linux/fs/f2fs/dir.c

944 lines
23 KiB
C
Raw Normal View History

/*
* fs/f2fs/dir.c
*
* Copyright (c) 2012 Samsung Electronics Co., Ltd.
* http://www.samsung.com/
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#include <linux/fs.h>
#include <linux/f2fs_fs.h>
#include <linux/sched/signal.h>
#include "f2fs.h"
#include "node.h"
#include "acl.h"
#include "xattr.h"
#include <trace/events/f2fs.h>
static unsigned long dir_blocks(struct inode *inode)
{
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 20:29:47 +08:00
return ((unsigned long long) (i_size_read(inode) + PAGE_SIZE - 1))
>> PAGE_SHIFT;
}
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
static unsigned int dir_buckets(unsigned int level, int dir_level)
{
if (level + dir_level < MAX_DIR_HASH_DEPTH / 2)
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
return 1 << (level + dir_level);
else
return MAX_DIR_BUCKETS;
}
static unsigned int bucket_blocks(unsigned int level)
{
if (level < MAX_DIR_HASH_DEPTH / 2)
return 2;
else
return 4;
}
static unsigned char f2fs_filetype_table[F2FS_FT_MAX] = {
[F2FS_FT_UNKNOWN] = DT_UNKNOWN,
[F2FS_FT_REG_FILE] = DT_REG,
[F2FS_FT_DIR] = DT_DIR,
[F2FS_FT_CHRDEV] = DT_CHR,
[F2FS_FT_BLKDEV] = DT_BLK,
[F2FS_FT_FIFO] = DT_FIFO,
[F2FS_FT_SOCK] = DT_SOCK,
[F2FS_FT_SYMLINK] = DT_LNK,
};
static unsigned char f2fs_type_by_mode[S_IFMT >> S_SHIFT] = {
[S_IFREG >> S_SHIFT] = F2FS_FT_REG_FILE,
[S_IFDIR >> S_SHIFT] = F2FS_FT_DIR,
[S_IFCHR >> S_SHIFT] = F2FS_FT_CHRDEV,
[S_IFBLK >> S_SHIFT] = F2FS_FT_BLKDEV,
[S_IFIFO >> S_SHIFT] = F2FS_FT_FIFO,
[S_IFSOCK >> S_SHIFT] = F2FS_FT_SOCK,
[S_IFLNK >> S_SHIFT] = F2FS_FT_SYMLINK,
};
void set_de_type(struct f2fs_dir_entry *de, umode_t mode)
{
de->file_type = f2fs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
}
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
unsigned char get_de_type(struct f2fs_dir_entry *de)
{
if (de->file_type < F2FS_FT_MAX)
return f2fs_filetype_table[de->file_type];
return DT_UNKNOWN;
}
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
static unsigned long dir_block_index(unsigned int level,
int dir_level, unsigned int idx)
{
unsigned long i;
unsigned long bidx = 0;
for (i = 0; i < level; i++)
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
bidx += idx * bucket_blocks(level);
return bidx;
}
static struct f2fs_dir_entry *find_in_block(struct page *dentry_page,
struct fscrypt_name *fname,
f2fs_hash_t namehash,
int *max_slots,
struct page **res_page)
{
struct f2fs_dentry_block *dentry_blk;
struct f2fs_dir_entry *de;
struct f2fs_dentry_ptr d;
dentry_blk = (struct f2fs_dentry_block *)kmap(dentry_page);
make_dentry_ptr_block(NULL, &d, dentry_blk);
de = find_target_dentry(fname, namehash, max_slots, &d);
if (de)
*res_page = dentry_page;
else
kunmap(dentry_page);
return de;
}
struct f2fs_dir_entry *find_target_dentry(struct fscrypt_name *fname,
f2fs_hash_t namehash, int *max_slots,
struct f2fs_dentry_ptr *d)
{
struct f2fs_dir_entry *de;
unsigned long bit_pos = 0;
int max_len = 0;
if (max_slots)
*max_slots = 0;
while (bit_pos < d->max) {
if (!test_bit_le(bit_pos, d->bitmap)) {
bit_pos++;
max_len++;
continue;
}
de = &d->dentry[bit_pos];
if (unlikely(!de->name_len)) {
bit_pos++;
continue;
}
if (de->hash_code == namehash &&
fscrypt_match_name(fname, d->filename[bit_pos],
le16_to_cpu(de->name_len)))
goto found;
if (max_slots && max_len > *max_slots)
*max_slots = max_len;
max_len = 0;
bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
}
de = NULL;
found:
if (max_slots && max_len > *max_slots)
*max_slots = max_len;
return de;
}
static struct f2fs_dir_entry *find_in_level(struct inode *dir,
unsigned int level,
struct fscrypt_name *fname,
struct page **res_page)
{
struct qstr name = FSTR_TO_QSTR(&fname->disk_name);
int s = GET_DENTRY_SLOTS(name.len);
unsigned int nbucket, nblock;
unsigned int bidx, end_block;
struct page *dentry_page;
struct f2fs_dir_entry *de = NULL;
bool room = false;
int max_slots;
f2fs_hash_t namehash = f2fs_dentry_hash(&name, fname);
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
nblock = bucket_blocks(level);
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
le32_to_cpu(namehash) % nbucket);
end_block = bidx + nblock;
for (; bidx < end_block; bidx++) {
/* no need to allocate new dentry pages to all the indices */
dentry_page = find_data_page(dir, bidx);
if (IS_ERR(dentry_page)) {
if (PTR_ERR(dentry_page) == -ENOENT) {
room = true;
continue;
} else {
*res_page = dentry_page;
break;
}
}
de = find_in_block(dentry_page, fname, namehash, &max_slots,
res_page);
if (de)
break;
if (max_slots >= s)
room = true;
f2fs_put_page(dentry_page, 0);
}
if (!de && room && F2FS_I(dir)->chash != namehash) {
F2FS_I(dir)->chash = namehash;
F2FS_I(dir)->clevel = level;
}
return de;
}
struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
struct fscrypt_name *fname, struct page **res_page)
{
unsigned long npages = dir_blocks(dir);
struct f2fs_dir_entry *de = NULL;
unsigned int max_depth;
unsigned int level;
if (f2fs_has_inline_dentry(dir)) {
*res_page = NULL;
de = find_in_inline_dir(dir, fname, res_page);
goto out;
}
if (npages == 0) {
*res_page = NULL;
goto out;
}
max_depth = F2FS_I(dir)->i_current_depth;
if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
f2fs_msg(F2FS_I_SB(dir)->sb, KERN_WARNING,
"Corrupted max_depth of %lu: %u",
dir->i_ino, max_depth);
max_depth = MAX_DIR_HASH_DEPTH;
f2fs_i_depth_write(dir, max_depth);
}
for (level = 0; level < max_depth; level++) {
*res_page = NULL;
de = find_in_level(dir, level, fname, res_page);
if (de || IS_ERR(*res_page))
break;
}
out:
/* This is to increase the speed of f2fs_create */
if (!de)
F2FS_I(dir)->task = current;
return de;
}
/*
* Find an entry in the specified directory with the wanted name.
