OSDN Git Service
Herbert Xu [Tue, 26 Jan 2016 14:14:36 +0000 (22:14 +0800)]
crypto: skcipher - Add helper to retrieve driver name
This patch adds the helper crypto_skcipher_driver_name which returns
the driver name of the alg object for a given tfm. This is needed by
ecryptfs.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Herbert Xu [Thu, 21 Jan 2016 09:10:56 +0000 (17:10 +0800)]
crypto: skcipher - Add default key size helper
While converting ecryptfs over to skcipher I found that it needs
to pick a default key size if one isn't given. Rather than having
it poke into the guts of the algorithm to get max_keysize, let's
provide a helper that is meant to give a sane default (just in
case we ever get an algorithm that has no maximum key size).
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Eric Biggers [Tue, 8 May 2018 00:22:08 +0000 (17:22 -0700)]
fscrypt: add Speck128/256 support
fscrypt currently only supports AES encryption. However, many low-end
mobile devices have older CPUs that don't have AES instructions, e.g.
the ARMv8 Cryptography Extensions. Currently, user data on such devices
is not encrypted at rest because AES is too slow, even when the NEON
bit-sliced implementation of AES is used. Unfortunately, it is
infeasible to encrypt these devices at all when AES is the only option.
Therefore, this patch updates fscrypt to support the Speck block cipher,
which was recently added to the crypto API. The C implementation of
Speck is not especially fast, but Speck can be implemented very
efficiently with general-purpose vector instructions, e.g. ARM NEON.
For example, on an ARMv7 processor, we measured the NEON-accelerated
Speck128/256-XTS at 69 MB/s for both encryption and decryption, while
AES-256-XTS with the NEON bit-sliced implementation was only 22 MB/s
encryption and 19 MB/s decryption.
There are multiple variants of Speck. This patch only adds support for
Speck128/256, which is the variant with a 128-bit block size and 256-bit
key size -- the same as AES-256. This is believed to be the most secure
variant of Speck, and it's only about 6% slower than Speck128/128.
Speck64/128 would be at least 20% faster because it has 20% rounds, and
it can be even faster on CPUs that can't efficiently do the 64-bit
operations needed for Speck128. However, Speck64's 64-bit block size is
not preferred security-wise. ARM NEON also supports the needed 64-bit
operations even on 32-bit CPUs, resulting in Speck128 being fast enough
for our targeted use cases so far.
The chosen modes of operation are XTS for contents and CTS-CBC for
filenames. These are the same modes of operation that fscrypt defaults
to for AES. Note that as with the other fscrypt modes, Speck will not
be used unless userspace chooses to use it. Nor are any of the existing
modes (which are all AES-based) being removed, of course.
We intentionally don't make CONFIG_FS_ENCRYPTION select
CONFIG_CRYPTO_SPECK, so people will have to enable Speck support
themselves if they need it. This is because we shouldn't bloat the
FS_ENCRYPTION dependencies with every new cipher, especially ones that
aren't recommended for most users. Moreover, CRYPTO_SPECK is just the
generic implementation, which won't be fast enough for many users; in
practice, they'll need to enable CRYPTO_SPECK_NEON to get acceptable
performance.
More details about our choice of Speck can be found in our patches that
added Speck to the crypto API, and the follow-on discussion threads.
We're planning a publication that explains the choice in more detail.
But briefly, we can't use ChaCha20 as we previously proposed, since it
would be insecure to use a stream cipher in this context, with potential
IV reuse during writes on f2fs and/or on wear-leveling flash storage.
We also evaluated many other lightweight and/or ARX-based block ciphers
such as Chaskey-LTS, RC5, LEA, CHAM, Threefish, RC6, NOEKEON, SPARX, and
XTEA. However, all had disadvantages vs. Speck, such as insufficient
performance with NEON, much less published cryptanalysis, or an
insufficient security level. Various design choices in Speck make it
perform better with NEON than competing ciphers while still having a
security margin similar to AES, and in the case of Speck128 also the
same available security levels. Unfortunately, Speck does have some
political baggage attached -- it's an NSA designed cipher, and was
rejected from an ISO standard (though for context, as far as I know none
of the above-mentioned alternatives are ISO standards either).
Nevertheless, we believe it is a good solution to the problem from a
technical perspective.
Certain algorithms constructed from ChaCha or the ChaCha permutation,
such as MEM (Masked Even-Mansour) or HPolyC, may also meet our
performance requirements. However, these are new constructions that
need more time to receive the cryptographic review and acceptance needed
to be confident in their security. HPolyC hasn't been published yet,
and we are concerned that MEM makes stronger assumptions about the
underlying permutation than the ChaCha stream cipher does. In contrast,
the XTS mode of operation is relatively well accepted, and Speck has
over 70 cryptanalysis papers. Of course, these ChaCha-based algorithms
can still be added later if they become ready.
The best known attack on Speck128/256 is a differential cryptanalysis
attack on 25 of 34 rounds with 2^253 time complexity and 2^125 chosen
plaintexts, i.e. only marginally faster than brute force. There is no
known attack on the full 34 rounds.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:49 +0000 (15:51 -0700)]
fscrypt: only derive the needed portion of the key
Currently the key derivation function in fscrypt uses the master key
length as the amount of output key material to derive. This works, but
it means we can waste time deriving more key material than is actually
used, e.g. most commonly, deriving 64 bytes for directories which only
take a 32-byte AES-256-CTS-CBC key. It also forces us to validate that
the master key length is a multiple of AES_BLOCK_SIZE, which wouldn't
otherwise be necessary.
Fix it to only derive the needed length key.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:48 +0000 (15:51 -0700)]
fscrypt: separate key lookup from key derivation
Refactor the confusingly-named function 'validate_user_key()' into a new
function 'find_and_derive_key()' which first finds the keyring key, then
does the key derivation. Among other benefits this avoids the strange
behavior we had previously where if key derivation failed for some
reason, then we would fall back to the alternate key prefix. Now, we'll
only fall back to the alternate key prefix if a valid key isn't found.
This patch also improves the warning messages that are logged when the
keyring key's payload is invalid.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:47 +0000 (15:51 -0700)]
fscrypt: use a common logging function
Use a common function for fscrypt warning and error messages so that all
the messages are consistently ratelimited, include the "fscrypt:"
prefix, and include the filesystem name if applicable.
