openwrt/staging/blogic.git
7 years agocrypto: cavium - switch to pci_alloc_irq_vectors
Christoph Hellwig [Wed, 15 Feb 2017 07:18:43 +0000 (08:18 +0100)]
crypto: cavium - switch to pci_alloc_irq_vectors

pci_enable_msix has been long deprecated, but this driver adds a new
instance.  Convert it to pci_alloc_irq_vectors and greatly simplify
the code, and make sure the prope code properly unwinds.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - switch to pci_alloc_irq_vectors
Christoph Hellwig [Wed, 15 Feb 2017 07:18:42 +0000 (08:18 +0100)]
crypto: cavium - switch to pci_alloc_irq_vectors

pci_enable_msix has been long deprecated, but this driver adds a new
instance.  Convert it to pci_alloc_irq_vectors and greatly simplify
the code.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - remove dead MSI-X related define
Christoph Hellwig [Wed, 15 Feb 2017 07:18:41 +0000 (08:18 +0100)]
crypto: cavium - remove dead MSI-X related define

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: brcm - Avoid double free in ahash_finup()
Rob Rice [Tue, 14 Feb 2017 17:45:52 +0000 (12:45 -0500)]
crypto: brcm - Avoid double free in ahash_finup()

In Broadcom SPU driver, in case where incremental hash
is done in software in ahash_finup(), tmpbuf was freed
twice.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Rob Rice <rob.rice@broadcom.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - fix Kconfig dependencies
Arnd Bergmann [Tue, 14 Feb 2017 17:07:31 +0000 (18:07 +0100)]
crypto: cavium - fix Kconfig dependencies

The driver fails to build if MSI support is disabled:

In file included from /git/arm-soc/drivers/crypto/cavium/cpt/cptpf_main.c:18:0:
drivers/crypto/cavium/cpt/cptpf.h:57:20: error: array type has incomplete element type 'struct msix_entry'
  struct msix_entry msix_entries[CPT_PF_MSIX_VECTORS];
                    ^~~~~~~~~~~~
drivers/crypto/cavium/cpt/cptpf_main.c: In function 'cpt_enable_msix':
drivers/crypto/cavium/cpt/cptpf_main.c:344:8: error: implicit declaration of function 'pci_enable_msix';did you mean 'cpt_enable_msix'? [-Werror=implicit-function-declaration]

On the other hand, it doesn't seem to have any build dependency on ARCH_THUNDER,
so let's allow compile-testing to catch this kind of problem more easily.
The 64-bit dependency is needed for the use of readq/writeq.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: David Daney <david.daney@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - cpt_bind_vq_to_grp could return an error code
George Cherian [Tue, 14 Feb 2017 09:23:17 +0000 (09:23 +0000)]
crypto: cavium - cpt_bind_vq_to_grp could return an error code

cpt_bind_vq_to_grp() could return an error code. However, it currently
returns a u8. This produce the static checker warning.

drivers/crypto/cavium/cpt/cptpf_mbox.c:70 cpt_bind_vq_to_grp() warn: signedness bug returning '(-22)'

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: George Cherian <george.cherian@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: doc - fix typo
Gilad Ben-Yossef [Tue, 14 Feb 2017 06:21:45 +0000 (08:21 +0200)]
crypto: doc - fix typo

Fix a single letter typo in api-skcipher.rst.

Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: omap - update Kconfig help description
Russell King [Mon, 13 Feb 2017 12:04:08 +0000 (12:04 +0000)]
hwrng: omap - update Kconfig help description

omap-rng also supports Marvell Armada 7k/8k SoCs, but no mention of this
is made in the help text, despite the dependency being added. Explicitly
mention these SoCs in the help description so people know that it covers
more than just TI SoCs.

Fixes: 383212425c92 ("hwrng: omap - Add device variant for SafeXcel IP-76 found in Armada 8K")
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccm - drop unnecessary minimum 32-bit alignment
Ard Biesheuvel [Sat, 11 Feb 2017 19:25:22 +0000 (19:25 +0000)]
crypto: ccm - drop unnecessary minimum 32-bit alignment

The CCM driver forces 32-bit alignment even if the underlying ciphers
don't care about alignment. This is because crypto_xor() used to require
this, but since this is no longer the case, drop the hardcoded minimum
of 32 bits.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccm - honour alignmask of subordinate MAC cipher
Ard Biesheuvel [Sat, 11 Feb 2017 19:25:21 +0000 (19:25 +0000)]
crypto: ccm - honour alignmask of subordinate MAC cipher

The CCM driver was recently updated to defer the MAC part of the algorithm
to a dedicated crypto transform, and a template for instantiating such
transforms was added at the same time.

However, this new cbcmac template fails to take the alignmask of the
encapsulated cipher into account, which may result in buffer addresses
being passed down that are not sufficiently aligned.

So update the code to ensure that the digest buffer in the desc ctx
appears at a sufficiently aligned offset, and tweak the code so that all
calls to crypto_cipher_encrypt_one() operate on this buffer exclusively.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix state buffer DMA (un)mapping
Horia Geantă [Fri, 10 Feb 2017 12:07:25 +0000 (14:07 +0200)]
crypto: caam - fix state buffer DMA (un)mapping

If we register the DMA API debug notification chain to
receive platform bus events:
    dma_debug_add_bus(&platform_bus_type);
we start receiving warnings after a simple test like "modprobe caam_jr &&
modprobe caamhash && modprobe -r caamhash && modprobe -r caam_jr":
platform ffe301000.jr: DMA-API: device driver has pending DMA allocations while released from device [count=1938]
One of leaked entries details: [device address=0x0000000173fda090] [size=63 bytes] [mapped with DMA_TO_DEVICE] [mapped as single]

It turns out there are several issues with handling buf_dma (mapping of buffer
holding the previous chunk smaller than hash block size):
-detection of buf_dma mapping failure occurs too late, after a job descriptor
using that value has been submitted for execution
-dma mapping leak - unmapping is not performed in all places: for e.g.
in ahash_export or in most ahash_fin* callbacks (due to current back-to-back
implementation of buf_dma unmapping/mapping)

Fix these by:
-calling dma_mapping_error() on buf_dma right after the mapping and providing
an error code if needed
-unmapping buf_dma during the "job done" (ahash_done_*) callbacks

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - abstract ahash request double buffering
Horia Geantă [Fri, 10 Feb 2017 12:07:24 +0000 (14:07 +0200)]
crypto: caam - abstract ahash request double buffering

caamhash uses double buffering for holding previous/current
and next chunks (data smaller than block size) to be hashed.

Add (inline) functions to abstract this mechanism.

