Clang sometimes makes very different inlining decisions from gcc.
In case of the aegis crypto algorithms, it decides to turn the innermost
primitives (and, xor, ...) into separate functions but inline most of
the rest.

This results in a huge amount of variables spilled on the stack, leading
to rather slow execution as well as kernel stack usage beyond the 32-bit
warning limit when CONFIG_KASAN is enabled:

crypto/aegis256.c:123:13: warning: stack frame size of 648 bytes in function 'crypto_aegis256_encrypt_chunk' [-Wframe-larger-than=]
crypto/aegis256.c:366:13: warning: stack frame size of 1264 bytes in function 'crypto_aegis256_crypt' [-Wframe-larger-than=]
crypto/aegis256.c:187:13: warning: stack frame size of 656 bytes in function 'crypto_aegis256_decrypt_chunk' [-Wframe-larger-than=]
crypto/aegis128l.c:135:13: warning: stack frame size of 832 bytes in function 'crypto_aegis128l_encrypt_chunk' [-Wframe-larger-than=]
crypto/aegis128l.c:415:13: warning: stack frame size of 1480 bytes in function 'crypto_aegis128l_crypt' [-Wframe-larger-than=]
crypto/aegis128l.c:218:13: warning: stack frame size of 848 bytes in function 'crypto_aegis128l_decrypt_chunk' [-Wframe-larger-than=]
crypto/aegis128.c:116:13: warning: stack frame size of 584 bytes in function 'crypto_aegis128_encrypt_chunk' [-Wframe-larger-than=]
crypto/aegis128.c:351:13: warning: stack frame size of 1064 bytes in function 'crypto_aegis128_crypt' [-Wframe-larger-than=]
crypto/aegis128.c:177:13: warning: stack frame size of 592 bytes in function 'crypto_aegis128_decrypt_chunk' [-Wframe-larger-than=]

Forcing the primitives to all get inlined avoids the issue and the
resulting code is similar to what gcc produces.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Acked-by: NNick Desaulniers <ndesaulniers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Currently, NETLINK_CRYPTO works only in the init network namespace. It
doesn't make much sense to cut it out of the other network namespaces,
so do the minor plumbing work necessary to make it work in any network
namespace. Code inspired by net/core/sock_diag.c.

Tested using kcapi-dgst from libkcapi [1]:
Before:
    # unshare -n kcapi-dgst -c sha256 </dev/null | wc -c
    libkcapi - Error: Netlink error: sendmsg failed
    libkcapi - Error: Netlink error: sendmsg failed
    libkcapi - Error: NETLINK_CRYPTO: cannot obtain cipher information for hmac(sha512) (is required crypto_user.c patch missing? see documentation)
    0

After:
    # unshare -n kcapi-dgst -c sha256 </dev/null | wc -c
    32

[1] https://github.com/smuellerDD/libkcapiSigned-off-by: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Reviewed-by: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert 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: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Add some plumbing to allow the AEGIS128 code to be built with SIMD
routines for acceleration.
Reviewed-by: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The generic AES code provides four sets of lookup tables, where each
set consists of four tables containing the same 32-bit values, but
rotated by 0, 8, 16 and 24 bits, respectively. This makes sense for
CISC architectures such as x86 which support memory operands, but
for other architectures, the rotates are quite cheap, and using all
four tables needlessly thrashes the D-cache, and actually hurts rather
than helps performance.

Since x86 already has its own implementation of AEGIS based on AES-NI
instructions, let's tweak the generic implementation towards other
architectures, and avoid the prerotated tables, and perform the
rotations inline. On ARM Cortex-A53, this results in a ~8% speedup.
Acked-by: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

TFM init/exit routines are optional, so no need to provide empty ones.
Reviewed-by: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert 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: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert 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.pdfReviewed-by: NOndrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Add self-tests for the lzo-rle algorithm.
Signed-off-by: NHannah Pan <hannahpan@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The versions of the AES lookup tables that are only used during the last
round are never used outside of the driver, so there is no need to
export their symbols.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Drop aes-generic's version of crypto_aes_expand_key(), and switch to
the key expansion routine provided by the AES library. AES key expansion
is not performance critical, and it is better to have a single version
shared by all AES implementations.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The AES assembler code for x86 isn't actually faster than code
generated by the compiler from aes_generic.c, and considering
the disproportionate maintenance burden of assembler code on
x86, it is better just to drop it entirely. Modern x86 systems
will use AES-NI anyway, and given that the modules being removed
have a dependency on aes_generic already, we can remove them
without running the risk of regressions.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The AES-NI code contains fallbacks for invocations that occur from a
context where the SIMD unit is unavailable, which really only occurs
when running in softirq context that was entered from a hard IRQ that
was taken while running kernel code that was already using the FPU.