Also fix up a few of the log messages to be more descriptive.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:46 +0000 (15:51 -0700)]
fscrypt: remove internal key size constants
With one exception, the internal key size constants such as
FS_AES_256_XTS_KEY_SIZE are only used for the 'available_modes' array,
where they really only serve to obfuscate what the values are. Also
some of the constants are unused, and the key sizes tend to be in the
names of the algorithms anyway. In the past these values were also
misused, e.g. we used to have FS_AES_256_XTS_KEY_SIZE in places that
technically should have been FS_MAX_KEY_SIZE.
The exception is that FS_AES_128_ECB_KEY_SIZE is used for key
derivation. But it's more appropriate to use
FS_KEY_DERIVATION_NONCE_SIZE for that instead.
Thus, just put the sizes directly in the 'available_modes' array.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:45 +0000 (15:51 -0700)]
fscrypt: remove unnecessary check for non-logon key type
We're passing 'key_type_logon' to request_key(), so the found key is
guaranteed to be of type "logon". Thus, there is no reason to check
later that the key is really a "logon" key.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:44 +0000 (15:51 -0700)]
fscrypt: make fscrypt_operations.max_namelen an integer
Now ->max_namelen() is only called to limit the filename length when
adding NUL padding, and only for real filenames -- not symlink targets.
It also didn't give the correct length for symlink targets anyway since
it forgot to subtract 'sizeof(struct fscrypt_symlink_data)'.
Thus, change ->max_namelen from a function to a simple 'unsigned int'
that gives the filesystem's maximum filename length.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:43 +0000 (15:51 -0700)]
fscrypt: drop empty name check from fname_decrypt()
fname_decrypt() is validating that the encrypted filename is nonempty.
However, earlier a stronger precondition was already enforced: the
encrypted filename must be at least 16 (FS_CRYPTO_BLOCK_SIZE) bytes.
Drop the redundant check for an empty filename.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:42 +0000 (15:51 -0700)]
fscrypt: drop max_namelen check from fname_decrypt()
fname_decrypt() returns an error if the input filename is longer than
the inode's ->max_namelen() as given by the filesystem. But, this
doesn't actually make sense because the filesystem provided the input
filename in the first place, where it was subject to the filesystem's
limits. And fname_decrypt() has no internal limit itself.
Thus, remove this unnecessary check.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:41 +0000 (15:51 -0700)]
fscrypt: don't special-case EOPNOTSUPP from fscrypt_get_encryption_info()
In fscrypt_setup_filename(), remove the unnecessary check for
fscrypt_get_encryption_info() returning EOPNOTSUPP. There's no reason
to handle this error differently from any other. I think there may have
been some confusion because the "notsupp" version of
fscrypt_get_encryption_info() returns EOPNOTSUPP -- but that's not
applicable from inside fs/crypto/.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:40 +0000 (15:51 -0700)]
fscrypt: don't clear flags on crypto transform
fscrypt is clearing the flags on the crypto_skcipher it allocates for
each inode. But, this is unnecessary and may cause problems in the
future because it will even clear flags that are meant to be internal to
the crypto API, e.g. CRYPTO_TFM_NEED_KEY.
Remove the unnecessary flag clearing.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:39 +0000 (15:51 -0700)]
fscrypt: remove stale comment from fscrypt_d_revalidate()
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:38 +0000 (15:51 -0700)]
fscrypt: remove error messages for skcipher_request_alloc() failure
skcipher_request_alloc() can only fail due to lack of memory, and in
that case the memory allocator will have already printed a detailed
error message. Thus, remove the redundant error messages from fscrypt.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:37 +0000 (15:51 -0700)]
fscrypt: remove unnecessary NULL check when allocating skcipher
crypto_alloc_skcipher() returns an ERR_PTR() on failure, not NULL.
Remove the unnecessary check for NULL.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Mon, 30 Apr 2018 22:51:36 +0000 (15:51 -0700)]
fscrypt: clean up after fscrypt_prepare_lookup() conversions
Now that all filesystems have been converted to use
fscrypt_prepare_lookup(), we can remove the fscrypt_set_d_op() and
fscrypt_set_encrypted_dentry() functions as well as un-export
fscrypt_d_ops.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 20 Apr 2018 23:30:02 +0000 (16:30 -0700)]
fscrypt: use unbound workqueue for decryption
Improve fscrypt read performance by switching the decryption workqueue
from bound to unbound. With the bound workqueue, when multiple bios
completed on the same CPU, they were decrypted on that same CPU. But
with the unbound queue, they are now decrypted in parallel on any CPU.
Although fscrypt read performance can be tough to measure due to the
many sources of variation, this change is most beneficial when
decryption is slow, e.g. on CPUs without AES instructions. For example,
I timed tarring up encrypted directories on f2fs. On x86 with AES-NI
instructions disabled, the unbound workqueue improved performance by
about 25-35%, using 1 to NUM_CPUs jobs with 4 or 8 CPUs available. But
with AES-NI enabled, performance was unchanged to within ~2%.
I also did the same test on a quad-core ARM CPU using xts-speck128-neon
encryption. There performance was usually about 10% better with the
unbound workqueue, bringing it closer to the unencrypted speed.
The unbound workqueue may be worse in some cases due to worse locality,
but I think it's still the better default. dm-crypt uses an unbound
workqueue by default too, so this change makes fscrypt match.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Jaegeuk Kim [Thu, 31 May 2018 17:20:48 +0000 (10:20 -0700)]
f2fs: run fstrim asynchronously if runtime discard is on
We don't need to wait for whole bunch of discard candidates in fstrim, since
runtime discard will issue them in idle time.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Tue, 10 Apr 2018 07:43:09 +0000 (15:43 +0800)]
f2fs: turn down IO priority of discard from background
In order to avoid interfering normal r/w IO, let's turn down IO
priority of discard issued from background.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Mon, 9 Apr 2018 02:25:23 +0000 (10:25 +0800)]
f2fs: don't split checkpoint in fstrim
Now, we issue discard asynchronously in separated thread instead of in
checkpoint, after that, we won't encounter long latency in checkpoint
due to huge number of synchronous discard command handling, so, we don't
need to split checkpoint to do trim in batch, merge it and obsolete
related sysfs entry.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Tue, 29 May 2018 16:58:42 +0000 (09:58 -0700)]
f2fs: issue discard commands proactively in high fs utilization
In the high utilization like over 80%, we don't expect huge # of large discard
commands, but do many small pending discards which affects FTL GCs a lot.
