1 // SPDX-License-Identifier: GPL-2.0 1 // SPDX-License-Identifier: GPL-2.0
2 /* 2 /*
3 * linux/fs/ext4/namei.c 3 * linux/fs/ext4/namei.c
4 * 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995 5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr) 6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal 7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI) 8 * Universite Pierre et Marie Curie (Paris VI)
9 * 9 *
10 * from 10 * from
11 * 11 *
12 * linux/fs/minix/namei.c 12 * linux/fs/minix/namei.c
13 * 13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds 14 * Copyright (C) 1991, 1992 Linus Torvalds
15 * 15 *
16 * Big-endian to little-endian byte-swapping/ 16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers. 17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 * Directory entry file type support and forw 18 * Directory entry file type support and forward compatibility hooks
19 * for B-tree directories by Theodore Ts' 19 * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998
20 * Hash Tree Directory indexing (c) 20 * Hash Tree Directory indexing (c)
21 * Daniel Phillips, 2001 21 * Daniel Phillips, 2001
22 * Hash Tree Directory indexing porting 22 * Hash Tree Directory indexing porting
23 * Christopher Li, 2002 23 * Christopher Li, 2002
24 * Hash Tree Directory indexing cleanup 24 * Hash Tree Directory indexing cleanup
25 * Theodore Ts'o, 2002 25 * Theodore Ts'o, 2002
26 */ 26 */
27 27
28 #include <linux/fs.h> 28 #include <linux/fs.h>
29 #include <linux/pagemap.h> 29 #include <linux/pagemap.h>
30 #include <linux/time.h> 30 #include <linux/time.h>
31 #include <linux/fcntl.h> 31 #include <linux/fcntl.h>
32 #include <linux/stat.h> 32 #include <linux/stat.h>
33 #include <linux/string.h> 33 #include <linux/string.h>
34 #include <linux/quotaops.h> 34 #include <linux/quotaops.h>
35 #include <linux/buffer_head.h> 35 #include <linux/buffer_head.h>
36 #include <linux/bio.h> 36 #include <linux/bio.h>
37 #include <linux/iversion.h> 37 #include <linux/iversion.h>
38 #include <linux/unicode.h> 38 #include <linux/unicode.h>
39 #include "ext4.h" 39 #include "ext4.h"
40 #include "ext4_jbd2.h" 40 #include "ext4_jbd2.h"
41 41
42 #include "xattr.h" 42 #include "xattr.h"
43 #include "acl.h" 43 #include "acl.h"
44 44
45 #include <trace/events/ext4.h> 45 #include <trace/events/ext4.h>
46 /* 46 /*
47 * define how far ahead to read directories wh 47 * define how far ahead to read directories while searching them.
48 */ 48 */
49 #define NAMEI_RA_CHUNKS 2 49 #define NAMEI_RA_CHUNKS 2
50 #define NAMEI_RA_BLOCKS 4 50 #define NAMEI_RA_BLOCKS 4
51 #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS 51 #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS)
52 52
53 static struct buffer_head *ext4_append(handle_ 53 static struct buffer_head *ext4_append(handle_t *handle,
54 struct 54 struct inode *inode,
55 ext4_l 55 ext4_lblk_t *block)
56 { 56 {
57 struct ext4_map_blocks map; 57 struct ext4_map_blocks map;
58 struct buffer_head *bh; 58 struct buffer_head *bh;
59 int err; 59 int err;
60 60
61 if (unlikely(EXT4_SB(inode->i_sb)->s_m 61 if (unlikely(EXT4_SB(inode->i_sb)->s_max_dir_size_kb &&
62 ((inode->i_size >> 10) >= 62 ((inode->i_size >> 10) >=
63 EXT4_SB(inode->i_sb)->s_ 63 EXT4_SB(inode->i_sb)->s_max_dir_size_kb)))
64 return ERR_PTR(-ENOSPC); 64 return ERR_PTR(-ENOSPC);
65 65
66 *block = inode->i_size >> inode->i_sb- 66 *block = inode->i_size >> inode->i_sb->s_blocksize_bits;
67 map.m_lblk = *block; 67 map.m_lblk = *block;
68 map.m_len = 1; 68 map.m_len = 1;
69 69
70 /* 70 /*
71 * We're appending new directory block 71 * We're appending new directory block. Make sure the block is not
72 * allocated yet, otherwise we will en 72 * allocated yet, otherwise we will end up corrupting the
73 * directory. 73 * directory.
74 */ 74 */
75 err = ext4_map_blocks(NULL, inode, &ma 75 err = ext4_map_blocks(NULL, inode, &map, 0);
76 if (err < 0) 76 if (err < 0)
77 return ERR_PTR(err); 77 return ERR_PTR(err);
78 if (err) { 78 if (err) {
79 EXT4_ERROR_INODE(inode, "Logic 79 EXT4_ERROR_INODE(inode, "Logical block already allocated");
80 return ERR_PTR(-EFSCORRUPTED); 80 return ERR_PTR(-EFSCORRUPTED);
81 } 81 }
82 82
83 bh = ext4_bread(handle, inode, *block, 83 bh = ext4_bread(handle, inode, *block, EXT4_GET_BLOCKS_CREATE);
84 if (IS_ERR(bh)) 84 if (IS_ERR(bh))
85 return bh; 85 return bh;
86 inode->i_size += inode->i_sb->s_blocks 86 inode->i_size += inode->i_sb->s_blocksize;
87 EXT4_I(inode)->i_disksize = inode->i_s 87 EXT4_I(inode)->i_disksize = inode->i_size;
88 err = ext4_mark_inode_dirty(handle, in 88 err = ext4_mark_inode_dirty(handle, inode);
89 if (err) 89 if (err)
90 goto out; 90 goto out;
91 BUFFER_TRACE(bh, "get_write_access"); 91 BUFFER_TRACE(bh, "get_write_access");
92 err = ext4_journal_get_write_access(ha 92 err = ext4_journal_get_write_access(handle, inode->i_sb, bh,
93 EX 93 EXT4_JTR_NONE);
94 if (err) 94 if (err)
95 goto out; 95 goto out;
96 return bh; 96 return bh;
97 97
98 out: 98 out:
99 brelse(bh); 99 brelse(bh);
100 ext4_std_error(inode->i_sb, err); 100 ext4_std_error(inode->i_sb, err);
101 return ERR_PTR(err); 101 return ERR_PTR(err);
102 } 102 }
103 103
104 static int ext4_dx_csum_verify(struct inode *i 104 static int ext4_dx_csum_verify(struct inode *inode,
105 struct ext4_dir 105 struct ext4_dir_entry *dirent);
106 106
107 /* 107 /*
108 * Hints to ext4_read_dirblock regarding wheth 108 * Hints to ext4_read_dirblock regarding whether we expect a directory
109 * block being read to be an index block, or a 109 * block being read to be an index block, or a block containing
110 * directory entries (and if the latter, wheth 110 * directory entries (and if the latter, whether it was found via a
111 * logical block in an htree index block). Th 111 * logical block in an htree index block). This is used to control
112 * what sort of sanity checkinig ext4_read_dir 112 * what sort of sanity checkinig ext4_read_dirblock() will do on the
113 * directory block read from the storage devic 113 * directory block read from the storage device. EITHER will means
114 * the caller doesn't know what kind of direct 114 * the caller doesn't know what kind of directory block will be read,
115 * so no specific verification will be done. 115 * so no specific verification will be done.
116 */ 116 */
117 typedef enum { 117 typedef enum {
118 EITHER, INDEX, DIRENT, DIRENT_HTREE 118 EITHER, INDEX, DIRENT, DIRENT_HTREE
119 } dirblock_type_t; 119 } dirblock_type_t;
120 120
121 #define ext4_read_dirblock(inode, block, type) 121 #define ext4_read_dirblock(inode, block, type) \
122 __ext4_read_dirblock((inode), (block), 122 __ext4_read_dirblock((inode), (block), (type), __func__, __LINE__)
123 123
124 static struct buffer_head *__ext4_read_dirbloc 124 static struct buffer_head *__ext4_read_dirblock(struct inode *inode,
125 125 ext4_lblk_t block,
126 126 dirblock_type_t type,
127 127 const char *func,
128 128 unsigned int line)
129 { 129 {
130 struct buffer_head *bh; 130 struct buffer_head *bh;
131 struct ext4_dir_entry *dirent; 131 struct ext4_dir_entry *dirent;
132 int is_dx_block = 0; 132 int is_dx_block = 0;
133 133
134 if (block >= inode->i_size >> inode->i 134 if (block >= inode->i_size >> inode->i_blkbits) {
135 ext4_error_inode(inode, func, 135 ext4_error_inode(inode, func, line, block,
136 "Attempting to read dir 136 "Attempting to read directory block (%u) that is past i_size (%llu)",
137 block, inode->i_size); 137 block, inode->i_size);
138 return ERR_PTR(-EFSCORRUPTED); 138 return ERR_PTR(-EFSCORRUPTED);
139 } 139 }
140 140
141 if (ext4_simulate_fail(inode->i_sb, EX 141 if (ext4_simulate_fail(inode->i_sb, EXT4_SIM_DIRBLOCK_EIO))
142 bh = ERR_PTR(-EIO); 142 bh = ERR_PTR(-EIO);
143 else 143 else
144 bh = ext4_bread(NULL, inode, b 144 bh = ext4_bread(NULL, inode, block, 0);
145 if (IS_ERR(bh)) { 145 if (IS_ERR(bh)) {
146 __ext4_warning(inode->i_sb, fu 146 __ext4_warning(inode->i_sb, func, line,
147 "inode #%lu: lb 147 "inode #%lu: lblock %lu: comm %s: "
148 "error %ld read 148 "error %ld reading directory block",
149 inode->i_ino, ( 149 inode->i_ino, (unsigned long)block,
150 current->comm, 150 current->comm, PTR_ERR(bh));
151 151
152 return bh; 152 return bh;
153 } 153 }
154 /* The first directory block must not 154 /* The first directory block must not be a hole. */
155 if (!bh && (type == INDEX || type == D 155 if (!bh && (type == INDEX || type == DIRENT_HTREE || block == 0)) {
156 ext4_error_inode(inode, func, 156 ext4_error_inode(inode, func, line, block,
157 "Directory ho 157 "Directory hole found for htree %s block %u",
158 (type == INDE 158 (type == INDEX) ? "index" : "leaf", block);
159 return ERR_PTR(-EFSCORRUPTED); 159 return ERR_PTR(-EFSCORRUPTED);
160 } 160 }
161 if (!bh) 161 if (!bh)
162 return NULL; 162 return NULL;
163 dirent = (struct ext4_dir_entry *) bh- 163 dirent = (struct ext4_dir_entry *) bh->b_data;
164 /* Determine whether or not we have an 164 /* Determine whether or not we have an index block */
165 if (is_dx(inode)) { 165 if (is_dx(inode)) {
166 if (block == 0) 166 if (block == 0)
167 is_dx_block = 1; 167 is_dx_block = 1;
168 else if (ext4_rec_len_from_dis 168 else if (ext4_rec_len_from_disk(dirent->rec_len,
169 169 inode->i_sb->s_blocksize) ==
170 inode->i_sb->s_blocks 170 inode->i_sb->s_blocksize)
171 is_dx_block = 1; 171 is_dx_block = 1;
172 } 172 }
173 if (!is_dx_block && type == INDEX) { 173 if (!is_dx_block && type == INDEX) {
174 ext4_error_inode(inode, func, 174 ext4_error_inode(inode, func, line, block,
175 "directory leaf block f 175 "directory leaf block found instead of index block");
176 brelse(bh); 176 brelse(bh);
177 return ERR_PTR(-EFSCORRUPTED); 177 return ERR_PTR(-EFSCORRUPTED);
178 } 178 }
179 if (!ext4_has_metadata_csum(inode->i_s 179 if (!ext4_has_metadata_csum(inode->i_sb) ||
180 buffer_verified(bh)) 180 buffer_verified(bh))
181 return bh; 181 return bh;
182 182
183 /* 183 /*
184 * An empty leaf block can get mistake 184 * An empty leaf block can get mistaken for a index block; for
185 * this reason, we can only check the 185 * this reason, we can only check the index checksum when the
186 * caller is sure it should be an inde 186 * caller is sure it should be an index block.
187 */ 187 */
188 if (is_dx_block && type == INDEX) { 188 if (is_dx_block && type == INDEX) {
189 if (ext4_dx_csum_verify(inode, 189 if (ext4_dx_csum_verify(inode, dirent) &&
190 !ext4_simulate_fail(inode- 190 !ext4_simulate_fail(inode->i_sb, EXT4_SIM_DIRBLOCK_CRC))
191 set_buffer_verified(bh 191 set_buffer_verified(bh);
192 else { 192 else {
193 ext4_error_inode_err(i 193 ext4_error_inode_err(inode, func, line, block,
194 E 194 EFSBADCRC,
195 " 195 "Directory index failed checksum");
196 brelse(bh); 196 brelse(bh);
197 return ERR_PTR(-EFSBAD 197 return ERR_PTR(-EFSBADCRC);
198 } 198 }
199 } 199 }
200 if (!is_dx_block) { 200 if (!is_dx_block) {
201 if (ext4_dirblock_csum_verify( 201 if (ext4_dirblock_csum_verify(inode, bh) &&
202 !ext4_simulate_fail(inode- 202 !ext4_simulate_fail(inode->i_sb, EXT4_SIM_DIRBLOCK_CRC))
203 set_buffer_verified(bh 203 set_buffer_verified(bh);
204 else { 204 else {
205 ext4_error_inode_err(i 205 ext4_error_inode_err(inode, func, line, block,
206 E 206 EFSBADCRC,
207 " 207 "Directory block failed checksum");
208 brelse(bh); 208 brelse(bh);
209 return ERR_PTR(-EFSBAD 209 return ERR_PTR(-EFSBADCRC);
210 } 210 }
211 } 211 }
212 return bh; 212 return bh;
213 } 213 }
214 214
215 #ifdef DX_DEBUG 215 #ifdef DX_DEBUG
216 #define dxtrace(command) command 216 #define dxtrace(command) command
217 #else 217 #else
218 #define dxtrace(command) 218 #define dxtrace(command)
219 #endif 219 #endif
220 220
221 struct fake_dirent 221 struct fake_dirent
222 { 222 {
223 __le32 inode; 223 __le32 inode;
224 __le16 rec_len; 224 __le16 rec_len;
225 u8 name_len; 225 u8 name_len;
226 u8 file_type; 226 u8 file_type;
227 }; 227 };
228 228
229 struct dx_countlimit 229 struct dx_countlimit
230 { 230 {
231 __le16 limit; 231 __le16 limit;
232 __le16 count; 232 __le16 count;
233 }; 233 };
234 234
235 struct dx_entry 235 struct dx_entry
236 { 236 {
237 __le32 hash; 237 __le32 hash;
238 __le32 block; 238 __le32 block;
239 }; 239 };
240 240
241 /* 241 /*
242 * dx_root_info is laid out so that if it shou 242 * dx_root_info is laid out so that if it should somehow get overlaid by a
243 * dirent the two low bits of the hash version 243 * dirent the two low bits of the hash version will be zero. Therefore, the
244 * hash version mod 4 should never be 0. Sinc 244 * hash version mod 4 should never be 0. Sincerely, the paranoia department.
245 */ 245 */
246 246
247 struct dx_root 247 struct dx_root
248 { 248 {
249 struct fake_dirent dot; 249 struct fake_dirent dot;
250 char dot_name[4]; 250 char dot_name[4];
251 struct fake_dirent dotdot; 251 struct fake_dirent dotdot;
252 char dotdot_name[4]; 252 char dotdot_name[4];
253 struct dx_root_info 253 struct dx_root_info
254 { 254 {
255 __le32 reserved_zero; 255 __le32 reserved_zero;
256 u8 hash_version; 256 u8 hash_version;
257 u8 info_length; /* 8 */ 257 u8 info_length; /* 8 */
258 u8 indirect_levels; 258 u8 indirect_levels;
259 u8 unused_flags; 259 u8 unused_flags;
260 } 260 }
261 info; 261 info;
262 struct dx_entry entries[]; 262 struct dx_entry entries[];
263 }; 263 };
264 264
265 struct dx_node 265 struct dx_node
266 { 266 {
267 struct fake_dirent fake; 267 struct fake_dirent fake;
268 struct dx_entry entries[]; 268 struct dx_entry entries[];
269 }; 269 };
270 270
271 271
272 struct dx_frame 272 struct dx_frame
273 { 273 {
274 struct buffer_head *bh; 274 struct buffer_head *bh;
275 struct dx_entry *entries; 275 struct dx_entry *entries;
276 struct dx_entry *at; 276 struct dx_entry *at;
277 }; 277 };
278 278
279 struct dx_map_entry 279 struct dx_map_entry
280 { 280 {
281 u32 hash; 281 u32 hash;
282 u16 offs; 282 u16 offs;
283 u16 size; 283 u16 size;
284 }; 284 };
285 285
286 /* 286 /*
287 * This goes at the end of each htree block. 287 * This goes at the end of each htree block.
288 */ 288 */
289 struct dx_tail { 289 struct dx_tail {
290 u32 dt_reserved; 290 u32 dt_reserved;
291 __le32 dt_checksum; /* crc32c(uuid 291 __le32 dt_checksum; /* crc32c(uuid+inum+dirblock) */
292 }; 292 };
293 293
294 static inline ext4_lblk_t dx_get_block(struct 294 static inline ext4_lblk_t dx_get_block(struct dx_entry *entry);
295 static void dx_set_block(struct dx_entry *entr 295 static void dx_set_block(struct dx_entry *entry, ext4_lblk_t value);
296 static inline unsigned dx_get_hash(struct dx_e 296 static inline unsigned dx_get_hash(struct dx_entry *entry);
297 static void dx_set_hash(struct dx_entry *entry 297 static void dx_set_hash(struct dx_entry *entry, unsigned value);
298 static unsigned dx_get_count(struct dx_entry * 298 static unsigned dx_get_count(struct dx_entry *entries);
299 static unsigned dx_get_limit(struct dx_entry * 299 static unsigned dx_get_limit(struct dx_entry *entries);
300 static void dx_set_count(struct dx_entry *entr 300 static void dx_set_count(struct dx_entry *entries, unsigned value);
301 static void dx_set_limit(struct dx_entry *entr 301 static void dx_set_limit(struct dx_entry *entries, unsigned value);
302 static unsigned dx_root_limit(struct inode *di 302 static unsigned dx_root_limit(struct inode *dir, unsigned infosize);
303 static unsigned dx_node_limit(struct inode *di 303 static unsigned dx_node_limit(struct inode *dir);
304 static struct dx_frame *dx_probe(struct ext4_f 304 static struct dx_frame *dx_probe(struct ext4_filename *fname,
305 struct inode 305 struct inode *dir,
306 struct dx_has 306 struct dx_hash_info *hinfo,
307 struct dx_fra 307 struct dx_frame *frame);
308 static void dx_release(struct dx_frame *frames 308 static void dx_release(struct dx_frame *frames);
309 static int dx_make_map(struct inode *dir, stru 309 static int dx_make_map(struct inode *dir, struct buffer_head *bh,
310 struct dx_hash_info *hi 310 struct dx_hash_info *hinfo,
311 struct dx_map_entry *ma 311 struct dx_map_entry *map_tail);
312 static void dx_sort_map(struct dx_map_entry *m 312 static void dx_sort_map(struct dx_map_entry *map, unsigned count);
313 static struct ext4_dir_entry_2 *dx_move_dirent 313 static struct ext4_dir_entry_2 *dx_move_dirents(struct inode *dir, char *from,
314 char * 314 char *to, struct dx_map_entry *offsets,
315 int co 315 int count, unsigned int blocksize);
316 static struct ext4_dir_entry_2 *dx_pack_dirent 316 static struct ext4_dir_entry_2 *dx_pack_dirents(struct inode *dir, char *base,
317 317 unsigned int blocksize);
318 static void dx_insert_block(struct dx_frame *f 318 static void dx_insert_block(struct dx_frame *frame,
319 u32 ha 319 u32 hash, ext4_lblk_t block);
320 static int ext4_htree_next_block(struct inode 320 static int ext4_htree_next_block(struct inode *dir, __u32 hash,
321 struct dx_fra 321 struct dx_frame *frame,
322 struct dx_fra 322 struct dx_frame *frames,
323 __u32 *start_ 323 __u32 *start_hash);
324 static struct buffer_head * ext4_dx_find_entry 324 static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
325 struct ext4_filename *fname, 325 struct ext4_filename *fname,
326 struct ext4_dir_entry_2 **res_ 326 struct ext4_dir_entry_2 **res_dir);
327 static int ext4_dx_add_entry(handle_t *handle, 327 static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname,
328 struct inode *dir 328 struct inode *dir, struct inode *inode);
329 329
330 /* checksumming functions */ 330 /* checksumming functions */
331 void ext4_initialize_dirent_tail(struct buffer 331 void ext4_initialize_dirent_tail(struct buffer_head *bh,
332 unsigned int 332 unsigned int blocksize)
333 { 333 {
334 struct ext4_dir_entry_tail *t = EXT4_D 334 struct ext4_dir_entry_tail *t = EXT4_DIRENT_TAIL(bh->b_data, blocksize);
335 335
336 memset(t, 0, sizeof(struct ext4_dir_en 336 memset(t, 0, sizeof(struct ext4_dir_entry_tail));
337 t->det_rec_len = ext4_rec_len_to_disk( 337 t->det_rec_len = ext4_rec_len_to_disk(
338 sizeof(struct ext4_dir 338 sizeof(struct ext4_dir_entry_tail), blocksize);
339 t->det_reserved_ft = EXT4_FT_DIR_CSUM; 339 t->det_reserved_ft = EXT4_FT_DIR_CSUM;
340 } 340 }
341 341
342 /* Walk through a dirent block to find a check 342 /* Walk through a dirent block to find a checksum "dirent" at the tail */
343 static struct ext4_dir_entry_tail *get_dirent_ 343 static struct ext4_dir_entry_tail *get_dirent_tail(struct inode *inode,
344 344 struct buffer_head *bh)
345 { 345 {
346 struct ext4_dir_entry_tail *t; 346 struct ext4_dir_entry_tail *t;
347 int blocksize = EXT4_BLOCK_SIZE(inode- 347 int blocksize = EXT4_BLOCK_SIZE(inode->i_sb);
348 348
349 #ifdef PARANOID 349 #ifdef PARANOID
350 struct ext4_dir_entry *d, *top; 350 struct ext4_dir_entry *d, *top;
351 351
352 d = (struct ext4_dir_entry *)bh->b_dat 352 d = (struct ext4_dir_entry *)bh->b_data;
353 top = (struct ext4_dir_entry *)(bh->b_ 353 top = (struct ext4_dir_entry *)(bh->b_data +
354 (blocksize - sizeof(struct ext 354 (blocksize - sizeof(struct ext4_dir_entry_tail)));
355 while (d < top && ext4_rec_len_from_di 355 while (d < top && ext4_rec_len_from_disk(d->rec_len, blocksize))
356 d = (struct ext4_dir_entry *)( 356 d = (struct ext4_dir_entry *)(((void *)d) +
357 ext4_rec_len_from_disk(d-> 357 ext4_rec_len_from_disk(d->rec_len, blocksize));
358 358
359 if (d != top) 359 if (d != top)
360 return NULL; 360 return NULL;
361 361
362 t = (struct ext4_dir_entry_tail *)d; 362 t = (struct ext4_dir_entry_tail *)d;
363 #else 363 #else
364 t = EXT4_DIRENT_TAIL(bh->b_data, EXT4_ 364 t = EXT4_DIRENT_TAIL(bh->b_data, EXT4_BLOCK_SIZE(inode->i_sb));
365 #endif 365 #endif
366 366
367 if (t->det_reserved_zero1 || 367 if (t->det_reserved_zero1 ||
368 (ext4_rec_len_from_disk(t->det_rec 368 (ext4_rec_len_from_disk(t->det_rec_len, blocksize) !=
369 sizeof(struct ext4_dir_entry_tail 369 sizeof(struct ext4_dir_entry_tail)) ||
370 t->det_reserved_zero2 || 370 t->det_reserved_zero2 ||
371 t->det_reserved_ft != EXT4_FT_DIR_ 371 t->det_reserved_ft != EXT4_FT_DIR_CSUM)
372 return NULL; 372 return NULL;
373 373
374 return t; 374 return t;
375 } 375 }
376 376
377 static __le32 ext4_dirblock_csum(struct inode 377 static __le32 ext4_dirblock_csum(struct inode *inode, void *dirent, int size)
378 { 378 {
379 struct ext4_sb_info *sbi = EXT4_SB(ino 379 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
380 struct ext4_inode_info *ei = EXT4_I(in 380 struct ext4_inode_info *ei = EXT4_I(inode);
381 __u32 csum; 381 __u32 csum;
382 382
383 csum = ext4_chksum(sbi, ei->i_csum_see 383 csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size);
384 return cpu_to_le32(csum); 384 return cpu_to_le32(csum);
385 } 385 }
386 386
387 #define warn_no_space_for_csum(inode) 387 #define warn_no_space_for_csum(inode) \
388 __warn_no_space_for_csum((inode), __fu 388 __warn_no_space_for_csum((inode), __func__, __LINE__)
389 389
390 static void __warn_no_space_for_csum(struct in 390 static void __warn_no_space_for_csum(struct inode *inode, const char *func,
391 unsigned 391 unsigned int line)
392 { 392 {
393 __ext4_warning_inode(inode, func, line 393 __ext4_warning_inode(inode, func, line,
394 "No space for directory leaf c 394 "No space for directory leaf checksum. Please run e2fsck -D.");
395 } 395 }
396 396
397 int ext4_dirblock_csum_verify(struct inode *in 397 int ext4_dirblock_csum_verify(struct inode *inode, struct buffer_head *bh)
398 { 398 {
399 struct ext4_dir_entry_tail *t; 399 struct ext4_dir_entry_tail *t;
400 400
401 if (!ext4_has_metadata_csum(inode->i_s 401 if (!ext4_has_metadata_csum(inode->i_sb))
402 return 1; 402 return 1;
403 403
404 t = get_dirent_tail(inode, bh); 404 t = get_dirent_tail(inode, bh);
405 if (!t) { 405 if (!t) {
406 warn_no_space_for_csum(inode); 406 warn_no_space_for_csum(inode);
407 return 0; 407 return 0;
408 } 408 }
409 409
410 if (t->det_checksum != ext4_dirblock_c 410 if (t->det_checksum != ext4_dirblock_csum(inode, bh->b_data,
411 411 (char *)t - bh->b_data))
412 return 0; 412 return 0;
413 413
414 return 1; 414 return 1;
415 } 415 }
416 416
417 static void ext4_dirblock_csum_set(struct inod 417 static void ext4_dirblock_csum_set(struct inode *inode,
418 struct buffer 418 struct buffer_head *bh)
419 { 419 {
420 struct ext4_dir_entry_tail *t; 420 struct ext4_dir_entry_tail *t;
421 421
422 if (!ext4_has_metadata_csum(inode->i_s 422 if (!ext4_has_metadata_csum(inode->i_sb))
423 return; 423 return;
424 424
425 t = get_dirent_tail(inode, bh); 425 t = get_dirent_tail(inode, bh);
426 if (!t) { 426 if (!t) {
427 warn_no_space_for_csum(inode); 427 warn_no_space_for_csum(inode);
428 return; 428 return;
429 } 429 }
430 430
431 t->det_checksum = ext4_dirblock_csum(i 431 t->det_checksum = ext4_dirblock_csum(inode, bh->b_data,
432 ( 432 (char *)t - bh->b_data);
433 } 433 }
434 434
435 int ext4_handle_dirty_dirblock(handle_t *handl 435 int ext4_handle_dirty_dirblock(handle_t *handle,
436 struct inode *i 436 struct inode *inode,
437 struct buffer_h 437 struct buffer_head *bh)
438 { 438 {
439 ext4_dirblock_csum_set(inode, bh); 439 ext4_dirblock_csum_set(inode, bh);
440 return ext4_handle_dirty_metadata(hand 440 return ext4_handle_dirty_metadata(handle, inode, bh);
441 } 441 }
442 442
443 static struct dx_countlimit *get_dx_countlimit 443 static struct dx_countlimit *get_dx_countlimit(struct inode *inode,
444 444 struct ext4_dir_entry *dirent,
445 445 int *offset)
446 { 446 {
447 struct ext4_dir_entry *dp; 447 struct ext4_dir_entry *dp;
448 struct dx_root_info *root; 448 struct dx_root_info *root;
449 int count_offset; 449 int count_offset;
450 int blocksize = EXT4_BLOCK_SIZE(inode- 450 int blocksize = EXT4_BLOCK_SIZE(inode->i_sb);
451 unsigned int rlen = ext4_rec_len_from_ 451 unsigned int rlen = ext4_rec_len_from_disk(dirent->rec_len, blocksize);
452 452
453 if (rlen == blocksize) 453 if (rlen == blocksize)
454 count_offset = 8; 454 count_offset = 8;
455 else if (rlen == 12) { 455 else if (rlen == 12) {
456 dp = (struct ext4_dir_entry *) 456 dp = (struct ext4_dir_entry *)(((void *)dirent) + 12);
457 if (ext4_rec_len_from_disk(dp- 457 if (ext4_rec_len_from_disk(dp->rec_len, blocksize) != blocksize - 12)
458 return NULL; 458 return NULL;
459 root = (struct dx_root_info *) 459 root = (struct dx_root_info *)(((void *)dp + 12));
460 if (root->reserved_zero || 460 if (root->reserved_zero ||
461 root->info_length != sizeo 461 root->info_length != sizeof(struct dx_root_info))
462 return NULL; 462 return NULL;
463 count_offset = 32; 463 count_offset = 32;
464 } else 464 } else
465 return NULL; 465 return NULL;
466 466
467 if (offset) 467 if (offset)
468 *offset = count_offset; 468 *offset = count_offset;
469 return (struct dx_countlimit *)(((void 469 return (struct dx_countlimit *)(((void *)dirent) + count_offset);
470 } 470 }
471 471
472 static __le32 ext4_dx_csum(struct inode *inode 472 static __le32 ext4_dx_csum(struct inode *inode, struct ext4_dir_entry *dirent,
473 int count_offset, i 473 int count_offset, int count, struct dx_tail *t)
474 { 474 {
475 struct ext4_sb_info *sbi = EXT4_SB(ino 475 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
476 struct ext4_inode_info *ei = EXT4_I(in 476 struct ext4_inode_info *ei = EXT4_I(inode);
477 __u32 csum; 477 __u32 csum;
478 int size; 478 int size;
479 __u32 dummy_csum = 0; 479 __u32 dummy_csum = 0;
480 int offset = offsetof(struct dx_tail, 480 int offset = offsetof(struct dx_tail, dt_checksum);
481 481
482 size = count_offset + (count * sizeof( 482 size = count_offset + (count * sizeof(struct dx_entry));
483 csum = ext4_chksum(sbi, ei->i_csum_see 483 csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size);
484 csum = ext4_chksum(sbi, csum, (__u8 *) 484 csum = ext4_chksum(sbi, csum, (__u8 *)t, offset);
485 csum = ext4_chksum(sbi, csum, (__u8 *) 485 csum = ext4_chksum(sbi, csum, (__u8 *)&dummy_csum, sizeof(dummy_csum));
486 486
487 return cpu_to_le32(csum); 487 return cpu_to_le32(csum);
488 } 488 }
489 489
490 static int ext4_dx_csum_verify(struct inode *i 490 static int ext4_dx_csum_verify(struct inode *inode,
491 struct ext4_dir 491 struct ext4_dir_entry *dirent)
492 { 492 {
493 struct dx_countlimit *c; 493 struct dx_countlimit *c;
494 struct dx_tail *t; 494 struct dx_tail *t;
495 int count_offset, limit, count; 495 int count_offset, limit, count;
496 496
497 if (!ext4_has_metadata_csum(inode->i_s 497 if (!ext4_has_metadata_csum(inode->i_sb))
498 return 1; 498 return 1;
499 499
500 c = get_dx_countlimit(inode, dirent, & 500 c = get_dx_countlimit(inode, dirent, &count_offset);
501 if (!c) { 501 if (!c) {
502 EXT4_ERROR_INODE(inode, "dir s 502 EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D.");
503 return 0; 503 return 0;
504 } 504 }
505 limit = le16_to_cpu(c->limit); 505 limit = le16_to_cpu(c->limit);
506 count = le16_to_cpu(c->count); 506 count = le16_to_cpu(c->count);
507 if (count_offset + (limit * sizeof(str 507 if (count_offset + (limit * sizeof(struct dx_entry)) >
508 EXT4_BLOCK_SIZE(inode->i_sb) - siz 508 EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) {
509 warn_no_space_for_csum(inode); 509 warn_no_space_for_csum(inode);
510 return 0; 510 return 0;
511 } 511 }
512 t = (struct dx_tail *)(((struct dx_ent 512 t = (struct dx_tail *)(((struct dx_entry *)c) + limit);
513 513
514 if (t->dt_checksum != ext4_dx_csum(ino 514 if (t->dt_checksum != ext4_dx_csum(inode, dirent, count_offset,
515 co 515 count, t))
516 return 0; 516 return 0;
517 return 1; 517 return 1;
518 } 518 }
519 519
520 static void ext4_dx_csum_set(struct inode *ino 520 static void ext4_dx_csum_set(struct inode *inode, struct ext4_dir_entry *dirent)
521 { 521 {
522 struct dx_countlimit *c; 522 struct dx_countlimit *c;
523 struct dx_tail *t; 523 struct dx_tail *t;
524 int count_offset, limit, count; 524 int count_offset, limit, count;
525 525
526 if (!ext4_has_metadata_csum(inode->i_s 526 if (!ext4_has_metadata_csum(inode->i_sb))
527 return; 527 return;
528 528
529 c = get_dx_countlimit(inode, dirent, & 529 c = get_dx_countlimit(inode, dirent, &count_offset);
530 if (!c) { 530 if (!c) {
531 EXT4_ERROR_INODE(inode, "dir s 531 EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D.");
532 return; 532 return;
533 } 533 }
534 limit = le16_to_cpu(c->limit); 534 limit = le16_to_cpu(c->limit);
535 count = le16_to_cpu(c->count); 535 count = le16_to_cpu(c->count);
536 if (count_offset + (limit * sizeof(str 536 if (count_offset + (limit * sizeof(struct dx_entry)) >
537 EXT4_BLOCK_SIZE(inode->i_sb) - siz 537 EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) {
538 warn_no_space_for_csum(inode); 538 warn_no_space_for_csum(inode);
539 return; 539 return;
540 } 540 }
541 t = (struct dx_tail *)(((struct dx_ent 541 t = (struct dx_tail *)(((struct dx_entry *)c) + limit);
542 542
543 t->dt_checksum = ext4_dx_csum(inode, d 543 t->dt_checksum = ext4_dx_csum(inode, dirent, count_offset, count, t);
544 } 544 }
545 545
546 static inline int ext4_handle_dirty_dx_node(ha 546 static inline int ext4_handle_dirty_dx_node(handle_t *handle,
547 st 547 struct inode *inode,
548 st 548 struct buffer_head *bh)
549 { 549 {
550 ext4_dx_csum_set(inode, (struct ext4_d 550 ext4_dx_csum_set(inode, (struct ext4_dir_entry *)bh->b_data);
551 return ext4_handle_dirty_metadata(hand 551 return ext4_handle_dirty_metadata(handle, inode, bh);
552 } 552 }
553 553
554 /* 554 /*
555 * p is at least 6 bytes before the end of pag 555 * p is at least 6 bytes before the end of page
556 */ 556 */
557 static inline struct ext4_dir_entry_2 * 557 static inline struct ext4_dir_entry_2 *
558 ext4_next_entry(struct ext4_dir_entry_2 *p, un 558 ext4_next_entry(struct ext4_dir_entry_2 *p, unsigned long blocksize)
559 { 559 {
560 return (struct ext4_dir_entry_2 *)((ch 560 return (struct ext4_dir_entry_2 *)((char *)p +
561 ext4_rec_len_from_disk(p->rec_ 561 ext4_rec_len_from_disk(p->rec_len, blocksize));
562 } 562 }
563 563
564 /* 564 /*
565 * Future: use high four bits of block for coa 565 * Future: use high four bits of block for coalesce-on-delete flags
566 * Mask them off for now. 566 * Mask them off for now.
