During defragment, we missed to trigger fragmented blocks migration
for below condition:
In defragment region:
- total number of valid blocks is smaller than 512;
- the tail part of the region are all holes;
In addtion, return zero to user via range->len if there is no
fragmented blocks.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
EOPNOTSUPP is widely used as error number indicating operation is
not supported in syscall, and ENOTSUPP was defined and only used
for NFSv3 protocol, so use EOPNOTSUPP instead.
Fixes: 0a2aa8fbb9 ("f2fs: refactor __exchange_data_block for speed up")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Modeled after commit b886ee3e77 ("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>
Add charset encoding to f2fs to support casefolding. It is modeled after
the same feature introduced in commit c83ad55eaa ("ext4: include charset
encoding information in the superblock")
Currently this is not compatible with encryption, similar to the current
ext4 imlpementation. This will change in the future.
>From the ext4 patch:
"""
The s_encoding field stores a magic number indicating the encoding
format and version used globally by file and directory names in the
filesystem. The s_encoding_flags defines policies for using the charset
encoding, like how to handle invalid sequences. The magic number is
mapped to the exact charset table, but the mapping is specific to ext4.
Since we don't have any commitment to support old encodings, the only
encoding I am supporting right now is utf8-12.1.0.
The current implementation prevents the user from enabling encoding and
per-directory encryption on the same filesystem at the same time. The
incompatibility between these features lies in how we do efficient
directory searches when we cannot be sure the encryption of the user
provided fname will match the actual hash stored in the disk without
decrypting every directory entry, because of normalization cases. My
quickest solution is to simply block the concurrent use of these
features for now, and enable it later, once we have a better solution.
"""
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In preparation for including the casefold feature within f2fs, elevate
the EXT4_CASEFOLD_FL flag to FS_CASEFOLD_FL.
Signed-off-by: Daniel Rosenberg <drosen@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
vfree() don't wish to be called from interrupt context, move it
out of spin_lock_irqsave() coverage.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
In fill_super() and put_super(), f2fs_destroy_stats() is called
in prior to f2fs_destroy_segment_manager(), so if current
sbi can still be visited in global stat list, SM_I(sbi) should be
released yet.
For this reason, SM_I(sbi) does not need to be checked in
update_general_status().
Thank Chao Yu for advice.
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Atomic write needs page cache to cache data of transaction,
direct IO should never be allowed in atomic write, detect
and deny it when open atomic write file.
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
With quota_ino feature on, generic/232 reports an inconsistence issue
on the image.
The root cause is that the testcase tries to:
- use quotactl to shutdown journalled quota based on sysfile;
- and then use quotactl to enable/turn on quota based on specific file
(aquota.user or aquota.group).
Eventually, quota sysfile will be out-of-update due to following specific
file creation.
Change as below to fix this issue:
- deny enabling quota based on specific file if quota sysfile exists.
- set SBI_QUOTA_NEED_REPAIR once sysfile based quota shutdowns via
ioctl.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
It needs to return -EIO if filesystem has been shutdown, fix the
miss case in f2fs_setxattr().
