blob: 0bb3f9ea08329a46d27f75160a6b42718fed6042 [file] [log] [blame]
Dave Kleikampac27a0e2006-10-11 01:20:50 -07001/*
Mingming Cao617ba132006-10-11 01:20:53 -07002 * linux/fs/ext4/dir.c
Dave Kleikampac27a0e2006-10-11 01:20:50 -07003 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/dir.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
Mingming Cao617ba132006-10-11 01:20:53 -070015 * ext4 directory handling functions
Dave Kleikampac27a0e2006-10-11 01:20:50 -070016 *
17 * Big-endian to little-endian byte-swapping/bitmaps by
18 * David S. Miller (davem@caip.rutgers.edu), 1995
19 *
20 * Hash Tree Directory indexing (c) 2001 Daniel Phillips
21 *
22 */
23
24#include <linux/fs.h>
Mingming Caodab291a2006-10-11 01:21:01 -070025#include <linux/jbd2.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070026#include <linux/buffer_head.h>
Dave Kleikampac27a0e2006-10-11 01:20:50 -070027#include <linux/slab.h>
28#include <linux/rbtree.h>
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040029#include "ext4.h"
Tao Ma65d165d2012-12-10 14:05:59 -050030#include "xattr.h"
Dave Kleikampac27a0e2006-10-11 01:20:50 -070031
Al Viro725bebb2013-05-17 16:08:53 -040032static int ext4_dx_readdir(struct file *, struct dir_context *);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070033
Fan Yongd1f52732012-03-18 22:44:40 -040034/**
35 * Check if the given dir-inode refers to an htree-indexed directory
Anatol Pomozov70261f52013-08-28 14:40:12 -040036 * (or a directory which could potentially get converted to use htree
Fan Yongd1f52732012-03-18 22:44:40 -040037 * indexing).
38 *
39 * Return 1 if it is a dx dir, 0 if not
40 */
41static int is_dx_dir(struct inode *inode)
42{
43 struct super_block *sb = inode->i_sb;
44
45 if (EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
46 EXT4_FEATURE_COMPAT_DIR_INDEX) &&
47 ((ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) ||
Tao Ma8af0f082013-04-19 17:53:09 -040048 ((inode->i_size >> sb->s_blocksize_bits) == 1) ||
49 ext4_has_inline_data(inode)))
Fan Yongd1f52732012-03-18 22:44:40 -040050 return 1;
51
52 return 0;
53}
54
Theodore Ts'ocad3f002010-12-19 22:07:02 -050055/*
56 * Return 0 if the directory entry is OK, and 1 if there is a problem
57 *
58 * Note: this is the opposite of what ext2 and ext3 historically returned...
Tao Ma65d165d2012-12-10 14:05:59 -050059 *
60 * bh passed here can be an inode block or a dir data block, depending
61 * on the inode inline data flag.
Theodore Ts'ocad3f002010-12-19 22:07:02 -050062 */
Theodore Ts'o60fd4da2010-07-27 11:54:40 -040063int __ext4_check_dir_entry(const char *function, unsigned int line,
Theodore Ts'of7c21172011-01-10 12:10:55 -050064 struct inode *dir, struct file *filp,
Theodore Ts'o60fd4da2010-07-27 11:54:40 -040065 struct ext4_dir_entry_2 *de,
Tao Ma226ba972012-12-10 14:05:58 -050066 struct buffer_head *bh, char *buf, int size,
Theodore Ts'o60fd4da2010-07-27 11:54:40 -040067 unsigned int offset)
Dave Kleikampac27a0e2006-10-11 01:20:50 -070068{
Theodore Ts'oaf5bc922008-09-08 22:25:24 -040069 const char *error_msg = NULL;
Wei Yongjun3d0518f2009-02-14 23:01:36 -050070 const int rlen = ext4_rec_len_from_disk(de->rec_len,
71 dir->i_sb->s_blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -070072
