Expose the generic Curve25519 library via the crypto API KPP interface.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Wire up our newly added Blake2s implementation via the shash API.
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
If aead is built as a module along with cryptomgr, it creates a
dependency loop due to the dependency chain aead => crypto_null =>
cryptomgr => aead.
This is due to the presence of the AEAD geniv code. This code is
not really part of the AEAD API but simply support code for IV
generators such as seqiv. This patch moves the geniv code into
its own module thus breaking the dependency loop.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Now that the blkcipher algorithm type has been removed in favor of
skcipher, rename the crypto_blkcipher kernel module to crypto_skcipher,
and rename the config options accordingly:
CONFIG_CRYPTO_BLKCIPHER => CONFIG_CRYPTO_SKCIPHER
CONFIG_CRYPTO_BLKCIPHER2 => CONFIG_CRYPTO_SKCIPHER2
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Now that all "blkcipher" algorithms have been converted to "skcipher",
remove the blkcipher algorithm type.
The skcipher (symmetric key cipher) algorithm type was introduced a few
years ago to replace both blkcipher and ablkcipher (synchronous and
asynchronous block cipher). The advantages of skcipher include:
- A much less confusing name, since none of these algorithm types have
ever actually been for raw block ciphers, but rather for all
length-preserving encryption modes including block cipher modes of
operation, stream ciphers, and other length-preserving modes.
- It unified blkcipher and ablkcipher into a single algorithm type
which supports both synchronous and asynchronous implementations.
Note, blkcipher already operated only on scatterlists, so the fact
that skcipher does too isn't a regression in functionality.
- Better type safety by using struct skcipher_alg, struct
crypto_skcipher, etc. instead of crypto_alg, crypto_tfm, etc.
- It sometimes simplifies the implementations of algorithms.
Also, the blkcipher API was no longer being tested.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The patch brings support of several BLAKE2 variants (2b with various
digest lengths). The keyed digest is supported, using tfm->setkey call.
The in-tree user will be btrfs (for checksumming), we're going to use
the BLAKE2b-256 variant.
The code is reference implementation taken from the official sources and
modified in terms of kernel coding style (whitespace, comments, uintXX_t
-> uXX types, removed unused prototypes and #ifdefs, removed testing
code, changed secure_zero_memory -> memzero_explicit, used own helpers
for unaligned reads/writes and rotations).
Further changes removed sanity checks of key length or output size,
these values are verified in the crypto API callbacks or hardcoded in
shash_alg and not exposed to users.
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The next version of Clang will start policing compiler command line
options, and will reject combinations of -march and -mfpu that it
thinks are incompatible.
This results in errors like
clang-10: warning: ignoring extension 'crypto' because the 'armv7-a'
architecture does not support it [-Winvalid-command-line-argument]
/tmp/aegis128-neon-inner-5ee428.s: Assembler messages:
/tmp/aegis128-neon-inner-5ee428.s:73: Error: selected
processor does not support `aese.8 q2,q14' in ARM mode
when buiding the SIMD aegis128 code for 32-bit ARM, given that the
'armv7-a' -march argument is considered to be compatible with the
ARM crypto extensions. Instead, we should use armv8-a, which does
allow the crypto extensions to be enabled.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
implementation details that do not belong in DM crypt. The wrapper
template for ESSIV generation that was factored out will also be used
by fscrypt in the future.
- Add root hash pkcs#7 signature verification to the DM verity target.
- Add a new "clone" DM target that allows for efficient remote
replication of a device.
- Enhance DM bufio's cache to be tailored to each client based on use.
Clients that make heavy use of the cache get more of it, and those
that use less have reduced cache usage.
- Add a new DM_GET_TARGET_VERSION ioctl to allow userspace to query the
version number of a DM target (even if the associated module isn't yet
loaded).
- Fix invalid memory access in DM zoned target.
- Fix the max_discard_sectors limit advertised by the DM raid target; it
was mistakenly storing the limit in bytes rather than sectors.
- Small optimizations and cleanups in DM writecache target.
- Various fixes and cleanups in DM core, DM raid1 and space map portion
of DM persistent data library.
-----BEGIN PGP SIGNATURE-----
iQFHBAABCAAxFiEEJfWUX4UqZ4x1O2wixSPxCi2dA1oFAl2D7ycTHHNuaXR6ZXJA
cmVkaGF0LmNvbQAKCRDFI/EKLZ0DWp9QCACwTkVGzPGMCbAaCVlCACo8B5JyY4OO
FNxucqUlt1MHKuBbzJd4XwNGlLg68xjMUKVPYPlgina7TaDl+wvlTbHchaJS8nak
x1zyhDSywy0F9f6HHiXJi/vshmAfa0xnIM6fQXVPM346S6xf9u7hqOJQMCrdvY92
w4FhuW9nVt5xizo8iC/3LzoWbhrWncT7dyZUZtG3/tmglhkEK7QwctlgQxcD7tXg
H1lhntQzHzpxQAVBefWWdw7ubuDd6XCHuQMaxRhyR++c62P3eKDR8ck9hhd3hZKv
E481gtxcsjKuYLxwULjqFJZaNFitWFNMJ7gppQyKRqCzn2zlGAL6npl8
=m6zD
-----END PGP SIGNATURE-----
Merge tag 'for-5.4/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm
Pull device mapper updates from Mike Snitzer:
- crypto and DM crypt advances that allow the crypto API to reclaim
implementation details that do not belong in DM crypt. The wrapper
template for ESSIV generation that was factored out will also be used
by fscrypt in the future.
- Add root hash pkcs#7 signature verification to the DM verity target.
- Add a new "clone" DM target that allows for efficient remote
replication of a device.
- Enhance DM bufio's cache to be tailored to each client based on use.
Clients that make heavy use of the cache get more of it, and those
that use less have reduced cache usage.
- Add a new DM_GET_TARGET_VERSION ioctl to allow userspace to query the
version number of a DM target (even if the associated module isn't
yet loaded).
- Fix invalid memory access in DM zoned target.
- Fix the max_discard_sectors limit advertised by the DM raid target;
it was mistakenly storing the limit in bytes rather than sectors.