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
We missed to call f2fs_is_checkpoint_ready() in several places, it may
allow space allocation even when free space was exhausted during
checkpoint is disabled, fix to add them.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Adjust f2fs_fiemap() to support fiemap() on directory inode.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
=============================================================================
BUG discard_cmd (Tainted: G B OE ): Objects remaining in discard_cmd on __kmem_cache_shutdown()
-----------------------------------------------------------------------------
INFO: Slab 0xffffe1ac481d22c0 objects=36 used=2 fp=0xffff936b4748bf50 flags=0x2ffff0000000100
Call Trace:
dump_stack+0x63/0x87
slab_err+0xa1/0xb0
__kmem_cache_shutdown+0x183/0x390
shutdown_cache+0x14/0x110
kmem_cache_destroy+0x195/0x1c0
f2fs_destroy_segment_manager_caches+0x21/0x40 [f2fs]
exit_f2fs_fs+0x35/0x641 [f2fs]
SyS_delete_module+0x155/0x230
? vtime_user_exit+0x29/0x70
do_syscall_64+0x6e/0x160
entry_SYSCALL64_slow_path+0x25/0x25
INFO: Object 0xffff936b4748b000 @offset=0
INFO: Object 0xffff936b4748b070 @offset=112
kmem_cache_destroy discard_cmd: Slab cache still has objects
Call Trace:
dump_stack+0x63/0x87
kmem_cache_destroy+0x1b4/0x1c0
f2fs_destroy_segment_manager_caches+0x21/0x40 [f2fs]
exit_f2fs_fs+0x35/0x641 [f2fs]
SyS_delete_module+0x155/0x230
do_syscall_64+0x6e/0x160
entry_SYSCALL64_slow_path+0x25/0x25
Recovery can cache discard commands, so in error path of fill_super(),
we need give a chance to handle them, otherwise it will lead to leak
of discard_cmd slab cache.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
On a quota disabled image, with fault injection, SBI_QUOTA_NEED_REPAIR
will be set incorrectly in error path of f2fs_evict_inode(), fix it.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
As reported in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=204193
A null pointer dereference bug is triggered in f2fs under kernel-5.1.3.
kasan_report.cold+0x5/0x32
f2fs_write_end_io+0x215/0x650
bio_endio+0x26e/0x320
blk_update_request+0x209/0x5d0
blk_mq_end_request+0x2e/0x230
lo_complete_rq+0x12c/0x190
blk_done_softirq+0x14a/0x1a0
__do_softirq+0x119/0x3e5
irq_exit+0x94/0xe0
call_function_single_interrupt+0xf/0x20
During umount, we will access NULL sbi->node_inode pointer in
f2fs_write_end_io():
f2fs_bug_on(sbi, page->mapping == NODE_MAPPING(sbi) &&
page->index != nid_of_node(page));
The reason is if disable_checkpoint mount option is on, meta dirty
pages can remain during umount, and then be flushed by iput() of
meta_inode, however node_inode has been iput()ed before
meta_inode's iput().
Since checkpoint is disabled, all meta/node datas are useless and
should be dropped in next mount, so in umount, let's adjust
drop_inode() to give a hint to iput_final() to drop all those dirty
datas correctly.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
If IO alignment feature is turned on after remount, we didn't
initialize mempool of it, it turns out we will encounter panic
during IO submission due to access NULL mempool pointer.
This feature should be set only at mount time, so simply deny
configuring during remount.
This fixes bug reported in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=204135
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Since 07173c3ec2 ("block: enable multipage bvecs"), one bio vector
can store multi pages, so that we can not calculate max IO size of
bio as PAGE_SIZE * bio->bi_max_vecs. However IO alignment feature of
f2fs always has that assumption, so finally, it may cause panic during
IO submission as below stack.
kernel BUG at fs/f2fs/data.c:317!
RIP: 0010:__submit_merged_bio+0x8b0/0x8c0
Call Trace:
f2fs_submit_page_write+0x3cd/0xdd0
do_write_page+0x15d/0x360
f2fs_outplace_write_data+0xd7/0x210
f2fs_do_write_data_page+0x43b/0xf30
__write_data_page+0xcf6/0x1140
f2fs_write_cache_pages+0x3ba/0xb40
f2fs_write_data_pages+0x3dd/0x8b0
do_writepages+0xbb/0x1e0
__writeback_single_inode+0xb6/0x800
writeback_sb_inodes+0x441/0x910
wb_writeback+0x261/0x650
wb_workfn+0x1f9/0x7a0
process_one_work+0x503/0x970
worker_thread+0x7d/0x820
kthread+0x1ad/0x210
ret_from_fork+0x35/0x40
This patch adds one extra condition to check left space in bio while
trying merging page to bio, to avoid panic.