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix error path for ctx_dma mapping failure
Horia Geantă [Fri, 10 Feb 2017 12:07:23 +0000 (14:07 +0200)]
crypto: caam - fix error path for ctx_dma mapping failure

In case ctx_dma dma mapping fails, ahash_unmap_ctx() tries to
dma unmap an invalid address:
map_seq_out_ptr_ctx() / ctx_map_to_sec4_sg() -> goto unmap_ctx ->
-> ahash_unmap_ctx() -> dma unmap ctx_dma

There is also possible to reach ahash_unmap_ctx() with ctx_dma
uninitialzed or to try to unmap the same address twice.

Fix these by setting ctx_dma = 0 where needed:
-initialize ctx_dma in ahash_init()
-clear ctx_dma in case of mapping error (instead of holding
the error code returned by the dma map function)
-clear ctx_dma after each unmapping

Fixes: 32686d34f8fb6 ("crypto: caam - ensure that we clean up after an error")
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix DMA API leaks for multiple setkey() calls
Horia Geantă [Fri, 10 Feb 2017 12:07:22 +0000 (14:07 +0200)]
crypto: caam - fix DMA API leaks for multiple setkey() calls

setkey() callback may be invoked multiple times for the same tfm.
In this case, DMA API leaks are caused by shared descriptors
(and key for caamalg) being mapped several times and unmapped only once.
Fix this by performing mapping / unmapping only in crypto algorithm's
cra_init() / cra_exit() callbacks and sync_for_device in the setkey()
tfm callback.

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - don't dma_map key for hash algorithms
Horia Geantă [Fri, 10 Feb 2017 12:07:21 +0000 (14:07 +0200)]
crypto: caam - don't dma_map key for hash algorithms

Shared descriptors for hash algorithms are small enough
for (split) keys to be inlined in all cases.
Since driver already does this, all what's left is to remove
unused ctx->key_dma.

Fixes: 045e36780f115 ("crypto: caam - ahash hmac support")
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - use dma_map_sg() return code
Horia Geantă [Fri, 10 Feb 2017 12:07:20 +0000 (14:07 +0200)]
crypto: caam - use dma_map_sg() return code

dma_map_sg() might coalesce S/G entries, so use the number of S/G
entries returned by it instead of what sg_nents_for_len() initially
returns.

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - replace sg_count() with sg_nents_for_len()
Horia Geantă [Fri, 10 Feb 2017 12:07:19 +0000 (14:07 +0200)]
crypto: caam - replace sg_count() with sg_nents_for_len()

Replace internal sg_count() function and the convoluted logic
around it with the standard sg_nents_for_len() function.
src_nents, dst_nents now hold the number of SW S/G entries,
instead of the HW S/G table entries.

With this change, null (zero length) input data for AEAD case
needs to be handled in a visible way. req->src is no longer
(un)mapped, pointer address is set to 0 in SEQ IN PTR command.

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - check sg_count() return value
Horia Geantă [Fri, 10 Feb 2017 12:07:18 +0000 (14:07 +0200)]
crypto: caam - check sg_count() return value

sg_count() internally calls sg_nents_for_len(), which could fail
in case the required number of bytes is larger than the total
bytes in the S/G.

Thus, add checks to validate the input.

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix HW S/G in ablkcipher_giv_edesc_alloc()
Horia Geantă [Fri, 10 Feb 2017 12:07:17 +0000 (14:07 +0200)]
crypto: caam - fix HW S/G in ablkcipher_giv_edesc_alloc()

HW S/G generation does not work properly when the following conditions
are met:
-src == dst
-src/dst is S/G
-IV is right before (contiguous with) the first src/dst S/G entry
since "iv_contig" is set to true (iv_contig is a misnomer here and
it actually refers to the whole output being contiguous)

Fix this by setting dst S/G nents equal to src S/G nents, instead of
leaving it set to init value (0).

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix JR IO mapping if one fails
Tudor Ambarus [Fri, 10 Feb 2017 12:07:16 +0000 (14:07 +0200)]
crypto: caam - fix JR IO mapping if one fails

If one of the JRs failed at init, the next JR used
the failed JR's IO space. The patch fixes this bug.

Signed-off-by: Tudor Ambarus <tudor-dan.ambarus@nxp.com>
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - check return code of dma_set_mask_and_coherent()
Horia Geantă [Fri, 10 Feb 2017 12:07:15 +0000 (14:07 +0200)]
crypto: caam - check return code of dma_set_mask_and_coherent()

Setting the dma mask could fail, thus make sure it succeeds
before going further.

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - don't include unneeded headers
Horia Geantă [Fri, 10 Feb 2017 12:07:14 +0000 (14:07 +0200)]
crypto: caam - don't include unneeded headers

intern.h, jr.h are not needed in error.c
error.h is not needed in ctrl.c

Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Simplify some buffer management routines
Gary R Hook [Thu, 9 Feb 2017 21:50:08 +0000 (15:50 -0600)]
crypto: ccp - Simplify some buffer management routines

The reverse-get/set functions can be simplified by
eliminating unused code.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Update the command queue on errors
Gary R Hook [Thu, 9 Feb 2017 21:49:57 +0000 (15:49 -0600)]
crypto: ccp - Update the command queue on errors

Move the command queue tail pointer when an error is
detected. Always return the error.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Change mode for detailed CCP init messages
Gary R Hook [Thu, 9 Feb 2017 21:49:48 +0000 (15:49 -0600)]
crypto: ccp - Change mode for detailed CCP init messages

The CCP initialization messages only need to be sent to
syslog in debug mode.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - fix error management in atmel_sha_start()
Cyrille Pitchen [Thu, 9 Feb 2017 16:51:21 +0000 (17:51 +0100)]
crypto: atmel-sha - fix error management in atmel_sha_start()

This patch clarifies and fixes how errors should be handled by
atmel_sha_start().

For update operations, the previous code wrongly assumed that
(err != -EINPROGRESS) implies (err == 0). It's wrong because that doesn't
take the error cases (err < 0) into account.

This patch also adds many comments to detail all the possible returned
values and what should be done in each case.

Especially, when an error occurs, since atmel_sha_complete() has already
been called, hence releasing the hardware, atmel_sha_start() must not call
atmel_sha_finish_req() later otherwise atmel_sha_complete() would be
called a second time.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - fix missing "return" instructions
Cyrille Pitchen [Thu, 9 Feb 2017 16:51:20 +0000 (17:51 +0100)]
crypto: atmel-sha - fix missing "return" instructions

This patch fixes a previous patch: "crypto: atmel-sha - update request
queue management to make it more generic".

