2018-09-12 01:16:07 +00:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
2012-11-29 04:28:09 +00:00
|
|
|
/*
|
2012-11-14 07:59:04 +00:00
|
|
|
* fs/f2fs/dir.c
|
|
|
|
*
|
|
|
|
* Copyright (c) 2012 Samsung Electronics Co., Ltd.
|
|
|
|
* http://www.samsung.com/
|
|
|
|
*/
|
2020-11-19 06:09:04 +00:00
|
|
|
#include <asm/unaligned.h>
|
2012-11-14 07:59:04 +00:00
|
|
|
#include <linux/fs.h>
|
|
|
|
#include <linux/f2fs_fs.h>
|
2017-10-13 10:01:34 +00:00
|
|
|
#include <linux/sched/signal.h>
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
#include <linux/unicode.h>
|
2012-11-14 07:59:04 +00:00
|
|
|
#include "f2fs.h"
|
f2fs: fix handling errors got by f2fs_write_inode
Ruslan reported that f2fs hangs with an infinite loop in f2fs_sync_file():
while (sync_node_pages(sbi, inode->i_ino, &wbc) == 0)
f2fs_write_inode(inode, NULL);
The reason was revealed that the cold flag is not set even thought this inode is
a normal file. Therefore, sync_node_pages() skips to write node blocks since it
only writes cold node blocks.
The cold flag is stored to the node_footer in node block, and whenever a new
node page is allocated, it is set according to its file type, file or directory.
But, after sudden-power-off, when recovering the inode page, f2fs doesn't recover
its cold flag.
So, let's assign the cold flag in more right places.
One more thing:
If f2fs_write_inode() returns an error due to whatever situations, there would
be no dirty node pages so that sync_node_pages() returns zero.
(i.e., zero means nothing was written.)
Reported-by: Ruslan N. Marchenko <me@ruff.mobi>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-12-19 06:28:39 +00:00
|
|
|
#include "node.h"
|
2012-11-14 07:59:04 +00:00
|
|
|
#include "acl.h"
|
2013-06-03 10:46:19 +00:00
|
|
|
#include "xattr.h"
|
2017-10-13 10:01:33 +00:00
|
|
|
#include <trace/events/f2fs.h>
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2022-01-18 06:56:14 +00:00
|
|
|
#if IS_ENABLED(CONFIG_UNICODE)
|
2021-06-10 23:46:30 +00:00
|
|
|
extern struct kmem_cache *f2fs_cf_name_slab;
|
|
|
|
#endif
|
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
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 12:29:47 +00:00
|
|
|
return ((unsigned long long) (i_size_read(inode) + PAGE_SIZE - 1))
|
|
|
|
>> PAGE_SHIFT;
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
2014-02-27 09:20:00 +00:00
|
|
|
static unsigned int dir_buckets(unsigned int level, int dir_level)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
2014-05-28 00:56:09 +00:00
|
|
|
if (level + dir_level < MAX_DIR_HASH_DEPTH / 2)
|
2014-02-27 09:20:00 +00:00
|
|
|
return 1 << (level + dir_level);
|
2012-11-14 07:59:04 +00:00
|
|
|
else
|
2014-05-28 00:56:09 +00:00
|
|
|
return MAX_DIR_BUCKETS;
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static unsigned int bucket_blocks(unsigned int level)
|
|
|
|
{
|
|
|
|
if (level < MAX_DIR_HASH_DEPTH / 2)
|
|
|
|
return 2;
|
|
|
|
else
|
|
|
|
return 4;
|
|
|
|
}
|
|
|
|
|
2016-09-18 15:30:03 +00:00
|
|
|
static unsigned char f2fs_filetype_table[F2FS_FT_MAX] = {
|
2012-11-14 07:59:04 +00:00
|
|
|
[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,
|
|
|
|
};
|
|
|
|
|
2018-05-29 16:20:40 +00:00
|
|
|
static void set_de_type(struct f2fs_dir_entry *de, umode_t mode)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
de->file_type = f2fs_type_by_mode[(mode & S_IFMT) >> S_SHIFT];
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
unsigned char f2fs_get_de_type(struct f2fs_dir_entry *de)
|
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 10:29:18 +00:00
|
|
|
{
|
|
|
|
if (de->file_type < F2FS_FT_MAX)
|
|
|
|
return f2fs_filetype_table[de->file_type];
|
|
|
|
return DT_UNKNOWN;
|
|
|
|
}
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
/* If @dir is casefolded, initialize @fname->cf_name from @fname->usr_fname. */
|
|
|
|
int f2fs_init_casefolded_name(const struct inode *dir,
|
|
|
|
struct f2fs_filename *fname)
|
|
|
|
{
|
2022-01-18 06:56:14 +00:00
|
|
|
#if IS_ENABLED(CONFIG_UNICODE)
|
2020-07-08 09:12:36 +00:00
|
|
|
struct super_block *sb = dir->i_sb;
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
|
2022-05-14 17:59:29 +00:00
|
|
|
if (IS_CASEFOLDED(dir) &&
|
|
|
|
!is_dot_dotdot(fname->usr_fname->name, fname->usr_fname->len)) {
|
2021-08-09 00:24:48 +00:00
|
|
|
fname->cf_name.name = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
|
|
|
|
GFP_NOFS, false, F2FS_SB(sb));
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
if (!fname->cf_name.name)
|
|
|
|
return -ENOMEM;
|
2020-07-08 09:12:36 +00:00
|
|
|
fname->cf_name.len = utf8_casefold(sb->s_encoding,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
fname->usr_fname,
|
|
|
|
fname->cf_name.name,
|
|
|
|
F2FS_NAME_LEN);
|
|
|
|
if ((int)fname->cf_name.len <= 0) {
|
2021-06-10 23:46:30 +00:00
|
|
|
kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
fname->cf_name.name = NULL;
|
2020-07-08 09:12:36 +00:00
|
|
|
if (sb_has_strict_encoding(sb))
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
return -EINVAL;
|
|
|
|
/* fall back to treating name as opaque byte sequence */
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __f2fs_setup_filename(const struct inode *dir,
|
|
|
|
const struct fscrypt_name *crypt_name,
|
|
|
|
struct f2fs_filename *fname)
|
|
|
|
{
|
|
|
|
int err;
|
|
|
|
|
|
|
|
memset(fname, 0, sizeof(*fname));
|
|
|
|
|
|
|
|
fname->usr_fname = crypt_name->usr_fname;
|
|
|
|
fname->disk_name = crypt_name->disk_name;
|
|
|
|
#ifdef CONFIG_FS_ENCRYPTION
|
|
|
|
fname->crypto_buf = crypt_name->crypto_buf;
|
|
|
|
#endif
|
2020-09-24 04:26:23 +00:00
|
|
|
if (crypt_name->is_nokey_name) {
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
/* hash was decoded from the no-key name */
|
|
|
|
fname->hash = cpu_to_le32(crypt_name->hash);
|
|
|
|
} else {
|
|
|
|
err = f2fs_init_casefolded_name(dir, fname);
|
|
|
|
if (err) {
|
|
|
|
f2fs_free_filename(fname);
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
f2fs_hash_filename(dir, fname);
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Prepare to search for @iname in @dir. This is similar to
|
|
|
|
* fscrypt_setup_filename(), but this also handles computing the casefolded name
|
|
|
|
* and the f2fs dirhash if needed, then packing all the information about this
|
|
|
|
* filename up into a 'struct f2fs_filename'.
|
|
|
|
*/
|
|
|
|
int f2fs_setup_filename(struct inode *dir, const struct qstr *iname,
|
|
|
|
int lookup, struct f2fs_filename *fname)
|
|
|
|
{
|
|
|
|
struct fscrypt_name crypt_name;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = fscrypt_setup_filename(dir, iname, lookup, &crypt_name);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
return __f2fs_setup_filename(dir, &crypt_name, fname);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Prepare to look up @dentry in @dir. This is similar to
|
|
|
|
* fscrypt_prepare_lookup(), but this also handles computing the casefolded name
|
|
|
|
* and the f2fs dirhash if needed, then packing all the information about this
|
|
|
|
* filename up into a 'struct f2fs_filename'.
|
|
|
|
*/
|
|
|
|
int f2fs_prepare_lookup(struct inode *dir, struct dentry *dentry,
|
|
|
|
struct f2fs_filename *fname)
|
|
|
|
{
|
|
|
|
struct fscrypt_name crypt_name;
|
|
|
|
int err;
|
|
|
|
|
|
|
|
err = fscrypt_prepare_lookup(dir, dentry, &crypt_name);
|
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
|
|
|
return __f2fs_setup_filename(dir, &crypt_name, fname);
|
|
|
|
}
|
|
|
|
|
|
|
|
void f2fs_free_filename(struct f2fs_filename *fname)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_FS_ENCRYPTION
|
|
|
|
kfree(fname->crypto_buf.name);
|
|
|
|
fname->crypto_buf.name = NULL;
|
|
|
|
#endif
|
2022-01-18 06:56:14 +00:00
|
|
|
#if IS_ENABLED(CONFIG_UNICODE)
|
2021-06-10 23:46:30 +00:00
|
|
|
if (fname->cf_name.name) {
|
|
|
|
kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
|
|
|
|
fname->cf_name.name = NULL;
|
|
|
|
}
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2014-02-27 09:20:00 +00:00
|
|
|
static unsigned long dir_block_index(unsigned int level,
|
|
|
|
int dir_level, unsigned int idx)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
unsigned long i;
|
|
|
|
unsigned long bidx = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < level; i++)
|
2014-02-27 09:20:00 +00:00
|
|
|
bidx += dir_buckets(i, dir_level) * bucket_blocks(i);
|
2012-11-14 07:59:04 +00:00
|
|
|
bidx += idx * bucket_blocks(level);
|
|
|
|
return bidx;
|
|
|
|
}
|
|
|
|
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
static struct f2fs_dir_entry *find_in_block(struct inode *dir,
|
|
|
|
struct page *dentry_page,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct f2fs_filename *fname,
|
2020-09-26 00:20:46 +00:00
|
|
|
int *max_slots)
|
2014-10-14 00:26:14 +00:00
|
|
|
{
|
|
|
|
struct f2fs_dentry_block *dentry_blk;
|
2014-10-19 05:52:52 +00:00
|
|
|
struct f2fs_dentry_ptr d;
|
2014-10-14 00:26:14 +00:00
|
|
|
|
2018-02-28 12:31:52 +00:00
|
|
|
dentry_blk = (struct f2fs_dentry_block *)page_address(dentry_page);
|
2014-10-19 05:52:52 +00:00
|
|
|
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
make_dentry_ptr_block(dir, &d, dentry_blk);
|
2020-09-26 00:20:46 +00:00
|
|
|
return f2fs_find_target_dentry(&d, fname, max_slots);
|
2014-10-14 00:26:14 +00:00
|
|
|
}
|
|
|
|
|
2022-01-18 06:56:14 +00:00
|
|
|
#if IS_ENABLED(CONFIG_UNICODE)
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
/*
|
|
|
|
* Test whether a case-insensitive directory entry matches the filename
|
|
|
|
* being searched for.
|
2020-11-19 06:09:04 +00:00
|
|
|
*
|
|
|
|
* Returns 1 for a match, 0 for no match, and -errno on an error.
