cascardo/linux.git
7 years agoPCI: Fix cavium quirk compile failure with PCI_ATS off
Herbert Xu [Mon, 5 Sep 2016 09:12:57 +0000 (17:12 +0800)]
PCI: Fix cavium quirk compile failure with PCI_ATS off

The newly added quirk_cavium_sriov_rnm_link doesn't compile if
PCI_ATS is off.  This patch adds a check for PCI_ATS.

Fixes: 21b5b8eebbae ("PCI: quirk fixup for cavium invalid sriov...")
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: pasemi - Migrate to managed API
PrasannaKumar Muralidharan [Sun, 4 Sep 2016 18:13:08 +0000 (23:43 +0530)]
hwrng: pasemi - Migrate to managed API

Use devm_ioremap and devm_hwrng_register instead of ioremap and
hwrng_register. This removes unregistering and error handling code.

Changes in v2:
Remove hardcoded resource size in ioremap, use resource struct obtained
by calling platform_get_resource.

Removing hardcoded resource size was suggested by LABBE Corentin.

CC: Darren Stevens <darren@stevens-zone.net>
Suggested-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: PrasannaKumar Muralidharan <prasannatsmkumar@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mv_cesa - remove NO_IRQ reference
Arnd Bergmann [Fri, 2 Sep 2016 23:26:40 +0000 (01:26 +0200)]
crypto: mv_cesa - remove NO_IRQ reference

Drivers should not use NO_IRQ, as we are trying to get rid of that.
In this case, the call to irq_of_parse_and_map() is both wrong
(as it returns '0' on failure, not NO_IRQ) and unnecessary
(as platform_get_irq() does the same thing)

This removes the call to irq_of_parse_and_map() and checks for
the error code correctly.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - add missing release in ccp_dmaengine_register
Quentin Lambert [Fri, 2 Sep 2016 09:48:53 +0000 (11:48 +0200)]
crypto: ccp - add missing release in ccp_dmaengine_register

ccp_dmaengine_register used to return with an error code before
releasing all resource. This patch adds a jump to the appropriate label
ensuring that the resources are properly released before returning.

This issue was found with Hector.

Signed-off-by: Quentin Lambert <lambert.quentin@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm/ghash-ce - add missing async import/export
Ard Biesheuvel [Thu, 1 Sep 2016 13:25:42 +0000 (14:25 +0100)]
crypto: arm/ghash-ce - add missing async import/export

Since commit 8996eafdcbad ("crypto: ahash - ensure statesize is non-zero"),
all ahash drivers are required to implement import()/export(), and must have
a non-zero statesize. Fix this for the ARM Crypto Extensions GHASH
implementation.

Fixes: 8996eafdcbad ("crypto: ahash - ensure statesize is non-zero")
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: arm/sha1-neon - add support for building in Thumb2 mode
Ard Biesheuvel [Thu, 1 Sep 2016 13:25:41 +0000 (14:25 +0100)]
crypto: arm/sha1-neon - add support for building in Thumb2 mode

The ARMv7 NEON module is explicitly built in ARM mode, which is not
supported by the Thumb2 kernel. So remove the explicit override, and
leave it up to the build environment to decide whether the core SHA1
routines are assembled as ARM or as Thumb2 code.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: algif_hash - Handle NULL hashes correctly
Herbert Xu [Thu, 1 Sep 2016 09:16:44 +0000 (17:16 +0800)]
crypto: algif_hash - Handle NULL hashes correctly

Right now attempting to read an empty hash simply returns zeroed
bytes, this patch corrects this by calling the digest function
using an empty input.

Reported-by: Russell King - ARM Linux <linux@armlinux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix rfc3686(ctr(aes)) IV load
Catalin Vasile [Wed, 31 Aug 2016 12:57:55 +0000 (15:57 +0300)]
crypto: caam - fix rfc3686(ctr(aes)) IV load

-nonce is being loaded using append_load_imm_u32() instead of
append_load_as_imm() (nonce is a byte array / stream, not a 4-byte
variable)
-counter is not being added in big endian format, as mandatated by
RFC3686 and expected by the crypto engine

Signed-off-by: Catalin Vasile <cata.vasile@nxp.com>
Reviewed-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: engine - permit to enqueue ashash_request
Corentin LABBE [Wed, 31 Aug 2016 12:02:58 +0000 (14:02 +0200)]
crypto: engine - permit to enqueue ashash_request

The current crypto engine allow only ablkcipher_request to be enqueued.
Thus denying any use of it for hardware that also handle hash algo.

This patch modify the API for allowing to enqueue ciphers and hash.

Since omap-aes/omap-des are the only users, this patch also convert them
to the new cryptoengine API.

Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: engine - move crypto engine to its own header
Corentin LABBE [Wed, 31 Aug 2016 12:02:57 +0000 (14:02 +0200)]
crypto: engine - move crypto engine to its own header

This patch move the whole crypto engine API to its own header
crypto/engine.h.

Signed-off-by: Corentin Labbe <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: qat - fix incorrect accelerator mask for C3X devices
Maksim Lukoshkov [Tue, 30 Aug 2016 17:56:00 +0000 (18:56 +0100)]
crypto: qat - fix incorrect accelerator mask for C3X devices

Fix incorrect value of ADF_C3XXX_ACCELERATORS_MASK.

Signed-off-by: Maksim Lukoshkov <maksim.lukoshkov@intel.com>
Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: bcm2835 - handle of_iomap failures
Arvind Yadav [Mon, 29 Aug 2016 17:10:16 +0000 (22:40 +0530)]
hwrng: bcm2835 - handle of_iomap failures

Check return value of of_iomap and handle errors correctly.

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Acked-by: Eric Anholt <eric@anholt.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: FIPS - allow tests to be disabled in FIPS mode
Stephan Mueller [Thu, 25 Aug 2016 13:15:01 +0000 (15:15 +0200)]
crypto: FIPS - allow tests to be disabled in FIPS mode

In FIPS mode, additional restrictions may apply. If these restrictions
are violated, the kernel will panic(). This patch allows test vectors
for symmetric ciphers to be marked as to be skipped in FIPS mode.

Together with the patch, the XTS test vectors where the AES key is
identical to the tweak key is disabled in FIPS mode. This test vector
violates the FIPS requirement that both keys must be different.

Reported-by: Tapas Sarangi <TSarangi@trustwave.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: qat - fix constants table DMA
Maksim Lukoshkov [Mon, 29 Aug 2016 12:28:31 +0000 (13:28 +0100)]
crypto: qat - fix constants table DMA

Copy const_tab array into DMA-able memory (accesible by qat hw).

