While in this particular case it would not be a (critical) issue,
the pattern itself is bad and error prone in case the location
of the parameter is changed.

Don't cast parameter to unsigned long pointer in the bit operations.
Instead copy to a local variable on stack of a proper type and use.

Fixes: cf718ea ("crypto: cavium/nitrox - Enabled Mailbox support")
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

The ARM version of the accelerated XOR routines are simply the 8-way C
routines passed through the auto-vectorizer with SIMD codegen enabled.
This used to require GCC version 4.6 at least, but given that 5.1 is now
the baseline, this check is no longer necessary, and actually
misidentifies Clang as GCC < 4.6 as Clang defines the GCC major/minor as
well, but makes no attempt at doing this in a way that conveys feature
parity with a certain version of GCC (which would not be a great idea in
the first place).

So let's drop the version check, and make the auto-vectorize pragma
(which is based on a GCC-specific command line option) GCC-only. Since
Clang performs SIMD auto-vectorization by default at -O2, no pragma is
necessary here.

Tested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Link: ClangBuiltLinux#496
Link: ClangBuiltLinux#503
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Modern compilers are perfectly capable of extracting parallelism from
the XOR routines, provided that the prototypes reflect the nature of the
input accurately, in particular, the fact that the input vectors are
expected not to overlap. This is not documented explicitly, but is
implied by the interchangeability of the various C routines, some of
which use temporary variables while others don't: this means that these
routines only behave identically for non-overlapping inputs.

So let's decorate these input vectors with the __restrict modifier,
which informs the compiler that there is no overlap. While at it, make
the input-only vectors pointer-to-const as well.

Tested-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Link: ClangBuiltLinux#563
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

… SoC

Add support for the TDES IP found on sama7g5.

Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

FIPS 140 requires a minimum security strength of 112 bits. This implies
that the HMAC key must not be smaller than 112 in FIPS mode.

This restriction implies that the test vectors for HMAC that have a key
that is smaller than 112 bits must be disabled when FIPS support is
compiled.

Signed-off-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

By adding the support for the flag fips_skip, hash / HMAC test vectors
may be marked to be not applicable in FIPS mode. Such vectors are
silently skipped in FIPS mode.

Signed-off-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

…nd out_sg

Pointers in_sg and out_sg are being initialized with values that are
never read, they are being re-assigned the same values later on. The
initializations are redundant, remove them in preference to the later
assignments that are closer to when the pointers are being used.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

…e c_size

Variable c_size is being initialized with a value that is never read, it
is being re-assigned with a different value later on. The initialization
is redundant and can be removed.

Signed-off-by: Colin Ian King <colin.i.king@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

No issues were found while using the driver with dm-crypt enabled. So
CONFIG_DM_CRYPT check in the driver can be removed.

This also fixes the NULL pointer dereference in driver release if
CONFIG_DM_CRYPT is enabled.

...
Unable to handle kernel NULL pointer dereference at virtual address 0000000000000008
...
Call trace:
 crypto_unregister_alg+0x68/0xfc
 crypto_unregister_skciphers+0x44/0x60
 otx2_cpt_crypto_exit+0x100/0x1a0
 otx2_cptvf_remove+0xf8/0x200
 pci_device_remove+0x3c/0xd4
 __device_release_driver+0x188/0x234
 device_release_driver+0x2c/0x4c
...

Fixes: 6f03f0e ("crypto: octeontx2 - register with linux crypto framework")
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The multibuffer algorithms was removed already in 2018, so it is
necessary to clear the test code left by tcrypt.

Suggested-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch adds support for hardware version of AES and SHA IPs
available on lan966x SoC.

Signed-off-by: Kavyasree Kotagiri <kavyasree.kotagiri@microchip.com>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Tested-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

In case the entropy quality is low, there may be less than one bit to
credit in the call to add_hwgenerator_randomness(): The number of bytes
returned by rng_get_data() multiplied by the current quality (in entropy
bits per 1024 bits of input) must be larger than 128 to credit at least
one bit. However, imx-rngc.c sets the quality to 19, but may return less
than 32 bytes; hid_u2fzero.c sets the quality to 1; and users may override
the quality setting manually.