Theodore Ts'ocad3f002010-12-19 22:07:02 -050073 if (unlikely(rlen < EXT4_DIR_REC_LEN(1)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -070074 error_msg = "rec_len is smaller than minimal";
Theodore Ts'ocad3f002010-12-19 22:07:02 -050075 else if (unlikely(rlen % 4 != 0))
Dave Kleikampac27a0e2006-10-11 01:20:50 -070076 error_msg = "rec_len % 4 != 0";
Theodore Ts'ocad3f002010-12-19 22:07:02 -050077 else if (unlikely(rlen < EXT4_DIR_REC_LEN(de->name_len)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -070078 error_msg = "rec_len is too small for name_len";
Tao Ma226ba972012-12-10 14:05:58 -050079 else if (unlikely(((char *) de - buf) + rlen > size))
80 error_msg = "directory entry across range";
Theodore Ts'ocad3f002010-12-19 22:07:02 -050081 else if (unlikely(le32_to_cpu(de->inode) >
82 le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count)))
Dave Kleikampac27a0e2006-10-11 01:20:50 -070083 error_msg = "inode out of bounds";
Theodore Ts'ocad3f002010-12-19 22:07:02 -050084 else
85 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -070086
Theodore Ts'of7c21172011-01-10 12:10:55 -050087 if (filp)
Dan Carpenter4fda4002012-02-20 17:53:05 -050088 ext4_error_file(filp, function, line, bh->b_blocknr,
Theodore Ts'of7c21172011-01-10 12:10:55 -050089 "bad entry in directory: %s - offset=%u(%u), "
90 "inode=%u, rec_len=%d, name_len=%d",
Tao Ma226ba972012-12-10 14:05:58 -050091 error_msg, (unsigned) (offset % size),
Theodore Ts'of7c21172011-01-10 12:10:55 -050092 offset, le32_to_cpu(de->inode),
93 rlen, de->name_len);
94 else
Dan Carpenter4fda4002012-02-20 17:53:05 -050095 ext4_error_inode(dir, function, line, bh->b_blocknr,
Theodore Ts'of7c21172011-01-10 12:10:55 -050096 "bad entry in directory: %s - offset=%u(%u), "
97 "inode=%u, rec_len=%d, name_len=%d",
Tao Ma226ba972012-12-10 14:05:58 -050098 error_msg, (unsigned) (offset % size),
Theodore Ts'of7c21172011-01-10 12:10:55 -050099 offset, le32_to_cpu(de->inode),
100 rlen, de->name_len);
101
Theodore Ts'ocad3f002010-12-19 22:07:02 -0500102 return 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700103}
104
Al Viro725bebb2013-05-17 16:08:53 -0400105static int ext4_readdir(struct file *file, struct dir_context *ctx)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700106{
Theodore Ts'o498e5f22008-11-05 00:14:04 -0500107 unsigned int offset;
Giedrius Rekasiusaa13d5f2014-05-27 12:48:55 -0400108 int i;
Mingming Cao617ba132006-10-11 01:20:53 -0700109 struct ext4_dir_entry_2 *de;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700110 int err;
Al Viro725bebb2013-05-17 16:08:53 -0400111 struct inode *inode = file_inode(file);
Fan Yongd1f52732012-03-18 22:44:40 -0400112 struct super_block *sb = inode->i_sb;
Eric Sandeen9d9f1772008-10-09 11:15:52 -0400113 int dir_has_error = 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700114
Fan Yongd1f52732012-03-18 22:44:40 -0400115 if (is_dx_dir(inode)) {
Al Viro725bebb2013-05-17 16:08:53 -0400116 err = ext4_dx_readdir(file, ctx);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700117 if (err != ERR_BAD_DX_DIR) {
Al Viro725bebb2013-05-17 16:08:53 -0400118 return err;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700119 }
120 /*
121 * We don't set the inode dirty flag since it's not
122 * critical that it get flushed back to the disk.