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
*/
|
2020-11-19 06:09:04 +00:00
|
|
|
static int f2fs_match_ci_name(const struct inode *dir, const struct qstr *name,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const u8 *de_name, u32 de_name_len)
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
{
|
2020-07-08 09:12:36 +00:00
|
|
|
const struct super_block *sb = dir->i_sb;
|
|
|
|
const struct unicode_map *um = sb->s_encoding;
|
2020-11-19 06:09:04 +00:00
|
|
|
struct fscrypt_str decrypted_name = FSTR_INIT(NULL, de_name_len);
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct qstr entry = QSTR_INIT(de_name, de_name_len);
|
2020-05-07 07:59:03 +00:00
|
|
|
int res;
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
|
2020-11-19 06:09:04 +00:00
|
|
|
if (IS_ENCRYPTED(dir)) {
|
|
|
|
const struct fscrypt_str encrypted_name =
|
|
|
|
FSTR_INIT((u8 *)de_name, de_name_len);
|
|
|
|
|
|
|
|
if (WARN_ON_ONCE(!fscrypt_has_encryption_key(dir)))
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
decrypted_name.name = kmalloc(de_name_len, GFP_KERNEL);
|
|
|
|
if (!decrypted_name.name)
|
|
|
|
return -ENOMEM;
|
|
|
|
res = fscrypt_fname_disk_to_usr(dir, 0, 0, &encrypted_name,
|
|
|
|
&decrypted_name);
|
|
|
|
if (res < 0)
|
|
|
|
goto out;
|
|
|
|
entry.name = decrypted_name.name;
|
|
|
|
entry.len = decrypted_name.len;
|
|
|
|
}
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
res = utf8_strncasecmp_folded(um, name, &entry);
|
2020-11-19 06:09:04 +00:00
|
|
|
/*
|
|
|
|
* In strict mode, ignore invalid names. In non-strict mode,
|
|
|
|
* fall back to treating them as opaque byte sequences.
|
|
|
|
*/
|
|
|
|
if (res < 0 && !sb_has_strict_encoding(sb)) {
|
|
|
|
res = name->len == entry.len &&
|
|
|
|
memcmp(name->name, entry.name, name->len) == 0;
|
|
|
|
} else {
|
|
|
|
/* utf8_strncasecmp_folded returns 0 on match */
|
|
|
|
res = (res == 0);
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
}
|
2020-11-19 06:09:04 +00:00
|
|
|
out:
|
|
|
|
kfree(decrypted_name.name);
|
|
|
|
return res;
|
f2fs: Support case-insensitive file name lookups
Modeled after commit b886ee3e778e ("ext4: Support case-insensitive file
name lookups")
"""
This patch implements the actual support for case-insensitive file name
lookups in f2fs, based on the feature bit and the encoding stored in the
superblock.
A filesystem that has the casefold feature set is able to configure
directories with the +F (F2FS_CASEFOLD_FL) attribute, enabling lookups
to succeed in that directory in a case-insensitive fashion, i.e: match
a directory entry even if the name used by userspace is not a byte per
byte match with the disk name, but is an equivalent case-insensitive
version of the Unicode string. This operation is called a
case-insensitive file name lookup.
The feature is configured as an inode attribute applied to directories
and inherited by its children. This attribute can only be enabled on
empty directories for filesystems that support the encoding feature,
thus preventing collision of file names that only differ by case.
* dcache handling:
For a +F directory, F2Fs only stores the first equivalent name dentry
used in the dcache. This is done to prevent unintentional duplication of
dentries in the dcache, while also allowing the VFS code to quickly find
the right entry in the cache despite which equivalent string was used in
a previous lookup, without having to resort to ->lookup().
d_hash() of casefolded directories is implemented as the hash of the
casefolded string, such that we always have a well-known bucket for all
the equivalencies of the same string. d_compare() uses the
utf8_strncasecmp() infrastructure, which handles the comparison of
equivalent, same case, names as well.
For now, negative lookups are not inserted in the dcache, since they
would need to be invalidated anyway, because we can't trust missing file
dentries. This is bad for performance but requires some leveraging of
the vfs layer to fix. We can live without that for now, and so does
everyone else.
* on-disk data:
Despite using a specific version of the name as the internal
representation within the dcache, the name stored and fetched from the
disk is a byte-per-byte match with what the user requested, making this
implementation 'name-preserving'. i.e. no actual information is lost
when writing to storage.
DX is supported by modifying the hashes used in +F directories to make
them case/encoding-aware. The new disk hashes are calculated as the
hash of the full casefolded string, instead of the string directly.
This allows us to efficiently search for file names in the htree without
requiring the user to provide an exact name.
* Dealing with invalid sequences:
By default, when a invalid UTF-8 sequence is identified, ext4 will treat
it as an opaque byte sequence, ignoring the encoding and reverting to
the old behavior for that unique file. This means that case-insensitive
file name lookup will not work only for that file. An optional bit can
be set in the superblock telling the filesystem code and userspace tools
to enforce the encoding. When that optional bit is set, any attempt to
create a file name using an invalid UTF-8 sequence will fail and return
an error to userspace.
* Normalization algorithm:
The UTF-8 algorithms used to compare strings in f2fs is implemented
in fs/unicode, and is based on a previous version developed by
SGI. It implements the Canonical decomposition (NFD) algorithm
described by the Unicode specification 12.1, or higher, combined with
the elimination of ignorable code points (NFDi) and full
case-folding (CF) as documented in fs/unicode/utf8_norm.c.
NFD seems to be the best normalization method for F2FS because:
- It has a lower cost than NFC/NFKC (which requires
decomposing to NFD as an intermediary step)
- It doesn't eliminate important semantic meaning like
compatibility decompositions.
Although:
- This implementation is not completely linguistic accurate, because
different languages have conflicting rules, which would require the
specialization of the filesystem to a given locale, which brings all
sorts of problems for removable media and for users who use more than
one language.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2019-07-23 23:05:29 +00:00
|
|
|
}
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
#endif /* CONFIG_UNICODE */
|
2019-08-21 15:13:35 +00:00
|
|
|
|
2020-11-19 06:09:04 +00:00
|
|
|
static inline int f2fs_match_name(const struct inode *dir,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct f2fs_filename *fname,
|
|
|
|
const u8 *de_name, u32 de_name_len)
|
2019-08-21 15:13:35 +00:00
|
|
|
{
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct fscrypt_name f;
|
2019-08-21 15:13:35 +00:00
|
|
|
|
2022-01-18 06:56:14 +00:00
|
|
|
#if IS_ENABLED(CONFIG_UNICODE)
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
if (fname->cf_name.name) {
|
|
|
|
struct qstr cf = FSTR_TO_QSTR(&fname->cf_name);
|
2019-08-21 15:13:34 +00:00
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
return f2fs_match_ci_name(dir, &cf, de_name, de_name_len);
|
2019-08-21 15:13:35 +00:00
|
|
|
}
|
2019-08-21 15:13:34 +00:00
|
|
|
#endif
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
f.usr_fname = fname->usr_fname;
|
|
|
|
f.disk_name = fname->disk_name;
|
|
|
|
#ifdef CONFIG_FS_ENCRYPTION
|
|
|
|
f.crypto_buf = fname->crypto_buf;
|
|
|
|
#endif
|
|
|
|
return fscrypt_match_name(&f, de_name, de_name_len);
|
2019-08-21 15:13:34 +00:00
|
|
|
}
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct f2fs_dir_entry *f2fs_find_target_dentry(const struct f2fs_dentry_ptr *d,
|
|
|
|
const struct f2fs_filename *fname, int *max_slots)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
struct f2fs_dir_entry *de;
|
2014-02-27 04:57:53 +00:00
|
|
|
unsigned long bit_pos = 0;
|
|
|
|
int max_len = 0;
|
2020-11-19 06:09:04 +00:00
|
|
|
int res = 0;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2014-10-19 05:52:52 +00:00
|
|
|
if (max_slots)
|
|
|
|
*max_slots = 0;
|
|
|
|
while (bit_pos < d->max) {
|
|
|
|
if (!test_bit_le(bit_pos, d->bitmap)) {
|
2014-02-27 04:57:53 +00:00
|
|
|
bit_pos++;
|
2015-03-09 09:33:16 +00:00
|
|
|
max_len++;
|
2014-02-27 04:57:53 +00:00
|
|
|
continue;
|
|
|
|
}
|
2015-03-09 09:33:16 +00:00
|
|
|
|
2014-10-19 05:52:52 +00:00
|
|
|
de = &d->dentry[bit_pos];
