Update the AES-XTS implementation based on NEON instructions so that it
can deal with inputs whose size is not a multiple of the cipher block
size. This is part of the original XTS specification, but was never
implemented before in the Linux kernel.
Since the bit slicing driver is only faster if it can operate on at
least 7 blocks of input at the same time, let's reuse the alternate
path we are adding for CTS to process any data tail whose size is
not a multiple of 128 bytes.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add the missing support for ciphertext stealing in the implementation
of AES-XTS, which is part of the XTS specification but was omitted up
until now due to lack of a need for it.
The asm helpers are updated so they can deal with any input size, as
long as the last full block and the final partial block are presented
at the same time. The glue code is updated so that the common case of
operating on a sector or page is mostly as before. When CTS is needed,
the walk is split up into two pieces, unless the entire input is covered
by a single step.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Since the CTS-CBC code completes synchronously, there is no point in
keeping part of the scratch data it uses in the request context, so
move it to the stack instead.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Optimize away one of the tbl instructions in the decryption path,
which turns out to be unnecessary.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The pure NEON AES implementation predates the bit-slicing one, and is
generally slower, unless the algorithm in question can only execute
sequentially.
So advertising the skciphers that the bit-slicing driver implements as
well serves no real purpose, and we can just disable them. Note that the
bit-slicing driver also has a link time dependency on the pure NEON
driver, for CBC encryption and for XTS tweak calculation, so we still
need both drivers on systems that do not implement the Crypto Extensions.
At the same time, expose those modaliases for the AES instruction based
driver. This is necessary since otherwise, we may end up loading the
wrong driver when any of the skciphers are instantiated before the CPU
capability based module loading has completed.
Finally, add the missing modalias for cts(cbc(aes)) so requests for
this algorithm will autoload the correct module.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace the vector load from memory sequence with a simple instruction
sequence to compose the tweak vector directly.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
PTR_ERR_OR_ZERO contains if(IS_ERR(...)) + PTR_ERR. It is better to
use it directly. hence just replace it.
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Rename static / file-local functions so that they do not conflict with
the functions declared in crypto/sha256.h.
This is a preparation patch for folding crypto/sha256.h into crypto/sha.h.
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Add an accelerated version of the 'essiv(cbc(aes),sha256)' skcipher,
which is used by fscrypt or dm-crypt on systems where CBC mode is
signficantly more performant than XTS mode (e.g., when using a h/w
accelerator which supports the former but not the latter) This avoids
a separate call into the AES cipher for every invocation.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The plain CBC driver and the CTS one share some code that iterates over
a scatterwalk and invokes the CBC asm code to do the processing. The
upcoming ESSIV/CBC mode will clone that pattern for the third time, so
let's factor it out first.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Instead of calling into the table based scalar AES code in situations
where the SIMD unit may not be used, use the generic AES code, which
is more appropriate since it is less likely to be susceptible to
timing attacks.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In preparation of duplicating the sync ctr(aes) functionality to modules
under arch/arm, move the helper function from a inline .h file to the
AES library, which is already depended upon by the drivers that use this
fallback.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Switch to the new AES library that also provides an implementation of
the AES key expansion routine. This removes the dependency on the
generic AES cipher, allowing it to be omitted entirely in the future.
While at it, remove some references to the table based arm64 version
of AES and replace them with AES library calls as well.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Switch to the new AES library that also provides an implementation of
the AES key expansion routine. This removes the dependency on the
generic AES cipher, allowing it to be omitted entirely in the future.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The CCM code calls directly into the scalar table based AES cipher for
arm64 from the fallback path, and since this implementation is known to
be non-time invariant, doing so from a time invariant SIMD cipher is a
bit nasty.
So let's switch to the AES library - this makes the code more robust,
and drops the dependency on the generic AES cipher, allowing us to
omit it entirely in the future.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The GHASH code uses the generic AES key expansion routines, and calls
directly into the scalar table based AES cipher for arm64 from the
fallback path, and since this implementation is known to be non-time
invariant, doing so from a time invariant SIMD cipher is a bit nasty.