Signed-off-by: Maksim Lukoshkov <maksim.lukoshkov@intel.com>
Signed-off-by: Giovanni Cabiddu <giovanni.cabiddu@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: tx4939 - Use devm_hwrng_register instead of hwrng_register
PrasannaKumar Muralidharan [Sun, 28 Aug 2016 08:49:12 +0000 (14:19 +0530)]
hwrng: tx4939 - Use devm_hwrng_register instead of hwrng_register

By using devm_hwrng_register instead of hwrng_register the .remove
callback in platform_driver can be removed. This reduces a few lines in
code.

Signed-off-by: PrasannaKumar Muralidharan <prasannatsmkumar@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: Remove check for max less than 4 bytes
PrasannaKumar Muralidharan [Fri, 26 Aug 2016 18:32:04 +0000 (00:02 +0530)]
hwrng: Remove check for max less than 4 bytes

HW RNG core never asks for data less than 4 bytes. The check whether max
is less than 4 bytes is unnecessary. Remove the check.

Signed-off-by: PrasannaKumar Muralidharan <prasannatsmkumar@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: xor - Fix warning when XOR_SELECT_TEMPLATE is unset
Herbert Xu [Fri, 26 Aug 2016 15:19:39 +0000 (23:19 +0800)]
crypto: xor - Fix warning when XOR_SELECT_TEMPLATE is unset

This patch fixes an unused label warning triggered when the macro
XOR_SELECT_TEMPLATE is not set.

Fixes: 39457acda913 ("crypto: xor - skip speed test if the xor...")
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Suggested-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - Convert to new hwrng read() API
Corentin LABBE [Fri, 26 Aug 2016 11:11:36 +0000 (13:11 +0200)]
hwrng: amd - Convert to new hwrng read() API

This patch convert the hwrng interface used by amd768-rng to its new API
by replacing data_read()/data_present() by read().

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - Access hardware via ioread32/iowrite32
Corentin LABBE [Fri, 26 Aug 2016 11:11:35 +0000 (13:11 +0200)]
hwrng: amd - Access hardware via ioread32/iowrite32

Instead of accessing hw directly via pmbase, it's better to
access after ioport_map() via ioread32/iowrite32.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - Replace global variable with private struct
Corentin LABBE [Fri, 26 Aug 2016 11:11:34 +0000 (13:11 +0200)]
hwrng: amd - Replace global variable with private struct

Instead of having two global variable, it's better to use a
private struct. This will permit to remove amd_pdev variable

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - release_region must be called after hwrng_unregister
Corentin LABBE [Fri, 26 Aug 2016 11:11:33 +0000 (13:11 +0200)]
hwrng: amd - release_region must be called after hwrng_unregister

The driver release the memory region before being sure that nobody use
it.

This patch made hwrng_unregister ran before any release was done.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - Remove asm/io.h
Corentin LABBE [Fri, 26 Aug 2016 11:11:32 +0000 (13:11 +0200)]
hwrng: amd - Remove asm/io.h

checkpatch complains about <asm/io.h> used instead of linux/io.h.
In fact it is not needed.
This patch remove it, and in the process, alphabetize the other headers.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - Be consitent with the driver name
Corentin LABBE [Fri, 26 Aug 2016 11:11:31 +0000 (13:11 +0200)]
hwrng: amd - Be consitent with the driver name

The driver name is displayed each time differently.
This patch make use of the same name everywhere.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - use the BIT macro
Corentin LABBE [Fri, 26 Aug 2016 11:11:30 +0000 (13:11 +0200)]
hwrng: amd - use the BIT macro

This patch add usage of the BIT() macro

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: amd - Fix style problem with blank line
Corentin LABBE [Fri, 26 Aug 2016 11:11:29 +0000 (13:11 +0200)]
hwrng: amd - Fix style problem with blank line

Some blank line are unncessary, and one is missing after declaration.
This patch fix thoses style problems.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - add missing header dependencies
Baoyou Xie [Fri, 26 Aug 2016 09:56:24 +0000 (17:56 +0800)]
crypto: caam - add missing header dependencies

We get 1 warning when biuld kernel with W=1:
drivers/crypto/caam/ctrl.c:398:5: warning: no previous prototype for 'caam_get_era' [-Wmissing-prototypes]

In fact, this function is declared in drivers/crypto/caam/ctrl.h,
so this patch add missing header dependencies.

Signed-off-by: Baoyou Xie <baoyou.xie@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: cavium - Add Cavium HWRNG driver for ThunderX SoC.
Omer Khaliq [Tue, 23 Aug 2016 23:27:15 +0000 (16:27 -0700)]
hwrng: cavium - Add Cavium HWRNG driver for ThunderX SoC.

The Cavium ThunderX SoC has a hardware random number generator.
This driver provides support using the HWRNG framework.

Signed-off-by: Omer Khaliq <okhaliq@caviumnetworks.com>
Signed-off-by: Ananth Jasty <Ananth.Jasty@cavium.com>
Acked-by: David Daney <david.daney@cavium.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agoPCI: quirk fixup for cavium invalid sriov link value.
Ananth Jasty [Tue, 23 Aug 2016 23:27:14 +0000 (16:27 -0700)]
PCI: quirk fixup for cavium invalid sriov link value.

Cavium cn88xx hardware presents an incorrect SR-IOV Function
Dependency Link, add a fixup quirk for the affected devices.

Acked-by: David Daney <david.daney@cavium.com>
Signed-off-by: Ananth Jasty <Ananth.Jasty@cavium.com>
Signed-off-by: Omer Khaliq <okhaliq@caviumnetworks.com>
Acked-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: rockchip - use devm_add_action_or_reset()
Sudip Mukherjee [Tue, 23 Aug 2016 14:58:54 +0000 (20:28 +0530)]
crypto: rockchip - use devm_add_action_or_reset()

If devm_add_action() fails we are explicitly calling the cleanup to free
the resources allocated.  Lets use the helper devm_add_action_or_reset()
and return directly in case of error, as we know that the cleanup function
has been already called by the helper if there was any error.

Signed-off-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: rsa - allow keys >= 2048 bits in FIPS mode
Stephan Mueller [Tue, 23 Aug 2016 08:09:32 +0000 (10:09 +0200)]
crypto: rsa - allow keys >= 2048 bits in FIPS mode

With a public notification, NIST now allows the use of RSA keys with a
modulus >= 2048 bits. The new rule allows any modulus size >= 2048 bits
provided that either 2048 or 3072 bits are supported at least so that
the entire RSA implementation can be CAVS tested.