567 */ 567 */
568 568
569 static inline ext4_lblk_t dx_get_block(struct 569 static inline ext4_lblk_t dx_get_block(struct dx_entry *entry)
570 { 570 {
571 return le32_to_cpu(entry->block) & 0x0 571 return le32_to_cpu(entry->block) & 0x0fffffff;
572 } 572 }
573 573
574 static inline void dx_set_block(struct dx_entr 574 static inline void dx_set_block(struct dx_entry *entry, ext4_lblk_t value)
575 { 575 {
576 entry->block = cpu_to_le32(value); 576 entry->block = cpu_to_le32(value);
577 } 577 }
578 578
579 static inline unsigned dx_get_hash(struct dx_e 579 static inline unsigned dx_get_hash(struct dx_entry *entry)
580 { 580 {
581 return le32_to_cpu(entry->hash); 581 return le32_to_cpu(entry->hash);
582 } 582 }
583 583
584 static inline void dx_set_hash(struct dx_entry 584 static inline void dx_set_hash(struct dx_entry *entry, unsigned value)
585 { 585 {
586 entry->hash = cpu_to_le32(value); 586 entry->hash = cpu_to_le32(value);
587 } 587 }
588 588
589 static inline unsigned dx_get_count(struct dx_ 589 static inline unsigned dx_get_count(struct dx_entry *entries)
590 { 590 {
591 return le16_to_cpu(((struct dx_countli 591 return le16_to_cpu(((struct dx_countlimit *) entries)->count);
592 } 592 }
593 593
594 static inline unsigned dx_get_limit(struct dx_ 594 static inline unsigned dx_get_limit(struct dx_entry *entries)
595 { 595 {
596 return le16_to_cpu(((struct dx_countli 596 return le16_to_cpu(((struct dx_countlimit *) entries)->limit);
597 } 597 }
598 598
599 static inline void dx_set_count(struct dx_entr 599 static inline void dx_set_count(struct dx_entry *entries, unsigned value)
600 { 600 {
601 ((struct dx_countlimit *) entries)->co 601 ((struct dx_countlimit *) entries)->count = cpu_to_le16(value);
602 } 602 }
603 603
604 static inline void dx_set_limit(struct dx_entr 604 static inline void dx_set_limit(struct dx_entry *entries, unsigned value)
605 { 605 {
606 ((struct dx_countlimit *) entries)->li 606 ((struct dx_countlimit *) entries)->limit = cpu_to_le16(value);
607 } 607 }
608 608
609 static inline unsigned dx_root_limit(struct in 609 static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize)
610 { 610 {
611 unsigned int entry_space = dir->i_sb-> 611 unsigned int entry_space = dir->i_sb->s_blocksize -
612 ext4_dir_rec_len(1, NU 612 ext4_dir_rec_len(1, NULL) -
613 ext4_dir_rec_len(2, NU 613 ext4_dir_rec_len(2, NULL) - infosize;
614 614
615 if (ext4_has_metadata_csum(dir->i_sb)) 615 if (ext4_has_metadata_csum(dir->i_sb))
616 entry_space -= sizeof(struct d 616 entry_space -= sizeof(struct dx_tail);
617 return entry_space / sizeof(struct dx_ 617 return entry_space / sizeof(struct dx_entry);
618 } 618 }
619 619
620 static inline unsigned dx_node_limit(struct in 620 static inline unsigned dx_node_limit(struct inode *dir)
621 { 621 {
622 unsigned int entry_space = dir->i_sb-> 622 unsigned int entry_space = dir->i_sb->s_blocksize -
623 ext4_dir_rec_len(0, di 623 ext4_dir_rec_len(0, dir);
624 624
625 if (ext4_has_metadata_csum(dir->i_sb)) 625 if (ext4_has_metadata_csum(dir->i_sb))
626 entry_space -= sizeof(struct d 626 entry_space -= sizeof(struct dx_tail);
627 return entry_space / sizeof(struct dx_ 627 return entry_space / sizeof(struct dx_entry);
628 } 628 }
629 629
630 /* 630 /*
631 * Debug 631 * Debug
632 */ 632 */
633 #ifdef DX_DEBUG 633 #ifdef DX_DEBUG
634 static void dx_show_index(char * label, struct 634 static void dx_show_index(char * label, struct dx_entry *entries)
635 { 635 {
636 int i, n = dx_get_count (entries); 636 int i, n = dx_get_count (entries);
637 printk(KERN_DEBUG "%s index", label); 637 printk(KERN_DEBUG "%s index", label);
638 for (i = 0; i < n; i++) { 638 for (i = 0; i < n; i++) {
639 printk(KERN_CONT " %x->%lu", 639 printk(KERN_CONT " %x->%lu",
640 i ? dx_get_hash(entries 640 i ? dx_get_hash(entries + i) : 0,
641 (unsigned long)dx_get_b 641 (unsigned long)dx_get_block(entries + i));
642 } 642 }
643 printk(KERN_CONT "\n"); 643 printk(KERN_CONT "\n");
644 } 644 }
645 645
646 struct stats 646 struct stats
647 { 647 {
648 unsigned names; 648 unsigned names;
649 unsigned space; 649 unsigned space;
650 unsigned bcount; 650 unsigned bcount;
651 }; 651 };
652 652
653 static struct stats dx_show_leaf(struct inode 653 static struct stats dx_show_leaf(struct inode *dir,
654 struct dx_hash 654 struct dx_hash_info *hinfo,
655 struct ext4_di 655 struct ext4_dir_entry_2 *de,
656 int size, int 656 int size, int show_names)
657 { 657 {
658 unsigned names = 0, space = 0; 658 unsigned names = 0, space = 0;
659 char *base = (char *) de; 659 char *base = (char *) de;
660 struct dx_hash_info h = *hinfo; 660 struct dx_hash_info h = *hinfo;
661 661
662 printk("names: "); 662 printk("names: ");
663 while ((char *) de < base + size) 663 while ((char *) de < base + size)
664 { 664 {
665 if (de->inode) 665 if (de->inode)
666 { 666 {
667 if (show_names) 667 if (show_names)
668 { 668 {
669 #ifdef CONFIG_FS_ENCRYPTION 669 #ifdef CONFIG_FS_ENCRYPTION
670 int len; 670 int len;
671 char *name; 671 char *name;
672 struct fscrypt 672 struct fscrypt_str fname_crypto_str =
673 FSTR_I 673 FSTR_INIT(NULL, 0);
674 int res = 0; 674 int res = 0;
675 675
676 name = de->na 676 name = de->name;
677 len = de->name 677 len = de->name_len;
678 if (!IS_ENCRYP 678 if (!IS_ENCRYPTED(dir)) {
679 /* Dir 679 /* Directory is not encrypted */
680 (void) 680 (void) ext4fs_dirhash(dir, de->name,
681 681 de->name_len, &h);
682 printk 682 printk("%*.s:(U)%x.%u ", len,
683 683 name, h.hash,
684 684 (unsigned) ((char *) de
685 685 - base));
686 } else { 686 } else {
687 struct 687 struct fscrypt_str de_name =
688 688 FSTR_INIT(name, len);
689 689
690 /* Dir 690 /* Directory is encrypted */
691 res = 691 res = fscrypt_fname_alloc_buffer(
692 692 len, &fname_crypto_str);
693 if (re 693 if (res)
694 694 printk(KERN_WARNING "Error "
695 695 "allocating crypto "
696 696 "buffer--skipping "
697 697 "crypto\n");
698 res = 698 res = fscrypt_fname_disk_to_usr(dir,
699 699 0, 0, &de_name,
700 700 &fname_crypto_str);
701 if (re 701 if (res) {
702 702 printk(KERN_WARNING "Error "
703 703 "converting filename "
704 704 "from disk to usr"
705 705 "\n");
706 706 name = "??";
707 707 len = 2;
708 } else 708 } else {
709 709 name = fname_crypto_str.name;
710 710 len = fname_crypto_str.len;
711 } 711 }
712 if (IS 712 if (IS_CASEFOLDED(dir))
713 713 h.hash = EXT4_DIRENT_HASH(de);
714 else 714 else
715 715 (void) ext4fs_dirhash(dir,
716 716 de->name,
717 717 de->name_len, &h);
718 printk 718 printk("%*.s:(E)%x.%u ", len, name,
719 719 h.hash, (unsigned) ((char *) de
720 720 - base));
721 fscryp 721 fscrypt_fname_free_buffer(
722 722 &fname_crypto_str);
723 } 723 }
724 #else 724 #else
725 int len = de-> 725 int len = de->name_len;
726 char *name = d 726 char *name = de->name;
727 (void) ext4fs_ 727 (void) ext4fs_dirhash(dir, de->name,
728 728 de->name_len, &h);
729 printk("%*.s:% 729 printk("%*.s:%x.%u ", len, name, h.hash,
730 (unsign 730 (unsigned) ((char *) de - base));
731 #endif 731 #endif
732 } 732 }
733 space += ext4_dir_rec_ 733 space += ext4_dir_rec_len(de->name_len, dir);
734 names++; 734 names++;
735 } 735 }
736 de = ext4_next_entry(de, size) 736 de = ext4_next_entry(de, size);
737 } 737 }
738 printk(KERN_CONT "(%i)\n", names); 738 printk(KERN_CONT "(%i)\n", names);
739 return (struct stats) { names, space, 739 return (struct stats) { names, space, 1 };
740 } 740 }
741 741
742 struct stats dx_show_entries(struct dx_hash_in 742 struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir,
743 struct dx_entry * 743 struct dx_entry *entries, int levels)
744 { 744 {
745 unsigned blocksize = dir->i_sb->s_bloc 745 unsigned blocksize = dir->i_sb->s_blocksize;
746 unsigned count = dx_get_count(entries) 746 unsigned count = dx_get_count(entries), names = 0, space = 0, i;
747 unsigned bcount = 0; 747 unsigned bcount = 0;
748 struct buffer_head *bh; 748 struct buffer_head *bh;
749 printk("%i indexed blocks...\n", count 749 printk("%i indexed blocks...\n", count);
750 for (i = 0; i < count; i++, entries++) 750 for (i = 0; i < count; i++, entries++)
751 { 751 {
752 ext4_lblk_t block = dx_get_blo 752 ext4_lblk_t block = dx_get_block(entries);
753 ext4_lblk_t hash = i ? dx_get 753 ext4_lblk_t hash = i ? dx_get_hash(entries): 0;
754 u32 range = i < count - 1? (dx 754 u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash;
755 struct stats stats; 755 struct stats stats;
756 printk("%s%3u:%03u hash %8x/%8 756 printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range);
757 bh = ext4_bread(NULL,dir, bloc 757 bh = ext4_bread(NULL,dir, block, 0);
758 if (!bh || IS_ERR(bh)) 758 if (!bh || IS_ERR(bh))
759 continue; 759 continue;
760 stats = levels? 760 stats = levels?
761 dx_show_entries(hinfo, dir, 761 dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1):
762 dx_show_leaf(dir, hinfo, (s 762 dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *)
763 bh->b_data, blocksize, 763 bh->b_data, blocksize, 0);
764 names += stats.names; 764 names += stats.names;
765 space += stats.space; 765 space += stats.space;
766 bcount += stats.bcount; 766 bcount += stats.bcount;
767 brelse(bh); 767 brelse(bh);
768 } 768 }
769 if (bcount) 769 if (bcount)
770 printk(KERN_DEBUG "%snames %u, 770 printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n",
771 levels ? "" : " ", na 771 levels ? "" : " ", names, space/bcount,
772 (space/bcount)*100/bloc 772 (space/bcount)*100/blocksize);
773 return (struct stats) { names, space, 773 return (struct stats) { names, space, bcount};
774 } 774 }
775 775
776 /* 776 /*
777 * Linear search cross check 777 * Linear search cross check
778 */ 778 */
779 static inline void htree_rep_invariant_check(s 779 static inline void htree_rep_invariant_check(struct dx_entry *at,
780 s 780 struct dx_entry *target,
781 u 781 u32 hash, unsigned int n)
782 { 782 {
783 while (n--) { 783 while (n--) {
784 dxtrace(printk(KERN_CONT ",")) 784 dxtrace(printk(KERN_CONT ","));
785 if (dx_get_hash(++at) > hash) 785 if (dx_get_hash(++at) > hash) {
786 at--; 786 at--;
787 break; 787 break;
788 } 788 }
789 } 789 }
790 ASSERT(at == target - 1); 790 ASSERT(at == target - 1);
791 } 791 }
792 #else /* DX_DEBUG */ 792 #else /* DX_DEBUG */
793 static inline void htree_rep_invariant_check(s 793 static inline void htree_rep_invariant_check(struct dx_entry *at,
794 s 794 struct dx_entry *target,
795 u 795 u32 hash, unsigned int n)
796 { 796 {
797 } 797 }
798 #endif /* DX_DEBUG */ 798 #endif /* DX_DEBUG */
799 799
800 /* 800 /*
801 * Probe for a directory leaf block to search. 801 * Probe for a directory leaf block to search.
802 * 802 *
803 * dx_probe can return ERR_BAD_DX_DIR, which m 803 * dx_probe can return ERR_BAD_DX_DIR, which means there was a format
804 * error in the directory index, and the calle 804 * error in the directory index, and the caller should fall back to
805 * searching the directory normally. The call 805 * searching the directory normally. The callers of dx_probe **MUST**
806 * check for this error code, and make sure it 806 * check for this error code, and make sure it never gets reflected
807 * back to userspace. 807 * back to userspace.
808 */ 808 */
809 static struct dx_frame * 809 static struct dx_frame *
810 dx_probe(struct ext4_filename *fname, struct i 810 dx_probe(struct ext4_filename *fname, struct inode *dir,
811 struct dx_hash_info *hinfo, struct dx 811 struct dx_hash_info *hinfo, struct dx_frame *frame_in)
812 { 812 {
813 unsigned count, indirect, level, i; 813 unsigned count, indirect, level, i;
814 struct dx_entry *at, *entries, *p, *q, 814 struct dx_entry *at, *entries, *p, *q, *m;
815 struct dx_root *root; 815 struct dx_root *root;
816 struct dx_frame *frame = frame_in; 816 struct dx_frame *frame = frame_in;
817 struct dx_frame *ret_err = ERR_PTR(ERR 817 struct dx_frame *ret_err = ERR_PTR(ERR_BAD_DX_DIR);
818 u32 hash; 818 u32 hash;
819 ext4_lblk_t block; 819 ext4_lblk_t block;
820 ext4_lblk_t blocks[EXT4_HTREE_LEVEL]; 820 ext4_lblk_t blocks[EXT4_HTREE_LEVEL];
821 821
822 memset(frame_in, 0, EXT4_HTREE_LEVEL * 822 memset(frame_in, 0, EXT4_HTREE_LEVEL * sizeof(frame_in[0]));
823 frame->bh = ext4_read_dirblock(dir, 0, 823 frame->bh = ext4_read_dirblock(dir, 0, INDEX);
824 if (IS_ERR(frame->bh)) 824 if (IS_ERR(frame->bh))
825 return (struct dx_frame *) fra 825 return (struct dx_frame *) frame->bh;
826 826
827 root = (struct dx_root *) frame->bh->b 827 root = (struct dx_root *) frame->bh->b_data;
828 if (root->info.hash_version != DX_HASH 828 if (root->info.hash_version != DX_HASH_TEA &&
829 root->info.hash_version != DX_HASH 829 root->info.hash_version != DX_HASH_HALF_MD4 &&
830 root->info.hash_version != DX_HASH 830 root->info.hash_version != DX_HASH_LEGACY &&
831 root->info.hash_version != DX_HASH 831 root->info.hash_version != DX_HASH_SIPHASH) {
832 ext4_warning_inode(dir, "Unrec 832 ext4_warning_inode(dir, "Unrecognised inode hash code %u",
833 root->info. 833 root->info.hash_version);
834 goto fail; 834 goto fail;
835 } 835 }
836 if (ext4_hash_in_dirent(dir)) { 836 if (ext4_hash_in_dirent(dir)) {
837 if (root->info.hash_version != 837 if (root->info.hash_version != DX_HASH_SIPHASH) {
838 ext4_warning_inode(dir 838 ext4_warning_inode(dir,
839 "Hash in diren 839 "Hash in dirent, but hash is not SIPHASH");
840 goto fail; 840 goto fail;
841 } 841 }
842 } else { 842 } else {
843 if (root->info.hash_version == 843 if (root->info.hash_version == DX_HASH_SIPHASH) {
844 ext4_warning_inode(dir 844 ext4_warning_inode(dir,
845 "Hash code is 845 "Hash code is SIPHASH, but hash not in dirent");
846 goto fail; 846 goto fail;
847 } 847 }
848 } 848 }
849 if (fname) 849 if (fname)
850 hinfo = &fname->hinfo; 850 hinfo = &fname->hinfo;
851 hinfo->hash_version = root->info.hash_ 851 hinfo->hash_version = root->info.hash_version;
852 if (hinfo->hash_version <= DX_HASH_TEA 852 if (hinfo->hash_version <= DX_HASH_TEA)
853 hinfo->hash_version += EXT4_SB 853 hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned;
854 hinfo->seed = EXT4_SB(dir->i_sb)->s_ha 854 hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed;
855 /* hash is already computed for encryp 855 /* hash is already computed for encrypted casefolded directory */
856 if (fname && fname_name(fname) && 856 if (fname && fname_name(fname) &&
857 !(IS_ENCRYPTED(dir) && IS_CASEFOLD 857 !(IS_ENCRYPTED(dir) && IS_CASEFOLDED(dir))) {
858 int ret = ext4fs_dirhash(dir, 858 int ret = ext4fs_dirhash(dir, fname_name(fname),
859 fname 859 fname_len(fname), hinfo);
860 if (ret < 0) { 860 if (ret < 0) {
861 ret_err = ERR_PTR(ret) 861 ret_err = ERR_PTR(ret);
862 goto fail; 862 goto fail;
863 } 863 }
864 } 864 }
865 hash = hinfo->hash; 865 hash = hinfo->hash;
866 866
867 if (root->info.unused_flags & 1) { 867 if (root->info.unused_flags & 1) {
868 ext4_warning_inode(dir, "Unimp 868 ext4_warning_inode(dir, "Unimplemented hash flags: %#06x",
869 root->info. 869 root->info.unused_flags);
870 goto fail; 870 goto fail;
871 } 871 }
872 872
873 indirect = root->info.indirect_levels; 873 indirect = root->info.indirect_levels;
874 if (indirect >= ext4_dir_htree_level(d 874 if (indirect >= ext4_dir_htree_level(dir->i_sb)) {
875 ext4_warning(dir->i_sb, 875 ext4_warning(dir->i_sb,
876 "Directory (ino: 876 "Directory (ino: %lu) htree depth %#06x exceed"
877 "supported value" 877 "supported value", dir->i_ino,
878 ext4_dir_htree_le 878 ext4_dir_htree_level(dir->i_sb));
879 if (ext4_dir_htree_level(dir-> 879 if (ext4_dir_htree_level(dir->i_sb) < EXT4_HTREE_LEVEL) {
880 ext4_warning(dir->i_sb 880 ext4_warning(dir->i_sb, "Enable large directory "
881 881 "feature to access it");
882 } 882 }
883 goto fail; 883 goto fail;
884 } 884 }
885 885
886 entries = (struct dx_entry *)(((char * 886 entries = (struct dx_entry *)(((char *)&root->info) +
887 root->in 887 root->info.info_length);
888 888
889 if (dx_get_limit(entries) != dx_root_l 889 if (dx_get_limit(entries) != dx_root_limit(dir,
890 890 root->info.info_length)) {
891 ext4_warning_inode(dir, "dx en 891 ext4_warning_inode(dir, "dx entry: limit %u != root limit %u",
892 dx_get_limi 892 dx_get_limit(entries),
893 dx_root_lim 893 dx_root_limit(dir, root->info.info_length));
894 goto fail; 894 goto fail;
895 } 895 }
896 896
897 dxtrace(printk("Look up %x", hash)); 897 dxtrace(printk("Look up %x", hash));
898 level = 0; 898 level = 0;
899 blocks[0] = 0; 899 blocks[0] = 0;
900 while (1) { 900 while (1) {
901 count = dx_get_count(entries); 901 count = dx_get_count(entries);
902 if (!count || count > dx_get_l 902 if (!count || count > dx_get_limit(entries)) {
903 ext4_warning_inode(dir 903 ext4_warning_inode(dir,
904 "dx 904 "dx entry: count %u beyond limit %u",
905 cou 905 count, dx_get_limit(entries));
906 goto fail; 906 goto fail;
907 } 907 }
908 908
909 p = entries + 1; 909 p = entries + 1;
910 q = entries + count - 1; 910 q = entries + count - 1;
911 while (p <= q) { 911 while (p <= q) {
912 m = p + (q - p) / 2; 912 m = p + (q - p) / 2;
913 dxtrace(printk(KERN_CO 913 dxtrace(printk(KERN_CONT "."));
914 if (dx_get_hash(m) > h 914 if (dx_get_hash(m) > hash)
915 q = m - 1; 915 q = m - 1;
916 else 916 else
917 p = m + 1; 917 p = m + 1;
918 } 918 }
919 919
920 htree_rep_invariant_check(entr 920 htree_rep_invariant_check(entries, p, hash, count - 1);
921 921
922 at = p - 1; 922 at = p - 1;
923 dxtrace(printk(KERN_CONT " %x- 923 dxtrace(printk(KERN_CONT " %x->%u\n",
924 at == entries ? 924 at == entries ? 0 : dx_get_hash(at),
925 dx_get_block(at 925 dx_get_block(at)));
926 frame->entries = entries; 926 frame->entries = entries;
927 frame->at = at; 927 frame->at = at;
928 928
929 block = dx_get_block(at); 929 block = dx_get_block(at);
930 for (i = 0; i <= level; i++) { 930 for (i = 0; i <= level; i++) {
931 if (blocks[i] == block 931 if (blocks[i] == block) {
932 ext4_warning_i 932 ext4_warning_inode(dir,
933 "dx en 933 "dx entry: tree cycle block %u points back to block %u",
934 blocks 934 blocks[level], block);
935 goto fail; 935 goto fail;
936 } 936 }
937 } 937 }
938 if (++level > indirect) 938 if (++level > indirect)
939 return frame; 939 return frame;
940 blocks[level] = block; 940 blocks[level] = block;
941 frame++; 941 frame++;
942 frame->bh = ext4_read_dirblock 942 frame->bh = ext4_read_dirblock(dir, block, INDEX);
943 if (IS_ERR(frame->bh)) { 943 if (IS_ERR(frame->bh)) {
944 ret_err = (struct dx_f 944 ret_err = (struct dx_frame *) frame->bh;
945 frame->bh = NULL; 945 frame->bh = NULL;
946 goto fail; 946 goto fail;
947 } 947 }
948 948
949 entries = ((struct dx_node *) 949 entries = ((struct dx_node *) frame->bh->b_data)->entries;
950 950
951 if (dx_get_limit(entries) != d 951 if (dx_get_limit(entries) != dx_node_limit(dir)) {
952 ext4_warning_inode(dir 952 ext4_warning_inode(dir,
953 "dx entry: lim 953 "dx entry: limit %u != node limit %u",
954 dx_get_limit(e 954 dx_get_limit(entries), dx_node_limit(dir));
955 goto fail; 955 goto fail;
956 } 956 }
957 } 957 }
958 fail: 958 fail:
959 while (frame >= frame_in) { 959 while (frame >= frame_in) {
960 brelse(frame->bh); 960 brelse(frame->bh);
961 frame--; 961 frame--;
962 } 962 }
963 963
964 if (ret_err == ERR_PTR(ERR_BAD_DX_DIR) 964 if (ret_err == ERR_PTR(ERR_BAD_DX_DIR))
965 ext4_warning_inode(dir, 965 ext4_warning_inode(dir,
966 "Corrupt directory, ru 966 "Corrupt directory, running e2fsck is recommended");
967 return ret_err; 967 return ret_err;
968 } 968 }
969 969
970 static void dx_release(struct dx_frame *frames 970 static void dx_release(struct dx_frame *frames)
971 { 971 {
972 struct dx_root_info *info; 972 struct dx_root_info *info;
973 int i; 973 int i;
974 unsigned int indirect_levels; 974 unsigned int indirect_levels;
975 975
976 if (frames[0].bh == NULL) 976 if (frames[0].bh == NULL)
977 return; 977 return;
978 978
979 info = &((struct dx_root *)frames[0].b 979 info = &((struct dx_root *)frames[0].bh->b_data)->info;
980 /* save local copy, "info" may be free 980 /* save local copy, "info" may be freed after brelse() */
981 indirect_levels = info->indirect_level 981 indirect_levels = info->indirect_levels;
982 for (i = 0; i <= indirect_levels; i++) 982 for (i = 0; i <= indirect_levels; i++) {
983 if (frames[i].bh == NULL) 983 if (frames[i].bh == NULL)
984 break; 984 break;
985 brelse(frames[i].bh); 985 brelse(frames[i].bh);
986 frames[i].bh = NULL; 986 frames[i].bh = NULL;
987 } 987 }
988 } 988 }
989 989
990 /* 990 /*
991 * This function increments the frame pointer 991 * This function increments the frame pointer to search the next leaf
992 * block, and reads in the necessary interveni 992 * block, and reads in the necessary intervening nodes if the search
993 * should be necessary. Whether or not the se 993 * should be necessary. Whether or not the search is necessary is
994 * controlled by the hash parameter. If the h 994 * controlled by the hash parameter. If the hash value is even, then
995 * the search is only continued if the next bl 995 * the search is only continued if the next block starts with that
996 * hash value. This is used if we are searchi 996 * hash value. This is used if we are searching for a specific file.
997 * 997 *
998 * If the hash value is HASH_NB_ALWAYS, then a 998 * If the hash value is HASH_NB_ALWAYS, then always go to the next block.
999 * 999 *
1000 * This function returns 1 if the caller shou 1000 * This function returns 1 if the caller should continue to search,
1001 * or 0 if it should not. If there is an err 1001 * or 0 if it should not. If there is an error reading one of the
1002 * index blocks, it will a negative error cod 1002 * index blocks, it will a negative error code.
1003 * 1003 *
1004 * If start_hash is non-null, it will be fill 1004 * If start_hash is non-null, it will be filled in with the starting
1005 * hash of the next page. 1005 * hash of the next page.
1006 */ 1006 */
1007 static int ext4_htree_next_block(struct inode 1007 static int ext4_htree_next_block(struct inode *dir, __u32 hash,
1008 struct dx_fr 1008 struct dx_frame *frame,
1009 struct dx_fr 1009 struct dx_frame *frames,
1010 __u32 *start 1010 __u32 *start_hash)
1011 { 1011 {
1012 struct dx_frame *p; 1012 struct dx_frame *p;
1013 struct buffer_head *bh; 1013 struct buffer_head *bh;
1014 int num_frames = 0; 1014 int num_frames = 0;
1015 __u32 bhash; 1015 __u32 bhash;
1016 1016
1017 p = frame; 1017 p = frame;
1018 /* 1018 /*
1019 * Find the next leaf page by increme 1019 * Find the next leaf page by incrementing the frame pointer.
1020 * If we run out of entries in the in 1020 * If we run out of entries in the interior node, loop around and
1021 * increment pointer in the parent no 1021 * increment pointer in the parent node. When we break out of
1022 * this loop, num_frames indicates th 1022 * this loop, num_frames indicates the number of interior
1023 * nodes need to be read. 1023 * nodes need to be read.
1024 */ 1024 */
1025 while (1) { 1025 while (1) {
1026 if (++(p->at) < p->entries + 1026 if (++(p->at) < p->entries + dx_get_count(p->entries))
1027 break; 1027 break;
1028 if (p == frames) 1028 if (p == frames)
1029 return 0; 1029 return 0;
1030 num_frames++; 1030 num_frames++;
1031 p--; 1031 p--;
1032 } 1032 }
1033 1033
1034 /* 1034 /*
1035 * If the hash is 1, then continue on 1035 * If the hash is 1, then continue only if the next page has a
1036 * continuation hash of any value. T 1036 * continuation hash of any value. This is used for readdir
1037 * handling. Otherwise, check to see 1037 * handling. Otherwise, check to see if the hash matches the
1038 * desired continuation hash. If it 1038 * desired continuation hash. If it doesn't, return since
1039 * there's no point to read in the su 1039 * there's no point to read in the successive index pages.
1040 */ 1040 */
1041 bhash = dx_get_hash(p->at); 1041 bhash = dx_get_hash(p->at);
1042 if (start_hash) 1042 if (start_hash)
1043 *start_hash = bhash; 1043 *start_hash = bhash;
1044 if ((hash & 1) == 0) { 1044 if ((hash & 1) == 0) {
1045 if ((bhash & ~1) != hash) 1045 if ((bhash & ~1) != hash)
1046 return 0; 1046 return 0;
1047 } 1047 }
1048 /* 1048 /*
1049 * If the hash is HASH_NB_ALWAYS, we 1049 * If the hash is HASH_NB_ALWAYS, we always go to the next
1050 * block so no check is necessary 1050 * block so no check is necessary
1051 */ 1051 */
1052 while (num_frames--) { 1052 while (num_frames--) {
1053 bh = ext4_read_dirblock(dir, 1053 bh = ext4_read_dirblock(dir, dx_get_block(p->at), INDEX);
1054 if (IS_ERR(bh)) 1054 if (IS_ERR(bh))
1055 return PTR_ERR(bh); 1055 return PTR_ERR(bh);
1056 p++; 1056 p++;
1057 brelse(p->bh); 1057 brelse(p->bh);
1058 p->bh = bh; 1058 p->bh = bh;
1059 p->at = p->entries = ((struct 1059 p->at = p->entries = ((struct dx_node *) bh->b_data)->entries;
1060 } 1060 }
1061 return 1; 1061 return 1;
1062 } 1062 }
1063 1063
1064 1064
1065 /* 1065 /*
1066 * This function fills a red-black tree with 1066 * This function fills a red-black tree with information from a
1067 * directory block. It returns the number di 1067 * directory block. It returns the number directory entries loaded
1068 * into the tree. If there is an error it is 1068 * into the tree. If there is an error it is returned in err.
1069 */ 1069 */
1070 static int htree_dirblock_to_tree(struct file 1070 static int htree_dirblock_to_tree(struct file *dir_file,
1071 struct inod 1071 struct inode *dir, ext4_lblk_t block,
1072 struct dx_h 1072 struct dx_hash_info *hinfo,
1073 __u32 start 1073 __u32 start_hash, __u32 start_minor_hash)
1074 { 1074 {
1075 struct buffer_head *bh; 1075 struct buffer_head *bh;
1076 struct ext4_dir_entry_2 *de, *top; 1076 struct ext4_dir_entry_2 *de, *top;
1077 int err = 0, count = 0; 1077 int err = 0, count = 0;
1078 struct fscrypt_str fname_crypto_str = 1078 struct fscrypt_str fname_crypto_str = FSTR_INIT(NULL, 0), tmp_str;
1079 int csum = ext4_has_metadata_csum(dir 1079 int csum = ext4_has_metadata_csum(dir->i_sb);
1080 1080
1081 dxtrace(printk(KERN_INFO "In htree di 1081 dxtrace(printk(KERN_INFO "In htree dirblock_to_tree: block %lu\n",
1082 1082 (unsigned long)block));
1083 bh = ext4_read_dirblock(dir, block, D 1083 bh = ext4_read_dirblock(dir, block, DIRENT_HTREE);
1084 if (IS_ERR(bh)) 1084 if (IS_ERR(bh))
1085 return PTR_ERR(bh); 1085 return PTR_ERR(bh);
1086 1086
1087 de = (struct ext4_dir_entry_2 *) bh-> 1087 de = (struct ext4_dir_entry_2 *) bh->b_data;
1088 /* csum entries are not larger in the 1088 /* csum entries are not larger in the casefolded encrypted case */
1089 top = (struct ext4_dir_entry_2 *) ((c 1089 top = (struct ext4_dir_entry_2 *) ((char *) de +
1090 di 1090 dir->i_sb->s_blocksize -
1091 ex 1091 ext4_dir_rec_len(0,
1092 1092 csum ? NULL : dir));
1093 /* Check if the directory is encrypte 1093 /* Check if the directory is encrypted */
1094 if (IS_ENCRYPTED(dir)) { 1094 if (IS_ENCRYPTED(dir)) {
1095 err = fscrypt_prepare_readdir 1095 err = fscrypt_prepare_readdir(dir);
1096 if (err < 0) { 1096 if (err < 0) {
1097 brelse(bh); 1097 brelse(bh);
1098 return err; 1098 return err;
1099 } 1099 }
1100 err = fscrypt_fname_alloc_buf 1100 err = fscrypt_fname_alloc_buffer(EXT4_NAME_LEN,
1101 1101 &fname_crypto_str);
1102 if (err < 0) { 1102 if (err < 0) {
1103 brelse(bh); 1103 brelse(bh);
1104 return err; 1104 return err;
1105 } 1105 }
1106 } 1106 }
1107 1107
1108 for (; de < top; de = ext4_next_entry 1108 for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) {
1109 if (ext4_check_dir_entry(dir, 1109 if (ext4_check_dir_entry(dir, NULL, de, bh,
1110 bh->b_data, b 1110 bh->b_data, bh->b_size,
1111 (block<<EXT4_ 1111 (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb))
1112 + (( 1112 + ((char *)de - bh->b_data))) {
1113 /* silently ignore th 1113 /* silently ignore the rest of the block */
1114 break; 1114 break;
1115 } 1115 }
1116 if (ext4_hash_in_dirent(dir)) 1116 if (ext4_hash_in_dirent(dir)) {
1117 if (de->name_len && d 1117 if (de->name_len && de->inode) {
1118 hinfo->hash = 1118 hinfo->hash = EXT4_DIRENT_HASH(de);
1119 hinfo->minor_ 1119 hinfo->minor_hash = EXT4_DIRENT_MINOR_HASH(de);
1120 } else { 1120 } else {
1121 hinfo->hash = 1121 hinfo->hash = 0;
1122 hinfo->minor_ 1122 hinfo->minor_hash = 0;
1123 } 1123 }
1124 } else { 1124 } else {
1125 err = ext4fs_dirhash( 1125 err = ext4fs_dirhash(dir, de->name,
1126 1126 de->name_len, hinfo);
1127 if (err < 0) { 1127 if (err < 0) {
1128 count = err; 1128 count = err;
1129 goto errout; 1129 goto errout;
1130 } 1130 }
1131 } 1131 }
1132 if ((hinfo->hash < start_hash 1132 if ((hinfo->hash < start_hash) ||
1133 ((hinfo->hash == start_ha 1133 ((hinfo->hash == start_hash) &&
1134 (hinfo->minor_hash < sta 1134 (hinfo->minor_hash < start_minor_hash)))
1135 continue; 1135 continue;
1136 if (de->inode == 0) 1136 if (de->inode == 0)
1137 continue; 1137 continue;
1138 if (!IS_ENCRYPTED(dir)) { 1138 if (!IS_ENCRYPTED(dir)) {
1139 tmp_str.name = de->na 1139 tmp_str.name = de->name;
1140 tmp_str.len = de->nam 1140 tmp_str.len = de->name_len;
1141 err = ext4_htree_stor 1141 err = ext4_htree_store_dirent(dir_file,
1142 hinfo->has 1142 hinfo->hash, hinfo->minor_hash, de,
1143 &tmp_str); 1143 &tmp_str);
1144 } else { 1144 } else {
1145 int save_len = fname_ 1145 int save_len = fname_crypto_str.len;
1146 struct fscrypt_str de 1146 struct fscrypt_str de_name = FSTR_INIT(de->name,
1147 1147 de->name_len);
1148 1148
1149 /* Directory is encry 1149 /* Directory is encrypted */
1150 err = fscrypt_fname_d 1150 err = fscrypt_fname_disk_to_usr(dir, hinfo->hash,
1151 hinfo 1151 hinfo->minor_hash, &de_name,
1152 &fnam 1152 &fname_crypto_str);
1153 if (err) { 1153 if (err) {
1154 count = err; 1154 count = err;
1155 goto errout; 1155 goto errout;
1156 } 1156 }
1157 err = ext4_htree_stor 1157 err = ext4_htree_store_dirent(dir_file,
1158 hinfo->has 1158 hinfo->hash, hinfo->minor_hash, de,
1159 &fnam 1159 &fname_crypto_str);
1160 fname_crypto_str.len 1160 fname_crypto_str.len = save_len;
1161 } 1161 }
1162 if (err != 0) { 1162 if (err != 0) {
1163 count = err; 1163 count = err;
1164 goto errout; 1164 goto errout;
1165 } 1165 }
1166 count++; 1166 count++;
1167 } 1167 }
1168 errout: 1168 errout:
1169 brelse(bh); 1169 brelse(bh);
1170 fscrypt_fname_free_buffer(&fname_cryp 1170 fscrypt_fname_free_buffer(&fname_crypto_str);
1171 return count; 1171 return count;
1172 } 1172 }
1173 1173
1174 1174
1175 /* 1175 /*
1176 * This function fills a red-black tree with 1176 * This function fills a red-black tree with information from a
1177 * directory. We start scanning the director 1177 * directory. We start scanning the directory in hash order, starting
1178 * at start_hash and start_minor_hash. 1178 * at start_hash and start_minor_hash.