That means performance is not really a consideration, and we can just
use the new library code for this use case, which has a smaller
footprint and is believed to be time invariant. This will allow us to
drop the non-SIMD asm routines in a subsequent patch.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Take the existing small footprint and mostly time invariant C code
and turn it into a AES library that can be used for non-performance
critical, casual use of AES, and as a fallback for, e.g., SIMD code
that needs a secondary path that can be taken in contexts where the
SIMD unit is off limits (e.g., in hard interrupts taken from kernel
context)
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The fixed time AES code mangles the key schedule so that xoring the
first round key with values at fixed offsets across the Sbox produces
the correct value. This primes the D-cache with the entire Sbox before
any data dependent lookups are done, making it more difficult to infer
key bits from timing variances when the plaintext is known.

The downside of this approach is that it renders the key schedule
incompatible with other implementations of AES in the kernel, which
makes it cumbersome to use this implementation as a fallback for SIMD
based AES in contexts where this is not allowed.

So let's tweak the fixed Sbox indexes so that they add up to zero under
the xor operation. While at it, increase the granularity to 16 bytes so
we cover the entire Sbox even on systems with 16 byte cachelines.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert 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: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Crypto test failures in FIPS mode cause an immediate panic, but
on some system the cryptographic boundary extends beyond just
the Linux controlled domain.

Add a simple atomic notification chain to allow interested parties
to register to receive notification prior to us kicking the bucket.
Signed-off-by: NGilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Michal Suchanek reported [1] that running the pcrypt_aead01 test from
LTP [2] in a loop and holding Ctrl-C causes a NULL dereference of
alg->cra_users.next in crypto_remove_spawns(), via crypto_del_alg().
The test repeatedly uses CRYPTO_MSG_NEWALG and CRYPTO_MSG_DELALG.

The crash occurs when the instance that CRYPTO_MSG_DELALG is trying to
unregister isn't a real registered algorithm, but rather is a "test
larval", which is a special "algorithm" added to the algorithms list
while the real algorithm is still being tested.  Larvals don't have
initialized cra_users, so that causes the crash.  Normally pcrypt_aead01
doesn't trigger this because CRYPTO_MSG_NEWALG waits for the algorithm
to be tested; however, CRYPTO_MSG_NEWALG returns early when interrupted.

Everything else in the "crypto user configuration" API has this same bug
too, i.e. it inappropriately allows operating on larval algorithms
(though it doesn't look like the other cases can cause a crash).

Fix this by making crypto_alg_match() exclude larval algorithms.

[1] https://lkml.kernel.org/r/20190625071624.27039-1-msuchanek@suse.de
[2] https://github.com/linux-test-project/ltp/blob/20190517/testcases/kernel/crypto/pcrypt_aead01.cReported-by: NMichal Suchanek <msuchanek@suse.de>
Fixes: a38f7907 ("crypto: Add userspace configuration API")
Cc: <stable@vger.kernel.org> # v3.2+
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

cryptd_skcipher_free() fails to free the struct skcipher_instance
allocated in cryptd_create_skcipher(), leading to a memory leak.  This
is detected by kmemleak on bootup on ARM64 platforms:

 unreferenced object 0xffff80003377b180 (size 1024):
   comm "cryptomgr_probe", pid 822, jiffies 4294894830 (age 52.760s)
   backtrace:
     kmem_cache_alloc_trace+0x270/0x2d0
     cryptd_create+0x990/0x124c
     cryptomgr_probe+0x5c/0x1e8
     kthread+0x258/0x318
     ret_from_fork+0x10/0x1c