Indeed the patch above should have replaced the "return -EINVAL;" lines by
"return atmel_sha_complete(dd, -EINVAL);" but instead replaced them by a
simple call of "atmel_sha_complete(dd, -EINVAL);".
Hence all "return" instructions were missing.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Set the AES size field for all modes
Gary R Hook [Wed, 8 Feb 2017 19:07:06 +0000 (13:07 -0600)]
crypto: ccp - Set the AES size field for all modes

Ensure that the size field is correctly populated for
all AES modes.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: brcm - Add Broadcom SPU driver
Rob Rice [Fri, 3 Feb 2017 17:55:33 +0000 (12:55 -0500)]
crypto: brcm - Add Broadcom SPU driver

Add Broadcom Secure Processing Unit (SPU) crypto driver for SPU
hardware crypto offload. The driver supports ablkcipher, ahash,
and aead symmetric crypto operations.

Signed-off-by: Steve Lin <steven.lin1@broadcom.com>
Signed-off-by: Rob Rice <rob.rice@broadcom.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: brcm - DT documentation for Broadcom SPU hardware
Rob Rice [Fri, 3 Feb 2017 17:55:32 +0000 (12:55 -0500)]
crypto: brcm - DT documentation for Broadcom SPU hardware

Device tree documentation for Broadcom Secure Processing Unit
(SPU) crypto hardware.

Signed-off-by: Steve Lin <steven.lin1@broadcom.com>
Signed-off-by: Rob Rice <rob.rice@broadcom.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - Enable CPT options crypto for build
George Cherian [Tue, 7 Feb 2017 14:51:15 +0000 (14:51 +0000)]
crypto: cavium - Enable CPT options crypto for build

Add the CPT options in crypto Kconfig and update the
crypto Makefile

Update the MAINTAINERS file too.

Signed-off-by: George Cherian <george.cherian@cavium.com>
Reviewed-by: David Daney <david.daney@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - Add the Virtual Function driver for CPT
George Cherian [Tue, 7 Feb 2017 14:51:14 +0000 (14:51 +0000)]
crypto: cavium - Add the Virtual Function driver for CPT

Enable the CPT VF driver. CPT is the cryptographic Acceleration Unit
in Octeon-tx series of processors.

Signed-off-by: George Cherian <george.cherian@cavium.com>
Reviewed-by: David Daney <david.daney@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: cavium - Add Support for Octeon-tx CPT Engine
George Cherian [Tue, 7 Feb 2017 14:51:13 +0000 (14:51 +0000)]
crypto: cavium - Add Support for Octeon-tx CPT Engine

Enable the Physical Function driver for the Cavium Crypto Engine (CPT)
found in Octeon-tx series of SoC's. CPT is the Cryptographic Accelaration
Unit. CPT includes microcoded GigaCypher symmetric engines (SEs) and
asymmetric engines (AEs).

Signed-off-by: George Cherian <george.cherian@cavium.com>
Reviewed-by: David Daney <david.daney@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: cavium - Use per device name to allow for multiple devices.
David Daney [Mon, 6 Feb 2017 22:28:46 +0000 (14:28 -0800)]
hwrng: cavium - Use per device name to allow for multiple devices.

Systems containing the Cavium HW RNG may have one device per NUMA
node.  A typical configuration is a 2-node NUMA system, which results
in 2 RNG devices.  The hwrng subsystem refuses (and rightly so) to
register more than one device with he same name, so we get failure
messages on these systems.

Make the hwrng name unique by including the underlying device name.
Also remove spaces from the name to make it possible to switch devices
via the sysfs knobs.

Signed-off-by: David Daney <david.daney@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel - fix 64-bit build warnings
Arnd Bergmann [Mon, 6 Feb 2017 12:32:16 +0000 (13:32 +0100)]
crypto: atmel - fix 64-bit build warnings

When we enable COMPILE_TEST building for the Atmel sha and tdes implementations,
we run into a couple of warnings about incorrect format strings, e.g.

In file included from include/linux/platform_device.h:14:0,
                 from drivers/crypto/atmel-sha.c:24:
drivers/crypto/atmel-sha.c: In function 'atmel_sha_xmit_cpu':
drivers/crypto/atmel-sha.c:571:19: error: format '%d' expects argument of type 'int', but argument 6 has type 'size_t {aka long unsigned int}' [-Werror=format=]
In file included from include/linux/printk.h:6:0,
                 from include/linux/kernel.h:13,
                 from drivers/crypto/atmel-tdes.c:17:
drivers/crypto/atmel-tdes.c: In function 'atmel_tdes_crypt_dma_stop':
include/linux/kern_levels.h:4:18: error: format '%u' expects argument of type 'unsigned int', but argument 2 has type 'size_t {aka long unsigned int}' [-Werror=format=]

These are all fixed by using the "%z" modifier for size_t data.

There are also a few uses of min()/max() with incompatible types:

drivers/crypto/atmel-tdes.c: In function 'atmel_tdes_crypt_start':
drivers/crypto/atmel-tdes.c:528:181: error: comparison of distinct pointer types lacks a cast [-Werror]

Where possible, we should use consistent types here, otherwise we can use
min_t()/max_t() to get well-defined behavior without a warning.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel - refine Kconfig dependencies
Arnd Bergmann [Mon, 6 Feb 2017 12:32:15 +0000 (13:32 +0100)]
crypto: atmel - refine Kconfig dependencies

With the new authenc support, we get a harmless Kconfig warning:

warning: (CRYPTO_DEV_ATMEL_AUTHENC) selects CRYPTO_DEV_ATMEL_SHA which has unmet direct dependencies (CRYPTO && CRYPTO_HW && ARCH_AT91)

The problem is that each of the options has slightly different dependencies,
although they all seem to want the same thing: allow building for real AT91
targets that actually have the hardware, and possibly for compile testing.

This makes all four options consistent: instead of depending on a particular
dmaengine implementation, we depend on the ARM platform, CONFIG_COMPILE_TEST
as an alternative when that is turned off. This makes the 'select' statements
work correctly.

Fixes: 89a82ef87e01 ("crypto: atmel-authenc - add support to authenc(hmac(shaX), Y(aes)) modes")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: algapi - make crypto_xor() and crypto_inc() alignment agnostic
Ard Biesheuvel [Sun, 5 Feb 2017 10:06:12 +0000 (10:06 +0000)]
crypto: algapi - make crypto_xor() and crypto_inc() alignment agnostic

Instead of unconditionally forcing 4 byte alignment for all generic
chaining modes that rely on crypto_xor() or crypto_inc() (which may
result in unnecessary copying of data when the underlying hardware
can perform unaligned accesses efficiently), make those functions
deal with unaligned input explicitly, but only if the Kconfig symbol
HAVE_EFFICIENT_UNALIGNED_ACCESS is set. This will allow us to drop
the alignmasks from the CBC, CMAC, CTR, CTS, PCBC and SEQIV drivers.

For crypto_inc(), this simply involves making the 4-byte stride
conditional on HAVE_EFFICIENT_UNALIGNED_ACCESS being set, given that
it typically operates on 16 byte buffers.