- Small optimizations and cleanups in DM writecache target.
- Various fixes and cleanups in DM core, DM raid1 and space map portion
of DM persistent data library.
* tag 'for-5.4/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm: (22 commits)
dm: introduce DM_GET_TARGET_VERSION
dm bufio: introduce a global cache replacement
dm bufio: remove old-style buffer cleanup
dm bufio: introduce a global queue
dm bufio: refactor adjust_total_allocated
dm bufio: call adjust_total_allocated from __link_buffer and __unlink_buffer
dm: add clone target
dm raid: fix updating of max_discard_sectors limit
dm writecache: skip writecache_wait for pmem mode
dm stats: use struct_size() helper
dm crypt: omit parsing of the encapsulated cipher
dm crypt: switch to ESSIV crypto API template
crypto: essiv - create wrapper template for ESSIV generation
dm space map common: remove check for impossible sm_find_free() return value
dm raid1: use struct_size() with kzalloc()
dm writecache: optimize performance by sorting the blocks for writeback_all
dm writecache: add unlikely for getting two block with same LBA
dm writecache: remove unused member pointer in writeback_struct
dm zoned: fix invalid memory access
dm verity: add root hash pkcs#7 signature verification
...
Implement a template that wraps a (skcipher,shash) or (aead,shash) tuple
so that we can consolidate the ESSIV handling in fscrypt and dm-crypt and
move it into the crypto API. This will result in better test coverage, and
will allow future changes to make the bare cipher interface internal to the
crypto subsystem, in order to increase robustness of the API against misuse.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Tested-by: Milan Broz <gmazyland@gmail.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Provide a version of the core AES transform to the aegis128 SIMD
code that does not rely on the special AES instructions, but uses
plain NEON instructions instead. This allows the SIMD version of
the aegis128 driver to be used on arm64 systems that do not
implement those instructions (which are not mandatory in the
architecture), such as the Raspberry Pi 3.
Since GCC makes a mess of this when using the tbl/tbx intrinsics
to perform the sbox substitution, preload the Sbox into v16..v31
in this case and use inline asm to emit the tbl/tbx instructions.
Clang does not support this approach, nor does it require it, since
it does a much better job at code generation, so there we use the
intrinsics as usual.
Cc: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Provide an accelerated implementation of aegis128 by wiring up the
SIMD hooks in the generic driver to an implementation based on NEON
intrinsics, which can be compiled to both ARM and arm64 code.
This results in a performance of 2.2 cycles per byte on Cortex-A53,
which is a performance increase of ~11x compared to the generic
code.
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add some plumbing to allow the AEGIS128 code to be built with SIMD
routines for acceleration.
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Recent clang-9 snapshots double the kernel stack usage when building
this file with -O0 -fsanitize=kernel-hwaddress, compared to clang-8
and older snapshots, this changed between commits svn364966 and
svn366056:
crypto/jitterentropy.c:516:5: error: stack frame size of 2640 bytes in function 'jent_entropy_init' [-Werror,-Wframe-larger-than=]
int jent_entropy_init(void)
^
crypto/jitterentropy.c:185:14: error: stack frame size of 2224 bytes in function 'jent_lfsr_time' [-Werror,-Wframe-larger-than=]
static __u64 jent_lfsr_time(struct rand_data *ec, __u64 time, __u64 loop_cnt)
^
I prepared a reduced test case in case any clang developers want to
take a closer look, but from looking at the earlier output it seems
that even with clang-8, something was very wrong here.
Turn off any KASAN and UBSAN sanitizing for this file, as that likely
clashes with -O0 anyway. Turning off just KASAN avoids the warning
already, but I suspect both of these have undesired side-effects
for jitterentropy.
Link: https://godbolt.org/z/fDcwZ5
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This reverts commit ecc8bc81f2
("crypto: aegis128 - provide a SIMD implementation based on NEON
intrinsics") and commit 7cdc0ddbf7
("crypto: aegis128 - add support for SIMD acceleration").
They cause compile errors on platforms other than ARM because
the mechanism to selectively compile the SIMD code is broken.
Repoted-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Provide an accelerated implementation of aegis128 by wiring up the
SIMD hooks in the generic driver to an implementation based on NEON
intrinsics, which can be compiled to both ARM and arm64 code.
This results in a performance of 2.2 cycles per byte on Cortex-A53,
which is a performance increase of ~11x compared to the generic
code.
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add some plumbing to allow the AEGIS128 code to be built with SIMD
routines for acceleration.
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Three variants of AEGIS were proposed for the CAESAR competition, and
only one was selected for the final portfolio: AEGIS128.
The other variants, AEGIS128L and AEGIS256, are not likely to ever turn
up in networking protocols or other places where interoperability
between Linux and other systems is a concern, nor are they likely to
be subjected to further cryptanalysis. However, uninformed users may
think that AEGIS128L (which is faster) is equally fit for use.
So let's remove them now, before anyone starts using them and we are
forced to support them forever.
Note that there are no known flaws in the algorithms or in any of these
implementations, but they have simply outlived their usefulness.
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
MORUS was not selected as a winner in the CAESAR competition, which
is not surprising since it is considered to be cryptographically
broken [0]. (Note that this is not an implementation defect, but a
flaw in the underlying algorithm). Since it is unlikely to be in use
currently, let's remove it before we're stuck with it.
[0] https://eprint.iacr.org/2019/172.pdf
Reviewed-by: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
xxhash is currently implemented as a self-contained module in /lib.
This patch enables that module to be used as part of the generic kernel
crypto framework. It adds a simple wrapper to the 64bit version.