Fixes: 4e0958d1 ("crypto: cryptd - Add support for skcipher")
Cc: <stable@vger.kernel.org>
Signed-off-by: NVincent Whitchurch <vincent.whitchurch@axis.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Build testing with some core crypto options disabled revealed
a few modules that are missing CRYPTO_HASH:

crypto/asymmetric_keys/x509_public_key.o: In function `x509_get_sig_params':
x509_public_key.c:(.text+0x4c7): undefined reference to `crypto_alloc_shash'
x509_public_key.c:(.text+0x5e5): undefined reference to `crypto_shash_digest'
crypto/asymmetric_keys/pkcs7_verify.o: In function `pkcs7_digest.isra.0':
pkcs7_verify.c:(.text+0xab): undefined reference to `crypto_alloc_shash'
pkcs7_verify.c:(.text+0x1b2): undefined reference to `crypto_shash_digest'
pkcs7_verify.c:(.text+0x3c1): undefined reference to `crypto_shash_update'
pkcs7_verify.c:(.text+0x411): undefined reference to `crypto_shash_finup'

This normally doesn't show up in randconfig tests because there is
a large number of other options that select CRYPTO_HASH.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The same bug that gcc hit in the past is apparently now showing
up with clang, which decides to inline __serpent_setkey_sbox:

crypto/serpent_generic.c:268:5: error: stack frame size of 2112 bytes in function '__serpent_setkey' [-Werror,-Wframe-larger-than=]

Marking it 'noinline' reduces the stack usage from 2112 bytes to
192 and 96 bytes, respectively, and seems to generate more
useful object code.

Fixes: c871c10e ("crypto: serpent - improve __serpent_setkey with UBSAN")
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Reviewed-by: NEric Biggers <ebiggers@kernel.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

The largest stack object in this file is now the shash descriptor.
Since there are many other stack variables, this can push it
over the 1024 byte warning limit, in particular with clang and
KASAN:

crypto/testmgr.c:1693:12: error: stack frame size of 1312 bytes in function '__alg_test_hash' [-Werror,-Wframe-larger-than=]

Make test_hash_vs_generic_impl() do the same thing as the
corresponding eaed and skcipher functions by allocating the
descriptor dynamically. We can still do better than this,
but it brings us well below the 1024 byte limit.
Suggested-by: NEric Biggers <ebiggers@kernel.org>
Fixes: 9a8a6b3f ("crypto: testmgr - fuzz hashes against their generic implementation")
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Reviewed-by: NEric Biggers <ebiggers@kernel.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

On arm32, we get warnings about high stack usage in some of the functions:

crypto/testmgr.c:2269:12: error: stack frame size of 1032 bytes in function 'alg_test_aead' [-Werror,-Wframe-larger-than=]
static int alg_test_aead(const struct alg_test_desc *desc, const char *driver,
           ^
crypto/testmgr.c:1693:12: error: stack frame size of 1312 bytes in function '__alg_test_hash' [-Werror,-Wframe-larger-than=]
static int __alg_test_hash(const struct hash_testvec *vecs,
           ^

On of the larger objects on the stack here is struct testvec_config, so
change that to dynamic allocation.

Fixes: 40153b10 ("crypto: testmgr - fuzz AEADs against their generic implementation")
Fixes: d435e10e ("crypto: testmgr - fuzz skciphers against their generic implementation")
Fixes: 9a8a6b3f ("crypto: testmgr - fuzz hashes against their generic implementation")
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Reviewed-by: NEric Biggers <ebiggers@kernel.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Add a 'recurse' flag for keyring searches so that the flag can be omitted
and recursion disabled, thereby allowing just the nominated keyring to be
searched and none of the children.
Signed-off-by: NDavid Howells <dhowells@redhat.com>

There are no remaining users of the cipher implementation, and there
are no meaningful ways in which the arc4 cipher can be combined with
templates other than ECB (and the way we do provide that combination
is highly dubious to begin with).