For crypto_xor(), an algorithm is implemented that simply runs through
the input using the largest strides possible if unaligned accesses are
allowed. If they are not, an optimal sequence of memory accesses is
emitted that takes the relative alignment of the input buffers into
account, e.g., if the relative misalignment of dst and src is 4 bytes,
the entire xor operation will be completed using 4 byte loads and stores
(modulo unaligned bits at the start and end). Note that all expressions
involving misalign are simply eliminated by the compiler when
HAVE_EFFICIENT_UNALIGNED_ACCESS is defined.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: improve gcc optimization flags for serpent and wp512
Arnd Bergmann [Fri, 3 Feb 2017 22:33:23 +0000 (23:33 +0100)]
crypto: improve gcc optimization flags for serpent and wp512

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>
7 years agocrypto: arm64/aes - add NEON/Crypto Extensions CBCMAC/CMAC/XCBC driver
Ard Biesheuvel [Fri, 3 Feb 2017 14:49:37 +0000 (14:49 +0000)]
crypto: arm64/aes - add NEON/Crypto Extensions CBCMAC/CMAC/XCBC driver

On ARMv8 implementations that do not support the Crypto Extensions,
such as the Raspberry Pi 3, the CCM driver falls back to the generic
table based AES implementation to perform the MAC part of the
algorithm, which is slow and not time invariant. So add a CBCMAC
implementation to the shared glue code between NEON AES and Crypto
Extensions AES, so that it can be used instead now that the CCM
driver has been updated to look for CBCMAC implementations other
than the one it supplies itself.

Also, given how these algorithms mostly only differ in the way the key
handling and the final encryption are implemented, expose CMAC and XCBC
algorithms as well based on the same core update code.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccm - switch to separate cbcmac driver
Ard Biesheuvel [Fri, 3 Feb 2017 14:49:36 +0000 (14:49 +0000)]
crypto: ccm - switch to separate cbcmac driver

Update the generic CCM driver to defer CBC-MAC processing to a
dedicated CBC-MAC ahash transform rather than open coding this
transform (and much of the associated scatterwalk plumbing) in
the CCM driver itself.

This cleans up the code considerably, but more importantly, it allows
the use of alternative CBC-MAC implementations that don't suffer from
performance degradation due to significant setup time (e.g., the NEON
based AES code needs to enable/disable the NEON, and load the S-box
into 16 SIMD registers, which cannot be amortized over the entire input
when using the cipher interface)

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: testmgr - add test cases for cbcmac(aes)
Ard Biesheuvel [Fri, 3 Feb 2017 14:49:35 +0000 (14:49 +0000)]
crypto: testmgr - add test cases for cbcmac(aes)

In preparation of splitting off the CBC-MAC transform in the CCM
driver into a separate algorithm, define some test cases for the
AES incarnation of cbcmac.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: aes - add generic time invariant AES cipher
Ard Biesheuvel [Thu, 2 Feb 2017 16:37:40 +0000 (16:37 +0000)]
crypto: aes - add generic time invariant AES cipher

Lookup table based AES is sensitive to timing attacks, which is due to
the fact that such table lookups are data dependent, and the fact that
8 KB worth of tables covers a significant number of cachelines on any
architecture, resulting in an exploitable correlation between the key
and the processing time for known plaintexts.

For network facing algorithms such as CTR, CCM or GCM, this presents a
security risk, which is why arch specific AES ports are typically time
invariant, either through the use of special instructions, or by using
SIMD algorithms that don't rely on table lookups.

For generic code, this is difficult to achieve without losing too much
performance, but we can improve the situation significantly by switching
to an implementation that only needs 256 bytes of table data (the actual
S-box itself), which can be prefetched at the start of each block to
eliminate data dependent latencies.

This code encrypts at ~25 cycles per byte on ARM Cortex-A57 (while the
ordinary generic AES driver manages 18 cycles per byte on this
hardware). Decryption is substantially slower.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: aes-generic - drop alignment requirement
Ard Biesheuvel [Thu, 2 Feb 2017 15:58:57 +0000 (15:58 +0000)]
crypto: aes-generic - drop alignment requirement

The generic AES code exposes a 32-bit align mask, which forces all
users of the code to use temporary buffers or take other measures to
ensure the alignment requirement is adhered to, even on architectures
that don't care about alignment for software algorithms such as this
one.

So drop the align mask, and fix the code to use get_unaligned_le32()
where appropriate, which will resolve to whatever is optimal for the
architecture.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sha512-mb - Protect sha512 mb ctx mgr access
Tim Chen [Wed, 1 Feb 2017 18:45:02 +0000 (10:45 -0800)]
crypto: sha512-mb - Protect sha512 mb ctx mgr access

The flusher and regular multi-buffer computation via mcryptd may race with another.
Add here a lock and turn off interrupt to to access multi-buffer
computation state cstate->mgr before a round of computation. This should
prevent the flusher code jumping in.

Signed-off-by: Tim Chen <tim.c.chen@linux.intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/crc32 - merge CRC32 and PMULL instruction based drivers
Ard Biesheuvel [Wed, 1 Feb 2017 15:35:40 +0000 (15:35 +0000)]
crypto: arm64/crc32 - merge CRC32 and PMULL instruction based drivers

The PMULL based CRC32 implementation already contains code based on the
separate, optional CRC32 instructions to fallback to when operating on
small quantities of data. We can expose these routines directly on systems
that lack the 64x64 PMULL instructions but do implement the CRC32 ones,
which makes the driver that is based solely on those CRC32 instructions
redundant. So remove it.

Note that this aligns arm64 with ARM, whose accelerated CRC32 driver
also combines the CRC32 extension based and the PMULL based versions.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Tested-by: Matthias Brugger <mbrugger@suse.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm/aes - don't use IV buffer to return final keystream block
Ard Biesheuvel [Thu, 2 Feb 2017 11:38:56 +0000 (11:38 +0000)]
crypto: arm/aes - don't use IV buffer to return final keystream block

The ARM bit sliced AES core code uses the IV buffer to pass the final
keystream block back to the glue code if the input is not a multiple of
the block size, so that the asm code does not have to deal with anything
except 16 byte blocks. This is done under the assumption that the outgoing
IV is meaningless anyway in this case, given that chaining is no longer
possible under these circumstances.

However, as it turns out, the CCM driver does expect the IV to retain
a value that is equal to the original IV except for the counter value,
and even interprets byte zero as a length indicator, which may result
in memory corruption if the IV is overwritten with something else.