|
2015-04-28 00:12:39 +00:00
|
|
|
|
2016-04-27 14:22:20 +00:00
|
|
|
if (unlikely(!de->name_len)) {
|
|
|
|
bit_pos++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
2020-11-19 06:09:04 +00:00
|
|
|
if (de->hash_code == fname->hash) {
|
|
|
|
res = f2fs_match_name(d->inode, fname,
|
|
|
|
d->filename[bit_pos],
|
|
|
|
le16_to_cpu(de->name_len));
|
|
|
|
if (res < 0)
|
|
|
|
return ERR_PTR(res);
|
|
|
|
if (res)
|
|
|
|
goto found;
|
|
|
|
}
|
2017-04-24 17:00:12 +00:00
|
|
|
|
2015-03-09 09:33:16 +00:00
|
|
|
if (max_slots && max_len > *max_slots)
|
2014-02-27 04:57:53 +00:00
|
|
|
*max_slots = max_len;
|
2015-03-09 09:33:16 +00:00
|
|
|
max_len = 0;
|
2014-07-10 04:37:46 +00:00
|
|
|
|
2014-02-27 04:57:53 +00:00
|
|
|
bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
de = NULL;
|
|
|
|
found:
|
2014-10-19 05:52:52 +00:00
|
|
|
if (max_slots && max_len > *max_slots)
|
2014-02-27 04:57:53 +00:00
|
|
|
*max_slots = max_len;
|
2012-11-14 07:59:04 +00:00
|
|
|
return de;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct f2fs_dir_entry *find_in_level(struct inode *dir,
|
2015-04-28 00:12:39 +00:00
|
|
|
unsigned int level,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct f2fs_filename *fname,
|
2015-04-28 00:12:39 +00:00
|
|
|
struct page **res_page)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
int s = GET_DENTRY_SLOTS(fname->disk_name.len);
|
2012-11-14 07:59:04 +00:00
|
|
|
unsigned int nbucket, nblock;
|
|
|
|
unsigned int bidx, end_block;
|
|
|
|
struct page *dentry_page;
|
|
|
|
struct f2fs_dir_entry *de = NULL;
|
|
|
|
bool room = false;
|
2014-10-14 00:26:14 +00:00
|
|
|
int max_slots;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2014-02-27 09:20:00 +00:00
|
|
|
nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
|
2012-11-14 07:59:04 +00:00
|
|
|
nblock = bucket_blocks(level);
|
|
|
|
|
2014-02-27 09:20:00 +00:00
|
|
|
bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
le32_to_cpu(fname->hash) % nbucket);
|
2012-11-14 07:59:04 +00:00
|
|
|
end_block = bidx + nblock;
|
|
|
|
|
|
|
|
for (; bidx < end_block; bidx++) {
|
|
|
|
/* no need to allocate new dentry pages to all the indices */
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
dentry_page = f2fs_find_data_page(dir, bidx);
|
2012-11-14 07:59:04 +00:00
|
|
|
if (IS_ERR(dentry_page)) {
|
2016-05-25 21:29:11 +00:00
|
|
|
if (PTR_ERR(dentry_page) == -ENOENT) {
|
|
|
|
room = true;
|
|
|
|
continue;
|
|
|
|
} else {
|
|
|
|
*res_page = dentry_page;
|
|
|
|
break;
|
|
|
|
}
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
2020-09-26 00:20:46 +00:00
|
|
|
de = find_in_block(dir, dentry_page, fname, &max_slots);
|
2020-11-19 06:09:04 +00:00
|
|
|
if (IS_ERR(de)) {
|
|
|
|
*res_page = ERR_CAST(de);
|
|
|
|
de = NULL;
|
|
|
|
break;
|
|
|
|
} else if (de) {
|
2020-09-26 00:20:46 +00:00
|
|
|
*res_page = dentry_page;
|
2012-11-14 07:59:04 +00:00
|
|
|
break;
|
2020-09-26 00:20:46 +00:00
|
|
|
}
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
if (max_slots >= s)
|
|
|
|
room = true;
|
|
|
|
f2fs_put_page(dentry_page, 0);
|
|
|
|
}
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
if (!de && room && F2FS_I(dir)->chash != fname->hash) {
|
|
|
|
F2FS_I(dir)->chash = fname->hash;
|
2017-04-22 02:39:20 +00:00
|
|
|
F2FS_I(dir)->clevel = level;
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return de;
|
|
|
|
}
|
|
|
|
|
2016-08-29 03:27:56 +00:00
|
|
|
struct f2fs_dir_entry *__f2fs_find_entry(struct inode *dir,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct f2fs_filename *fname,
|
|
|
|
struct page **res_page)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
unsigned long npages = dir_blocks(dir);
|
|
|
|
struct f2fs_dir_entry *de = NULL;
|
|
|
|
unsigned int max_depth;
|
|
|
|
unsigned int level;
|
2015-04-28 00:12:39 +00:00
|
|
|
|
2020-09-29 01:22:50 +00:00
|
|
|
*res_page = NULL;
|
|
|
|
|
2015-04-28 00:12:39 +00:00
|
|
|
if (f2fs_has_inline_dentry(dir)) {
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
de = f2fs_find_in_inline_dir(dir, fname, res_page);
|
2015-04-28 00:12:39 +00:00
|
|
|
goto out;
|
|
|
|
}
|
2014-09-24 10:19:10 +00:00
|
|
|
|
2020-09-29 01:22:50 +00:00
|
|
|
if (npages == 0)
|
2015-04-28 00:12:39 +00:00
|
|
|
goto out;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
max_depth = F2FS_I(dir)->i_current_depth;
|
2015-12-31 18:28:52 +00:00
|
|
|
if (unlikely(max_depth > MAX_DIR_HASH_DEPTH)) {
|
2019-06-18 09:48:42 +00:00
|
|
|
f2fs_warn(F2FS_I_SB(dir), "Corrupted max_depth of %lu: %u",
|
|
|
|
dir->i_ino, max_depth);
|
2015-12-31 18:28:52 +00:00
|
|
|
max_depth = MAX_DIR_HASH_DEPTH;
|
2016-05-20 16:52:20 +00:00
|
|
|
f2fs_i_depth_write(dir, max_depth);
|
2015-12-31 18:28:52 +00:00
|
|
|
}
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
for (level = 0; level < max_depth; level++) {
|
2016-08-29 03:27:56 +00:00
|
|
|
de = find_in_level(dir, level, fname, res_page);
|
2016-05-25 21:29:11 +00:00
|
|
|
if (de || IS_ERR(*res_page))
|
2012-11-14 07:59:04 +00:00
|
|
|
break;
|
|
|
|
}
|
2015-04-28 00:12:39 +00:00
|
|
|
out:
|
2017-04-22 02:39:20 +00:00
|
|
|
/* This is to increase the speed of f2fs_create */
|
|
|
|
if (!de)
|
|
|
|
F2FS_I(dir)->task = current;
|
2016-08-29 03:27:56 +00:00
|
|
|
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;
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct f2fs_filename fname;
|
2016-08-29 03:27:56 +00:00
|
|
|
int err;
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
err = f2fs_setup_filename(dir, child, 1, &fname);
|
2016-08-29 03:27:56 +00:00
|
|
|
if (err) {
|
2016-12-05 19:12:44 +00:00
|
|
|
if (err == -ENOENT)
|
|
|
|
*res_page = NULL;
|
|
|
|
else
|
|
|
|
*res_page = ERR_PTR(err);
|
2016-08-29 03:27:56 +00:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
de = __f2fs_find_entry(dir, &fname, res_page);
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
f2fs_free_filename(&fname);
|
2012-11-14 07:59:04 +00:00
|
|
|
return de;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct f2fs_dir_entry *f2fs_parent_dir(struct inode *dir, struct page **p)
|
|
|
|
{
|
2021-04-15 23:46:50 +00:00
|
|
|
return f2fs_find_entry(dir, &dotdot_name, p);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
2016-08-06 13:49:02 +00:00
|
|
|
ino_t f2fs_inode_by_name(struct inode *dir, const struct qstr *qstr,
|
2016-07-19 00:27:47 +00:00
|
|
|
struct page **page)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
ino_t res = 0;
|
|
|
|
struct f2fs_dir_entry *de;
|
|
|
|
|
2016-07-19 00:27:47 +00:00
|
|
|
de = f2fs_find_entry(dir, qstr, page);
|
2012-11-14 07:59:04 +00:00
|
|
|
if (de) {
|
|
|
|
res = le32_to_cpu(de->ino);
|
2016-07-19 00:27:47 +00:00
|
|
|
f2fs_put_page(*page, 0);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
return res;
|
|
|
|
}
|
|
|
|
|
|
|
|
void f2fs_set_link(struct inode *dir, struct f2fs_dir_entry *de,
|
|
|
|
struct page *page, struct inode *inode)
|
|
|
|
{
|
2014-10-14 02:34:26 +00:00
|
|
|
enum page_type type = f2fs_has_inline_dentry(dir) ? NODE : DATA;
|
2021-04-06 01:47:35 +00:00
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
lock_page(page);
|
2018-12-25 09:43:42 +00:00
|
|
|
f2fs_wait_on_page_writeback(page, type, true, true);
|
2012-11-14 07:59:04 +00:00
|
|
|
de->ino = cpu_to_le32(inode->i_ino);
|
2015-03-30 22:07:16 +00:00
|
|
|
set_de_type(de, inode->i_mode);
|
2012-11-14 07:59:04 +00:00
|
|
|
set_page_dirty(page);
|
f2fs: fix tracking parent inode number
Previously, f2fs didn't track the parent inode number correctly which is stored
in each f2fs_inode. In the case of the following scenario, a bug can be occured.
Let's suppose there are one directory, "/b", and two files, "/a" and "/b/a".
- pino of "/a" is ROOT_INO.
- pino of "/b/a" is DIR_B_INO.
Then,
# sync
: The inode pages of "/a" and "/b/a" contain the parent inode numbers as
ROOT_INO and DIR_B_INO respectively.
# mv /a /b/a
: The parent inode number of "/a" should be changed to DIR_B_INO, but f2fs
didn't do that. Ref. f2fs_set_link().
In order to fix this clearly, I added i_pino in f2fs_inode_info, and whenever
it needs to be changed like in f2fs_add_link() and f2fs_set_link(), it is
updated temporarily in f2fs_inode_info.
And later, f2fs_write_inode() stores the latest information to the inode pages.
For power-off-recovery, f2fs_sync_file() triggers simply f2fs_write_inode().