* It returns the page where the entry was found (as a parameter - res_page),
* and the entry itself. Page is returned mapped and unlocked.
* Entry is guaranteed to be valid.
*/
struct f2fs_dir_entry *f2fs_find_entry(struct inode *dir,
const struct qstr *child, struct page **res_page)
{
struct f2fs_dir_entry *de = NULL;
struct fscrypt_name fname;
int err;
err = fscrypt_setup_filename(dir, child, 1, &fname);
if (err) {
if (err == -ENOENT)
*res_page = NULL;
else
*res_page = ERR_PTR(err);
return NULL;
}
de = __f2fs_find_entry(dir, &fname, res_page);
fscrypt_free_filename(&fname);
return de;
}
struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
{
struct qstr dotdot = QSTR_INIT("..", 2);
return f2fs_find_entry(dir, &dotdot, p);
}
ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
struct page **page)
{
ino_t res = 0;
struct f2fs_dir_entry *de;
de = f2fs_find_entry(dir, qstr, page);
if (de) {
res = le32_to_cpu(de->ino);
f2fs_dentry_kunmap(dir, *page);
f2fs_put_page(*page, 0);
}
return res;
}
void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
struct page *page, struct inode *inode)
{
enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
lock_page(page);
f2fs_wait_on_page_writeback(page, type, true);
de->ino = cpu_to_le32(inode->i_ino);
set_de_type(de, inode->i_mode);
f2fs_dentry_kunmap(dir, page);
set_page_dirty(page);
dir->i_mtime = dir->i_ctime = current_time(dir);
f2fs_mark_inode_dirty_sync(dir, false);
f2fs_put_page(page, 1);
}
static void init_dent_inode(const struct qstr *name, struct page *ipage)
{
struct f2fs_inode *ri;
f2fs_wait_on_page_writeback(ipage, NODE, true);
/* copy name info. to this inode page */
ri = F2FS_INODE(ipage);
ri->i_namelen = cpu_to_le32(name->len);
memcpy(ri->i_name, name->name, name->len);
set_page_dirty(ipage);
}
void do_make_empty_dir(struct inode *inode, struct inode *parent,
struct f2fs_dentry_ptr *d)
{
struct qstr dot = QSTR_INIT(".", 1);
struct qstr dotdot = QSTR_INIT("..", 2);
/* update dirent of "." */
f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
/* update dirent of ".." */
f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
}
static int make_empty_dir(struct inode *inode,
struct inode *parent, struct page *page)
f2fs: introduce a new global lock scheme In the previous version, f2fs uses global locks according to the usage types, such as directory operations, block allocation, block write, and so on. Reference the following lock types in f2fs.h. enum lock_type { RENAME, /* for renaming operations */ DENTRY_OPS, /* for directory operations */ DATA_WRITE, /* for data write */ DATA_NEW, /* for data allocation */ DATA_TRUNC, /* for data truncate */ NODE_NEW, /* for node allocation */ NODE_TRUNC, /* for node truncate */ NODE_WRITE, /* for node write */ NR_LOCK_TYPE, }; In that case, we lose the performance under the multi-threading environment, since every types of operations must be conducted one at a time. In order to address the problem, let's share the locks globally with a mutex array regardless of any types. So, let users grab a mutex and perform their jobs in parallel as much as possbile. For this, I propose a new global lock scheme as follows. 0. Data structure - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS] - f2fs_sb_info -> node_write 1. mutex_lock_op(sbi) - try to get an avaiable lock from the array. - returns the index of the gottern lock variable. 2. mutex_unlock_op(sbi, index of the lock) - unlock the given index of the lock. 3. mutex_lock_all(sbi) - grab all the locks in the array before the checkpoint. 4. mutex_unlock_all(sbi) - release all the locks in the array after checkpoint. 5. block_operations() - call mutex_lock_all() - sync_dirty_dir_inodes() - grab node_write - sync_node_pages() Note that, the pairs of mutex_lock_op()/mutex_unlock_op() and mutex_lock_all()/mutex_unlock_all() should be used together. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 15:21:29 +08:00
{
struct page *dentry_page;
struct f2fs_dentry_block *dentry_blk;
struct f2fs_dentry_ptr d;