I've also added test vectors (with help from Nick Terrell). The upstream
xxhash code is tested by running hashing operation on random 222 byte
data with seed values of 0 and a prime number. The upstream test
suite can be found at https://github.com/Cyan4973/xxHash/blob/cf46e0c/xxhsum.c#L664
Essentially hashing is run on data of length 0,1,14,222 with the
aforementioned seed values 0 and prime 2654435761. The particular random
222 byte string was provided to me by Nick Terrell by reading
/dev/random and the checksums were calculated by the upstream xxsum
utility with the following bash script:
dd if=/dev/random of=TEST_VECTOR bs=1 count=222
for a in 0 1; do
for l in 0 1 14 222; do
for s in 0 2654435761; do
echo algo $a length $l seed $s;
head -c $l TEST_VECTOR | ~/projects/kernel/xxHash/xxhsum -H$a -s$s
done
done
done
This produces output as follows:
algo 0 length 0 seed 0
02cc5d05 stdin
algo 0 length 0 seed 2654435761
02cc5d05 stdin
algo 0 length 1 seed 0
25201171 stdin
algo 0 length 1 seed 2654435761
25201171 stdin
algo 0 length 14 seed 0
c1d95975 stdin
algo 0 length 14 seed 2654435761
c1d95975 stdin
algo 0 length 222 seed 0
b38662a6 stdin
algo 0 length 222 seed 2654435761
b38662a6 stdin
algo 1 length 0 seed 0
ef46db3751d8e999 stdin
algo 1 length 0 seed 2654435761
ac75fda2929b17ef stdin
algo 1 length 1 seed 0
27c3f04c2881203a stdin
algo 1 length 1 seed 2654435761
4a15ed26415dfe4d stdin
algo 1 length 14 seed 0
3d33dc700231dfad stdin
algo 1 length 14 seed 2654435761
ea5f7ddef9a64f80 stdin
algo 1 length 222 seed 0
5f3d3c08ec2bef34 stdin
algo 1 length 222 seed 2654435761
6a9df59664c7ed62 stdin
algo 1 is xx64 variant, algo 0 is the 32 bit variant which is currently
not hooked up.
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
kcrypto_wq is only used by cryptd, so move it into cryptd.c and change
the workqueue name from "crypto" to "cryptd".
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add Elliptic Curve Russian Digital Signature Algorithm (GOST R
34.10-2012, RFC 7091, ISO/IEC 14888-3) is one of the Russian (and since
2018 the CIS countries) cryptographic standard algorithms (called GOST
algorithms). Only signature verification is supported, with intent to be
used in the IMA.
Summary of the changes:
* crypto/Kconfig:
- EC-RDSA is added into Public-key cryptography section.
* crypto/Makefile:
- ecrdsa objects are added.
* crypto/asymmetric_keys/x509_cert_parser.c:
- Recognize EC-RDSA and Streebog OIDs.
* include/linux/oid_registry.h:
- EC-RDSA OIDs are added to the enum. Also, a two currently not
implemented curve OIDs are added for possible extension later (to
not change numbering and grouping).
* crypto/ecc.c:
- Kenneth MacKay copyright date is updated to 2014, because
vli_mmod_slow, ecc_point_add, ecc_point_mult_shamir are based on his
code from micro-ecc.
- Functions needed for ecrdsa are EXPORT_SYMBOL'ed.
- New functions:
vli_is_negative - helper to determine sign of vli;
vli_from_be64 - unpack big-endian array into vli (used for
a signature);
vli_from_le64 - unpack little-endian array into vli (used for
a public key);
vli_uadd, vli_usub - add/sub u64 value to/from vli (used for
increment/decrement);
mul_64_64 - optimized to use __int128 where appropriate, this speeds
up point multiplication (and as a consequence signature
verification) by the factor of 1.5-2;
vli_umult - multiply vli by a small value (speeds up point
multiplication by another factor of 1.5-2, depending on vli sizes);
vli_mmod_special - module reduction for some form of Pseudo-Mersenne
primes (used for the curves A);
vli_mmod_special2 - module reduction for another form of
Pseudo-Mersenne primes (used for the curves B);
vli_mmod_barrett - module reduction using pre-computed value (used
for the curve C);
vli_mmod_slow - more general module reduction which is much slower
(used when the modulus is subgroup order);
vli_mod_mult_slow - modular multiplication;
ecc_point_add - add two points;
ecc_point_mult_shamir - add two points multiplied by scalars in one
combined multiplication (this gives speed up by another factor 2 in
compare to two separate multiplications).
ecc_is_pubkey_valid_partial - additional samity check is added.
- Updated vli_mmod_fast with non-strict heuristic to call optimal
module reduction function depending on the prime value;
- All computations for the previously defined (two NIST) curves should
not unaffected.
* crypto/ecc.h:
- Newly exported functions are documented.
* crypto/ecrdsa_defs.h
- Five curves are defined.
* crypto/ecrdsa.c:
- Signature verification is implemented.
* crypto/ecrdsa_params.asn1, crypto/ecrdsa_pub_key.asn1:
- Templates for BER decoder for EC-RDSA parameters and public key.
Cc: linux-integrity@vger.kernel.org
Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
ecc.c have algorithms that could be used togeter by ecdh and ecrdsa.
Make it separate module. Add CRYPTO_ECC into Kconfig. EXPORT_SYMBOL and
document to what seems appropriate. Move structs ecc_point and ecc_curve
from ecc_curve_defs.h into ecc.h.
No code changes.
Signed-off-by: Vitaly Chikunov <vt@altlinux.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
To prevent any issues with persistent data, separate lzo-rle from lzo so
that it is treated as a separate algorithm, and lzo is still available.
Link: http://lkml.kernel.org/r/20190205155944.16007-3-dave.rodgman@arm.com
Signed-off-by: Dave Rodgman <dave.rodgman@arm.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Markus F.X.J. Oberhumer <markus@oberhumer.com>
Cc: Matt Sealey <matt.sealey@arm.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <nitingupta910@gmail.com>
Cc: Richard Purdie <rpurdie@openedhand.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Sonny Rao <sonnyrao@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Even if CRYPTO_STATS is set to n, some part of CRYPTO_STATS are
compiled.
This patch made all part of crypto_user_stat uncompiled in that case.
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add support for the Adiantum encryption mode. Adiantum was designed by
Paul Crowley and is specified by our paper:
Adiantum: length-preserving encryption for entry-level processors
(https://eprint.iacr.org/2018/720.pdf)
See our paper for full details; this patch only provides an overview.