Let's issue them in that case.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sat, 26 May 2018 01:02:58 +0000 (18:02 -0700)]
f2fs: add fsync_mode=nobarrier for non-atomic files
For non-atomic files, this patch adds an option to give nobarrier which
doesn't issue flush commands to the device.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Thu, 24 May 2018 20:57:26 +0000 (13:57 -0700)]
f2fs: let fstrim issue discard commands in lower priority
The fstrim gathers huge number of large discard commands, and tries to issue
without IO awareness, which results in long user-perceive IO latencies on
READ, WRITE, and FLUSH in UFS. We've observed some of commands take several
seconds due to long discard latency.
This patch limits the maximum size to 2MB per candidate, and check IO congestion
when issuing them to disk.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Fri, 4 May 2018 06:26:02 +0000 (23:26 -0700)]
f2fs: avoid fsync() failure caused by EAGAIN in writepage()
pageout() in MM traslates EAGAIN, so calls handle_write_error()
-> mapping_set_error() -> set_bit(AS_EIO, ...).
file_write_and_wait_range() will see EIO error, which is critical
to return value of fsync() followed by atomic_write failure to user.
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Thu, 12 Apr 2018 06:09:04 +0000 (23:09 -0700)]
f2fs: clear PageError on writepage - part 2
This patch clears PageError in some pages tagged by read path, but when we
write the pages with valid contents, writepage should clear the bit likewise
ext4.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sat, 21 Apr 2018 06:44:59 +0000 (23:44 -0700)]
f2fs: check cap_resource only for data blocks
This patch changes the rule to check cap_resource for data blocks, not inode
or node blocks in order to avoid selinux denial.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sat, 21 Apr 2018 02:29:52 +0000 (19:29 -0700)]
Revert "f2fs: introduce f2fs_set_page_dirty_nobuffer"
This patch reverts copied f2fs_set_page_dirty_nobuffer to use generic function
for stability.
This reverts commit
fe76b796fc5194cc3d57265002e3a748566d073f.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Thu, 12 Apr 2018 06:09:04 +0000 (23:09 -0700)]
f2fs: clear PageError on writepage
This patch clears PageError in some pages tagged by read path, but when we
write the pages with valid contents, writepage should clear the bit likewise
ext4.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Eric Biggers [Wed, 18 Apr 2018 22:48:42 +0000 (15:48 -0700)]
f2fs: call unlock_new_inode() before d_instantiate()
xfstest generic/429 sometimes hangs on f2fs, caused by a thread being
unable to take a directory's i_rwsem for write in vfs_rmdir(). In the
test, one thread repeatedly creates and removes a directory, and other
threads repeatedly look up a file in the directory. The bug is that
f2fs_mkdir() calls d_instantiate() before unlock_new_inode(), resulting
in the directory inode being exposed to lookups before it has been fully
initialized. And with CONFIG_DEBUG_LOCK_ALLOC, unlock_new_inode()
reinitializes ->i_rwsem, corrupting its state when it is already held.
Fix it by calling unlock_new_inode() before d_instantiate(). This
matches what other filesystems do.
Fixes:
57397d86c62d ("f2fs: add inode operations for special inodes")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Eric Biggers [Wed, 18 Apr 2018 18:09:48 +0000 (11:09 -0700)]
f2fs: refactor read path to allow multiple postprocessing steps
Currently f2fs's ->readpage() and ->readpages() assume that either the
data undergoes no postprocessing, or decryption only. But with
fs-verity, there will be an additional authenticity verification step,
and it may be needed either by itself, or combined with decryption.
To support this, store a 'struct bio_post_read_ctx' in ->bi_private
which contains a work struct, a bitmask of postprocessing steps that are
enabled, and an indicator of the current step. The bio completion
routine, if there was no I/O error, enqueues the first postprocessing
step. When that completes, it continues to the next step. Pages that
fail any postprocessing step have PageError set. Once all steps have
completed, pages without PageError set are set Uptodate, and all pages
are unlocked.
Also replace f2fs_encrypted_file() with a new function
f2fs_post_read_required() in places like direct I/O and garbage
collection that really should be testing whether the file needs special
I/O processing, not whether it is encrypted specifically.
This may also be useful for other future f2fs features such as
compression.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Eric Biggers [Wed, 18 Apr 2018 18:09:47 +0000 (11:09 -0700)]
fscrypt: allow synchronous bio decryption
Currently, fscrypt provides fscrypt_decrypt_bio_pages() which decrypts a
bio's pages asynchronously, then unlocks them afterwards. But, this
assumes that decryption is the last "postprocessing step" for the bio,
so it's incompatible with additional postprocessing steps such as
authenticity verification after decryption.
Therefore, rename the existing fscrypt_decrypt_bio_pages() to
fscrypt_enqueue_decrypt_bio(). Then, add fscrypt_decrypt_bio() which
decrypts the pages in the bio synchronously without unlocking the pages,
nor setting them Uptodate; and add fscrypt_enqueue_decrypt_work(), which
enqueues work on the fscrypt_read_workqueue. The new functions will be
used by filesystems that support both fscrypt and fs-verity.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Fri, 30 Mar 2018 05:50:41 +0000 (22:50 -0700)]
f2fs: remain written times to update inode during fsync
This fixes xfstests/generic/392.
The failure was caused by different times between 1) one marked in the last
fsync(2) call and 2) the other given by roll-forward recovery after power-cut.
The reason was that we skipped updating inode block at 1), since its i_size
was recoverable along with 4KB-aligned data writes, which was fixed by:
"f2fs: fix a wrong condition in f2fs_skip_inode_update"
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Yunlong Song [Mon, 2 Apr 2018 12:22:20 +0000 (20:22 +0800)]
f2fs: make assignment of t->dentry_bitmap more readable
In make_dentry_ptr_block, it is confused with "&" for t->dentry_bitmap
but without "&" for t->dentry, so delete "&" to make code more readable.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sat, 31 Mar 2018 00:58:13 +0000 (17:58 -0700)]
f2fs: truncate preallocated blocks in error case
If write is failed, we must deallocate the blocks that we couldn't write.