f2fs: introduce a new global lock scheme In the previous version, f2fs uses global locks according to the usage types, such as directory operations, block allocation, block write, and so on. Reference the following lock types in f2fs.h. enum lock_type { RENAME, /* for renaming operations */ DENTRY_OPS, /* for directory operations */ DATA_WRITE, /* for data write */ DATA_NEW, /* for data allocation */ DATA_TRUNC, /* for data truncate */ NODE_NEW, /* for node allocation */ NODE_TRUNC, /* for node truncate */ NODE_WRITE, /* for node write */ NR_LOCK_TYPE, }; In that case, we lose the performance under the multi-threading environment, since every types of operations must be conducted one at a time. In order to address the problem, let's share the locks globally with a mutex array regardless of any types. So, let users grab a mutex and perform their jobs in parallel as much as possbile. For this, I propose a new global lock scheme as follows. 0. Data structure - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS] - f2fs_sb_info -> node_write 1. mutex_lock_op(sbi) - try to get an avaiable lock from the array. - returns the index of the gottern lock variable. 2. mutex_unlock_op(sbi, index of the lock) - unlock the given index of the lock. 3. mutex_lock_all(sbi) - grab all the locks in the array before the checkpoint. 4. mutex_unlock_all(sbi) - release all the locks in the array after checkpoint. 5. block_operations() - call mutex_lock_all() - sync_dirty_dir_inodes() - grab node_write - sync_node_pages() Note that, the pairs of mutex_lock_op()/mutex_unlock_op() and mutex_lock_all()/mutex_unlock_all() should be used together. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 15:21:29 +08:00
if (f2fs_has_inline_dentry(inode))
return make_empty_inline_dir(inode, parent, page);
dentry_page = get_new_data_page(inode, page, 0, true);
f2fs: introduce a new global lock scheme In the previous version, f2fs uses global locks according to the usage types, such as directory operations, block allocation, block write, and so on. Reference the following lock types in f2fs.h. enum lock_type { RENAME, /* for renaming operations */ DENTRY_OPS, /* for directory operations */ DATA_WRITE, /* for data write */ DATA_NEW, /* for data allocation */ DATA_TRUNC, /* for data truncate */ NODE_NEW, /* for node allocation */ NODE_TRUNC, /* for node truncate */ NODE_WRITE, /* for node write */ NR_LOCK_TYPE, }; In that case, we lose the performance under the multi-threading environment, since every types of operations must be conducted one at a time. In order to address the problem, let's share the locks globally with a mutex array regardless of any types. So, let users grab a mutex and perform their jobs in parallel as much as possbile. For this, I propose a new global lock scheme as follows. 0. Data structure - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS] - f2fs_sb_info -> node_write 1. mutex_lock_op(sbi) - try to get an avaiable lock from the array. - returns the index of the gottern lock variable. 2. mutex_unlock_op(sbi, index of the lock) - unlock the given index of the lock. 3. mutex_lock_all(sbi) - grab all the locks in the array before the checkpoint. 4. mutex_unlock_all(sbi) - release all the locks in the array after checkpoint. 5. block_operations() - call mutex_lock_all() - sync_dirty_dir_inodes() - grab node_write - sync_node_pages() Note that, the pairs of mutex_lock_op()/mutex_unlock_op() and mutex_lock_all()/mutex_unlock_all() should be used together. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 15:21:29 +08:00
if (IS_ERR(dentry_page))
return PTR_ERR(dentry_page);
dentry_blk = kmap_atomic(dentry_page);
f2fs: introduce a new global lock scheme In the previous version, f2fs uses global locks according to the usage types, such as directory operations, block allocation, block write, and so on. Reference the following lock types in f2fs.h. enum lock_type { RENAME, /* for renaming operations */ DENTRY_OPS, /* for directory operations */ DATA_WRITE, /* for data write */ DATA_NEW, /* for data allocation */ DATA_TRUNC, /* for data truncate */ NODE_NEW, /* for node allocation */ NODE_TRUNC, /* for node truncate */ NODE_WRITE, /* for node write */ NR_LOCK_TYPE, }; In that case, we lose the performance under the multi-threading environment, since every types of operations must be conducted one at a time. In order to address the problem, let's share the locks globally with a mutex array regardless of any types. So, let users grab a mutex and perform their jobs in parallel as much as possbile. For this, I propose a new global lock scheme as follows. 0. Data structure - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS] - f2fs_sb_info -> node_write 1. mutex_lock_op(sbi) - try to get an avaiable lock from the array. - returns the index of the gottern lock variable. 2. mutex_unlock_op(sbi, index of the lock) - unlock the given index of the lock. 3. mutex_lock_all(sbi) - grab all the locks in the array before the checkpoint. 4. mutex_unlock_all(sbi) - release all the locks in the array after checkpoint. 5. block_operations() - call mutex_lock_all() - sync_dirty_dir_inodes() - grab node_write - sync_node_pages() Note that, the pairs of mutex_lock_op()/mutex_unlock_op() and mutex_lock_all()/mutex_unlock_all() should be used together. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 15:21:29 +08:00