Adiantum is a tweakable, length-preserving encryption mode designed for
fast and secure disk encryption, especially on CPUs without dedicated
crypto instructions. Adiantum encrypts each sector using the XChaCha12
stream cipher, two passes of an ε-almost-∆-universal (εA∆U) hash
function, and an invocation of the AES-256 block cipher on a single
16-byte block. On CPUs without AES instructions, Adiantum is much
faster than AES-XTS; for example, on ARM Cortex-A7, on 4096-byte sectors
Adiantum encryption is about 4 times faster than AES-256-XTS encryption,
and decryption about 5 times faster.
Adiantum is a specialization of the more general HBSH construction. Our
earlier proposal, HPolyC, was also a HBSH specialization, but it used a
different εA∆U hash function, one based on Poly1305 only. Adiantum's
εA∆U hash function, which is based primarily on the "NH" hash function
like that used in UMAC (RFC4418), is about twice as fast as HPolyC's;
consequently, Adiantum is about 20% faster than HPolyC.
This speed comes with no loss of security: Adiantum is provably just as
secure as HPolyC, in fact slightly *more* secure. Like HPolyC,
Adiantum's security is reducible to that of XChaCha12 and AES-256,
subject to a security bound. XChaCha12 itself has a security reduction
to ChaCha12. Therefore, one need not "trust" Adiantum; one need only
trust ChaCha12 and AES-256. Note that the εA∆U hash function is only
used for its proven combinatorical properties so cannot be "broken".
Adiantum is also a true wide-block encryption mode, so flipping any
plaintext bit in the sector scrambles the entire ciphertext, and vice
versa. No other such mode is available in the kernel currently; doing
the same with XTS scrambles only 16 bytes. Adiantum also supports
arbitrary-length tweaks and naturally supports any length input >= 16
bytes without needing "ciphertext stealing".
For the stream cipher, Adiantum uses XChaCha12 rather than XChaCha20 in
order to make encryption feasible on the widest range of devices.
Although the 20-round variant is quite popular, the best known attacks
on ChaCha are on only 7 rounds, so ChaCha12 still has a substantial
security margin; in fact, larger than AES-256's. 12-round Salsa20 is
also the eSTREAM recommendation. For the block cipher, Adiantum uses
AES-256, despite it having a lower security margin than XChaCha12 and
needing table lookups, due to AES's extensive adoption and analysis
making it the obvious first choice. Nevertheless, for flexibility this
patch also permits the "adiantum" template to be instantiated with
XChaCha20 and/or with an alternate block cipher.
We need Adiantum support in the kernel for use in dm-crypt and fscrypt,
where currently the only other suitable options are block cipher modes
such as AES-XTS. A big problem with this is that many low-end mobile
devices (e.g. Android Go phones sold primarily in developing countries,
as well as some smartwatches) still have CPUs that lack AES
instructions, e.g. ARM Cortex-A7. Sadly, AES-XTS encryption is much too
slow to be viable on these devices. We did find that some "lightweight"
block ciphers are fast enough, but these suffer from problems such as
not having much cryptanalysis or being too controversial.
The ChaCha stream cipher has excellent performance but is insecure to
use directly for disk encryption, since each sector's IV is reused each
time it is overwritten. Even restricting the threat model to offline
attacks only isn't enough, since modern flash storage devices don't
guarantee that "overwrites" are really overwrites, due to wear-leveling.
Adiantum avoids this problem by constructing a
"tweakable super-pseudorandom permutation"; this is the strongest
possible security model for length-preserving encryption.
Of course, storing random nonces along with the ciphertext would be the
ideal solution. But doing that with existing hardware and filesystems
runs into major practical problems; in most cases it would require data
journaling (like dm-integrity) which severely degrades performance.
Thus, for now length-preserving encryption is still needed.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add a generic implementation of NHPoly1305, an ε-almost-∆-universal hash
function used in the Adiantum encryption mode.
CONFIG_NHPOLY1305 is not selectable by itself since there won't be any
real reason to enable it without also enabling Adiantum support.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In preparation for adding XChaCha12 support, rename/refactor
chacha20-generic to support different numbers of rounds. The
justification for needing XChaCha12 support is explained in more detail
in the patch "crypto: chacha - add XChaCha12 support".
The only difference between ChaCha{8,12,20} are the number of rounds
itself; all other parts of the algorithm are the same. Therefore,
remove the "20" from all definitions, structures, functions, files, etc.
that will be shared by all ChaCha versions.
Also make ->setkey() store the round count in the chacha_ctx (previously
chacha20_ctx). The generic code then passes the round count through to
chacha_block(). There will be a ->setkey() function for each explicitly
allowed round count; the encrypt/decrypt functions will be the same. I
decided not to do it the opposite way (same ->setkey() function for all
round counts, with different encrypt/decrypt functions) because that
would have required more boilerplate code in architecture-specific
implementations of ChaCha and XChaCha.
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Acked-by: Martin Willi <martin@strongswan.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add a generic version of output feedback mode. We already have support of
several hardware based transformations of this mode and the needed test
vectors but we somehow missed adding a generic software one. Fix this now.
Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch implement a generic way to get statistics about all crypto
usages.
Signed-off-by: Corentin Labbe <clabbe@baylibre.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As it turns out, the AVX2 multibuffer SHA routines are currently
broken [0], in a way that would have likely been noticed if this
code were in wide use. Since the code is too complicated to be
maintained by anyone except the original authors, and since the
performance benefits for real-world use cases are debatable to
begin with, it is better to drop it entirely for the moment.
[0] https://marc.info/?l=linux-crypto-vger&m=153476243825350&w=2
Suggested-by: Eric Biggers <ebiggers@google.com>
Cc: Megha Dey <megha.dey@linux.intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
These are unused, undesired, and have never actually been used by
anybody. The original authors of this code have changed their mind about
its inclusion. While originally proposed for disk encryption on low-end
devices, the idea was discarded [1] in favor of something else before
that could really get going. Therefore, this patch removes Speck.
