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