So use a separate buffer to return the final keystream block.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes - don't use IV buffer to return final keystream block
Ard Biesheuvel [Thu, 2 Feb 2017 11:38:55 +0000 (11:38 +0000)]
crypto: arm64/aes - don't use IV buffer to return final keystream block

The arm64 bit sliced AES core code uses the IV buffer to pass the final
keystream block back to the glue code if the input is not a multiple of
the block size, so that the asm code does not have to deal with anything
except 16 byte blocks. This is done under the assumption that the outgoing
IV is meaningless anyway in this case, given that chaining is no longer
possible under these circumstances.

However, as it turns out, the CCM driver does expect the IV to retain
a value that is equal to the original IV except for the counter value,
and even interprets byte zero as a length indicator, which may result
in memory corruption if the IV is overwritten with something else.

So use a separate buffer to return the final keystream block.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes - replace scalar fallback with plain NEON fallback
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:39 +0000 (23:25 +0000)]
crypto: arm64/aes - replace scalar fallback with plain NEON fallback

The new bitsliced NEON implementation of AES uses a fallback in two
places: CBC encryption (which is strictly sequential, whereas this
driver can only operate efficiently on 8 blocks at a time), and the
XTS tweak generation, which involves encrypting a single AES block
with a different key schedule.

The plain (i.e., non-bitsliced) NEON code is more suitable as a fallback,
given that it is faster than scalar on low end cores (which is what
the NEON implementations target, since high end cores have dedicated
instructions for AES), and shows similar behavior in terms of D-cache
footprint and sensitivity to cache timing attacks. So switch the fallback
handling to the plain NEON driver.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes-neon-blk - tweak performance for low end cores
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:38 +0000 (23:25 +0000)]
crypto: arm64/aes-neon-blk - tweak performance for low end cores

The non-bitsliced AES implementation using the NEON is highly sensitive
to micro-architectural details, and, as it turns out, the Cortex-A53 on
the Raspberry Pi 3 is a core that can benefit from this code, given that
its scalar AES performance is abysmal (32.9 cycles per byte).

The new bitsliced AES code manages 19.8 cycles per byte on this core,
but can only operate on 8 blocks at a time, which is not supported by
all chaining modes. With a bit of tweaking, we can get the plain NEON
code to run at 22.0 cycles per byte, making it useful for sequential
modes like CBC encryption. (Like bitsliced NEON, the plain NEON
implementation does not use any lookup tables, which makes it easy on
the D-cache, and invulnerable to cache timing attacks)

So tweak the plain NEON AES code to use tbl instructions rather than
shl/sri pairs, and to avoid the need to reload permutation vectors or
other constants from memory in every round. Also, improve the decryption
performance by switching to 16x8 pmul instructions for the performing
the multiplications in GF(2^8).

To allow the ECB and CBC encrypt routines to be reused by the bitsliced
NEON code in a subsequent patch, export them from the module.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes - performance tweak
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:37 +0000 (23:25 +0000)]
crypto: arm64/aes - performance tweak

Shuffle some instructions around in the __hround macro to shave off
0.1 cycles per byte on Cortex-A57.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes - avoid literals for cross-module symbol references
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:36 +0000 (23:25 +0000)]
crypto: arm64/aes - avoid literals for cross-module symbol references

Using simple adrp/add pairs to refer to the AES lookup tables exposed by
the generic AES driver (which could be loaded far away from this driver
when KASLR is in effect) was unreliable at module load time before commit
41c066f2c4d4 ("arm64: assembler: make adr_l work in modules under KASLR"),
which is why the AES code used literals instead.

So now we can get rid of the literals, and switch to the adr_l macro.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/chacha20 - remove cra_alignmask
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:35 +0000 (23:25 +0000)]
crypto: arm64/chacha20 - remove cra_alignmask

Remove the unnecessary alignmask: it is much more efficient to deal with
the misalignment in the core algorithm than relying on the crypto API to
copy the data to a suitably aligned buffer.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes-blk - remove cra_alignmask
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:34 +0000 (23:25 +0000)]
crypto: arm64/aes-blk - remove cra_alignmask

Remove the unnecessary alignmask: it is much more efficient to deal with
the misalignment in the core algorithm than relying on the crypto API to
copy the data to a suitably aligned buffer.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm64/aes-ce-ccm - remove cra_alignmask
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:33 +0000 (23:25 +0000)]
crypto: arm64/aes-ce-ccm - remove cra_alignmask

Remove the unnecessary alignmask: it is much more efficient to deal with
the misalignment in the core algorithm than relying on the crypto API to
copy the data to a suitably aligned buffer.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm/chacha20 - remove cra_alignmask
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:32 +0000 (23:25 +0000)]
crypto: arm/chacha20 - remove cra_alignmask

Remove the unnecessary alignmask: it is much more efficient to deal with
the misalignment in the core algorithm than relying on the crypto API to
copy the data to a suitably aligned buffer.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm/aes-ce - remove cra_alignmask
Ard Biesheuvel [Sat, 28 Jan 2017 23:25:31 +0000 (23:25 +0000)]
crypto: arm/aes-ce - remove cra_alignmask

Remove the unnecessary alignmask: it is much more efficient to deal with
the misalignment in the core algorithm than relying on the crypto API to
copy the data to a suitably aligned buffer.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Fix Smatch Complaint
Harsh Jain [Fri, 27 Jan 2017 10:39:12 +0000 (16:09 +0530)]
crypto: chcr - Fix Smatch Complaint

Initialise variable after null check.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Harsh Jain <harsh@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Fix wrong typecasting
Harsh Jain [Fri, 27 Jan 2017 10:39:11 +0000 (16:09 +0530)]
crypto: chcr - Fix wrong typecasting

Typecast the pointer with correct structure.

Signed-off-by: Atul Gupta <atul.gupta@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Change algo priority
Harsh Jain [Fri, 27 Jan 2017 10:39:10 +0000 (16:09 +0530)]
crypto: chcr - Change algo priority

Update priorities to 3000

Signed-off-by: Harsh Jain <harsh@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Change cra_flags for cipher algos
Harsh Jain [Fri, 27 Jan 2017 10:39:09 +0000 (16:09 +0530)]
crypto: chcr - Change cra_flags for cipher algos

Change cipher algos flags to CRYPTO_ALG_TYPE_ABLKCIPHER.

Signed-off-by: Harsh Jain <harsh@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Use cipher instead of Block Cipher in gcm setkey
Harsh Jain [Fri, 27 Jan 2017 10:39:08 +0000 (16:09 +0530)]
crypto: chcr - Use cipher instead of Block Cipher in gcm setkey

1 Block of encrption can be done with aes-generic. no need of
cbc(aes). This patch replaces cbc(aes-generic) with aes-generic.

Signed-off-by: Harsh Jain <harsh@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - fix itnull.cocci warnings
Harsh Jain [Fri, 27 Jan 2017 10:39:07 +0000 (16:09 +0530)]
crypto: chcr - fix itnull.cocci warnings

The first argument to list_for_each_entry cannot be NULL.