So let's switch to the AES library - this makes the code more robust,
and drops the dependency on the generic AES cipher, allowing us to
omit it entirely in the future.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Rename some local AES encrypt/decrypt routines so they don't clash with
the names we are about to introduce for the routines exposed by the
generic AES library.
Signed-off-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 5.3:
API:
- Test shash interface directly in testmgr
- cra_driver_name is now mandatory
Algorithms:
- Replace arc4 crypto_cipher with library helper
- Implement 5 way interleave for ECB, CBC and CTR on arm64
- Add xxhash
- Add continuous self-test on noise source to drbg
- Update jitter RNG
Drivers:
- Add support for SHA204A random number generator
- Add support for 7211 in iproc-rng200
- Fix fuzz test failures in inside-secure
- Fix fuzz test failures in talitos
- Fix fuzz test failures in qat"
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (143 commits)
crypto: stm32/hash - remove interruptible condition for dma
crypto: stm32/hash - Fix hmac issue more than 256 bytes
crypto: stm32/crc32 - rename driver file
crypto: amcc - remove memset after dma_alloc_coherent
crypto: ccp - Switch to SPDX license identifiers
crypto: ccp - Validate the the error value used to index error messages
crypto: doc - Fix formatting of new crypto engine content
crypto: doc - Add parameter documentation
crypto: arm64/aes-ce - implement 5 way interleave for ECB, CBC and CTR
crypto: arm64/aes-ce - add 5 way interleave routines
crypto: talitos - drop icv_ool
crypto: talitos - fix hash on SEC1.
crypto: talitos - move struct talitos_edesc into talitos.h
lib/scatterlist: Fix mapping iterator when sg->offset is greater than PAGE_SIZE
crypto/NX: Set receive window credits to max number of CRBs in RxFIFO
crypto: asymmetric_keys - select CRYPTO_HASH where needed
crypto: serpent - mark __serpent_setkey_sbox noinline
crypto: testmgr - dynamically allocate crypto_shash
crypto: testmgr - dynamically allocate testvec_config
crypto: talitos - eliminate unneeded 'done' functions at build time
...
This implements 5-way interleaving for ECB, CBC decryption and CTR,
resulting in a speedup of ~11% on Marvell ThunderX2, which has a
very deep pipeline and therefore a high issue latency for NEON
instructions operating on the same registers.
Note that XTS is left alone: implementing 5-way interleave there
would either involve spilling of the calculated tweaks to the
stack, or recalculating them after the encryption operation, and
doing either of those would most likely penalize low end cores.
For ECB, this is not a concern at all, given that we have plenty
of spare registers. For CTR and CBC decryption, we take advantage
of the fact that v16 is not used by the CE version of the code
(which is the only one targeted by the optimization), and so we
can reshuffle the code a bit and avoid having to spill to memory
(with the exception of one extra reload in the CBC routine)
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In preparation of tweaking the accelerated AES chaining mode routines
to be able to use a 5-way stride, implement the core routines to
support processing 5 blocks of input at a time. While at it, drop
the 2 way versions, which have been unused for a while now.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Based on 2 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation #
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 4122 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Enrico Weigelt <info@metux.net>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Allison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190604081206.933168790@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Constify the ctx and iv arguments to crypto_chacha_init() and the
various chacha*_stream_xor() functions. This makes it clear that they
are not modified.
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>
The sha256-ce finup implementation for ARM64 produces wrong digest
for empty input (len=0). Expected: the actual digest, result: initial
value of SHA internal state. The error is in sha256_ce_finup:
for empty data `finalize` will be 1, so the code is relying on
sha2_ce_transform to make the final round. However, in
sha256_base_do_update, the block function will not be called when
len == 0.
Fix it by setting finalize to 0 if data is empty.