This patch fixes the inability to boot the kernel in FIPS mode, because
certs/x509.genkey defines a 4096 bit RSA key per default. This key causes
the RSA signature verification to fail in FIPS mode without the patch
below.

Signed-off-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: mxc-scc - check clk_prepare_enable() error
Fabio Estevam [Mon, 22 Aug 2016 02:37:27 +0000 (23:37 -0300)]
crypto: mxc-scc - check clk_prepare_enable() error

clk_prepare_enable() may fail, so we should better check its return
value and propagate it in the case of failure.

Signed-off-by: Fabio Estevam <fabio.estevam@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: omap3-rom - Remove obsoleted functions
PrasannaKumar Muralidharan [Sat, 20 Aug 2016 16:01:06 +0000 (21:31 +0530)]
hwrng: omap3-rom - Remove obsoleted functions

Remove omap3_rom_rng_data_present method as it was returning 1 always.
Use .read callback instead of .data_read callback. This avoids use of
obsolete callbacks.

This patch is not tested with hardware as I don't have access to it.

Signed-off-by: PrasannaKumar Muralidharan <prasannatsmkumar@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: drbg - fix error return code
Wei Yongjun [Sat, 20 Aug 2016 15:06:51 +0000 (15:06 +0000)]
crypto: drbg - fix error return code

Fix to return a negative error code from the error handling
case instead of 0.

Signed-off-by: Wei Yongjun <weiyj.lk@gmail.com>
Acked-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - fix missing unlock on error in sun4i_hash()
Wei Yongjun [Sat, 20 Aug 2016 10:48:53 +0000 (10:48 +0000)]
crypto: sun4i-ss - fix missing unlock on error in sun4i_hash()

Add the missing unlock before return from function sun4i_hash()
in the error handling case.

Fixes: 477d9b2e591b ("crypto: sun4i-ss - unify update/final function")
Signed-off-by: Wei Yongjun <weiyj.lk@gmail.com>
Acked-by: Corentin LABBE <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: xor - skip speed test if the xor function is selected automatically
Martin Schwidefsky [Fri, 19 Aug 2016 12:19:30 +0000 (14:19 +0200)]
crypto: xor - skip speed test if the xor function is selected automatically

If the architecture selected the xor function with XOR_SELECT_TEMPLATE
the speed result of the do_xor_speed benchmark is of limited value.
The speed measurement increases the bootup time a little, which can
makes a difference for kernels used in container like virtual machines.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: core - Improve description of the ->read() interface
Daniel Thompson [Thu, 18 Aug 2016 12:37:21 +0000 (13:37 +0100)]
hwrng: core - Improve description of the ->read() interface

Currently, very few RNG drivers support single byte reads using the
->read() interface. Of the 14 drivers in drivers/char/hw_random that
support this interface only three of these actually support max == 1.
The other behaviours vary between return 0, return 2, return 4 and return
-EIO).

This is not a problem in practice because the core hw_random code never
performs a read shorter than 16 bytes. The documentation for this function
already contrains the alignment of the buffer pointer, so let's also
guarantee that the buffer is at least as large as its alignment.

This constraint is intended to be the weakest guarantee neccessary to
allow driver writers to safely simplify their code.

Signed-off-by: Daniel Thompson <daniel.thompson@linaro.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: doc - fix documentation for bulk registration functions
Eric Biggers [Wed, 17 Aug 2016 22:47:11 +0000 (15:47 -0700)]
crypto: doc - fix documentation for bulk registration functions

Update the documentation for crypto_register_algs() and
crypto_unregister_algs() to match the actual behavior.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: drbg - do not call drbg_instantiate in healt test
Stephan Mueller [Tue, 9 Aug 2016 19:02:36 +0000 (21:02 +0200)]
crypto: drbg - do not call drbg_instantiate in healt test

When calling the DRBG health test in FIPS mode, the Jitter RNG is not
yet present in the kernel crypto API which will cause the instantiation
to fail and thus the health test to fail.

As the health tests cover the enforcement of various thresholds, invoke
the functions that are supposed to enforce the thresholds directly.

This patch also saves precious seed.

Reported-by: Tapas Sarangi <TSarangi@trustwave.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agohwrng: pic32 - Delete unnecessary assignment for the field "owner"
Markus Elfring [Tue, 16 Aug 2016 05:51:21 +0000 (07:51 +0200)]
hwrng: pic32 - Delete unnecessary assignment for the field "owner"

The field "owner" is set by the core.
Thus delete an unneeded initialisation.

Generated by: scripts/coccinelle/api/platform_no_drv_owner.cocci
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Fix non static symbol warning
Wei Yongjun [Fri, 12 Aug 2016 00:00:09 +0000 (00:00 +0000)]
crypto: ccp - Fix non static symbol warning

Fixes the following sparse warning:

drivers/crypto/ccp/ccp-dev.c:62:14: warning:
 symbol 'ccp_increment_unit_ordinal' was not declared. Should it be static?

Signed-off-by: Wei Yongjun <weiyj.lk@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - fix indentation of two crypto alg
Corentin LABBE [Wed, 10 Aug 2016 09:45:34 +0000 (11:45 +0200)]
crypto: sun4i-ss - fix indentation of two crypto alg

Two crypto alg are badly indented, this patch fix this style issue.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - Always use sun4i_tfm_ctx for storing pointer to dev ss
Corentin LABBE [Wed, 10 Aug 2016 09:45:33 +0000 (11:45 +0200)]
crypto: sun4i-ss - Always use sun4i_tfm_ctx for storing pointer to dev ss

The dev *ss is stored both in sun4i_tfm_ctx and sun4i_req_ctx.
Since this pointer will never be changed during tfm life, it is better
to remove it from sun4i_req_ctx.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - fix spelling
Corentin LABBE [Wed, 10 Aug 2016 09:45:32 +0000 (11:45 +0200)]
crypto: sun4i-ss - fix spelling

Two words are badly spelled, this patch respell them.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - clean unused ss
Corentin LABBE [Wed, 10 Aug 2016 09:45:31 +0000 (11:45 +0200)]
crypto: sun4i-ss - clean unused ss

The ss variable is never used, remove it.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - unify update/final function
Corentin LABBE [Wed, 10 Aug 2016 09:45:30 +0000 (11:45 +0200)]
crypto: sun4i-ss - unify update/final function

The update and final functions have lots of common action.
This patch mix them in one function.
This will give some improvements:
- This will permit asynchronous support more easily
- This will permit to use finup/digest functions with some performance
  improvements

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sun4i-ss - fix a few signed warning
Corentin LABBE [Wed, 10 Aug 2016 09:45:29 +0000 (11:45 +0200)]
crypto: sun4i-ss - fix a few signed warning

The variable i is always checked against unsigned value and cannot be
negative.
This patch set it as unsigned.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: xts - fix a little typo
Corentin LABBE [Wed, 10 Aug 2016 09:29:33 +0000 (11:29 +0200)]
crypto: xts - fix a little typo

The sentence 'Based on' is misspelled, respell it.