So let's drop the arc4 cipher altogether, and only keep the ecb(arc4)
skcipher, which is used in various places in the kernel.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Refactor the core rc4 handling so we can move most users to a library
interface, permitting us to drop the cipher interface entirely in a
future patch. This is part of an effort to simplify the crypto API
and improve its robustness against incorrect use.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert 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: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NEnrico Weigelt <info@metux.net>
Reviewed-by: NKate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: NAllison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190604081206.933168790@linutronix.deSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

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 version 2 as
  published by the free software foundation this program is
  distributed in the hope that it will be useful but without any
  warranty without even the implied warranty of merchantability or
  fitness for a particular purpose see the gnu general public license
  for more details you should have received a copy of the gnu general
  public license along with this program if not see http www gnu org
  licenses

extracted by the scancode license scanner the SPDX license identifier

  GPL-2.0-only

has been chosen to replace the boilerplate/reference in 503 file(s).
Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NAlexios Zavras <alexios.zavras@intel.com>
Reviewed-by: NAllison Randal <allison@lohutok.net>
Reviewed-by: NEnrico Weigelt <info@metux.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190602204653.811534538@linutronix.deSigned-off-by: NGreg Kroah-Hartman <gregkh@linuxfoundation.org>

The wusb code takes a very peculiar approach at implementing CBC-MAC,
by using plain CBC into a scratch buffer, and taking the output IV
as the MAC.

We can clean up this code substantially by switching to the cbcmac
shash, as exposed by the CCM template. To ensure that the module is
loaded on demand, add the cbcmac template name as a module alias.
Signed-off-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NGreg 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: NEric Biggers <ebiggers@google.com>
Acked-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

- Use sg_init_one() instead of sg_init_table() then sg_set_buf().

- Remove unneeded calls to sg_init_table() prior to scatterwalk_ffwd().

- Simplify initializing the poly tail block.

- Simplify computing padlen.

This doesn't change any actual behavior.

Cc: Martin Willi <martin@strongswan.org>
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

crypto_skcipher_encrypt() and crypto_skcipher_decrypt() have grown to be
more than a single indirect function call.  They now also check whether
a key has been set, and with CONFIG_CRYPTO_STATS=y they also update the
crypto statistics.  That can add up to a lot of bloat at every call
site.  Moreover, these always involve a function call anyway, which
greatly limits the benefits of inlining.

So change them to be non-inline.
Signed-off-by: NEric Biggers <ebiggers@google.com>
Acked-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

crypto_aead_encrypt() and crypto_aead_decrypt() have grown to be more
than a single indirect function call.  They now also check whether a key
has been set, the decryption side checks whether the input is at least
as long as the authentication tag length, and with CONFIG_CRYPTO_STATS=y
they also update the crypto statistics.  That can add up to a lot of
bloat at every call site.  Moreover, these always involve a function
call anyway, which greatly limits the benefits of inlining.

So change them to be non-inline.
Signed-off-by: NEric Biggers <ebiggers@google.com>
Acked-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Call cond_resched() after each fuzz test iteration.  This avoids stall
warnings if fuzz_iterations is set very high for testing purposes.

While we're at it, also call cond_resched() after finishing testing each
test vector.
Signed-off-by: NEric Biggers <ebiggers@google.com>
Acked-by: NArd Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Now that all algorithms explicitly set cra_driver_name, make it required
for algorithm registration and remove the code that generated a default
cra_driver_name.

Also add an explicit check that cra_name is set too, since that's
obviously required too, yet it didn't seem to be checked anywhere.
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Most generic crypto algorithms declare a driver name ending in
"-generic".  The rest don't declare a driver name and instead rely on
the crypto API automagically appending "-generic" upon registration.

Having multiple conventions is unnecessarily confusing and makes it
harder to grep for all generic algorithms in the kernel source tree.
But also, allowing NULL driver names is problematic because sometimes
people fail to set it, e.g. the case fixed by commit 41798036
("crypto: cavium/zip - fix collision with generic cra_driver_name").

Of course, people can also incorrectly name their drivers "-generic".
But that's much easier to notice / grep for.

Therefore, let's make cra_driver_name mandatory.  In preparation for
this, this patch makes all generic algorithms set cra_driver_name.
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Clear the CRYPTO_TFM_REQ_MAY_SLEEP flag when the chacha20poly1305
operation is being continued from an async completion callback, since
sleeping may not be allowed in that context.

