Pull f2fs updates from Jaegeuk Kim:
 "In this round, we've mainly focused on discard, aka unmap, control
  along with fstrim for Android-specific usage model. In addition, we've
  fixed writepage flow which returned EAGAIN previously resulting in EIO
  of fsync(2) due to mapping's error state. In order to avoid old MM bug
  [1], we decided not to use __GFP_ZERO for the mapping for node and
  meta page caches. As always, we've cleaned up many places for future
  fsverity and symbol conflicts.

  Enhancements:
   - do discard/fstrim in lower priority considering fs utilization
   - split large discard commands into smaller ones for better responsiveness
   - add more sanity checks to address syzbot reports
   - add a mount option, fsync_mode=nobarrier, which can reduce # of cache flushes
   - clean up symbol namespace with modified function names
   - be strict on block allocation and IO control in corner cases

  Bug fixes:
   - don't use __GFP_ZERO for mappings
   - fix error reports in writepage to avoid fsync() failure
   - avoid selinux denial on CAP_RESOURCE on resgid/resuid
   - fix some subtle race conditions in GC/atomic writes/shutdown
   - fix overflow bugs in sanity_check_raw_super
   - fix missing bits on get_flags

  Clean-ups:
   - prepare the generic flow for future fsverity integration
   - fix some broken coding standard"

[1] https://lkml.org/lkml/2018/4/8/661

* tag 'f2fs-for-4.18' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (79 commits)
  f2fs: fix to clear FI_VOLATILE_FILE correctly
  f2fs: let sync node IO interrupt async one
  f2fs: don't change wbc->sync_mode
  f2fs: fix to update mtime correctly
  fs: f2fs: insert space around that ':' and ', '
  fs: f2fs: add missing blank lines after declarations
  fs: f2fs: changed variable type of offset "unsigned" to "loff_t"
  f2fs: clean up symbol namespace
  f2fs: make set_de_type() static
  f2fs: make __f2fs_write_data_pages() static
  f2fs: fix to avoid accessing cross the boundary
  f2fs: fix to let caller retry allocating block address
  disable loading f2fs module on PAGE_SIZE > 4KB
  f2fs: fix error path of move_data_page
  f2fs: don't drop dentry pages after fs shutdown
  f2fs: fix to avoid race during access gc_thread pointer
  f2fs: clean up with clear_radix_tree_dirty_tag
  f2fs: fix to don't trigger writeback during recovery
  f2fs: clear discard_wake earlier
  f2fs: let discard thread wait a little longer if dev is busy
  ...

Log the crypto algorithm driver name for each fscrypt encryption mode on
its first use, also showing a friendly name for the mode.

This will help people determine whether the expected implementations are
being used.  In some cases we've seen people do benchmarks and reject
using encryption for performance reasons, when in fact they used a much
slower implementation of AES-XTS than was possible on the hardware.  It
can make an enormous difference; e.g., AES-XTS on ARM is about 10x
faster with the crypto extensions (AES instructions) than without.

This also makes it more obvious which modes are being used, now that
fscrypt supports multiple combinations of modes.

Example messages (with default modes, on x86_64):

[   35.492057] fscrypt: AES-256-CTS-CBC using implementation "cts(cbc-aes-aesni)"
[   35.492171] fscrypt: AES-256-XTS using implementation "xts-aes-aesni"

Note: algorithms can be dynamically added to the crypto API, which can
result in different implementations being used at different times.  But
this is rare; for most users, showing the first will be good enough.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>

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>

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>

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>

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>

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>

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>

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>

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>