123 */
Al Viro725bebb2013-05-17 16:08:53 -0400124 ext4_clear_inode_flag(file_inode(file),
Theodore Ts'o07a03822010-06-14 09:54:48 -0400125 EXT4_INODE_INDEX);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700126 }
Tao Ma8af0f082013-04-19 17:53:09 -0400127
128 if (ext4_has_inline_data(inode)) {
129 int has_inline_data = 1;
Al Viro725bebb2013-05-17 16:08:53 -0400130 int ret = ext4_read_inline_dir(file, ctx,
Tao Ma8af0f082013-04-19 17:53:09 -0400131 &has_inline_data);
132 if (has_inline_data)
133 return ret;
134 }
135
Al Viro725bebb2013-05-17 16:08:53 -0400136 offset = ctx->pos & (sb->s_blocksize - 1);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700137
Al Viro725bebb2013-05-17 16:08:53 -0400138 while (ctx->pos < inode->i_size) {
Theodore Ts'o2ed88682010-05-16 20:00:00 -0400139 struct ext4_map_blocks map;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700140 struct buffer_head *bh = NULL;
141
Al Viro725bebb2013-05-17 16:08:53 -0400142 map.m_lblk = ctx->pos >> EXT4_BLOCK_SIZE_BITS(sb);
Theodore Ts'o2ed88682010-05-16 20:00:00 -0400143 map.m_len = 1;
144 err = ext4_map_blocks(NULL, inode, &map, 0);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700145 if (err > 0) {
Theodore Ts'o2ed88682010-05-16 20:00:00 -0400146 pgoff_t index = map.m_pblk >>
Fengguang Wudc7868f2007-07-19 01:48:04 -0700147 (PAGE_CACHE_SHIFT - inode->i_blkbits);
Al Viro725bebb2013-05-17 16:08:53 -0400148 if (!ra_has_index(&file->f_ra, index))
Rusty Russellcf914a72007-07-19 01:48:08 -0700149 page_cache_sync_readahead(
Fengguang Wudc7868f2007-07-19 01:48:04 -0700150 sb->s_bdev->bd_inode->i_mapping,
Al Viro725bebb2013-05-17 16:08:53 -0400151 &file->f_ra, file,
Rusty Russellcf914a72007-07-19 01:48:08 -0700152 index, 1);
Al Viro725bebb2013-05-17 16:08:53 -0400153 file->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
Theodore Ts'o2ed88682010-05-16 20:00:00 -0400154 bh = ext4_bread(NULL, inode, map.m_lblk, 0, &err);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700155 }
156
157 /*
158 * We ignore I/O errors on directories so users have a chance
159 * of recovering data when there's a bad sector
160 */
161 if (!bh) {
Eric Sandeen9d9f1772008-10-09 11:15:52 -0400162 if (!dir_has_error) {
Al Viro725bebb2013-05-17 16:08:53 -0400163 EXT4_ERROR_FILE(file, 0,
Theodore Ts'of7c21172011-01-10 12:10:55 -0500164 "directory contains a "
165 "hole at offset %llu",
Al Viro725bebb2013-05-17 16:08:53 -0400166 (unsigned long long) ctx->pos);
Eric Sandeen9d9f1772008-10-09 11:15:52 -0400167 dir_has_error = 1;
168 }
Eric Sandeene6c40212006-12-06 20:36:28 -0800169 /* corrupt size? Maybe no more blocks to read */
Al Viro725bebb2013-05-17 16:08:53 -0400170 if (ctx->pos > inode->i_blocks << 9)
Eric Sandeene6c40212006-12-06 20:36:28 -0800171 break;
Al Viro725bebb2013-05-17 16:08:53 -0400172 ctx->pos += sb->s_blocksize - offset;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700173 continue;
174 }
175
Darrick J. Wongb0336e82012-04-29 18:41:10 -0400176 /* Check the checksum */
177 if (!buffer_verified(bh) &&
178 !ext4_dirent_csum_verify(inode,
179 (struct ext4_dir_entry *)bh->b_data)) {
Al Viro725bebb2013-05-17 16:08:53 -0400180 EXT4_ERROR_FILE(file, 0, "directory fails checksum "
Darrick J. Wongb0336e82012-04-29 18:41:10 -0400181 "at offset %llu",
Al Viro725bebb2013-05-17 16:08:53 -0400182 (unsigned long long)ctx->pos);
183 ctx->pos += sb->s_blocksize - offset;
Guo Chaod5ac7772013-01-28 21:23:24 -0500184 brelse(bh);
Darrick J. Wongb0336e82012-04-29 18:41:10 -0400185 continue;
186 }
187 set_buffer_verified(bh);
188
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700189 /* If the dir block has changed since the last call to
190 * readdir(2), then we might be pointing to an invalid
191 * dirent right now. Scan from the start of the block
192 * to make sure. */
Al Viro725bebb2013-05-17 16:08:53 -0400193 if (file->f_version != inode->i_version) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700194 for (i = 0; i < sb->s_blocksize && i < offset; ) {