make_dentry_ptr_block(NULL, &d, dentry_blk);
do_make_empty_dir(inode, parent, &d);
f2fs: introduce a new global lock scheme In the previous version, f2fs uses global locks according to the usage types, such as directory operations, block allocation, block write, and so on. Reference the following lock types in f2fs.h. enum lock_type { RENAME, /* for renaming operations */ DENTRY_OPS, /* for directory operations */ DATA_WRITE, /* for data write */ DATA_NEW, /* for data allocation */ DATA_TRUNC, /* for data truncate */ NODE_NEW, /* for node allocation */ NODE_TRUNC, /* for node truncate */ NODE_WRITE, /* for node write */ NR_LOCK_TYPE, }; In that case, we lose the performance under the multi-threading environment, since every types of operations must be conducted one at a time. In order to address the problem, let's share the locks globally with a mutex array regardless of any types. So, let users grab a mutex and perform their jobs in parallel as much as possbile. For this, I propose a new global lock scheme as follows. 0. Data structure - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS] - f2fs_sb_info -> node_write 1. mutex_lock_op(sbi) - try to get an avaiable lock from the array. - returns the index of the gottern lock variable. 2. mutex_unlock_op(sbi, index of the lock) - unlock the given index of the lock. 3. mutex_lock_all(sbi) - grab all the locks in the array before the checkpoint. 4. mutex_unlock_all(sbi) - release all the locks in the array after checkpoint. 5. block_operations() - call mutex_lock_all() - sync_dirty_dir_inodes() - grab node_write - sync_node_pages() Note that, the pairs of mutex_lock_op()/mutex_unlock_op() and mutex_lock_all()/mutex_unlock_all() should be used together. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 15:21:29 +08:00
kunmap_atomic(dentry_blk);
f2fs: introduce a new global lock scheme In the previous version, f2fs uses global locks according to the usage types, such as directory operations, block allocation, block write, and so on. Reference the following lock types in f2fs.h. enum lock_type { RENAME, /* for renaming operations */ DENTRY_OPS, /* for directory operations */ DATA_WRITE, /* for data write */ DATA_NEW, /* for data allocation */ DATA_TRUNC, /* for data truncate */ NODE_NEW, /* for node allocation */ NODE_TRUNC, /* for node truncate */ NODE_WRITE, /* for node write */ NR_LOCK_TYPE, }; In that case, we lose the performance under the multi-threading environment, since every types of operations must be conducted one at a time. In order to address the problem, let's share the locks globally with a mutex array regardless of any types. So, let users grab a mutex and perform their jobs in parallel as much as possbile. For this, I propose a new global lock scheme as follows. 0. Data structure - f2fs_sb_info -> mutex_lock[NR_GLOBAL_LOCKS] - f2fs_sb_info -> node_write 1. mutex_lock_op(sbi) - try to get an avaiable lock from the array. - returns the index of the gottern lock variable. 2. mutex_unlock_op(sbi, index of the lock) - unlock the given index of the lock. 3. mutex_lock_all(sbi) - grab all the locks in the array before the checkpoint. 4. mutex_unlock_all(sbi) - release all the locks in the array after checkpoint. 5. block_operations() - call mutex_lock_all() - sync_dirty_dir_inodes() - grab node_write - sync_node_pages() Note that, the pairs of mutex_lock_op()/mutex_unlock_op() and mutex_lock_all()/mutex_unlock_all() should be used together. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-11-22 15:21:29 +08:00
set_page_dirty(dentry_page);
f2fs_put_page(dentry_page, 1);
return 0;
}