In case there is less than one bit to credit, keep track of it and add
that credit to the next iteration.

Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

…t sizes

Even though the kernel's implementations of AES-XTS were updated to
implement ciphertext stealing and can operate on inputs of any size
larger than or equal to the AES block size, this feature is rarely used
in practice.

In fact, in the kernel, AES-XTS is only used to operate on 4096 or 512
byte blocks, which means that not only the ciphertext stealing is
effectively dead code, the logic in the bit sliced NEON implementation
to deal with fewer than 8 blocks at a time is also never used.

Since the bit-sliced NEON driver already depends on the plain NEON
version, which is slower but can operate on smaller data quantities more
straightforwardly, let's fallback to the plain NEON implementation of
XTS for any residual inputs that are not multiples of 128 bytes. This
allows us to remove a lot of complicated logic that rarely gets
exercised in practice.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Instead of processing the entire input with the 8-way bit sliced
algorithm, which is sub-optimal for inputs that are not a multiple of
128 bytes in size, invoke the plain NEON version of CTR for the
remainder of the input after processing the bulk using 128 byte strides.

This allows us to greatly simplify the asm code that implements CTR, and
get rid of all the branches and special code paths. It also gains us a
couple of percent of performance.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Instead of falling back to C code to deal with the final bit of input
that is not a round multiple of the block size, handle this in the asm
code, permitting us to use overlapping loads and stores for performance,
and implement the 16-byte wide XOR using a single NEON instruction.

Since NEON loads and stores have a natural width of 16 bytes, we need to
handle inputs of less than 16 bytes in a special way, but this rarely
occurs in practice so it does not impact performance. All other input
sizes can be consumed directly by the NEON asm code, although it should
be noted that the core AES transform can still only process 128 bytes (8
AES blocks) at a time.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Instead of falling back to C code to do a memcpy of the output of the
last block, handle this in the asm code directly if possible, which is
the case if the entire input is longer than 16 bytes.

Cc: Nathan Huckleberry <nhuck@google.com>
Cc: Eric Biggers <ebiggers@google.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

LDWB is getting incorrectly used in HW when
CPT_AF_LF()_PTR_CTL[IQB_LDWB]=1 and CPT instruction queue has less than
320 free entries. So, increase HW instruction queue size by 320 and give
320 entries less for SW/NIX RX as a SW workaround.

Signed-off-by: Srujana Challa <schalla@marvell.com>
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

When CPT_AF_DIAG[FLT_DIS] = 0 and a CPT engine access to
LLC/DRAM encounters a fault/poison, a rare case may result
in unpredictable data being delivered to a CPT engine.
So, this patch adds code to set FLT_DIS as a workaround.

Signed-off-by: Srujana Challa <schalla@marvell.com>
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

When software sets CPT_AF_CTL[RNM_REQ_EN]=1 and RNM in not producing
entropy(i.e., RNM_ENTROPY_STATUS[NORMAL_CNT] < 0x40), the first cycle of
the response may be lost due to a conditional clocking issue. Due to
this, the subsequent random number stream will be corrupted. So, this
patch adds support to ensure RNM_ENTROPY_STATUS[NORMAL_CNT] = 0x40
before writing CPT_AF_CTL[RNM_REQ_EN] = 1, as a workaround.

Signed-off-by: Srujana Challa <schalla@marvell.com>
Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

For two reasons, current_quality may become zero within the rngd
kernel thread: (1) The user lowers current_quality to 0 by writing
to the sysfs module parameter file (note that increasing the quality
from zero is without effect at the moment), or (2) there are two or
more hwrng devices registered, and those which provide quality>0 are
unregistered, but one with quality==0 remains.

If current_quality is 0, the randomness is not trusted and cannot help
to increase the entropy count. That will lead to continuous calls to
the hwrngd thread and continuous stirring of the input pool with
untrusted bits.

Cc: Matt Mackall <mpm@selenic.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

In case the user-specified rng device is not working, it is not used;
therefore cur_rng_set_by_user must not be set to 1.