Mingming Cao617ba132006-10-11 01:20:53 -0700195 de = (struct ext4_dir_entry_2 *)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700196 (bh->b_data + i);
197 /* It's too expensive to do a full
198 * dirent test each time round this
199 * loop, but we do have to test at
200 * least that it is non-zero. A
201 * failure will be detected in the
202 * dirent test below. */
Wei Yongjun3d0518f2009-02-14 23:01:36 -0500203 if (ext4_rec_len_from_disk(de->rec_len,
204 sb->s_blocksize) < EXT4_DIR_REC_LEN(1))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700205 break;
Wei Yongjun3d0518f2009-02-14 23:01:36 -0500206 i += ext4_rec_len_from_disk(de->rec_len,
207 sb->s_blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700208 }
209 offset = i;
Al Viro725bebb2013-05-17 16:08:53 -0400210 ctx->pos = (ctx->pos & ~(sb->s_blocksize - 1))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700211 | offset;
Al Viro725bebb2013-05-17 16:08:53 -0400212 file->f_version = inode->i_version;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700213 }
214
Al Viro725bebb2013-05-17 16:08:53 -0400215 while (ctx->pos < inode->i_size
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700216 && offset < sb->s_blocksize) {
Mingming Cao617ba132006-10-11 01:20:53 -0700217 de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
Al Viro725bebb2013-05-17 16:08:53 -0400218 if (ext4_check_dir_entry(inode, file, de, bh,
Tao Ma226ba972012-12-10 14:05:58 -0500219 bh->b_data, bh->b_size,
220 offset)) {
Andrew Morton63f57932006-10-11 01:21:24 -0700221 /*
Al Viro725bebb2013-05-17 16:08:53 -0400222 * On error, skip to the next block
Andrew Morton63f57932006-10-11 01:21:24 -0700223 */
Al Viro725bebb2013-05-17 16:08:53 -0400224 ctx->pos = (ctx->pos |
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700225 (sb->s_blocksize - 1)) + 1;
Al Viro725bebb2013-05-17 16:08:53 -0400226 break;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700227 }
Wei Yongjun3d0518f2009-02-14 23:01:36 -0500228 offset += ext4_rec_len_from_disk(de->rec_len,
229 sb->s_blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700230 if (le32_to_cpu(de->inode)) {
Al Viro725bebb2013-05-17 16:08:53 -0400231 if (!dir_emit(ctx, de->name,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700232 de->name_len,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700233 le32_to_cpu(de->inode),
Al Viro725bebb2013-05-17 16:08:53 -0400234 get_dtype(sb, de->file_type))) {
235 brelse(bh);
236 return 0;
237 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700238 }
Al Viro725bebb2013-05-17 16:08:53 -0400239 ctx->pos += ext4_rec_len_from_disk(de->rec_len,
Wei Yongjun3d0518f2009-02-14 23:01:36 -0500240 sb->s_blocksize);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700241 }
242 offset = 0;
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400243 brelse(bh);
Al Viro725bebb2013-05-17 16:08:53 -0400244 if (ctx->pos < inode->i_size) {
245 if (!dir_relax(inode))
246 return 0;
247 }
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700248 }
Al Viro725bebb2013-05-17 16:08:53 -0400249 return 0;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700250}
251
Fan Yongd1f52732012-03-18 22:44:40 -0400252static inline int is_32bit_api(void)
253{
254#ifdef CONFIG_COMPAT
255 return is_compat_task();
256#else
257 return (BITS_PER_LONG == 32);
258#endif
259}
260
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700261/*
262 * These functions convert from the major/minor hash to an f_pos
Fan Yongd1f52732012-03-18 22:44:40 -0400263 * value for dx directories
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700264 *
Fan Yongd1f52732012-03-18 22:44:40 -0400265 * Upper layer (for example NFS) should specify FMODE_32BITHASH or
266 * FMODE_64BITHASH explicitly. On the other hand, we allow ext4 to be mounted
267 * directly on both 32-bit and 64-bit nodes, under such case, neither
268 * FMODE_32BITHASH nor FMODE_64BITHASH is specified.