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-12-10 08:52:48 +00:00
|
|
|
|
2016-09-14 14:48:04 +00:00
|
|
|
dir->i_mtime = dir->i_ctime = current_time(dir);
|
2016-10-14 18:51:23 +00:00
|
|
|
f2fs_mark_inode_dirty_sync(dir, false);
|
2012-11-14 07:59:04 +00:00
|
|
|
f2fs_put_page(page, 1);
|
|
|
|
}
|
|
|
|
|
2020-11-19 06:09:04 +00:00
|
|
|
static void init_dent_inode(struct inode *dir, struct inode *inode,
|
|
|
|
const struct f2fs_filename *fname,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct page *ipage)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
2013-12-26 07:30:41 +00:00
|
|
|
struct f2fs_inode *ri;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2020-11-19 06:09:04 +00:00
|
|
|
if (!fname) /* tmpfile case? */
|
|
|
|
return;
|
|
|
|
|
2018-12-25 09:43:42 +00:00
|
|
|
f2fs_wait_on_page_writeback(ipage, NODE, true, true);
|
2014-04-29 08:28:32 +00:00
|
|
|
|
2013-01-25 21:01:21 +00:00
|
|
|
/* copy name info. to this inode page */
|
2013-12-26 07:30:41 +00:00
|
|
|
ri = F2FS_INODE(ipage);
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
ri->i_namelen = cpu_to_le32(fname->disk_name.len);
|
|
|
|
memcpy(ri->i_name, fname->disk_name.name, fname->disk_name.len);
|
2020-11-19 06:09:04 +00:00
|
|
|
if (IS_ENCRYPTED(dir)) {
|
|
|
|
file_set_enc_name(inode);
|
|
|
|
/*
|
|
|
|
* Roll-forward recovery doesn't have encryption keys available,
|
|
|
|
* so it can't compute the dirhash for encrypted+casefolded
|
|
|
|
* filenames. Append it to i_name if possible. Else, disable
|
|
|
|
* roll-forward recovery of the dentry (i.e., make fsync'ing the
|
|
|
|
* file force a checkpoint) by setting LOST_PINO.
|
|
|
|
*/
|
|
|
|
if (IS_CASEFOLDED(dir)) {
|
|
|
|
if (fname->disk_name.len + sizeof(f2fs_hash_t) <=
|
|
|
|
F2FS_NAME_LEN)
|
|
|
|
put_unaligned(fname->hash, (f2fs_hash_t *)
|
|
|
|
&ri->i_name[fname->disk_name.len]);
|
|
|
|
else
|
|
|
|
file_lost_pino(inode);
|
|
|
|
}
|
|
|
|
}
|
2012-11-14 07:59:04 +00:00
|
|
|
set_page_dirty(ipage);
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
void f2fs_do_make_empty_dir(struct inode *inode, struct inode *parent,
|
2014-10-19 06:06:41 +00:00
|
|
|
struct f2fs_dentry_ptr *d)
|
|
|
|
{
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct fscrypt_str dot = FSTR_INIT(".", 1);
|
|
|
|
struct fscrypt_str dotdot = FSTR_INIT("..", 2);
|
2014-10-19 06:06:41 +00:00
|
|
|
|
2016-03-09 14:07:28 +00:00
|
|
|
/* update dirent of "." */
|
|
|
|
f2fs_update_dentry(inode->i_ino, inode->i_mode, d, &dot, 0, 0);
|
2014-10-19 06:06:41 +00:00
|
|
|
|
2016-03-09 14:07:28 +00:00
|
|
|
/* update dirent of ".." */
|
|
|
|
f2fs_update_dentry(parent->i_ino, parent->i_mode, d, &dotdot, 0, 1);
|
2014-10-19 06:06:41 +00:00
|
|
|
}
|
|
|
|
|
2013-05-20 01:10:29 +00:00
|
|
|
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 07:21:29 +00:00
|
|
|
{
|
|
|
|
struct page *dentry_page;
|
|
|
|
struct f2fs_dentry_block *dentry_blk;
|
2014-10-19 06:06:41 +00:00
|
|
|
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 07:21:29 +00:00
|
|
|
|
2014-09-24 10:19:10 +00:00
|
|
|
if (f2fs_has_inline_dentry(inode))
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
return f2fs_make_empty_inline_dir(inode, parent, page);
|
2014-09-24 10:19:10 +00:00
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
dentry_page = f2fs_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 07:21:29 +00:00
|
|
|
if (IS_ERR(dentry_page))
|
|
|
|
return PTR_ERR(dentry_page);
|
|
|
|
|
2018-02-28 12:31:52 +00:00
|
|
|
dentry_blk = page_address(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 07:21:29 +00:00
|
|
|
|
2017-04-04 10:01:22 +00:00
|
|
|
make_dentry_ptr_block(NULL, &d, dentry_blk);
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_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 07:21:29 +00:00
|
|
|
|
|
|
|
set_page_dirty(dentry_page);
|
|
|
|
f2fs_put_page(dentry_page, 1);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
struct page *f2fs_init_inode_metadata(struct inode *inode, struct inode *dir,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct f2fs_filename *fname, struct page *dpage)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
2013-05-20 01:10:29 +00:00
|
|
|
struct page *page;
|
|
|
|
int err;
|
|
|
|
|
2016-05-20 17:13:22 +00:00
|
|
|
if (is_inode_flag_set(inode, FI_NEW_INODE)) {
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
page = f2fs_new_inode_page(inode);
|
2013-05-20 01:10:29 +00:00
|
|
|
if (IS_ERR(page))
|
|
|
|
return page;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
if (S_ISDIR(inode->i_mode)) {
|
2016-05-02 19:34:48 +00:00
|
|
|
/* in order to handle error case */
|
|
|
|
get_page(page);
|
2013-05-20 01:10:29 +00:00
|
|
|
err = make_empty_dir(inode, dir, page);
|
2016-05-02 19:34:48 +00:00
|
|
|
if (err) {
|
|
|
|
lock_page(page);
|
|
|
|
goto put_error;
|
|
|
|
}
|
|
|
|
put_page(page);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
2014-10-14 02:42:53 +00:00
|
|
|
err = f2fs_init_acl(inode, dir, page, dpage);
|
2013-05-20 01:10:29 +00:00
|
|
|
if (err)
|
2013-12-27 08:04:17 +00:00
|
|
|
goto put_error;
|
2013-05-20 01:10:29 +00:00
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
err = f2fs_init_security(inode, dir,
|
|
|
|
fname ? fname->usr_fname : NULL, page);
|
2013-06-03 10:46:19 +00:00
|
|
|
if (err)
|
2013-12-27 08:04:17 +00:00
|
|
|
goto put_error;
|
2015-04-22 03:39:58 +00:00
|
|
|
|
2020-03-21 12:19:33 +00:00
|
|
|
if (IS_ENCRYPTED(inode)) {
|
2020-09-17 04:11:27 +00:00
|
|
|
err = fscrypt_set_context(inode, page);
|
2015-04-22 03:39:58 +00:00
|
|
|
if (err)
|
|
|
|
goto put_error;
|
|
|
|
}
|
2012-11-14 07:59:04 +00:00
|
|
|
} else {
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
page = f2fs_get_node_page(F2FS_I_SB(dir), inode->i_ino);
|
2013-05-20 01:10:29 +00:00
|
|
|
if (IS_ERR(page))
|
|
|
|
return page;
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
2013-05-20 01:10:29 +00:00
|
|
|
|
2020-11-19 06:09:04 +00:00
|
|
|
init_dent_inode(dir, inode, fname, page);
|
2013-05-20 01:10:29 +00:00
|
|
|
|
2013-05-28 03:25:47 +00:00
|
|
|
/*
|
|
|
|
* This file should be checkpointed during fsync.
|
|
|
|
* We lost i_pino from now on.
|
|
|
|
*/
|
2016-05-20 17:13:22 +00:00
|
|
|
if (is_inode_flag_set(inode, FI_INC_LINK)) {
|
2017-06-26 02:41:35 +00:00
|
|
|
if (!S_ISDIR(inode->i_mode))
|
|
|
|
file_lost_pino(inode);
|
2014-06-19 08:23:19 +00:00
|
|
|
/*
|
|
|
|
* If link the tmpfile to alias through linkat path,
|
|
|
|
* we should remove this inode from orphan list.
