8x : %u]\n",
3207 			fsck->chk.valid_nat_entry_cnt,
3208 			fsck->chk.valid_nat_entry_cnt);
3209 }
3210 
check_sector_size(struct f2fs_super_block * sb)3211 static int check_sector_size(struct f2fs_super_block *sb)
3212 {
3213 	u_int32_t log_sectorsize, log_sectors_per_block;
3214 
3215 	log_sectorsize = log_base_2(c.sector_size);
3216 	log_sectors_per_block = log_base_2(c.sectors_per_blk);
3217 
3218 	if (log_sectorsize == get_sb(log_sectorsize) &&
3219 			log_sectors_per_block == get_sb(log_sectors_per_block))
3220 		return 0;
3221 
3222 	set_sb(log_sectorsize, log_sectorsize);
3223 	set_sb(log_sectors_per_block, log_sectors_per_block);
3224 
3225 	update_superblock(sb, SB_MASK_ALL);
3226 	return 0;
3227 }
3228 
tune_sb_features(struct f2fs_sb_info * sbi)3229 static int tune_sb_features(struct f2fs_sb_info *sbi)
3230 {
3231 	int sb_changed = 0;
3232 	struct f2fs_super_block *sb = F2FS_RAW_SUPER(sbi);
3233 
3234 	if (!(sb->feature & cpu_to_le32(F2FS_FEATURE_ENCRYPT)) &&
3235 			c.feature & cpu_to_le32(F2FS_FEATURE_ENCRYPT)) {
3236 		sb->feature |= cpu_to_le32(F2FS_FEATURE_ENCRYPT);
3237 		MSG(0, "Info: Set Encryption feature\n");
3238 		sb_changed = 1;
3239 	}
3240 	if (!(sb->feature & cpu_to_le32(F2FS_FEATURE_CASEFOLD)) &&
3241 		c.feature & cpu_to_le32(F2FS_FEATURE_CASEFOLD)) {
3242 		if (!c.s_encoding) {
3243 			ERR_MSG("ERROR: Must specify encoding to enable casefolding.\n");
3244 			return -1;
3245 		}
3246 		sb->feature |= cpu_to_le32(F2FS_FEATURE_CASEFOLD);
3247 		MSG(0, "Info: Set Casefold feature\n");
3248 		sb_changed = 1;
3249 	}
3250 	/* TODO: quota needs to allocate inode numbers */
3251 
3252 	c.feature = sb->feature;
3253 	if (!sb_changed)
3254 		return 0;
3255 
3256 	update_superblock(sb, SB_MASK_ALL);
3257 	return 0;
3258 }
3259 
get_fsync_inode(struct list_head * head,nid_t ino)3260 static struct fsync_inode_entry *get_fsync_inode(struct list_head *head,
3261 								nid_t ino)
3262 {
3263 	struct fsync_inode_entry *entry;
3264 
3265 	list_for_each_entry(entry, head, list)
3266 		if (entry->ino == ino)
3267 			return entry;
3268 
3269 	return NULL;
3270 }
3271 
add_fsync_inode(struct list_head * head,nid_t ino)3272 static struct fsync_inode_entry *add_fsync_inode(struct list_head *head,
3273 								nid_t ino)
3274 {
3275 	struct fsync_inode_entry *entry;
3276 
3277 	entry = calloc(sizeof(struct fsync_inode_entry), 1);
3278 	if (!entry)
3279 		return NULL;
3280 	entry->ino = ino;
3281 	list_add_tail(&entry->list, head);
3282 	return entry;
3283 }
3284 
del_fsync_inode(struct fsync_inode_entry * entry)3285 static void del_fsync_inode(struct fsync_inode_entry *entry)
3286 {
3287 	list_del(&entry->list);
3288 	free(entry);
3289 }
3290 
destroy_fsync_dnodes(struct list_head * head)3291 static void destroy_fsync_dnodes(struct list_head *head)
3292 {
3293 	struct fsync_inode_entry *entry, *tmp;
3294 
3295 	list_for_each_entry_safe(entry, tmp, head, list)
3296 		del_fsync_inode(entry);
3297 }
3298 
find_fsync_inode(struct f2fs_sb_info * sbi,struct list_head * head)3299 static int find_fsync_inode(struct f2fs_sb_info *sbi, struct list_head *head)