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700269 */
Fan Yongd1f52732012-03-18 22:44:40 -0400270static inline loff_t hash2pos(struct file *filp, __u32 major, __u32 minor)
271{
272 if ((filp->f_mode & FMODE_32BITHASH) ||
273 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
274 return major >> 1;
275 else
276 return ((__u64)(major >> 1) << 32) | (__u64)minor;
277}
278
279static inline __u32 pos2maj_hash(struct file *filp, loff_t pos)
280{
281 if ((filp->f_mode & FMODE_32BITHASH) ||
282 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
283 return (pos << 1) & 0xffffffff;
284 else
285 return ((pos >> 32) << 1) & 0xffffffff;
286}
287
288static inline __u32 pos2min_hash(struct file *filp, loff_t pos)
289{
290 if ((filp->f_mode & FMODE_32BITHASH) ||
291 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
292 return 0;
293 else
294 return pos & 0xffffffff;
295}
296
297/*
298 * Return 32- or 64-bit end-of-file for dx directories
299 */
300static inline loff_t ext4_get_htree_eof(struct file *filp)
301{
302 if ((filp->f_mode & FMODE_32BITHASH) ||
303 (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api()))
304 return EXT4_HTREE_EOF_32BIT;
305 else
306 return EXT4_HTREE_EOF_64BIT;
307}
308
309
310/*
Eric Sandeenec7268c2012-04-30 13:14:03 -0500311 * ext4_dir_llseek() calls generic_file_llseek_size to handle htree
312 * directories, where the "offset" is in terms of the filename hash
313 * value instead of the byte offset.
Fan Yongd1f52732012-03-18 22:44:40 -0400314 *
Eric Sandeenec7268c2012-04-30 13:14:03 -0500315 * Because we may return a 64-bit hash that is well beyond offset limits,
316 * we need to pass the max hash as the maximum allowable offset in
317 * the htree directory case.
318 *
319 * For non-htree, ext4_llseek already chooses the proper max offset.
Fan Yongd1f52732012-03-18 22:44:40 -0400320 */
Zheng Liud4e43952013-03-02 17:24:05 -0500321static loff_t ext4_dir_llseek(struct file *file, loff_t offset, int whence)
Fan Yongd1f52732012-03-18 22:44:40 -0400322{
323 struct inode *inode = file->f_mapping->host;
Fan Yongd1f52732012-03-18 22:44:40 -0400324 int dx_dir = is_dx_dir(inode);
Eric Sandeenec7268c2012-04-30 13:14:03 -0500325 loff_t htree_max = ext4_get_htree_eof(file);
Fan Yongd1f52732012-03-18 22:44:40 -0400326
Eric Sandeenec7268c2012-04-30 13:14:03 -0500327 if (likely(dx_dir))
Andrew Morton965c8e52012-12-17 15:59:39 -0800328 return generic_file_llseek_size(file, offset, whence,
Eric Sandeenec7268c2012-04-30 13:14:03 -0500329 htree_max, htree_max);
330 else
Andrew Morton965c8e52012-12-17 15:59:39 -0800331 return ext4_llseek(file, offset, whence);
Fan Yongd1f52732012-03-18 22:44:40 -0400332}
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700333
334/*
335 * This structure holds the nodes of the red-black tree used to store
336 * the directory entry in hash order.