|
|
|
|
*/
|
|
|
|
if (inode->i_nlink == 0)
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_remove_orphan_inode(F2FS_I_SB(dir), inode->i_ino);
|
2016-05-20 16:43:20 +00:00
|
|
|
f2fs_i_links_write(inode, true);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
2013-05-20 01:10:29 +00:00
|
|
|
return page;
|
|
|
|
|
2013-12-27 08:04:17 +00:00
|
|
|
put_error:
|
2016-05-02 19:34:48 +00:00
|
|
|
clear_nlink(inode);
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_update_inode(inode, page);
|
2016-05-02 19:34:48 +00:00
|
|
|
f2fs_put_page(page, 1);
|
2013-05-20 01:10:29 +00:00
|
|
|
return ERR_PTR(err);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
void f2fs_update_parent_metadata(struct inode *dir, struct inode *inode,
|
2012-11-14 07:59:04 +00:00
|
|
|
unsigned int current_depth)
|
|
|
|
{
|
2016-05-20 17:13:22 +00:00
|
|
|
if (inode && is_inode_flag_set(inode, FI_NEW_INODE)) {
|
2016-05-20 23:32:49 +00:00
|
|
|
if (S_ISDIR(inode->i_mode))
|
2016-05-20 16:43:20 +00:00
|
|
|
f2fs_i_links_write(dir, true);
|
2016-05-20 17:13:22 +00:00
|
|
|
clear_inode_flag(inode, FI_NEW_INODE);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
2016-09-14 14:48:04 +00:00
|
|
|
dir->i_mtime = dir->i_ctime = current_time(dir);
|
2016-10-14 18:51:23 +00:00
|
|
|
f2fs_mark_inode_dirty_sync(dir, false);
|
2014-01-21 04:32:12 +00:00
|
|
|
|
2016-05-20 23:32:49 +00:00
|
|
|
if (F2FS_I(dir)->i_current_depth != current_depth)
|
2016-05-20 16:52:20 +00:00
|
|
|
f2fs_i_depth_write(dir, current_depth);
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2016-05-20 17:13:22 +00:00
|
|
|
if (inode && is_inode_flag_set(inode, FI_INC_LINK))
|
|
|
|
clear_inode_flag(inode, FI_INC_LINK);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
int f2fs_room_for_filename(const void *bitmap, int slots, int max_slots)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
int bit_start = 0;
|
|
|
|
int zero_start, zero_end;
|
|
|
|
next:
|
2014-10-13 23:28:13 +00:00
|
|
|
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);
|
2012-11-14 07:59:04 +00:00
|
|
|
if (zero_end - zero_start >= slots)
|
|
|
|
return zero_start;
|
|
|
|
|
|
|
|
bit_start = zero_end + 1;
|
|
|
|
|
2014-10-13 23:28:13 +00:00
|
|
|
if (zero_end + 1 >= max_slots)
|
|
|
|
return max_slots;
|
2012-11-14 07:59:04 +00:00
|
|
|
goto next;
|
|
|
|
}
|
|
|
|
|
2019-12-10 03:03:05 +00:00
|
|
|
bool f2fs_has_enough_room(struct inode *dir, struct page *ipage,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct f2fs_filename *fname)
|
2019-12-10 03:03:05 +00:00
|
|
|
{
|
|
|
|
struct f2fs_dentry_ptr d;
|
|
|
|
unsigned int bit_pos;
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
int slots = GET_DENTRY_SLOTS(fname->disk_name.len);
|
2019-12-10 03:03:05 +00:00
|
|
|
|
|
|
|
make_dentry_ptr_inline(dir, &d, inline_data_addr(dir, ipage));
|
|
|
|
|
|
|
|
bit_pos = f2fs_room_for_filename(d.bitmap, slots, d.max);
|
|
|
|
|
|
|
|
return bit_pos < d.max;
|
|
|
|
}
|
|
|
|
|
2015-03-30 22:07:16 +00:00
|
|
|
void f2fs_update_dentry(nid_t ino, umode_t mode, struct f2fs_dentry_ptr *d,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
const struct fscrypt_str *name, f2fs_hash_t name_hash,
|
|
|
|
unsigned int bit_pos)
|
2015-02-16 08:17:20 +00:00
|
|
|
{
|
|
|
|
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);
|
2015-03-30 22:07:16 +00:00
|
|
|
de->ino = cpu_to_le32(ino);
|
|
|
|
set_de_type(de, mode);
|
2016-02-12 22:29:28 +00:00
|
|
|
for (i = 0; i < slots; i++) {
|
2016-08-31 23:20:37 +00:00
|
|
|
__set_bit_le(bit_pos + i, (void *)d->bitmap);
|
2016-02-12 22:29:28 +00:00
|
|
|
/* avoid wrong garbage data for readdir */
|
|
|
|
if (i)
|
|
|
|
(de + i)->name_len = 0;
|
|
|
|
}
|
2015-02-16 08:17:20 +00:00
|
|
|
}
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
int f2fs_add_regular_entry(struct inode *dir, const struct f2fs_filename *fname,
|
|
|
|
struct inode *inode, nid_t ino, umode_t mode)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
unsigned int bit_pos;
|
|
|
|
unsigned int level;
|
|
|
|
unsigned int current_depth;
|
|
|
|
unsigned long bidx, block;
|
|
|
|
unsigned int nbucket, nblock;
|
|
|
|
struct page *dentry_page = NULL;
|
|
|
|
struct f2fs_dentry_block *dentry_blk = NULL;
|
2015-02-16 08:17:20 +00:00
|
|
|
struct f2fs_dentry_ptr d;
|
2015-03-30 22:07:16 +00:00
|
|
|
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 10:29:18 +00:00
|
|
|
int slots, err = 0;
|
2014-09-24 10:19:10 +00:00
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
level = 0;
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
slots = GET_DENTRY_SLOTS(fname->disk_name.len);
|
2015-04-27 21:51:02 +00:00
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
current_depth = F2FS_I(dir)->i_current_depth;
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
if (F2FS_I(dir)->chash == fname->hash) {
|
2012-11-14 07:59:04 +00:00
|
|
|
level = F2FS_I(dir)->clevel;
|
|
|
|
F2FS_I(dir)->chash = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
start:
|
2017-02-25 03:08:28 +00:00
|
|
|
if (time_to_inject(F2FS_I_SB(dir), FAULT_DIR_DEPTH)) {
|
2019-11-01 09:53:23 +00:00
|
|
|
f2fs_show_injection_info(F2FS_I_SB(dir), FAULT_DIR_DEPTH);
|
2016-04-29 23:29:22 +00:00
|
|
|
return -ENOSPC;
|
2017-02-25 03:08:28 +00:00
|
|
|
}
|
2018-08-13 21:38:06 +00:00
|
|
|
|
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 10:29:18 +00:00
|
|
|
if (unlikely(current_depth == MAX_DIR_HASH_DEPTH))
|
|
|
|
return -ENOSPC;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
/* Increase the depth, if required */
|
|
|
|
if (level == current_depth)
|
|
|
|
++current_depth;
|
|
|
|
|
2014-02-27 09:20:00 +00:00
|
|
|
nbucket = dir_buckets(level, F2FS_I(dir)->i_dir_level);
|
2012-11-14 07:59:04 +00:00
|
|
|
nblock = bucket_blocks(level);
|
|
|
|
|
2014-02-27 09:20:00 +00:00
|
|
|
bidx = dir_block_index(level, F2FS_I(dir)->i_dir_level,
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
(le32_to_cpu(fname->hash) % nbucket));
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
for (block = bidx; block <= (bidx + nblock - 1); block++) {
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
dentry_page = f2fs_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 10:29:18 +00:00
|
|
|
if (IS_ERR(dentry_page))
|
|
|
|
return PTR_ERR(dentry_page);
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2018-02-28 12:31:52 +00:00
|
|
|
dentry_blk = page_address(dentry_page);
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
bit_pos = f2fs_room_for_filename(&dentry_blk->dentry_bitmap,
|
2014-10-13 23:28:13 +00:00
|
|
|
slots, NR_DENTRY_IN_BLOCK);
|
2012-11-14 07:59:04 +00:00
|
|
|
if (bit_pos < NR_DENTRY_IN_BLOCK)
|
|
|
|
goto add_dentry;
|
|
|
|
|
|
|
|
f2fs_put_page(dentry_page, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Move to next level to find the empty slot for new dentry */
|
|
|
|
++level;
|
|
|
|
goto start;
|
|
|
|
add_dentry:
|
2018-12-25 09:43:42 +00:00
|
|
|
f2fs_wait_on_page_writeback(dentry_page, DATA, true, true);
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2015-03-30 22:07:16 +00:00
|
|
|
if (inode) {
|
2022-01-07 20:48:44 +00:00
|
|
|
f2fs_down_write(&F2FS_I(inode)->i_sem);
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
page = f2fs_init_inode_metadata(inode, dir, fname, NULL);
|
2015-03-30 22:07:16 +00:00
|
|
|
if (IS_ERR(page)) {
|
|
|
|
err = PTR_ERR(page);
|
|
|
|
goto fail;
|
|
|
|
}
|
2013-05-20 01:10:29 +00:00
|
|
|
}
|
2015-02-16 08:17:20 +00:00
|
|
|
|
2017-04-04 10:01:22 +00:00
|
|
|
make_dentry_ptr_block(NULL, &d, dentry_blk);
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
f2fs_update_dentry(ino, mode, &d, &fname->disk_name, fname->hash,
|
|
|
|
bit_pos);
|
2015-02-16 08:17:20 +00:00
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
set_page_dirty(dentry_page);
|
f2fs: fix tracking parent inode number
Previously, f2fs didn't track the parent inode number correctly which is stored
in each f2fs_inode. In the case of the following scenario, a bug can be occured.
Let's suppose there are one directory, "/b", and two files, "/a" and "/b/a".
- pino of "/a" is ROOT_INO.
- pino of "/b/a" is DIR_B_INO.
Then,
# sync
: The inode pages of "/a" and "/b/a" contain the parent inode numbers as
ROOT_INO and DIR_B_INO respectively.
# mv /a /b/a
: The parent inode number of "/a" should be changed to DIR_B_INO, but f2fs
didn't do that. Ref. f2fs_set_link().
In order to fix this clearly, I added i_pino in f2fs_inode_info, and whenever
it needs to be changed like in f2fs_add_link() and f2fs_set_link(), it is
updated temporarily in f2fs_inode_info.
And later, f2fs_write_inode() stores the latest information to the inode pages.
For power-off-recovery, f2fs_sync_file() triggers simply f2fs_write_inode().
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
2012-12-10 08:52:48 +00:00
|
|
|
|
2015-03-30 22:07:16 +00:00
|
|
|
if (inode) {
|
2016-05-20 16:52:20 +00:00
|
|
|
f2fs_i_pino_write(inode, dir->i_ino);
|
2019-09-10 01:14:16 +00:00
|
|
|
|
|
|
|
/* synchronize inode page's data from inode cache */
|
|
|
|
if (is_inode_flag_set(inode, FI_NEW_INODE))
|
|
|
|
f2fs_update_inode(inode, page);
|
|
|
|
|
2015-03-30 22:07:16 +00:00
|
|
|
f2fs_put_page(page, 1);
|
|
|
|
}
|
2013-05-20 01:10:29 +00:00
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_update_parent_metadata(dir, inode, current_depth);
|
2012-11-14 07:59:04 +00:00
|
|
|
fail:
|
2015-03-30 22:07:16 +00:00
|
|
|
if (inode)
|
2022-01-07 20:48:44 +00:00
|
|
|
f2fs_up_write(&F2FS_I(inode)->i_sem);
|
2014-03-20 10:10:08 +00:00
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
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 10:29:18 +00:00
|
|
|
|
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
int f2fs_add_dentry(struct inode *dir, const struct f2fs_filename *fname,
|
|
|
|
struct inode *inode, nid_t ino, umode_t mode)
|
2016-08-29 03:27:56 +00:00
|
|
|
{
|
|
|
|
int err = -EAGAIN;
|
|
|
|
|
|
|
|
if (f2fs_has_inline_dentry(dir))
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
err = f2fs_add_inline_entry(dir, fname, inode, ino, mode);
|
2016-08-29 03:27:56 +00:00
|
|
|
if (err == -EAGAIN)
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
err = f2fs_add_regular_entry(dir, fname, inode, ino, mode);
|
2016-08-29 03:27:56 +00:00
|
|
|
|
|
|
|
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 10:29:18 +00:00
|
|
|
/*
|
|
|
|
* Caller should grab and release a rwsem by calling f2fs_lock_op() and
|
|
|
|
* f2fs_unlock_op().