1179 * 1179 *
1180 * This function returns the number of entrie 1180 * This function returns the number of entries inserted into the tree,
1181 * or a negative error code. 1181 * or a negative error code.
1182 */ 1182 */
1183 int ext4_htree_fill_tree(struct file *dir_fil 1183 int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1184 __u32 start_minor_ha 1184 __u32 start_minor_hash, __u32 *next_hash)
1185 { 1185 {
1186 struct dx_hash_info hinfo; 1186 struct dx_hash_info hinfo;
1187 struct ext4_dir_entry_2 *de; 1187 struct ext4_dir_entry_2 *de;
1188 struct dx_frame frames[EXT4_HTREE_LEV 1188 struct dx_frame frames[EXT4_HTREE_LEVEL], *frame;
1189 struct inode *dir; 1189 struct inode *dir;
1190 ext4_lblk_t block; 1190 ext4_lblk_t block;
1191 int count = 0; 1191 int count = 0;
1192 int ret, err; 1192 int ret, err;
1193 __u32 hashval; 1193 __u32 hashval;
1194 struct fscrypt_str tmp_str; 1194 struct fscrypt_str tmp_str;
1195 1195
1196 dxtrace(printk(KERN_DEBUG "In htree_f 1196 dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n",
1197 start_hash, start_mino 1197 start_hash, start_minor_hash));
1198 dir = file_inode(dir_file); 1198 dir = file_inode(dir_file);
1199 if (!(ext4_test_inode_flag(dir, EXT4_ 1199 if (!(ext4_test_inode_flag(dir, EXT4_INODE_INDEX))) {
1200 if (ext4_hash_in_dirent(dir)) 1200 if (ext4_hash_in_dirent(dir))
1201 hinfo.hash_version = 1201 hinfo.hash_version = DX_HASH_SIPHASH;
1202 else 1202 else
1203 hinfo.hash_version = 1203 hinfo.hash_version =
1204 EXT4_ 1204 EXT4_SB(dir->i_sb)->s_def_hash_version;
1205 if (hinfo.hash_version <= DX_ 1205 if (hinfo.hash_version <= DX_HASH_TEA)
1206 hinfo.hash_version += 1206 hinfo.hash_version +=
1207 EXT4_SB(dir-> 1207 EXT4_SB(dir->i_sb)->s_hash_unsigned;
1208 hinfo.seed = EXT4_SB(dir->i_s 1208 hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
1209 if (ext4_has_inline_data(dir) 1209 if (ext4_has_inline_data(dir)) {
1210 int has_inline_data = 1210 int has_inline_data = 1;
1211 count = ext4_inlinedi 1211 count = ext4_inlinedir_to_tree(dir_file, dir, 0,
1212 1212 &hinfo, start_hash,
1213 1213 start_minor_hash,
1214 1214 &has_inline_data);
1215 if (has_inline_data) 1215 if (has_inline_data) {
1216 *next_hash = 1216 *next_hash = ~0;
1217 return count; 1217 return count;
1218 } 1218 }
1219 } 1219 }
1220 count = htree_dirblock_to_tre 1220 count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo,
1221 1221 start_hash, start_minor_hash);
1222 *next_hash = ~0; 1222 *next_hash = ~0;
1223 return count; 1223 return count;
1224 } 1224 }
1225 hinfo.hash = start_hash; 1225 hinfo.hash = start_hash;
1226 hinfo.minor_hash = 0; 1226 hinfo.minor_hash = 0;
1227 frame = dx_probe(NULL, dir, &hinfo, f 1227 frame = dx_probe(NULL, dir, &hinfo, frames);
1228 if (IS_ERR(frame)) 1228 if (IS_ERR(frame))
1229 return PTR_ERR(frame); 1229 return PTR_ERR(frame);
1230 1230
1231 /* Add '.' and '..' from the htree he 1231 /* Add '.' and '..' from the htree header */
1232 if (!start_hash && !start_minor_hash) 1232 if (!start_hash && !start_minor_hash) {
1233 de = (struct ext4_dir_entry_2 1233 de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data;
1234 tmp_str.name = de->name; 1234 tmp_str.name = de->name;
1235 tmp_str.len = de->name_len; 1235 tmp_str.len = de->name_len;
1236 err = ext4_htree_store_dirent 1236 err = ext4_htree_store_dirent(dir_file, 0, 0,
1237 1237 de, &tmp_str);
1238 if (err != 0) 1238 if (err != 0)
1239 goto errout; 1239 goto errout;
1240 count++; 1240 count++;
1241 } 1241 }
1242 if (start_hash < 2 || (start_hash ==2 1242 if (start_hash < 2 || (start_hash ==2 && start_minor_hash==0)) {
1243 de = (struct ext4_dir_entry_2 1243 de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data;
1244 de = ext4_next_entry(de, dir- 1244 de = ext4_next_entry(de, dir->i_sb->s_blocksize);
1245 tmp_str.name = de->name; 1245 tmp_str.name = de->name;
1246 tmp_str.len = de->name_len; 1246 tmp_str.len = de->name_len;
1247 err = ext4_htree_store_dirent 1247 err = ext4_htree_store_dirent(dir_file, 2, 0,
1248 1248 de, &tmp_str);
1249 if (err != 0) 1249 if (err != 0)
1250 goto errout; 1250 goto errout;
1251 count++; 1251 count++;
1252 } 1252 }
1253 1253
1254 while (1) { 1254 while (1) {
1255 if (fatal_signal_pending(curr 1255 if (fatal_signal_pending(current)) {
1256 err = -ERESTARTSYS; 1256 err = -ERESTARTSYS;
1257 goto errout; 1257 goto errout;
1258 } 1258 }
1259 cond_resched(); 1259 cond_resched();
1260 block = dx_get_block(frame->a 1260 block = dx_get_block(frame->at);
1261 ret = htree_dirblock_to_tree( 1261 ret = htree_dirblock_to_tree(dir_file, dir, block, &hinfo,
1262 1262 start_hash, start_minor_hash);
1263 if (ret < 0) { 1263 if (ret < 0) {
1264 err = ret; 1264 err = ret;
1265 goto errout; 1265 goto errout;
1266 } 1266 }
1267 count += ret; 1267 count += ret;
1268 hashval = ~0; 1268 hashval = ~0;
1269 ret = ext4_htree_next_block(d 1269 ret = ext4_htree_next_block(dir, HASH_NB_ALWAYS,
1270 f 1270 frame, frames, &hashval);
1271 *next_hash = hashval; 1271 *next_hash = hashval;
1272 if (ret < 0) { 1272 if (ret < 0) {
1273 err = ret; 1273 err = ret;
1274 goto errout; 1274 goto errout;
1275 } 1275 }
1276 /* 1276 /*
1277 * Stop if: (a) there are no 1277 * Stop if: (a) there are no more entries, or
1278 * (b) we have inserted at le 1278 * (b) we have inserted at least one entry and the
1279 * next hash value is not a c 1279 * next hash value is not a continuation
1280 */ 1280 */
1281 if ((ret == 0) || 1281 if ((ret == 0) ||
1282 (count && ((hashval & 1) 1282 (count && ((hashval & 1) == 0)))
1283 break; 1283 break;
1284 } 1284 }
1285 dx_release(frames); 1285 dx_release(frames);
1286 dxtrace(printk(KERN_DEBUG "Fill tree: 1286 dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, "
1287 "next hash: %x\n", cou 1287 "next hash: %x\n", count, *next_hash));
1288 return count; 1288 return count;
1289 errout: 1289 errout:
1290 dx_release(frames); 1290 dx_release(frames);
1291 return (err); 1291 return (err);
1292 } 1292 }
1293 1293
1294 static inline int search_dirblock(struct buff 1294 static inline int search_dirblock(struct buffer_head *bh,
1295 struct inod 1295 struct inode *dir,
1296 struct ext4 1296 struct ext4_filename *fname,
1297 unsigned in 1297 unsigned int offset,
1298 struct ext4 1298 struct ext4_dir_entry_2 **res_dir)
1299 { 1299 {
1300 return ext4_search_dir(bh, bh->b_data 1300 return ext4_search_dir(bh, bh->b_data, dir->i_sb->s_blocksize, dir,
1301 fname, offset, 1301 fname, offset, res_dir);
1302 } 1302 }
1303 1303
1304 /* 1304 /*
1305 * Directory block splitting, compacting 1305 * Directory block splitting, compacting
1306 */ 1306 */
1307 1307
1308 /* 1308 /*
1309 * Create map of hash values, offsets, and si 1309 * Create map of hash values, offsets, and sizes, stored at end of block.
1310 * Returns number of entries mapped. 1310 * Returns number of entries mapped.
1311 */ 1311 */
1312 static int dx_make_map(struct inode *dir, str 1312 static int dx_make_map(struct inode *dir, struct buffer_head *bh,
1313 struct dx_hash_info *h 1313 struct dx_hash_info *hinfo,
1314 struct dx_map_entry *m 1314 struct dx_map_entry *map_tail)
1315 { 1315 {
1316 int count = 0; 1316 int count = 0;
1317 struct ext4_dir_entry_2 *de = (struct 1317 struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *)bh->b_data;
1318 unsigned int buflen = bh->b_size; 1318 unsigned int buflen = bh->b_size;
1319 char *base = bh->b_data; 1319 char *base = bh->b_data;
1320 struct dx_hash_info h = *hinfo; 1320 struct dx_hash_info h = *hinfo;
1321 int blocksize = EXT4_BLOCK_SIZE(dir-> 1321 int blocksize = EXT4_BLOCK_SIZE(dir->i_sb);
1322 1322
1323 if (ext4_has_metadata_csum(dir->i_sb) 1323 if (ext4_has_metadata_csum(dir->i_sb))
1324 buflen -= sizeof(struct ext4_ 1324 buflen -= sizeof(struct ext4_dir_entry_tail);
1325 1325
1326 while ((char *) de < base + buflen) { 1326 while ((char *) de < base + buflen) {
1327 if (ext4_check_dir_entry(dir, 1327 if (ext4_check_dir_entry(dir, NULL, de, bh, base, buflen,
1328 ((ch 1328 ((char *)de) - base))
1329 return -EFSCORRUPTED; 1329 return -EFSCORRUPTED;
1330 if (de->name_len && de->inode 1330 if (de->name_len && de->inode) {
1331 if (ext4_hash_in_dire 1331 if (ext4_hash_in_dirent(dir))
1332 h.hash = EXT4 1332 h.hash = EXT4_DIRENT_HASH(de);
1333 else { 1333 else {
1334 int err = ext 1334 int err = ext4fs_dirhash(dir, de->name,
1335 1335 de->name_len, &h);
1336 if (err < 0) 1336 if (err < 0)
1337 retur 1337 return err;
1338 } 1338 }
1339 map_tail--; 1339 map_tail--;
1340 map_tail->hash = h.ha 1340 map_tail->hash = h.hash;
1341 map_tail->offs = ((ch 1341 map_tail->offs = ((char *) de - base)>>2;
1342 map_tail->size = ext4 1342 map_tail->size = ext4_rec_len_from_disk(de->rec_len,
1343 1343 blocksize);
1344 count++; 1344 count++;
1345 cond_resched(); 1345 cond_resched();
1346 } 1346 }
1347 de = ext4_next_entry(de, bloc 1347 de = ext4_next_entry(de, blocksize);
1348 } 1348 }
1349 return count; 1349 return count;
1350 } 1350 }
1351 1351
1352 /* Sort map by hash value */ 1352 /* Sort map by hash value */
1353 static void dx_sort_map (struct dx_map_entry 1353 static void dx_sort_map (struct dx_map_entry *map, unsigned count)
1354 { 1354 {
1355 struct dx_map_entry *p, *q, *top = ma 1355 struct dx_map_entry *p, *q, *top = map + count - 1;
1356 int more; 1356 int more;
1357 /* Combsort until bubble sort doesn't 1357 /* Combsort until bubble sort doesn't suck */
1358 while (count > 2) { 1358 while (count > 2) {
1359 count = count*10/13; 1359 count = count*10/13;
1360 if (count - 9 < 2) /* 9, 10 - 1360 if (count - 9 < 2) /* 9, 10 -> 11 */
1361 count = 11; 1361 count = 11;
1362 for (p = top, q = p - count; 1362 for (p = top, q = p - count; q >= map; p--, q--)
1363 if (p->hash < q->hash 1363 if (p->hash < q->hash)
1364 swap(*p, *q); 1364 swap(*p, *q);
1365 } 1365 }
1366 /* Garden variety bubble sort */ 1366 /* Garden variety bubble sort */
1367 do { 1367 do {
1368 more = 0; 1368 more = 0;
1369 q = top; 1369 q = top;
1370 while (q-- > map) { 1370 while (q-- > map) {
1371 if (q[1].hash >= q[0] 1371 if (q[1].hash >= q[0].hash)
1372 continue; 1372 continue;
1373 swap(*(q+1), *q); 1373 swap(*(q+1), *q);
1374 more = 1; 1374 more = 1;
1375 } 1375 }
1376 } while(more); 1376 } while(more);
1377 } 1377 }
1378 1378
1379 static void dx_insert_block(struct dx_frame * 1379 static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block)
1380 { 1380 {
1381 struct dx_entry *entries = frame->ent 1381 struct dx_entry *entries = frame->entries;
1382 struct dx_entry *old = frame->at, *ne 1382 struct dx_entry *old = frame->at, *new = old + 1;
1383 int count = dx_get_count(entries); 1383 int count = dx_get_count(entries);
1384 1384
1385 ASSERT(count < dx_get_limit(entries)) 1385 ASSERT(count < dx_get_limit(entries));
1386 ASSERT(old < entries + count); 1386 ASSERT(old < entries + count);
1387 memmove(new + 1, new, (char *)(entrie 1387 memmove(new + 1, new, (char *)(entries + count) - (char *)(new));
1388 dx_set_hash(new, hash); 1388 dx_set_hash(new, hash);
1389 dx_set_block(new, block); 1389 dx_set_block(new, block);
1390 dx_set_count(entries, count + 1); 1390 dx_set_count(entries, count + 1);
1391 } 1391 }
1392 1392
1393 #if IS_ENABLED(CONFIG_UNICODE) 1393 #if IS_ENABLED(CONFIG_UNICODE)
1394 int ext4_fname_setup_ci_filename(struct inode 1394 int ext4_fname_setup_ci_filename(struct inode *dir, const struct qstr *iname,
1395 struct ext4 1395 struct ext4_filename *name)
1396 { 1396 {
1397 struct qstr *cf_name = &name->cf_name 1397 struct qstr *cf_name = &name->cf_name;
1398 unsigned char *buf; 1398 unsigned char *buf;
1399 struct dx_hash_info *hinfo = &name->h 1399 struct dx_hash_info *hinfo = &name->hinfo;
1400 int len; 1400 int len;
1401 1401
1402 if (!IS_CASEFOLDED(dir) || 1402 if (!IS_CASEFOLDED(dir) ||
1403 (IS_ENCRYPTED(dir) && !fscrypt_ha 1403 (IS_ENCRYPTED(dir) && !fscrypt_has_encryption_key(dir))) {
1404 cf_name->name = NULL; 1404 cf_name->name = NULL;
1405 return 0; 1405 return 0;
1406 } 1406 }
1407 1407
1408 buf = kmalloc(EXT4_NAME_LEN, GFP_NOFS 1408 buf = kmalloc(EXT4_NAME_LEN, GFP_NOFS);
1409 if (!buf) 1409 if (!buf)
1410 return -ENOMEM; 1410 return -ENOMEM;
1411 1411
1412 len = utf8_casefold(dir->i_sb->s_enco 1412 len = utf8_casefold(dir->i_sb->s_encoding, iname, buf, EXT4_NAME_LEN);
1413 if (len <= 0) { 1413 if (len <= 0) {
1414 kfree(buf); 1414 kfree(buf);
1415 buf = NULL; 1415 buf = NULL;
1416 } 1416 }
1417 cf_name->name = buf; 1417 cf_name->name = buf;
1418 cf_name->len = (unsigned) len; 1418 cf_name->len = (unsigned) len;
1419 1419
1420 if (!IS_ENCRYPTED(dir)) 1420 if (!IS_ENCRYPTED(dir))
1421 return 0; 1421 return 0;
1422 1422
1423 hinfo->hash_version = DX_HASH_SIPHASH 1423 hinfo->hash_version = DX_HASH_SIPHASH;
1424 hinfo->seed = NULL; 1424 hinfo->seed = NULL;
1425 if (cf_name->name) 1425 if (cf_name->name)
1426 return ext4fs_dirhash(dir, cf 1426 return ext4fs_dirhash(dir, cf_name->name, cf_name->len, hinfo);
1427 else 1427 else
1428 return ext4fs_dirhash(dir, in 1428 return ext4fs_dirhash(dir, iname->name, iname->len, hinfo);
1429 } 1429 }
1430 #endif 1430 #endif
1431 1431
1432 /* 1432 /*
1433 * Test whether a directory entry matches the 1433 * Test whether a directory entry matches the filename being searched for.
1434 * 1434 *
1435 * Return: %true if the directory entry match 1435 * Return: %true if the directory entry matches, otherwise %false.
1436 */ 1436 */
1437 static bool ext4_match(struct inode *parent, 1437 static bool ext4_match(struct inode *parent,
1438 const struct ex 1438 const struct ext4_filename *fname,
1439 struct ext4_dir 1439 struct ext4_dir_entry_2 *de)
1440 { 1440 {
1441 struct fscrypt_name f; 1441 struct fscrypt_name f;
1442 1442
1443 if (!de->inode) 1443 if (!de->inode)
1444 return false; 1444 return false;
1445 1445
1446 f.usr_fname = fname->usr_fname; 1446 f.usr_fname = fname->usr_fname;
1447 f.disk_name = fname->disk_name; 1447 f.disk_name = fname->disk_name;
1448 #ifdef CONFIG_FS_ENCRYPTION 1448 #ifdef CONFIG_FS_ENCRYPTION
1449 f.crypto_buf = fname->crypto_buf; 1449 f.crypto_buf = fname->crypto_buf;
1450 #endif 1450 #endif
1451 1451
1452 #if IS_ENABLED(CONFIG_UNICODE) 1452 #if IS_ENABLED(CONFIG_UNICODE)
1453 if (IS_CASEFOLDED(parent) && 1453 if (IS_CASEFOLDED(parent) &&
1454 (!IS_ENCRYPTED(parent) || fscrypt 1454 (!IS_ENCRYPTED(parent) || fscrypt_has_encryption_key(parent))) {
1455 /* 1455 /*
1456 * Just checking IS_ENCRYPTED 1456 * Just checking IS_ENCRYPTED(parent) below is not
1457 * sufficient to decide wheth 1457 * sufficient to decide whether one can use the hash for
1458 * skipping the string compar 1458 * skipping the string comparison, because the key might
1459 * have been added right afte 1459 * have been added right after
1460 * ext4_fname_setup_ci_filena 1460 * ext4_fname_setup_ci_filename(). In this case, a hash
1461 * mismatch will be a false n 1461 * mismatch will be a false negative. Therefore, make
1462 * sure cf_name was properly 1462 * sure cf_name was properly initialized before
1463 * considering the calculated 1463 * considering the calculated hash.
1464 */ 1464 */
1465 if (IS_ENCRYPTED(parent) && f 1465 if (IS_ENCRYPTED(parent) && fname->cf_name.name &&
1466 (fname->hinfo.hash != EXT 1466 (fname->hinfo.hash != EXT4_DIRENT_HASH(de) ||
1467 fname->hinfo.minor_hash 1467 fname->hinfo.minor_hash != EXT4_DIRENT_MINOR_HASH(de)))
1468 return false; 1468 return false;
1469 /* 1469 /*
1470 * Treat comparison errors as 1470 * Treat comparison errors as not a match. The
1471 * only case where it happens 1471 * only case where it happens is on a disk
1472 * corruption or ENOMEM. 1472 * corruption or ENOMEM.
1473 */ 1473 */
1474 1474
1475 return generic_ci_match(paren 1475 return generic_ci_match(parent, fname->usr_fname,
1476 &fnam 1476 &fname->cf_name, de->name,
1477 de->n 1477 de->name_len) > 0;
1478 } 1478 }
1479 #endif 1479 #endif
1480 1480
1481 return fscrypt_match_name(&f, de->nam 1481 return fscrypt_match_name(&f, de->name, de->name_len);
1482 } 1482 }
1483 1483
1484 /* 1484 /*
1485 * Returns 0 if not found, -EFSCORRUPTED on f 1485 * Returns 0 if not found, -EFSCORRUPTED on failure, and 1 on success
1486 */ 1486 */
1487 int ext4_search_dir(struct buffer_head *bh, c 1487 int ext4_search_dir(struct buffer_head *bh, char *search_buf, int buf_size,
1488 struct inode *dir, struct 1488 struct inode *dir, struct ext4_filename *fname,
1489 unsigned int offset, stru 1489 unsigned int offset, struct ext4_dir_entry_2 **res_dir)
1490 { 1490 {
1491 struct ext4_dir_entry_2 * de; 1491 struct ext4_dir_entry_2 * de;
1492 char * dlimit; 1492 char * dlimit;
1493 int de_len; 1493 int de_len;
1494 1494
1495 de = (struct ext4_dir_entry_2 *)searc 1495 de = (struct ext4_dir_entry_2 *)search_buf;
1496 dlimit = search_buf + buf_size; 1496 dlimit = search_buf + buf_size;
1497 while ((char *) de < dlimit - EXT4_BA 1497 while ((char *) de < dlimit - EXT4_BASE_DIR_LEN) {
1498 /* this code is executed quad 1498 /* this code is executed quadratically often */
1499 /* do minimal checking `by ha 1499 /* do minimal checking `by hand' */
1500 if (de->name + de->name_len < 1500 if (de->name + de->name_len <= dlimit &&
1501 ext4_match(dir, fname, de 1501 ext4_match(dir, fname, de)) {
1502 /* found a match - ju 1502 /* found a match - just to be sure, do
1503 * a full check */ 1503 * a full check */
1504 if (ext4_check_dir_en 1504 if (ext4_check_dir_entry(dir, NULL, de, bh, search_buf,
1505 1505 buf_size, offset))
1506 return -EFSCO 1506 return -EFSCORRUPTED;
1507 *res_dir = de; 1507 *res_dir = de;
1508 return 1; 1508 return 1;
1509 } 1509 }
1510 /* prevent looping on a bad b 1510 /* prevent looping on a bad block */
1511 de_len = ext4_rec_len_from_di 1511 de_len = ext4_rec_len_from_disk(de->rec_len,
1512 1512 dir->i_sb->s_blocksize);
1513 if (de_len <= 0) 1513 if (de_len <= 0)
1514 return -EFSCORRUPTED; 1514 return -EFSCORRUPTED;
1515 offset += de_len; 1515 offset += de_len;
1516 de = (struct ext4_dir_entry_2 1516 de = (struct ext4_dir_entry_2 *) ((char *) de + de_len);
1517 } 1517 }
1518 return 0; 1518 return 0;
1519 } 1519 }
1520 1520
1521 static int is_dx_internal_node(struct inode * 1521 static int is_dx_internal_node(struct inode *dir, ext4_lblk_t block,
1522 struct ext4_di 1522 struct ext4_dir_entry *de)
1523 { 1523 {
1524 struct super_block *sb = dir->i_sb; 1524 struct super_block *sb = dir->i_sb;
1525 1525
1526 if (!is_dx(dir)) 1526 if (!is_dx(dir))
1527 return 0; 1527 return 0;
1528 if (block == 0) 1528 if (block == 0)
1529 return 1; 1529 return 1;
1530 if (de->inode == 0 && 1530 if (de->inode == 0 &&
1531 ext4_rec_len_from_disk(de->rec_le 1531 ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) ==
1532 sb->s_blocksize) 1532 sb->s_blocksize)
1533 return 1; 1533 return 1;
1534 return 0; 1534 return 0;
1535 } 1535 }
1536 1536
1537 /* 1537 /*
1538 * __ext4_find_entry() 1538 * __ext4_find_entry()
1539 * 1539 *
1540 * finds an entry in the specified directory 1540 * finds an entry in the specified directory with the wanted name. It
1541 * returns the cache buffer in which the entr 1541 * returns the cache buffer in which the entry was found, and the entry
1542 * itself (as a parameter - res_dir). It does 1542 * itself (as a parameter - res_dir). It does NOT read the inode of the
1543 * entry - you'll have to do that yourself if 1543 * entry - you'll have to do that yourself if you want to.
1544 * 1544 *
1545 * The returned buffer_head has ->b_count ele 1545 * The returned buffer_head has ->b_count elevated. The caller is expected
1546 * to brelse() it when appropriate. 1546 * to brelse() it when appropriate.
1547 */ 1547 */
1548 static struct buffer_head *__ext4_find_entry( 1548 static struct buffer_head *__ext4_find_entry(struct inode *dir,
1549 1549 struct ext4_filename *fname,
1550 1550 struct ext4_dir_entry_2 **res_dir,
1551 1551 int *inlined)
1552 { 1552 {
1553 struct super_block *sb; 1553 struct super_block *sb;
1554 struct buffer_head *bh_use[NAMEI_RA_S 1554 struct buffer_head *bh_use[NAMEI_RA_SIZE];
1555 struct buffer_head *bh, *ret = NULL; 1555 struct buffer_head *bh, *ret = NULL;
1556 ext4_lblk_t start, block; 1556 ext4_lblk_t start, block;
1557 const u8 *name = fname->usr_fname->na 1557 const u8 *name = fname->usr_fname->name;
1558 size_t ra_max = 0; /* Number of 1558 size_t ra_max = 0; /* Number of bh's in the readahead
1559 buffer, bh 1559 buffer, bh_use[] */
1560 size_t ra_ptr = 0; /* Current in 1560 size_t ra_ptr = 0; /* Current index into readahead
1561 buffer */ 1561 buffer */
1562 ext4_lblk_t nblocks; 1562 ext4_lblk_t nblocks;
1563 int i, namelen, retval; 1563 int i, namelen, retval;
1564 1564
1565 *res_dir = NULL; 1565 *res_dir = NULL;
1566 sb = dir->i_sb; 1566 sb = dir->i_sb;
1567 namelen = fname->usr_fname->len; 1567 namelen = fname->usr_fname->len;
1568 if (namelen > EXT4_NAME_LEN) 1568 if (namelen > EXT4_NAME_LEN)
1569 return NULL; 1569 return NULL;
1570 1570
1571 if (ext4_has_inline_data(dir)) { 1571 if (ext4_has_inline_data(dir)) {
1572 int has_inline_data = 1; 1572 int has_inline_data = 1;
1573 ret = ext4_find_inline_entry( 1573 ret = ext4_find_inline_entry(dir, fname, res_dir,
1574 1574 &has_inline_data);
1575 if (inlined) 1575 if (inlined)
1576 *inlined = has_inline 1576 *inlined = has_inline_data;
1577 if (has_inline_data || IS_ERR 1577 if (has_inline_data || IS_ERR(ret))
1578 goto cleanup_and_exit 1578 goto cleanup_and_exit;
1579 } 1579 }
1580 1580
1581 if ((namelen <= 2) && (name[0] == '.' 1581 if ((namelen <= 2) && (name[0] == '.') &&
1582 (name[1] == '.' || name[1] == '\0 1582 (name[1] == '.' || name[1] == '\0')) {
1583 /* 1583 /*
1584 * "." or ".." will only be i 1584 * "." or ".." will only be in the first block
1585 * NFS may look up ".."; "." 1585 * NFS may look up ".."; "." should be handled by the VFS
1586 */ 1586 */
1587 block = start = 0; 1587 block = start = 0;
1588 nblocks = 1; 1588 nblocks = 1;
1589 goto restart; 1589 goto restart;
1590 } 1590 }
1591 if (is_dx(dir)) { 1591 if (is_dx(dir)) {
1592 ret = ext4_dx_find_entry(dir, 1592 ret = ext4_dx_find_entry(dir, fname, res_dir);
1593 /* 1593 /*
1594 * On success, or if the erro 1594 * On success, or if the error was file not found,
1595 * return. Otherwise, fall b 1595 * return. Otherwise, fall back to doing a search the
1596 * old fashioned way. 1596 * old fashioned way.
1597 */ 1597 */
1598 if (!IS_ERR(ret) || PTR_ERR(r 1598 if (!IS_ERR(ret) || PTR_ERR(ret) != ERR_BAD_DX_DIR)
1599 goto cleanup_and_exit 1599 goto cleanup_and_exit;
1600 dxtrace(printk(KERN_DEBUG "ex 1600 dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, "
1601 "falling back\ 1601 "falling back\n"));
1602 ret = NULL; 1602 ret = NULL;
1603 } 1603 }
1604 nblocks = dir->i_size >> EXT4_BLOCK_S 1604 nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb);
1605 if (!nblocks) { 1605 if (!nblocks) {
1606 ret = NULL; 1606 ret = NULL;
1607 goto cleanup_and_exit; 1607 goto cleanup_and_exit;
1608 } 1608 }
1609 start = EXT4_I(dir)->i_dir_start_look 1609 start = EXT4_I(dir)->i_dir_start_lookup;
1610 if (start >= nblocks) 1610 if (start >= nblocks)
1611 start = 0; 1611 start = 0;
1612 block = start; 1612 block = start;
1613 restart: 1613 restart:
1614 do { 1614 do {
1615 /* 1615 /*
1616 * We deal with the read-ahea 1616 * We deal with the read-ahead logic here.
1617 */ 1617 */
1618 cond_resched(); 1618 cond_resched();
1619 if (ra_ptr >= ra_max) { 1619 if (ra_ptr >= ra_max) {
1620 /* Refill the readahe 1620 /* Refill the readahead buffer */
1621 ra_ptr = 0; 1621 ra_ptr = 0;
1622 if (block < start) 1622 if (block < start)
1623 ra_max = star 1623 ra_max = start - block;
1624 else 1624 else
1625 ra_max = nblo 1625 ra_max = nblocks - block;
1626 ra_max = min(ra_max, 1626 ra_max = min(ra_max, ARRAY_SIZE(bh_use));
1627 retval = ext4_bread_b 1627 retval = ext4_bread_batch(dir, block, ra_max,
1628 1628 false /* wait */, bh_use);
1629 if (retval) { 1629 if (retval) {
1630 ret = ERR_PTR 1630 ret = ERR_PTR(retval);
1631 ra_max = 0; 1631 ra_max = 0;
1632 goto cleanup_ 1632 goto cleanup_and_exit;
1633 } 1633 }
1634 } 1634 }
1635 if ((bh = bh_use[ra_ptr++]) = 1635 if ((bh = bh_use[ra_ptr++]) == NULL)
1636 goto next; 1636 goto next;
1637 wait_on_buffer(bh); 1637 wait_on_buffer(bh);
1638 if (!buffer_uptodate(bh)) { 1638 if (!buffer_uptodate(bh)) {
1639 EXT4_ERROR_INODE_ERR( 1639 EXT4_ERROR_INODE_ERR(dir, EIO,
1640 1640 "reading directory lblock %lu",
1641 1641 (unsigned long) block);
1642 brelse(bh); 1642 brelse(bh);
1643 ret = ERR_PTR(-EIO); 1643 ret = ERR_PTR(-EIO);
1644 goto cleanup_and_exit 1644 goto cleanup_and_exit;
1645 } 1645 }
1646 if (!buffer_verified(bh) && 1646 if (!buffer_verified(bh) &&
1647 !is_dx_internal_node(dir, 1647 !is_dx_internal_node(dir, block,
1648 (str 1648 (struct ext4_dir_entry *)bh->b_data) &&
1649 !ext4_dirblock_csum_verif 1649 !ext4_dirblock_csum_verify(dir, bh)) {
1650 EXT4_ERROR_INODE_ERR( 1650 EXT4_ERROR_INODE_ERR(dir, EFSBADCRC,
1651 1651 "checksumming directory "
1652 1652 "block %lu", (unsigned long)block);
1653 brelse(bh); 1653 brelse(bh);
1654 ret = ERR_PTR(-EFSBAD 1654 ret = ERR_PTR(-EFSBADCRC);
1655 goto cleanup_and_exit 1655 goto cleanup_and_exit;
1656 } 1656 }
1657 set_buffer_verified(bh); 1657 set_buffer_verified(bh);
1658 i = search_dirblock(bh, dir, 1658 i = search_dirblock(bh, dir, fname,
1659 block << EXT4_BLO 1659 block << EXT4_BLOCK_SIZE_BITS(sb), res_dir);
1660 if (i == 1) { 1660 if (i == 1) {
1661 EXT4_I(dir)->i_dir_st 1661 EXT4_I(dir)->i_dir_start_lookup = block;
1662 ret = bh; 1662 ret = bh;
1663 goto cleanup_and_exit 1663 goto cleanup_and_exit;
1664 } else { 1664 } else {
1665 brelse(bh); 1665 brelse(bh);
1666 if (i < 0) { 1666 if (i < 0) {
1667 ret = ERR_PTR 1667 ret = ERR_PTR(i);
1668 goto cleanup_ 1668 goto cleanup_and_exit;
1669 } 1669 }
1670 } 1670 }
1671 next: 1671 next:
1672 if (++block >= nblocks) 1672 if (++block >= nblocks)
1673 block = 0; 1673 block = 0;
1674 } while (block != start); 1674 } while (block != start);
1675 1675
1676 /* 1676 /*
1677 * If the directory has grown while w 1677 * If the directory has grown while we were searching, then
1678 * search the last part of the direct 1678 * search the last part of the directory before giving up.