This bug was reported in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=204043
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Wrap merge condition into function for readability, no logic change.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Add fs-verity support to f2fs. fs-verity is a filesystem feature that
enables transparent integrity protection and authentication of read-only
files. It uses a dm-verity like mechanism at the file level: a Merkle
tree is used to verify any block in the file in log(filesize) time. It
is implemented mainly by helper functions in fs/verity/. See
Documentation/filesystems/fsverity.rst for the full documentation.
The f2fs support for fs-verity consists of:
- Adding a filesystem feature flag and an inode flag for fs-verity.
- Implementing the fsverity_operations to support enabling verity on an
inode and reading/writing the verity metadata.
- Updating ->readpages() to verify data as it's read from verity files
and to support reading verity metadata pages.
- Updating ->write_begin(), ->write_end(), and ->writepages() to support
writing verity metadata pages.
- Calling the fs-verity hooks for ->open(), ->setattr(), and ->ioctl().
Like ext4, f2fs stores the verity metadata (Merkle tree and
fsverity_descriptor) past the end of the file, starting at the first 64K
boundary beyond i_size. This approach works because (a) verity files
are readonly, and (b) pages fully beyond i_size aren't visible to
userspace but can be read/written internally by f2fs with only some
relatively small changes to f2fs. Extended attributes cannot be used
because (a) f2fs limits the total size of an inode's xattr entries to
4096 bytes, which wouldn't be enough for even a single Merkle tree
block, and (b) f2fs encryption doesn't encrypt xattrs, yet the verity
metadata *must* be encrypted when the file is because it contains hashes
of the plaintext data.
Acked-by: Jaegeuk Kim <jaegeuk@kernel.org>
Acked-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Document the format of verity files on ext4, and the corresponding inode
and superblock flags.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Make ext4_mpage_readpages() verify data as it is read from fs-verity
files, using the helper functions from fs/verity/.
To support both encryption and verity simultaneously, this required
refactoring the decryption workflow into a generic "post-read
processing" workflow which can do decryption, verification, or both.
The case where the ext4 block size is not equal to the PAGE_SIZE is not
supported yet, since in that case ext4_mpage_readpages() sometimes falls
back to block_read_full_page(), which does not support fs-verity yet.
Co-developed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add most of fs-verity support to ext4. fs-verity is a filesystem
feature that enables transparent integrity protection and authentication
of read-only files. It uses a dm-verity like mechanism at the file
level: a Merkle tree is used to verify any block in the file in
log(filesize) time. It is implemented mainly by helper functions in
fs/verity/. See Documentation/filesystems/fsverity.rst for the full
documentation.
This commit adds all of ext4 fs-verity support except for the actual
data verification, including:
- Adding a filesystem feature flag and an inode flag for fs-verity.
- Implementing the fsverity_operations to support enabling verity on an
inode and reading/writing the verity metadata.
- Updating ->write_begin(), ->write_end(), and ->writepages() to support
writing verity metadata pages.
- Calling the fs-verity hooks for ->open(), ->setattr(), and ->ioctl().
ext4 stores the verity metadata (Merkle tree and fsverity_descriptor)
past the end of the file, starting at the first 64K boundary beyond
i_size. This approach works because (a) verity files are readonly, and
(b) pages fully beyond i_size aren't visible to userspace but can be
read/written internally by ext4 with only some relatively small changes
to ext4. This approach avoids having to depend on the EA_INODE feature
and on rearchitecturing ext4's xattr support to support paging
multi-gigabyte xattrs into memory, and to support encrypting xattrs.
Note that the verity metadata *must* be encrypted when the file is,
since it contains hashes of the plaintext data.