[1] https://marc.info/?l=linux-crypto-vger&m=153359499015659
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Acked-by: Eric Biggers <ebiggers@google.com>
Cc: stable@vger.kernel.org
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Commit 56e8e57fc3 ("crypto: morus - Add common SIMD glue code for
MORUS") accidetally consiedered the glue code to be usable by different
architectures, but it seems to be only usable on x86.
This patch moves it under arch/x86/crypto and adds 'depends on X86' to
the Kconfig options and also removes the prompt to hide these internal
options from the user.
Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Ondrej Mosnacek <omosnacek@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds a common glue code for optimized implementations of
MORUS AEAD algorithms.
Signed-off-by: Ondrej Mosnacek <omosnacek@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the generic implementation of the MORUS family of AEAD
algorithms (MORUS-640 and MORUS-1280). The original authors of MORUS
are Hongjun Wu and Tao Huang.
At the time of writing, MORUS is one of the finalists in CAESAR, an
open competition intended to select a portfolio of alternatives to
the problematic AES-GCM:
https://competitions.cr.yp.to/caesar-submissions.htmlhttps://competitions.cr.yp.to/round3/morusv2.pdf
Signed-off-by: Ondrej Mosnacek <omosnacek@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the generic implementation of the AEGIS family of AEAD
algorithms (AEGIS-128, AEGIS-128L, and AEGIS-256). The original
authors of AEGIS are Hongjun Wu and Bart Preneel.
At the time of writing, AEGIS is one of the finalists in CAESAR, an
open competition intended to select a portfolio of alternatives to
the problematic AES-GCM:
https://competitions.cr.yp.to/caesar-submissions.htmlhttps://competitions.cr.yp.to/round3/aegisv11.pdf
Signed-off-by: Ondrej Mosnacek <omosnacek@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Adds zstd support to crypto and scompress. Only supports the default
level.
Previously we held off on this patch, since there weren't any users.
Now zram is ready for zstd support, but depends on CONFIG_CRYPTO_ZSTD,
which isn't defined until this patch is in. I also see a patch adding
zstd to pstore [0], which depends on crypto zstd.
[0] lkml.kernel.org/r/9c9416b2dff19f05fb4c35879aaa83d11ff72c92.1521626182.git.geliangtang@gmail.com
Signed-off-by: Nick Terrell <terrelln@fb.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Our convention is to distinguish file types by suffixes with a period
as a separator.
*-asn1.[ch] is a different pattern from other generated sources such
as *.lex.c, *.tab.[ch], *.dtb.S, etc. More confusing, files with
'-asn1.[ch]' are generated files, but '_asn1.[ch]' are checked-in
files:
net/netfilter/nf_conntrack_h323_asn1.c
include/linux/netfilter/nf_conntrack_h323_asn1.h
include/linux/sunrpc/gss_asn1.h
Rename generated files to *.asn1.[ch] for consistency.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Introduce the SM4 cipher algorithms (OSCCA GB/T 32907-2016).
SM4 (GBT.32907-2016) is a cryptographic standard issued by the
Organization of State Commercial Administration of China (OSCCA)
as an authorized cryptographic algorithms for the use within China.
SMS4 was originally created for use in protecting wireless
networks, and is mandated in the Chinese National Standard for
Wireless LAN WAPI (Wired Authentication and Privacy Infrastructure)
(GB.15629.11-2003).
Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
TPM security routines require encryption and decryption with AES in
CFB mode, so add it to the Linux Crypto schemes. CFB is basically a
one time pad where the pad is generated initially from the encrypted
IV and then subsequently from the encrypted previous block of
ciphertext. The pad is XOR'd into the plain text to get the final
ciphertext.
https://en.wikipedia.org/wiki/Block_cipher_mode_of_operation#CFB
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
All users of ablk_helper have been converted over to crypto_simd, so
remove ablk_helper.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add a generic implementation of Speck, including the Speck128 and
Speck64 variants. Speck is a lightweight block cipher that can be much
faster than AES on processors that don't have AES instructions.
We are planning to offer Speck-XTS (probably Speck128/256-XTS) as an
option for dm-crypt and fscrypt on Android, for low-end mobile devices
with older CPUs such as ARMv7 which don't have the Cryptography
Extensions. Currently, such devices are unencrypted because AES is not
fast enough, even when the NEON bit-sliced implementation of AES is
used. Other AES alternatives such as Twofish, Threefish, Camellia,
CAST6, and Serpent aren't fast enough either; it seems that only a
modern ARX cipher can provide sufficient performance on these devices.
This is a replacement for our original proposal
(https://patchwork.kernel.org/patch/10101451/) which was to offer
ChaCha20 for these devices. However, the use of a stream cipher for
disk/file encryption with no space to store nonces would have been much
more insecure than we thought initially, given that it would be used on
top of flash storage as well as potentially on top of F2FS, neither of
which is guaranteed to overwrite data in-place.
Speck has been somewhat controversial due to its origin. Nevertheless,
it has a straightforward design (it's an ARX cipher), and it appears to
be the leading software-optimized lightweight block cipher currently,
with the most cryptanalysis. It's also easy to implement without side
channels, unlike AES. Moreover, we only intend Speck to be used when
the status quo is no encryption, due to AES not being fast enough.
We've also considered a novel length-preserving encryption mode based on
ChaCha20 and Poly1305. While theoretically attractive, such a mode
would be a brand new crypto construction and would be more complicated
and difficult to implement efficiently in comparison to Speck-XTS.
There is confusion about the byte and word orders of Speck, since the
original paper doesn't specify them. But we have implemented it using
the orders the authors recommended in a correspondence with them. The
test vectors are taken from the original paper but were mapped to byte
arrays using the recommended byte and word orders.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
My last bugfix added -Os on the command line, which unfortunately caused
a build regression on powerpc in some configurations.
I've done some more analysis of the original problem and found slightly
different workaround that avoids this regression and also results in
better performance on gcc-7.0: -fcode-hoisting is an optimization step
that got added in gcc-7 and that for all gcc-7 versions causes worse
performance.
This disables -fcode-hoisting on all compilers that understand the option.