1679 */ 1679 */
1680 block = nblocks; 1680 block = nblocks;
1681 nblocks = dir->i_size >> EXT4_BLOCK_S 1681 nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb);
1682 if (block < nblocks) { 1682 if (block < nblocks) {
1683 start = 0; 1683 start = 0;
1684 goto restart; 1684 goto restart;
1685 } 1685 }
1686 1686
1687 cleanup_and_exit: 1687 cleanup_and_exit:
1688 /* Clean up the read-ahead blocks */ 1688 /* Clean up the read-ahead blocks */
1689 for (; ra_ptr < ra_max; ra_ptr++) 1689 for (; ra_ptr < ra_max; ra_ptr++)
1690 brelse(bh_use[ra_ptr]); 1690 brelse(bh_use[ra_ptr]);
1691 return ret; 1691 return ret;
1692 } 1692 }
1693 1693
1694 static struct buffer_head *ext4_find_entry(st 1694 static struct buffer_head *ext4_find_entry(struct inode *dir,
1695 co 1695 const struct qstr *d_name,
1696 st 1696 struct ext4_dir_entry_2 **res_dir,
1697 in 1697 int *inlined)
1698 { 1698 {
1699 int err; 1699 int err;
1700 struct ext4_filename fname; 1700 struct ext4_filename fname;
1701 struct buffer_head *bh; 1701 struct buffer_head *bh;
1702 1702
1703 err = ext4_fname_setup_filename(dir, 1703 err = ext4_fname_setup_filename(dir, d_name, 1, &fname);
1704 if (err == -ENOENT) 1704 if (err == -ENOENT)
1705 return NULL; 1705 return NULL;
1706 if (err) 1706 if (err)
1707 return ERR_PTR(err); 1707 return ERR_PTR(err);
1708 1708
1709 bh = __ext4_find_entry(dir, &fname, r 1709 bh = __ext4_find_entry(dir, &fname, res_dir, inlined);
1710 1710
1711 ext4_fname_free_filename(&fname); 1711 ext4_fname_free_filename(&fname);
1712 return bh; 1712 return bh;
1713 } 1713 }
1714 1714
1715 static struct buffer_head *ext4_lookup_entry( 1715 static struct buffer_head *ext4_lookup_entry(struct inode *dir,
1716 1716 struct dentry *dentry,
1717 1717 struct ext4_dir_entry_2 **res_dir)
1718 { 1718 {
1719 int err; 1719 int err;
1720 struct ext4_filename fname; 1720 struct ext4_filename fname;
1721 struct buffer_head *bh; 1721 struct buffer_head *bh;
1722 1722
1723 err = ext4_fname_prepare_lookup(dir, 1723 err = ext4_fname_prepare_lookup(dir, dentry, &fname);
1724 if (err == -ENOENT) 1724 if (err == -ENOENT)
1725 return NULL; 1725 return NULL;
1726 if (err) 1726 if (err)
1727 return ERR_PTR(err); 1727 return ERR_PTR(err);
1728 1728
1729 bh = __ext4_find_entry(dir, &fname, r 1729 bh = __ext4_find_entry(dir, &fname, res_dir, NULL);
1730 1730
1731 ext4_fname_free_filename(&fname); 1731 ext4_fname_free_filename(&fname);
1732 return bh; 1732 return bh;
1733 } 1733 }
1734 1734
1735 static struct buffer_head * ext4_dx_find_entr 1735 static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
1736 struct ext4_filename 1736 struct ext4_filename *fname,
1737 struct ext4_dir_entry 1737 struct ext4_dir_entry_2 **res_dir)
1738 { 1738 {
1739 struct super_block * sb = dir->i_sb; 1739 struct super_block * sb = dir->i_sb;
1740 struct dx_frame frames[EXT4_HTREE_LEV 1740 struct dx_frame frames[EXT4_HTREE_LEVEL], *frame;
1741 struct buffer_head *bh; 1741 struct buffer_head *bh;
1742 ext4_lblk_t block; 1742 ext4_lblk_t block;
1743 int retval; 1743 int retval;
1744 1744
1745 #ifdef CONFIG_FS_ENCRYPTION 1745 #ifdef CONFIG_FS_ENCRYPTION
1746 *res_dir = NULL; 1746 *res_dir = NULL;
1747 #endif 1747 #endif
1748 frame = dx_probe(fname, dir, NULL, fr 1748 frame = dx_probe(fname, dir, NULL, frames);
1749 if (IS_ERR(frame)) 1749 if (IS_ERR(frame))
1750 return (struct buffer_head *) 1750 return (struct buffer_head *) frame;
1751 do { 1751 do {
1752 block = dx_get_block(frame->a 1752 block = dx_get_block(frame->at);
1753 bh = ext4_read_dirblock(dir, 1753 bh = ext4_read_dirblock(dir, block, DIRENT_HTREE);
1754 if (IS_ERR(bh)) 1754 if (IS_ERR(bh))
1755 goto errout; 1755 goto errout;
1756 1756
1757 retval = search_dirblock(bh, 1757 retval = search_dirblock(bh, dir, fname,
1758 bloc 1758 block << EXT4_BLOCK_SIZE_BITS(sb),
1759 res_ 1759 res_dir);
1760 if (retval == 1) 1760 if (retval == 1)
1761 goto success; 1761 goto success;
1762 brelse(bh); 1762 brelse(bh);
1763 if (retval < 0) { 1763 if (retval < 0) {
1764 bh = ERR_PTR(ERR_BAD_ 1764 bh = ERR_PTR(ERR_BAD_DX_DIR);
1765 goto errout; 1765 goto errout;
1766 } 1766 }
1767 1767
1768 /* Check to see if we should 1768 /* Check to see if we should continue to search */
1769 retval = ext4_htree_next_bloc 1769 retval = ext4_htree_next_block(dir, fname->hinfo.hash, frame,
1770 1770 frames, NULL);
1771 if (retval < 0) { 1771 if (retval < 0) {
1772 ext4_warning_inode(di 1772 ext4_warning_inode(dir,
1773 "error %d rea 1773 "error %d reading directory index block",
1774 retval); 1774 retval);
1775 bh = ERR_PTR(retval); 1775 bh = ERR_PTR(retval);
1776 goto errout; 1776 goto errout;
1777 } 1777 }
1778 } while (retval == 1); 1778 } while (retval == 1);
1779 1779
1780 bh = NULL; 1780 bh = NULL;
1781 errout: 1781 errout:
1782 dxtrace(printk(KERN_DEBUG "%s not fou 1782 dxtrace(printk(KERN_DEBUG "%s not found\n", fname->usr_fname->name));
1783 success: 1783 success:
1784 dx_release(frames); 1784 dx_release(frames);
1785 return bh; 1785 return bh;
1786 } 1786 }
1787 1787
1788 static struct dentry *ext4_lookup(struct inod 1788 static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags)
1789 { 1789 {
1790 struct inode *inode; 1790 struct inode *inode;
1791 struct ext4_dir_entry_2 *de; 1791 struct ext4_dir_entry_2 *de;
1792 struct buffer_head *bh; 1792 struct buffer_head *bh;
1793 1793
1794 if (dentry->d_name.len > EXT4_NAME_LE 1794 if (dentry->d_name.len > EXT4_NAME_LEN)
1795 return ERR_PTR(-ENAMETOOLONG) 1795 return ERR_PTR(-ENAMETOOLONG);
1796 1796
1797 bh = ext4_lookup_entry(dir, dentry, & 1797 bh = ext4_lookup_entry(dir, dentry, &de);
1798 if (IS_ERR(bh)) 1798 if (IS_ERR(bh))
1799 return ERR_CAST(bh); 1799 return ERR_CAST(bh);
1800 inode = NULL; 1800 inode = NULL;
1801 if (bh) { 1801 if (bh) {
1802 __u32 ino = le32_to_cpu(de->i 1802 __u32 ino = le32_to_cpu(de->inode);
1803 brelse(bh); 1803 brelse(bh);
1804 if (!ext4_valid_inum(dir->i_s 1804 if (!ext4_valid_inum(dir->i_sb, ino)) {
1805 EXT4_ERROR_INODE(dir, 1805 EXT4_ERROR_INODE(dir, "bad inode number: %u", ino);
1806 return ERR_PTR(-EFSCO 1806 return ERR_PTR(-EFSCORRUPTED);
1807 } 1807 }
1808 if (unlikely(ino == dir->i_in 1808 if (unlikely(ino == dir->i_ino)) {
1809 EXT4_ERROR_INODE(dir, 1809 EXT4_ERROR_INODE(dir, "'%pd' linked to parent dir",
1810 dent 1810 dentry);
1811 return ERR_PTR(-EFSCO 1811 return ERR_PTR(-EFSCORRUPTED);
1812 } 1812 }
1813 inode = ext4_iget(dir->i_sb, 1813 inode = ext4_iget(dir->i_sb, ino, EXT4_IGET_NORMAL);
1814 if (inode == ERR_PTR(-ESTALE) 1814 if (inode == ERR_PTR(-ESTALE)) {
1815 EXT4_ERROR_INODE(dir, 1815 EXT4_ERROR_INODE(dir,
1816 "del 1816 "deleted inode referenced: %u",
1817 ino) 1817 ino);
1818 return ERR_PTR(-EFSCO 1818 return ERR_PTR(-EFSCORRUPTED);
1819 } 1819 }
1820 if (!IS_ERR(inode) && IS_ENCR 1820 if (!IS_ERR(inode) && IS_ENCRYPTED(dir) &&
1821 (S_ISDIR(inode->i_mode) | 1821 (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) &&
1822 !fscrypt_has_permitted_co 1822 !fscrypt_has_permitted_context(dir, inode)) {
1823 ext4_warning(inode->i 1823 ext4_warning(inode->i_sb,
1824 "Inconsi 1824 "Inconsistent encryption contexts: %lu/%lu",
1825 dir->i_i 1825 dir->i_ino, inode->i_ino);
1826 iput(inode); 1826 iput(inode);
1827 return ERR_PTR(-EPERM 1827 return ERR_PTR(-EPERM);
1828 } 1828 }
1829 } 1829 }
1830 1830
1831 if (IS_ENABLED(CONFIG_UNICODE) && !in 1831 if (IS_ENABLED(CONFIG_UNICODE) && !inode && IS_CASEFOLDED(dir)) {
1832 /* Eventually we want to call 1832 /* Eventually we want to call d_add_ci(dentry, NULL)
1833 * for negative dentries in t 1833 * for negative dentries in the encoding case as
1834 * well. For now, prevent th 1834 * well. For now, prevent the negative dentry
1835 * from being cached. 1835 * from being cached.
1836 */ 1836 */
1837 return NULL; 1837 return NULL;
1838 } 1838 }
1839 1839
1840 return d_splice_alias(inode, dentry); 1840 return d_splice_alias(inode, dentry);
1841 } 1841 }
1842 1842
1843 1843
1844 struct dentry *ext4_get_parent(struct dentry 1844 struct dentry *ext4_get_parent(struct dentry *child)
1845 { 1845 {
1846 __u32 ino; 1846 __u32 ino;
1847 struct ext4_dir_entry_2 * de; 1847 struct ext4_dir_entry_2 * de;
1848 struct buffer_head *bh; 1848 struct buffer_head *bh;
1849 1849
1850 bh = ext4_find_entry(d_inode(child), 1850 bh = ext4_find_entry(d_inode(child), &dotdot_name, &de, NULL);
1851 if (IS_ERR(bh)) 1851 if (IS_ERR(bh))
1852 return ERR_CAST(bh); 1852 return ERR_CAST(bh);
1853 if (!bh) 1853 if (!bh)
1854 return ERR_PTR(-ENOENT); 1854 return ERR_PTR(-ENOENT);
1855 ino = le32_to_cpu(de->inode); 1855 ino = le32_to_cpu(de->inode);
1856 brelse(bh); 1856 brelse(bh);
1857 1857
1858 if (!ext4_valid_inum(child->d_sb, ino 1858 if (!ext4_valid_inum(child->d_sb, ino)) {
1859 EXT4_ERROR_INODE(d_inode(chil 1859 EXT4_ERROR_INODE(d_inode(child),
1860 "bad parent 1860 "bad parent inode number: %u", ino);
1861 return ERR_PTR(-EFSCORRUPTED) 1861 return ERR_PTR(-EFSCORRUPTED);
1862 } 1862 }
1863 1863
1864 return d_obtain_alias(ext4_iget(child 1864 return d_obtain_alias(ext4_iget(child->d_sb, ino, EXT4_IGET_NORMAL));
1865 } 1865 }
1866 1866
1867 /* 1867 /*
1868 * Move count entries from end of map between 1868 * Move count entries from end of map between two memory locations.
1869 * Returns pointer to last entry moved. 1869 * Returns pointer to last entry moved.
1870 */ 1870 */
1871 static struct ext4_dir_entry_2 * 1871 static struct ext4_dir_entry_2 *
1872 dx_move_dirents(struct inode *dir, char *from 1872 dx_move_dirents(struct inode *dir, char *from, char *to,
1873 struct dx_map_entry *map, int 1873 struct dx_map_entry *map, int count,
1874 unsigned blocksize) 1874 unsigned blocksize)
1875 { 1875 {
1876 unsigned rec_len = 0; 1876 unsigned rec_len = 0;
1877 1877
1878 while (count--) { 1878 while (count--) {
1879 struct ext4_dir_entry_2 *de = 1879 struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *)
1880 1880 (from + (map->offs<<2));
1881 rec_len = ext4_dir_rec_len(de 1881 rec_len = ext4_dir_rec_len(de->name_len, dir);
1882 1882
1883 memcpy (to, de, rec_len); 1883 memcpy (to, de, rec_len);
1884 ((struct ext4_dir_entry_2 *) 1884 ((struct ext4_dir_entry_2 *) to)->rec_len =
1885 ext4_rec_len_ 1885 ext4_rec_len_to_disk(rec_len, blocksize);
1886 1886
1887 /* wipe dir_entry excluding t 1887 /* wipe dir_entry excluding the rec_len field */
1888 de->inode = 0; 1888 de->inode = 0;
1889 memset(&de->name_len, 0, ext4 1889 memset(&de->name_len, 0, ext4_rec_len_from_disk(de->rec_len,
1890 1890 blocksize) -
1891 offs 1891 offsetof(struct ext4_dir_entry_2,
1892 1892 name_len));
1893 1893
1894 map++; 1894 map++;
1895 to += rec_len; 1895 to += rec_len;
1896 } 1896 }
1897 return (struct ext4_dir_entry_2 *) (t 1897 return (struct ext4_dir_entry_2 *) (to - rec_len);
1898 } 1898 }
1899 1899
1900 /* 1900 /*
1901 * Compact each dir entry in the range to the 1901 * Compact each dir entry in the range to the minimal rec_len.
1902 * Returns pointer to last entry in range. 1902 * Returns pointer to last entry in range.
1903 */ 1903 */
1904 static struct ext4_dir_entry_2 *dx_pack_diren 1904 static struct ext4_dir_entry_2 *dx_pack_dirents(struct inode *dir, char *base,
1905 1905 unsigned int blocksize)
1906 { 1906 {
1907 struct ext4_dir_entry_2 *next, *to, * 1907 struct ext4_dir_entry_2 *next, *to, *prev, *de = (struct ext4_dir_entry_2 *) base;
1908 unsigned rec_len = 0; 1908 unsigned rec_len = 0;
1909 1909
1910 prev = to = de; 1910 prev = to = de;
1911 while ((char*)de < base + blocksize) 1911 while ((char*)de < base + blocksize) {
1912 next = ext4_next_entry(de, bl 1912 next = ext4_next_entry(de, blocksize);
1913 if (de->inode && de->name_len 1913 if (de->inode && de->name_len) {
1914 rec_len = ext4_dir_re 1914 rec_len = ext4_dir_rec_len(de->name_len, dir);
1915 if (de > to) 1915 if (de > to)
1916 memmove(to, d 1916 memmove(to, de, rec_len);
1917 to->rec_len = ext4_re 1917 to->rec_len = ext4_rec_len_to_disk(rec_len, blocksize);
1918 prev = to; 1918 prev = to;
1919 to = (struct ext4_dir 1919 to = (struct ext4_dir_entry_2 *) (((char *) to) + rec_len);
1920 } 1920 }
1921 de = next; 1921 de = next;
1922 } 1922 }
1923 return prev; 1923 return prev;
1924 } 1924 }
1925 1925
1926 /* 1926 /*
1927 * Split a full leaf block to make room for a 1927 * Split a full leaf block to make room for a new dir entry.
1928 * Allocate a new block, and move entries so 1928 * Allocate a new block, and move entries so that they are approx. equally full.
1929 * Returns pointer to de in block into which 1929 * Returns pointer to de in block into which the new entry will be inserted.
1930 */ 1930 */
1931 static struct ext4_dir_entry_2 *do_split(hand 1931 static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir,
1932 struct buffer_head ** 1932 struct buffer_head **bh,struct dx_frame *frame,
1933 struct dx_hash_info * 1933 struct dx_hash_info *hinfo)
1934 { 1934 {
1935 unsigned blocksize = dir->i_sb->s_blo 1935 unsigned blocksize = dir->i_sb->s_blocksize;
1936 unsigned continued; 1936 unsigned continued;
1937 int count; 1937 int count;
1938 struct buffer_head *bh2; 1938 struct buffer_head *bh2;
1939 ext4_lblk_t newblock; 1939 ext4_lblk_t newblock;
1940 u32 hash2; 1940 u32 hash2;
1941 struct dx_map_entry *map; 1941 struct dx_map_entry *map;
1942 char *data1 = (*bh)->b_data, *data2; 1942 char *data1 = (*bh)->b_data, *data2;
1943 unsigned split, move, size; 1943 unsigned split, move, size;
1944 struct ext4_dir_entry_2 *de = NULL, * 1944 struct ext4_dir_entry_2 *de = NULL, *de2;
1945 int csum_size = 0; 1945 int csum_size = 0;
1946 int err = 0, i; 1946 int err = 0, i;
1947 1947
1948 if (ext4_has_metadata_csum(dir->i_sb) 1948 if (ext4_has_metadata_csum(dir->i_sb))
1949 csum_size = sizeof(struct ext 1949 csum_size = sizeof(struct ext4_dir_entry_tail);
1950 1950
1951 bh2 = ext4_append(handle, dir, &newbl 1951 bh2 = ext4_append(handle, dir, &newblock);
1952 if (IS_ERR(bh2)) { 1952 if (IS_ERR(bh2)) {
1953 brelse(*bh); 1953 brelse(*bh);
1954 *bh = NULL; 1954 *bh = NULL;
1955 return (struct ext4_dir_entry 1955 return (struct ext4_dir_entry_2 *) bh2;
1956 } 1956 }
1957 1957
1958 BUFFER_TRACE(*bh, "get_write_access") 1958 BUFFER_TRACE(*bh, "get_write_access");
1959 err = ext4_journal_get_write_access(h 1959 err = ext4_journal_get_write_access(handle, dir->i_sb, *bh,
1960 E 1960 EXT4_JTR_NONE);
1961 if (err) 1961 if (err)
1962 goto journal_error; 1962 goto journal_error;
1963 1963
1964 BUFFER_TRACE(frame->bh, "get_write_ac 1964 BUFFER_TRACE(frame->bh, "get_write_access");
1965 err = ext4_journal_get_write_access(h 1965 err = ext4_journal_get_write_access(handle, dir->i_sb, frame->bh,
1966 E 1966 EXT4_JTR_NONE);
1967 if (err) 1967 if (err)
1968 goto journal_error; 1968 goto journal_error;
1969 1969
1970 data2 = bh2->b_data; 1970 data2 = bh2->b_data;
1971 1971
1972 /* create map in the end of data2 blo 1972 /* create map in the end of data2 block */
1973 map = (struct dx_map_entry *) (data2 1973 map = (struct dx_map_entry *) (data2 + blocksize);
1974 count = dx_make_map(dir, *bh, hinfo, 1974 count = dx_make_map(dir, *bh, hinfo, map);
1975 if (count < 0) { 1975 if (count < 0) {
1976 err = count; 1976 err = count;
1977 goto journal_error; 1977 goto journal_error;
1978 } 1978 }
1979 map -= count; 1979 map -= count;
1980 dx_sort_map(map, count); 1980 dx_sort_map(map, count);
1981 /* Ensure that neither split block is 1981 /* Ensure that neither split block is over half full */
1982 size = 0; 1982 size = 0;
1983 move = 0; 1983 move = 0;
1984 for (i = count-1; i >= 0; i--) { 1984 for (i = count-1; i >= 0; i--) {
1985 /* is more than half of this 1985 /* is more than half of this entry in 2nd half of the block? */
1986 if (size + map[i].size/2 > bl 1986 if (size + map[i].size/2 > blocksize/2)
1987 break; 1987 break;
1988 size += map[i].size; 1988 size += map[i].size;
1989 move++; 1989 move++;
1990 } 1990 }
1991 /* 1991 /*
1992 * map index at which we will split 1992 * map index at which we will split
1993 * 1993 *
1994 * If the sum of active entries didn' 1994 * If the sum of active entries didn't exceed half the block size, just
1995 * split it in half by count; each re 1995 * split it in half by count; each resulting block will have at least
1996 * half the space free. 1996 * half the space free.
1997 */ 1997 */
1998 if (i > 0) 1998 if (i > 0)
1999 split = count - move; 1999 split = count - move;
2000 else 2000 else
2001 split = count/2; 2001 split = count/2;
2002 2002
2003 hash2 = map[split].hash; 2003 hash2 = map[split].hash;
2004 continued = split > 0 ? hash2 == map[ 2004 continued = split > 0 ? hash2 == map[split - 1].hash : 0;
2005 dxtrace(printk(KERN_INFO "Split block 2005 dxtrace(printk(KERN_INFO "Split block %lu at %x, %i/%i\n",
2006 (unsigned long)dx_get 2006 (unsigned long)dx_get_block(frame->at),
2007 hash2 2007 hash2, split, count-split));
2008 2008
2009 /* Fancy dance to stay within two buf 2009 /* Fancy dance to stay within two buffers */
2010 de2 = dx_move_dirents(dir, data1, dat 2010 de2 = dx_move_dirents(dir, data1, data2, map + split, count - split,
2011 blocksize); 2011 blocksize);
2012 de = dx_pack_dirents(dir, data1, bloc 2012 de = dx_pack_dirents(dir, data1, blocksize);
2013 de->rec_len = ext4_rec_len_to_disk(da 2013 de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) -
2014 (c 2014 (char *) de,
2015 bl 2015 blocksize);
2016 de2->rec_len = ext4_rec_len_to_disk(d 2016 de2->rec_len = ext4_rec_len_to_disk(data2 + (blocksize - csum_size) -
2017 ( 2017 (char *) de2,
2018 b 2018 blocksize);
2019 if (csum_size) { 2019 if (csum_size) {
2020 ext4_initialize_dirent_tail(* 2020 ext4_initialize_dirent_tail(*bh, blocksize);
2021 ext4_initialize_dirent_tail(b 2021 ext4_initialize_dirent_tail(bh2, blocksize);
2022 } 2022 }
2023 2023
2024 dxtrace(dx_show_leaf(dir, hinfo, (str 2024 dxtrace(dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) data1,
2025 blocksize, 1)); 2025 blocksize, 1));
2026 dxtrace(dx_show_leaf(dir, hinfo, (str 2026 dxtrace(dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) data2,
2027 blocksize, 1)); 2027 blocksize, 1));
2028 2028
2029 /* Which block gets the new entry? */ 2029 /* Which block gets the new entry? */
2030 if (hinfo->hash >= hash2) { 2030 if (hinfo->hash >= hash2) {
2031 swap(*bh, bh2); 2031 swap(*bh, bh2);
2032 de = de2; 2032 de = de2;
2033 } 2033 }
2034 dx_insert_block(frame, hash2 + contin 2034 dx_insert_block(frame, hash2 + continued, newblock);
2035 err = ext4_handle_dirty_dirblock(hand 2035 err = ext4_handle_dirty_dirblock(handle, dir, bh2);
2036 if (err) 2036 if (err)
2037 goto journal_error; 2037 goto journal_error;
2038 err = ext4_handle_dirty_dx_node(handl 2038 err = ext4_handle_dirty_dx_node(handle, dir, frame->bh);
2039 if (err) 2039 if (err)
2040 goto journal_error; 2040 goto journal_error;
2041 brelse(bh2); 2041 brelse(bh2);
2042 dxtrace(dx_show_index("frame", frame- 2042 dxtrace(dx_show_index("frame", frame->entries));
2043 return de; 2043 return de;
2044 2044
2045 journal_error: 2045 journal_error:
2046 brelse(*bh); 2046 brelse(*bh);
2047 brelse(bh2); 2047 brelse(bh2);
2048 *bh = NULL; 2048 *bh = NULL;
2049 ext4_std_error(dir->i_sb, err); 2049 ext4_std_error(dir->i_sb, err);
2050 return ERR_PTR(err); 2050 return ERR_PTR(err);
2051 } 2051 }
2052 2052
2053 int ext4_find_dest_de(struct inode *dir, stru 2053 int ext4_find_dest_de(struct inode *dir, struct inode *inode,
2054 struct buffer_head *bh, 2054 struct buffer_head *bh,
2055 void *buf, int buf_size 2055 void *buf, int buf_size,
2056 struct ext4_filename *f 2056 struct ext4_filename *fname,
2057 struct ext4_dir_entry_2 2057 struct ext4_dir_entry_2 **dest_de)
2058 { 2058 {
2059 struct ext4_dir_entry_2 *de; 2059 struct ext4_dir_entry_2 *de;
2060 unsigned short reclen = ext4_dir_rec_ 2060 unsigned short reclen = ext4_dir_rec_len(fname_len(fname), dir);
2061 int nlen, rlen; 2061 int nlen, rlen;
2062 unsigned int offset = 0; 2062 unsigned int offset = 0;
2063 char *top; 2063 char *top;
2064 2064
2065 de = buf; 2065 de = buf;
2066 top = buf + buf_size - reclen; 2066 top = buf + buf_size - reclen;
2067 while ((char *) de <= top) { 2067 while ((char *) de <= top) {
2068 if (ext4_check_dir_entry(dir, 2068 if (ext4_check_dir_entry(dir, NULL, de, bh,
2069 buf, 2069 buf, buf_size, offset))
2070 return -EFSCORRUPTED; 2070 return -EFSCORRUPTED;
2071 if (ext4_match(dir, fname, de 2071 if (ext4_match(dir, fname, de))
2072 return -EEXIST; 2072 return -EEXIST;
2073 nlen = ext4_dir_rec_len(de->n 2073 nlen = ext4_dir_rec_len(de->name_len, dir);
2074 rlen = ext4_rec_len_from_disk 2074 rlen = ext4_rec_len_from_disk(de->rec_len, buf_size);
2075 if ((de->inode ? rlen - nlen 2075 if ((de->inode ? rlen - nlen : rlen) >= reclen)
2076 break; 2076 break;
2077 de = (struct ext4_dir_entry_2 2077 de = (struct ext4_dir_entry_2 *)((char *)de + rlen);
2078 offset += rlen; 2078 offset += rlen;
2079 } 2079 }
2080 if ((char *) de > top) 2080 if ((char *) de > top)
2081 return -ENOSPC; 2081 return -ENOSPC;
2082 2082
2083 *dest_de = de; 2083 *dest_de = de;
2084 return 0; 2084 return 0;
2085 } 2085 }
2086 2086
2087 void ext4_insert_dentry(struct inode *dir, 2087 void ext4_insert_dentry(struct inode *dir,
2088 struct inode *inode, 2088 struct inode *inode,
2089 struct ext4_dir_entry 2089 struct ext4_dir_entry_2 *de,
2090 int buf_size, 2090 int buf_size,
2091 struct ext4_filename 2091 struct ext4_filename *fname)
2092 { 2092 {
2093 2093
2094 int nlen, rlen; 2094 int nlen, rlen;
2095 2095
2096 nlen = ext4_dir_rec_len(de->name_len, 2096 nlen = ext4_dir_rec_len(de->name_len, dir);
2097 rlen = ext4_rec_len_from_disk(de->rec 2097 rlen = ext4_rec_len_from_disk(de->rec_len, buf_size);
2098 if (de->inode) { 2098 if (de->inode) {
2099 struct ext4_dir_entry_2 *de1 2099 struct ext4_dir_entry_2 *de1 =
2100 (struct ext4_dir_entr 2100 (struct ext4_dir_entry_2 *)((char *)de + nlen);
2101 de1->rec_len = ext4_rec_len_t 2101 de1->rec_len = ext4_rec_len_to_disk(rlen - nlen, buf_size);
2102 de->rec_len = ext4_rec_len_to 2102 de->rec_len = ext4_rec_len_to_disk(nlen, buf_size);
2103 de = de1; 2103 de = de1;
2104 } 2104 }
2105 de->file_type = EXT4_FT_UNKNOWN; 2105 de->file_type = EXT4_FT_UNKNOWN;
2106 de->inode = cpu_to_le32(inode->i_ino) 2106 de->inode = cpu_to_le32(inode->i_ino);
2107 ext4_set_de_type(inode->i_sb, de, ino 2107 ext4_set_de_type(inode->i_sb, de, inode->i_mode);
2108 de->name_len = fname_len(fname); 2108 de->name_len = fname_len(fname);
2109 memcpy(de->name, fname_name(fname), f 2109 memcpy(de->name, fname_name(fname), fname_len(fname));
2110 if (ext4_hash_in_dirent(dir)) { 2110 if (ext4_hash_in_dirent(dir)) {
2111 struct dx_hash_info *hinfo = 2111 struct dx_hash_info *hinfo = &fname->hinfo;
2112 2112
2113 EXT4_DIRENT_HASHES(de)->hash 2113 EXT4_DIRENT_HASHES(de)->hash = cpu_to_le32(hinfo->hash);
2114 EXT4_DIRENT_HASHES(de)->minor 2114 EXT4_DIRENT_HASHES(de)->minor_hash =
2115 2115 cpu_to_le32(hinfo->minor_hash);
2116 } 2116 }
2117 } 2117 }
2118 2118
2119 /* 2119 /*
2120 * Add a new entry into a directory (leaf) bl 2120 * Add a new entry into a directory (leaf) block. If de is non-NULL,
2121 * it points to a directory entry which is gu 2121 * it points to a directory entry which is guaranteed to be large
2122 * enough for new directory entry. If de is 2122 * enough for new directory entry. If de is NULL, then
2123 * add_dirent_to_buf will attempt search the 2123 * add_dirent_to_buf will attempt search the directory block for
2124 * space. It will return -ENOSPC if no space 2124 * space. It will return -ENOSPC if no space is available, and -EIO
2125 * and -EEXIST if directory entry already exi 2125 * and -EEXIST if directory entry already exists.
2126 */ 2126 */
2127 static int add_dirent_to_buf(handle_t *handle 2127 static int add_dirent_to_buf(handle_t *handle, struct ext4_filename *fname,
2128 struct inode *di 2128 struct inode *dir,
2129 struct inode *in 2129 struct inode *inode, struct ext4_dir_entry_2 *de,
2130 struct buffer_he 2130 struct buffer_head *bh)
2131 { 2131 {
2132 unsigned int blocksize = dir->i_sb 2132 unsigned int blocksize = dir->i_sb->s_blocksize;
2133 int csum_size = 0; 2133 int csum_size = 0;
2134 int err, err2; 2134 int err, err2;
2135 2135
2136 if (ext4_has_metadata_csum(inode->i_s 2136 if (ext4_has_metadata_csum(inode->i_sb))
2137 csum_size = sizeof(struct ext 2137 csum_size = sizeof(struct ext4_dir_entry_tail);
2138 2138
2139 if (!de) { 2139 if (!de) {
2140 err = ext4_find_dest_de(dir, 2140 err = ext4_find_dest_de(dir, inode, bh, bh->b_data,
2141 block 2141 blocksize - csum_size, fname, &de);
2142 if (err) 2142 if (err)
2143 return err; 2143 return err;
2144 } 2144 }
2145 BUFFER_TRACE(bh, "get_write_access"); 2145 BUFFER_TRACE(bh, "get_write_access");
2146 err = ext4_journal_get_write_access(h 2146 err = ext4_journal_get_write_access(handle, dir->i_sb, bh,
2147 E 2147 EXT4_JTR_NONE);
2148 if (err) { 2148 if (err) {
2149 ext4_std_error(dir->i_sb, err 2149 ext4_std_error(dir->i_sb, err);
2150 return err; 2150 return err;
2151 } 2151 }
2152 2152
2153 /* By now the buffer is marked for jo 2153 /* By now the buffer is marked for journaling */
2154 ext4_insert_dentry(dir, inode, de, bl 2154 ext4_insert_dentry(dir, inode, de, blocksize, fname);
2155 2155
2156 /* 2156 /*
2157 * XXX shouldn't update any times unt 2157 * XXX shouldn't update any times until successful
2158 * completion of syscall, but too man 2158 * completion of syscall, but too many callers depend
2159 * on this. 2159 * on this.
2160 * 2160 *
2161 * XXX similarly, too many callers de 2161 * XXX similarly, too many callers depend on
2162 * ext4_new_inode() setting the times 2162 * ext4_new_inode() setting the times, but error
2163 * recovery deletes the inode, so the 2163 * recovery deletes the inode, so the worst that can
2164 * happen is that the times are sligh 2164 * happen is that the times are slightly out of date
2165 * and/or different from the director 2165 * and/or different from the directory change time.