Fixes: 03802f6a80 ("crypto: arm64/sha2-ce - move SHA-224/256 ARMv8 implementation to base layer")
Cc: stable@vger.kernel.org
Signed-off-by: Elena Petrova <lenaptr@google.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The sha1-ce finup implementation for ARM64 produces wrong digest
for empty input (len=0). Expected: da39a3ee..., result: 67452301...
(initial value of SHA internal state). The error is in sha1_ce_finup:
for empty data `finalize` will be 1, so the code is relying on
sha1_ce_transform to make the final round. However, in
sha1_base_do_update, the block function will not be called when
len == 0.
Fix it by setting finalize to 0 if data is empty.
Fixes: 07eb54d306 ("crypto: arm64/sha1-ce - move SHA-1 ARMv8 implementation to base layer")
Cc: stable@vger.kernel.org
Signed-off-by: Elena Petrova <lenaptr@google.com>
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Based on 1 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license as published by
the free software foundation either version 2 of the license or at
your option any later version
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-or-later
has been chosen to replace the boilerplate/reference in 3029 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Allison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190527070032.746973796@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Pull crypto update from Herbert Xu:
"API:
- Add support for AEAD in simd
- Add fuzz testing to testmgr
- Add panic_on_fail module parameter to testmgr
- Use per-CPU struct instead multiple variables in scompress
- Change verify API for akcipher
Algorithms:
- Convert x86 AEAD algorithms over to simd
- Forbid 2-key 3DES in FIPS mode
- Add EC-RDSA (GOST 34.10) algorithm
Drivers:
- Set output IV with ctr-aes in crypto4xx
- Set output IV in rockchip
- Fix potential length overflow with hashing in sun4i-ss
- Fix computation error with ctr in vmx
- Add SM4 protected keys support in ccree
- Remove long-broken mxc-scc driver
- Add rfc4106(gcm(aes)) cipher support in cavium/nitrox"
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (179 commits)
crypto: ccree - use a proper le32 type for le32 val
crypto: ccree - remove set but not used variable 'du_size'
crypto: ccree - Make cc_sec_disable static
crypto: ccree - fix spelling mistake "protedcted" -> "protected"
crypto: caam/qi2 - generate hash keys in-place
crypto: caam/qi2 - fix DMA mapping of stack memory
crypto: caam/qi2 - fix zero-length buffer DMA mapping
crypto: stm32/cryp - update to return iv_out
crypto: stm32/cryp - remove request mutex protection
crypto: stm32/cryp - add weak key check for DES
crypto: atmel - remove set but not used variable 'alg_name'
crypto: picoxcell - Use dev_get_drvdata()
crypto: crypto4xx - get rid of redundant using_sd variable
crypto: crypto4xx - use sync skcipher for fallback
crypto: crypto4xx - fix cfb and ofb "overran dst buffer" issues
crypto: crypto4xx - fix ctr-aes missing output IV
crypto: ecrdsa - select ASN1 and OID_REGISTRY for EC-RDSA
crypto: ux500 - use ccflags-y instead of CFLAGS_<basename>.o
crypto: ccree - handle tee fips error during power management resume
crypto: ccree - add function to handle cryptocell tee fips error
...
If the user-provided IV needs to be aligned to the algorithm's
alignmask, then skcipher_walk_virt() copies the IV into a new aligned
buffer walk.iv. But skcipher_walk_virt() can fail afterwards, and then
if the caller unconditionally accesses walk.iv, it's a use-after-free.
xts-aes-neonbs doesn't set an alignmask, so currently it isn't affected
by this despite unconditionally accessing walk.iv. However this is more
subtle than desired, and unconditionally accessing walk.iv has caused a
real problem in other algorithms. Thus, update xts-aes-neonbs to start
checking the return value of skcipher_walk_virt().