Cc: stable@vger.kernel.org
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Junling Zheng [Thu, 29 Mar 2018 11:27:12 +0000 (19:27 +0800)]
f2fs: fix a wrong condition in f2fs_skip_inode_update
Fix commit
97dd26ad8347 (f2fs: fix wrong AUTO_RECOVER condition).
We should use ~PAGE_MASK to determine whether i_size is aligned to
the f2fs's block size or not.
Signed-off-by: Junling Zheng <zhengjunling@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Eric Biggers [Wed, 28 Mar 2018 18:15:09 +0000 (11:15 -0700)]
f2fs: reserve bits for fs-verity
Reserve an F2FS feature flag and inode flag for fs-verity. This is an
in-development feature that is planned be discussed at LSF/MM 2018 [1].
It will provide file-based integrity and authenticity for read-only
files. Most code will be in a filesystem-independent module, with
smaller changes needed to individual filesystems that opt-in to
supporting the feature. An early prototype supporting F2FS is available
[2]. Reserving the F2FS on-disk bits for fs-verity will prevent users
of the prototype from conflicting with other new F2FS features.
Note that we're reserving the inode flag in f2fs_inode.i_advise, which
isn't really appropriate since it's not a hint or advice. But
->i_advise is already being used to hold the 'encrypt' flag; and F2FS's
->i_flags uses the generic FS_* values, so it seems ->i_flags can't be
used for an F2FS-specific flag without additional work to remove the
assumption that ->i_flags uses the generic flags namespace.
[1] https://marc.info/?l=linux-fsdevel&m=
151690752225644
[2] https://git.kernel.org/pub/scm/linux/kernel/git/mhalcrow/linux.git/log/?h=fs-verity-dev
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Yunlei He [Mon, 26 Mar 2018 09:32:23 +0000 (17:32 +0800)]
f2fs: Add a segment type check in inplace write
This patch add a segment type check in IPU, in
case of something wrong with blkadd in dnode.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Yunlong Song [Thu, 22 Mar 2018 02:08:40 +0000 (10:08 +0800)]
f2fs: no need to initialize zero value for GFP_F2FS_ZERO
Since f2fs_inode_info is allocated with flag GFP_F2FS_ZERO, so we do not
need to initialize zero value for its member any more.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Tue, 20 Mar 2018 15:08:30 +0000 (23:08 +0800)]
f2fs: don't track new nat entry in nat set
Nat entry set is used only in checkpoint(), and during checkpoint() we
won't flush new nat entry with unallocated address, so we don't need to
add new nat entry into nat set, then nat_entry_set::entry_cnt can
indicate actual entry count we need to flush in checkpoint().
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Tue, 20 Mar 2018 15:08:29 +0000 (23:08 +0800)]
f2fs: clean up with F2FS_BLK_ALIGN
Clean up F2FS_BYTES_TO_BLK(x + F2FS_BLKSIZE - 1) with F2FS_BLK_ALIGN(x).
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Yunlei He [Thu, 8 Mar 2018 08:29:13 +0000 (16:29 +0800)]
f2fs: check blkaddr more accuratly before issue a bio
This patch check blkaddr more accuratly before issue a
write or read bio.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Ritesh Harjani [Fri, 16 Mar 2018 13:23:53 +0000 (18:53 +0530)]
f2fs: Set GF_NOFS in read_cache_page_gfp while doing f2fs_quota_read
Quota code itself is serializing the operations by taking mutex_lock.
It seems a below deadlock can happen if GF_NOFS is not used in
f2fs_quota_read
__switch_to+0x88
__schedule+0x5b0
schedule+0x78
schedule_preempt_disabled+0x20
__mutex_lock_slowpath+0xdc //mutex owner is itself
mutex_lock+0x2c
dquot_commit+0x30 //mutex_lock(&dqopt->dqio_mutex);
dqput+0xe0
__dquot_drop+0x80
dquot_drop+0x48
f2fs_evict_inode+0x218
evict+0xa8
dispose_list+0x3c
prune_icache_sb+0x58
super_cache_scan+0xf4
do_shrink_slab+0x208
shrink_slab.part.40+0xac
shrink_zone+0x1b0
do_try_to_free_pages+0x25c
try_to_free_pages+0x164
__alloc_pages_nodemask+0x534
do_read_cache_page+0x6c
read_cache_page+0x14
f2fs_quota_read+0xa4
read_blk+0x54
find_tree_dqentry+0xe4
find_tree_dqentry+0xb8
find_tree_dqentry+0xb8
find_tree_dqentry+0xb8
qtree_read_dquot+0x68
v2_read_dquot+0x24
dquot_acquire+0x5c // mutex_lock(&dqopt->dqio_mutex);
dqget+0x238
__dquot_initialize+0xd4
dquot_initialize+0x10
dquot_file_open+0x34
f2fs_file_open+0x6c
do_dentry_open+0x1e4
vfs_open+0x6c
path_openat+0xa20
do_filp_open+0x4c
do_sys_open+0x178
Signed-off-by: Ritesh Harjani <riteshh@codeaurora.org>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Sheng Yong [Thu, 15 Mar 2018 10:51:42 +0000 (18:51 +0800)]
f2fs: introduce a new mount option test_dummy_encryption
This patch introduces a new mount option `test_dummy_encryption'
to allow fscrypt to create a fake fscrypt context. This is used
by xfstests.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Sheng Yong [Thu, 15 Mar 2018 10:51:41 +0000 (18:51 +0800)]
f2fs: introduce F2FS_FEATURE_LOST_FOUND feature
This patch introduces a new feature, F2FS_FEATURE_LOST_FOUND, which
is set by mkfs. mkfs creates a directory named lost+found, which saves
unreachable files. If fsck finds a file which has no parent, or its
parent is removed by fsck, the file will be placed under lost+found
directory by fsck.
lost+found directory could not be encrypted. As a result, the root
directory cannot be encrypted too. So if LOST_FOUND feature is enabled,
let's avoid to encrypt root directory.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Qiuyang Sun [Tue, 13 Mar 2018 11:42:50 +0000 (19:42 +0800)]
f2fs: release locks before return in f2fs_ioc_gc_range()
Currently, we will leave the kernel with locks still held when the gc_range
is invalid. This patch fixes the bug.