2166 */ 2166 */
2167 inode_set_mtime_to_ts(dir, inode_set_ 2167 inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
2168 ext4_update_dx_flag(dir); 2168 ext4_update_dx_flag(dir);
2169 inode_inc_iversion(dir); 2169 inode_inc_iversion(dir);
2170 err2 = ext4_mark_inode_dirty(handle, 2170 err2 = ext4_mark_inode_dirty(handle, dir);
2171 BUFFER_TRACE(bh, "call ext4_handle_di 2171 BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
2172 err = ext4_handle_dirty_dirblock(hand 2172 err = ext4_handle_dirty_dirblock(handle, dir, bh);
2173 if (err) 2173 if (err)
2174 ext4_std_error(dir->i_sb, err 2174 ext4_std_error(dir->i_sb, err);
2175 return err ? err : err2; 2175 return err ? err : err2;
2176 } 2176 }
2177 2177
2178 static bool ext4_check_dx_root(struct inode * 2178 static bool ext4_check_dx_root(struct inode *dir, struct dx_root *root)
2179 { 2179 {
2180 struct fake_dirent *fde; 2180 struct fake_dirent *fde;
2181 const char *error_msg; 2181 const char *error_msg;
2182 unsigned int rlen; 2182 unsigned int rlen;
2183 unsigned int blocksize = dir->i_sb->s 2183 unsigned int blocksize = dir->i_sb->s_blocksize;
2184 char *blockend = (char *)root + dir-> 2184 char *blockend = (char *)root + dir->i_sb->s_blocksize;
2185 2185
2186 fde = &root->dot; 2186 fde = &root->dot;
2187 if (unlikely(fde->name_len != 1)) { 2187 if (unlikely(fde->name_len != 1)) {
2188 error_msg = "invalid name_len 2188 error_msg = "invalid name_len for '.'";
2189 goto corrupted; 2189 goto corrupted;
2190 } 2190 }
2191 if (unlikely(strncmp(root->dot_name, 2191 if (unlikely(strncmp(root->dot_name, ".", fde->name_len))) {
2192 error_msg = "invalid name for 2192 error_msg = "invalid name for '.'";
2193 goto corrupted; 2193 goto corrupted;
2194 } 2194 }
2195 rlen = ext4_rec_len_from_disk(fde->re 2195 rlen = ext4_rec_len_from_disk(fde->rec_len, blocksize);
2196 if (unlikely((char *)fde + rlen >= bl 2196 if (unlikely((char *)fde + rlen >= blockend)) {
2197 error_msg = "invalid rec_len 2197 error_msg = "invalid rec_len for '.'";
2198 goto corrupted; 2198 goto corrupted;
2199 } 2199 }
2200 2200
2201 fde = &root->dotdot; 2201 fde = &root->dotdot;
2202 if (unlikely(fde->name_len != 2)) { 2202 if (unlikely(fde->name_len != 2)) {
2203 error_msg = "invalid name_len 2203 error_msg = "invalid name_len for '..'";
2204 goto corrupted; 2204 goto corrupted;
2205 } 2205 }
2206 if (unlikely(strncmp(root->dotdot_nam 2206 if (unlikely(strncmp(root->dotdot_name, "..", fde->name_len))) {
2207 error_msg = "invalid name for 2207 error_msg = "invalid name for '..'";
2208 goto corrupted; 2208 goto corrupted;
2209 } 2209 }
2210 rlen = ext4_rec_len_from_disk(fde->re 2210 rlen = ext4_rec_len_from_disk(fde->rec_len, blocksize);
2211 if (unlikely((char *)fde + rlen >= bl 2211 if (unlikely((char *)fde + rlen >= blockend)) {
2212 error_msg = "invalid rec_len 2212 error_msg = "invalid rec_len for '..'";
2213 goto corrupted; 2213 goto corrupted;
2214 } 2214 }
2215 2215
2216 return true; 2216 return true;
2217 2217
2218 corrupted: 2218 corrupted:
2219 EXT4_ERROR_INODE(dir, "Corrupt dir, % 2219 EXT4_ERROR_INODE(dir, "Corrupt dir, %s, running e2fsck is recommended",
2220 error_msg); 2220 error_msg);
2221 return false; 2221 return false;
2222 } 2222 }
2223 2223
2224 /* 2224 /*
2225 * This converts a one block unindexed direct 2225 * This converts a one block unindexed directory to a 3 block indexed
2226 * directory, and adds the dentry to the inde 2226 * directory, and adds the dentry to the indexed directory.
2227 */ 2227 */
2228 static int make_indexed_dir(handle_t *handle, 2228 static int make_indexed_dir(handle_t *handle, struct ext4_filename *fname,
2229 struct inode *dir 2229 struct inode *dir,
2230 struct inode *ino 2230 struct inode *inode, struct buffer_head *bh)
2231 { 2231 {
2232 struct buffer_head *bh2; 2232 struct buffer_head *bh2;
2233 struct dx_root *root; 2233 struct dx_root *root;
2234 struct dx_frame frames[EXT4_HTREE_LEV 2234 struct dx_frame frames[EXT4_HTREE_LEVEL], *frame;
2235 struct dx_entry *entries; 2235 struct dx_entry *entries;
2236 struct ext4_dir_entry_2 *de, *de2; 2236 struct ext4_dir_entry_2 *de, *de2;
2237 char *data2, *top; 2237 char *data2, *top;
2238 unsigned len; 2238 unsigned len;
2239 int retval; 2239 int retval;
2240 unsigned blocksize; 2240 unsigned blocksize;
2241 ext4_lblk_t block; 2241 ext4_lblk_t block;
2242 struct fake_dirent *fde; 2242 struct fake_dirent *fde;
2243 int csum_size = 0; 2243 int csum_size = 0;
2244 2244
2245 if (ext4_has_metadata_csum(inode->i_s 2245 if (ext4_has_metadata_csum(inode->i_sb))
2246 csum_size = sizeof(struct ext 2246 csum_size = sizeof(struct ext4_dir_entry_tail);
2247 2247
2248 blocksize = dir->i_sb->s_blocksize; 2248 blocksize = dir->i_sb->s_blocksize;
2249 dxtrace(printk(KERN_DEBUG "Creating i 2249 dxtrace(printk(KERN_DEBUG "Creating index: inode %lu\n", dir->i_ino));
2250 BUFFER_TRACE(bh, "get_write_access"); 2250 BUFFER_TRACE(bh, "get_write_access");
2251 retval = ext4_journal_get_write_acces 2251 retval = ext4_journal_get_write_access(handle, dir->i_sb, bh,
2252 2252 EXT4_JTR_NONE);
2253 if (retval) { 2253 if (retval) {
2254 ext4_std_error(dir->i_sb, ret 2254 ext4_std_error(dir->i_sb, retval);
2255 brelse(bh); 2255 brelse(bh);
2256 return retval; 2256 return retval;
2257 } 2257 }
2258 2258
2259 root = (struct dx_root *) bh->b_data; 2259 root = (struct dx_root *) bh->b_data;
2260 if (!ext4_check_dx_root(dir, root)) { 2260 if (!ext4_check_dx_root(dir, root)) {
2261 brelse(bh); 2261 brelse(bh);
2262 return -EFSCORRUPTED; 2262 return -EFSCORRUPTED;
2263 } 2263 }
2264 2264
2265 /* The 0th block becomes the root, mo 2265 /* The 0th block becomes the root, move the dirents out */
2266 fde = &root->dotdot; 2266 fde = &root->dotdot;
2267 de = (struct ext4_dir_entry_2 *)((cha 2267 de = (struct ext4_dir_entry_2 *)((char *)fde +
2268 ext4_rec_len_from_disk(fde->r 2268 ext4_rec_len_from_disk(fde->rec_len, blocksize));
2269 len = ((char *) root) + (blocksize - 2269 len = ((char *) root) + (blocksize - csum_size) - (char *) de;
2270 2270
2271 /* Allocate new block for the 0th blo 2271 /* Allocate new block for the 0th block's dirents */
2272 bh2 = ext4_append(handle, dir, &block 2272 bh2 = ext4_append(handle, dir, &block);
2273 if (IS_ERR(bh2)) { 2273 if (IS_ERR(bh2)) {
2274 brelse(bh); 2274 brelse(bh);
2275 return PTR_ERR(bh2); 2275 return PTR_ERR(bh2);
2276 } 2276 }
2277 ext4_set_inode_flag(dir, EXT4_INODE_I 2277 ext4_set_inode_flag(dir, EXT4_INODE_INDEX);
2278 data2 = bh2->b_data; 2278 data2 = bh2->b_data;
2279 2279
2280 memcpy(data2, de, len); 2280 memcpy(data2, de, len);
2281 memset(de, 0, len); /* wipe old data 2281 memset(de, 0, len); /* wipe old data */
2282 de = (struct ext4_dir_entry_2 *) data 2282 de = (struct ext4_dir_entry_2 *) data2;
2283 top = data2 + len; 2283 top = data2 + len;
2284 while ((char *)(de2 = ext4_next_entry 2284 while ((char *)(de2 = ext4_next_entry(de, blocksize)) < top) {
2285 if (ext4_check_dir_entry(dir, 2285 if (ext4_check_dir_entry(dir, NULL, de, bh2, data2, len,
2286 (char 2286 (char *)de - data2)) {
2287 brelse(bh2); 2287 brelse(bh2);
2288 brelse(bh); 2288 brelse(bh);
2289 return -EFSCORRUPTED; 2289 return -EFSCORRUPTED;
2290 } 2290 }
2291 de = de2; 2291 de = de2;
2292 } 2292 }
2293 de->rec_len = ext4_rec_len_to_disk(da 2293 de->rec_len = ext4_rec_len_to_disk(data2 + (blocksize - csum_size) -
2294 (c 2294 (char *) de, blocksize);
2295 2295
2296 if (csum_size) 2296 if (csum_size)
2297 ext4_initialize_dirent_tail(b 2297 ext4_initialize_dirent_tail(bh2, blocksize);
2298 2298
2299 /* Initialize the root; the dot diren 2299 /* Initialize the root; the dot dirents already exist */
2300 de = (struct ext4_dir_entry_2 *) (&ro 2300 de = (struct ext4_dir_entry_2 *) (&root->dotdot);
2301 de->rec_len = ext4_rec_len_to_disk( 2301 de->rec_len = ext4_rec_len_to_disk(
2302 blocksize - ext4_dir_ 2302 blocksize - ext4_dir_rec_len(2, NULL), blocksize);
2303 memset (&root->info, 0, sizeof(root-> 2303 memset (&root->info, 0, sizeof(root->info));
2304 root->info.info_length = sizeof(root- 2304 root->info.info_length = sizeof(root->info);
2305 if (ext4_hash_in_dirent(dir)) 2305 if (ext4_hash_in_dirent(dir))
2306 root->info.hash_version = DX_ 2306 root->info.hash_version = DX_HASH_SIPHASH;
2307 else 2307 else
2308 root->info.hash_version = 2308 root->info.hash_version =
2309 EXT4_SB(dir-> 2309 EXT4_SB(dir->i_sb)->s_def_hash_version;
2310 2310
2311 entries = root->entries; 2311 entries = root->entries;
2312 dx_set_block(entries, 1); 2312 dx_set_block(entries, 1);
2313 dx_set_count(entries, 1); 2313 dx_set_count(entries, 1);
2314 dx_set_limit(entries, dx_root_limit(d 2314 dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info)));
2315 2315
2316 /* Initialize as for dx_probe */ 2316 /* Initialize as for dx_probe */
2317 fname->hinfo.hash_version = root->inf 2317 fname->hinfo.hash_version = root->info.hash_version;
2318 if (fname->hinfo.hash_version <= DX_H 2318 if (fname->hinfo.hash_version <= DX_HASH_TEA)
2319 fname->hinfo.hash_version += 2319 fname->hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned;
2320 fname->hinfo.seed = EXT4_SB(dir->i_sb 2320 fname->hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed;
2321 2321
2322 /* casefolded encrypted hashes are co 2322 /* casefolded encrypted hashes are computed on fname setup */
2323 if (!ext4_hash_in_dirent(dir)) { 2323 if (!ext4_hash_in_dirent(dir)) {
2324 int err = ext4fs_dirhash(dir, 2324 int err = ext4fs_dirhash(dir, fname_name(fname),
2325 fnam 2325 fname_len(fname), &fname->hinfo);
2326 if (err < 0) { 2326 if (err < 0) {
2327 brelse(bh2); 2327 brelse(bh2);
2328 brelse(bh); 2328 brelse(bh);
2329 return err; 2329 return err;
2330 } 2330 }
2331 } 2331 }
2332 memset(frames, 0, sizeof(frames)); 2332 memset(frames, 0, sizeof(frames));
2333 frame = frames; 2333 frame = frames;
2334 frame->entries = entries; 2334 frame->entries = entries;
2335 frame->at = entries; 2335 frame->at = entries;
2336 frame->bh = bh; 2336 frame->bh = bh;
2337 2337
2338 retval = ext4_handle_dirty_dx_node(ha 2338 retval = ext4_handle_dirty_dx_node(handle, dir, frame->bh);
2339 if (retval) 2339 if (retval)
2340 goto out_frames; 2340 goto out_frames;
2341 retval = ext4_handle_dirty_dirblock(h 2341 retval = ext4_handle_dirty_dirblock(handle, dir, bh2);
2342 if (retval) 2342 if (retval)
2343 goto out_frames; 2343 goto out_frames;
2344 2344
2345 de = do_split(handle,dir, &bh2, frame 2345 de = do_split(handle,dir, &bh2, frame, &fname->hinfo);
2346 if (IS_ERR(de)) { 2346 if (IS_ERR(de)) {
2347 retval = PTR_ERR(de); 2347 retval = PTR_ERR(de);
2348 goto out_frames; 2348 goto out_frames;
2349 } 2349 }
2350 2350
2351 retval = add_dirent_to_buf(handle, fn 2351 retval = add_dirent_to_buf(handle, fname, dir, inode, de, bh2);
2352 out_frames: 2352 out_frames:
2353 /* 2353 /*
2354 * Even if the block split failed, we 2354 * Even if the block split failed, we have to properly write
2355 * out all the changes we did so far. 2355 * out all the changes we did so far. Otherwise we can end up
2356 * with corrupted filesystem. 2356 * with corrupted filesystem.
2357 */ 2357 */
2358 if (retval) 2358 if (retval)
2359 ext4_mark_inode_dirty(handle, 2359 ext4_mark_inode_dirty(handle, dir);
2360 dx_release(frames); 2360 dx_release(frames);
2361 brelse(bh2); 2361 brelse(bh2);
2362 return retval; 2362 return retval;
2363 } 2363 }
2364 2364
2365 /* 2365 /*
2366 * ext4_add_entry() 2366 * ext4_add_entry()
2367 * 2367 *
2368 * adds a file entry to the specified directo 2368 * adds a file entry to the specified directory, using the same
2369 * semantics as ext4_find_entry(). It returns 2369 * semantics as ext4_find_entry(). It returns NULL if it failed.
2370 * 2370 *
2371 * NOTE!! The inode part of 'de' is left at 0 2371 * NOTE!! The inode part of 'de' is left at 0 - which means you
2372 * may not sleep between calling this and put 2372 * may not sleep between calling this and putting something into
2373 * the entry, as someone else might have used 2373 * the entry, as someone else might have used it while you slept.
2374 */ 2374 */
2375 static int ext4_add_entry(handle_t *handle, s 2375 static int ext4_add_entry(handle_t *handle, struct dentry *dentry,
2376 struct inode *inode 2376 struct inode *inode)
2377 { 2377 {
2378 struct inode *dir = d_inode(dentry->d 2378 struct inode *dir = d_inode(dentry->d_parent);
2379 struct buffer_head *bh = NULL; 2379 struct buffer_head *bh = NULL;
2380 struct ext4_dir_entry_2 *de; 2380 struct ext4_dir_entry_2 *de;
2381 struct super_block *sb; 2381 struct super_block *sb;
2382 struct ext4_filename fname; 2382 struct ext4_filename fname;
2383 int retval; 2383 int retval;
2384 int dx_fallback=0; 2384 int dx_fallback=0;
2385 unsigned blocksize; 2385 unsigned blocksize;
2386 ext4_lblk_t block, blocks; 2386 ext4_lblk_t block, blocks;
2387 int csum_size = 0; 2387 int csum_size = 0;
2388 2388
2389 if (ext4_has_metadata_csum(inode->i_s 2389 if (ext4_has_metadata_csum(inode->i_sb))
2390 csum_size = sizeof(struct ext 2390 csum_size = sizeof(struct ext4_dir_entry_tail);
2391 2391
2392 sb = dir->i_sb; 2392 sb = dir->i_sb;
2393 blocksize = sb->s_blocksize; 2393 blocksize = sb->s_blocksize;
2394 2394
2395 if (fscrypt_is_nokey_name(dentry)) 2395 if (fscrypt_is_nokey_name(dentry))
2396 return -ENOKEY; 2396 return -ENOKEY;
2397 2397
2398 #if IS_ENABLED(CONFIG_UNICODE) 2398 #if IS_ENABLED(CONFIG_UNICODE)
2399 if (sb_has_strict_encoding(sb) && IS_ 2399 if (sb_has_strict_encoding(sb) && IS_CASEFOLDED(dir) &&
2400 utf8_validate(sb->s_encoding, &de 2400 utf8_validate(sb->s_encoding, &dentry->d_name))
2401 return -EINVAL; 2401 return -EINVAL;
2402 #endif 2402 #endif
2403 2403
2404 retval = ext4_fname_setup_filename(di 2404 retval = ext4_fname_setup_filename(dir, &dentry->d_name, 0, &fname);
2405 if (retval) 2405 if (retval)
2406 return retval; 2406 return retval;
2407 2407
2408 if (ext4_has_inline_data(dir)) { 2408 if (ext4_has_inline_data(dir)) {
2409 retval = ext4_try_add_inline_ 2409 retval = ext4_try_add_inline_entry(handle, &fname, dir, inode);
2410 if (retval < 0) 2410 if (retval < 0)
2411 goto out; 2411 goto out;
2412 if (retval == 1) { 2412 if (retval == 1) {
2413 retval = 0; 2413 retval = 0;
2414 goto out; 2414 goto out;
2415 } 2415 }
2416 } 2416 }
2417 2417
2418 if (is_dx(dir)) { 2418 if (is_dx(dir)) {
2419 retval = ext4_dx_add_entry(ha 2419 retval = ext4_dx_add_entry(handle, &fname, dir, inode);
2420 if (!retval || (retval != ERR 2420 if (!retval || (retval != ERR_BAD_DX_DIR))
2421 goto out; 2421 goto out;
2422 /* Can we just ignore htree d 2422 /* Can we just ignore htree data? */
2423 if (ext4_has_metadata_csum(sb 2423 if (ext4_has_metadata_csum(sb)) {
2424 EXT4_ERROR_INODE(dir, 2424 EXT4_ERROR_INODE(dir,
2425 "Directory ha 2425 "Directory has corrupted htree index.");
2426 retval = -EFSCORRUPTE 2426 retval = -EFSCORRUPTED;
2427 goto out; 2427 goto out;
2428 } 2428 }
2429 ext4_clear_inode_flag(dir, EX 2429 ext4_clear_inode_flag(dir, EXT4_INODE_INDEX);
2430 dx_fallback++; 2430 dx_fallback++;
2431 retval = ext4_mark_inode_dirt 2431 retval = ext4_mark_inode_dirty(handle, dir);
2432 if (unlikely(retval)) 2432 if (unlikely(retval))
2433 goto out; 2433 goto out;
2434 } 2434 }
2435 blocks = dir->i_size >> sb->s_blocksi 2435 blocks = dir->i_size >> sb->s_blocksize_bits;
2436 for (block = 0; block < blocks; block 2436 for (block = 0; block < blocks; block++) {
2437 bh = ext4_read_dirblock(dir, 2437 bh = ext4_read_dirblock(dir, block, DIRENT);
2438 if (bh == NULL) { 2438 if (bh == NULL) {
2439 bh = ext4_bread(handl 2439 bh = ext4_bread(handle, dir, block,
2440 EXT4_ 2440 EXT4_GET_BLOCKS_CREATE);
2441 goto add_to_new_block 2441 goto add_to_new_block;
2442 } 2442 }
2443 if (IS_ERR(bh)) { 2443 if (IS_ERR(bh)) {
2444 retval = PTR_ERR(bh); 2444 retval = PTR_ERR(bh);
2445 bh = NULL; 2445 bh = NULL;
2446 goto out; 2446 goto out;
2447 } 2447 }
2448 retval = add_dirent_to_buf(ha 2448 retval = add_dirent_to_buf(handle, &fname, dir, inode,
2449 NU 2449 NULL, bh);
2450 if (retval != -ENOSPC) 2450 if (retval != -ENOSPC)
2451 goto out; 2451 goto out;
2452 2452
2453 if (blocks == 1 && !dx_fallba 2453 if (blocks == 1 && !dx_fallback &&
2454 ext4_has_feature_dir_inde 2454 ext4_has_feature_dir_index(sb)) {
2455 retval = make_indexed 2455 retval = make_indexed_dir(handle, &fname, dir,
2456 2456 inode, bh);
2457 bh = NULL; /* make_in 2457 bh = NULL; /* make_indexed_dir releases bh */
2458 goto out; 2458 goto out;
2459 } 2459 }
2460 brelse(bh); 2460 brelse(bh);
2461 } 2461 }
2462 bh = ext4_append(handle, dir, &block) 2462 bh = ext4_append(handle, dir, &block);
2463 add_to_new_block: 2463 add_to_new_block:
2464 if (IS_ERR(bh)) { 2464 if (IS_ERR(bh)) {
2465 retval = PTR_ERR(bh); 2465 retval = PTR_ERR(bh);
2466 bh = NULL; 2466 bh = NULL;
2467 goto out; 2467 goto out;
2468 } 2468 }
2469 de = (struct ext4_dir_entry_2 *) bh-> 2469 de = (struct ext4_dir_entry_2 *) bh->b_data;
2470 de->inode = 0; 2470 de->inode = 0;
2471 de->rec_len = ext4_rec_len_to_disk(bl 2471 de->rec_len = ext4_rec_len_to_disk(blocksize - csum_size, blocksize);
2472 2472
2473 if (csum_size) 2473 if (csum_size)
2474 ext4_initialize_dirent_tail(b 2474 ext4_initialize_dirent_tail(bh, blocksize);
2475 2475
2476 retval = add_dirent_to_buf(handle, &f 2476 retval = add_dirent_to_buf(handle, &fname, dir, inode, de, bh);
2477 out: 2477 out:
2478 ext4_fname_free_filename(&fname); 2478 ext4_fname_free_filename(&fname);
2479 brelse(bh); 2479 brelse(bh);
2480 if (retval == 0) 2480 if (retval == 0)
2481 ext4_set_inode_state(inode, E 2481 ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY);
2482 return retval; 2482 return retval;
2483 } 2483 }
2484 2484
2485 /* 2485 /*
2486 * Returns 0 for success, or a negative error 2486 * Returns 0 for success, or a negative error value
2487 */ 2487 */
2488 static int ext4_dx_add_entry(handle_t *handle 2488 static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname,
2489 struct inode *di 2489 struct inode *dir, struct inode *inode)
2490 { 2490 {
2491 struct dx_frame frames[EXT4_HTREE_LEV 2491 struct dx_frame frames[EXT4_HTREE_LEVEL], *frame;
2492 struct dx_entry *entries, *at; 2492 struct dx_entry *entries, *at;
2493 struct buffer_head *bh; 2493 struct buffer_head *bh;
2494 struct super_block *sb = dir->i_sb; 2494 struct super_block *sb = dir->i_sb;
2495 struct ext4_dir_entry_2 *de; 2495 struct ext4_dir_entry_2 *de;
2496 int restart; 2496 int restart;
2497 int err; 2497 int err;
2498 2498
2499 again: 2499 again:
2500 restart = 0; 2500 restart = 0;
2501 frame = dx_probe(fname, dir, NULL, fr 2501 frame = dx_probe(fname, dir, NULL, frames);
2502 if (IS_ERR(frame)) 2502 if (IS_ERR(frame))
2503 return PTR_ERR(frame); 2503 return PTR_ERR(frame);
2504 entries = frame->entries; 2504 entries = frame->entries;
2505 at = frame->at; 2505 at = frame->at;
2506 bh = ext4_read_dirblock(dir, dx_get_b 2506 bh = ext4_read_dirblock(dir, dx_get_block(frame->at), DIRENT_HTREE);
2507 if (IS_ERR(bh)) { 2507 if (IS_ERR(bh)) {
2508 err = PTR_ERR(bh); 2508 err = PTR_ERR(bh);
2509 bh = NULL; 2509 bh = NULL;
2510 goto cleanup; 2510 goto cleanup;
2511 } 2511 }
2512 2512
2513 BUFFER_TRACE(bh, "get_write_access"); 2513 BUFFER_TRACE(bh, "get_write_access");
2514 err = ext4_journal_get_write_access(h 2514 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
2515 if (err) 2515 if (err)
2516 goto journal_error; 2516 goto journal_error;
2517 2517
2518 err = add_dirent_to_buf(handle, fname 2518 err = add_dirent_to_buf(handle, fname, dir, inode, NULL, bh);
2519 if (err != -ENOSPC) 2519 if (err != -ENOSPC)
2520 goto cleanup; 2520 goto cleanup;
2521 2521
2522 err = 0; 2522 err = 0;
2523 /* Block full, should compress but fo 2523 /* Block full, should compress but for now just split */
2524 dxtrace(printk(KERN_DEBUG "using %u o 2524 dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n",
2525 dx_get_count(entries), 2525 dx_get_count(entries), dx_get_limit(entries)));
2526 /* Need to split index? */ 2526 /* Need to split index? */
2527 if (dx_get_count(entries) == dx_get_l 2527 if (dx_get_count(entries) == dx_get_limit(entries)) {
2528 ext4_lblk_t newblock; 2528 ext4_lblk_t newblock;
2529 int levels = frame - frames + 2529 int levels = frame - frames + 1;
2530 unsigned int icount; 2530 unsigned int icount;
2531 int add_level = 1; 2531 int add_level = 1;
2532 struct dx_entry *entries2; 2532 struct dx_entry *entries2;
2533 struct dx_node *node2; 2533 struct dx_node *node2;
2534 struct buffer_head *bh2; 2534 struct buffer_head *bh2;
2535 2535
2536 while (frame > frames) { 2536 while (frame > frames) {
2537 if (dx_get_count((fra 2537 if (dx_get_count((frame - 1)->entries) <
2538 dx_get_limit((fra 2538 dx_get_limit((frame - 1)->entries)) {
2539 add_level = 0 2539 add_level = 0;
2540 break; 2540 break;
2541 } 2541 }
2542 frame--; /* split hig 2542 frame--; /* split higher index block */
2543 at = frame->at; 2543 at = frame->at;
2544 entries = frame->entr 2544 entries = frame->entries;
2545 restart = 1; 2545 restart = 1;
2546 } 2546 }
2547 if (add_level && levels == ex 2547 if (add_level && levels == ext4_dir_htree_level(sb)) {
2548 ext4_warning(sb, "Dir 2548 ext4_warning(sb, "Directory (ino: %lu) index full, "
2549 "rea 2549 "reach max htree level :%d",
2550 dir- 2550 dir->i_ino, levels);
2551 if (ext4_dir_htree_le 2551 if (ext4_dir_htree_level(sb) < EXT4_HTREE_LEVEL) {
2552 ext4_warning( 2552 ext4_warning(sb, "Large directory feature is "
2553 2553 "not enabled on this "
2554 2554 "filesystem");
2555 } 2555 }
2556 err = -ENOSPC; 2556 err = -ENOSPC;
2557 goto cleanup; 2557 goto cleanup;
2558 } 2558 }
2559 icount = dx_get_count(entries 2559 icount = dx_get_count(entries);
2560 bh2 = ext4_append(handle, dir 2560 bh2 = ext4_append(handle, dir, &newblock);
2561 if (IS_ERR(bh2)) { 2561 if (IS_ERR(bh2)) {
2562 err = PTR_ERR(bh2); 2562 err = PTR_ERR(bh2);
2563 goto cleanup; 2563 goto cleanup;
2564 } 2564 }
2565 node2 = (struct dx_node *)(bh 2565 node2 = (struct dx_node *)(bh2->b_data);
2566 entries2 = node2->entries; 2566 entries2 = node2->entries;
2567 memset(&node2->fake, 0, sizeo 2567 memset(&node2->fake, 0, sizeof(struct fake_dirent));
2568 node2->fake.rec_len = ext4_re 2568 node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize,
2569 2569 sb->s_blocksize);
2570 BUFFER_TRACE(frame->bh, "get_ 2570 BUFFER_TRACE(frame->bh, "get_write_access");
2571 err = ext4_journal_get_write_ 2571 err = ext4_journal_get_write_access(handle, sb, frame->bh,
2572 2572 EXT4_JTR_NONE);
2573 if (err) 2573 if (err)
2574 goto journal_error; 2574 goto journal_error;
2575 if (!add_level) { 2575 if (!add_level) {
2576 unsigned icount1 = ic 2576 unsigned icount1 = icount/2, icount2 = icount - icount1;
2577 unsigned hash2 = dx_g 2577 unsigned hash2 = dx_get_hash(entries + icount1);
2578 dxtrace(printk(KERN_D 2578 dxtrace(printk(KERN_DEBUG "Split index %i/%i\n",
2579 icount 2579 icount1, icount2));
2580 2580
2581 BUFFER_TRACE(frame->b 2581 BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */
2582 err = ext4_journal_ge 2582 err = ext4_journal_get_write_access(handle, sb,
2583 2583 (frame - 1)->bh,
2584 2584 EXT4_JTR_NONE);
2585 if (err) 2585 if (err)
2586 goto journal_ 2586 goto journal_error;
2587 2587
2588 memcpy((char *) entri 2588 memcpy((char *) entries2, (char *) (entries + icount1),
2589 icount2 * size 2589 icount2 * sizeof(struct dx_entry));
2590 dx_set_count(entries, 2590 dx_set_count(entries, icount1);
2591 dx_set_count(entries2 2591 dx_set_count(entries2, icount2);
2592 dx_set_limit(entries2 2592 dx_set_limit(entries2, dx_node_limit(dir));
2593 2593
2594 /* Which index block 2594 /* Which index block gets the new entry? */
2595 if (at - entries >= i 2595 if (at - entries >= icount1) {
2596 frame->at = a 2596 frame->at = at - entries - icount1 + entries2;
2597 frame->entrie 2597 frame->entries = entries = entries2;
2598 swap(frame->b 2598 swap(frame->bh, bh2);
2599 } 2599 }
2600 dx_insert_block((fram 2600 dx_insert_block((frame - 1), hash2, newblock);
2601 dxtrace(dx_show_index 2601 dxtrace(dx_show_index("node", frame->entries));
2602 dxtrace(dx_show_index 2602 dxtrace(dx_show_index("node",
2603 ((struct dx_no 2603 ((struct dx_node *) bh2->b_data)->entries));
2604 err = ext4_handle_dir 2604 err = ext4_handle_dirty_dx_node(handle, dir, bh2);
2605 if (err) 2605 if (err)
2606 goto journal_ 2606 goto journal_error;
2607 brelse (bh2); 2607 brelse (bh2);
2608 err = ext4_handle_dir 2608 err = ext4_handle_dirty_dx_node(handle, dir,
2609 2609 (frame - 1)->bh);
2610 if (err) 2610 if (err)
2611 goto journal_ 2611 goto journal_error;
2612 err = ext4_handle_dir 2612 err = ext4_handle_dirty_dx_node(handle, dir,
2613 2613 frame->bh);
2614 if (restart || err) 2614 if (restart || err)
2615 goto journal_ 2615 goto journal_error;
2616 } else { 2616 } else {
2617 struct dx_root *dxroo 2617 struct dx_root *dxroot;
2618 memcpy((char *) entri 2618 memcpy((char *) entries2, (char *) entries,
2619 icount * sizeo 2619 icount * sizeof(struct dx_entry));
2620 dx_set_limit(entries2 2620 dx_set_limit(entries2, dx_node_limit(dir));
2621 2621
2622 /* Set up root */ 2622 /* Set up root */
2623 dx_set_count(entries, 2623 dx_set_count(entries, 1);
2624 dx_set_block(entries 2624 dx_set_block(entries + 0, newblock);
2625 dxroot = (struct dx_r 2625 dxroot = (struct dx_root *)frames[0].bh->b_data;
2626 dxroot->info.indirect 2626 dxroot->info.indirect_levels += 1;
2627 dxtrace(printk(KERN_D 2627 dxtrace(printk(KERN_DEBUG
2628 "Creat 2628 "Creating %d level index...\n",
2629 dxroot 2629 dxroot->info.indirect_levels));
2630 err = ext4_handle_dir 2630 err = ext4_handle_dirty_dx_node(handle, dir, frame->bh);
2631 if (err) 2631 if (err)
2632 goto journal_ 2632 goto journal_error;
2633 err = ext4_handle_dir 2633 err = ext4_handle_dirty_dx_node(handle, dir, bh2);
2634 brelse(bh2); 2634 brelse(bh2);
2635 restart = 1; 2635 restart = 1;
2636 goto journal_error; 2636 goto journal_error;
2637 } 2637 }
2638 } 2638 }
2639 de = do_split(handle, dir, &bh, frame 2639 de = do_split(handle, dir, &bh, frame, &fname->hinfo);
2640 if (IS_ERR(de)) { 2640 if (IS_ERR(de)) {
2641 err = PTR_ERR(de); 2641 err = PTR_ERR(de);
2642 goto cleanup; 2642 goto cleanup;
2643 } 2643 }
2644 err = add_dirent_to_buf(handle, fname 2644 err = add_dirent_to_buf(handle, fname, dir, inode, de, bh);
2645 goto cleanup; 2645 goto cleanup;
2646 2646
2647 journal_error: 2647 journal_error:
2648 ext4_std_error(dir->i_sb, err); /* th 2648 ext4_std_error(dir->i_sb, err); /* this is a no-op if err == 0 */
2649 cleanup: 2649 cleanup:
2650 brelse(bh); 2650 brelse(bh);
2651 dx_release(frames); 2651 dx_release(frames);
2652 /* @restart is true means htree-path 2652 /* @restart is true means htree-path has been changed, we need to
2653 * repeat dx_probe() to find out vali 2653 * repeat dx_probe() to find out valid htree-path
2654 */ 2654 */
2655 if (restart && err == 0) 2655 if (restart && err == 0)
2656 goto again; 2656 goto again;
2657 return err; 2657 return err;
2658 } 2658 }
2659 2659
2660 /* 2660 /*
2661 * ext4_generic_delete_entry deletes a direct 2661 * ext4_generic_delete_entry deletes a directory entry by merging it
2662 * with the previous entry 2662 * with the previous entry
2663 */ 2663 */
2664 int ext4_generic_delete_entry(struct inode *d 2664 int ext4_generic_delete_entry(struct inode *dir,
2665 struct ext4_dir 2665 struct ext4_dir_entry_2 *de_del,
2666 struct buffer_h 2666 struct buffer_head *bh,
2667 void *entry_buf 2667 void *entry_buf,
2668 int buf_size, 2668 int buf_size,
2669 int csum_size) 2669 int csum_size)
2670 { 2670 {
2671 struct ext4_dir_entry_2 *de, *pde; 2671 struct ext4_dir_entry_2 *de, *pde;
2672 unsigned int blocksize = dir->i_sb->s 2672 unsigned int blocksize = dir->i_sb->s_blocksize;
2673 int i; 2673 int i;
2674 2674
2675 i = 0; 2675 i = 0;
2676 pde = NULL; 2676 pde = NULL;
2677 de = entry_buf; 2677 de = entry_buf;
2678 while (i < buf_size - csum_size) { 2678 while (i < buf_size - csum_size) {
2679 if (ext4_check_dir_entry(dir, 2679 if (ext4_check_dir_entry(dir, NULL, de, bh,
2680 entr 2680 entry_buf, buf_size, i))
2681 return -EFSCORRUPTED; 2681 return -EFSCORRUPTED;
2682 if (de == de_del) { 2682 if (de == de_del) {
2683 if (pde) { 2683 if (pde) {
2684 pde->rec_len 2684 pde->rec_len = ext4_rec_len_to_disk(
2685 ext4_ 2685 ext4_rec_len_from_disk(pde->rec_len,
2686 2686 blocksize) +
2687 ext4_ 2687 ext4_rec_len_from_disk(de->rec_len,
2688 2688 blocksize),
2689 block 2689 blocksize);
2690 2690
2691 /* wipe entir 2691 /* wipe entire dir_entry */
2692 memset(de, 0, 2692 memset(de, 0, ext4_rec_len_from_disk(de->rec_len,
2693 2693 blocksize));
2694 } else { 2694 } else {
2695 /* wipe dir_e 2695 /* wipe dir_entry excluding the rec_len field */
2696 de->inode = 0 2696 de->inode = 0;
2697 memset(&de->n 2697 memset(&de->name_len, 0,
2698 ext4_ 2698 ext4_rec_len_from_disk(de->rec_len,
2699 2699 blocksize) -
2700 offse 2700 offsetof(struct ext4_dir_entry_2,
2701 2701 name_len));
2702 } 2702 }
2703 2703
2704 inode_inc_iversion(di 2704 inode_inc_iversion(dir);
2705 return 0; 2705 return 0;
2706 } 2706 }
2707 i += ext4_rec_len_from_disk(d 2707 i += ext4_rec_len_from_disk(de->rec_len, blocksize);
2708 pde = de; 2708 pde = de;
2709 de = ext4_next_entry(de, bloc 2709 de = ext4_next_entry(de, blocksize);
2710 } 2710 }
2711 return -ENOENT; 2711 return -ENOENT;
2712 } 2712 }
2713 2713
2714 static int ext4_delete_entry(handle_t *handle 2714 static int ext4_delete_entry(handle_t *handle,
2715 struct inode *di 2715 struct inode *dir,
2716 struct ext4_dir_ 2716 struct ext4_dir_entry_2 *de_del,
2717 struct buffer_he 2717 struct buffer_head *bh)
2718 { 2718 {
2719 int err, csum_size = 0; 2719 int err, csum_size = 0;
2720 2720
2721 if (ext4_has_inline_data(dir)) { 2721 if (ext4_has_inline_data(dir)) {
2722 int has_inline_data = 1; 2722 int has_inline_data = 1;
2723 err = ext4_delete_inline_entr 2723 err = ext4_delete_inline_entry(handle, dir, de_del, bh,
2724 2724 &has_inline_data);
2725 if (has_inline_data) 2725 if (has_inline_data)
2726 return err; 2726 return err;
2727 } 2727 }
2728 2728
2729 if (ext4_has_metadata_csum(dir->i_sb) 2729 if (ext4_has_metadata_csum(dir->i_sb))
2730 csum_size = sizeof(struct ext 2730 csum_size = sizeof(struct ext4_dir_entry_tail);
2731 2731
2732 BUFFER_TRACE(bh, "get_write_access"); 2732 BUFFER_TRACE(bh, "get_write_access");
2733 err = ext4_journal_get_write_access(h 2733 err = ext4_journal_get_write_access(handle, dir->i_sb, bh,
2734 E 2734 EXT4_JTR_NONE);
2735 if (unlikely(err)) 2735 if (unlikely(err))
2736 goto out; 2736 goto out;
2737 2737
2738 err = ext4_generic_delete_entry(dir, 2738 err = ext4_generic_delete_entry(dir, de_del, bh, bh->b_data,
2739 dir-> 2739 dir->i_sb->s_blocksize, csum_size);
2740 if (err) 2740 if (err)
2741 goto out; 2741 goto out;
2742 2742
2743 BUFFER_TRACE(bh, "call ext4_handle_di 2743 BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata");
2744 err = ext4_handle_dirty_dirblock(hand 2744 err = ext4_handle_dirty_dirblock(handle, dir, bh);
2745 if (unlikely(err)) 2745 if (unlikely(err))
2746 goto out; 2746 goto out;
2747 2747
2748 return 0; 2748 return 0;
2749 out: 2749 out:
2750 if (err != -ENOENT) 2750 if (err != -ENOENT)
2751 ext4_std_error(dir->i_sb, err 2751 ext4_std_error(dir->i_sb, err);
2752 return err; 2752 return err;
2753 } 2753 }
2754 2754
2755 /* 2755 /*
2756 * Set directory link count to 1 if nlinks > 2756 * Set directory link count to 1 if nlinks > EXT4_LINK_MAX, or if nlinks == 2
2757 * since this indicates that nlinks count was 2757 * since this indicates that nlinks count was previously 1 to avoid overflowing
2758 * the 16-bit i_links_count field on disk. D 2758 * the 16-bit i_links_count field on disk. Directories with i_nlink == 1 mean
2759 * that subdirectory link counts are not bein 2759 * that subdirectory link counts are not being maintained accurately.