3300 {
3301 	struct curseg_info *curseg;
3302 	struct f2fs_node *node_blk;
3303 	block_t blkaddr;
3304 	unsigned int loop_cnt = 0;
3305 	unsigned int free_blocks = TOTAL_SEGS(sbi) * sbi->blocks_per_seg -
3306 						sbi->total_valid_block_count;
3307 	int err = 0;
3308 
3309 	/* get node pages in the current segment */
3310 	curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
3311 	blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
3312 
3313 	node_blk = calloc(F2FS_BLKSIZE, 1);
3314 	ASSERT(node_blk);
3315 
3316 	while (1) {
3317 		struct fsync_inode_entry *entry;
3318 
3319 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
3320 			break;
3321 
3322 		err = dev_read_block(node_blk, blkaddr);
3323 		if (err)
3324 			break;
3325 
3326 		if (!is_recoverable_dnode(sbi, node_blk))
3327 			break;
3328 
3329 		if (!is_fsync_dnode(node_blk))
3330 			goto next;
3331 
3332 		entry = get_fsync_inode(head, ino_of_node(node_blk));
3333 		if (!entry) {
3334 			entry = add_fsync_inode(head, ino_of_node(node_blk));
3335 			if (!entry) {
3336 				err = -1;
3337 				break;
3338 			}
3339 		}
3340 		entry->blkaddr = blkaddr;
3341 
3342 		if (IS_INODE(node_blk) && is_dent_dnode(node_blk))
3343 			entry->last_dentry = blkaddr;
3344 next:
3345 		/* sanity check in order to detect looped node chain */
3346 		if (++loop_cnt >= free_blocks ||
3347 			blkaddr == next_blkaddr_of_node(node_blk)) {
3348 			MSG(0, "\tdetect looped node chain, blkaddr:%u, next:%u\n",
3349 				    blkaddr,
3350 				    next_blkaddr_of_node(node_blk));
3351 			err = -1;
3352 			break;
3353 		}
3354 
3355 		blkaddr = next_blkaddr_of_node(node_blk);
3356 	}
3357 
3358 	free(node_blk);
3359 	return err;
3360 }
3361 
do_record_fsync_data(struct f2fs_sb_info * sbi,struct f2fs_node * node_blk,block_t blkaddr)3362 static int do_record_fsync_data(struct f2fs_sb_info *sbi,
3363 					struct f2fs_node *node_blk,
3364 					block_t blkaddr)
3365 {
3366 	unsigned int segno, offset;
3367 	struct seg_entry *se;
3368 	unsigned int ofs_in_node = 0;
3369 	unsigned int start, end;
3370 	int err = 0, recorded = 0;
3371 
3372 	segno = GET_SEGNO(sbi, blkaddr);
3373 	se = get_seg_entry(sbi, segno);
3374 	offset = OFFSET_IN_SEG(sbi, blkaddr);
3375 
3376 	if (f2fs_test_bit(offset, (char *)se->cur_valid_map)) {
3377 		ASSERT(0);
3378 		return -1;
3379 	}
3380 	if (f2fs_test_bit(offset, (char *)se->ckpt_valid_map)) {
3381 		ASSERT(0);
3382 		return -1;
3383 	}
3384 
3385 	if (!se->ckpt_valid_blocks)
3386 		se->ckpt_type = CURSEG_WARM_NODE;
3387 
3388 	se->ckpt_valid_blocks++;
3389 	f2fs_set_bit(offset, (char *)se->ckpt_valid_map);
3390 
3391 	MSG(1, "do_record_fsync_data: [node] ino = %u, nid = %u, blkaddr = %u\n",
3392 	    ino_of_node(node_blk), ofs_of_node(node_blk), blkaddr);
3393 
3394 	/* inline data */
3395 	if (IS_INODE(node_blk) && (node_blk->i.i_inline & F2FS_INLINE_DATA))
3396 		return 0;
3397 	/* xattr node */
3398 	if (ofs_of_node(node_blk) == XATTR_NODE_OFFSET)
3399 		return 0;
3400 
3401 	/* step 3: recover data indices */