Signed-off-by: LABBE Corentin <clabbe.montjoie@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - Don't hardcode block size in mv_cesa_ahash_cache_req
Romain Perier [Tue, 9 Aug 2016 09:03:20 +0000 (11:03 +0200)]
crypto: marvell - Don't hardcode block size in mv_cesa_ahash_cache_req

Don't use 64 'as is', as max block size in mv_cesa_ahash_cache_req. Use
CESA_MAX_HASH_BLOCK_SIZE instead, this is better for readability.

Signed-off-by: Romain Perier <romain.perier@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - Don't overwrite default creq->state during initialization
Romain Perier [Tue, 9 Aug 2016 09:03:19 +0000 (11:03 +0200)]
crypto: marvell - Don't overwrite default creq->state during initialization

Currently, in mv_cesa_{md5,sha1,sha256}_init creq->state is initialized
before the call to mv_cesa_ahash_init. This is wrong because this
function fills creq with zero by using memset, so its 'state' that
contains the default DIGEST is overwritten. This commit fixes the issue
by initializing creq->state just after the call to mv_cesa_ahash_init.

Fixes: commit b0ef51067cb4 ("crypto: marvell/cesa - initialize hash...")
Signed-off-by: Romain Perier <romain.perier@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - Update transformation context for each dequeued req
Romain Perier [Tue, 9 Aug 2016 09:03:18 +0000 (11:03 +0200)]
crypto: marvell - Update transformation context for each dequeued req

So far, sub part of mv_cesa_int was responsible of dequeuing complete
requests, then call the 'cleanup' operation on these reqs and call the
crypto api callback 'complete'. The problem is that the transformation
context 'ctx' is retrieved only once before the while loop. Which means
that the wrong 'cleanup' operation might be called on the wrong type of
cesa requests, it can lead to memory corruptions with this message:

marvell-cesa f1090000.crypto: dma_pool_free cesa_padding, 5a5a5a5a/5a5a5a5a (bad dma)

This commit fixes the issue, by updating the transformation context for
each dequeued cesa request.

Fixes: commit 85030c5168f1 ("crypto: marvell - Add support for chai...")
Signed-off-by: Romain Perier <romain.perier@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - make mv_cesa_ahash_cache_req() return bool
Thomas Petazzoni [Tue, 9 Aug 2016 09:03:17 +0000 (11:03 +0200)]
crypto: marvell - make mv_cesa_ahash_cache_req() return bool

The mv_cesa_ahash_cache_req() function always returns 0, which makes
its return value pretty much useless. However, in addition to
returning a useless value, it also returns a boolean in a variable
passed by reference to indicate if the request was already cached.

So, this commit changes mv_cesa_ahash_cache_req() to return this
boolean. It consequently simplifies the only call site of
mv_cesa_ahash_cache_req(), where the "ret" variable is no longer
needed.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - turn mv_cesa_ahash_init() into a function returning void
Thomas Petazzoni [Tue, 9 Aug 2016 09:03:16 +0000 (11:03 +0200)]
crypto: marvell - turn mv_cesa_ahash_init() into a function returning void

The mv_cesa_ahash_init() function always returns 0, and the return
value is anyway never checked. Turn it into a function returning void.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - remove unused parameter in mv_cesa_ahash_dma_add_cache()
Thomas Petazzoni [Tue, 9 Aug 2016 09:03:15 +0000 (11:03 +0200)]
crypto: marvell - remove unused parameter in mv_cesa_ahash_dma_add_cache()

The dma_iter parameter of mv_cesa_ahash_dma_add_cache() is never used,
so get rid of it.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: marvell - be explicit about destination in mv_cesa_dma_add_op()
Thomas Petazzoni [Tue, 9 Aug 2016 09:03:14 +0000 (11:03 +0200)]
crypto: marvell - be explicit about destination in mv_cesa_dma_add_op()

The mv_cesa_dma_add_op() function builds a mv_cesa_tdma_desc structure
to copy the operation description to the SRAM, but doesn't explicitly
initialize the destination of the copy. It works fine because the
operatin description must be copied at the beginning of the SRAM, and
the mv_cesa_tdma_desc structure is initialized to zero when
allocated. However, it is somewhat confusing to not have a destination
defined.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - avoid kernel warnings on probe failure
Russell King [Tue, 9 Aug 2016 07:30:10 +0000 (08:30 +0100)]
crypto: caam - avoid kernel warnings on probe failure

While debugging setkey issues, the following warnings were found while
trying to reinsert the caam module.  Fix this by avoiding the duplicated
cleanup in the probe path after caam_remove(), which has already cleaned
up the resources.