For gcc-7.1 and 7.2 I found the same performance as my previous patch
(using -Os), in gcc-7.0 it was even better. On gcc-8 I could see no
change in performance from this patch. In theory, code hoisting should
not be able make things better for the AES cipher, so leaving it
disabled for gcc-8 only serves to simplify the Makefile change.
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Link: https://www.mail-archive.com/linux-crypto@vger.kernel.org/msg30418.html
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83356
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83651
Fixes: 148b974dee ("crypto: aes-generic - build with -Os on gcc-7+")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
While testing other changes, I discovered that gcc-7.2.1 produces badly
optimized code for aes_encrypt/aes_decrypt. This is especially true when
CONFIG_UBSAN_SANITIZE_ALL is enabled, where it leads to extremely
large stack usage that in turn might cause kernel stack overflows:
crypto/aes_generic.c: In function 'aes_encrypt':
crypto/aes_generic.c:1371:1: warning: the frame size of 4880 bytes is larger than 2048 bytes [-Wframe-larger-than=]
crypto/aes_generic.c: In function 'aes_decrypt':
crypto/aes_generic.c:1441:1: warning: the frame size of 4864 bytes is larger than 2048 bytes [-Wframe-larger-than=]
I verified that this problem exists on all architectures that are
supported by gcc-7.2, though arm64 in particular is less affected than
the others. I also found that gcc-7.1 and gcc-8 do not show the extreme
stack usage but still produce worse code than earlier versions for this
file, apparently because of optimization passes that generally provide
a substantial improvement in object code quality but understandably fail
to find any shortcuts in the AES algorithm.
Possible workarounds include
a) disabling -ftree-pre and -ftree-sra optimizations, this was an earlier
patch I tried, which reliably fixed the stack usage, but caused a
serious performance regression in some versions, as later testing
found.
b) disabling UBSAN on this file or all ciphers, as suggested by Ard
Biesheuvel. This would lead to massively better crypto performance in
UBSAN-enabled kernels and avoid the stack usage, but there is a concern
over whether we should exclude arbitrary files from UBSAN at all.
c) Forcing the optimization level in a different way. Similar to a),
but rather than deselecting specific optimization stages,
this now uses "gcc -Os" for this file, regardless of the
CONFIG_CC_OPTIMIZE_FOR_PERFORMANCE/SIZE option. This is a reliable
workaround for the stack consumption on all architecture, and I've
retested the performance results now on x86, cycles/byte (lower is
better) for cbc(aes-generic) with 256 bit keys:
-O2 -Os
gcc-6.3.1 14.9 15.1
gcc-7.0.1 14.7 15.3
gcc-7.1.1 15.3 14.7
gcc-7.2.1 16.8 15.9
gcc-8.0.0 15.5 15.6
This implements the option c) by enabling forcing -Os on all compiler
versions starting with gcc-7.1. As a workaround for PR83356, it would
only be needed for gcc-7.2+ with UBSAN enabled, but since it also shows
better performance on gcc-7.1 without UBSAN, it seems appropriate to
use the faster version here as well.
Side note: during testing, I also played with the AES code in libressl,
which had a similar performance regression from gcc-6 to gcc-7.2,
but was three times slower overall. It might be interesting to
investigate that further and possibly port the Linux implementation
into that.
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83356
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83651
Cc: Richard Biener <rguenther@suse.de>
Cc: Jakub Jelinek <jakub@gcc.gnu.org>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Pull crypto updates from Herbert Xu:
"Here is the crypto update for 4.15:
API:
- Disambiguate EBUSY when queueing crypto request by adding ENOSPC.
This change touches code outside the crypto API.
- Reset settings when empty string is written to rng_current.
Algorithms:
- Add OSCCA SM3 secure hash.
Drivers:
- Remove old mv_cesa driver (replaced by marvell/cesa).
- Enable rfc3686/ecb/cfb/ofb AES in crypto4xx.
- Add ccm/gcm AES in crypto4xx.
- Add support for BCM7278 in iproc-rng200.
- Add hash support on Exynos in s5p-sss.
- Fix fallback-induced error in vmx.
- Fix output IV in atmel-aes.
- Fix empty GCM hash in mediatek.
Others:
- Fix DoS potential in lib/mpi.
- Fix potential out-of-order issues with padata"
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (162 commits)
lib/mpi: call cond_resched() from mpi_powm() loop
crypto: stm32/hash - Fix return issue on update
crypto: dh - Remove pointless checks for NULL 'p' and 'g'
crypto: qat - Clean up error handling in qat_dh_set_secret()
crypto: dh - Don't permit 'key' or 'g' size longer than 'p'
crypto: dh - Don't permit 'p' to be 0
crypto: dh - Fix double free of ctx->p
hwrng: iproc-rng200 - Add support for BCM7278
dt-bindings: rng: Document BCM7278 RNG200 compatible
crypto: chcr - Replace _manual_ swap with swap macro
crypto: marvell - Add a NULL entry at the end of mv_cesa_plat_id_table[]
hwrng: virtio - Virtio RNG devices need to be re-registered after suspend/resume
crypto: atmel - remove empty functions
crypto: ecdh - remove empty exit()
MAINTAINERS: update maintainer for qat
crypto: caam - remove unused param of ctx_map_to_sec4_sg()
crypto: caam - remove unneeded edesc zeroization
crypto: atmel-aes - Reset the controller before each use
crypto: atmel-aes - properly set IV after {en,de}crypt
hwrng: core - Reset user selected rng by writing "" to rng_current
...
Many source files in the tree are missing licensing information, which
makes it harder for compliance tools to determine the correct license.
By default all files without license information are under the default
license of the kernel, which is GPL version 2.
Update the files which contain no license information with the 'GPL-2.0'
SPDX license identifier. The SPDX identifier is a legally binding
shorthand, which can be used instead of the full boiler plate text.
This patch is based on work done by Thomas Gleixner and Kate Stewart and
Philippe Ombredanne.
How this work was done:
Patches were generated and checked against linux-4.14-rc6 for a subset of
the use cases:
- file had no licensing information it it.