2760 * 2760 *
2761 * The caller has already checked for i_nlink 2761 * The caller has already checked for i_nlink overflow in case the DIR_LINK
2762 * feature is not enabled and returned -EMLIN 2762 * feature is not enabled and returned -EMLINK. The is_dx() check is a proxy
2763 * for checking S_ISDIR(inode) (since the INO 2763 * for checking S_ISDIR(inode) (since the INODE_INDEX feature will not be set
2764 * on regular files) and to avoid creating hu 2764 * on regular files) and to avoid creating huge/slow non-HTREE directories.
2765 */ 2765 */
2766 static void ext4_inc_count(struct inode *inod 2766 static void ext4_inc_count(struct inode *inode)
2767 { 2767 {
2768 inc_nlink(inode); 2768 inc_nlink(inode);
2769 if (is_dx(inode) && 2769 if (is_dx(inode) &&
2770 (inode->i_nlink > EXT4_LINK_MAX | 2770 (inode->i_nlink > EXT4_LINK_MAX || inode->i_nlink == 2))
2771 set_nlink(inode, 1); 2771 set_nlink(inode, 1);
2772 } 2772 }
2773 2773
2774 /* 2774 /*
2775 * If a directory had nlink == 1, then we sho 2775 * If a directory had nlink == 1, then we should let it be 1. This indicates
2776 * directory has >EXT4_LINK_MAX subdirs. 2776 * directory has >EXT4_LINK_MAX subdirs.
2777 */ 2777 */
2778 static void ext4_dec_count(struct inode *inod 2778 static void ext4_dec_count(struct inode *inode)
2779 { 2779 {
2780 if (!S_ISDIR(inode->i_mode) || inode- 2780 if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 2)
2781 drop_nlink(inode); 2781 drop_nlink(inode);
2782 } 2782 }
2783 2783
2784 2784
2785 /* 2785 /*
2786 * Add non-directory inode to a directory. On 2786 * Add non-directory inode to a directory. On success, the inode reference is
2787 * consumed by dentry is instantiation. This 2787 * consumed by dentry is instantiation. This is also indicated by clearing of
2788 * *inodep pointer. On failure, the caller is 2788 * *inodep pointer. On failure, the caller is responsible for dropping the
2789 * inode reference in the safe context. 2789 * inode reference in the safe context.
2790 */ 2790 */
2791 static int ext4_add_nondir(handle_t *handle, 2791 static int ext4_add_nondir(handle_t *handle,
2792 struct dentry *dentry, struct 2792 struct dentry *dentry, struct inode **inodep)
2793 { 2793 {
2794 struct inode *dir = d_inode(dentry->d 2794 struct inode *dir = d_inode(dentry->d_parent);
2795 struct inode *inode = *inodep; 2795 struct inode *inode = *inodep;
2796 int err = ext4_add_entry(handle, dent 2796 int err = ext4_add_entry(handle, dentry, inode);
2797 if (!err) { 2797 if (!err) {
2798 err = ext4_mark_inode_dirty(h 2798 err = ext4_mark_inode_dirty(handle, inode);
2799 if (IS_DIRSYNC(dir)) 2799 if (IS_DIRSYNC(dir))
2800 ext4_handle_sync(hand 2800 ext4_handle_sync(handle);
2801 d_instantiate_new(dentry, ino 2801 d_instantiate_new(dentry, inode);
2802 *inodep = NULL; 2802 *inodep = NULL;
2803 return err; 2803 return err;
2804 } 2804 }
2805 drop_nlink(inode); 2805 drop_nlink(inode);
2806 ext4_mark_inode_dirty(handle, inode); 2806 ext4_mark_inode_dirty(handle, inode);
2807 ext4_orphan_add(handle, inode); 2807 ext4_orphan_add(handle, inode);
2808 unlock_new_inode(inode); 2808 unlock_new_inode(inode);
2809 return err; 2809 return err;
2810 } 2810 }
2811 2811
2812 /* 2812 /*
2813 * By the time this is called, we already hav 2813 * By the time this is called, we already have created
2814 * the directory cache entry for the new file 2814 * the directory cache entry for the new file, but it
2815 * is so far negative - it has no inode. 2815 * is so far negative - it has no inode.
2816 * 2816 *
2817 * If the create succeeds, we fill in the ino 2817 * If the create succeeds, we fill in the inode information
2818 * with d_instantiate(). 2818 * with d_instantiate().
2819 */ 2819 */
2820 static int ext4_create(struct mnt_idmap *idma 2820 static int ext4_create(struct mnt_idmap *idmap, struct inode *dir,
2821 struct dentry *dentry, 2821 struct dentry *dentry, umode_t mode, bool excl)
2822 { 2822 {
2823 handle_t *handle; 2823 handle_t *handle;
2824 struct inode *inode; 2824 struct inode *inode;
2825 int err, credits, retries = 0; 2825 int err, credits, retries = 0;
2826 2826
2827 err = dquot_initialize(dir); 2827 err = dquot_initialize(dir);
2828 if (err) 2828 if (err)
2829 return err; 2829 return err;
2830 2830
2831 credits = (EXT4_DATA_TRANS_BLOCKS(dir 2831 credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
2832 EXT4_INDEX_EXTRA_TRANS_BLO 2832 EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3);
2833 retry: 2833 retry:
2834 inode = ext4_new_inode_start_handle(i 2834 inode = ext4_new_inode_start_handle(idmap, dir, mode, &dentry->d_name,
2835 0 2835 0, NULL, EXT4_HT_DIR, credits);
2836 handle = ext4_journal_current_handle( 2836 handle = ext4_journal_current_handle();
2837 err = PTR_ERR(inode); 2837 err = PTR_ERR(inode);
2838 if (!IS_ERR(inode)) { 2838 if (!IS_ERR(inode)) {
2839 inode->i_op = &ext4_file_inod 2839 inode->i_op = &ext4_file_inode_operations;
2840 inode->i_fop = &ext4_file_ope 2840 inode->i_fop = &ext4_file_operations;
2841 ext4_set_aops(inode); 2841 ext4_set_aops(inode);
2842 err = ext4_add_nondir(handle, 2842 err = ext4_add_nondir(handle, dentry, &inode);
2843 if (!err) 2843 if (!err)
2844 ext4_fc_track_create( 2844 ext4_fc_track_create(handle, dentry);
2845 } 2845 }
2846 if (handle) 2846 if (handle)
2847 ext4_journal_stop(handle); 2847 ext4_journal_stop(handle);
2848 if (!IS_ERR_OR_NULL(inode)) 2848 if (!IS_ERR_OR_NULL(inode))
2849 iput(inode); 2849 iput(inode);
2850 if (err == -ENOSPC && ext4_should_ret 2850 if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2851 goto retry; 2851 goto retry;
2852 return err; 2852 return err;
2853 } 2853 }
2854 2854
2855 static int ext4_mknod(struct mnt_idmap *idmap 2855 static int ext4_mknod(struct mnt_idmap *idmap, struct inode *dir,
2856 struct dentry *dentry, 2856 struct dentry *dentry, umode_t mode, dev_t rdev)
2857 { 2857 {
2858 handle_t *handle; 2858 handle_t *handle;
2859 struct inode *inode; 2859 struct inode *inode;
2860 int err, credits, retries = 0; 2860 int err, credits, retries = 0;
2861 2861
2862 err = dquot_initialize(dir); 2862 err = dquot_initialize(dir);
2863 if (err) 2863 if (err)
2864 return err; 2864 return err;
2865 2865
2866 credits = (EXT4_DATA_TRANS_BLOCKS(dir 2866 credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
2867 EXT4_INDEX_EXTRA_TRANS_BLO 2867 EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3);
2868 retry: 2868 retry:
2869 inode = ext4_new_inode_start_handle(i 2869 inode = ext4_new_inode_start_handle(idmap, dir, mode, &dentry->d_name,
2870 0 2870 0, NULL, EXT4_HT_DIR, credits);
2871 handle = ext4_journal_current_handle( 2871 handle = ext4_journal_current_handle();
2872 err = PTR_ERR(inode); 2872 err = PTR_ERR(inode);
2873 if (!IS_ERR(inode)) { 2873 if (!IS_ERR(inode)) {
2874 init_special_inode(inode, ino 2874 init_special_inode(inode, inode->i_mode, rdev);
2875 inode->i_op = &ext4_special_i 2875 inode->i_op = &ext4_special_inode_operations;
2876 err = ext4_add_nondir(handle, 2876 err = ext4_add_nondir(handle, dentry, &inode);
2877 if (!err) 2877 if (!err)
2878 ext4_fc_track_create( 2878 ext4_fc_track_create(handle, dentry);
2879 } 2879 }
2880 if (handle) 2880 if (handle)
2881 ext4_journal_stop(handle); 2881 ext4_journal_stop(handle);
2882 if (!IS_ERR_OR_NULL(inode)) 2882 if (!IS_ERR_OR_NULL(inode))
2883 iput(inode); 2883 iput(inode);
2884 if (err == -ENOSPC && ext4_should_ret 2884 if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2885 goto retry; 2885 goto retry;
2886 return err; 2886 return err;
2887 } 2887 }
2888 2888
2889 static int ext4_tmpfile(struct mnt_idmap *idm 2889 static int ext4_tmpfile(struct mnt_idmap *idmap, struct inode *dir,
2890 struct file *file, um 2890 struct file *file, umode_t mode)
2891 { 2891 {
2892 handle_t *handle; 2892 handle_t *handle;
2893 struct inode *inode; 2893 struct inode *inode;
2894 int err, retries = 0; 2894 int err, retries = 0;
2895 2895
2896 err = dquot_initialize(dir); 2896 err = dquot_initialize(dir);
2897 if (err) 2897 if (err)
2898 return err; 2898 return err;
2899 2899
2900 retry: 2900 retry:
2901 inode = ext4_new_inode_start_handle(i 2901 inode = ext4_new_inode_start_handle(idmap, dir, mode,
2902 N 2902 NULL, 0, NULL,
2903 E 2903 EXT4_HT_DIR,
2904 EXT4_MAXQUOTAS_TRANS_ 2904 EXT4_MAXQUOTAS_TRANS_BLOCKS(dir->i_sb) +
2905 4 + EXT4_XATTR_TRAN 2905 4 + EXT4_XATTR_TRANS_BLOCKS);
2906 handle = ext4_journal_current_handle( 2906 handle = ext4_journal_current_handle();
2907 err = PTR_ERR(inode); 2907 err = PTR_ERR(inode);
2908 if (!IS_ERR(inode)) { 2908 if (!IS_ERR(inode)) {
2909 inode->i_op = &ext4_file_inod 2909 inode->i_op = &ext4_file_inode_operations;
2910 inode->i_fop = &ext4_file_ope 2910 inode->i_fop = &ext4_file_operations;
2911 ext4_set_aops(inode); 2911 ext4_set_aops(inode);
2912 d_tmpfile(file, inode); 2912 d_tmpfile(file, inode);
2913 err = ext4_orphan_add(handle, 2913 err = ext4_orphan_add(handle, inode);
2914 if (err) 2914 if (err)
2915 goto err_unlock_inode 2915 goto err_unlock_inode;
2916 mark_inode_dirty(inode); 2916 mark_inode_dirty(inode);
2917 unlock_new_inode(inode); 2917 unlock_new_inode(inode);
2918 } 2918 }
2919 if (handle) 2919 if (handle)
2920 ext4_journal_stop(handle); 2920 ext4_journal_stop(handle);
2921 if (err == -ENOSPC && ext4_should_ret 2921 if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
2922 goto retry; 2922 goto retry;
2923 return finish_open_simple(file, err); 2923 return finish_open_simple(file, err);
2924 err_unlock_inode: 2924 err_unlock_inode:
2925 ext4_journal_stop(handle); 2925 ext4_journal_stop(handle);
2926 unlock_new_inode(inode); 2926 unlock_new_inode(inode);
2927 return err; 2927 return err;
2928 } 2928 }
2929 2929
2930 struct ext4_dir_entry_2 *ext4_init_dot_dotdot 2930 struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode,
2931 struct ext4_dir_ent 2931 struct ext4_dir_entry_2 *de,
2932 int blocksize, int 2932 int blocksize, int csum_size,
2933 unsigned int parent 2933 unsigned int parent_ino, int dotdot_real_len)
2934 { 2934 {
2935 de->inode = cpu_to_le32(inode->i_ino) 2935 de->inode = cpu_to_le32(inode->i_ino);
2936 de->name_len = 1; 2936 de->name_len = 1;
2937 de->rec_len = ext4_rec_len_to_disk(ex 2937 de->rec_len = ext4_rec_len_to_disk(ext4_dir_rec_len(de->name_len, NULL),
2938 bl 2938 blocksize);
2939 strcpy(de->name, "."); 2939 strcpy(de->name, ".");
2940 ext4_set_de_type(inode->i_sb, de, S_I 2940 ext4_set_de_type(inode->i_sb, de, S_IFDIR);
2941 2941
2942 de = ext4_next_entry(de, blocksize); 2942 de = ext4_next_entry(de, blocksize);
2943 de->inode = cpu_to_le32(parent_ino); 2943 de->inode = cpu_to_le32(parent_ino);
2944 de->name_len = 2; 2944 de->name_len = 2;
2945 if (!dotdot_real_len) 2945 if (!dotdot_real_len)
2946 de->rec_len = ext4_rec_len_to 2946 de->rec_len = ext4_rec_len_to_disk(blocksize -
2947 (csum 2947 (csum_size + ext4_dir_rec_len(1, NULL)),
2948 block 2948 blocksize);
2949 else 2949 else
2950 de->rec_len = ext4_rec_len_to 2950 de->rec_len = ext4_rec_len_to_disk(
2951 ext4_ 2951 ext4_dir_rec_len(de->name_len, NULL),
2952 block 2952 blocksize);
2953 strcpy(de->name, ".."); 2953 strcpy(de->name, "..");
2954 ext4_set_de_type(inode->i_sb, de, S_I 2954 ext4_set_de_type(inode->i_sb, de, S_IFDIR);
2955 2955
2956 return ext4_next_entry(de, blocksize) 2956 return ext4_next_entry(de, blocksize);
2957 } 2957 }
2958 2958
2959 int ext4_init_new_dir(handle_t *handle, struc 2959 int ext4_init_new_dir(handle_t *handle, struct inode *dir,
2960 struct inode *in 2960 struct inode *inode)
2961 { 2961 {
2962 struct buffer_head *dir_block = NULL; 2962 struct buffer_head *dir_block = NULL;
2963 struct ext4_dir_entry_2 *de; 2963 struct ext4_dir_entry_2 *de;
2964 ext4_lblk_t block = 0; 2964 ext4_lblk_t block = 0;
2965 unsigned int blocksize = dir->i_sb->s 2965 unsigned int blocksize = dir->i_sb->s_blocksize;
2966 int csum_size = 0; 2966 int csum_size = 0;
2967 int err; 2967 int err;
2968 2968
2969 if (ext4_has_metadata_csum(dir->i_sb) 2969 if (ext4_has_metadata_csum(dir->i_sb))
2970 csum_size = sizeof(struct ext 2970 csum_size = sizeof(struct ext4_dir_entry_tail);
2971 2971
2972 if (ext4_test_inode_state(inode, EXT4 2972 if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) {
2973 err = ext4_try_create_inline_ 2973 err = ext4_try_create_inline_dir(handle, dir, inode);
2974 if (err < 0 && err != -ENOSPC 2974 if (err < 0 && err != -ENOSPC)
2975 goto out; 2975 goto out;
2976 if (!err) 2976 if (!err)
2977 goto out; 2977 goto out;
2978 } 2978 }
2979 2979
2980 inode->i_size = 0; 2980 inode->i_size = 0;
2981 dir_block = ext4_append(handle, inode 2981 dir_block = ext4_append(handle, inode, &block);
2982 if (IS_ERR(dir_block)) 2982 if (IS_ERR(dir_block))
2983 return PTR_ERR(dir_block); 2983 return PTR_ERR(dir_block);
2984 de = (struct ext4_dir_entry_2 *)dir_b 2984 de = (struct ext4_dir_entry_2 *)dir_block->b_data;
2985 ext4_init_dot_dotdot(inode, de, block 2985 ext4_init_dot_dotdot(inode, de, blocksize, csum_size, dir->i_ino, 0);
2986 set_nlink(inode, 2); 2986 set_nlink(inode, 2);
2987 if (csum_size) 2987 if (csum_size)
2988 ext4_initialize_dirent_tail(d 2988 ext4_initialize_dirent_tail(dir_block, blocksize);
2989 2989
2990 BUFFER_TRACE(dir_block, "call ext4_ha 2990 BUFFER_TRACE(dir_block, "call ext4_handle_dirty_metadata");
2991 err = ext4_handle_dirty_dirblock(hand 2991 err = ext4_handle_dirty_dirblock(handle, inode, dir_block);
2992 if (err) 2992 if (err)
2993 goto out; 2993 goto out;
2994 set_buffer_verified(dir_block); 2994 set_buffer_verified(dir_block);
2995 out: 2995 out:
2996 brelse(dir_block); 2996 brelse(dir_block);
2997 return err; 2997 return err;
2998 } 2998 }
2999 2999
3000 static int ext4_mkdir(struct mnt_idmap *idmap 3000 static int ext4_mkdir(struct mnt_idmap *idmap, struct inode *dir,
3001 struct dentry *dentry, 3001 struct dentry *dentry, umode_t mode)
3002 { 3002 {
3003 handle_t *handle; 3003 handle_t *handle;
3004 struct inode *inode; 3004 struct inode *inode;
3005 int err, err2 = 0, credits, retries = 3005 int err, err2 = 0, credits, retries = 0;
3006 3006
3007 if (EXT4_DIR_LINK_MAX(dir)) 3007 if (EXT4_DIR_LINK_MAX(dir))
3008 return -EMLINK; 3008 return -EMLINK;
3009 3009
3010 err = dquot_initialize(dir); 3010 err = dquot_initialize(dir);
3011 if (err) 3011 if (err)
3012 return err; 3012 return err;
3013 3013
3014 credits = (EXT4_DATA_TRANS_BLOCKS(dir 3014 credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
3015 EXT4_INDEX_EXTRA_TRANS_BLO 3015 EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3);
3016 retry: 3016 retry:
3017 inode = ext4_new_inode_start_handle(i 3017 inode = ext4_new_inode_start_handle(idmap, dir, S_IFDIR | mode,
3018 & 3018 &dentry->d_name,
3019 0 3019 0, NULL, EXT4_HT_DIR, credits);
3020 handle = ext4_journal_current_handle( 3020 handle = ext4_journal_current_handle();
3021 err = PTR_ERR(inode); 3021 err = PTR_ERR(inode);
3022 if (IS_ERR(inode)) 3022 if (IS_ERR(inode))
3023 goto out_stop; 3023 goto out_stop;
3024 3024
3025 inode->i_op = &ext4_dir_inode_operati 3025 inode->i_op = &ext4_dir_inode_operations;
3026 inode->i_fop = &ext4_dir_operations; 3026 inode->i_fop = &ext4_dir_operations;
3027 err = ext4_init_new_dir(handle, dir, 3027 err = ext4_init_new_dir(handle, dir, inode);
3028 if (err) 3028 if (err)
3029 goto out_clear_inode; 3029 goto out_clear_inode;
3030 err = ext4_mark_inode_dirty(handle, i 3030 err = ext4_mark_inode_dirty(handle, inode);
3031 if (!err) 3031 if (!err)
3032 err = ext4_add_entry(handle, 3032 err = ext4_add_entry(handle, dentry, inode);
3033 if (err) { 3033 if (err) {
3034 out_clear_inode: 3034 out_clear_inode:
3035 clear_nlink(inode); 3035 clear_nlink(inode);
3036 ext4_orphan_add(handle, inode 3036 ext4_orphan_add(handle, inode);
3037 unlock_new_inode(inode); 3037 unlock_new_inode(inode);
3038 err2 = ext4_mark_inode_dirty( 3038 err2 = ext4_mark_inode_dirty(handle, inode);
3039 if (unlikely(err2)) 3039 if (unlikely(err2))
3040 err = err2; 3040 err = err2;
3041 ext4_journal_stop(handle); 3041 ext4_journal_stop(handle);
3042 iput(inode); 3042 iput(inode);
3043 goto out_retry; 3043 goto out_retry;
3044 } 3044 }
3045 ext4_inc_count(dir); 3045 ext4_inc_count(dir);
3046 3046
3047 ext4_update_dx_flag(dir); 3047 ext4_update_dx_flag(dir);
3048 err = ext4_mark_inode_dirty(handle, d 3048 err = ext4_mark_inode_dirty(handle, dir);
3049 if (err) 3049 if (err)
3050 goto out_clear_inode; 3050 goto out_clear_inode;
3051 d_instantiate_new(dentry, inode); 3051 d_instantiate_new(dentry, inode);
3052 ext4_fc_track_create(handle, dentry); 3052 ext4_fc_track_create(handle, dentry);
3053 if (IS_DIRSYNC(dir)) 3053 if (IS_DIRSYNC(dir))
3054 ext4_handle_sync(handle); 3054 ext4_handle_sync(handle);
3055 3055
3056 out_stop: 3056 out_stop:
3057 if (handle) 3057 if (handle)
3058 ext4_journal_stop(handle); 3058 ext4_journal_stop(handle);
3059 out_retry: 3059 out_retry:
3060 if (err == -ENOSPC && ext4_should_ret 3060 if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
3061 goto retry; 3061 goto retry;
3062 return err; 3062 return err;
3063 } 3063 }
3064 3064
3065 /* 3065 /*
3066 * routine to check that the specified direct 3066 * routine to check that the specified directory is empty (for rmdir)
3067 */ 3067 */
3068 bool ext4_empty_dir(struct inode *inode) 3068 bool ext4_empty_dir(struct inode *inode)
3069 { 3069 {
3070 unsigned int offset; 3070 unsigned int offset;
3071 struct buffer_head *bh; 3071 struct buffer_head *bh;
3072 struct ext4_dir_entry_2 *de; 3072 struct ext4_dir_entry_2 *de;
3073 struct super_block *sb; 3073 struct super_block *sb;
3074 3074
3075 if (ext4_has_inline_data(inode)) { 3075 if (ext4_has_inline_data(inode)) {
3076 int has_inline_data = 1; 3076 int has_inline_data = 1;
3077 int ret; 3077 int ret;
3078 3078
3079 ret = empty_inline_dir(inode, 3079 ret = empty_inline_dir(inode, &has_inline_data);
3080 if (has_inline_data) 3080 if (has_inline_data)
3081 return ret; 3081 return ret;
3082 } 3082 }
3083 3083
3084 sb = inode->i_sb; 3084 sb = inode->i_sb;
3085 if (inode->i_size < ext4_dir_rec_len( 3085 if (inode->i_size < ext4_dir_rec_len(1, NULL) +
3086 ext4_ 3086 ext4_dir_rec_len(2, NULL)) {
3087 EXT4_ERROR_INODE(inode, "inva 3087 EXT4_ERROR_INODE(inode, "invalid size");
3088 return false; 3088 return false;
3089 } 3089 }
3090 bh = ext4_read_dirblock(inode, 0, EIT 3090 bh = ext4_read_dirblock(inode, 0, EITHER);
3091 if (IS_ERR(bh)) 3091 if (IS_ERR(bh))
3092 return false; 3092 return false;
3093 3093
3094 de = (struct ext4_dir_entry_2 *) bh-> 3094 de = (struct ext4_dir_entry_2 *) bh->b_data;
3095 if (ext4_check_dir_entry(inode, NULL, 3095 if (ext4_check_dir_entry(inode, NULL, de, bh, bh->b_data, bh->b_size,
3096 0) || 3096 0) ||
3097 le32_to_cpu(de->inode) != inode-> 3097 le32_to_cpu(de->inode) != inode->i_ino || strcmp(".", de->name)) {
3098 ext4_warning_inode(inode, "di 3098 ext4_warning_inode(inode, "directory missing '.'");
3099 brelse(bh); 3099 brelse(bh);
3100 return false; 3100 return false;
3101 } 3101 }
3102 offset = ext4_rec_len_from_disk(de->r 3102 offset = ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize);
3103 de = ext4_next_entry(de, sb->s_blocks 3103 de = ext4_next_entry(de, sb->s_blocksize);
3104 if (ext4_check_dir_entry(inode, NULL, 3104 if (ext4_check_dir_entry(inode, NULL, de, bh, bh->b_data, bh->b_size,
3105 offset) || 3105 offset) ||
3106 le32_to_cpu(de->inode) == 0 || st 3106 le32_to_cpu(de->inode) == 0 || strcmp("..", de->name)) {
3107 ext4_warning_inode(inode, "di 3107 ext4_warning_inode(inode, "directory missing '..'");
3108 brelse(bh); 3108 brelse(bh);
3109 return false; 3109 return false;
3110 } 3110 }
3111 offset += ext4_rec_len_from_disk(de-> 3111 offset += ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize);
3112 while (offset < inode->i_size) { 3112 while (offset < inode->i_size) {
3113 if (!(offset & (sb->s_blocksi 3113 if (!(offset & (sb->s_blocksize - 1))) {
3114 unsigned int lblock; 3114 unsigned int lblock;
3115 brelse(bh); 3115 brelse(bh);
3116 lblock = offset >> EX 3116 lblock = offset >> EXT4_BLOCK_SIZE_BITS(sb);
3117 bh = ext4_read_dirblo 3117 bh = ext4_read_dirblock(inode, lblock, EITHER);
3118 if (bh == NULL) { 3118 if (bh == NULL) {
3119 offset += sb- 3119 offset += sb->s_blocksize;
3120 continue; 3120 continue;
3121 } 3121 }
3122 if (IS_ERR(bh)) 3122 if (IS_ERR(bh))
3123 return false; 3123 return false;
3124 } 3124 }
3125 de = (struct ext4_dir_entry_2 3125 de = (struct ext4_dir_entry_2 *) (bh->b_data +
3126 (offs 3126 (offset & (sb->s_blocksize - 1)));
3127 if (ext4_check_dir_entry(inod 3127 if (ext4_check_dir_entry(inode, NULL, de, bh,
3128 bh-> 3128 bh->b_data, bh->b_size, offset) ||
3129 le32_to_cpu(de->inode)) { 3129 le32_to_cpu(de->inode)) {
3130 brelse(bh); 3130 brelse(bh);
3131 return false; 3131 return false;
3132 } 3132 }
3133 offset += ext4_rec_len_from_d 3133 offset += ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize);
3134 } 3134 }
3135 brelse(bh); 3135 brelse(bh);
3136 return true; 3136 return true;
3137 } 3137 }
3138 3138
3139 static int ext4_rmdir(struct inode *dir, stru 3139 static int ext4_rmdir(struct inode *dir, struct dentry *dentry)
3140 { 3140 {
3141 int retval; 3141 int retval;
3142 struct inode *inode; 3142 struct inode *inode;
3143 struct buffer_head *bh; 3143 struct buffer_head *bh;
3144 struct ext4_dir_entry_2 *de; 3144 struct ext4_dir_entry_2 *de;
3145 handle_t *handle = NULL; 3145 handle_t *handle = NULL;
3146 3146
3147 if (unlikely(ext4_forced_shutdown(dir 3147 if (unlikely(ext4_forced_shutdown(dir->i_sb)))
3148 return -EIO; 3148 return -EIO;
3149 3149
3150 /* Initialize quotas before so that e 3150 /* Initialize quotas before so that eventual writes go in
3151 * separate transaction */ 3151 * separate transaction */
3152 retval = dquot_initialize(dir); 3152 retval = dquot_initialize(dir);
3153 if (retval) 3153 if (retval)
3154 return retval; 3154 return retval;
3155 retval = dquot_initialize(d_inode(den 3155 retval = dquot_initialize(d_inode(dentry));
3156 if (retval) 3156 if (retval)
3157 return retval; 3157 return retval;
3158 3158
3159 retval = -ENOENT; 3159 retval = -ENOENT;
3160 bh = ext4_find_entry(dir, &dentry->d_ 3160 bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL);
3161 if (IS_ERR(bh)) 3161 if (IS_ERR(bh))
3162 return PTR_ERR(bh); 3162 return PTR_ERR(bh);
3163 if (!bh) 3163 if (!bh)
3164 goto end_rmdir; 3164 goto end_rmdir;
3165 3165
3166 inode = d_inode(dentry); 3166 inode = d_inode(dentry);
3167 3167
3168 retval = -EFSCORRUPTED; 3168 retval = -EFSCORRUPTED;
3169 if (le32_to_cpu(de->inode) != inode-> 3169 if (le32_to_cpu(de->inode) != inode->i_ino)
3170 goto end_rmdir; 3170 goto end_rmdir;
3171 3171
3172 retval = -ENOTEMPTY; 3172 retval = -ENOTEMPTY;
3173 if (!ext4_empty_dir(inode)) 3173 if (!ext4_empty_dir(inode))
3174 goto end_rmdir; 3174 goto end_rmdir;
3175 3175
3176 handle = ext4_journal_start(dir, EXT4 3176 handle = ext4_journal_start(dir, EXT4_HT_DIR,
3177 EXT4_DATA 3177 EXT4_DATA_TRANS_BLOCKS(dir->i_sb));
3178 if (IS_ERR(handle)) { 3178 if (IS_ERR(handle)) {
3179 retval = PTR_ERR(handle); 3179 retval = PTR_ERR(handle);
3180 handle = NULL; 3180 handle = NULL;
3181 goto end_rmdir; 3181 goto end_rmdir;
3182 } 3182 }
3183 3183
3184 if (IS_DIRSYNC(dir)) 3184 if (IS_DIRSYNC(dir))
3185 ext4_handle_sync(handle); 3185 ext4_handle_sync(handle);
3186 3186
3187 retval = ext4_delete_entry(handle, di 3187 retval = ext4_delete_entry(handle, dir, de, bh);
3188 if (retval) 3188 if (retval)
3189 goto end_rmdir; 3189 goto end_rmdir;
3190 if (!EXT4_DIR_LINK_EMPTY(inode)) 3190 if (!EXT4_DIR_LINK_EMPTY(inode))
3191 ext4_warning_inode(inode, 3191 ext4_warning_inode(inode,
3192 "empty directory 3192 "empty directory '%.*s' has too many links (%u)",
3193 dentry->d_name.l 3193 dentry->d_name.len, dentry->d_name.name,
3194 inode->i_nlink); 3194 inode->i_nlink);
3195 inode_inc_iversion(inode); 3195 inode_inc_iversion(inode);
3196 clear_nlink(inode); 3196 clear_nlink(inode);
3197 /* There's no need to set i_disksize: 3197 /* There's no need to set i_disksize: the fact that i_nlink is
3198 * zero will ensure that the right th 3198 * zero will ensure that the right thing happens during any
3199 * recovery. */ 3199 * recovery. */
3200 inode->i_size = 0; 3200 inode->i_size = 0;
3201 ext4_orphan_add(handle, inode); 3201 ext4_orphan_add(handle, inode);
3202 inode_set_mtime_to_ts(dir, inode_set_ 3202 inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
3203 inode_set_ctime_current(inode); 3203 inode_set_ctime_current(inode);
3204 retval = ext4_mark_inode_dirty(handle 3204 retval = ext4_mark_inode_dirty(handle, inode);
3205 if (retval) 3205 if (retval)
3206 goto end_rmdir; 3206 goto end_rmdir;
3207 ext4_dec_count(dir); 3207 ext4_dec_count(dir);
3208 ext4_update_dx_flag(dir); 3208 ext4_update_dx_flag(dir);
3209 ext4_fc_track_unlink(handle, dentry); 3209 ext4_fc_track_unlink(handle, dentry);
3210 retval = ext4_mark_inode_dirty(handle 3210 retval = ext4_mark_inode_dirty(handle, dir);
3211 3211
3212 /* VFS negative dentries are incompat 3212 /* VFS negative dentries are incompatible with Encoding and
3213 * Case-insensitiveness. Eventually w 3213 * Case-insensitiveness. Eventually we'll want avoid
3214 * invalidating the dentries here, al 3214 * invalidating the dentries here, alongside with returning the
3215 * negative dentries at ext4_lookup() 3215 * negative dentries at ext4_lookup(), when it is better
3216 * supported by the VFS for the CI ca 3216 * supported by the VFS for the CI case.