337 */
338struct fname {
339 __u32 hash;
340 __u32 minor_hash;
341 struct rb_node rb_hash;
342 struct fname *next;
343 __u32 inode;
344 __u8 name_len;
345 __u8 file_type;
346 char name[0];
347};
348
349/*
350 * This functoin implements a non-recursive way of freeing all of the
351 * nodes in the red-black tree.
352 */
353static void free_rb_tree_fname(struct rb_root *root)
354{
Cody P Schaferd1866bd2014-01-23 15:56:10 -0800355 struct fname *fname, *next;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700356
Cody P Schaferd1866bd2014-01-23 15:56:10 -0800357 rbtree_postorder_for_each_entry_safe(fname, next, root, rb_hash)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700358 while (fname) {
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400359 struct fname *old = fname;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700360 fname = fname->next;
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400361 kfree(old);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700362 }
Cody P Schaferd1866bd2014-01-23 15:56:10 -0800363
364 *root = RB_ROOT;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700365}
366
367
Fan Yongd1f52732012-03-18 22:44:40 -0400368static struct dir_private_info *ext4_htree_create_dir_info(struct file *filp,
369 loff_t pos)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700370{
371 struct dir_private_info *p;
372
Shen Feng69baee02008-07-11 19:27:31 -0400373 p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700374 if (!p)
375 return NULL;
Fan Yongd1f52732012-03-18 22:44:40 -0400376 p->curr_hash = pos2maj_hash(filp, pos);
377 p->curr_minor_hash = pos2min_hash(filp, pos);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700378 return p;
379}
380
Mingming Cao617ba132006-10-11 01:20:53 -0700381void ext4_htree_free_dir_info(struct dir_private_info *p)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700382{
383 free_rb_tree_fname(&p->root);
384 kfree(p);
385}
386
387/*
388 * Given a directory entry, enter it into the fname rb tree.
389 */
Mingming Cao617ba132006-10-11 01:20:53 -0700390int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700391 __u32 minor_hash,
Mingming Cao617ba132006-10-11 01:20:53 -0700392 struct ext4_dir_entry_2 *dirent)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700393{
394 struct rb_node **p, *parent = NULL;
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400395 struct fname *fname, *new_fn;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700396 struct dir_private_info *info;
397 int len;
398
Joe Perchesa271fe82010-07-27 11:56:04 -0400399 info = dir_file->private_data;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700400 p = &info->root.rb_node;
401
402 /* Create and allocate the fname structure */
403 len = sizeof(struct fname) + dirent->name_len + 1;
404 new_fn = kzalloc(len, GFP_KERNEL);
405 if (!new_fn)
406 return -ENOMEM;
407 new_fn->hash = hash;
408 new_fn->minor_hash = minor_hash;
409 new_fn->inode = le32_to_cpu(dirent->inode);
410 new_fn->name_len = dirent->name_len;
411 new_fn->file_type = dirent->file_type;
412 memcpy(new_fn->name, dirent->name, dirent->name_len);
413 new_fn->name[dirent->name_len] = 0;
414
415 while (*p) {
416 parent = *p;
417 fname = rb_entry(parent, struct fname, rb_hash);
418
419 /*
420 * If the hash and minor hash match up, then we put
421 * them on a linked list. This rarely happens...
422 */
423 if ((new_fn->hash == fname->hash) &&
424 (new_fn->minor_hash == fname->minor_hash)) {
425 new_fn->next = fname->next;
426 fname->next = new_fn;
427 return 0;
428 }
429
430 if (new_fn->hash < fname->hash)
431 p = &(*p)->rb_left;
432 else if (new_fn->hash > fname->hash)
433 p = &(*p)->rb_right;
434 else if (new_fn->minor_hash < fname->minor_hash)
435 p = &(*p)->rb_left;
436 else /* if (new_fn->minor_hash > fname->minor_hash) */
437 p = &(*p)->rb_right;
438 }
439
440 rb_link_node(&new_fn->rb_hash, parent, p);
441 rb_insert_color(&new_fn->rb_hash, &info->root);
442 return 0;
443}
444
445
446
447/*
Mingming Cao617ba132006-10-11 01:20:53 -0700448 * This is a helper function for ext4_dx_readdir. It calls filldir
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700449 * for all entres on the fname linked list. (Normally there is only
450 * one entry on the linked list, unless there are 62 bit hash collisions.)