Signed-off-by: Qiuyang Sun <sunqiuyang@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sat, 10 Mar 2018 01:42:28 +0000 (17:42 -0800)]
f2fs: align memory boundary for bitops
For example, in arm64, free_nid_bitmap should be aligned to word size in order
to use bit operations.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Fri, 9 Mar 2018 06:24:22 +0000 (14:24 +0800)]
f2fs: remove unneeded set_cold_node()
When setting COLD_BIT_SHIFT flag in node block, we only need to call
set_cold_node() in new_node_page() and recover_inode_page() during
node page initialization. So remove unneeded set_cold_node() in other
places.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Hyunchul Lee [Thu, 8 Mar 2018 10:34:38 +0000 (19:34 +0900)]
f2fs: add nowait aio support
This patch adds nowait aio support[1].
Return EAGAIN if any of the following checks fail for direct I/O:
- i_rwsem is not lockable
- Blocks are not allocated at the write location
And xfstests generic/471 is passed.
[1]: 6be96d "Introduce RWF_NOWAIT and FMODE_AIO_NOWAIT"
Signed-off-by: Hyunchul Lee <cheol.lee@lge.com>
Reviewed-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Thu, 8 Mar 2018 06:22:56 +0000 (14:22 +0800)]
f2fs: wrap all options with f2fs_sb_info.mount_opt
This patch merges miscellaneous mount options into struct f2fs_mount_info,
After this patch, once we add new mount option, we don't need to worry
about recovery of it in remount_fs(), since we will recover the
f2fs_sb_info.mount_opt including all options.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Yunlei He [Wed, 7 Mar 2018 08:22:50 +0000 (16:22 +0800)]
f2fs: Don't overwrite all types of node to keep node chain
Currently, we enable node SSR by default, and mixed
different types of node segment to do SSR more intensively.
Although reuse warm node is not allowed, warm node chain
will be destroyed by errors introduced by other types
node chain. So we'd better forbid reusing all types
of node to keep warm node chain.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Junling Zheng [Wed, 7 Mar 2018 04:07:49 +0000 (12:07 +0800)]
f2fs: introduce mount option for fsync mode
Commit "
0a007b97aad6"(f2fs: recover directory operations by fsync)
fixed xfstest generic/342 case, but it also increased the written
data and caused the performance degradation. In most cases, there's
no need to do so heavy fsync actually.
So we introduce new mount option "fsync_mode={posix,strict}" to
control the policy of fsync. "fsync_mode=posix" is set by default,
and means that f2fs uses a light fsync, which follows POSIX semantics.
And "fsync_mode=strict" means that it's a heavy fsync, which behaves
in line with xfs, ext4 and btrfs, where generic/342 will pass, but
the performance will regress.
Signed-off-by: Junling Zheng <zhengjunling@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Thu, 1 Mar 2018 15:40:32 +0000 (23:40 +0800)]
f2fs: fix to restore old mount option in ->remount_fs
This patch fixes to restore old mount option once we encounter failure
in ->remount_fs.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Thu, 1 Mar 2018 15:40:31 +0000 (23:40 +0800)]
f2fs: wrap sb_rdonly with f2fs_readonly
Use f2fs_readonly to wrap sb_rdonly for cleanup, and spread it in
all places.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Fri, 9 Mar 2018 04:47:33 +0000 (20:47 -0800)]
f2fs: avoid selinux denial on CAP_SYS_RESOURCE
This fixes CAP_SYS_RESOURCE denial of selinux when using resgid, since it
seems selinux reports it at the first place, but mostly we don't need to
check this condition first.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Wed, 28 Feb 2018 09:07:27 +0000 (17:07 +0800)]
f2fs: support hot file extension
This patch supports to recognize hot file extension in f2fs, so that we
can allocate proper hot segment location for its data, which can lead to
better hot/cold seperation in filesystem.
In addition, we changes a bit on query/add/del operation method for
extension_list sysfs entry as below:
- Query: cat /sys/fs/f2fs/<disk>/extension_list
- Add: echo 'extension' > /sys/fs/f2fs/<disk>/extension_list
- Del: echo '!extension' > /sys/fs/f2fs/<disk>/extension_list
- Add: echo '[h/c]extension' > /sys/fs/f2fs/<disk>/extension_list
- Del: echo '[h/c]!extension' > /sys/fs/f2fs/<disk>/extension_list
- [h] means add/del hot file extension
- [c] means add/del cold file extension
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Tue, 27 Feb 2018 14:45:24 +0000 (22:45 +0800)]
f2fs: fix to avoid race in between atomic write and background GC
Sqlite user Background GC
- move_data_block
: move page #1
- f2fs_is_atomic_file
- f2fs_ioc_start_atomic_write
- f2fs_ioc_commit_atomic_write
- commit_inmem_pages
: commit page #1 & set node #2 dirty
- f2fs_submit_page_write
- f2fs_update_data_blkaddr
- set_page_dirty
: set node #2 dirty
- f2fs_do_sync_file
- fsync_node_pages
: commit node #1 & node #2, then sudden power-cut
In a race case, we may check FI_ATOMIC_FILE flag before starting atomic
write flow, then we will commit meta data before data with reversed
order, after a sudden pow-cut, database transaction will be inconsistent.
So we'd better to exclude gc/atomic_write to each other by using lock
instead of flag checking.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Mon, 26 Feb 2018 23:40:30 +0000 (15:40 -0800)]
f2fs: do gc in greedy mode for whole range if gc_urgent mode is set
Otherwise, f2fs conducts GC on 8GB range only based on slow cost-benefit.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Fri, 23 Feb 2018 07:30:55 +0000 (23:30 -0800)]
f2fs: issue discard aggressively in the gc_urgent mode
This patch avoids to skip discard commands when user sets gc_urgent mode.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sun, 25 Feb 2018 09:04:57 +0000 (01:04 -0800)]
f2fs: set readdir_ra by default
It gives general readdir improvement.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Thu, 22 Feb 2018 22:09:30 +0000 (14:09 -0800)]
f2fs: add auto tuning for small devices
If f2fs is running on top of very small devices, it's worth to avoid abusing
free LBAs. In order to achieve that, this patch introduces some parameter
tuning.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Sun, 18 Feb 2018 16:50:49 +0000 (08:50 -0800)]
f2fs: add mount option for segment allocation policy
This patch adds an mount option, "alloc_mode=%s" having two options, "default"
and "reuse".