|
|
|
|
*/
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
int f2fs_do_add_link(struct inode *dir, const struct qstr *name,
|
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 10:29:18 +00:00
|
|
|
struct inode *inode, nid_t ino, umode_t mode)
|
|
|
|
{
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
struct f2fs_filename fname;
|
2017-02-14 17:54:37 +00:00
|
|
|
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 10:29:18 +00:00
|
|
|
int err;
|
|
|
|
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
err = f2fs_setup_filename(dir, name, 0, &fname);
|
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 10:29:18 +00:00
|
|
|
if (err)
|
|
|
|
return err;
|
|
|
|
|
2017-02-14 17:54:37 +00:00
|
|
|
/*
|
2020-06-25 12:40:11 +00:00
|
|
|
* An immature stackable filesystem shows a race condition between lookup
|
2017-02-14 17:54:37 +00:00
|
|
|
* 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_put_page(page, 0);
|
|
|
|
err = -EEXIST;
|
|
|
|
} else if (IS_ERR(page)) {
|
|
|
|
err = PTR_ERR(page);
|
|
|
|
} else {
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
err = f2fs_add_dentry(dir, &fname, inode, ino, mode);
|
2017-02-14 17:54:37 +00:00
|
|
|
}
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
f2fs_free_filename(&fname);
|
2012-11-14 07:59:04 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2014-06-21 04:37:02 +00:00
|
|
|
int f2fs_do_tmpfile(struct inode *inode, struct inode *dir)
|
|
|
|
{
|
|
|
|
struct page *page;
|
|
|
|
int err = 0;
|
|
|
|
|
2022-01-07 20:48:44 +00:00
|
|
|
f2fs_down_write(&F2FS_I(inode)->i_sem);
|
f2fs: rework filename handling
Rework f2fs's handling of filenames to use a new 'struct f2fs_filename'.
Similar to 'struct ext4_filename', this stores the usr_fname, disk_name,
dirhash, crypto_buf, and casefolded name. Some of these names can be
NULL in some cases. 'struct f2fs_filename' differs from
'struct fscrypt_name' mainly in that the casefolded name is included.
For user-initiated directory operations like lookup() and create(),
initialize the f2fs_filename by translating the corresponding
fscrypt_name, then computing the dirhash and casefolded name if needed.
This makes the dirhash and casefolded name be cached for each syscall,
so we don't have to recompute them repeatedly. (Previously, f2fs
computed the dirhash once per directory level, and the casefolded name
once per directory block.) This improves performance.
This rework also makes it much easier to correctly handle all
combinations of normal, encrypted, casefolded, and encrypted+casefolded
directories. (The fourth isn't supported yet but is being worked on.)
The only other cases where an f2fs_filename gets initialized are for two
filesystem-internal operations: (1) when converting an inline directory
to a regular one, we grab the needed disk_name and hash from an existing
f2fs_dir_entry; and (2) when roll-forward recovering a new dentry, we
grab the needed disk_name from f2fs_inode::i_name and compute the hash.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2020-05-07 07:59:04 +00:00
|
|
|
page = f2fs_init_inode_metadata(inode, dir, NULL, NULL);
|
2014-06-21 04:37:02 +00:00
|
|
|
if (IS_ERR(page)) {
|
|
|
|
err = PTR_ERR(page);
|
|
|
|
goto fail;
|
|
|
|
}
|
|
|
|
f2fs_put_page(page, 1);
|
|
|
|
|
2016-05-20 17:13:22 +00:00
|
|
|
clear_inode_flag(inode, FI_NEW_INODE);
|
2018-10-05 05:17:39 +00:00
|
|
|
f2fs_update_time(F2FS_I_SB(inode), REQ_TIME);
|
2014-06-21 04:37:02 +00:00
|
|
|
fail:
|
2022-01-07 20:48:44 +00:00
|
|
|
f2fs_up_write(&F2FS_I(inode)->i_sem);
|
2014-06-21 04:37:02 +00:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2016-06-02 04:18:25 +00:00
|
|
|
void f2fs_drop_nlink(struct inode *dir, struct inode *inode)
|
2014-09-24 10:17:04 +00:00
|
|
|
{
|
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(dir);
|
|
|
|
|
2022-01-07 20:48:44 +00:00
|
|
|
f2fs_down_write(&F2FS_I(inode)->i_sem);
|
2014-09-24 10:17:04 +00:00
|
|
|
|
2016-05-20 23:32:49 +00:00
|
|
|
if (S_ISDIR(inode->i_mode))
|
2016-05-20 16:43:20 +00:00
|
|
|
f2fs_i_links_write(dir, false);
|
2016-09-14 14:48:04 +00:00
|
|
|
inode->i_ctime = current_time(inode);
|
2014-09-24 10:17:04 +00:00
|
|
|
|
2016-05-20 16:43:20 +00:00
|
|
|
f2fs_i_links_write(inode, false);
|
2014-09-24 10:17:04 +00:00
|
|
|
if (S_ISDIR(inode->i_mode)) {
|
2016-05-20 16:43:20 +00:00
|
|
|
f2fs_i_links_write(inode, false);
|
2016-05-20 16:22:03 +00:00
|
|
|
f2fs_i_size_write(inode, 0);
|
2014-09-24 10:17:04 +00:00
|
|
|
}
|
2022-01-07 20:48:44 +00:00
|
|
|
f2fs_up_write(&F2FS_I(inode)->i_sem);
|
2014-09-24 10:17:04 +00:00
|
|
|
|
|
|
|
if (inode->i_nlink == 0)
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_add_orphan_inode(inode);
|
2014-09-24 10:17:04 +00:00
|
|
|
else
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_release_orphan_inode(sbi);
|
2014-09-24 10:17:04 +00:00
|
|
|
}
|
|
|
|
|
2012-11-29 04:28:09 +00:00
|
|
|
/*
|
2014-08-06 14:22:50 +00:00
|
|
|
* It only removes the dentry from the dentry page, corresponding name
|
2012-11-14 07:59:04 +00:00
|
|
|
* entry in name page does not need to be touched during deletion.
|
|
|
|
*/
|
|
|
|
void f2fs_delete_entry(struct f2fs_dir_entry *dentry, struct page *page,
|
2014-09-24 10:17:04 +00:00
|
|
|
struct inode *dir, struct inode *inode)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
|
|
|
struct f2fs_dentry_block *dentry_blk;
|
|
|
|
unsigned int bit_pos;
|
2012-12-08 05:54:50 +00:00
|
|
|
int slots = GET_DENTRY_SLOTS(le16_to_cpu(dentry->name_len));
|
2012-11-14 07:59:04 +00:00
|
|
|
int i;
|
|
|
|
|
2016-01-09 00:57:48 +00:00
|
|
|
f2fs_update_time(F2FS_I_SB(dir), REQ_TIME);
|
|
|
|
|
2018-03-08 06:22:56 +00:00
|
|
|
if (F2FS_OPTION(F2FS_I_SB(dir)).fsync_mode == FSYNC_MODE_STRICT)
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_add_ino_entry(F2FS_I_SB(dir), dir->i_ino, TRANS_DIR_INO);
|
2017-12-28 16:09:44 +00:00
|
|
|
|
2014-09-24 10:19:10 +00:00
|
|
|
if (f2fs_has_inline_dentry(dir))
|
|
|
|
return f2fs_delete_inline_entry(dentry, page, dir, inode);
|
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
lock_page(page);
|
2018-12-25 09:43:42 +00:00
|
|
|
f2fs_wait_on_page_writeback(page, DATA, true, true);
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2014-06-27 09:57:04 +00:00
|
|
|
dentry_blk = page_address(page);
|
|
|
|
bit_pos = dentry - dentry_blk->dentry;
|
2012-11-14 07:59:04 +00:00
|
|
|
for (i = 0; i < slots; i++)
|
2017-03-07 22:11:06 +00:00
|
|
|
__clear_bit_le(bit_pos + i, &dentry_blk->dentry_bitmap);
|
2012-11-14 07:59:04 +00:00
|
|
|
|
|
|
|
/* Let's check and deallocate this dentry page */
|
|
|
|
bit_pos = find_next_bit_le(&dentry_blk->dentry_bitmap,
|
|
|
|
NR_DENTRY_IN_BLOCK,
|
|
|
|
0);
|
|
|
|
set_page_dirty(page);
|
|
|
|
|
2015-08-12 09:48:21 +00:00
|
|
|
if (bit_pos == NR_DENTRY_IN_BLOCK &&
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
!f2fs_truncate_hole(dir, page->index, page->index + 1)) {
|
2017-12-05 01:25:25 +00:00
|
|
|
f2fs_clear_page_cache_dirty_tag(page);
|
2012-11-14 07:59:04 +00:00
|
|
|
clear_page_dirty_for_io(page);
|
|
|
|
ClearPageUptodate(page);
|
2021-04-28 09:20:31 +00:00
|
|
|
|
|
|
|
clear_page_private_gcing(page);
|
|
|
|
|
2014-09-12 22:53:45 +00:00
|
|
|
inode_dec_dirty_pages(dir);
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_remove_dirty_inode(dir);
|
2021-04-28 09:20:31 +00:00
|
|
|
|
|
|
|
detach_page_private(page);
|
|
|
|
set_page_private(page, 0);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
2012-12-08 05:54:35 +00:00
|
|
|
f2fs_put_page(page, 1);
|
2020-03-20 10:17:54 +00:00
|
|
|
|
|
|
|
dir->i_ctime = dir->i_mtime = current_time(dir);
|
|
|
|
f2fs_mark_inode_dirty_sync(dir, false);
|
|
|
|
|
|
|
|
if (inode)
|
|
|
|
f2fs_drop_nlink(dir, inode);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
bool f2fs_empty_dir(struct inode *dir)
|
|
|
|
{
|
|
|
|
unsigned long bidx;
|
|
|
|
struct page *dentry_page;
|
|
|
|
unsigned int bit_pos;
|
2014-09-24 10:17:04 +00:00
|
|
|
struct f2fs_dentry_block *dentry_blk;
|
2012-11-14 07:59:04 +00:00
|
|
|
unsigned long nblock = dir_blocks(dir);
|
|
|
|
|
2014-09-24 10:19:10 +00:00
|
|
|
if (f2fs_has_inline_dentry(dir))
|
|
|
|
return f2fs_empty_inline_dir(dir);
|
|
|
|
|
2012-11-14 07:59:04 +00:00
|
|
|
for (bidx = 0; bidx < nblock; bidx++) {
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
dentry_page = f2fs_get_lock_data_page(dir, bidx, false);
|
2012-11-14 07:59:04 +00:00
|
|
|
if (IS_ERR(dentry_page)) {
|
|
|
|
if (PTR_ERR(dentry_page) == -ENOENT)
|
|
|
|
continue;
|
|
|
|
else
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2018-02-28 12:31:52 +00:00
|
|
|
dentry_blk = page_address(dentry_page);
|
2012-11-14 07:59:04 +00:00
|
|
|
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);
|
|
|
|
|
|
|
|
f2fs_put_page(dentry_page, 1);
|
|
|
|
|
|
|
|
if (bit_pos < NR_DENTRY_IN_BLOCK)
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2016-10-29 10:46:34 +00:00
|
|
|
int f2fs_fill_dentries(struct dir_context *ctx, struct f2fs_dentry_ptr *d,
|
2015-05-15 23:26:10 +00:00
|
|
|
unsigned int start_pos, struct fscrypt_str *fstr)
|
2014-10-16 04:29:51 +00:00
|
|
|
{
|
|
|
|
unsigned char d_type = DT_UNKNOWN;
|
|
|
|
unsigned int bit_pos;
|
|
|
|
struct f2fs_dir_entry *de = NULL;
|
2015-05-15 23:26:10 +00:00
|
|
|
struct fscrypt_str de_name = FSTR_INIT(NULL, 0);
|
2017-11-22 10:23:38 +00:00
|
|
|
struct f2fs_sb_info *sbi = F2FS_I_SB(d->inode);
|
2018-09-07 11:49:07 +00:00
|
|
|
struct blk_plug plug;
|
|
|
|
bool readdir_ra = sbi->readdir_ra == 1;
|
f2fs: reduce the scope of setting fsck tag when de->name_len is zero
I recently found a case where de->name_len is 0 in f2fs_fill_dentries()
easily reproduced, and finally set the fsck flag.