------------[ cut here ]------------
WARNING: CPU: 0 PID: 2346 at /home/rmk/git/linux-rmk/mm/vmalloc.c:1490 __vunmap+0xcc/0xf4
Trying to vfree() nonexistent vm area (f2400000)
Modules linked in: caam(+) cbc rfcomm bnep bluetooth nfsd em28xx_rc si2157 si2168 em28xx_dvb uvcvideo snd_soc_imx_sgtl5000 em28xx snd_soc_imx_spdif tveeprom snd_soc_fsl_asoc_card snd_soc_imx_audmux snd_soc_sgtl5000 imx_sdma imx2_wdt coda v4l2_mem2mem videobuf2_dma_contig snd_soc_fsl_ssi rc_cec snd_soc_fsl_spdif imx_pcm_dma videobuf2_vmalloc videobuf2_memops imx_thermal dw_hdmi_ahb_audio dw_hdmi_cec etnaviv fuse rc_pinnacle_pctv_hd [last unloaded: caam]
CPU: 0 PID: 2346 Comm: modprobe Tainted: G        W       4.8.0-rc1+ #2014
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[<c0013bb0>] (dump_backtrace) from [<c0013d4c>] (show_stack+0x18/0x1c)
[<c0013d34>] (show_stack) from [<c0357c00>] (dump_stack+0xa4/0xdc)
[<c0357b5c>] (dump_stack) from [<c002e650>] (__warn+0xdc/0x108)
[<c002e574>] (__warn) from [<c002e734>] (warn_slowpath_fmt+0x40/0x48)
[<c002e6f8>] (warn_slowpath_fmt) from [<c0151708>] (__vunmap+0xcc/0xf4)
[<c015163c>] (__vunmap) from [<c015177c>] (vunmap+0x4c/0x54)
[<c0151730>] (vunmap) from [<c001f48c>] (__iounmap+0x2c/0x30)
[<c001f460>] (__iounmap) from [<c001f118>] (iounmap+0x1c/0x20)
[<c001f0fc>] (iounmap) from [<bf247ae4>] (caam_probe+0x3dc/0x1498 [caam])
[<bf247708>] (caam_probe [caam]) from [<c042da8c>] (platform_drv_probe+0x58/0xb8)
[<c042da34>] (platform_drv_probe) from [<c042bb4c>] (driver_probe_device+0x1fc/0x2b8)
[<c042b950>] (driver_probe_device) from [<c042bcc4>] (__driver_attach+0xbc/0xc0) r10:00000000 r8:bf24b000 r7:00000000 r6:ef215844 r5:bf2490c4 r4:ef215810
[<c042bc08>] (__driver_attach) from [<c0429f14>] (bus_for_each_dev+0x5c/0x90)
[<c0429eb8>] (bus_for_each_dev) from [<c042b358>] (driver_attach+0x24/0x28)
[<c042b334>] (driver_attach) from [<c042b058>] (bus_add_driver+0xf4/0x200)
[<c042af64>] (bus_add_driver) from [<c042cadc>] (driver_register+0x80/0xfc)
[<c042ca5c>] (driver_register) from [<c042d960>] (__platform_driver_register+0x48/0x4c)
[<c042d918>] (__platform_driver_register) from [<bf24b018>] (caam_driver_init+0x18/0x24 [caam])
[<bf24b000>] (caam_driver_init [caam]) from [<c00098ac>] (do_one_initcall+0x44/0x178)
[<c0009868>] (do_one_initcall) from [<c010e034>] (do_init_module+0x68/0x1d8)
[<c010dfcc>] (do_init_module) from [<c00c8fbc>] (load_module+0x1974/0x20b0)
[<c00c7648>] (load_module) from [<c00c98d0>] (SyS_finit_module+0x94/0xa0)
[<c00c983c>] (SyS_finit_module) from [<c000fda0>] (ret_fast_syscall+0x0/0x1c)
---[ end trace 34e3370d88bb1786 ]---
------------[ cut here ]------------
WARNING: CPU: 0 PID: 2346 at /home/rmk/git/linux-rmk/drivers/clk/clk.c:594 clk_core_disable+0xe4/0x26c
Modules linked in: caam(+) cbc rfcomm bnep bluetooth nfsd em28xx_rc si2157 si2168 em28xx_dvb uvcvideo snd_soc_imx_sgtl5000 em28xx snd_soc_imx_spdif tveeprom snd_soc_fsl_asoc_card snd_soc_imx_audmux snd_soc_sgtl5000 imx_sdma imx2_wdt coda v4l2_mem2mem videobuf2_dma_contig snd_soc_fsl_ssi rc_cec snd_soc_fsl_spdif imx_pcm_dma videobuf2_vmalloc videobuf2_memops imx_thermal dw_hdmi_ahb_audio dw_hdmi_cec etnaviv fuse rc_pinnacle_pctv_hd [last unloaded: caam]
CPU: 0 PID: 2346 Comm: modprobe Tainted: G        W       4.8.0-rc1+ #2014
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[<c0013bb0>] (dump_backtrace) from [<c0013d4c>] (show_stack+0x18/0x1c)
[<c0013d34>] (show_stack) from [<c0357c00>] (dump_stack+0xa4/0xdc)
[<c0357b5c>] (dump_stack) from [<c002e650>] (__warn+0xdc/0x108)
[<c002e574>] (__warn) from [<c002e6a4>] (warn_slowpath_null+0x28/0x30)
[<c002e67c>] (warn_slowpath_null) from [<c05b113c>] (clk_core_disable+0xe4/0x26c)
[<c05b1058>] (clk_core_disable) from [<c05b2e3c>] (clk_core_disable_lock+0x20/0x2c)
[<c05b2e1c>] (clk_core_disable_lock) from [<c05b2e6c>] (clk_disable+0x24/0x28)
[<c05b2e48>] (clk_disable) from [<bf247b04>] (caam_probe+0x3fc/0x1498 [caam])
[<bf247708>] (caam_probe [caam]) from [<c042da8c>] (platform_drv_probe+0x58/0xb8)
[<c042da34>] (platform_drv_probe) from [<c042bb4c>] (driver_probe_device+0x1fc/0x2b8)
[<c042b950>] (driver_probe_device) from [<c042bcc4>] (__driver_attach+0xbc/0xc0) r10:00000000 r8:bf24b000 r7:00000000 r6:ef215844 r5:bf2490c4 r4:ef215810
[<c042bc08>] (__driver_attach) from [<c0429f14>] (bus_for_each_dev+0x5c/0x90)
[<c0429eb8>] (bus_for_each_dev) from [<c042b358>] (driver_attach+0x24/0x28)
[<c042b334>] (driver_attach) from [<c042b058>] (bus_add_driver+0xf4/0x200)
[<c042af64>] (bus_add_driver) from [<c042cadc>] (driver_register+0x80/0xfc)
[<c042ca5c>] (driver_register) from [<c042d960>] (__platform_driver_register+0x48/0x4c)
[<c042d918>] (__platform_driver_register) from [<bf24b018>] (caam_driver_init+0x18/0x24 [caam])
[<bf24b000>] (caam_driver_init [caam]) from [<c00098ac>] (do_one_initcall+0x44/0x178)
[<c0009868>] (do_one_initcall) from [<c010e034>] (do_init_module+0x68/0x1d8)
[<c010dfcc>] (do_init_module) from [<c00c8fbc>] (load_module+0x1974/0x20b0)
[<c00c7648>] (load_module) from [<c00c98d0>] (SyS_finit_module+0x94/0xa0)
[<c00c983c>] (SyS_finit_module) from [<c000fda0>] (ret_fast_syscall+0x0/0x1c)
---[ end trace 34e3370d88bb1787 ]---
------------[ cut here ]------------
WARNING: CPU: 0 PID: 2346 at /home/rmk/git/linux-rmk/drivers/clk/clk.c:476 clk_core_unprepare+0x204/0x388
Modules linked in: caam(+) cbc rfcomm bnep bluetooth nfsd em28xx_rc si2157 si2168 em28xx_dvb uvcvideo snd_soc_imx_sgtl5000 em28xx snd_soc_imx_spdif tveeprom snd_soc_fsl_asoc_card snd_soc_imx_audmux snd_soc_sgtl5000 imx_sdma imx2_wdt coda v4l2_mem2mem videobuf2_dma_contig snd_soc_fsl_ssi rc_cec snd_soc_fsl_spdif imx_pcm_dma videobuf2_vmalloc videobuf2_memops imx_thermal dw_hdmi_ahb_audio dw_hdmi_cec etnaviv fuse rc_pinnacle_pctv_hd [last unloaded: caam]
CPU: 0 PID: 2346 Comm: modprobe Tainted: G        W       4.8.0-rc1+ #2014
Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
Backtrace:
[<c0013bb0>] (dump_backtrace) from [<c0013d4c>] (show_stack+0x18/0x1c)
[<c0013d34>] (show_stack) from [<c0357c00>] (dump_stack+0xa4/0xdc)
[<c0357b5c>] (dump_stack) from [<c002e650>] (__warn+0xdc/0x108)
[<c002e574>] (__warn) from [<c002e6a4>] (warn_slowpath_null+0x28/0x30)
[<c002e67c>] (warn_slowpath_null) from [<c05b0834>] (clk_core_unprepare+0x204/0x388)
[<c05b0630>] (clk_core_unprepare) from [<c05b4c0c>] (clk_unprepare+0x2c/0x34)
[<c05b4be0>] (clk_unprepare) from [<bf247b0c>] (caam_probe+0x404/0x1498 [caam])
[<bf247708>] (caam_probe [caam]) from [<c042da8c>] (platform_drv_probe+0x58/0xb8)
[<c042da34>] (platform_drv_probe) from [<c042bb4c>] (driver_probe_device+0x1fc/0x2b8)
[<c042b950>] (driver_probe_device) from [<c042bcc4>] (__driver_attach+0xbc/0xc0) r10:00000000 r8:bf24b000 r7:00000000 r6:ef215844 r5:bf2490c4 r4:ef215810
[<c042bc08>] (__driver_attach) from [<c0429f14>] (bus_for_each_dev+0x5c/0x90)
[<c0429eb8>] (bus_for_each_dev) from [<c042b358>] (driver_attach+0x24/0x28)
[<c042b334>] (driver_attach) from [<c042b058>] (bus_add_driver+0xf4/0x200)
[<c042af64>] (bus_add_driver) from [<c042cadc>] (driver_register+0x80/0xfc)
[<c042ca5c>] (driver_register) from [<c042d960>] (__platform_driver_register+0x48/0x4c)
[<c042d918>] (__platform_driver_register) from [<bf24b018>] (caam_driver_init+0x18/0x24 [caam])
[<bf24b000>] (caam_driver_init [caam]) from [<c00098ac>] (do_one_initcall+0x44/0x178)
[<c0009868>] (do_one_initcall) from [<c010e034>] (do_init_module+0x68/0x1d8)
[<c010dfcc>] (do_init_module) from [<c00c8fbc>] (load_module+0x1974/0x20b0)
[<c00c7648>] (load_module) from [<c00c98d0>] (SyS_finit_module+0x94/0xa0)
[<c00c983c>] (SyS_finit_module) from [<c000fda0>] (ret_fast_syscall+0x0/0x1c)
---[ end trace 34e3370d88bb1788 ]---

Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - get rid of tasklet
Russell King [Mon, 8 Aug 2016 17:05:24 +0000 (18:05 +0100)]
crypto: caam - get rid of tasklet

Threaded interrupts can perform the function of the tasklet, and much
more safely too - without races when trying to take the tasklet and
interrupt down on device removal.

With the old code, there is a window where we call tasklet_kill().  If
the interrupt handler happens to be running on a different CPU, and
subsequently calls tasklet_schedule(), the tasklet will be re-scheduled
for execution.

Switching to a hardirq/threadirq combination implementation avoids this,
and it also means generic code deals with the teardown sequencing of the
threaded and non-threaded parts.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - add ahash_edesc_add_src()
Russell King [Mon, 8 Aug 2016 17:05:19 +0000 (18:05 +0100)]
crypto: caam - add ahash_edesc_add_src()

Add a helper to map the source scatterlist into the descriptor.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - move job descriptor initialisation to ahash_edesc_alloc()
Russell King [Mon, 8 Aug 2016 17:05:13 +0000 (18:05 +0100)]
crypto: caam - move job descriptor initialisation to ahash_edesc_alloc()

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - add ahash_edesc_alloc() for descriptor allocation
Russell King [Mon, 8 Aug 2016 17:05:08 +0000 (18:05 +0100)]
crypto: caam - add ahash_edesc_alloc() for descriptor allocation

Add a helper function to perform the descriptor allocation.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - check and use dma_map_sg() return code
Russell King [Mon, 8 Aug 2016 17:05:03 +0000 (18:05 +0100)]
crypto: caam - check and use dma_map_sg() return code

Strictly, dma_map_sg() may coalesce SG entries, but in practise on iMX
hardware, this will never happen.  However, dma_map_sg() can fail, and
we completely fail to check its return value.  So, fix this properly.

Arrange the code to map the scatterlist early, so we know how many
scatter table entries to allocate, and then fill them in.  This allows
us to keep relatively simple error cleanup paths.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - ensure that we clean up after an error
Russell King [Mon, 8 Aug 2016 17:04:58 +0000 (18:04 +0100)]
crypto: caam - ensure that we clean up after an error

Ensure that we clean up allocations and DMA mappings after encountering
an error rather than just giving up and leaking memory and resources.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - replace sec4_sg pointer with array
Russell King [Mon, 8 Aug 2016 17:04:52 +0000 (18:04 +0100)]
crypto: caam - replace sec4_sg pointer with array

Since the extended descriptor includes the hardware descriptor, and the
sec4 scatterlist immediately follows this, we can declare it as a array
at the very end of the extended descriptor.  This allows us to get rid
of an initialiser for every site where we allocate an extended
descriptor.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - mark the hardware descriptor as cache line aligned
Russell King [Mon, 8 Aug 2016 17:04:47 +0000 (18:04 +0100)]
crypto: caam - mark the hardware descriptor as cache line aligned

Mark the hardware descriptor as being cache line aligned; on DMA
incoherent architectures, the hardware descriptor should sit in a
separate cache line from the CPU accessed data to avoid polluting
the caches.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - incorporate job descriptor into struct ahash_edesc
Russell King [Mon, 8 Aug 2016 17:04:42 +0000 (18:04 +0100)]
crypto: caam - incorporate job descriptor into struct ahash_edesc