Fixes: 1abee99eaf ("crypto: arm64/aes - reimplement bit-sliced ARM/NEON implementation for arm64")
Cc: <stable@vger.kernel.org> # v4.11+
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As we will exhaust the first 32 bits of AT_HWCAP let's start
exposing AT_HWCAP2 to userspace to give us up to 64 caps.
Whilst it's possible to use the remaining 32 bits of AT_HWCAP, we
prefer to expand into AT_HWCAP2 in order to provide a consistent
view to userspace between ILP32 and LP64. However internal to the
kernel we prefer to continue to use the full space of elf_hwcap.
To reduce complexity and allow for future expansion, we now
represent hwcaps in the kernel as ordinals and use a
KERNEL_HWCAP_ prefix. This allows us to support automatic feature
based module loading for all our hwcaps.
We introduce cpu_set_feature to set hwcaps which complements the
existing cpu_have_feature helper. These helpers allow us to clean
up existing direct uses of elf_hwcap and reduce any future effort
required to move beyond 64 caps.
For convenience we also introduce cpu_{have,set}_named_feature which
makes use of the cpu_feature macro to allow providing a hwcap name
without a {KERNEL_}HWCAP_ prefix.
Signed-off-by: Andrew Murray <andrew.murray@arm.com>
[will: use const_ilog2() and tweak documentation]
Signed-off-by: Will Deacon <will.deacon@arm.com>
My patches to make testmgr fuzz algorithms against their generic
implementation detected that the arm64 implementations of "cbcmac(aes)"
handle empty messages differently from the cbcmac template. Namely, the
arm64 implementations return the encrypted initial value, but the cbcmac
template returns the initial value directly.
This isn't actually a meaningful case because any user of cbcmac needs
to prepend the message length, as CCM does; otherwise it's insecure.
However, we should keep the behavior consistent; at the very least this
makes testing easier.
Do it the easy way, which is to change the arm64 implementations to have
the same behavior as the cbcmac template.
For what it's worth, ghash does things essentially the same way: it
returns its initial value when given an empty message, even though in
practice ghash is never passed an empty message.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Replace all calls to may_use_simd() in the arm64 crypto code with
crypto_simd_usable(), in order to allow testing the no-SIMD code paths.
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>
The arm64 gcm-aes-ce algorithm is failing the extra crypto self-tests
following my patches to test the !may_use_simd() code paths, which
previously were untested. The problem is that in the !may_use_simd()
case, an odd number of AES blocks can be processed within each step of
the skcipher_walk. However, the skcipher_walk is being done with a
"stride" of 2 blocks and is advanced by an even number of blocks after
each step. This causes the encryption to produce the wrong ciphertext
and authentication tag, and causes the decryption to incorrectly fail.
Fix it by only processing an even number of blocks per step.
Fixes: c2b24c36e0 ("crypto: arm64/aes-gcm-ce - fix scatterwalk API violation")
Fixes: 71e52c278c ("crypto: arm64/aes-ce-gcm - operate on two input blocks at a time")
Cc: <stable@vger.kernel.org> # v4.19+
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>
Pull crypto update from Herbert Xu:
"API:
- Add helper for simple skcipher modes.
- Add helper to register multiple templates.
- Set CRYPTO_TFM_NEED_KEY when setkey fails.
- Require neither or both of export/import in shash.
- AEAD decryption test vectors are now generated from encryption
ones.
- New option CONFIG_CRYPTO_MANAGER_EXTRA_TESTS that includes random
fuzzing.
Algorithms:
- Conversions to skcipher and helper for many templates.
- Add more test vectors for nhpoly1305 and adiantum.
Drivers:
- Add crypto4xx prng support.
- Add xcbc/cmac/ecb support in caam.
- Add AES support for Exynos5433 in s5p.