Generated by: scripts/coccinelle/iterators/itnull.cocci

Signed-off-by: Julia Lawall <julia.lawall@lip6.fr>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Harsh Jain <harsh@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Change flow IDs
Harsh Jain [Fri, 27 Jan 2017 10:39:05 +0000 (16:09 +0530)]
crypto: chcr - Change flow IDs

Change assign flowc id to each outgoing request.Firmware use flowc id
to schedule each request onto HW. FW reply may miss without this change.

Reviewed-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Atul Gupta <atul.gupta@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - add verbose debug facilities to print hw register names
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:57 +0000 (17:07 +0100)]
crypto: atmel-sha - add verbose debug facilities to print hw register names

When VERBOSE_DEBUG is defined and SHA_FLAGS_DUMP_REG flag is set in
dd->flags, this patch prints the register names and values when performing
IO accesses.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-authenc - add support to authenc(hmac(shaX), Y(aes)) modes
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:56 +0000 (17:07 +0100)]
crypto: atmel-authenc - add support to authenc(hmac(shaX), Y(aes)) modes

This patchs allows to combine the AES and SHA hardware accelerators on
some Atmel SoCs. Doing so, AES blocks are only written to/read from the
AES hardware. Those blocks are also transferred from the AES to the SHA
accelerator internally, without additionnal accesses to the system busses.

Hence, the AES and SHA accelerators work in parallel to process all the
data blocks, instead of serializing the process by (de)crypting those
blocks first then authenticating them after like the generic
crypto/authenc.c driver does.

Of course, both the AES and SHA hardware accelerators need to be available
before we can start to process the data blocks. Hence we use their crypto
request queue to synchronize both drivers.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-aes - fix atmel_aes_handle_queue()
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:55 +0000 (17:07 +0100)]
crypto: atmel-aes - fix atmel_aes_handle_queue()

This patch fixes the value returned by atmel_aes_handle_queue(), which
could have been wrong previously when the crypto request was started
synchronously but became asynchronous during the ctx->start() call.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - add support to hmac(shaX)
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:54 +0000 (17:07 +0100)]
crypto: atmel-sha - add support to hmac(shaX)

This patch adds support to the hmac(shaX) algorithms.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - add simple DMA transfers
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:53 +0000 (17:07 +0100)]
crypto: atmel-sha - add simple DMA transfers

This patch adds a simple function to perform data transfer with the DMA
controller.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - add atmel_sha_cpu_start()
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:52 +0000 (17:07 +0100)]
crypto: atmel-sha - add atmel_sha_cpu_start()

This patch adds a simple function to perform data transfer with PIO, hence
handled by the CPU.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - add SHA_MR_MODE_IDATAR0
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:51 +0000 (17:07 +0100)]
crypto: atmel-sha - add SHA_MR_MODE_IDATAR0

This patch defines an alias macro to SHA_MR_MODE_PDC, which is not suited
for DMA usage.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - add atmel_sha_wait_for_data_ready()
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:50 +0000 (17:07 +0100)]
crypto: atmel-sha - add atmel_sha_wait_for_data_ready()

This patch simply defines a helper function to test the 'Data Ready' flag
of the Status Register. It also gives a chance for the crypto request to
be processed synchronously if this 'Data Ready' flag is already set when
polling the Status Register. Indeed, running synchronously avoid the
latency of the 'Data Ready' interrupt.

When the 'Data Ready' flag has not been set yet, we enable the associated
interrupt and resume processing the crypto request asynchronously from the
'done' task just as before.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - redefine SHA_FLAGS_SHA* flags to match SHA_MR_ALGO_SHA*
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:49 +0000 (17:07 +0100)]
crypto: atmel-sha - redefine SHA_FLAGS_SHA* flags to match SHA_MR_ALGO_SHA*

This patch modifies the SHA_FLAGS_SHA* flags: those algo flags are now
organized as values of a single bitfield instead of individual bits.
This allows to reduce the number of bits needed to encode all possible
values. Also the new values match the SHA_MR_ALGO_SHA* values hence
the algorithm bitfield of the SHA_MR register could simply be set with:

mr = (mr & ~SHA_FLAGS_ALGO_MASK) | (ctx->flags & SHA_FLAGS_ALGO_MASK)

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - make atmel_sha_done_task more generic
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:48 +0000 (17:07 +0100)]
crypto: atmel-sha - make atmel_sha_done_task more generic

This patch is a transitional patch. It updates atmel_sha_done_task() to
make it more generic. Indeed, it adds a new .resume() member in the
atmel_sha_dev structure. This hook is called from atmel_sha_done_task()
to resume processing an asynchronous request.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - update request queue management to make it more generic
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:47 +0000 (17:07 +0100)]
crypto: atmel-sha - update request queue management to make it more generic

This patch is a transitional patch. It splits the atmel_sha_handle_queue()
function. Now atmel_sha_handle_queue() only manages the request queue and
calls a new .start() hook from the atmel_sha_ctx structure.
This hook allows to implement different kind of requests still handled by
a single queue.

Also when the req parameter of atmel_sha_handle_queue() refers to the very
same request as the one returned by crypto_dequeue_request(), the queue
management now gives a chance to this crypto request to be handled
synchronously, hence reducing latencies. The .start() hook returns 0 if
the crypto request was handled synchronously and -EINPROGRESS if the
crypto request still need to be handled asynchronously.

Besides, the new .is_async member of the atmel_sha_dev structure helps
tagging this asynchronous state. Indeed, the req->base.complete() callback
should not be called if the crypto request is handled synchronously.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: atmel-sha - create function to get an Atmel SHA device
Cyrille Pitchen [Thu, 26 Jan 2017 16:07:46 +0000 (17:07 +0100)]
crypto: atmel-sha - create function to get an Atmel SHA device

This is a transitional patch: it creates the atmel_sha_find_dev() function,
which will be used in further patches to share the source code responsible
for finding a Atmel SHA device.

Signed-off-by: Cyrille Pitchen <cyrille.pitchen@atmel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: doc - Fix hash export state information
Rabin Vincent [Thu, 26 Jan 2017 15:33:00 +0000 (16:33 +0100)]
crypto: doc - Fix hash export state information

The documentation states that crypto_ahash_reqsize() provides the size
of the state structure used by crypto_ahash_export().  But it's actually
crypto_ahash_statesize() which provides this size.

Signed-off-by: Rabin Vincent <rabinv@axis.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Herbert Xu [Fri, 3 Feb 2017 10:14:10 +0000 (18:14 +0800)]
Merge git://git./linux/kernel/git/herbert/crypto-2.6

Merge the crypto tree to pick up arm64 output IV patch.