3402 	start = start_bidx_of_node(ofs_of_node(node_blk), node_blk);
3403 	end = start + ADDRS_PER_PAGE(sbi, node_blk, NULL);
3404 
3405 	for (; start < end; start++, ofs_in_node++) {
3406 		blkaddr = datablock_addr(node_blk, ofs_in_node);
3407 
3408 		if (!is_valid_data_blkaddr(blkaddr))
3409 			continue;
3410 
3411 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR)) {
3412 			err = -1;
3413 			goto out;
3414 		}
3415 
3416 		segno = GET_SEGNO(sbi, blkaddr);
3417 		se = get_seg_entry(sbi, segno);
3418 		offset = OFFSET_IN_SEG(sbi, blkaddr);
3419 
3420 		if (f2fs_test_bit(offset, (char *)se->cur_valid_map))
3421 			continue;
3422 		if (f2fs_test_bit(offset, (char *)se->ckpt_valid_map))
3423 			continue;
3424 
3425 		if (!se->ckpt_valid_blocks)
3426 			se->ckpt_type = CURSEG_WARM_DATA;
3427 
3428 		se->ckpt_valid_blocks++;
3429 		f2fs_set_bit(offset, (char *)se->ckpt_valid_map);
3430 
3431 		MSG(1, "do_record_fsync_data: [data] ino = %u, nid = %u, blkaddr = %u\n",
3432 		    ino_of_node(node_blk), ofs_of_node(node_blk), blkaddr);
3433 
3434 		recorded++;
3435 	}
3436 out:
3437 	MSG(1, "recover_data: ino = %u, nid = %u, recorded = %d, err = %d\n",
3438 		    ino_of_node(node_blk), ofs_of_node(node_blk),
3439 		    recorded, err);
3440 	return err;
3441 }
3442 
traverse_dnodes(struct f2fs_sb_info * sbi,struct list_head * inode_list)3443 static int traverse_dnodes(struct f2fs_sb_info *sbi,
3444 				struct list_head *inode_list)
3445 {
3446 	struct curseg_info *curseg;
3447 	struct f2fs_node *node_blk;
3448 	block_t blkaddr;
3449 	int err = 0;
3450 
3451 	/* get node pages in the current segment */
3452 	curseg = CURSEG_I(sbi, CURSEG_WARM_NODE);
3453 	blkaddr = NEXT_FREE_BLKADDR(sbi, curseg);
3454 
3455 	node_blk = calloc(F2FS_BLKSIZE, 1);
3456 	ASSERT(node_blk);
3457 
3458 	while (1) {
3459 		struct fsync_inode_entry *entry;
3460 
3461 		if (!f2fs_is_valid_blkaddr(sbi, blkaddr, META_POR))
3462 			break;
3463 
3464 		err = dev_read_block(node_blk, blkaddr);
3465 		if (err)
3466 			break;
3467 
3468 		if (!is_recoverable_dnode(sbi, node_blk))
3469 			break;
3470 
3471 		entry = get_fsync_inode(inode_list,
3472 					ino_of_node(node_blk));
3473 		if (!entry)
3474 			goto next;
3475 
3476 		err = do_record_fsync_data(sbi, node_blk, blkaddr);
3477 		if (err)
3478 			break;
3479 
3480 		if (entry->blkaddr == blkaddr)
3481 			del_fsync_inode(entry);
3482 next:
3483 		blkaddr = next_blkaddr_of_node(node_blk);
3484 	}
3485 
3486 	free(node_blk);
3487 	return err;
3488 }
3489 
record_fsync_data(struct f2fs_sb_info * sbi)3490 static int record_fsync_data(struct f2fs_sb_info *sbi)
3491 {
3492 	struct list_head inode_list = LIST_HEAD_INIT(inode_list);
3493 	int ret;
3494 
3495 	if (!need_fsync_data_record(sbi))
3496 		return 0;
3497 
3498 	ret = find_fsync_inode(sbi, &inode_list);
3499 	if (ret)
3500 		goto out;
3501 
3502 	ret = late_build_segment_manager(sbi);
3503 	if (ret < 0) {
3504 		ERR_MSG("late_build_segment_manager failed\n");
3505 		goto out;
3506 	}
3507 
3508 	ret = traverse_dnodes(sbi, &inode_list);
3509 out:
3510 	destroy_fsync_dnodes(&inode_list);
3511 	return ret;
3512 }
3513 
f2fs_do_mount(struct f2fs_sb_info * sbi)3514 int f2fs_do_mount(struct f2fs_sb_info *sbi)
3515 {
3516 	struct f2fs_checkpoint *cp = NULL;
3517 	struct f2fs_super_block *sb = NULL;
3518 	int ret;
3519 
3520 	sbi->active_logs = NR_CURSEG_TYPE;
3521 	ret = validate_super_block(sbi, SB0_ADDR);
3522 	if (ret) {
3523 		ret = validate_super_block(sbi, SB1_ADDR);
3524 		if (ret)
3525 			return -1;
3526 	}
3527 	sb = F2FS_RAW_SUPER(sbi);
3528 
3529 	ret = check_sector_size(sb);
3530 	if (ret)
3531 		return -1;
3532 
3533 	print_raw_sb_info(sb);
3534 
3535 	init_sb_info(sbi);
3536 
3537 	ret = get_valid_checkpoint(sbi);
3538 	if (ret) {
3539 		ERR_MSG("Can't find valid checkpoint\n");
3540 		return -1;
3541 	}
3542 
3543 	c.bug_on = 0;
3544 
3545 	if (sanity_check_ckpt(sbi)) {
3546 		ERR_MSG("Checkpoint is polluted\n");
3547 		return -1;
3548 	}
3549 	cp = F2FS_CKPT(sbi);
3550 
3551 	if (c.func != FSCK && c.func != DUMP &&
3552 		!is_set_ckpt_flags(F2FS_CKPT(sbi), CP_UMOUNT_FLAG)) {
3553 		ERR_MSG("Mount unclean image to replay log first\n");
3554 		return -1;
3555 	}
3556 
3557 	print_ckpt_info(sbi);
3558 
3559 	if (c.quota_fix) {
3560 		if (get_cp(ckpt_flags) & CP_QUOTA_NEED_FSCK_FLAG)
3561 			c.fix_on = 1;
3562 	}
3563 	if (c.layout)
3564 		return 1;
3565 
3566 	if (tune_sb_features(sbi))
3567 		return -1;
3568 
3569 	/* precompute checksum seed for metadata */
3570 	if (c.feature & cpu_to_le32(F2FS_FEATURE_INODE_CHKSUM))
3571 		c.chksum_seed = f2fs_cal_crc32(~0, sb->uuid, sizeof(sb->uuid));
3572 
3573 	sbi->total_valid_node_count = get_cp(valid_node_count);
3574 	sbi->total_valid_inode_count = get_cp(valid_inode_count);
3575 	sbi->user_block_count = get_cp(user_block_count);
3576 	sbi->total_valid_block_count = get_cp(valid_block_count);
3577 	sbi->last_valid_block_count = sbi->total_valid_block_count;
3578 	sbi->alloc_valid_block_count = 0;
3579 
3580 	if (early_build_segment_manager(sbi)) {
3581 		ERR_MSG("early_build_segment_manager failed\n");
3582 		return -1;
3583 	}
3584 
3585 	if (build_node_manager(sbi)) {
3586 		ERR_MSG("build_node_manager failed\n");
3587 		return -1;
3588 	}
3589 
3590 	if (record_fsync_data(sbi)) {
3591 		ERR_MSG("record_fsync_data failed\n");
3592 		return -1;
3593 	}
3594 
3595 	if (!f2fs_should_proceed(sb, get_cp(ckpt_flags)))
3596 		return 1;
3597 
3598 	if (late_build_segment_manager(sbi) < 0) {
3599 		ERR_MSG("late_build_segment_manager failed\n");
3600 		return -1;
3601 	}
3602 
3603 	if (f2fs_late_init_nid_bitmap(sbi)) {
3604 		ERR_MSG("f2fs_late_init_nid_bitmap failed\n");
3605 		return -1;
3606 	}
3607 
3608 	/* Check nat_bits */
3609 	if (c.func == FSCK && is_set_ckpt_flags(cp, CP_NAT_BITS_FLAG)) {
3610 		if (check_nat_bits(sbi, sb, cp) && c.fix_on)
3611 			write_nat_bits(sbi, sb, cp, sbi->cur_cp);
3612 	}
3613 	return 0;
3614 }
3615 
f2fs_do_umount(struct f2fs_sb_info * sbi)3616 void f2fs_do_umount(struct f2fs_sb_info *sbi)
3617 {
3618 	struct sit_info *sit_i = SIT_I(sbi);
3619 	struct f2fs_sm_info *sm_i = SM_I(sbi);