This patch incorporates work by Theodore Ts'o and Chandan Rajendra.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
To meet some users' needs, add optional support for having fs-verity
handle a portion of the authentication policy in the kernel. An
".fs-verity" keyring is created to which X.509 certificates can be
added; then a sysctl 'fs.verity.require_signatures' can be set to cause
the kernel to enforce that all fs-verity files contain a signature of
their file measurement by a key in this keyring.
See the "Built-in signature verification" section of
Documentation/filesystems/fsverity.rst for the full documentation.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add SHA-512 support to fs-verity. This is primarily a demonstration of
the trivial changes needed to support a new hash algorithm in fs-verity;
most users will still use SHA-256, due to the smaller space required to
store the hashes. But some users may prefer SHA-512.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add a function for filesystems to call to implement the
FS_IOC_MEASURE_VERITY ioctl. This ioctl retrieves the file measurement
that fs-verity calculated for the given file and is enforcing for reads;
i.e., reads that don't match this hash will fail. This ioctl can be
used for authentication or logging of file measurements in userspace.
See the "FS_IOC_MEASURE_VERITY" section of
Documentation/filesystems/fsverity.rst for the documentation.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add a function for filesystems to call to implement the
FS_IOC_ENABLE_VERITY ioctl. This ioctl enables fs-verity on a file.
See the "FS_IOC_ENABLE_VERITY" section of
Documentation/filesystems/fsverity.rst for the documentation.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add functions that verify data pages that have been read from a
fs-verity file, against that file's Merkle tree. These will be called
from filesystems' ->readpage() and ->readpages() methods.
Since data verification can block, a workqueue is provided for these
methods to enqueue verification work from their bio completion callback.
See the "Verifying data" section of
Documentation/filesystems/fsverity.rst for more information.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add a function fsverity_prepare_setattr() which filesystems that support
fs-verity must call to deny truncates of verity files.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add the fsverity_file_open() function, which prepares an fs-verity file
to be read from. If not already done, it loads the fs-verity descriptor
from the filesystem and sets up an fsverity_info structure for the inode
which describes the Merkle tree and contains the file measurement. It
also denies all attempts to open verity files for writing.
This commit also begins the include/linux/fsverity.h header, which
declares the interface between fs/verity/ and filesystems.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Analogous to fs/crypto/, add fields to the VFS inode and superblock for
use by the fs/verity/ support layer:
- ->s_vop: points to the fsverity_operations if the filesystem supports
fs-verity, otherwise is NULL.
- ->i_verity_info: points to cached fs-verity information for the inode
after someone opens it, otherwise is NULL.
- S_VERITY: bit in ->i_flags that identifies verity inodes, even when
they haven't been opened yet and thus still have NULL ->i_verity_info.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add the beginnings of the fs/verity/ support layer, including the
Kconfig option and various helper functions for hashing. To start, only
SHA-256 is supported, but other hash algorithms can easily be added.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Add FS_VERITY_FL to the flags for FS_IOC_GETFLAGS, so that applications
can easily determine whether a file is a verity file at the same time as
they're checking other file flags. This flag will be gettable only;
FS_IOC_SETFLAGS won't allow setting it, since an ioctl must be used
instead to provide more parameters.
This flag matches the on-disk bit that was already allocated for ext4.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
fs-verity will be jointly maintained by Eric Biggers and Theodore Ts'o.
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
If an directory has the a casefold flag set without the casefold
feature set, s_encoding will not be initialized, and this will cause
the kernel to dereference a NULL pointer. In addition to adding
checks to avoid these kernel oops, attempts to load inodes with the
casefold flag when the casefold feature is not enable will cause the
file system to be declared corrupted.
[Jaegeuk Kim: use EXT4_ERROR_INODE]
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Temporarily cache a casefolded version of the file name under lookup in
ext4_filename, to avoid repeatedly casefolding it. I got up to 30%
speedup on lookups of large directories (>100k entries), depending on
the length of the string under lookup.