Cc: Matt Mackall <mpm@selenic.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Using rng_buffer in add_early_randomness() may race with rng_dev_read().
Use rng_fillbuf instead, as it is otherwise only used within the kernel
by hwrng_fillfn() and therefore never exposed to userspace.

Cc: Matt Mackall <mpm@selenic.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

…ytes

According to <linux/hw_random.h>, the @max parameter of the ->read
callback "is a multiple of 4 and >= 32 bytes". That promise was not
kept by add_early_randomness(), which only asked for 16 bytes. As
rng_buffer_size() is at least 32, we can simply ask for 32 bytes.

Cc: Matt Mackall <mpm@selenic.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

hw-random device drivers depend on the hw-random core being
initialized. Make this ordering explicit, also for the case
these drivers are built-in. As the core itself depends on
misc_register() which is set up at subsys_initcall time,
advance the initialization of the core (only) to the
fs_initcall() level.

Cc: Matt Mackall <mpm@selenic.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

padata_do_parallel() calls cpumask_weight() to check if any bit of a
given cpumask is set. We can do it more efficiently with cpumask_empty()
because cpumask_empty() stops traversing the cpumask as soon as it finds
first set bit, while cpumask_weight() counts all bits unconditionally.

Signed-off-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch fixes a bug in scatterlist processing that may cause incorrect AES block encryption/decryption.

Fixes: 2e6d793 ("crypto: mxs-dcp - Use sg_mapping_iter to copy data")
Signed-off-by: Tomas Paukrt <tomaspaukrt@email.cz>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The kernel test rebot report this warning: Uninitialized variable: ret.
The code flow may return value of ret directly. This value is an
uninitialized variable, here is fix it.

Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Use the correct print format. Printing an unsigned int value should
use %u instead of %d.

Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The CTR counter is 32bit rollover default on the BD.
But the NIST standard is 128bit rollover. it cause the
testing failed, so need to fix the BD configuration.

Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Fix the maximum length of AAD for the CCM mode due to the hardware limited.

Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Modify the print of information that might lead to user misunderstanding.
Currently only XTS mode need the fallback tfm when using 192bit key.
Others algs not need soft fallback tfm. So others algs can return
directly.

Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Fixup icv(integrity check value) checking enabled wrong on
Kunpeng 930

Signed-off-by: Kai Ye <yekai13@huawei.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

OcteonTX2 CPT driver will fail to link without devlink support.

aarch64-linux-gnu-ld: otx2_cpt_devlink.o: in function `otx2_cpt_dl_egrp_delete':
otx2_cpt_devlink.c:18: undefined reference to `devlink_priv'
aarch64-linux-gnu-ld: otx2_cpt_devlink.o: in function `otx2_cpt_dl_egrp_create':
otx2_cpt_devlink.c:9: undefined reference to `devlink_priv'
aarch64-linux-gnu-ld: otx2_cpt_devlink.o: in function `otx2_cpt_dl_uc_info':
otx2_cpt_devlink.c:27: undefined reference to `devlink_priv'

Fixes: fed8f4d ("crypto: octeontx2 - parameters for custom engine groups")

Signed-off-by: Shijith Thotton <sthotton@marvell.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The C standard does not support dereferencing pointers that are not
aligned with respect to the pointed-to type, and doing so is technically
undefined behavior, even if the underlying hardware supports it.

This means that conditionally dereferencing such pointers based on
whether CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS=y is not the right thing
to do, and actually results in alignment faults on ARM, which are fixed
up on a slow path. Instead, we should use the unaligned accessors in
such cases: on architectures that don't care about alignment, they will
result in identical codegen whereas, e.g., codegen on ARM will avoid
doubleword loads and stores but use ordinary ones, which are able to
tolerate misalignment.

Link: https://lore.kernel.org/linux-crypto/CAHk-=wiKkdYLY0bv+nXrcJz3NH9mAqPAafX7PpW5EwVtxsEu7Q@mail.gmail.com/
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The function crypto_authenc_decrypt_tail discards its flags
argument and always relies on the flags from the original request
when starting its sub-request.

This is clearly wrong as it may cause the SLEEPABLE flag to be
set when it shouldn't.

Fixes: 92d95ba ("crypto: authenc - Convert to new AEAD interface")
Reported-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Tested-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>