3620 	struct f2fs_nm_info *nm_i = NM_I(sbi);
3621 	unsigned int i;
3622 
3623 	/* free nm_info */
3624 	if (c.func == SLOAD || c.func == FSCK)
3625 		free(nm_i->nid_bitmap);
3626 	free(nm_i->nat_bitmap);
3627 	free(sbi->nm_info);
3628 
3629 	/* free sit_info */
3630 	free(sit_i->bitmap);
3631 	free(sit_i->sit_bitmap);
3632 	free(sit_i->sentries);
3633 	free(sm_i->sit_info);
3634 
3635 	/* free sm_info */
3636 	for (i = 0; i < NR_CURSEG_TYPE; i++)
3637 		free(sm_i->curseg_array[i].sum_blk);
3638 
3639 	free(sm_i->curseg_array);
3640 	free(sbi->sm_info);
3641 
3642 	free(sbi->ckpt);
3643 	free(sbi->raw_super);
3644 }
3645 
3646 #ifdef WITH_ANDROID
f2fs_sparse_initialize_meta(struct f2fs_sb_info * sbi)3647 int f2fs_sparse_initialize_meta(struct f2fs_sb_info *sbi)
3648 {
3649 	struct f2fs_super_block *sb = sbi->raw_super;
3650 	u_int32_t sit_seg_count, sit_size;
3651 	u_int32_t nat_seg_count, nat_size;
3652 	u_int64_t sit_seg_addr, nat_seg_addr, payload_addr;
3653 	u_int32_t seg_size = 1 << get_sb(log_blocks_per_seg);
3654 	int ret;
3655 
3656 	if (!c.sparse_mode)
3657 		return 0;
3658 
3659 	sit_seg_addr = get_sb(sit_blkaddr);
3660 	sit_seg_count = get_sb(segment_count_sit);
3661 	sit_size = sit_seg_count * seg_size;
3662 
3663 	DBG(1, "\tSparse: filling sit area at block offset: 0x%08"PRIx64" len: %u\n",
3664 							sit_seg_addr, sit_size);
3665 	ret = dev_fill(NULL, sit_seg_addr * F2FS_BLKSIZE,
3666 					sit_size * F2FS_BLKSIZE);
3667 	if (ret) {
3668 		MSG(1, "\tError: While zeroing out the sit area "
3669 				"on disk!!!\n");
3670 		return -1;
3671 	}
3672 
3673 	nat_seg_addr = get_sb(nat_blkaddr);
3674 	nat_seg_count = get_sb(segment_count_nat);
3675 	nat_size = nat_seg_count * seg_size;
3676 
3677 	DBG(1, "\tSparse: filling nat area at block offset 0x%08"PRIx64" len: %u\n",
3678 							nat_seg_addr, nat_size);
3679 	ret = dev_fill(NULL, nat_seg_addr * F2FS_BLKSIZE,
3680 					nat_size * F2FS_BLKSIZE);
3681 	if (ret) {
3682 		MSG(1, "\tError: While zeroing out the nat area "
3683 				"on disk!!!\n");
3684 		return -1;
3685 	}
3686 
3687 	payload_addr = get_sb(segment0_blkaddr) + 1;
3688 
3689 	DBG(1, "\tSparse: filling bitmap area at block offset 0x%08"PRIx64" len: %u\n",
3690 					payload_addr, get_sb(cp_payload));
3691 	ret = dev_fill(NULL, payload_addr * F2FS_BLKSIZE,
3692 					get_sb(cp_payload) * F2FS_BLKSIZE);
3693 	if (ret) {
3694 		MSG(1, "\tError: While zeroing out the nat/sit bitmap area "
3695 				"on disk!!!\n");
3696 		return -1;
3697 	}
3698 
3699 	payload_addr += seg_size;
3700 
3701 	DBG(1, "\tSparse: filling bitmap area at block offset 0x%08"PRIx64" len: %u\n",
3702 					payload_addr, get_sb(cp_payload));
3703 	ret = dev_fill(NULL, payload_addr * F2FS_BLKSIZE,
3704 					get_sb(cp_payload) * F2FS_BLKSIZE);
3705 	if (ret) {
3706 		MSG(1, "\tError: While zeroing out the nat/sit bitmap area "
3707 				"on disk!!!\n");
3708 		return -1;
3709 	}
3710 	return 0;
3711 }
3712 #else
f2fs_sparse_initialize_meta(struct f2fs_sb_info * sbi)3713 int f2fs_sparse_initialize_meta(struct f2fs_sb_info *sbi) { return 0; }
3714 #endif
3715