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Found by visual inspection, this wasn't caught by my xfstest, since it's
effect is ignoring positive dentries in the cache the fallback just goes
to the disk. it was introduced in the last iteration of the
case-insensitive patch.
d_compare should return 0 when the entries match, so make sure we are
correctly comparing the entire string if the encoding feature is set and
we are on a case-INsensitive directory.
Fixes: b886ee3e77 ("ext4: Support case-insensitive file name lookups")
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
scripts/mkutf8data is used only when regenerating utf8data.h,
which never happens in the normal kernel build. However, it is
irrespectively built if CONFIG_UNICODE is enabled.
Moreover, there is no good reason for it to reside in the scripts/
directory since it is only used in fs/unicode/.
Hence, move it from scripts/ to fs/unicode/.
In some cases, we bypass build artifacts in the normal build. The
conventional way to do so is to surround the code with ifdef REGENERATE_*.
For example,
- 7373f4f83c ("kbuild: add implicit rules for parser generation")
- 6aaf49b495 ("crypto: arm,arm64 - Fix random regeneration of S_shipped")
I rewrote the rule in a more kbuild'ish style.
In the normal build, utf8data.h is just shipped from the check-in file.
$ make
[ snip ]
SHIPPED fs/unicode/utf8data.h
CC fs/unicode/utf8-norm.o
CC fs/unicode/utf8-core.o
CC fs/unicode/utf8-selftest.o
AR fs/unicode/built-in.a
If you want to generate utf8data.h based on UCD, put *.txt files into
fs/unicode/, then pass REGENERATE_UTF8DATA=1 from the command line.
The mkutf8data tool will be automatically compiled to generate the
utf8data.h from the *.txt files.
$ make REGENERATE_UTF8DATA=1
[ snip ]
HOSTCC fs/unicode/mkutf8data
GEN fs/unicode/utf8data.h
CC fs/unicode/utf8-norm.o
CC fs/unicode/utf8-core.o
CC fs/unicode/utf8-selftest.o
AR fs/unicode/built-in.a
I renamed the check-in utf8data.h to utf8data.h_shipped so that this
will work for the out-of-tree build.
You can update it based on the latest UCD like this:
$ make REGENERATE_UTF8DATA=1 fs/unicode/
$ cp fs/unicode/utf8data.h fs/unicode/utf8data.h_shipped
Also, I added entries to .gitignore and dontdiff.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
This patch implements the actual support for case-insensitive file name
lookups in ext4, 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 (EXT4_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, Ext4 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 ext4 is implemented
lives 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 EXT4 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: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Support for encoding is considered an incompatible feature, since it has
potential to create collisions of file names in existing filesystems.
If the feature flag is not enabled, the entire filesystem will operate
on opaque byte sequences, respecting the original behavior.
The s_encoding field stores a magic number indicating the encoding
format and version used globally by file and directory names in the
filesystem. The s_encoding_flags defines policies for using the charset
encoding, like how to handle invalid sequences. The magic number is
mapped to the exact charset table, but the mapping is specific to ext4.
Since we don't have any commitment to support old encodings, the only
encoding I am supporting right now is utf8-12.1.0.
The current implementation prevents the user from enabling encoding and
per-directory encryption on the same filesystem at the same time. The
incompatibility between these features lies in how we do efficient
directory searches when we cannot be sure the encryption of the user
provided fname will match the actual hash stored in the disk without
decrypting every directory entry, because of normalization cases. My
quickest solution is to simply block the concurrent use of these
features for now, and enable it later, once we have a better solution.
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Regenerate utf8data.h based on the latest UCD files and run tests
against the latest version.
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
This implements a in-kernel sanity test module for the utf8
normalization core. At probe time, it will run basic sequences through
the utf8n core, to identify problems will equivalent sequences and
normalization/casefold code. This is supposed to be useful for
regression testing when adding support for a new version of utf8 to
linux.
Signed-off-by: Gabriel Krisman Bertazi <krisman@collabora.co.uk>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