In "alloc_mode=reuse" case, f2fs starts to allocate segments from 0'th segment
all the time to reassign segments. It'd be useful for small-sized eMMC parts.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Mon, 26 Feb 2018 17:19:47 +0000 (09:19 -0800)]
f2fs: don't stop GC if GC is contended
Let's do GC as much as possible, while gc_urgent is set.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Mon, 26 Feb 2018 14:04:13 +0000 (22:04 +0800)]
f2fs: expose extension_list sysfs entry
This patch adds a sysfs entry 'extension_list' to support
query/add/del item in extension list.
Query:
cat /sys/fs/f2fs/<device>/extension_list
Add:
echo 'extension' > /sys/fs/f2fs/<device>/extension_list
Del:
echo '!extension' > /sys/fs/f2fs/<device>/extension_list
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Sun, 25 Feb 2018 15:38:21 +0000 (23:38 +0800)]
f2fs: fix to set KEEP_SIZE bit in f2fs_zero_range
As Jayashree Mohan reported:
A simple workload to reproduce this would be :
1. create foo
2. Write (8K - 16K) // foo size = 16K now
3. fsync()
4. falloc zero_range , keep_size (
4202496 -
4210688) // foo size must be 16K
5. fdatasync()
Crash now
On recovery, we see that the file size is
4210688 and not 16K, which
violates the semantics of keep_size flag. We have a test case to
reproduce this using CrashMonkey on 4.15 kernel. Try this out by
simply running :
./c_harness -f /dev/sda -d /dev/cow_ram0 -t f2fs -e 102400 -P -v
tests/generic_468_zero.so
The root cause is that we miss to set KEEP_SIZE bit correctly in zero_range
when zeroing block cross EOF with FALLOC_FL_KEEP_SIZE, let's fix this
missing case.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Sun, 11 Feb 2018 14:53:20 +0000 (22:53 +0800)]
f2fs: introduce sb_lock to make encrypt pwsalt update exclusive
f2fs_super_block.encrypt_pw_salt can be udpated and persisted
concurrently, result in getting different pwsalt in separated
threads, so let's introduce sb_lock to exclude concurrent
accessers.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Colin Ian King [Wed, 21 Feb 2018 18:13:40 +0000 (18:13 +0000)]
f2fs: remove redundant initialization of pointer 'p'
Pointer p is initialized with a value that is never read and is later
re-assigned a new value, hence the initialization is redundant and can
be removed.
Cleans up clang warning:
fs/f2fs/extent_cache.c:463:19: warning: Value stored to 'p' during
its initialization is never read
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Gao Xiang [Sat, 10 Feb 2018 04:12:51 +0000 (12:12 +0800)]
f2fs: flush cp pack except cp pack 2 page at first
Previously, we attempt to flush the whole cp pack in a single bio,
however, when suddenly powering off at this time, we could get into
an extreme scenario that cp pack 1 page and cp pack 2 page are updated
and latest, but payload or current summaries are still partially
outdated. (see reliable write in the UFS specification)
This patch submits the whole cp pack except cp pack 2 page at first,
and then writes the cp pack 2 page with an extra independent
bio with pre-io barrier.
Signed-off-by: Gao Xiang <gaoxiang25@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Sheng Yong [Tue, 6 Feb 2018 04:31:17 +0000 (12:31 +0800)]
f2fs: clean up f2fs_sb_has_xxx functions
This patch introduces F2FS_FEATURE_FUNCS to clean up the definitions of
different f2fs_sb_has_xxx functions.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Tiezhu Yang [Tue, 6 Feb 2018 00:21:45 +0000 (08:21 +0800)]
f2fs: remove redundant check of page type when submit bio
This patch removes redundant check of page type when submit bio to
make the logic more clear.
Signed-off-by: Tiezhu Yang <kernelpatch@126.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Sat, 3 Feb 2018 09:44:39 +0000 (17:44 +0800)]
f2fs: fix to handle looped node chain during recovery
There is no checksum in node block now, so bit-transition from hardware
can make node_footer.next_blkaddr being corrupted w/o any detection,
result in node chain becoming looped one.
For this condition, during recovery, in order to avoid running into dead
loop, let's detect it and just skip out.
Signed-off-by: Yunlei He <heyunlei@huawei.com>
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Jaegeuk Kim [Thu, 8 Feb 2018 01:01:48 +0000 (17:01 -0800)]
f2fs: handle quota for orphan inodes
This is to detect dquot_initialize errors early from evict_inode
for orphan inodes.
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Hyunchul Lee [Wed, 31 Jan 2018 02:36:58 +0000 (11:36 +0900)]
f2fs: support passing down write hints to block layer with F2FS policy
Add 'whint_mode=fs-based' mount option. In this mode, F2FS passes
down write hints with its policy.
* whint_mode=fs-based. F2FS passes down hints with its policy.
User F2FS Block
---- ---- -----
META WRITE_LIFE_MEDIUM;
HOT_NODE WRITE_LIFE_NOT_SET
WARM_NODE "
COLD_NODE WRITE_LIFE_NONE
ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
extension list " "
-- buffered io
WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_LONG
WRITE_LIFE_NONE " "
WRITE_LIFE_MEDIUM " "
WRITE_LIFE_LONG " "
-- direct io
WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
WRITE_LIFE_NONE " WRITE_LIFE_NONE
WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
WRITE_LIFE_LONG " WRITE_LIFE_LONG
Many thanks to Chao Yu and Jaegeuk Kim for comments to
implement this patch.
Signed-off-by: Hyunchul Lee <cheol.lee@lge.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Hyunchul Lee [Wed, 31 Jan 2018 02:36:57 +0000 (11:36 +0900)]
f2fs: support passing down write hints given by users to block layer
Add the 'whint_mode' mount option that controls which write
hints are passed down to block layer. There are "off" and
"user-based" mode. The default mode is "off".
1) whint_mode=off. F2FS only passes down WRITE_LIFE_NOT_SET.
2) whint_mode=user-based. F2FS tries to pass down hints given
by users.