451 */
Al Viro725bebb2013-05-17 16:08:53 -0400452static int call_filldir(struct file *file, struct dir_context *ctx,
453 struct fname *fname)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700454{
Al Viro725bebb2013-05-17 16:08:53 -0400455 struct dir_private_info *info = file->private_data;
456 struct inode *inode = file_inode(file);
457 struct super_block *sb = inode->i_sb;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700458
459 if (!fname) {
Theodore Ts'o92b97812012-03-19 23:41:49 -0400460 ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: comm %s: "
461 "called with null fname?!?", __func__, __LINE__,
462 inode->i_ino, current->comm);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700463 return 0;
464 }
Al Viro725bebb2013-05-17 16:08:53 -0400465 ctx->pos = hash2pos(file, fname->hash, fname->minor_hash);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700466 while (fname) {
Al Viro725bebb2013-05-17 16:08:53 -0400467 if (!dir_emit(ctx, fname->name,
468 fname->name_len,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700469 fname->inode,
Al Viro725bebb2013-05-17 16:08:53 -0400470 get_dtype(sb, fname->file_type))) {
Theodore Ts'od0156412008-08-19 21:57:43 -0400471 info->extra_fname = fname;
Al Viro725bebb2013-05-17 16:08:53 -0400472 return 1;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700473 }
474 fname = fname->next;
475 }
476 return 0;
477}
478
Al Viro725bebb2013-05-17 16:08:53 -0400479static int ext4_dx_readdir(struct file *file, struct dir_context *ctx)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700480{
Al Viro725bebb2013-05-17 16:08:53 -0400481 struct dir_private_info *info = file->private_data;
482 struct inode *inode = file_inode(file);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700483 struct fname *fname;
484 int ret;
485
486 if (!info) {
Al Viro725bebb2013-05-17 16:08:53 -0400487 info = ext4_htree_create_dir_info(file, ctx->pos);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700488 if (!info)
489 return -ENOMEM;
Al Viro725bebb2013-05-17 16:08:53 -0400490 file->private_data = info;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700491 }
492
Al Viro725bebb2013-05-17 16:08:53 -0400493 if (ctx->pos == ext4_get_htree_eof(file))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700494 return 0; /* EOF */
495
496 /* Some one has messed with f_pos; reset the world */
Al Viro725bebb2013-05-17 16:08:53 -0400497 if (info->last_pos != ctx->pos) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700498 free_rb_tree_fname(&info->root);
499 info->curr_node = NULL;
500 info->extra_fname = NULL;
Al Viro725bebb2013-05-17 16:08:53 -0400501 info->curr_hash = pos2maj_hash(file, ctx->pos);
502 info->curr_minor_hash = pos2min_hash(file, ctx->pos);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700503 }
504
505 /*
506 * If there are any leftover names on the hash collision
507 * chain, return them first.
508 */
Theodore Ts'od0156412008-08-19 21:57:43 -0400509 if (info->extra_fname) {
Al Viro725bebb2013-05-17 16:08:53 -0400510 if (call_filldir(file, ctx, info->extra_fname))
Theodore Ts'od0156412008-08-19 21:57:43 -0400511 goto finished;
Theodore Ts'od0156412008-08-19 21:57:43 -0400512 info->extra_fname = NULL;
Theodore Ts'o3c37fc82008-10-25 11:39:08 -0400513 goto next_node;
Theodore Ts'od0156412008-08-19 21:57:43 -0400514 } else if (!info->curr_node)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700515 info->curr_node = rb_first(&info->root);
516
517 while (1) {
518 /*
519 * Fill the rbtree if we have no more entries,
520 * or the inode has changed since we last read in the
521 * cached entries.