3217 */ 3217 */
3218 if (IS_ENABLED(CONFIG_UNICODE) && IS_ 3218 if (IS_ENABLED(CONFIG_UNICODE) && IS_CASEFOLDED(dir))
3219 d_invalidate(dentry); 3219 d_invalidate(dentry);
3220 3220
3221 end_rmdir: 3221 end_rmdir:
3222 brelse(bh); 3222 brelse(bh);
3223 if (handle) 3223 if (handle)
3224 ext4_journal_stop(handle); 3224 ext4_journal_stop(handle);
3225 return retval; 3225 return retval;
3226 } 3226 }
3227 3227
3228 int __ext4_unlink(struct inode *dir, const st 3228 int __ext4_unlink(struct inode *dir, const struct qstr *d_name,
3229 struct inode *inode, 3229 struct inode *inode,
3230 struct dentry *dentry /* NU 3230 struct dentry *dentry /* NULL during fast_commit recovery */)
3231 { 3231 {
3232 int retval = -ENOENT; 3232 int retval = -ENOENT;
3233 struct buffer_head *bh; 3233 struct buffer_head *bh;
3234 struct ext4_dir_entry_2 *de; 3234 struct ext4_dir_entry_2 *de;
3235 handle_t *handle; 3235 handle_t *handle;
3236 int skip_remove_dentry = 0; 3236 int skip_remove_dentry = 0;
3237 3237
3238 /* 3238 /*
3239 * Keep this outside the transaction; 3239 * Keep this outside the transaction; it may have to set up the
3240 * directory's encryption key, which 3240 * directory's encryption key, which isn't GFP_NOFS-safe.
3241 */ 3241 */
3242 bh = ext4_find_entry(dir, d_name, &de 3242 bh = ext4_find_entry(dir, d_name, &de, NULL);
3243 if (IS_ERR(bh)) 3243 if (IS_ERR(bh))
3244 return PTR_ERR(bh); 3244 return PTR_ERR(bh);
3245 3245
3246 if (!bh) 3246 if (!bh)
3247 return -ENOENT; 3247 return -ENOENT;
3248 3248
3249 if (le32_to_cpu(de->inode) != inode-> 3249 if (le32_to_cpu(de->inode) != inode->i_ino) {
3250 /* 3250 /*
3251 * It's okay if we find dont 3251 * It's okay if we find dont find dentry which matches
3252 * the inode. That's because 3252 * the inode. That's because it might have gotten
3253 * renamed to a different ino 3253 * renamed to a different inode number
3254 */ 3254 */
3255 if (EXT4_SB(inode->i_sb)->s_m 3255 if (EXT4_SB(inode->i_sb)->s_mount_state & EXT4_FC_REPLAY)
3256 skip_remove_dentry = 3256 skip_remove_dentry = 1;
3257 else 3257 else
3258 goto out_bh; 3258 goto out_bh;
3259 } 3259 }
3260 3260
3261 handle = ext4_journal_start(dir, EXT4 3261 handle = ext4_journal_start(dir, EXT4_HT_DIR,
3262 EXT4_DATA 3262 EXT4_DATA_TRANS_BLOCKS(dir->i_sb));
3263 if (IS_ERR(handle)) { 3263 if (IS_ERR(handle)) {
3264 retval = PTR_ERR(handle); 3264 retval = PTR_ERR(handle);
3265 goto out_bh; 3265 goto out_bh;
3266 } 3266 }
3267 3267
3268 if (IS_DIRSYNC(dir)) 3268 if (IS_DIRSYNC(dir))
3269 ext4_handle_sync(handle); 3269 ext4_handle_sync(handle);
3270 3270
3271 if (!skip_remove_dentry) { 3271 if (!skip_remove_dentry) {
3272 retval = ext4_delete_entry(ha 3272 retval = ext4_delete_entry(handle, dir, de, bh);
3273 if (retval) 3273 if (retval)
3274 goto out_handle; 3274 goto out_handle;
3275 inode_set_mtime_to_ts(dir, in 3275 inode_set_mtime_to_ts(dir, inode_set_ctime_current(dir));
3276 ext4_update_dx_flag(dir); 3276 ext4_update_dx_flag(dir);
3277 retval = ext4_mark_inode_dirt 3277 retval = ext4_mark_inode_dirty(handle, dir);
3278 if (retval) 3278 if (retval)
3279 goto out_handle; 3279 goto out_handle;
3280 } else { 3280 } else {
3281 retval = 0; 3281 retval = 0;
3282 } 3282 }
3283 if (inode->i_nlink == 0) 3283 if (inode->i_nlink == 0)
3284 ext4_warning_inode(inode, "De 3284 ext4_warning_inode(inode, "Deleting file '%.*s' with no links",
3285 d_name->le 3285 d_name->len, d_name->name);
3286 else 3286 else
3287 drop_nlink(inode); 3287 drop_nlink(inode);
3288 if (!inode->i_nlink) 3288 if (!inode->i_nlink)
3289 ext4_orphan_add(handle, inode 3289 ext4_orphan_add(handle, inode);
3290 inode_set_ctime_current(inode); 3290 inode_set_ctime_current(inode);
3291 retval = ext4_mark_inode_dirty(handle 3291 retval = ext4_mark_inode_dirty(handle, inode);
3292 if (dentry && !retval) 3292 if (dentry && !retval)
3293 ext4_fc_track_unlink(handle, 3293 ext4_fc_track_unlink(handle, dentry);
3294 out_handle: 3294 out_handle:
3295 ext4_journal_stop(handle); 3295 ext4_journal_stop(handle);
3296 out_bh: 3296 out_bh:
3297 brelse(bh); 3297 brelse(bh);
3298 return retval; 3298 return retval;
3299 } 3299 }
3300 3300
3301 static int ext4_unlink(struct inode *dir, str 3301 static int ext4_unlink(struct inode *dir, struct dentry *dentry)
3302 { 3302 {
3303 int retval; 3303 int retval;
3304 3304
3305 if (unlikely(ext4_forced_shutdown(dir 3305 if (unlikely(ext4_forced_shutdown(dir->i_sb)))
3306 return -EIO; 3306 return -EIO;
3307 3307
3308 trace_ext4_unlink_enter(dir, dentry); 3308 trace_ext4_unlink_enter(dir, dentry);
3309 /* 3309 /*
3310 * Initialize quotas before so that e 3310 * Initialize quotas before so that eventual writes go
3311 * in separate transaction 3311 * in separate transaction
3312 */ 3312 */
3313 retval = dquot_initialize(dir); 3313 retval = dquot_initialize(dir);
3314 if (retval) 3314 if (retval)
3315 goto out_trace; 3315 goto out_trace;
3316 retval = dquot_initialize(d_inode(den 3316 retval = dquot_initialize(d_inode(dentry));
3317 if (retval) 3317 if (retval)
3318 goto out_trace; 3318 goto out_trace;
3319 3319
3320 retval = __ext4_unlink(dir, &dentry-> 3320 retval = __ext4_unlink(dir, &dentry->d_name, d_inode(dentry), dentry);
3321 3321
3322 /* VFS negative dentries are incompat 3322 /* VFS negative dentries are incompatible with Encoding and
3323 * Case-insensitiveness. Eventually w 3323 * Case-insensitiveness. Eventually we'll want avoid
3324 * invalidating the dentries here, al 3324 * invalidating the dentries here, alongside with returning the
3325 * negative dentries at ext4_lookup() 3325 * negative dentries at ext4_lookup(), when it is better
3326 * supported by the VFS for the CI ca 3326 * supported by the VFS for the CI case.
3327 */ 3327 */
3328 if (IS_ENABLED(CONFIG_UNICODE) && IS_ 3328 if (IS_ENABLED(CONFIG_UNICODE) && IS_CASEFOLDED(dir))
3329 d_invalidate(dentry); 3329 d_invalidate(dentry);
3330 3330
3331 out_trace: 3331 out_trace:
3332 trace_ext4_unlink_exit(dentry, retval 3332 trace_ext4_unlink_exit(dentry, retval);
3333 return retval; 3333 return retval;
3334 } 3334 }
3335 3335
3336 static int ext4_init_symlink_block(handle_t * 3336 static int ext4_init_symlink_block(handle_t *handle, struct inode *inode,
3337 struct fsc 3337 struct fscrypt_str *disk_link)
3338 { 3338 {
3339 struct buffer_head *bh; 3339 struct buffer_head *bh;
3340 char *kaddr; 3340 char *kaddr;
3341 int err = 0; 3341 int err = 0;
3342 3342
3343 bh = ext4_bread(handle, inode, 0, EXT 3343 bh = ext4_bread(handle, inode, 0, EXT4_GET_BLOCKS_CREATE);
3344 if (IS_ERR(bh)) 3344 if (IS_ERR(bh))
3345 return PTR_ERR(bh); 3345 return PTR_ERR(bh);
3346 3346
3347 BUFFER_TRACE(bh, "get_write_access"); 3347 BUFFER_TRACE(bh, "get_write_access");
3348 err = ext4_journal_get_write_access(h 3348 err = ext4_journal_get_write_access(handle, inode->i_sb, bh, EXT4_JTR_NONE);
3349 if (err) 3349 if (err)
3350 goto out; 3350 goto out;
3351 3351
3352 kaddr = (char *)bh->b_data; 3352 kaddr = (char *)bh->b_data;
3353 memcpy(kaddr, disk_link->name, disk_l 3353 memcpy(kaddr, disk_link->name, disk_link->len);
3354 inode->i_size = disk_link->len - 1; 3354 inode->i_size = disk_link->len - 1;
3355 EXT4_I(inode)->i_disksize = inode->i_ 3355 EXT4_I(inode)->i_disksize = inode->i_size;
3356 err = ext4_handle_dirty_metadata(hand 3356 err = ext4_handle_dirty_metadata(handle, inode, bh);
3357 out: 3357 out:
3358 brelse(bh); 3358 brelse(bh);
3359 return err; 3359 return err;
3360 } 3360 }
3361 3361
3362 static int ext4_symlink(struct mnt_idmap *idm 3362 static int ext4_symlink(struct mnt_idmap *idmap, struct inode *dir,
3363 struct dentry *dentry 3363 struct dentry *dentry, const char *symname)
3364 { 3364 {
3365 handle_t *handle; 3365 handle_t *handle;
3366 struct inode *inode; 3366 struct inode *inode;
3367 int err, len = strlen(symname); 3367 int err, len = strlen(symname);
3368 int credits; 3368 int credits;
3369 struct fscrypt_str disk_link; 3369 struct fscrypt_str disk_link;
3370 int retries = 0; 3370 int retries = 0;
3371 3371
3372 if (unlikely(ext4_forced_shutdown(dir 3372 if (unlikely(ext4_forced_shutdown(dir->i_sb)))
3373 return -EIO; 3373 return -EIO;
3374 3374
3375 err = fscrypt_prepare_symlink(dir, sy 3375 err = fscrypt_prepare_symlink(dir, symname, len, dir->i_sb->s_blocksize,
3376 &disk_l 3376 &disk_link);
3377 if (err) 3377 if (err)
3378 return err; 3378 return err;
3379 3379
3380 err = dquot_initialize(dir); 3380 err = dquot_initialize(dir);
3381 if (err) 3381 if (err)
3382 return err; 3382 return err;
3383 3383
3384 /* 3384 /*
3385 * EXT4_INDEX_EXTRA_TRANS_BLOCKS for 3385 * EXT4_INDEX_EXTRA_TRANS_BLOCKS for addition of entry into the
3386 * directory. +3 for inode, inode bit 3386 * directory. +3 for inode, inode bitmap, group descriptor allocation.
3387 * EXT4_DATA_TRANS_BLOCKS for the dat 3387 * EXT4_DATA_TRANS_BLOCKS for the data block allocation and
3388 * modification. 3388 * modification.
3389 */ 3389 */
3390 credits = EXT4_DATA_TRANS_BLOCKS(dir- 3390 credits = EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
3391 EXT4_INDEX_EXTRA_TRANS_BLOC 3391 EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3;
3392 retry: 3392 retry:
3393 inode = ext4_new_inode_start_handle(i 3393 inode = ext4_new_inode_start_handle(idmap, dir, S_IFLNK|S_IRWXUGO,
3394 & 3394 &dentry->d_name, 0, NULL,
3395 E 3395 EXT4_HT_DIR, credits);
3396 handle = ext4_journal_current_handle( 3396 handle = ext4_journal_current_handle();
3397 if (IS_ERR(inode)) { 3397 if (IS_ERR(inode)) {
3398 if (handle) 3398 if (handle)
3399 ext4_journal_stop(han 3399 ext4_journal_stop(handle);
3400 err = PTR_ERR(inode); 3400 err = PTR_ERR(inode);
3401 goto out_retry; 3401 goto out_retry;
3402 } 3402 }
3403 3403
3404 if (IS_ENCRYPTED(inode)) { 3404 if (IS_ENCRYPTED(inode)) {
3405 err = fscrypt_encrypt_symlink 3405 err = fscrypt_encrypt_symlink(inode, symname, len, &disk_link);
3406 if (err) 3406 if (err)
3407 goto err_drop_inode; 3407 goto err_drop_inode;
3408 inode->i_op = &ext4_encrypted 3408 inode->i_op = &ext4_encrypted_symlink_inode_operations;
3409 } else { 3409 } else {
3410 if ((disk_link.len > EXT4_N_B 3410 if ((disk_link.len > EXT4_N_BLOCKS * 4)) {
3411 inode->i_op = &ext4_s 3411 inode->i_op = &ext4_symlink_inode_operations;
3412 } else { 3412 } else {
3413 inode->i_op = &ext4_f 3413 inode->i_op = &ext4_fast_symlink_inode_operations;
3414 inode->i_link = (char 3414 inode->i_link = (char *)&EXT4_I(inode)->i_data;
3415 } 3415 }
3416 } 3416 }
3417 3417
3418 if ((disk_link.len > EXT4_N_BLOCKS * 3418 if ((disk_link.len > EXT4_N_BLOCKS * 4)) {
3419 /* alloc symlink block and fi 3419 /* alloc symlink block and fill it */
3420 err = ext4_init_symlink_block 3420 err = ext4_init_symlink_block(handle, inode, &disk_link);
3421 if (err) 3421 if (err)
3422 goto err_drop_inode; 3422 goto err_drop_inode;
3423 } else { 3423 } else {
3424 /* clear the extent format fo 3424 /* clear the extent format for fast symlink */
3425 ext4_clear_inode_flag(inode, 3425 ext4_clear_inode_flag(inode, EXT4_INODE_EXTENTS);
3426 memcpy((char *)&EXT4_I(inode) 3426 memcpy((char *)&EXT4_I(inode)->i_data, disk_link.name,
3427 disk_link.len); 3427 disk_link.len);
3428 inode->i_size = disk_link.len 3428 inode->i_size = disk_link.len - 1;
3429 EXT4_I(inode)->i_disksize = i 3429 EXT4_I(inode)->i_disksize = inode->i_size;
3430 } 3430 }
3431 err = ext4_add_nondir(handle, dentry, 3431 err = ext4_add_nondir(handle, dentry, &inode);
3432 if (handle) 3432 if (handle)
3433 ext4_journal_stop(handle); 3433 ext4_journal_stop(handle);
3434 iput(inode); 3434 iput(inode);
3435 goto out_retry; 3435 goto out_retry;
3436 3436
3437 err_drop_inode: 3437 err_drop_inode:
3438 clear_nlink(inode); 3438 clear_nlink(inode);
3439 ext4_mark_inode_dirty(handle, inode); 3439 ext4_mark_inode_dirty(handle, inode);
3440 ext4_orphan_add(handle, inode); 3440 ext4_orphan_add(handle, inode);
3441 unlock_new_inode(inode); 3441 unlock_new_inode(inode);
3442 if (handle) 3442 if (handle)
3443 ext4_journal_stop(handle); 3443 ext4_journal_stop(handle);
3444 iput(inode); 3444 iput(inode);
3445 out_retry: 3445 out_retry:
3446 if (err == -ENOSPC && ext4_should_ret 3446 if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
3447 goto retry; 3447 goto retry;
3448 if (disk_link.name != (unsigned char 3448 if (disk_link.name != (unsigned char *)symname)
3449 kfree(disk_link.name); 3449 kfree(disk_link.name);
3450 return err; 3450 return err;
3451 } 3451 }
3452 3452
3453 int __ext4_link(struct inode *dir, struct ino 3453 int __ext4_link(struct inode *dir, struct inode *inode, struct dentry *dentry)
3454 { 3454 {
3455 handle_t *handle; 3455 handle_t *handle;
3456 int err, retries = 0; 3456 int err, retries = 0;
3457 retry: 3457 retry:
3458 handle = ext4_journal_start(dir, EXT4 3458 handle = ext4_journal_start(dir, EXT4_HT_DIR,
3459 (EXT4_DATA_TRANS_BLOCKS(dir-> 3459 (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) +
3460 EXT4_INDEX_EXTRA_TRANS_BLOCK 3460 EXT4_INDEX_EXTRA_TRANS_BLOCKS) + 1);
3461 if (IS_ERR(handle)) 3461 if (IS_ERR(handle))
3462 return PTR_ERR(handle); 3462 return PTR_ERR(handle);
3463 3463
3464 if (IS_DIRSYNC(dir)) 3464 if (IS_DIRSYNC(dir))
3465 ext4_handle_sync(handle); 3465 ext4_handle_sync(handle);
3466 3466
3467 inode_set_ctime_current(inode); 3467 inode_set_ctime_current(inode);
3468 ext4_inc_count(inode); 3468 ext4_inc_count(inode);
3469 ihold(inode); 3469 ihold(inode);
3470 3470
3471 err = ext4_add_entry(handle, dentry, 3471 err = ext4_add_entry(handle, dentry, inode);
3472 if (!err) { 3472 if (!err) {
3473 err = ext4_mark_inode_dirty(h 3473 err = ext4_mark_inode_dirty(handle, inode);
3474 /* this can happen only for t 3474 /* this can happen only for tmpfile being
3475 * linked the first time 3475 * linked the first time
3476 */ 3476 */
3477 if (inode->i_nlink == 1) 3477 if (inode->i_nlink == 1)
3478 ext4_orphan_del(handl 3478 ext4_orphan_del(handle, inode);
3479 d_instantiate(dentry, inode); 3479 d_instantiate(dentry, inode);
3480 ext4_fc_track_link(handle, de 3480 ext4_fc_track_link(handle, dentry);
3481 } else { 3481 } else {
3482 drop_nlink(inode); 3482 drop_nlink(inode);
3483 iput(inode); 3483 iput(inode);
3484 } 3484 }
3485 ext4_journal_stop(handle); 3485 ext4_journal_stop(handle);
3486 if (err == -ENOSPC && ext4_should_ret 3486 if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries))
3487 goto retry; 3487 goto retry;
3488 return err; 3488 return err;
3489 } 3489 }
3490 3490
3491 static int ext4_link(struct dentry *old_dentr 3491 static int ext4_link(struct dentry *old_dentry,
3492 struct inode *dir, struc 3492 struct inode *dir, struct dentry *dentry)
3493 { 3493 {
3494 struct inode *inode = d_inode(old_den 3494 struct inode *inode = d_inode(old_dentry);
3495 int err; 3495 int err;
3496 3496
3497 if (inode->i_nlink >= EXT4_LINK_MAX) 3497 if (inode->i_nlink >= EXT4_LINK_MAX)
3498 return -EMLINK; 3498 return -EMLINK;
3499 3499
3500 err = fscrypt_prepare_link(old_dentry 3500 err = fscrypt_prepare_link(old_dentry, dir, dentry);
3501 if (err) 3501 if (err)
3502 return err; 3502 return err;
3503 3503
3504 if ((ext4_test_inode_flag(dir, EXT4_I 3504 if ((ext4_test_inode_flag(dir, EXT4_INODE_PROJINHERIT)) &&
3505 (!projid_eq(EXT4_I(dir)->i_projid 3505 (!projid_eq(EXT4_I(dir)->i_projid,
3506 EXT4_I(old_dentry->d_ 3506 EXT4_I(old_dentry->d_inode)->i_projid)))
3507 return -EXDEV; 3507 return -EXDEV;
3508 3508
3509 err = dquot_initialize(dir); 3509 err = dquot_initialize(dir);
3510 if (err) 3510 if (err)
3511 return err; 3511 return err;
3512 return __ext4_link(dir, inode, dentry 3512 return __ext4_link(dir, inode, dentry);
3513 } 3513 }
3514 3514
3515 /* 3515 /*
3516 * Try to find buffer head where contains the 3516 * Try to find buffer head where contains the parent block.
3517 * It should be the inode block if it is inli 3517 * It should be the inode block if it is inlined or the 1st block
3518 * if it is a normal dir. 3518 * if it is a normal dir.
3519 */ 3519 */
3520 static struct buffer_head *ext4_get_first_dir 3520 static struct buffer_head *ext4_get_first_dir_block(handle_t *handle,
3521 struc 3521 struct inode *inode,
3522 int * 3522 int *retval,
3523 struc 3523 struct ext4_dir_entry_2 **parent_de,
3524 int * 3524 int *inlined)
3525 { 3525 {
3526 struct buffer_head *bh; 3526 struct buffer_head *bh;
3527 3527
3528 if (!ext4_has_inline_data(inode)) { 3528 if (!ext4_has_inline_data(inode)) {
3529 struct ext4_dir_entry_2 *de; 3529 struct ext4_dir_entry_2 *de;
3530 unsigned int offset; 3530 unsigned int offset;
3531 3531
3532 bh = ext4_read_dirblock(inode 3532 bh = ext4_read_dirblock(inode, 0, EITHER);
3533 if (IS_ERR(bh)) { 3533 if (IS_ERR(bh)) {
3534 *retval = PTR_ERR(bh) 3534 *retval = PTR_ERR(bh);
3535 return NULL; 3535 return NULL;
3536 } 3536 }
3537 3537
3538 de = (struct ext4_dir_entry_2 3538 de = (struct ext4_dir_entry_2 *) bh->b_data;
3539 if (ext4_check_dir_entry(inod 3539 if (ext4_check_dir_entry(inode, NULL, de, bh, bh->b_data,
3540 bh-> 3540 bh->b_size, 0) ||
3541 le32_to_cpu(de->inode) != 3541 le32_to_cpu(de->inode) != inode->i_ino ||
3542 strcmp(".", de->name)) { 3542 strcmp(".", de->name)) {
3543 EXT4_ERROR_INODE(inod 3543 EXT4_ERROR_INODE(inode, "directory missing '.'");
3544 brelse(bh); 3544 brelse(bh);
3545 *retval = -EFSCORRUPT 3545 *retval = -EFSCORRUPTED;
3546 return NULL; 3546 return NULL;
3547 } 3547 }
3548 offset = ext4_rec_len_from_di 3548 offset = ext4_rec_len_from_disk(de->rec_len,
3549 3549 inode->i_sb->s_blocksize);
3550 de = ext4_next_entry(de, inod 3550 de = ext4_next_entry(de, inode->i_sb->s_blocksize);
3551 if (ext4_check_dir_entry(inod 3551 if (ext4_check_dir_entry(inode, NULL, de, bh, bh->b_data,
3552 bh-> 3552 bh->b_size, offset) ||
3553 le32_to_cpu(de->inode) == 3553 le32_to_cpu(de->inode) == 0 || strcmp("..", de->name)) {
3554 EXT4_ERROR_INODE(inod 3554 EXT4_ERROR_INODE(inode, "directory missing '..'");
3555 brelse(bh); 3555 brelse(bh);
3556 *retval = -EFSCORRUPT 3556 *retval = -EFSCORRUPTED;
3557 return NULL; 3557 return NULL;
3558 } 3558 }
3559 *parent_de = de; 3559 *parent_de = de;
3560 3560
3561 return bh; 3561 return bh;
3562 } 3562 }
3563 3563
3564 *inlined = 1; 3564 *inlined = 1;
3565 return ext4_get_first_inline_block(in 3565 return ext4_get_first_inline_block(inode, parent_de, retval);
3566 } 3566 }
3567 3567
3568 struct ext4_renament { 3568 struct ext4_renament {
3569 struct inode *dir; 3569 struct inode *dir;
3570 struct dentry *dentry; 3570 struct dentry *dentry;
3571 struct inode *inode; 3571 struct inode *inode;
3572 bool is_dir; 3572 bool is_dir;
3573 int dir_nlink_delta; 3573 int dir_nlink_delta;
3574 3574
3575 /* entry for "dentry" */ 3575 /* entry for "dentry" */
3576 struct buffer_head *bh; 3576 struct buffer_head *bh;
3577 struct ext4_dir_entry_2 *de; 3577 struct ext4_dir_entry_2 *de;
3578 int inlined; 3578 int inlined;
3579 3579
3580 /* entry for ".." in inode if it's a 3580 /* entry for ".." in inode if it's a directory */
3581 struct buffer_head *dir_bh; 3581 struct buffer_head *dir_bh;
3582 struct ext4_dir_entry_2 *parent_de; 3582 struct ext4_dir_entry_2 *parent_de;
3583 int dir_inlined; 3583 int dir_inlined;
3584 }; 3584 };
3585 3585
3586 static int ext4_rename_dir_prepare(handle_t * 3586 static int ext4_rename_dir_prepare(handle_t *handle, struct ext4_renament *ent, bool is_cross)
3587 { 3587 {
3588 int retval; 3588 int retval;
3589 3589
3590 ent->is_dir = true; 3590 ent->is_dir = true;
3591 if (!is_cross) 3591 if (!is_cross)
3592 return 0; 3592 return 0;
3593 3593
3594 ent->dir_bh = ext4_get_first_dir_bloc 3594 ent->dir_bh = ext4_get_first_dir_block(handle, ent->inode,
3595 3595 &retval, &ent->parent_de,
3596 3596 &ent->dir_inlined);
3597 if (!ent->dir_bh) 3597 if (!ent->dir_bh)
3598 return retval; 3598 return retval;
3599 if (le32_to_cpu(ent->parent_de->inode 3599 if (le32_to_cpu(ent->parent_de->inode) != ent->dir->i_ino)
3600 return -EFSCORRUPTED; 3600 return -EFSCORRUPTED;
3601 BUFFER_TRACE(ent->dir_bh, "get_write_ 3601 BUFFER_TRACE(ent->dir_bh, "get_write_access");
3602 return ext4_journal_get_write_access( 3602 return ext4_journal_get_write_access(handle, ent->dir->i_sb,
3603 3603 ent->dir_bh, EXT4_JTR_NONE);
3604 } 3604 }
3605 3605
3606 static int ext4_rename_dir_finish(handle_t *h 3606 static int ext4_rename_dir_finish(handle_t *handle, struct ext4_renament *ent,
3607 unsigned di 3607 unsigned dir_ino)
3608 { 3608 {
3609 int retval; 3609 int retval;
3610 3610
3611 if (!ent->dir_bh) 3611 if (!ent->dir_bh)
3612 return 0; 3612 return 0;
3613 3613
3614 ent->parent_de->inode = cpu_to_le32(d 3614 ent->parent_de->inode = cpu_to_le32(dir_ino);
3615 BUFFER_TRACE(ent->dir_bh, "call ext4_ 3615 BUFFER_TRACE(ent->dir_bh, "call ext4_handle_dirty_metadata");
3616 if (!ent->dir_inlined) { 3616 if (!ent->dir_inlined) {
3617 if (is_dx(ent->inode)) { 3617 if (is_dx(ent->inode)) {
3618 retval = ext4_handle_ 3618 retval = ext4_handle_dirty_dx_node(handle,
3619 3619 ent->inode,
3620 3620 ent->dir_bh);
3621 } else { 3621 } else {
3622 retval = ext4_handle_ 3622 retval = ext4_handle_dirty_dirblock(handle, ent->inode,
3623 3623 ent->dir_bh);
3624 } 3624 }
3625 } else { 3625 } else {
3626 retval = ext4_mark_inode_dirt 3626 retval = ext4_mark_inode_dirty(handle, ent->inode);
3627 } 3627 }
3628 if (retval) { 3628 if (retval) {
3629 ext4_std_error(ent->dir->i_sb 3629 ext4_std_error(ent->dir->i_sb, retval);
3630 return retval; 3630 return retval;
3631 } 3631 }
3632 return 0; 3632 return 0;
3633 } 3633 }
3634 3634
3635 static int ext4_setent(handle_t *handle, stru 3635 static int ext4_setent(handle_t *handle, struct ext4_renament *ent,
3636 unsigned ino, unsigned 3636 unsigned ino, unsigned file_type)
3637 { 3637 {
3638 int retval, retval2; 3638 int retval, retval2;
3639 3639
3640 BUFFER_TRACE(ent->bh, "get write acce 3640 BUFFER_TRACE(ent->bh, "get write access");
3641 retval = ext4_journal_get_write_acces 3641 retval = ext4_journal_get_write_access(handle, ent->dir->i_sb, ent->bh,
3642 3642 EXT4_JTR_NONE);
3643 if (retval) 3643 if (retval)
3644 return retval; 3644 return retval;
3645 ent->de->inode = cpu_to_le32(ino); 3645 ent->de->inode = cpu_to_le32(ino);
3646 if (ext4_has_feature_filetype(ent->di 3646 if (ext4_has_feature_filetype(ent->dir->i_sb))
3647 ent->de->file_type = file_typ 3647 ent->de->file_type = file_type;
3648 inode_inc_iversion(ent->dir); 3648 inode_inc_iversion(ent->dir);
3649 inode_set_mtime_to_ts(ent->dir, inode 3649 inode_set_mtime_to_ts(ent->dir, inode_set_ctime_current(ent->dir));
3650 retval = ext4_mark_inode_dirty(handle 3650 retval = ext4_mark_inode_dirty(handle, ent->dir);
3651 BUFFER_TRACE(ent->bh, "call ext4_hand 3651 BUFFER_TRACE(ent->bh, "call ext4_handle_dirty_metadata");
3652 if (!ent->inlined) { 3652 if (!ent->inlined) {
3653 retval2 = ext4_handle_dirty_d 3653 retval2 = ext4_handle_dirty_dirblock(handle, ent->dir, ent->bh);
3654 if (unlikely(retval2)) { 3654 if (unlikely(retval2)) {
3655 ext4_std_error(ent->d 3655 ext4_std_error(ent->dir->i_sb, retval2);
3656 return retval2; 3656 return retval2;
3657 } 3657 }
3658 } 3658 }
3659 return retval; 3659 return retval;
3660 } 3660 }
3661 3661
3662 static void ext4_resetent(handle_t *handle, s 3662 static void ext4_resetent(handle_t *handle, struct ext4_renament *ent,
3663 unsigned ino, unsig 3663 unsigned ino, unsigned file_type)
3664 { 3664 {
3665 struct ext4_renament old = *ent; 3665 struct ext4_renament old = *ent;
3666 int retval = 0; 3666 int retval = 0;
3667 3667
3668 /* 3668 /*
3669 * old->de could have moved from unde 3669 * old->de could have moved from under us during make indexed dir,
3670 * so the old->de may no longer valid 3670 * so the old->de may no longer valid and need to find it again
3671 * before reset old inode info. 3671 * before reset old inode info.
3672 */ 3672 */
3673 old.bh = ext4_find_entry(old.dir, &ol 3673 old.bh = ext4_find_entry(old.dir, &old.dentry->d_name, &old.de,
3674 &old.inlined 3674 &old.inlined);
3675 if (IS_ERR(old.bh)) 3675 if (IS_ERR(old.bh))
3676 retval = PTR_ERR(old.bh); 3676 retval = PTR_ERR(old.bh);
3677 if (!old.bh) 3677 if (!old.bh)
3678 retval = -ENOENT; 3678 retval = -ENOENT;
3679 if (retval) { 3679 if (retval) {
3680 ext4_std_error(old.dir->i_sb, 3680 ext4_std_error(old.dir->i_sb, retval);
3681 return; 3681 return;
3682 } 3682 }
3683 3683
3684 ext4_setent(handle, &old, ino, file_t 3684 ext4_setent(handle, &old, ino, file_type);
3685 brelse(old.bh); 3685 brelse(old.bh);
3686 } 3686 }
3687 3687
3688 static int ext4_find_delete_entry(handle_t *h 3688 static int ext4_find_delete_entry(handle_t *handle, struct inode *dir,
3689 const struc 3689 const struct qstr *d_name)
3690 { 3690 {
3691 int retval = -ENOENT; 3691 int retval = -ENOENT;
3692 struct buffer_head *bh; 3692 struct buffer_head *bh;
3693 struct ext4_dir_entry_2 *de; 3693 struct ext4_dir_entry_2 *de;
3694 3694
3695 bh = ext4_find_entry(dir, d_name, &de 3695 bh = ext4_find_entry(dir, d_name, &de, NULL);
3696 if (IS_ERR(bh)) 3696 if (IS_ERR(bh))
3697 return PTR_ERR(bh); 3697 return PTR_ERR(bh);
3698 if (bh) { 3698 if (bh) {
3699 retval = ext4_delete_entry(ha 3699 retval = ext4_delete_entry(handle, dir, de, bh);
3700 brelse(bh); 3700 brelse(bh);
3701 } 3701 }
3702 return retval; 3702 return retval;
3703 } 3703 }
3704 3704
3705 static void ext4_rename_delete(handle_t *hand 3705 static void ext4_rename_delete(handle_t *handle, struct ext4_renament *ent,
3706 int force_rere 3706 int force_reread)
3707 { 3707 {
3708 int retval; 3708 int retval;
3709 /* 3709 /*
3710 * ent->de could have moved from unde 3710 * ent->de could have moved from under us during htree split, so make
3711 * sure that we are deleting the righ 3711 * sure that we are deleting the right entry. We might also be pointing
3712 * to a stale entry in the unused par 3712 * to a stale entry in the unused part of ent->bh so just checking inum
3713 * and the name isn't enough. 3713 * and the name isn't enough.