7 years agocrypto: chcr - Fix key length for RFC4106
Harsh Jain [Fri, 27 Jan 2017 10:39:06 +0000 (16:09 +0530)]
crypto: chcr - Fix key length for RFC4106

Check keylen before copying salt to avoid wrap around of Integer.

Signed-off-by: Harsh Jain <harsh@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: algif_aead - Fix kernel panic on list_del
Harsh Jain [Wed, 1 Feb 2017 15:40:28 +0000 (21:10 +0530)]
crypto: algif_aead - Fix kernel panic on list_del

Kernel panics when userspace program try to access AEAD interface.
Remove node from Linked List before freeing its memory.

Cc: <stable@vger.kernel.org>
Signed-off-by: Harsh Jain <harsh@chelsio.com>
Reviewed-by: Stephan Müller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: aesni - Fix failure when pcbc module is absent
Herbert Xu [Wed, 1 Feb 2017 14:17:39 +0000 (22:17 +0800)]
crypto: aesni - Fix failure when pcbc module is absent

When aesni is built as a module together with pcbc, the pcbc module
must be present for aesni to load.  However, the pcbc module may not
be present for reasons such as its absence on initramfs.  This patch
allows the aesni to function even if the pcbc module is enabled but
not present.

Reported-by: Arkadiusz Miśkiewicz <arekm@maven.pl>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Fix double add when creating new DMA command
Gary R Hook [Fri, 27 Jan 2017 23:09:04 +0000 (17:09 -0600)]
crypto: ccp - Fix double add when creating new DMA command

Eliminate a double-add by creating a new list to manage
command descriptors when created; move the descriptor to
the pending list when the command is submitted.

Cc: <stable@vger.kernel.org>
Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Fix DMA operations when IOMMU is enabled
Gary R Hook [Fri, 27 Jan 2017 21:28:45 +0000 (15:28 -0600)]
crypto: ccp - Fix DMA operations when IOMMU is enabled

An I/O page fault occurs when the IOMMU is enabled on a
system that supports the v5 CCP.  DMA operations use a
Request ID value that does not match what is expected by
the IOMMU, resulting in the I/O page fault.  Setting the
Request ID value to 0 corrects this issue.

Cc: <stable@vger.kernel.org>
Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Check device is allocated before use
Harsh Jain [Tue, 24 Jan 2017 05:04:33 +0000 (10:34 +0530)]
crypto: chcr - Check device is allocated before use

Ensure dev is allocated for crypto uld context before using the device
for crypto operations.

Cc: <stable@vger.kernel.org>
Signed-off-by: Atul Gupta <atul.gupta@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: chcr - Fix panic on dma_unmap_sg
Harsh Jain [Tue, 24 Jan 2017 05:04:32 +0000 (10:34 +0530)]
crypto: chcr - Fix panic on dma_unmap_sg

Save DMA mapped sg list addresses to request context buffer.

Signed-off-by: Atul Gupta <atul.gupta@chelsio.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: qat - zero esram only for DH85x devices
Giovanni Cabiddu [Thu, 22 Dec 2016 15:00:24 +0000 (15:00 +0000)]
crypto: qat - zero esram only for DH85x devices

Zero embedded ram in DH85x devices. This is not
needed for newer generations as it is done by HW.

Cc: <stable@vger.kernel.org>
Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: qat - fix bar discovery for c62x
Giovanni Cabiddu [Thu, 22 Dec 2016 15:00:12 +0000 (15:00 +0000)]
crypto: qat - fix bar discovery for c62x

Some accelerators of the c62x series have only two bars.
This patch skips BAR0 if the accelerator does not have it.

Cc: <stable@vger.kernel.org>
Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: vmx - disable preemption to enable vsx in aes_ctr.c
Li Zhong [Fri, 20 Jan 2017 08:35:33 +0000 (16:35 +0800)]
crypto: vmx - disable preemption to enable vsx in aes_ctr.c

Some preemptible check warnings were reported from enable_kernel_vsx(). This
patch disables preemption in aes_ctr.c before enabling vsx, and they are now
consistent with other files in the same directory.

Signed-off-by: Li Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - add support to GCM mode
Ryder Lee [Fri, 20 Jan 2017 05:41:15 +0000 (13:41 +0800)]
crypto: mediatek - add support to GCM mode

This patch adds support to the GCM mode.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - add support to CTR mode
Ryder Lee [Fri, 20 Jan 2017 05:41:14 +0000 (13:41 +0800)]
crypto: mediatek - add support to CTR mode

This patch adds support to the CTR mode.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - fix typo and indentation
Ryder Lee [Fri, 20 Jan 2017 05:41:13 +0000 (13:41 +0800)]
crypto: mediatek - fix typo and indentation

Dummy patch to fix typo and indentation.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - regroup functions by usage
Ryder Lee [Fri, 20 Jan 2017 05:41:12 +0000 (13:41 +0800)]
crypto: mediatek - regroup functions by usage

This patch only regroup functions by usage.
This will help to integrate the GCM support patch later by
adjusting some shared code section, such as common code which
will be reused by GCM, AES mode setting, and DMA transfer.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - rework crypto request completion
Ryder Lee [Fri, 20 Jan 2017 05:41:11 +0000 (13:41 +0800)]
crypto: mediatek - rework crypto request completion

This patch introduces a new callback 'resume' in the struct mtk_aes_rec.
This callback is run to resume/complete the processing of the crypto
request when woken up by AES interrupts when DMA completion.

This callback will help implementing the GCM mode support in further
patches.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - make crypto request queue management more generic
Ryder Lee [Fri, 20 Jan 2017 05:41:10 +0000 (13:41 +0800)]
crypto: mediatek - make crypto request queue management more generic

This patch changes mtk_aes_handle_queue() to make it more generic.
The function argument is now a pointer to struct crypto_async_request,
which is the common base of struct ablkcipher_request and
struct aead_request.

Also this patch introduces struct mtk_aes_base_ctx which will be the
common base of all the transformation contexts.

Hence the very same queue will be used to manage both block cipher and
AEAD requests (such as gcm and authenc implemented in further patches).

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - fix incorrect data transfer result
Ryder Lee [Fri, 20 Jan 2017 05:41:09 +0000 (13:41 +0800)]
crypto: mediatek - fix incorrect data transfer result

This patch fixes mtk_aes_xmit() data transfer bug.

The original function uses the same loop and ring->pos
to handle both command and result descriptors. But this
produces incomplete results when src.sg_len != dst.sg_len.

To solve the problem, we splits the descriptors into different
loops and uses cmd_pos and res_pos to record them respectively.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mediatek - move HW control data to transformation context
Ryder Lee [Fri, 20 Jan 2017 05:41:08 +0000 (13:41 +0800)]
crypto: mediatek - move HW control data to transformation context

This patch moves hardware control block members from
mtk_*_rec to transformation context and refines related
definition. This makes operational context to manage its
own control information easily for each DMA transfer.