f2fs: avoid deadlock on init_inode_metadata Previously, init_inode_metadata does not hold any parent directory's inode page. So, f2fs_init_acl can grab its parent inode page without any problem. But, when we use inline_dentry, that page is grabbed during f2fs_add_link, so that we can fall into deadlock condition like below. INFO: task mknod:11006 blocked for more than 120 seconds. Tainted: G OE 3.17.0-rc1+ #13 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. mknod D ffff88003fc94580 0 11006 11004 0x00000000 ffff880007717b10 0000000000000002 ffff88003c323220 ffff880007717fd8 0000000000014580 0000000000014580 ffff88003daecb30 ffff88003c323220 ffff88003fc94e80 ffff88003ffbb4e8 ffff880007717ba0 0000000000000002 Call Trace: [<ffffffff8173dc40>] ? bit_wait+0x50/0x50 [<ffffffff8173d4cd>] io_schedule+0x9d/0x130 [<ffffffff8173dc6c>] bit_wait_io+0x2c/0x50 [<ffffffff8173da3b>] __wait_on_bit_lock+0x4b/0xb0 [<ffffffff811640a7>] __lock_page+0x67/0x70 [<ffffffff810acf50>] ? autoremove_wake_function+0x40/0x40 [<ffffffff811652cc>] pagecache_get_page+0x14c/0x1e0 [<ffffffffa029afa9>] get_node_page+0x59/0x130 [f2fs] [<ffffffffa02a63ad>] read_all_xattrs+0x24d/0x430 [f2fs] [<ffffffffa02a6ca2>] f2fs_getxattr+0x52/0xe0 [f2fs] [<ffffffffa02a7481>] f2fs_get_acl+0x41/0x2d0 [f2fs] [<ffffffff8122d847>] get_acl+0x47/0x70 [<ffffffff8122db5a>] posix_acl_create+0x5a/0x150 [<ffffffffa02a7759>] f2fs_init_acl+0x29/0xcb [f2fs] [<ffffffffa0286a8d>] init_inode_metadata+0x5d/0x340 [f2fs] [<ffffffffa029253a>] f2fs_add_inline_entry+0x12a/0x2e0 [f2fs] [<ffffffffa0286ea5>] __f2fs_add_link+0x45/0x4a0 [f2fs] [<ffffffffa028b5b6>] ? f2fs_new_inode+0x146/0x220 [f2fs] [<ffffffffa028b816>] f2fs_mknod+0x86/0xf0 [f2fs] [<ffffffff811e3ec1>] vfs_mknod+0xe1/0x160 [<ffffffff811e4b26>] SyS_mknod+0x1f6/0x200 [<ffffffff81741d7f>] tracesys+0xe1/0xe6 Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2014-10-14 10:42:53 +08:00
struct page *init_inode_metadata(struct inode *inode, struct inode *dir,
const struct qstr *new_name, const struct qstr *orig_name,
struct page *dpage)
{
struct page *page;
int err;
if (is_inode_flag_set(inode, FI_NEW_INODE)) {
page = new_inode_page(inode);
if (IS_ERR(page))
return page;
if (S_ISDIR(inode->i_mode)) {
/* in order to handle error case */
get_page(page);
err = make_empty_dir(inode, dir, page);
if (err) {
lock_page(page);
goto put_error;
}
put_page(page);
}
f2fs: avoid deadlock on init_inode_metadata Previously, init_inode_metadata does not hold any parent directory's inode page. So, f2fs_init_acl can grab its parent inode page without any problem. But, when we use inline_dentry, that page is grabbed during f2fs_add_link, so that we can fall into deadlock condition like below. INFO: task mknod:11006 blocked for more than 120 seconds. Tainted: G OE 3.17.0-rc1+ #13 "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. mknod D ffff88003fc94580 0 11006 11004 0x00000000 ffff880007717b10 0000000000000002 ffff88003c323220 ffff880007717fd8 0000000000014580 0000000000014580 ffff88003daecb30 ffff88003c323220 ffff88003fc94e80 ffff88003ffbb4e8 ffff880007717ba0 0000000000000002 Call Trace: [<ffffffff8173dc40>] ? bit_wait+0x50/0x50 [<ffffffff8173d4cd>] io_schedule+0x9d/0x130 [<ffffffff8173dc6c>] bit_wait_io+0x2c/0x50 [<ffffffff8173da3b>] __wait_on_bit_lock+0x4b/0xb0 [<ffffffff811640a7>] __lock_page+0x67/0x70 [<ffffffff810acf50>] ? autoremove_wake_function+0x40/0x40 [<ffffffff811652cc>] pagecache_get_page+0x14c/0x1e0 [<ffffffffa029afa9>] get_node_page+0x59/0x130 [f2fs] [<ffffffffa02a63ad>] read_all_xattrs+0x24d/0x430 [f2fs] [<ffffffffa02a6ca2>] f2fs_getxattr+0x52/0xe0 [f2fs] [<ffffffffa02a7481>] f2fs_get_acl+0x41/0x2d0 [f2fs] [<ffffffff8122d847>] get_acl+0x47/0x70 [<ffffffff8122db5a>] posix_acl_create+0x5a/0x150 [<ffffffffa02a7759>] f2fs_init_acl+0x29/0xcb [f2fs] [<ffffffffa0286a8d>] init_inode_metadata+0x5d/0x340 [f2fs] [<ffffffffa029253a>] f2fs_add_inline_entry+0x12a/0x2e0 [f2fs] [<ffffffffa0286ea5>] __f2fs_add_link+0x45/0x4a0 [f2fs] [<ffffffffa028b5b6>] ? f2fs_new_inode+0x146/0x220 [f2fs] [<ffffffffa028b816>] f2fs_mknod+0x86/0xf0 [f2fs] [<ffffffff811e3ec1>] vfs_mknod+0xe1/0x160 [<ffffffff811e4b26>] SyS_mknod+0x1f6/0x200 [<ffffffff81741d7f>] tracesys+0xe1/0xe6 Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2014-10-14 10:42:53 +08:00
err = f2fs_init_acl(inode, dir, page, dpage);
if (err)
goto put_error;
err = f2fs_init_security(inode, dir, orig_name, page);
if (err)
goto put_error;
if (f2fs_encrypted_inode(dir) && f2fs_may_encrypt(inode)) {
err = fscrypt_inherit_context(dir, inode, page, false);
if (err)
goto put_error;
}
} else {
page = get_node_page(F2FS_I_SB(dir), inode->i_ino);
if (IS_ERR(page))
return page;
set_cold_node(inode, page);
}
if (new_name) {
init_dent_inode(new_name, page);
if (f2fs_encrypted_inode(dir))
file_set_enc_name(inode);
}
/*
* This file should be checkpointed during fsync.
* We lost i_pino from now on.
*/
if (is_inode_flag_set(inode, FI_INC_LINK)) {
if (!S_ISDIR(inode->i_mode))
file_lost_pino(inode);
/*
* If link the tmpfile to alias through linkat path,
* we should remove this inode from orphan list.