- file was a */uapi/* one with no licensing information in it,
- file was a */uapi/* one with existing licensing information,
Further patches will be generated in subsequent months to fix up cases
where non-standard license headers were used, and references to license
had to be inferred by heuristics based on keywords.
The analysis to determine which SPDX License Identifier to be applied to
a file was done in a spreadsheet of side by side results from of the
output of two independent scanners (ScanCode & Windriver) producing SPDX
tag:value files created by Philippe Ombredanne. Philippe prepared the
base worksheet, and did an initial spot review of a few 1000 files.
The 4.13 kernel was the starting point of the analysis with 60,537 files
assessed. Kate Stewart did a file by file comparison of the scanner
results in the spreadsheet to determine which SPDX license identifier(s)
to be applied to the file. She confirmed any determination that was not
immediately clear with lawyers working with the Linux Foundation.
Criteria used to select files for SPDX license identifier tagging was:
- Files considered eligible had to be source code files.
- Make and config files were included as candidates if they contained >5
lines of source
- File already had some variant of a license header in it (even if <5
lines).
All documentation files were explicitly excluded.
The following heuristics were used to determine which SPDX license
identifiers to apply.
- when both scanners couldn't find any license traces, file was
considered to have no license information in it, and the top level
COPYING file license applied.
For non */uapi/* files that summary was:
SPDX license identifier # files
---------------------------------------------------|-------
GPL-2.0 11139
and resulted in the first patch in this series.
If that file was a */uapi/* path one, it was "GPL-2.0 WITH
Linux-syscall-note" otherwise it was "GPL-2.0". Results of that was:
SPDX license identifier # files
---------------------------------------------------|-------
GPL-2.0 WITH Linux-syscall-note 930
and resulted in the second patch in this series.
- if a file had some form of licensing information in it, and was one
of the */uapi/* ones, it was denoted with the Linux-syscall-note if
any GPL family license was found in the file or had no licensing in
it (per prior point). Results summary:
SPDX license identifier # files
---------------------------------------------------|------
GPL-2.0 WITH Linux-syscall-note 270
GPL-2.0+ WITH Linux-syscall-note 169
((GPL-2.0 WITH Linux-syscall-note) OR BSD-2-Clause) 21
((GPL-2.0 WITH Linux-syscall-note) OR BSD-3-Clause) 17
LGPL-2.1+ WITH Linux-syscall-note 15
GPL-1.0+ WITH Linux-syscall-note 14
((GPL-2.0+ WITH Linux-syscall-note) OR BSD-3-Clause) 5
LGPL-2.0+ WITH Linux-syscall-note 4
LGPL-2.1 WITH Linux-syscall-note 3
((GPL-2.0 WITH Linux-syscall-note) OR MIT) 3
((GPL-2.0 WITH Linux-syscall-note) AND MIT) 1
and that resulted in the third patch in this series.
- when the two scanners agreed on the detected license(s), that became
the concluded license(s).
- when there was disagreement between the two scanners (one detected a
license but the other didn't, or they both detected different
licenses) a manual inspection of the file occurred.
- In most cases a manual inspection of the information in the file
resulted in a clear resolution of the license that should apply (and
which scanner probably needed to revisit its heuristics).
- When it was not immediately clear, the license identifier was
confirmed with lawyers working with the Linux Foundation.
- If there was any question as to the appropriate license identifier,
the file was flagged for further research and to be revisited later
in time.
In total, over 70 hours of logged manual review was done on the
spreadsheet to determine the SPDX license identifiers to apply to the
source files by Kate, Philippe, Thomas and, in some cases, confirmation
by lawyers working with the Linux Foundation.
Kate also obtained a third independent scan of the 4.13 code base from
FOSSology, and compared selected files where the other two scanners
disagreed against that SPDX file, to see if there was new insights. The
Windriver scanner is based on an older version of FOSSology in part, so
they are related.
Thomas did random spot checks in about 500 files from the spreadsheets
for the uapi headers and agreed with SPDX license identifier in the
files he inspected. For the non-uapi files Thomas did random spot checks
in about 15000 files.
In initial set of patches against 4.14-rc6, 3 files were found to have
copy/paste license identifier errors, and have been fixed to reflect the
correct identifier.
Additionally Philippe spent 10 hours this week doing a detailed manual
inspection and review of the 12,461 patched files from the initial patch
version early this week with:
- a full scancode scan run, collecting the matched texts, detected
license ids and scores
- reviewing anything where there was a license detected (about 500+
files) to ensure that the applied SPDX license was correct
- reviewing anything where there was no detection but the patch license
was not GPL-2.0 WITH Linux-syscall-note to ensure that the applied
SPDX license was correct
This produced a worksheet with 20 files needing minor correction. This
worksheet was then exported into 3 different .csv files for the
different types of files to be modified.
These .csv files were then reviewed by Greg. Thomas wrote a script to
parse the csv files and add the proper SPDX tag to the file, in the
format that the file expected. This script was further refined by Greg
based on the output to detect more types of files automatically and to
distinguish between header and source .c files (which need different
comment types.) Finally Greg ran the script using the .csv files to
generate the patches.
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Philippe Ombredanne <pombredanne@nexb.com>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add support for generating ecc private keys.
Generation of ecc private keys is helpful in a user-space to kernel
ecdh offload because the keys are not revealed to user-space. Private
key generation is also helpful to implement forward secrecy.
If the user provides a NULL ecc private key, the kernel will generate it
and further use it for ecdh.
Move ecdh's object files below drbg's. drbg must be present in the kernel
at the time of calling.
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Reviewed-by: Stephan Müller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
An ancient gcc bug (first reported in 2003) has apparently resurfaced
on MIPS, where kernelci.org reports an overly large stack frame in the
whirlpool hash algorithm:
crypto/wp512.c:987:1: warning: the frame size of 1112 bytes is larger than 1024 bytes [-Wframe-larger-than=]
With some testing in different configurations, I'm seeing large
variations in stack frames size up to 1500 bytes for what should have
around 300 bytes at most. I also checked the reference implementation,
which is essentially the same code but also comes with some test and
benchmarking infrastructure.