Signed-off-by: Ryder Lee <ryder.lee@mediatek.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: x86 - make constants readonly, allow linker to merge them
Denys Vlasenko [Thu, 19 Jan 2017 21:33:04 +0000 (22:33 +0100)]
crypto: x86 - make constants readonly, allow linker to merge them

A lot of asm-optimized routines in arch/x86/crypto/ keep its
constants in .data. This is wrong, they should be on .rodata.

Mnay of these constants are the same in different modules.
For example, 128-bit shuffle mask 0x000102030405060708090A0B0C0D0E0F
exists in at least half a dozen places.

There is a way to let linker merge them and use just one copy.
The rules are as follows: mergeable objects of different sizes
should not share sections. You can't put them all in one .rodata
section, they will lose "mergeability".

GCC puts its mergeable constants in ".rodata.cstSIZE" sections,
or ".rodata.cstSIZE.<object_name>" if -fdata-sections is used.
This patch does the same:

.section .rodata.cst16.SHUF_MASK, "aM", @progbits, 16

It is important that all data in such section consists of
16-byte elements, not larger ones, and there are no implicit
use of one element from another.

When this is not the case, use non-mergeable section:

.section .rodata[.VAR_NAME], "a", @progbits

This reduces .data by ~15 kbytes:

    text    data     bss     dec      hex filename
11097415 2705840 2630712 16433967  fac32f vmlinux-prev.o
11112095 2690672 2630712 16433479  fac147 vmlinux.o

Merged objects are visible in System.map:

ffffffff81a28810 r POLY
ffffffff81a28810 r POLY
ffffffff81a28820 r TWOONE
ffffffff81a28820 r TWOONE
ffffffff81a28830 r PSHUFFLE_BYTE_FLIP_MASK <- merged regardless of
ffffffff81a28830 r SHUF_MASK   <------------- the name difference
ffffffff81a28830 r SHUF_MASK
ffffffff81a28830 r SHUF_MASK
..
ffffffff81a28d00 r K512 <- merged three identical 640-byte tables
ffffffff81a28d00 r K512
ffffffff81a28d00 r K512

Use of object names in section name suffixes is not strictly necessary,
but might help if someday link stage will use garbage collection
to eliminate unused sections (ld --gc-sections).

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Josh Poimboeuf <jpoimboe@redhat.com>
CC: Xiaodong Liu <xiaodong.liu@intel.com>
CC: Megha Dey <megha.dey@intel.com>
CC: linux-crypto@vger.kernel.org
CC: x86@kernel.org
CC: linux-kernel@vger.kernel.org
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: x86/crc32c - fix %progbits -> @progbits
Denys Vlasenko [Thu, 19 Jan 2017 21:28:05 +0000 (22:28 +0100)]
crypto: x86/crc32c - fix %progbits -> @progbits

%progbits form is used on ARM (where @ is a comment char).

x86 consistently uses @progbits everywhere else.

Signed-off-by: Denys Vlasenko <dvlasenk@redhat.com>
CC: Herbert Xu <herbert@gondor.apana.org.au>
CC: Josh Poimboeuf <jpoimboe@redhat.com>
CC: Xiaodong Liu <xiaodong.liu@intel.com>
CC: Megha Dey <megha.dey@intel.com>
CC: George Spelvin <linux@horizon.com>
CC: linux-crypto@vger.kernel.org
CC: x86@kernel.org
CC: linux-kernel@vger.kernel.org
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm/aes-neonbs - fix issue with v2.22 and older assembler
Ard Biesheuvel [Thu, 19 Jan 2017 12:23:32 +0000 (12:23 +0000)]
crypto: arm/aes-neonbs - fix issue with v2.22 and older assembler

The GNU assembler for ARM version 2.22 or older fails to infer the
element size from the vmov instructions, and aborts the build in
the following way;

.../aes-neonbs-core.S: Assembler messages:
.../aes-neonbs-core.S:817: Error: bad type for scalar -- `vmov q1h[1],r10'
.../aes-neonbs-core.S:817: Error: bad type for scalar -- `vmov q1h[0],r9'
.../aes-neonbs-core.S:817: Error: bad type for scalar -- `vmov q1l[1],r8'
.../aes-neonbs-core.S:817: Error: bad type for scalar -- `vmov q1l[0],r7'
.../aes-neonbs-core.S:818: Error: bad type for scalar -- `vmov q2h[1],r10'
.../aes-neonbs-core.S:818: Error: bad type for scalar -- `vmov q2h[0],r9'
.../aes-neonbs-core.S:818: Error: bad type for scalar -- `vmov q2l[1],r8'
.../aes-neonbs-core.S:818: Error: bad type for scalar -- `vmov q2l[0],r7'

Fix this by setting the element size explicitly, by replacing vmov with
vmov.32.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: tcrypt - Add debug prints
Rabin Vincent [Wed, 18 Jan 2017 13:54:05 +0000 (14:54 +0100)]
crypto: tcrypt - Add debug prints

tcrypt is very tight-lipped when it succeeds, but a bit more feedback
would be useful when developing or debugging crypto drivers, especially
since even a successful run ends with the module failing to insert. Add
a couple of debug prints, which can be enabled with dynamic debug:

Before:

 # insmod tcrypt.ko mode=10
 insmod: can't insert 'tcrypt.ko': Resource temporarily unavailable

After:

 # insmod tcrypt.ko mode=10 dyndbg
 tcrypt: testing ecb(aes)
 tcrypt: testing cbc(aes)
 tcrypt: testing lrw(aes)
 tcrypt: testing xts(aes)
 tcrypt: testing ctr(aes)
 tcrypt: testing rfc3686(ctr(aes))
 tcrypt: all tests passed
 insmod: can't insert 'tcrypt.ko': Resource temporarily unavailable

Signed-off-by: Rabin Vincent <rabinv@axis.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - use dma_data_direction when calling dma_map_sg
Nicolas Iooss [Sun, 15 Jan 2017 12:37:50 +0000 (13:37 +0100)]
crypto: img-hash - use dma_data_direction when calling dma_map_sg

The fourth argument of dma_map_sg() and dma_unmap_sg() is an item of
dma_data_direction enum. Function img_hash_xmit_dma() wrongly used
DMA_MEM_TO_DEV, which is an item of dma_transfer_direction enum.

Replace DMA_MEM_TO_DEV (which value is 1) with DMA_TO_DEVICE (which
value is fortunately also 1) when calling dma_map_sg() and
dma_unmap_sg().

Signed-off-by: Nicolas Iooss <nicolas.iooss_linux@m4x.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>