- Remove sha384/sha512 from artpec7 as hardware cannot do partial
hash"
[ There is a merge of the Freescale SoC tree in order to pull in changes
required by patches to the caam/qi2 driver. ]
* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (174 commits)
crypto: s5p - add AES support for Exynos5433
dt-bindings: crypto: document Exynos5433 SlimSSS
crypto: crypto4xx - add missing of_node_put after of_device_is_available
crypto: cavium/zip - fix collision with generic cra_driver_name
crypto: af_alg - use struct_size() in sock_kfree_s()
crypto: caam - remove redundant likely/unlikely annotation
crypto: s5p - update iv after AES-CBC op end
crypto: x86/poly1305 - Clear key material from stack in SSE2 variant
crypto: caam - generate hash keys in-place
crypto: caam - fix DMA mapping xcbc key twice
crypto: caam - fix hash context DMA unmap size
hwrng: bcm2835 - fix probe as platform device
crypto: s5p-sss - Use AES_BLOCK_SIZE define instead of number
crypto: stm32 - drop pointless static qualifier in stm32_hash_remove()
crypto: chelsio - Fixed Traffic Stall
crypto: marvell - Remove set but not used variable 'ivsize'
crypto: ccp - Update driver messages to remove some confusion
crypto: adiantum - add 1536 and 4096-byte test vectors
crypto: nhpoly1305 - add a test vector with len % 16 != 0
crypto: arm/aes-ce - update IV after partial final CTR block
...
On big endian arm64 kernels, the xchacha20-neon and xchacha12-neon
self-tests fail because hchacha_block_neon() outputs little endian words
but the C code expects native endianness. Fix it to output the words in
native endianness (which also makes it match the arm32 version).
Fixes: cc7cf991e9 ("crypto: arm64/chacha20 - add XChaCha20 support")
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>
The change to encrypt a fifth ChaCha block using scalar instructions
caused the chacha20-neon, xchacha20-neon, and xchacha12-neon self-tests
to start failing on big endian arm64 kernels. The bug is that the
keystream block produced in 32-bit scalar registers is directly XOR'd
with the data words, which are loaded and stored in native endianness.
Thus in big endian mode the data bytes end up XOR'd with the wrong
bytes. Fix it by byte-swapping the keystream words in big endian mode.
Fixes: 2fe55987b2 ("crypto: arm64/chacha - use combined SIMD/ALU routine for more speed")
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>
Make the arm64 ctr-aes-neon and ctr-aes-ce algorithms update the IV
buffer to contain the next counter after processing a partial final
block, rather than leave it as the last counter. This makes these
algorithms pass the updated AES-CTR tests.
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The arm64 NEON bit-sliced implementation of AES-CTR fails the improved
skcipher tests because it sometimes produces the wrong ciphertext. The
bug is that the final keystream block isn't returned from the assembly
code when the number of non-final blocks is zero. This can happen if
the input data ends a few bytes after a page boundary. In this case the
last bytes get "encrypted" by XOR'ing them with uninitialized memory.
Fix the assembly code to return the final keystream block when needed.
Fixes: 88a3f582be ("crypto: arm64/aes - don't use IV buffer to return final keystream block")
Cc: <stable@vger.kernel.org> # v4.11+
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The x86, arm, and arm64 asm implementations of crct10dif are very
difficult to understand partly because many of the comments, labels, and
macros are named incorrectly: the lengths mentioned are usually off by a
factor of two from the actual code. Many other things are unnecessarily
convoluted as well, e.g. there are many more fold constants than
actually needed and some aren't fully reduced.
This series therefore cleans up all these implementations to be much
more maintainable. I also made some small optimizations where I saw
opportunities, resulting in slightly better performance.
This patch cleans up the arm64 version.
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The arm64 CRC-T10DIF implementation either uses 8-bit or 64-bit
polynomial multiplication instructions, since the latter are
faster but not mandatory in the architecture.
Since that prevents us from testing both implementations on the
same system, let's expose both implementations to the crypto API,
with the priorities reflecting that the P64 version is the
preferred one if available.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Remove some code that is no longer called now that we make sure never
to invoke the SIMD routine with less than 16 bytes of input.