Rather than giving the descriptor as hw_desc[0], give it's real size.
All places where we allocate an ahash_edesc incorporate DESC_JOB_IO_LEN
bytes of job descriptor.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - ensure descriptor buffers are cacheline aligned
Russell King [Mon, 8 Aug 2016 17:04:36 +0000 (18:04 +0100)]
crypto: caam - ensure descriptor buffers are cacheline aligned

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix DMA API mapping leak
Russell King [Mon, 8 Aug 2016 17:04:31 +0000 (18:04 +0100)]
crypto: caam - fix DMA API mapping leak

caamhash contains this weird code:

src_nents = sg_count(req->src, req->nbytes);
dma_map_sg(jrdev, req->src, src_nents ? : 1, DMA_TO_DEVICE);
...
edesc->src_nents = src_nents;

sg_count() returns zero when sg_nents_for_len() returns zero or one.
This means we don't need to use a hardware scatterlist.  However,
setting src_nents to zero causes problems when we unmap:

if (edesc->src_nents)
dma_unmap_sg_chained(dev, req->src, edesc->src_nents,
     DMA_TO_DEVICE, edesc->chained);

as zero here means that we have no entries to unmap.  This causes us
to leak DMA mappings, where we map one scatterlist entry and then
fail to unmap it.

This can be fixed in two ways: either by writing the number of entries
that were requested of dma_map_sg(), or by reworking the "no SG
required" case.

We adopt the re-work solution here - we replace sg_count() with
sg_nents_for_len(), so src_nents now contains the real number of
scatterlist entries, and we then change the test for using the
hardware scatterlist to src_nents > 1 rather than just non-zero.

This change passes my sshd, openssl tests hashing /bin and tcrypt
tests.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - Fix set_reqsize call
Will Thomas [Fri, 5 Aug 2016 13:00:20 +0000 (14:00 +0100)]
crypto: img-hash - Fix set_reqsize call

Properly allocate enough memory to respect the fallback.

Signed-off-by: Will Thomas <will.thomas@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - log a successful probe
James Hartley [Fri, 5 Aug 2016 13:00:19 +0000 (14:00 +0100)]
crypto: img-hash - log a successful probe

Currently the probe function only emits an output on success
when debug is specifically enabled. It would be more useful
if this happens by default.

Signed-off-by: James Hartley <james.hartley@imgtec.com>
Reviewed-by: Will Thomas <will.thomas@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - Add support for export and import
James Hartley [Fri, 5 Aug 2016 13:00:18 +0000 (14:00 +0100)]
crypto: img-hash - Add support for export and import

Currently the img-hash accelerator does not probe
successfully due to a change in the checks made during
registration with the crypto framework. This is due to
import and export functions not being defined. Correct
this.

Signed-off-by: James Hartley <james.hartley@imgtec.com>
Signed-off-by: Will Thomas <will.thomas@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - Add suspend resume hooks for img hash
Govindraj Raja [Fri, 5 Aug 2016 13:00:17 +0000 (14:00 +0100)]
crypto: img-hash - Add suspend resume hooks for img hash

Current img hash claims sys and periph gate clocks
and this can be gated in system suspend scenarios.

Add support for Device pm ops for img hash to gate
the clocks claimed by img hash.

Signed-off-by: Govindraj Raja <Govindraj.Raja@imgtec.com>
Reviewed-by: Will Thomas <will.thomas@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - Reconfigure DMA Burst length
Will Thomas [Fri, 5 Aug 2016 13:00:16 +0000 (14:00 +0100)]
crypto: img-hash - Reconfigure DMA Burst length

Burst length of 16 drives the hash accelerator out of spec
and causes stability issues in some cases. Reduce this to
stop data being lost.

Signed-off-by: Will Thomas <will.thomas@imgtec.com>
Reviewed-by: James Hartley <james.hartley@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - Fix hash request context
Will Thomas [Fri, 5 Aug 2016 13:00:15 +0000 (14:00 +0100)]
crypto: img-hash - Fix hash request context

Move 0 length buffer to end of structure to stop overwriting
fallback request data. This doesn't cause a bug itself as the
buffer is never used alongside the fallback but should be
changed.

Signed-off-by: Will Thomas <will.thomas@imgtec.com>
Reviewed-by: James Hartley <james.hartley@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: img-hash - Fix null pointer exception
Will Thomas [Fri, 5 Aug 2016 13:00:14 +0000 (14:00 +0100)]
crypto: img-hash - Fix null pointer exception

Sporadic null pointer exceptions came from here. Fix them.

Signed-off-by: Will Thomas <will.thomas@imgtec.com>
Reviewed-by: James Hartley <james.hartley@imgtec.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: testmgr - fix !x==y confusion
Yanjiang Jin [Fri, 29 Jul 2016 08:32:09 +0000 (16:32 +0800)]
crypto: testmgr - fix !x==y confusion

"if (!ret == template[i].fail)" is confusing to compilers (gcc5):

crypto/testmgr.c: In function '__test_aead':
crypto/testmgr.c:531:12: warning: logical not is only applied to the
left hand side of comparison [-Wlogical-not-parentheses]
   if (!ret == template[i].fail) {
            ^

Let there be 'if (template[i].fail == !ret) '.

Signed-off-by: Yanjiang Jin <yanjiang.jin@windriver.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Enable use of the additional CCP
Gary R Hook [Wed, 27 Jul 2016 00:10:49 +0000 (19:10 -0500)]
crypto: ccp - Enable use of the additional CCP

A second CCP is available, identical to the first, with
its ownn PCI ID. Make it available for use by the crypto
subsystem, as well as for DMA activity and random
number generation.

This device is not pre-configured at at boot time. The
driver must configure it (during the probe) for use.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Enable DMA service on a v5 CCP
Gary R Hook [Wed, 27 Jul 2016 00:10:40 +0000 (19:10 -0500)]
crypto: ccp - Enable DMA service on a v5 CCP

Every CCP is capable of providing general DMA services.
Register the device as a provider.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Add support for the RNG in a version 5 CCP
Gary R Hook [Wed, 27 Jul 2016 00:10:31 +0000 (19:10 -0500)]
crypto: ccp - Add support for the RNG in a version 5 CCP

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Let a v5 CCP provide the same function as v3
Gary R Hook [Wed, 27 Jul 2016 00:10:21 +0000 (19:10 -0500)]
crypto: ccp - Let a v5 CCP provide the same function as v3

Enable equivalent function on a v5 CCP. Add support for a
version 5 CCP which enables AES/XTS/SHA services. Also,
more work on the data structures to virtualize
functionality.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Refactor code to enable checks for queue space.
Gary R Hook [Wed, 27 Jul 2016 00:10:13 +0000 (19:10 -0500)]
crypto: ccp - Refactor code to enable checks for queue space.