This is basically the same bug that was recently fixed in the xts and
lrw templates.  But, it's always been broken in chacha20poly1305 too.
This was found using syzkaller in combination with the updated crypto
self-tests which actually test the MAY_SLEEP flag now.

Reproducer:

    python -c 'import socket; socket.socket(socket.AF_ALG, 5, 0).bind(
    	       ("aead", "rfc7539(cryptd(chacha20-generic),poly1305-generic)"))'

Kernel output:

    BUG: sleeping function called from invalid context at include/crypto/algapi.h:426
    in_atomic(): 1, irqs_disabled(): 0, pid: 1001, name: kworker/2:2
    [...]
    CPU: 2 PID: 1001 Comm: kworker/2:2 Not tainted 5.2.0-rc2 #5
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-20181126_142135-anatol 04/01/2014
    Workqueue: crypto cryptd_queue_worker
    Call Trace:
     __dump_stack lib/dump_stack.c:77 [inline]
     dump_stack+0x4d/0x6a lib/dump_stack.c:113
     ___might_sleep kernel/sched/core.c:6138 [inline]
     ___might_sleep.cold.19+0x8e/0x9f kernel/sched/core.c:6095
     crypto_yield include/crypto/algapi.h:426 [inline]
     crypto_hash_walk_done+0xd6/0x100 crypto/ahash.c:113
     shash_ahash_update+0x41/0x60 crypto/shash.c:251
     shash_async_update+0xd/0x10 crypto/shash.c:260
     crypto_ahash_update include/crypto/hash.h:539 [inline]
     poly_setkey+0xf6/0x130 crypto/chacha20poly1305.c:337
     poly_init+0x51/0x60 crypto/chacha20poly1305.c:364
     async_done_continue crypto/chacha20poly1305.c:78 [inline]
     poly_genkey_done+0x15/0x30 crypto/chacha20poly1305.c:369
     cryptd_skcipher_complete+0x29/0x70 crypto/cryptd.c:279
     cryptd_skcipher_decrypt+0xcd/0x110 crypto/cryptd.c:339
     cryptd_queue_worker+0x70/0xa0 crypto/cryptd.c:184
     process_one_work+0x1ed/0x420 kernel/workqueue.c:2269
     worker_thread+0x3e/0x3a0 kernel/workqueue.c:2415
     kthread+0x11f/0x140 kernel/kthread.c:255
     ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:352

Fixes: 71ebc4d1 ("crypto: chacha20poly1305 - Add a ChaCha20-Poly1305 AEAD construction, RFC7539")
Cc: <stable@vger.kernel.org> # v4.2+
Cc: Martin Willi <martin@strongswan.org>
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Commit c778f96b ("crypto: lrw - Optimize tweak computation")
incorrectly reduced the alignmask of LRW instances from
'__alignof__(u64) - 1' to '__alignof__(__be32) - 1'.

However, xor_tweak() and setkey() assume that the data and key,
respectively, are aligned to 'be128', which has u64 alignment.

Fix the alignmask to be at least '__alignof__(be128) - 1'.

Fixes: c778f96b ("crypto: lrw - Optimize tweak computation")
Cc: <stable@vger.kernel.org> # v4.20+
Cc: Ondrej Mosnacek <omosnace@redhat.com>
Signed-off-by: NEric Biggers <ebiggers@google.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>

Changing ghash_mod_init() to be subsys_initcall made it start running
before the alignment fault handler has been installed on ARM.  In kernel
builds where the keys in the ghash test vectors happened to be
misaligned in the kernel image, this exposed the longstanding bug that
ghash_setkey() is incorrectly casting the key buffer (which can have any
alignment) to be128 for passing to gf128mul_init_4k_lle().

Fix this by memcpy()ing the key to a temporary buffer.

Don't fix it by setting an alignmask on the algorithm instead because
that would unnecessarily force alignment of the data too.

Fixes: 2cdc6899 ("crypto: ghash - Add GHASH digest algorithm for GCM")
Reported-by: NPeter Robinson <pbrobinson@gmail.com>
Cc: stable@vger.kernel.org
Signed-off-by: NEric Biggers <ebiggers@google.com>
Tested-by: NPeter Robinson <pbrobinson@gmail.com>
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>