User F2FS Block
---- ---- -----
META WRITE_LIFE_NOT_SET
HOT_NODE "
WARM_NODE "
COLD_NODE "
ioctl(COLD) COLD_DATA WRITE_LIFE_EXTREME
extension list " "
-- buffered io
WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
WRITE_LIFE_NONE " "
WRITE_LIFE_MEDIUM " "
WRITE_LIFE_LONG " "
-- direct io
WRITE_LIFE_EXTREME COLD_DATA WRITE_LIFE_EXTREME
WRITE_LIFE_SHORT HOT_DATA WRITE_LIFE_SHORT
WRITE_LIFE_NOT_SET WARM_DATA WRITE_LIFE_NOT_SET
WRITE_LIFE_NONE " WRITE_LIFE_NONE
WRITE_LIFE_MEDIUM " WRITE_LIFE_MEDIUM
WRITE_LIFE_LONG " WRITE_LIFE_LONG
Many thanks to Chao Yu and Jaegeuk Kim for comments to
implement this patch.
Signed-off-by: Hyunchul Lee <cheol.lee@lge.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
[Jaegeuk Kim: avoid build warning]
[Chao Yu: fix to restore whint_mode in ->remount_fs]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Wed, 31 Jan 2018 01:30:34 +0000 (09:30 +0800)]
f2fs: fix to clear CP_TRIMMED_FLAG
Once CP_TRIMMED_FLAG is set, after a reboot, we will never issue discard
before LBA becomes invalid again, fix it by clearing the flag in
checkpoint without CP_TRIMMED reason.
Fixes:
1f43e2ad7bff ("f2fs: introduce CP_TRIMMED_FLAG to avoid unneeded discard")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Thu, 25 Jan 2018 11:40:08 +0000 (19:40 +0800)]
f2fs: support large nat bitmap
Previously, we will store all nat version bitmap in checkpoint pack block,
so our total node entry number has a limitation which caused total node
number can not exceed (3900 * 8) block * 455 node/block =
14196000. So
that once user wants to create more nodes in large size image, it becomes
a bottleneck, that's unreasonable.
This patch detects the new layout of nat/sit version bitmap in image in
order to enable supporting large nat bitmap.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Sat, 27 Jan 2018 09:29:49 +0000 (17:29 +0800)]
f2fs: fix to check extent cache in f2fs_drop_extent_tree
If noextent_cache mount option is on, we will never initialize extent tree
in inode, but still we're going to access it in f2fs_drop_extent_tree,
result in kernel panic as below:
BUG: unable to handle kernel NULL pointer dereference at
0000000000000038
IP: _raw_write_lock+0xc/0x30
Call Trace:
? f2fs_drop_extent_tree+0x41/0x70 [f2fs]
f2fs_fallocate+0x5a0/0xdd0 [f2fs]
? common_file_perm+0x47/0xc0
? apparmor_file_permission+0x1a/0x20
vfs_fallocate+0x15b/0x290
SyS_fallocate+0x44/0x70
do_syscall_64+0x6e/0x160
entry_SYSCALL64_slow_path+0x25/0x25
This patch fixes to check extent cache status before using in
f2fs_drop_extent_tree.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Sat, 27 Jan 2018 09:29:48 +0000 (17:29 +0800)]
f2fs: restrict inline_xattr_size configuration
This patch limits to enable inline_xattr_size mount option only if
both extra_attr and flexible_inline_xattr feature is on in current
image.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Yunlong Song [Mon, 29 Jan 2018 03:37:45 +0000 (11:37 +0800)]
f2fs: fix heap mode to reset it back
Commit
7a20b8a61eff81bdb7097a578752a74860e9d142 ("f2fs: allocate node
and hot data in the beginning of partition") introduces another mount
option, heap, to reset it back. But it does not do anything for heap
mode, so fix it.
Cc: stable@vger.kernel.org
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Sheng Yong [Mon, 29 Jan 2018 11:13:15 +0000 (19:13 +0800)]
f2fs: fix potential corruption in area before F2FS_SUPER_OFFSET
sb_getblk does not guarantee the buffer head is uptodate. If bh is not
uptodate, the data (may be used as boot code) in area before
F2FS_SUPER_OFFSET may get corrupted when super block is committed.
Signed-off-by: Sheng Yong <shengyong1@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Eric Biggers [Fri, 19 Jan 2018 21:45:24 +0000 (13:45 -0800)]
fscrypt: fix build with pre-4.6 gcc versions
gcc versions prior to 4.6 require an extra level of braces when using a
designated initializer for a member in an anonymous struct or union.
This caused a compile error with the 'struct qstr' initialization in
__fscrypt_encrypt_symlink().
Fix it by using QSTR_INIT().
Reported-by: Andrew Morton <akpm@linux-foundation.org>
Fixes:
76e81d6d5048 ("fscrypt: new helper functions for ->symlink()")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:30:08 +0000 (23:30 -0500)]
fscrypt: fix up fscrypt_fname_encrypted_size() for internal use
Filesystems don't need fscrypt_fname_encrypted_size() anymore, so
unexport it and move it to fscrypt_private.h.
We also never calculate the encrypted size of a filename without having
the fscrypt_info present since it is needed to know the amount of
NUL-padding which is determined by the encryption policy, and also we
will always truncate the NUL-padding to the maximum filename length.
Therefore, also make fscrypt_fname_encrypted_size() assume that the
fscrypt_info is present, and make it truncate the returned length to the
specified max_len.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:30:08 +0000 (23:30 -0500)]
fscrypt: define fscrypt_fname_alloc_buffer() to be for presented names
Previously fscrypt_fname_alloc_buffer() was used to allocate buffers for
both presented (decrypted or encoded) and encrypted filenames. That was
confusing, because it had to allocate the worst-case size for either,
e.g. including NUL-padding even when it was meaningless.
But now that fscrypt_setup_filename() no longer calls it, it is only
used in the ->get_link() and ->readdir() paths, which specifically want
a buffer for presented filenames. Therefore, switch the behavior over
to allocating the buffer for presented filenames only.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:30:08 +0000 (23:30 -0500)]
fscrypt: calculate NUL-padding length in one place only
Currently, when encrypting a filename (either a real filename or a
symlink target) we calculate the amount of NUL-padding twice: once
before encryption and once during encryption in fname_encrypt(). It is
needed before encryption to allocate the needed buffer size as well as
calculate the size the symlink target will take up on-disk before
creating the symlink inode. Calculating the size during encryption as
well is redundant.