Thread A Thread B
- f2fs_readdir
- f2fs_read_inline_dir
- ctx->pos = d.max
- f2fs_add_dentry
- f2fs_add_inline_entry
- do_convert_inline_dir
- f2fs_add_regular_entry
- f2fs_readdir
- f2fs_fill_dentries
- set_sbi_flag(sbi, SBI_NEED_FSCK)
Process A opens the folder, and has been reading without closing it.
During this period, Process B created a file under the folder (occupying
multiple f2fs_dir_entry, exceeding the d.max of the inline dir). After
creation, process A uses the d.max of inline dir to read it again, and
it will read that de->name_len is 0.
And Chao pointed out that w/o inline conversion, the race condition still
can happen as below:
dir_entry1: A
dir_entry2: B
dir_entry3: C
free slot: _
ctx->pos: ^
Thread A is traversing directory,
ctx-pos moves to below position after readdir() by thread A:
AAAABBBB___
^
Then thread B delete dir_entry2, and create dir_entry3.
Thread A calls readdir() to lookup dirents starting from middle
of new dirent slots as below:
AAAACCCCCC_
^
In these scenarios, the file system is not damaged, and it's hard to
avoid it. But we can bypass tagging FSCK flag if:
a) bit_pos (:= ctx->pos % d->max) is non-zero and
b) before bit_pos moves to first valid dir_entry.
Fixes: ddf06b753a85 ("f2fs: fix to trigger fsck if dirent.name_len is zero")
Signed-off-by: Yangtao Li <frank.li@vivo.com>
[Chao: clean up description]
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2021-08-04 03:29:46 +00:00
|
|
|
bool found_valid_dirent = false;
|
2018-09-07 11:49:07 +00:00
|
|
|
int err = 0;
|
2014-10-16 04:29:51 +00:00
|
|
|
|
2014-10-19 05:52:52 +00:00
|
|
|
bit_pos = ((unsigned long)ctx->pos % d->max);
|
2014-10-16 04:29:51 +00:00
|
|
|
|
2018-09-07 11:49:07 +00:00
|
|
|
if (readdir_ra)
|
|
|
|
blk_start_plug(&plug);
|
|
|
|
|
2014-10-19 05:52:52 +00:00
|
|
|
while (bit_pos < d->max) {
|
|
|
|
bit_pos = find_next_bit_le(d->bitmap, d->max, bit_pos);
|
|
|
|
if (bit_pos >= d->max)
|
2014-10-16 04:29:51 +00:00
|
|
|
break;
|
|
|
|
|
2014-10-19 05:52:52 +00:00
|
|
|
de = &d->dentry[bit_pos];
|
2016-02-12 22:29:28 +00:00
|
|
|
if (de->name_len == 0) {
|
f2fs: reduce the scope of setting fsck tag when de->name_len is zero
I recently found a case where de->name_len is 0 in f2fs_fill_dentries()
easily reproduced, and finally set the fsck flag.
Thread A Thread B
- f2fs_readdir
- f2fs_read_inline_dir
- ctx->pos = d.max
- f2fs_add_dentry
- f2fs_add_inline_entry
- do_convert_inline_dir
- f2fs_add_regular_entry
- f2fs_readdir
- f2fs_fill_dentries
- set_sbi_flag(sbi, SBI_NEED_FSCK)
Process A opens the folder, and has been reading without closing it.
During this period, Process B created a file under the folder (occupying
multiple f2fs_dir_entry, exceeding the d.max of the inline dir). After
creation, process A uses the d.max of inline dir to read it again, and
it will read that de->name_len is 0.
And Chao pointed out that w/o inline conversion, the race condition still
can happen as below:
dir_entry1: A
dir_entry2: B
dir_entry3: C
free slot: _
ctx->pos: ^
Thread A is traversing directory,
ctx-pos moves to below position after readdir() by thread A:
AAAABBBB___
^
Then thread B delete dir_entry2, and create dir_entry3.
Thread A calls readdir() to lookup dirents starting from middle
of new dirent slots as below:
AAAACCCCCC_
^
In these scenarios, the file system is not damaged, and it's hard to
avoid it. But we can bypass tagging FSCK flag if:
a) bit_pos (:= ctx->pos % d->max) is non-zero and
b) before bit_pos moves to first valid dir_entry.
Fixes: ddf06b753a85 ("f2fs: fix to trigger fsck if dirent.name_len is zero")
Signed-off-by: Yangtao Li <frank.li@vivo.com>
[Chao: clean up description]
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2021-08-04 03:29:46 +00:00
|
|
|
if (found_valid_dirent || !bit_pos) {
|
|
|
|
printk_ratelimited(
|
|
|
|
"%sF2FS-fs (%s): invalid namelen(0), ino:%u, run fsck to fix.",
|
|
|
|
KERN_WARNING, sbi->sb->s_id,
|
|
|
|
le32_to_cpu(de->ino));
|
|
|
|
set_sbi_flag(sbi, SBI_NEED_FSCK);
|
|
|
|
}
|
2016-02-12 22:29:28 +00:00
|
|
|
bit_pos++;
|
|
|
|
ctx->pos = start_pos + bit_pos;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
d_type = f2fs_get_de_type(de);
|
2015-04-27 23:26:24 +00:00
|
|
|
|
|
|
|
de_name.name = d->filename[bit_pos];
|
|
|
|
de_name.len = le16_to_cpu(de->name_len);
|
|
|
|
|
2018-11-14 20:40:30 +00:00
|
|
|
/* check memory boundary before moving forward */
|
|
|
|
bit_pos += GET_DENTRY_SLOTS(le16_to_cpu(de->name_len));
|
2019-01-07 07:02:34 +00:00
|
|
|
if (unlikely(bit_pos > d->max ||
|
|
|
|
le16_to_cpu(de->name_len) > F2FS_NAME_LEN)) {
|
2019-06-18 09:48:42 +00:00
|
|
|
f2fs_warn(sbi, "%s: corrupted namelen=%d, run fsck to fix.",
|
|
|
|
__func__, le16_to_cpu(de->name_len));
|
2018-11-14 20:40:30 +00:00
|
|
|
set_sbi_flag(sbi, SBI_NEED_FSCK);
|
2019-06-20 03:36:14 +00:00
|
|
|
err = -EFSCORRUPTED;
|
2022-09-28 15:38:54 +00:00
|
|
|
f2fs_handle_error(sbi, ERROR_CORRUPTED_DIRENT);
|
2018-11-14 20:40:30 +00:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
2018-12-12 09:50:11 +00:00
|
|
|
if (IS_ENCRYPTED(d->inode)) {
|
2015-04-27 23:26:24 +00:00
|
|
|
int save_len = fstr->len;
|
|
|
|
|
2016-09-15 21:25:55 +00:00
|
|
|
err = fscrypt_fname_disk_to_usr(d->inode,
|
2019-05-28 09:23:33 +00:00
|
|
|
(u32)le32_to_cpu(de->hash_code),
|
|
|
|
0, &de_name, fstr);
|
2016-09-15 21:25:55 +00:00
|
|
|
if (err)
|
2018-09-07 11:49:07 +00:00
|
|
|
goto out;
|
2015-09-03 20:38:23 +00:00
|
|
|
|
|
|
|
de_name = *fstr;
|
|
|
|
fstr->len = save_len;
|
2015-04-27 23:26:24 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!dir_emit(ctx, de_name.name, de_name.len,
|
2018-09-07 11:49:07 +00:00
|
|
|
le32_to_cpu(de->ino), d_type)) {
|
|
|
|
err = 1;
|
|
|
|
goto out;
|
|
|
|
}
|
2014-10-16 04:29:51 +00:00
|
|
|
|
2018-09-07 11:49:07 +00:00
|
|
|
if (readdir_ra)
|
f2fs: clean up symbol namespace
As Ted reported:
"Hi, I was looking at f2fs's sources recently, and I noticed that there
is a very large number of non-static symbols which don't have a f2fs
prefix. There's well over a hundred (see attached below).