3714 */ 3714 */
3715 if (le32_to_cpu(ent->de->inode) != en 3715 if (le32_to_cpu(ent->de->inode) != ent->inode->i_ino ||
3716 ent->de->name_len != ent->dentry- 3716 ent->de->name_len != ent->dentry->d_name.len ||
3717 strncmp(ent->de->name, ent->dentr 3717 strncmp(ent->de->name, ent->dentry->d_name.name,
3718 ent->de->name_len) || 3718 ent->de->name_len) ||
3719 force_reread) { 3719 force_reread) {
3720 retval = ext4_find_delete_ent 3720 retval = ext4_find_delete_entry(handle, ent->dir,
3721 3721 &ent->dentry->d_name);
3722 } else { 3722 } else {
3723 retval = ext4_delete_entry(ha 3723 retval = ext4_delete_entry(handle, ent->dir, ent->de, ent->bh);
3724 if (retval == -ENOENT) { 3724 if (retval == -ENOENT) {
3725 retval = ext4_find_de 3725 retval = ext4_find_delete_entry(handle, ent->dir,
3726 3726 &ent->dentry->d_name);
3727 } 3727 }
3728 } 3728 }
3729 3729
3730 if (retval) { 3730 if (retval) {
3731 ext4_warning_inode(ent->dir, 3731 ext4_warning_inode(ent->dir,
3732 "Deleting 3732 "Deleting old file: nlink %d, error=%d",
3733 ent->dir-> 3733 ent->dir->i_nlink, retval);
3734 } 3734 }
3735 } 3735 }
3736 3736
3737 static void ext4_update_dir_count(handle_t *h 3737 static void ext4_update_dir_count(handle_t *handle, struct ext4_renament *ent)
3738 { 3738 {
3739 if (ent->dir_nlink_delta) { 3739 if (ent->dir_nlink_delta) {
3740 if (ent->dir_nlink_delta == - 3740 if (ent->dir_nlink_delta == -1)
3741 ext4_dec_count(ent->d 3741 ext4_dec_count(ent->dir);
3742 else 3742 else
3743 ext4_inc_count(ent->d 3743 ext4_inc_count(ent->dir);
3744 ext4_mark_inode_dirty(handle, 3744 ext4_mark_inode_dirty(handle, ent->dir);
3745 } 3745 }
3746 } 3746 }
3747 3747
3748 static struct inode *ext4_whiteout_for_rename 3748 static struct inode *ext4_whiteout_for_rename(struct mnt_idmap *idmap,
3749 3749 struct ext4_renament *ent,
3750 3750 int credits, handle_t **h)
3751 { 3751 {
3752 struct inode *wh; 3752 struct inode *wh;
3753 handle_t *handle; 3753 handle_t *handle;
3754 int retries = 0; 3754 int retries = 0;
3755 3755
3756 /* 3756 /*
3757 * for inode block, sb block, group s 3757 * for inode block, sb block, group summaries,
3758 * and inode bitmap 3758 * and inode bitmap
3759 */ 3759 */
3760 credits += (EXT4_MAXQUOTAS_TRANS_BLOC 3760 credits += (EXT4_MAXQUOTAS_TRANS_BLOCKS(ent->dir->i_sb) +
3761 EXT4_XATTR_TRANS_BLOCKS + 3761 EXT4_XATTR_TRANS_BLOCKS + 4);
3762 retry: 3762 retry:
3763 wh = ext4_new_inode_start_handle(idma 3763 wh = ext4_new_inode_start_handle(idmap, ent->dir,
3764 S_IF 3764 S_IFCHR | WHITEOUT_MODE,
3765 &ent 3765 &ent->dentry->d_name, 0, NULL,
3766 EXT4 3766 EXT4_HT_DIR, credits);
3767 3767
3768 handle = ext4_journal_current_handle( 3768 handle = ext4_journal_current_handle();
3769 if (IS_ERR(wh)) { 3769 if (IS_ERR(wh)) {
3770 if (handle) 3770 if (handle)
3771 ext4_journal_stop(han 3771 ext4_journal_stop(handle);
3772 if (PTR_ERR(wh) == -ENOSPC && 3772 if (PTR_ERR(wh) == -ENOSPC &&
3773 ext4_should_retry_alloc(e 3773 ext4_should_retry_alloc(ent->dir->i_sb, &retries))
3774 goto retry; 3774 goto retry;
3775 } else { 3775 } else {
3776 *h = handle; 3776 *h = handle;
3777 init_special_inode(wh, wh->i_ 3777 init_special_inode(wh, wh->i_mode, WHITEOUT_DEV);
3778 wh->i_op = &ext4_special_inod 3778 wh->i_op = &ext4_special_inode_operations;
3779 } 3779 }
3780 return wh; 3780 return wh;
3781 } 3781 }
3782 3782
3783 /* 3783 /*
3784 * Anybody can rename anything with this: the 3784 * Anybody can rename anything with this: the permission checks are left to the
3785 * higher-level routines. 3785 * higher-level routines.
3786 * 3786 *
3787 * n.b. old_{dentry,inode) refers to the sou 3787 * n.b. old_{dentry,inode) refers to the source dentry/inode
3788 * while new_{dentry,inode) refers to the des 3788 * while new_{dentry,inode) refers to the destination dentry/inode
3789 * This comes from rename(const char *oldpath 3789 * This comes from rename(const char *oldpath, const char *newpath)
3790 */ 3790 */
3791 static int ext4_rename(struct mnt_idmap *idma 3791 static int ext4_rename(struct mnt_idmap *idmap, struct inode *old_dir,
3792 struct dentry *old_den 3792 struct dentry *old_dentry, struct inode *new_dir,
3793 struct dentry *new_den 3793 struct dentry *new_dentry, unsigned int flags)
3794 { 3794 {
3795 handle_t *handle = NULL; 3795 handle_t *handle = NULL;
3796 struct ext4_renament old = { 3796 struct ext4_renament old = {
3797 .dir = old_dir, 3797 .dir = old_dir,
3798 .dentry = old_dentry, 3798 .dentry = old_dentry,
3799 .inode = d_inode(old_dentry), 3799 .inode = d_inode(old_dentry),
3800 }; 3800 };
3801 struct ext4_renament new = { 3801 struct ext4_renament new = {
3802 .dir = new_dir, 3802 .dir = new_dir,
3803 .dentry = new_dentry, 3803 .dentry = new_dentry,
3804 .inode = d_inode(new_dentry), 3804 .inode = d_inode(new_dentry),
3805 }; 3805 };
3806 int force_reread; 3806 int force_reread;
3807 int retval; 3807 int retval;
3808 struct inode *whiteout = NULL; 3808 struct inode *whiteout = NULL;
3809 int credits; 3809 int credits;
3810 u8 old_file_type; 3810 u8 old_file_type;
3811 3811
3812 if (new.inode && new.inode->i_nlink = 3812 if (new.inode && new.inode->i_nlink == 0) {
3813 EXT4_ERROR_INODE(new.inode, 3813 EXT4_ERROR_INODE(new.inode,
3814 "target of r 3814 "target of rename is already freed");
3815 return -EFSCORRUPTED; 3815 return -EFSCORRUPTED;
3816 } 3816 }
3817 3817
3818 if ((ext4_test_inode_flag(new_dir, EX 3818 if ((ext4_test_inode_flag(new_dir, EXT4_INODE_PROJINHERIT)) &&
3819 (!projid_eq(EXT4_I(new_dir)->i_pr 3819 (!projid_eq(EXT4_I(new_dir)->i_projid,
3820 EXT4_I(old_dentry->d_ 3820 EXT4_I(old_dentry->d_inode)->i_projid)))
3821 return -EXDEV; 3821 return -EXDEV;
3822 3822
3823 retval = dquot_initialize(old.dir); 3823 retval = dquot_initialize(old.dir);
3824 if (retval) 3824 if (retval)
3825 return retval; 3825 return retval;
3826 retval = dquot_initialize(old.inode); 3826 retval = dquot_initialize(old.inode);
3827 if (retval) 3827 if (retval)
3828 return retval; 3828 return retval;
3829 retval = dquot_initialize(new.dir); 3829 retval = dquot_initialize(new.dir);
3830 if (retval) 3830 if (retval)
3831 return retval; 3831 return retval;
3832 3832
3833 /* Initialize quotas before so that e 3833 /* Initialize quotas before so that eventual writes go
3834 * in separate transaction */ 3834 * in separate transaction */
3835 if (new.inode) { 3835 if (new.inode) {
3836 retval = dquot_initialize(new 3836 retval = dquot_initialize(new.inode);
3837 if (retval) 3837 if (retval)
3838 return retval; 3838 return retval;
3839 } 3839 }
3840 3840
3841 old.bh = ext4_find_entry(old.dir, &ol 3841 old.bh = ext4_find_entry(old.dir, &old.dentry->d_name, &old.de,
3842 &old.inlined 3842 &old.inlined);
3843 if (IS_ERR(old.bh)) 3843 if (IS_ERR(old.bh))
3844 return PTR_ERR(old.bh); 3844 return PTR_ERR(old.bh);
3845 3845
3846 /* 3846 /*
3847 * Check for inode number is _not_ d 3847 * Check for inode number is _not_ due to possible IO errors.
3848 * We might rmdir the source, keep i 3848 * We might rmdir the source, keep it as pwd of some process
3849 * and merrily kill the link to what 3849 * and merrily kill the link to whatever was created under the
3850 * same name. Goodbye sticky bit ;-< 3850 * same name. Goodbye sticky bit ;-<
3851 */ 3851 */
3852 retval = -ENOENT; 3852 retval = -ENOENT;
3853 if (!old.bh || le32_to_cpu(old.de->in 3853 if (!old.bh || le32_to_cpu(old.de->inode) != old.inode->i_ino)
3854 goto release_bh; 3854 goto release_bh;
3855 3855
3856 new.bh = ext4_find_entry(new.dir, &ne 3856 new.bh = ext4_find_entry(new.dir, &new.dentry->d_name,
3857 &new.de, &ne 3857 &new.de, &new.inlined);
3858 if (IS_ERR(new.bh)) { 3858 if (IS_ERR(new.bh)) {
3859 retval = PTR_ERR(new.bh); 3859 retval = PTR_ERR(new.bh);
3860 new.bh = NULL; 3860 new.bh = NULL;
3861 goto release_bh; 3861 goto release_bh;
3862 } 3862 }
3863 if (new.bh) { 3863 if (new.bh) {
3864 if (!new.inode) { 3864 if (!new.inode) {
3865 brelse(new.bh); 3865 brelse(new.bh);
3866 new.bh = NULL; 3866 new.bh = NULL;
3867 } 3867 }
3868 } 3868 }
3869 if (new.inode && !test_opt(new.dir->i 3869 if (new.inode && !test_opt(new.dir->i_sb, NO_AUTO_DA_ALLOC))
3870 ext4_alloc_da_blocks(old.inod 3870 ext4_alloc_da_blocks(old.inode);
3871 3871
3872 credits = (2 * EXT4_DATA_TRANS_BLOCKS 3872 credits = (2 * EXT4_DATA_TRANS_BLOCKS(old.dir->i_sb) +
3873 EXT4_INDEX_EXTRA_TRANS_BLO 3873 EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2);
3874 if (!(flags & RENAME_WHITEOUT)) { 3874 if (!(flags & RENAME_WHITEOUT)) {
3875 handle = ext4_journal_start(o 3875 handle = ext4_journal_start(old.dir, EXT4_HT_DIR, credits);
3876 if (IS_ERR(handle)) { 3876 if (IS_ERR(handle)) {
3877 retval = PTR_ERR(hand 3877 retval = PTR_ERR(handle);
3878 goto release_bh; 3878 goto release_bh;
3879 } 3879 }
3880 } else { 3880 } else {
3881 whiteout = ext4_whiteout_for_ 3881 whiteout = ext4_whiteout_for_rename(idmap, &old, credits, &handle);
3882 if (IS_ERR(whiteout)) { 3882 if (IS_ERR(whiteout)) {
3883 retval = PTR_ERR(whit 3883 retval = PTR_ERR(whiteout);
3884 goto release_bh; 3884 goto release_bh;
3885 } 3885 }
3886 } 3886 }
3887 3887
3888 old_file_type = old.de->file_type; 3888 old_file_type = old.de->file_type;
3889 if (IS_DIRSYNC(old.dir) || IS_DIRSYNC 3889 if (IS_DIRSYNC(old.dir) || IS_DIRSYNC(new.dir))
3890 ext4_handle_sync(handle); 3890 ext4_handle_sync(handle);
3891 3891
3892 if (S_ISDIR(old.inode->i_mode)) { 3892 if (S_ISDIR(old.inode->i_mode)) {
3893 if (new.inode) { 3893 if (new.inode) {
3894 retval = -ENOTEMPTY; 3894 retval = -ENOTEMPTY;
3895 if (!ext4_empty_dir(n 3895 if (!ext4_empty_dir(new.inode))
3896 goto end_rena 3896 goto end_rename;
3897 } else { 3897 } else {
3898 retval = -EMLINK; 3898 retval = -EMLINK;
3899 if (new.dir != old.di 3899 if (new.dir != old.dir && EXT4_DIR_LINK_MAX(new.dir))
3900 goto end_rena 3900 goto end_rename;
3901 } 3901 }
3902 retval = ext4_rename_dir_prep 3902 retval = ext4_rename_dir_prepare(handle, &old, new.dir != old.dir);
3903 if (retval) 3903 if (retval)
3904 goto end_rename; 3904 goto end_rename;
3905 } 3905 }
3906 /* 3906 /*
3907 * If we're renaming a file within an 3907 * If we're renaming a file within an inline_data dir and adding or
3908 * setting the new dirent causes a co 3908 * setting the new dirent causes a conversion from inline_data to
3909 * extents/blockmap, we need to force 3909 * extents/blockmap, we need to force the dirent delete code to
3910 * re-read the directory, or else we 3910 * re-read the directory, or else we end up trying to delete a dirent
3911 * from what is now the extent tree r 3911 * from what is now the extent tree root (or a block map).
3912 */ 3912 */
3913 force_reread = (new.dir->i_ino == old 3913 force_reread = (new.dir->i_ino == old.dir->i_ino &&
3914 ext4_test_inode_flag( 3914 ext4_test_inode_flag(new.dir, EXT4_INODE_INLINE_DATA));
3915 3915
3916 if (whiteout) { 3916 if (whiteout) {
3917 /* 3917 /*
3918 * Do this before adding a ne 3918 * Do this before adding a new entry, so the old entry is sure
3919 * to be still pointing to th 3919 * to be still pointing to the valid old entry.
3920 */ 3920 */
3921 retval = ext4_setent(handle, 3921 retval = ext4_setent(handle, &old, whiteout->i_ino,
3922 EXT4_FT_ 3922 EXT4_FT_CHRDEV);
3923 if (retval) 3923 if (retval)
3924 goto end_rename; 3924 goto end_rename;
3925 retval = ext4_mark_inode_dirt 3925 retval = ext4_mark_inode_dirty(handle, whiteout);
3926 if (unlikely(retval)) 3926 if (unlikely(retval))
3927 goto end_rename; 3927 goto end_rename;
3928 3928
3929 } 3929 }
3930 if (!new.bh) { 3930 if (!new.bh) {
3931 retval = ext4_add_entry(handl 3931 retval = ext4_add_entry(handle, new.dentry, old.inode);
3932 if (retval) 3932 if (retval)
3933 goto end_rename; 3933 goto end_rename;
3934 } else { 3934 } else {
3935 retval = ext4_setent(handle, 3935 retval = ext4_setent(handle, &new,
3936 old.inod 3936 old.inode->i_ino, old_file_type);
3937 if (retval) 3937 if (retval)
3938 goto end_rename; 3938 goto end_rename;
3939 } 3939 }
3940 if (force_reread) 3940 if (force_reread)
3941 force_reread = !ext4_test_ino 3941 force_reread = !ext4_test_inode_flag(new.dir,
3942 3942 EXT4_INODE_INLINE_DATA);
3943 3943
3944 /* 3944 /*
3945 * Like most other Unix systems, set 3945 * Like most other Unix systems, set the ctime for inodes on a
3946 * rename. 3946 * rename.
3947 */ 3947 */
3948 inode_set_ctime_current(old.inode); 3948 inode_set_ctime_current(old.inode);
3949 retval = ext4_mark_inode_dirty(handle 3949 retval = ext4_mark_inode_dirty(handle, old.inode);
3950 if (unlikely(retval)) 3950 if (unlikely(retval))
3951 goto end_rename; 3951 goto end_rename;
3952 3952
3953 if (!whiteout) { 3953 if (!whiteout) {
3954 /* 3954 /*
3955 * ok, that's it 3955 * ok, that's it
3956 */ 3956 */
3957 ext4_rename_delete(handle, &o 3957 ext4_rename_delete(handle, &old, force_reread);
3958 } 3958 }
3959 3959
3960 if (new.inode) { 3960 if (new.inode) {
3961 ext4_dec_count(new.inode); 3961 ext4_dec_count(new.inode);
3962 inode_set_ctime_current(new.i 3962 inode_set_ctime_current(new.inode);
3963 } 3963 }
3964 inode_set_mtime_to_ts(old.dir, inode_ 3964 inode_set_mtime_to_ts(old.dir, inode_set_ctime_current(old.dir));
3965 ext4_update_dx_flag(old.dir); 3965 ext4_update_dx_flag(old.dir);
3966 if (old.is_dir) { 3966 if (old.is_dir) {
3967 retval = ext4_rename_dir_fini 3967 retval = ext4_rename_dir_finish(handle, &old, new.dir->i_ino);
3968 if (retval) 3968 if (retval)
3969 goto end_rename; 3969 goto end_rename;
3970 3970
3971 ext4_dec_count(old.dir); 3971 ext4_dec_count(old.dir);
3972 if (new.inode) { 3972 if (new.inode) {
3973 /* checked ext4_empty 3973 /* checked ext4_empty_dir above, can't have another
3974 * parent, ext4_dec_c 3974 * parent, ext4_dec_count() won't work for many-linked
3975 * dirs */ 3975 * dirs */
3976 clear_nlink(new.inode 3976 clear_nlink(new.inode);
3977 } else { 3977 } else {
3978 ext4_inc_count(new.di 3978 ext4_inc_count(new.dir);
3979 ext4_update_dx_flag(n 3979 ext4_update_dx_flag(new.dir);
3980 retval = ext4_mark_in 3980 retval = ext4_mark_inode_dirty(handle, new.dir);
3981 if (unlikely(retval)) 3981 if (unlikely(retval))
3982 goto end_rena 3982 goto end_rename;
3983 } 3983 }
3984 } 3984 }
3985 retval = ext4_mark_inode_dirty(handle 3985 retval = ext4_mark_inode_dirty(handle, old.dir);
3986 if (unlikely(retval)) 3986 if (unlikely(retval))
3987 goto end_rename; 3987 goto end_rename;
3988 3988
3989 if (old.is_dir) { 3989 if (old.is_dir) {
3990 /* 3990 /*
3991 * We disable fast commits he 3991 * We disable fast commits here that's because the
3992 * replay code is not yet cap 3992 * replay code is not yet capable of changing dot dot
3993 * dirents in directories. 3993 * dirents in directories.
3994 */ 3994 */
3995 ext4_fc_mark_ineligible(old.i 3995 ext4_fc_mark_ineligible(old.inode->i_sb,
3996 EXT4_FC_REASON_RENAME 3996 EXT4_FC_REASON_RENAME_DIR, handle);
3997 } else { 3997 } else {
3998 struct super_block *sb = old. 3998 struct super_block *sb = old.inode->i_sb;
3999 3999
4000 if (new.inode) 4000 if (new.inode)
4001 ext4_fc_track_unlink( 4001 ext4_fc_track_unlink(handle, new.dentry);
4002 if (test_opt2(sb, JOURNAL_FAS 4002 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4003 !(EXT4_SB(sb)->s_mount_st 4003 !(EXT4_SB(sb)->s_mount_state & EXT4_FC_REPLAY) &&
4004 !(ext4_test_mount_flag(sb 4004 !(ext4_test_mount_flag(sb, EXT4_MF_FC_INELIGIBLE))) {
4005 __ext4_fc_track_link( 4005 __ext4_fc_track_link(handle, old.inode, new.dentry);
4006 __ext4_fc_track_unlin 4006 __ext4_fc_track_unlink(handle, old.inode, old.dentry);
4007 if (whiteout) 4007 if (whiteout)
4008 __ext4_fc_tra 4008 __ext4_fc_track_create(handle, whiteout,
4009 4009 old.dentry);
4010 } 4010 }
4011 } 4011 }
4012 4012
4013 if (new.inode) { 4013 if (new.inode) {
4014 retval = ext4_mark_inode_dirt 4014 retval = ext4_mark_inode_dirty(handle, new.inode);
4015 if (unlikely(retval)) 4015 if (unlikely(retval))
4016 goto end_rename; 4016 goto end_rename;
4017 if (!new.inode->i_nlink) 4017 if (!new.inode->i_nlink)
4018 ext4_orphan_add(handl 4018 ext4_orphan_add(handle, new.inode);
4019 } 4019 }
4020 retval = 0; 4020 retval = 0;
4021 4021
4022 end_rename: 4022 end_rename:
4023 if (whiteout) { 4023 if (whiteout) {
4024 if (retval) { 4024 if (retval) {
4025 ext4_resetent(handle, 4025 ext4_resetent(handle, &old,
4026 old.ino 4026 old.inode->i_ino, old_file_type);
4027 drop_nlink(whiteout); 4027 drop_nlink(whiteout);
4028 ext4_mark_inode_dirty 4028 ext4_mark_inode_dirty(handle, whiteout);
4029 ext4_orphan_add(handl 4029 ext4_orphan_add(handle, whiteout);
4030 } 4030 }
4031 unlock_new_inode(whiteout); 4031 unlock_new_inode(whiteout);
4032 ext4_journal_stop(handle); 4032 ext4_journal_stop(handle);
4033 iput(whiteout); 4033 iput(whiteout);
4034 } else { 4034 } else {
4035 ext4_journal_stop(handle); 4035 ext4_journal_stop(handle);
4036 } 4036 }
4037 release_bh: 4037 release_bh:
4038 brelse(old.dir_bh); 4038 brelse(old.dir_bh);
4039 brelse(old.bh); 4039 brelse(old.bh);
4040 brelse(new.bh); 4040 brelse(new.bh);
4041 4041
4042 return retval; 4042 return retval;
4043 } 4043 }
4044 4044
4045 static int ext4_cross_rename(struct inode *ol 4045 static int ext4_cross_rename(struct inode *old_dir, struct dentry *old_dentry,
4046 struct inode *ne 4046 struct inode *new_dir, struct dentry *new_dentry)
4047 { 4047 {
4048 handle_t *handle = NULL; 4048 handle_t *handle = NULL;
4049 struct ext4_renament old = { 4049 struct ext4_renament old = {
4050 .dir = old_dir, 4050 .dir = old_dir,
4051 .dentry = old_dentry, 4051 .dentry = old_dentry,
4052 .inode = d_inode(old_dentry), 4052 .inode = d_inode(old_dentry),
4053 }; 4053 };
4054 struct ext4_renament new = { 4054 struct ext4_renament new = {
4055 .dir = new_dir, 4055 .dir = new_dir,
4056 .dentry = new_dentry, 4056 .dentry = new_dentry,
4057 .inode = d_inode(new_dentry), 4057 .inode = d_inode(new_dentry),
4058 }; 4058 };
4059 u8 new_file_type; 4059 u8 new_file_type;
4060 int retval; 4060 int retval;
4061 4061
4062 if ((ext4_test_inode_flag(new_dir, EX 4062 if ((ext4_test_inode_flag(new_dir, EXT4_INODE_PROJINHERIT) &&
4063 !projid_eq(EXT4_I(new_dir)->i_pr 4063 !projid_eq(EXT4_I(new_dir)->i_projid,
4064 EXT4_I(old_dentry->d_ 4064 EXT4_I(old_dentry->d_inode)->i_projid)) ||
4065 (ext4_test_inode_flag(old_dir, EX 4065 (ext4_test_inode_flag(old_dir, EXT4_INODE_PROJINHERIT) &&
4066 !projid_eq(EXT4_I(old_dir)->i_pr 4066 !projid_eq(EXT4_I(old_dir)->i_projid,
4067 EXT4_I(new_dentry->d_ 4067 EXT4_I(new_dentry->d_inode)->i_projid)))
4068 return -EXDEV; 4068 return -EXDEV;
4069 4069
4070 retval = dquot_initialize(old.dir); 4070 retval = dquot_initialize(old.dir);
4071 if (retval) 4071 if (retval)
4072 return retval; 4072 return retval;
4073 retval = dquot_initialize(new.dir); 4073 retval = dquot_initialize(new.dir);
4074 if (retval) 4074 if (retval)
4075 return retval; 4075 return retval;
4076 4076
4077 old.bh = ext4_find_entry(old.dir, &ol 4077 old.bh = ext4_find_entry(old.dir, &old.dentry->d_name,
4078 &old.de, &ol 4078 &old.de, &old.inlined);
4079 if (IS_ERR(old.bh)) 4079 if (IS_ERR(old.bh))
4080 return PTR_ERR(old.bh); 4080 return PTR_ERR(old.bh);
4081 /* 4081 /*
4082 * Check for inode number is _not_ d 4082 * Check for inode number is _not_ due to possible IO errors.
4083 * We might rmdir the source, keep i 4083 * We might rmdir the source, keep it as pwd of some process
4084 * and merrily kill the link to what 4084 * and merrily kill the link to whatever was created under the
4085 * same name. Goodbye sticky bit ;-< 4085 * same name. Goodbye sticky bit ;-<
4086 */ 4086 */
4087 retval = -ENOENT; 4087 retval = -ENOENT;
4088 if (!old.bh || le32_to_cpu(old.de->in 4088 if (!old.bh || le32_to_cpu(old.de->inode) != old.inode->i_ino)
4089 goto end_rename; 4089 goto end_rename;
4090 4090
4091 new.bh = ext4_find_entry(new.dir, &ne 4091 new.bh = ext4_find_entry(new.dir, &new.dentry->d_name,
4092 &new.de, &ne 4092 &new.de, &new.inlined);
4093 if (IS_ERR(new.bh)) { 4093 if (IS_ERR(new.bh)) {
4094 retval = PTR_ERR(new.bh); 4094 retval = PTR_ERR(new.bh);
4095 new.bh = NULL; 4095 new.bh = NULL;
4096 goto end_rename; 4096 goto end_rename;
4097 } 4097 }
4098 4098
4099 /* RENAME_EXCHANGE case: old *and* ne 4099 /* RENAME_EXCHANGE case: old *and* new must both exist */
4100 if (!new.bh || le32_to_cpu(new.de->in 4100 if (!new.bh || le32_to_cpu(new.de->inode) != new.inode->i_ino)
4101 goto end_rename; 4101 goto end_rename;
4102 4102
4103 handle = ext4_journal_start(old.dir, 4103 handle = ext4_journal_start(old.dir, EXT4_HT_DIR,
4104 (2 * EXT4_DATA_TRANS_BLOCKS(o 4104 (2 * EXT4_DATA_TRANS_BLOCKS(old.dir->i_sb) +
4105 2 * EXT4_INDEX_EXTRA_TRANS_B 4105 2 * EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2));
4106 if (IS_ERR(handle)) { 4106 if (IS_ERR(handle)) {
4107 retval = PTR_ERR(handle); 4107 retval = PTR_ERR(handle);
4108 handle = NULL; 4108 handle = NULL;
4109 goto end_rename; 4109 goto end_rename;
4110 } 4110 }
4111 4111
4112 if (IS_DIRSYNC(old.dir) || IS_DIRSYNC 4112 if (IS_DIRSYNC(old.dir) || IS_DIRSYNC(new.dir))
4113 ext4_handle_sync(handle); 4113 ext4_handle_sync(handle);
4114 4114
4115 if (S_ISDIR(old.inode->i_mode)) { 4115 if (S_ISDIR(old.inode->i_mode)) {
4116 retval = ext4_rename_dir_prep 4116 retval = ext4_rename_dir_prepare(handle, &old, new.dir != old.dir);
4117 if (retval) 4117 if (retval)
4118 goto end_rename; 4118 goto end_rename;
4119 } 4119 }
4120 if (S_ISDIR(new.inode->i_mode)) { 4120 if (S_ISDIR(new.inode->i_mode)) {
4121 retval = ext4_rename_dir_prep 4121 retval = ext4_rename_dir_prepare(handle, &new, new.dir != old.dir);
4122 if (retval) 4122 if (retval)
4123 goto end_rename; 4123 goto end_rename;
4124 } 4124 }
4125 4125
4126 /* 4126 /*
4127 * Other than the special case of ove 4127 * Other than the special case of overwriting a directory, parents'
4128 * nlink only needs to be modified if 4128 * nlink only needs to be modified if this is a cross directory rename.
4129 */ 4129 */
4130 if (old.dir != new.dir && old.is_dir 4130 if (old.dir != new.dir && old.is_dir != new.is_dir) {
4131 old.dir_nlink_delta = old.is_ 4131 old.dir_nlink_delta = old.is_dir ? -1 : 1;
4132 new.dir_nlink_delta = -old.di 4132 new.dir_nlink_delta = -old.dir_nlink_delta;
4133 retval = -EMLINK; 4133 retval = -EMLINK;
4134 if ((old.dir_nlink_delta > 0 4134 if ((old.dir_nlink_delta > 0 && EXT4_DIR_LINK_MAX(old.dir)) ||
4135 (new.dir_nlink_delta > 0 4135 (new.dir_nlink_delta > 0 && EXT4_DIR_LINK_MAX(new.dir)))
4136 goto end_rename; 4136 goto end_rename;
4137 } 4137 }
4138 4138
4139 new_file_type = new.de->file_type; 4139 new_file_type = new.de->file_type;
4140 retval = ext4_setent(handle, &new, ol 4140 retval = ext4_setent(handle, &new, old.inode->i_ino, old.de->file_type);
4141 if (retval) 4141 if (retval)
4142 goto end_rename; 4142 goto end_rename;
4143 4143
4144 retval = ext4_setent(handle, &old, ne 4144 retval = ext4_setent(handle, &old, new.inode->i_ino, new_file_type);
4145 if (retval) 4145 if (retval)
4146 goto end_rename; 4146 goto end_rename;
4147 4147
4148 /* 4148 /*
4149 * Like most other Unix systems, set 4149 * Like most other Unix systems, set the ctime for inodes on a
4150 * rename. 4150 * rename.
4151 */ 4151 */
4152 inode_set_ctime_current(old.inode); 4152 inode_set_ctime_current(old.inode);
4153 inode_set_ctime_current(new.inode); 4153 inode_set_ctime_current(new.inode);
4154 retval = ext4_mark_inode_dirty(handle 4154 retval = ext4_mark_inode_dirty(handle, old.inode);
4155 if (unlikely(retval)) 4155 if (unlikely(retval))
4156 goto end_rename; 4156 goto end_rename;
4157 retval = ext4_mark_inode_dirty(handle 4157 retval = ext4_mark_inode_dirty(handle, new.inode);
4158 if (unlikely(retval)) 4158 if (unlikely(retval))
4159 goto end_rename; 4159 goto end_rename;
4160 ext4_fc_mark_ineligible(new.inode->i_ 4160 ext4_fc_mark_ineligible(new.inode->i_sb,
4161 EXT4_FC_REASO 4161 EXT4_FC_REASON_CROSS_RENAME, handle);
4162 if (old.dir_bh) { 4162 if (old.dir_bh) {
4163 retval = ext4_rename_dir_fini 4163 retval = ext4_rename_dir_finish(handle, &old, new.dir->i_ino);
4164 if (retval) 4164 if (retval)
4165 goto end_rename; 4165 goto end_rename;
4166 } 4166 }
4167 if (new.dir_bh) { 4167 if (new.dir_bh) {
4168 retval = ext4_rename_dir_fini 4168 retval = ext4_rename_dir_finish(handle, &new, old.dir->i_ino);
4169 if (retval) 4169 if (retval)
4170 goto end_rename; 4170 goto end_rename;
4171 } 4171 }
4172 ext4_update_dir_count(handle, &old); 4172 ext4_update_dir_count(handle, &old);
4173 ext4_update_dir_count(handle, &new); 4173 ext4_update_dir_count(handle, &new);
4174 retval = 0; 4174 retval = 0;
4175 4175
4176 end_rename: 4176 end_rename:
4177 brelse(old.dir_bh); 4177 brelse(old.dir_bh);
4178 brelse(new.dir_bh); 4178 brelse(new.dir_bh);
4179 brelse(old.bh); 4179 brelse(old.bh);
4180 brelse(new.bh); 4180 brelse(new.bh);
4181 if (handle) 4181 if (handle)
4182 ext4_journal_stop(handle); 4182 ext4_journal_stop(handle);
4183 return retval; 4183 return retval;
4184 } 4184 }
4185 4185
4186 static int ext4_rename2(struct mnt_idmap *idm 4186 static int ext4_rename2(struct mnt_idmap *idmap,
4187 struct inode *old_dir 4187 struct inode *old_dir, struct dentry *old_dentry,
4188 struct inode *new_dir 4188 struct inode *new_dir, struct dentry *new_dentry,
4189 unsigned int flags) 4189 unsigned int flags)
4190 { 4190 {
4191 int err; 4191 int err;
4192 4192
4193 if (unlikely(ext4_forced_shutdown(old 4193 if (unlikely(ext4_forced_shutdown(old_dir->i_sb)))
4194 return -EIO; 4194 return -EIO;
4195 4195
4196 if (flags & ~(RENAME_NOREPLACE | RENA 4196 if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT))
4197 return -EINVAL; 4197 return -EINVAL;
4198 4198
4199 err = fscrypt_prepare_rename(old_dir, 4199 err = fscrypt_prepare_rename(old_dir, old_dentry, new_dir, new_dentry,
4200 flags); 4200 flags);
4201 if (err) 4201 if (err)
4202 return err; 4202 return err;
4203 4203
4204 if (flags & RENAME_EXCHANGE) { 4204 if (flags & RENAME_EXCHANGE) {
4205 return ext4_cross_rename(old_ 4205 return ext4_cross_rename(old_dir, old_dentry,
4206 new_ 4206 new_dir, new_dentry);
4207 } 4207 }
4208 4208
4209 return ext4_rename(idmap, old_dir, ol 4209 return ext4_rename(idmap, old_dir, old_dentry, new_dir, new_dentry, flags);
4210 } 4210 }
4211 4211
4212 /* 4212 /*
4213 * directories can handle most operations... 4213 * directories can handle most operations...
4214 */ 4214 */
4215 const struct inode_operations ext4_dir_inode_ 4215 const struct inode_operations ext4_dir_inode_operations = {
4216 .create = ext4_create, 4216 .create = ext4_create,
4217 .lookup = ext4_lookup, 4217 .lookup = ext4_lookup,
4218 .link = ext4_link, 4218 .link = ext4_link,
4219 .unlink = ext4_unlink, 4219 .unlink = ext4_unlink,
4220 .symlink = ext4_symlink, 4220 .symlink = ext4_symlink,
4221 .mkdir = ext4_mkdir, 4221 .mkdir = ext4_mkdir,
4222 .rmdir = ext4_rmdir, 4222 .rmdir = ext4_rmdir,
4223 .mknod = ext4_mknod, 4223 .mknod = ext4_mknod,
4224 .tmpfile = ext4_tmpfile, 4224 .tmpfile = ext4_tmpfile,
4225 .rename = ext4_rename2, 4225 .rename = ext4_rename2,
4226 .setattr = ext4_setattr, 4226 .setattr = ext4_setattr,
4227 .getattr = ext4_getattr, 4227 .getattr = ext4_getattr,
4228 .listxattr = ext4_listxattr, 4228 .listxattr = ext4_listxattr,
4229 .get_inode_acl = ext4_get_acl, 4229 .get_inode_acl = ext4_get_acl,
4230 .set_acl = ext4_set_acl, 4230 .set_acl = ext4_set_acl,
4231 .fiemap = ext4_fiemap, 4231 .fiemap = ext4_fiemap,
4232 .fileattr_get = ext4_fileattr_get, 4232 .fileattr_get = ext4_fileattr_get,
4233 .fileattr_set = ext4_fileattr_set, 4233 .fileattr_set = ext4_fileattr_set,
4234 }; 4234 };
4235 4235
4236 const struct inode_operations ext4_special_in 4236 const struct inode_operations ext4_special_inode_operations = {
4237 .setattr = ext4_setattr, 4237 .setattr = ext4_setattr,
4238 .getattr = ext4_getattr, 4238 .getattr = ext4_getattr,
4239 .listxattr = ext4_listxattr, 4239 .listxattr = ext4_listxattr,
4240 .get_inode_acl = ext4_get_acl, 4240 .get_inode_acl = ext4_get_acl,
4241 .set_acl = ext4_set_acl, 4241 .set_acl = ext4_set_acl,
4242 }; 4242 };
4243 4243
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.