It seems that recent compiler versions on at least arm, arm64 and powerpc
have a partial fix for this problem, but enabling "-fsched-pressure", but
even with that fix they suffer from the issue to a certain degree. Some
testing on arm64 shows that the time needed to hash a given amount of
data is roughly proportional to the stack frame size here, which makes
sense given that the wp512 implementation is doing lots of loads for
table lookups, and the problem with the overly large stack is a result
of doing a lot more loads and stores for spilled registers (as seen from
inspecting the object code).
Disabling -fschedule-insns consistently fixes the problem for wp512,
in my collection of cross-compilers, the results are consistently better
or identical when comparing the stack sizes in this function, though
some architectures (notable x86) have schedule-insns disabled by
default.
The four columns are:
default: -O2
press: -O2 -fsched-pressure
nopress: -O2 -fschedule-insns -fno-sched-pressure
nosched: -O2 -no-schedule-insns (disables sched-pressure)
default press nopress nosched
alpha-linux-gcc-4.9.3 1136 848 1136 176
am33_2.0-linux-gcc-4.9.3 2100 2076 2100 2104
arm-linux-gnueabi-gcc-4.9.3 848 848 1048 352
cris-linux-gcc-4.9.3 272 272 272 272
frv-linux-gcc-4.9.3 1128 1000 1128 280
hppa64-linux-gcc-4.9.3 1128 336 1128 184
hppa-linux-gcc-4.9.3 644 308 644 276
i386-linux-gcc-4.9.3 352 352 352 352
m32r-linux-gcc-4.9.3 720 656 720 268
microblaze-linux-gcc-4.9.3 1108 604 1108 256
mips64-linux-gcc-4.9.3 1328 592 1328 208
mips-linux-gcc-4.9.3 1096 624 1096 240
powerpc64-linux-gcc-4.9.3 1088 432 1088 160
powerpc-linux-gcc-4.9.3 1080 584 1080 224
s390-linux-gcc-4.9.3 456 456 624 360
sh3-linux-gcc-4.9.3 292 292 292 292
sparc64-linux-gcc-4.9.3 992 240 992 208
sparc-linux-gcc-4.9.3 680 592 680 312
x86_64-linux-gcc-4.9.3 224 240 272 224
xtensa-linux-gcc-4.9.3 1152 704 1152 304
aarch64-linux-gcc-7.0.0 224 224 1104 208
arm-linux-gnueabi-gcc-7.0.1 824 824 1048 352
mips-linux-gcc-7.0.0 1120 648 1120 272
x86_64-linux-gcc-7.0.1 240 240 304 240
arm-linux-gnueabi-gcc-4.4.7 840 392
arm-linux-gnueabi-gcc-4.5.4 784 728 784 320
arm-linux-gnueabi-gcc-4.6.4 736 728 736 304
arm-linux-gnueabi-gcc-4.7.4 944 784 944 352
arm-linux-gnueabi-gcc-4.8.5 464 464 760 352
arm-linux-gnueabi-gcc-4.9.3 848 848 1048 352
arm-linux-gnueabi-gcc-5.3.1 824 824 1064 336
arm-linux-gnueabi-gcc-6.1.1 808 808 1056 344
arm-linux-gnueabi-gcc-7.0.1 824 824 1048 352
Trying the same test for serpent-generic, the picture is a bit different,
and while -fno-schedule-insns is generally better here than the default,
-fsched-pressure wins overall, so I picked that instead.
default press nopress nosched
alpha-linux-gcc-4.9.3 1392 864 1392 960
am33_2.0-linux-gcc-4.9.3 536 524 536 528
arm-linux-gnueabi-gcc-4.9.3 552 552 776 536
cris-linux-gcc-4.9.3 528 528 528 528
frv-linux-gcc-4.9.3 536 400 536 504
hppa64-linux-gcc-4.9.3 524 208 524 480
hppa-linux-gcc-4.9.3 768 472 768 508
i386-linux-gcc-4.9.3 564 564 564 564
m32r-linux-gcc-4.9.3 712 576 712 532
microblaze-linux-gcc-4.9.3 724 392 724 512
mips64-linux-gcc-4.9.3 720 384 720 496
mips-linux-gcc-4.9.3 728 384 728 496
powerpc64-linux-gcc-4.9.3 704 304 704 480
powerpc-linux-gcc-4.9.3 704 296 704 480
s390-linux-gcc-4.9.3 560 560 592 536
sh3-linux-gcc-4.9.3 540 540 540 540
sparc64-linux-gcc-4.9.3 544 352 544 496
sparc-linux-gcc-4.9.3 544 344 544 496
x86_64-linux-gcc-4.9.3 528 536 576 528
xtensa-linux-gcc-4.9.3 752 544 752 544
aarch64-linux-gcc-7.0.0 432 432 656 480
arm-linux-gnueabi-gcc-7.0.1 616 616 808 536
mips-linux-gcc-7.0.0 720 464 720 488
x86_64-linux-gcc-7.0.1 536 528 600 536
arm-linux-gnueabi-gcc-4.4.7 592 440
arm-linux-gnueabi-gcc-4.5.4 776 448 776 544
arm-linux-gnueabi-gcc-4.6.4 776 448 776 544
arm-linux-gnueabi-gcc-4.7.4 768 448 768 544
arm-linux-gnueabi-gcc-4.8.5 488 488 776 544
arm-linux-gnueabi-gcc-4.9.3 552 552 776 536
arm-linux-gnueabi-gcc-5.3.1 552 552 776 536
arm-linux-gnueabi-gcc-6.1.1 560 560 776 536
arm-linux-gnueabi-gcc-7.0.1 616 616 808 536
I did not do any runtime tests with serpent, so it is possible that stack
frame size does not directly correlate with runtime performance here and
it actually makes things worse, but it's more likely to help here, and
the reduced stack frame size is probably enough reason to apply the patch,
especially given that the crypto code is often used in deep call chains.
Link: https://kernelci.org/build/id/58797d7559b5149efdf6c3a9/logs/
Link: http://www.larc.usp.br/~pbarreto/WhirlpoolPage.html
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=11488
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79149
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