*/
if (inode->i_nlink == 0)
remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
f2fs_i_links_write(inode, true);
}
return page;
put_error:
clear_nlink(inode);
update_inode(inode, page);
f2fs_put_page(page, 1);
return ERR_PTR(err);
}
void update_parent_metadata(struct inode *dir, struct inode *inode,
unsigned int current_depth)
{
if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
if (S_ISDIR(inode->i_mode))
f2fs_i_links_write(dir, true);
clear_inode_flag(inode, FI_NEW_INODE);
}
dir->i_mtime = dir->i_ctime = current_time(dir);
f2fs_mark_inode_dirty_sync(dir, false);
if (F2FS_I(dir)->i_current_depth != current_depth)
f2fs_i_depth_write(dir, current_depth);
if (inode && is_inode_flag_set(inode, FI_INC_LINK))
clear_inode_flag(inode, FI_INC_LINK);
}
int room_for_filename(const void *bitmap, int slots, int max_slots)
{
int bit_start = 0;
int zero_start, zero_end;
next:
zero_start = find_next_zero_bit_le(bitmap, max_slots, bit_start);
if (zero_start >= max_slots)
return max_slots;
zero_end = find_next_bit_le(bitmap, max_slots, zero_start);
if (zero_end - zero_start >= slots)
return zero_start;
bit_start = zero_end + 1;
if (zero_end + 1 >= max_slots)
return max_slots;
goto next;
}
void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
const struct qstr *name, f2fs_hash_t name_hash,
unsigned int bit_pos)
{
struct f2fs_dir_entry *de;
int slots = GET_DENTRY_SLOTS(name->len);
int i;
de = &d->dentry[bit_pos];
de->hash_code = name_hash;
de->name_len = cpu_to_le16(name->len);
memcpy(d->filename[bit_pos], name->name, name->len);
de->ino = cpu_to_le32(ino);
set_de_type(de, mode);
for (i = 0; i < slots; i++) {
__set_bit_le(bit_pos + i, (void *)d->bitmap);
/* avoid wrong garbage data for readdir */
if (i)
(de + i)->name_len = 0;
}
}
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
int f2fs_add_regular_entry(struct inode *dir, const struct qstr *new_name,
const struct qstr *orig_name,
struct inode *inode, nid_t ino, umode_t mode)
{
unsigned int bit_pos;
unsigned int level;
unsigned int current_depth;
unsigned long bidx, block;
f2fs_hash_t dentry_hash;
unsigned int nbucket, nblock;
struct page *dentry_page = NULL;
struct f2fs_dentry_block *dentry_blk = NULL;
struct f2fs_dentry_ptr d;
struct page *page = NULL;
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
int slots, err = 0;
level = 0;
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
slots = GET_DENTRY_SLOTS(new_name->len);
dentry_hash = f2fs_dentry_hash(new_name, NULL);
current_depth = F2FS_I(dir)->i_current_depth;
if (F2FS_I(dir)->chash == dentry_hash) {
level = F2FS_I(dir)->clevel;
F2FS_I(dir)->chash = 0;
}
start:
#ifdef CONFIG_F2FS_FAULT_INJECTION
if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH)) {
f2fs_show_injection_info(FAULT_DIR_DEPTH);
return -ENOSPC;
}
#endif
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
return -ENOSPC;
/* Increase the depth, if required */
if (level == current_depth)
++current_depth;
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
nblock = bucket_blocks(level);
f2fs: introduce large directory support This patch introduces an i_dir_level field to support large directory. Previously, f2fs maintains multi-level hash tables to find a dentry quickly from a bunch of chiild dentries in a directory, and the hash tables consist of the following tree structure as below. In Documentation/filesystems/f2fs.txt, ---------------------- A : bucket B : block N : MAX_DIR_HASH_DEPTH ---------------------- level #0 | A(2B) | level #1 | A(2B) - A(2B) | level #2 | A(2B) - A(2B) - A(2B) - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) But, if we can guess that a directory will handle a number of child files, we don't need to traverse the tree from level #0 to #N all the time. Since the lower level tables contain relatively small number of dentries, the miss ratio of the target dentry is likely to be high. In order to avoid that, we can configure the hash tables sparsely from level #0 like this. level #0 | A(2B) - A(2B) - A(2B) - A(2B) level #1 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N/2 | A(2B) - A(2B) - A(2B) - A(2B) - A(2B) - ... - A(2B) . | . . . . level #N | A(4B) - A(4B) - A(4B) - A(4B) - A(4B) - ... - A(4B) With this structure, we can skip the ineffective tree searches in lower level hash tables. This patch adds just a facility for this by introducing i_dir_level in f2fs_inode. Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2014-02-27 17:20:00 +08:00
bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
(le32_to_cpu(dentry_hash) % nbucket));
for (block = bidx; block <= (bidx + nblock - 1); block++) {
dentry_page = get_new_data_page(dir, NULL, block, true);
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
if (IS_ERR(dentry_page))
return PTR_ERR(dentry_page);
dentry_blk = kmap(dentry_page);
bit_pos = room_for_filename(&dentry_blk->dentry_bitmap,
slots, NR_DENTRY_IN_BLOCK);
if (bit_pos < NR_DENTRY_IN_BLOCK)
goto add_dentry;
kunmap(dentry_page);
f2fs_put_page(dentry_page, 1);
}
/* Move to next level to find the empty slot for new dentry */
++level;
goto start;
add_dentry:
f2fs_wait_on_page_writeback(dentry_page, DATA, true);
if (inode) {
down_write(&F2FS_I(inode)->i_sem);
page = init_inode_metadata(inode, dir, new_name,
orig_name, NULL);
if (IS_ERR(page)) {
err = PTR_ERR(page);
goto fail;
}
}
make_dentry_ptr_block(NULL, &d, dentry_blk);
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
f2fs_update_dentry(ino, mode, &d, new_name, dentry_hash, bit_pos);
set_page_dirty(dentry_page);
if (inode) {
f2fs_i_pino_write(inode, dir->i_ino);
f2fs_put_page(page, 1);
}
update_parent_metadata(dir, inode, current_depth);
fail:
if (inode)
up_write(&F2FS_I(inode)->i_sem);
kunmap(dentry_page);
f2fs_put_page(dentry_page, 1);
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
return err;
}
int __f2fs_do_add_link(struct inode *dir, struct fscrypt_name *fname,
struct inode *inode, nid_t ino, umode_t mode)
{
struct qstr new_name;
int err = -EAGAIN;
new_name.name = fname_name(fname);
new_name.len = fname_len(fname);
if (f2fs_has_inline_dentry(dir))
err = f2fs_add_inline_entry(dir, &new_name, fname->usr_fname,
inode, ino, mode);
if (err == -EAGAIN)
err = f2fs_add_regular_entry(dir, &new_name, fname->usr_fname,
inode, ino, mode);
f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
return err;
}
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
/*
* Caller should grab and release a rwsem by calling f2fs_lock_op() and
* f2fs_unlock_op().