Remove this redundancy by always calculating the exact size beforehand,
and making fname_encrypt() just add as much NUL padding as is needed to
fill the output buffer.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:30:08 +0000 (23:30 -0500)]
fscrypt: move fscrypt_symlink_data to fscrypt_private.h
Now that all filesystems have been converted to use the symlink helper
functions, they no longer need the declaration of 'struct
fscrypt_symlink_data'. Move it from fscrypt.h to fscrypt_private.h.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:30:08 +0000 (23:30 -0500)]
fscrypt: remove fscrypt_fname_usr_to_disk()
fscrypt_fname_usr_to_disk() sounded very generic but was actually only
used to encrypt symlinks. Remove it now that all filesystems have been
switched over to fscrypt_encrypt_symlink().
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:26:49 +0000 (23:26 -0500)]
f2fs: switch to fscrypt_get_symlink()
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 12 Jan 2018 04:26:49 +0000 (23:26 -0500)]
f2fs: switch to fscrypt ->symlink() helper functions
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:45:02 +0000 (10:45 -0800)]
fscrypt: new helper function - fscrypt_get_symlink()
Filesystems also have duplicate code to support ->get_link() on
encrypted symlinks. Factor it out into a new function
fscrypt_get_symlink(). It takes in the contents of the encrypted
symlink on-disk and provides the target (decrypted or encoded) that
should be returned from ->get_link().
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:45:01 +0000 (10:45 -0800)]
fscrypt: new helper functions for ->symlink()
Currently, filesystems supporting fscrypt need to implement some tricky
logic when creating encrypted symlinks, including handling a peculiar
on-disk format (struct fscrypt_symlink_data) and correctly calculating
the size of the encrypted symlink. Introduce helper functions to make
things a bit easier:
- fscrypt_prepare_symlink() computes and validates the size the symlink
target will require on-disk.
- fscrypt_encrypt_symlink() creates the encrypted target if needed.
The new helpers actually fix some subtle bugs. First, when checking
whether the symlink target was too long, filesystems didn't account for
the fact that the NUL padding is meant to be truncated if it would cause
the maximum length to be exceeded, as is done for filenames in
directories. Consequently users would receive ENAMETOOLONG when
creating symlinks close to what is supposed to be the maximum length.
For example, with EXT4 with a 4K block size, the maximum symlink target
length in an encrypted directory is supposed to be 4093 bytes (in
comparison to 4095 in an unencrypted directory), but in
FS_POLICY_FLAGS_PAD_32-mode only up to 4064 bytes were accepted.
Second, symlink targets of "." and ".." were not being encrypted, even
though they should be, as these names are special in *directory entries*
but not in symlink targets. Fortunately, we can fix this simply by
starting to encrypt them, as old kernels already accept them in
encrypted form.
Third, the output string length the filesystems were providing when
doing the actual encryption was incorrect, as it was forgotten to
exclude 'sizeof(struct fscrypt_symlink_data)'. Fortunately though, this
bug didn't make a difference.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:45:00 +0000 (10:45 -0800)]
fscrypt: trim down fscrypt.h includes
fscrypt.h included way too many other headers, given that it is included
by filesystems both with and without encryption support. Trim down the
includes list by moving the needed includes into more appropriate
places, and removing the unneeded ones.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:59 +0000 (10:44 -0800)]
fscrypt: move fscrypt_is_dot_dotdot() to fs/crypto/fname.c
Only fs/crypto/fname.c cares about treating the "." and ".." filenames
specially with regards to encryption, so move fscrypt_is_dot_dotdot()
from fscrypt.h to there.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:58 +0000 (10:44 -0800)]
fscrypt: move fscrypt_valid_enc_modes() to fscrypt_private.h
The encryption modes are validated by fs/crypto/, not by individual
filesystems. Therefore, move fscrypt_valid_enc_modes() from fscrypt.h
to fscrypt_private.h.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:57 +0000 (10:44 -0800)]
fscrypt: move fscrypt_operations declaration to fscrypt_supp.h
Filesystems now only define their fscrypt_operations when they are
compiled with encryption support, so move the fscrypt_operations
declaration from fscrypt.h to fscrypt_supp.h.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:56 +0000 (10:44 -0800)]
fscrypt: split fscrypt_dummy_context_enabled() into supp/notsupp versions
fscrypt_dummy_context_enabled() accesses ->s_cop, which now is only set
when the filesystem is built with encryption support. This didn't
actually matter because no filesystems called it. However, it will
start being used soon, so fix it by moving it from fscrypt.h to
fscrypt_supp.h and stubbing it out in fscrypt_notsupp.h.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:55 +0000 (10:44 -0800)]
fscrypt: move fscrypt_ctx declaration to fscrypt_supp.h
Filesystems only ever access 'struct fscrypt_ctx' through fscrypt
functions. But when a filesystem is built without encryption support,
these functions are all stubbed out, so the declaration of fscrypt_ctx
is unneeded. Therefore, move it from fscrypt.h to fscrypt_supp.h.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:54 +0000 (10:44 -0800)]
fscrypt: move fscrypt_info_cachep declaration to fscrypt_private.h
The fscrypt_info kmem_cache is internal to fscrypt; filesystems don't
need to access it. So move its declaration into fscrypt_private.h.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:53 +0000 (10:44 -0800)]
fscrypt: move fscrypt_control_page() to supp/notsupp headers
fscrypt_control_page() is already split into two versions depending on
whether the filesystem is being built with encryption support or not.
Move them into the appropriate headers.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Eric Biggers [Fri, 5 Jan 2018 18:44:52 +0000 (10:44 -0800)]
fscrypt: move fscrypt_has_encryption_key() to supp/notsupp headers
fscrypt_has_encryption_key() is already split into two versions
depending on whether the filesystem is being built with encryption
support or not. Move them into the appropriate headers.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Yunlong Song [Wed, 28 Feb 2018 12:31:52 +0000 (20:31 +0800)]
f2fs: don't put dentry page in pagecache into highmem
Previous dentry page uses highmem, which will cause panic in platforms
using highmem (such as arm), since the address space of dentry pages
from highmem directly goes into the decryption path via the function
fscrypt_fname_disk_to_usr. But sg_init_one assumes the address is not
from highmem, and then cause panic since it doesn't call kmap_high but
kunmap_high is triggered at the end. To fix this problem in a simple
way, this patch avoids to put dentry page in pagecache into highmem.
Signed-off-by: Yunlong Song <yunlong.song@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
[Jaegeuk Kim: fix coding style]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>