As one example, in fs/f2fs/dir.c there is:
unsigned char get_de_type(struct f2fs_dir_entry *de)
This function is clearly only useful for f2fs, but it has a generic
name. This means that if any other file system tries to have the same
symbol name, there will be a symbol conflict and the kernel would not
successfully build. It also means that when someone is looking f2fs
sources, it's not at all obvious whether a function such as
read_data_page(), invalidate_blocks(), is a generic kernel function
found in the fs, mm, or block layers, or a f2fs specific function.
You might want to fix this at some point. Hopefully Kent's bcachefs
isn't similarly using genericly named functions, since that might
cause conflicts with f2fs's functions --- but just as this would be a
problem that we would rightly insist that Kent fix, this is something
that we should have rightly insisted that f2fs should have fixed
before it was integrated into the mainline kernel.
acquire_orphan_inode
add_ino_entry
add_orphan_inode
allocate_data_block
allocate_new_segments
alloc_nid
alloc_nid_done
alloc_nid_failed
available_free_memory
...."
This patch adds "f2fs_" prefix for all non-static symbols in order to:
a) avoid conflict with other kernel generic symbols;
b) to indicate the function is f2fs specific one instead of generic
one;
Reported-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2018-05-29 16:20:41 +00:00
|
|
|
f2fs_ra_node_page(sbi, le32_to_cpu(de->ino));
|
2017-11-22 10:23:38 +00:00
|
|
|
|
2014-10-16 04:29:51 +00:00
|
|
|
ctx->pos = start_pos + bit_pos;
|
f2fs: reduce the scope of setting fsck tag when de->name_len is zero
I recently found a case where de->name_len is 0 in f2fs_fill_dentries()
easily reproduced, and finally set the fsck flag.
Thread A Thread B
- f2fs_readdir
- f2fs_read_inline_dir
- ctx->pos = d.max
- f2fs_add_dentry
- f2fs_add_inline_entry
- do_convert_inline_dir
- f2fs_add_regular_entry
- f2fs_readdir
- f2fs_fill_dentries
- set_sbi_flag(sbi, SBI_NEED_FSCK)
Process A opens the folder, and has been reading without closing it.
During this period, Process B created a file under the folder (occupying
multiple f2fs_dir_entry, exceeding the d.max of the inline dir). After
creation, process A uses the d.max of inline dir to read it again, and
it will read that de->name_len is 0.
And Chao pointed out that w/o inline conversion, the race condition still
can happen as below:
dir_entry1: A
dir_entry2: B
dir_entry3: C
free slot: _
ctx->pos: ^
Thread A is traversing directory,
ctx-pos moves to below position after readdir() by thread A:
AAAABBBB___
^
Then thread B delete dir_entry2, and create dir_entry3.
Thread A calls readdir() to lookup dirents starting from middle
of new dirent slots as below:
AAAACCCCCC_
^
In these scenarios, the file system is not damaged, and it's hard to
avoid it. But we can bypass tagging FSCK flag if:
a) bit_pos (:= ctx->pos % d->max) is non-zero and
b) before bit_pos moves to first valid dir_entry.
Fixes: ddf06b753a85 ("f2fs: fix to trigger fsck if dirent.name_len is zero")
Signed-off-by: Yangtao Li <frank.li@vivo.com>
[Chao: clean up description]
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
2021-08-04 03:29:46 +00:00
|
|
|
found_valid_dirent = true;
|
2014-10-16 04:29:51 +00:00
|
|
|
}
|
2018-09-07 11:49:07 +00:00
|
|
|
out:
|
|
|
|
if (readdir_ra)
|
|
|
|
blk_finish_plug(&plug);
|
|
|
|
return err;
|
2014-10-16 04:29:51 +00:00
|
|
|
}
|
|
|
|
|
2013-05-17 22:02:17 +00:00
|
|
|
static int f2fs_readdir(struct file *file, struct dir_context *ctx)
|
2012-11-14 07:59:04 +00:00
|
|
|
{
|
2013-01-23 22:07:38 +00:00
|
|
|
struct inode *inode = file_inode(file);
|
2012-11-14 07:59:04 +00:00
|
|
|
unsigned long npages = dir_blocks(inode);
|
|
|
|
struct f2fs_dentry_block *dentry_blk = NULL;
|
|
|
|
struct page *dentry_page = NULL;
|
2014-04-28 09:59:43 +00:00
|
|
|
struct file_ra_state *ra = &file->f_ra;
|
2017-10-13 10:01:33 +00:00
|
|
|
loff_t start_pos = ctx->pos;
|
2013-05-17 22:02:17 +00:00
|
|
|
unsigned int n = ((unsigned long)ctx->pos / NR_DENTRY_IN_BLOCK);
|
2014-10-19 05:52:52 +00:00
|
|
|
struct f2fs_dentry_ptr d;
|
2015-05-15 23:26:10 +00:00
|
|
|
struct fscrypt_str fstr = FSTR_INIT(NULL, 0);
|
2015-04-27 23:26:24 +00:00
|
|
|
int err = 0;
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2018-12-12 09:50:11 +00:00
|
|
|
if (IS_ENCRYPTED(inode)) {
|
2020-12-03 02:20:37 +00:00
|
|
|
err = fscrypt_prepare_readdir(inode);
|
2019-12-09 21:23:48 +00:00
|
|
|
if (err)
|
2017-10-13 10:01:33 +00:00
|
|
|
goto out;
|
2015-05-20 05:26:54 +00:00
|
|
|
|
2020-08-10 14:21:39 +00:00
|
|
|
err = fscrypt_fname_alloc_buffer(F2FS_NAME_LEN, &fstr);
|
2015-04-27 23:26:24 +00:00
|
|
|
if (err < 0)
|
2017-10-13 10:01:33 +00:00
|
|
|
goto out;
|
2015-04-27 23:26:24 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
if (f2fs_has_inline_dentry(inode)) {
|
|
|
|
err = f2fs_read_inline_dir(file, ctx, &fstr);
|
2017-10-13 10:01:33 +00:00
|
|
|
goto out_free;
|
2015-04-27 23:26:24 +00:00
|
|
|
}
|
2014-09-24 10:19:10 +00:00
|
|
|
|
2017-10-13 10:01:36 +00:00
|
|
|
for (; n < npages; n++, ctx->pos = n * NR_DENTRY_IN_BLOCK) {
|
2017-10-13 10:01:34 +00:00
|
|
|
|
|
|
|
/* allow readdir() to be interrupted */
|
|
|
|
if (fatal_signal_pending(current)) {
|
|
|
|
err = -ERESTARTSYS;
|
|
|
|
goto out_free;
|
|
|
|
}
|
|
|
|
cond_resched();
|
|
|
|
|
2017-10-13 10:01:35 +00:00
|
|
|
/* 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));
|
|
|
|
|
2019-02-21 04:57:35 +00:00
|
|
|
dentry_page = f2fs_find_data_page(inode, n);
|
2015-11-19 08:09:07 +00:00
|
|
|
if (IS_ERR(dentry_page)) {
|
|
|
|
err = PTR_ERR(dentry_page);
|
2016-10-29 10:46:34 +00:00
|
|
|
if (err == -ENOENT) {
|
|
|
|
err = 0;
|
2015-11-19 08:09:07 +00:00
|
|
|
continue;
|
2016-10-29 10:46:34 +00:00
|
|
|
} else {
|
2017-10-13 10:01:33 +00:00
|
|
|
goto out_free;
|
2016-10-29 10:46:34 +00:00
|
|
|
}
|
2015-11-19 08:09:07 +00:00
|
|
|
}
|
2012-11-14 07:59:04 +00:00
|
|
|
|
2018-02-28 12:31:52 +00:00
|
|
|
dentry_blk = page_address(dentry_page);
|
2013-07-05 08:28:12 +00:00
|
|
|
|
2017-04-04 10:01:22 +00:00
|
|
|
make_dentry_ptr_block(inode, &d, dentry_blk);
|
2014-10-19 05:52:52 +00:00
|
|
|
|
2016-10-29 10:46:34 +00:00
|
|
|
err = f2fs_fill_dentries(ctx, &d,
|
|
|
|
n * NR_DENTRY_IN_BLOCK, &fstr);
|
|
|
|
if (err) {
|
2019-02-21 04:57:35 +00:00
|
|
|
f2fs_put_page(dentry_page, 0);
|
2015-12-01 03:41:50 +00:00
|
|
|
break;
|
|
|
|
}
|
2014-10-16 04:29:51 +00:00
|
|
|
|
2019-02-21 04:57:35 +00:00
|
|
|
f2fs_put_page(dentry_page, 0);
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
2017-10-13 10:01:33 +00:00
|
|
|
out_free:
|
2015-05-15 23:26:10 +00:00
|
|
|
fscrypt_fname_free_buffer(&fstr);
|
2017-10-13 10:01:33 +00:00
|
|
|
out:
|
|
|
|
trace_f2fs_readdir(inode, start_pos, ctx->pos, err);
|
2016-10-29 10:46:34 +00:00
|
|
|
return err < 0 ? err : 0;
|
2012-11-14 07:59:04 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
const struct file_operations f2fs_dir_operations = {
|
|
|
|
.llseek = generic_file_llseek,
|
|
|
|
.read = generic_read_dir,
|
2016-05-10 20:41:13 +00:00
|
|
|
.iterate_shared = f2fs_readdir,
|
2012-11-14 07:59:04 +00:00
|
|
|
.fsync = f2fs_sync_file,
|
|
|
|
.unlocked_ioctl = f2fs_ioctl,
|
2015-05-12 08:05:57 +00:00
|
|
|
#ifdef CONFIG_COMPAT
|
|
|
|
.compat_ioctl = f2fs_compat_ioctl,
|
|
|
|
#endif
|
2012-11-14 07:59:04 +00:00
|
|
|
};
|