*/
int __f2fs_add_link(struct inode *dir, const struct qstr *name,
struct inode *inode, nid_t ino, umode_t mode)
{
struct fscrypt_name fname;
struct page *page = NULL;
struct f2fs_dir_entry *de = NULL;
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
int err;
err = fscrypt_setup_filename(dir, name, 0, &fname);
if (err)
return err;
/*
* An immature stakable filesystem shows a race condition between lookup
* and create. If we have same task when doing lookup and create, it's
* definitely fine as expected by VFS normally. Otherwise, let's just
* verify on-disk dentry one more time, which guarantees filesystem
* consistency more.
*/
if (current != F2FS_I(dir)->task) {
de = __f2fs_find_entry(dir, &fname, &page);
F2FS_I(dir)->task = NULL;
}
if (de) {
f2fs_dentry_kunmap(dir, page);
f2fs_put_page(page, 0);
err = -EEXIST;
} else if (IS_ERR(page)) {
err = PTR_ERR(page);
} else {
err = __f2fs_do_add_link(dir, &fname, inode, ino, mode);
}
fscrypt_free_filename(&fname);
return err;
}
int f2fs_do_tmpfile(struct inode *inode, struct inode *dir)
{
struct page *page;
int err = 0;
down_write(&F2FS_I(inode)->i_sem);
page = init_inode_metadata(inode, dir, NULL, NULL, NULL);
if (IS_ERR(page)) {
err = PTR_ERR(page);
goto fail;
}
f2fs_put_page(page, 1);
clear_inode_flag(inode, FI_NEW_INODE);
fail:
up_write(&F2FS_I(inode)->i_sem);
f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
return err;
}
void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
{
struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
down_write(&F2FS_I(inode)->i_sem);
if (S_ISDIR(inode->i_mode))
f2fs_i_links_write(dir, false);
inode->i_ctime = current_time(inode);
f2fs_i_links_write(inode, false);
if (S_ISDIR(inode->i_mode)) {
f2fs_i_links_write(inode, false);
f2fs_i_size_write(inode, 0);
}
up_write(&F2FS_I(inode)->i_sem);
if (inode->i_nlink == 0)
add_orphan_inode(inode);
else
release_orphan_inode(sbi);
}
/*
* It only removes the dentry from the dentry page, corresponding name
* entry in name page does not need to be touched during deletion.
*/
void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
struct inode *dir, struct inode *inode)
{
struct f2fs_dentry_block *dentry_blk;
unsigned int bit_pos;
int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
struct address_space *mapping = page_mapping(page);
unsigned long flags;
int i;
f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
if (f2fs_has_inline_dentry(dir))
return f2fs_delete_inline_entry(dentry, page, dir, inode);
lock_page(page);
f2fs_wait_on_page_writeback(page, DATA, true);
dentry_blk = page_address(page);
bit_pos = dentry - dentry_blk->dentry;
for (i = 0; i < slots; i++)
__clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
/* Let's check and deallocate this dentry page */
bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
NR_DENTRY_IN_BLOCK,
0);
kunmap(page); /* kunmap - pair of f2fs_find_entry */
set_page_dirty(page);
dir->i_ctime = dir->i_mtime = current_time(dir);
f2fs_mark_inode_dirty_sync(dir, false);
if (inode)
f2fs_drop_nlink(dir, inode);
if (bit_pos == NR_DENTRY_IN_BLOCK &&
!truncate_hole(dir, page->index, page->index + 1)) {
spin_lock_irqsave(&mapping->tree_lock, flags);
radix_tree_tag_clear(&mapping->page_tree, page_index(page),
PAGECACHE_TAG_DIRTY);
spin_unlock_irqrestore(&mapping->tree_lock, flags);
clear_page_dirty_for_io(page);
ClearPagePrivate(page);
ClearPageUptodate(page);
inode_dec_dirty_pages(dir);
remove_dirty_inode(dir);
}
f2fs_put_page(page, 1);
}
bool f2fs_empty_dir(struct inode *dir)
{
unsigned long bidx;
struct page *dentry_page;
unsigned int bit_pos;
struct f2fs_dentry_block *dentry_blk;
unsigned long nblock = dir_blocks(dir);
if (f2fs_has_inline_dentry(dir))
return f2fs_empty_inline_dir(dir);
for (bidx = 0; bidx < nblock; bidx++) {
dentry_page = get_lock_data_page(dir, bidx, false);
if (IS_ERR(dentry_page)) {
if (PTR_ERR(dentry_page) == -ENOENT)
continue;
else
return false;
}
dentry_blk = kmap_atomic(dentry_page);
if (bidx == 0)
bit_pos = 2;
else
bit_pos = 0;
bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
NR_DENTRY_IN_BLOCK,
bit_pos);
kunmap_atomic(dentry_blk);
f2fs_put_page(dentry_page, 1);
if (bit_pos < NR_DENTRY_IN_BLOCK)
return false;
}
return true;
}
int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
unsigned int start_pos, struct fscrypt_str *fstr)
{
unsigned char d_type = DT_UNKNOWN;
unsigned int bit_pos;
struct f2fs_dir_entry *de = NULL;
struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
bit_pos = ((unsigned long)ctx->pos % d->max);
while (bit_pos < d->max) {
bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
if (bit_pos >= d->max)
break;
de = &d->dentry[bit_pos];
if (de->name_len == 0) {
bit_pos++;
ctx->pos = start_pos + bit_pos;
continue;
}