522 */
523 if ((!info->curr_node) ||
Al Viro725bebb2013-05-17 16:08:53 -0400524 (file->f_version != inode->i_version)) {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700525 info->curr_node = NULL;
526 free_rb_tree_fname(&info->root);
Al Viro725bebb2013-05-17 16:08:53 -0400527 file->f_version = inode->i_version;
528 ret = ext4_htree_fill_tree(file, info->curr_hash,
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700529 info->curr_minor_hash,
530 &info->next_hash);
531 if (ret < 0)
532 return ret;
533 if (ret == 0) {
Al Viro725bebb2013-05-17 16:08:53 -0400534 ctx->pos = ext4_get_htree_eof(file);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700535 break;
536 }
537 info->curr_node = rb_first(&info->root);
538 }
539
540 fname = rb_entry(info->curr_node, struct fname, rb_hash);
541 info->curr_hash = fname->hash;
542 info->curr_minor_hash = fname->minor_hash;
Al Viro725bebb2013-05-17 16:08:53 -0400543 if (call_filldir(file, ctx, fname))
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700544 break;
Theodore Ts'o3c37fc82008-10-25 11:39:08 -0400545 next_node:
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700546 info->curr_node = rb_next(info->curr_node);
Theodore Ts'o3c37fc82008-10-25 11:39:08 -0400547 if (info->curr_node) {
548 fname = rb_entry(info->curr_node, struct fname,
549 rb_hash);
550 info->curr_hash = fname->hash;
551 info->curr_minor_hash = fname->minor_hash;
552 } else {
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700553 if (info->next_hash == ~0) {
Al Viro725bebb2013-05-17 16:08:53 -0400554 ctx->pos = ext4_get_htree_eof(file);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700555 break;
556 }
557 info->curr_hash = info->next_hash;
558 info->curr_minor_hash = 0;
559 }
560 }
561finished:
Al Viro725bebb2013-05-17 16:08:53 -0400562 info->last_pos = ctx->pos;
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700563 return 0;
564}
565
Theodore Ts'oaf5bc922008-09-08 22:25:24 -0400566static int ext4_release_dir(struct inode *inode, struct file *filp)
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700567{
Andrew Morton63f57932006-10-11 01:21:24 -0700568 if (filp->private_data)
Mingming Cao617ba132006-10-11 01:20:53 -0700569 ext4_htree_free_dir_info(filp->private_data);
Dave Kleikampac27a0e2006-10-11 01:20:50 -0700570
571 return 0;
572}
Fan Yongd1f52732012-03-18 22:44:40 -0400573
Darrick J. Wong40b163f2014-07-28 13:06:26 -0400574int ext4_check_all_de(struct inode *dir, struct buffer_head *bh, void *buf,
575 int buf_size)
576{
577 struct ext4_dir_entry_2 *de;
578 int nlen, rlen;
579 unsigned int offset = 0;
580 char *top;
581
582 de = (struct ext4_dir_entry_2 *)buf;
583 top = buf + buf_size;
584 while ((char *) de < top) {
585 if (ext4_check_dir_entry(dir, NULL, de, bh,
586 buf, buf_size, offset))
587 return -EIO;
588 nlen = EXT4_DIR_REC_LEN(de->name_len);
589 rlen = ext4_rec_len_from_disk(de->rec_len, buf_size);
590 de = (struct ext4_dir_entry_2 *)((char *)de + rlen);
591 offset += rlen;
592 }
593 if ((char *) de > top)
594 return -EIO;
595
596 return 0;
597}
598
Fan Yongd1f52732012-03-18 22:44:40 -0400599const struct file_operations ext4_dir_operations = {
600 .llseek = ext4_dir_llseek,
601 .read = generic_read_dir,
Al Viro725bebb2013-05-17 16:08:53 -0400602 .iterate = ext4_readdir,
Fan Yongd1f52732012-03-18 22:44:40 -0400603 .unlocked_ioctl = ext4_ioctl,
604#ifdef CONFIG_COMPAT
605 .compat_ioctl = ext4_compat_ioctl,
606#endif
607 .fsync = ext4_sync_file,
608 .release = ext4_release_dir,
609};