Reviewed-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The SIMD routine ported from x86 used to have a special code path
for inputs < 16 bytes, which got lost somewhere along the way.
Instead, the current glue code aligns the input pointer to 16 bytes,
which is not really necessary on this architecture (although it
could be beneficial to performance to expose aligned data to the
the NEON routine), but this could result in inputs of less than
16 bytes to be passed in. This not only fails the new extended
tests that Eric has implemented, it also results in the code
reading past the end of the input, which could potentially result
in crashes when dealing with less than 16 bytes of input at the
end of a page which is followed by an unmapped page.
So update the glue code to only invoke the NEON routine if the
input is at least 16 bytes.
Reported-by: Eric Biggers <ebiggers@kernel.org>
Reviewed-by: Eric Biggers <ebiggers@kernel.org>
Fixes: 6ef5737f39 ("crypto: arm64/crct10dif - port x86 SSE implementation to arm64")
Cc: <stable@vger.kernel.org> # v4.10+
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The arm64 GHASH implementation either uses 8-bit or 64-bit
polynomial multiplication instructions, since the latter are
faster but not mandatory in the architecture.
Since that prevents us from testing both implementations on the
same system, let's expose both implementations to the crypto API,
with the priorities reflecting that the P64 version is the
preferred one if available.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When the AES-CCM code was first added, the NEON register were saved
and restored eagerly, and so the code avoided doing so, and executed
the scatterwalk in atomic context inside the kernel_neon_begin/end
section.
This has been changed in the meantime, so switch to non-atomic
scatterwalks.
Fixes: bd2ad885e3 ("crypto: arm64/aes-ce-ccm - move kernel mode neon ...")
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Commit 5092fcf349 ("crypto: arm64/aes-ce-ccm: add non-SIMD generic
fallback") introduced C fallback code to replace the NEON routines
when invoked from a context where the NEON is not available (i.e.,
from the context of a softirq taken while the NEON is already being
used in kernel process context)
Fix two logical flaws in the MAC calculation of the associated data.
Reported-by: Eric Biggers <ebiggers@kernel.org>
Fixes: 5092fcf349 ("crypto: arm64/aes-ce-ccm: add non-SIMD generic fallback")
Cc: stable@vger.kernel.org
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The NEON MAC calculation routine fails to handle the case correctly
where there is some data in the buffer, and the input fills it up
exactly. In this case, we enter the loop at the end with w8 == 0,
while a negative value is assumed, and so the loop carries on until
the increment of the 32-bit counter wraps around, which is quite
obviously wrong.
So omit the loop altogether in this case, and exit right away.
Reported-by: Eric Biggers <ebiggers@kernel.org>
Fixes: a3fd82105b ("arm64/crypto: AES in CCM mode using ARMv8 Crypto ...")
Cc: stable@vger.kernel.org
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
To some degree, most known AArch64 micro-architectures appear to be
able to issue ALU instructions in parellel to SIMD instructions
without affecting the SIMD throughput. This means we can use the ALU
to process a fifth ChaCha block while the SIMD is processing four
blocks in parallel.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Update the 4-way NEON ChaCha routine so it can handle input of any
length >64 bytes in its entirety, rather than having to call into
the 1-way routine and/or memcpy()s via temp buffers to handle the
tail of a ChaCha invocation that is not a multiple of 256 bytes.
On inputs that are a multiple of 256 bytes (and thus in tcrypt
benchmarks), performance drops by around 1% on Cortex-A57, while
performance for inputs drawn randomly from the range [64, 1024)
increases by around 30%.
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Now that the ARM64 NEON implementation of ChaCha20 and XChaCha20 has
been refactored to support varying the number of rounds, add support for
XChaCha12. This is identical to XChaCha20 except for the number of
rounds, which is 12 instead of 20. This can be used by Adiantum.
Reviewed-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Ard Biesheuvel