f2fs: fix to convert inline directory correctly With below serials, we will lose parts of dirents: 1) mount f2fs with inline_dentry option 2) echo 1 > /sys/fs/f2fs/sdX/dir_level 3) mkdir dir 4) touch 180 files named [1-180] in dir 5) touch 181 in dir 6) echo 3 > /proc/sys/vm/drop_caches 7) ll dir ls: cannot access 2: No such file or directory ls: cannot access 4: No such file or directory ls: cannot access 5: No such file or directory ls: cannot access 6: No such file or directory ls: cannot access 8: No such file or directory ls: cannot access 9: No such file or directory ... total 360 drwxr-xr-x 2 root root 4096 Feb 19 15:12 ./ drwxr-xr-x 3 root root 4096 Feb 19 15:11 ../ -rw-r--r-- 1 root root 0 Feb 19 15:12 1 -rw-r--r-- 1 root root 0 Feb 19 15:12 10 -rw-r--r-- 1 root root 0 Feb 19 15:12 100 -????????? ? ? ? ? ? 101 -????????? ? ? ? ? ? 102 -????????? ? ? ? ? ? 103 ... The reason is: when doing the inline dir conversion, we didn't consider that directory has hierarchical hash structure which can be configured through sysfs interface 'dir_level'. By default, dir_level of directory inode is 0, it means we have one bucket in hash table located in first level, all dirents will be hashed in this bucket, so it has no problem for us to do the duplication simply between inline dentry page and converted normal dentry page. However, if we configured dir_level with the value N (greater than 0), it will expand the bucket number of first level hash table by 2^N - 1, it hashs dirents into different buckets according their hash value, if we still move all dirents to first bucket, it makes incorrent locating for inline dirents, the result is, although we can iterate all dirents through ->readdir, we can't stat some of them in ->lookup which based on hash table searching. This patch fixes this issue by rehashing dirents into correct position when converting inline directory. Signed-off-by: Chao Yu <chao2.yu@samsung.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2016-02-22 18:29:18 +08:00
d_type = get_de_type(de);
de_name.name = d->filename[bit_pos];
de_name.len = le16_to_cpu(de->name_len);
if (f2fs_encrypted_inode(d->inode)) {
int save_len = fstr->len;
int err;
err = fscrypt_fname_disk_to_usr(d->inode,
(u32)de->hash_code, 0,
&de_name, fstr);
if (err)
return err;
de_name = *fstr;
fstr->len = save_len;
}
if (!dir_emit(ctx, de_name.name, de_name.len,
le32_to_cpu(de->ino), d_type))
return 1;
if (sbi->readdir_ra == 1)
ra_node_page(sbi, le32_to_cpu(de->ino));
bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
ctx->pos = start_pos + bit_pos;
}
return 0;
}
static int f2fs_readdir(struct file *file, struct dir_context *ctx)
{
struct inode *inode = file_inode(file);
unsigned long npages = dir_blocks(inode);
struct f2fs_dentry_block *dentry_blk = NULL;
struct page *dentry_page = NULL;
struct file_ra_state *ra = &file->f_ra;
loff_t start_pos = ctx->pos;
unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
struct f2fs_dentry_ptr d;
struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
int err = 0;
if (f2fs_encrypted_inode(inode)) {
err = fscrypt_get_encryption_info(inode);
if (err && err != -ENOKEY)
goto out;
err = fscrypt_fname_alloc_buffer(inode, F2FS_NAME_LEN, &fstr);
if (err < 0)
goto out;
}
if (f2fs_has_inline_dentry(inode)) {
err = f2fs_read_inline_dir(file, ctx, &fstr);
goto out_free;
}
for (; n < npages; n++, ctx->pos = n * NR_DENTRY_IN_BLOCK) {
/* allow readdir() to be interrupted */
if (fatal_signal_pending(current)) {
err = -ERESTARTSYS;
goto out_free;
}
cond_resched();
/* readahead for multi pages of dir */
if (npages - n > 1 && !ra_has_index(ra, n))
page_cache_sync_readahead(inode->i_mapping, ra, file, n,
min(npages - n, (pgoff_t)MAX_DIR_RA_PAGES));
dentry_page = get_lock_data_page(inode, n, false);
if (IS_ERR(dentry_page)) {
err = PTR_ERR(dentry_page);
if (err == -ENOENT) {
err = 0;
continue;
} else {
goto out_free;
}
}
dentry_blk = kmap(dentry_page);
make_dentry_ptr_block(inode, &d, dentry_blk);
err = f2fs_fill_dentries(ctx, &d,
n * NR_DENTRY_IN_BLOCK, &fstr);
if (err) {
kunmap(dentry_page);
f2fs_put_page(dentry_page, 1);
break;
}
kunmap(dentry_page);
f2fs_put_page(dentry_page, 1);
}
out_free:
fscrypt_fname_free_buffer(&fstr);
out:
trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
return err < 0 ? err : 0;
}
static int f2fs_dir_open(struct inode *inode, struct file *filp)
{
if (f2fs_encrypted_inode(inode))
return fscrypt_get_encryption_info(inode) ? -EACCES : 0;
return 0;
}
const struct file_operations f2fs_dir_operations = {
.llseek = generic_file_llseek,
.read = generic_read_dir,
.iterate_shared = f2fs_readdir,
.fsync = f2fs_sync_file,
.open = f2fs_dir_open,
.unlocked_ioctl = f2fs_ioctl,
#ifdef CONFIG_COMPAT
.compat_ioctl = f2fs_compat_ioctl,
#endif
};