Available queue space is used to decide (by counting free slots)
if we have to put a command on hold or if it can be sent
to the engine immediately.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Refactor code supporting the CCP's RNG
Gary R Hook [Wed, 27 Jul 2016 00:10:02 +0000 (19:10 -0500)]
crypto: ccp - Refactor code supporting the CCP's RNG

Make the RNG support code common (where possible) in
preparation for adding a v5 device.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Refactor the storage block allocation code
Gary R Hook [Wed, 27 Jul 2016 00:09:50 +0000 (19:09 -0500)]
crypto: ccp - Refactor the storage block allocation code

Move the KSB access/management functions to the v3
device file, and add function pointers to the actions
structure. At the operations layer all of the references
to the storage block will be generic (virtual). This is
in preparation for a version 5 device, in which the
private storage block is managed differently.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Refactoring: symbol cleanup
Gary R Hook [Wed, 27 Jul 2016 00:09:40 +0000 (19:09 -0500)]
crypto: ccp - Refactoring: symbol cleanup

Form and use of the local storage block in the CCP is
particular to the device version. Much of the code that
accesses the storage block can treat it as a virtual
resource, and will under go some renaming. Device-specific
access to the memory will be moved into device file.
Service functions will be added to the actions
structure.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Shorten the fields of the action structure
Gary R Hook [Wed, 27 Jul 2016 00:09:31 +0000 (19:09 -0500)]
crypto: ccp - Shorten the fields of the action structure

Use more concise field names; "perform_" is too verbose.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ccp - Abstract PCI info for the CCP
Gary R Hook [Wed, 27 Jul 2016 00:09:20 +0000 (19:09 -0500)]
crypto: ccp - Abstract PCI info for the CCP

Device-specific values for the BAR and offset should be found
in the version data structure.

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 non-conforming comment style
Gary R Hook [Tue, 26 Jul 2016 23:09:46 +0000 (18:09 -0500)]
crypto: ccp - Fix non-conforming comment style

Adhere to the cryptodev comment convention.

Signed-off-by: Gary R Hook <gary.hook@amd.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ixp4xx - Add missing npe_c release in error branches
Quentin Lambert [Fri, 22 Jul 2016 13:32:41 +0000 (15:32 +0200)]
crypto: ixp4xx - Add missing npe_c release in error branches

Most error branches following the call to npe_request contain a call to
npe_request. This patch add a call to npe_release to error branches
following the call to npe_request that do not have it.

This issue was found with Hector.

Signed-off-by: Quentin Lambert <lambert.quentin@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: ixp4xx - Fix a "simple if" coding style warning
Quentin Lambert [Fri, 22 Jul 2016 13:32:42 +0000 (15:32 +0200)]
crypto: ixp4xx - Fix a "simple if" coding style warning

Signed-off-by: Quentin Lambert <lambert.quentin@gmail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix non-hmac hashes
Russell King [Tue, 9 Aug 2016 07:27:17 +0000 (08:27 +0100)]
crypto: caam - fix non-hmac hashes

Since 6de62f15b581 ("crypto: algif_hash - Require setkey before
accept(2)"), the AF_ALG interface requires userspace to provide a key
to any algorithm that has a setkey method.  However, the non-HMAC
algorithms are not keyed, so setting a key is unnecessary.

Fix this by removing the setkey method from the non-keyed hash
algorithms.

Fixes: 6de62f15b581 ("crypto: algif_hash - Require setkey before accept(2)")
Cc: <stable@vger.kernel.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: powerpc - CRYPT_CRC32C_VPMSUM should depend on ALTIVEC
Michael Ellerman [Mon, 8 Aug 2016 22:46:15 +0000 (08:46 +1000)]
crypto: powerpc - CRYPT_CRC32C_VPMSUM should depend on ALTIVEC

The optimised crc32c implementation depends on VMX (aka. Altivec)
instructions, so the kernel must be built with Altivec support in order
for the crc32c code to build.

Fixes: 6dd7a82cc54e ("crypto: powerpc - Add POWER8 optimised crc32c")
Acked-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - defer aead_set_sh_desc in case of zero authsize
Horia Geantă [Thu, 4 Aug 2016 17:02:47 +0000 (20:02 +0300)]
crypto: caam - defer aead_set_sh_desc in case of zero authsize

To be able to generate shared descriptors for AEAD, the authentication size
needs to be known. However, there is no imposed order of calling .setkey,
.setauthsize callbacks.

Thus, in case authentication size is not known at .setkey time, defer it
until .setauthsize is called.

The authsize != 0 check was incorrectly removed when converting the driver
to the new AEAD interface.

Cc: <stable@vger.kernel.org> # 4.3+
Fixes: 479bcc7c5b9e ("crypto: caam - Convert authenc to new AEAD interface")
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: caam - fix echainiv(authenc) encrypt shared descriptor
Horia Geantă [Thu, 4 Aug 2016 17:02:46 +0000 (20:02 +0300)]
crypto: caam - fix echainiv(authenc) encrypt shared descriptor

There are a few things missed by the conversion to the
new AEAD interface:

1 - echainiv(authenc) encrypt shared descriptor

The shared descriptor is incorrect: due to the order of operations,
at some point in time MATH3 register is being overwritten.

2 - buffer used for echainiv(authenc) encrypt shared descriptor

Encrypt and givencrypt shared descriptors (for AEAD ops) are mutually
exclusive and thus use the same buffer in context state: sh_desc_enc.

However, there's one place missed by s/sh_desc_givenc/sh_desc_enc,
leading to errors when echainiv(authenc(...)) algorithms are used:
DECO: desc idx 14: Header Error. Invalid length or parity, or
certain other problems.

While here, also fix a typo: dma_mapping_error() is checking
for validity of sh_desc_givenc_dma instead of sh_desc_enc_dma.

Cc: <stable@vger.kernel.org> # 4.3+
Fixes: 479bcc7c5b9e ("crypto: caam - Convert authenc to new AEAD interface")
Signed-off-by: Horia Geantă <horia.geanta@nxp.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agocrypto: sha3 - Add missing ULL suffixes for 64-bit constants
Geert Uytterhoeven [Wed, 3 Aug 2016 17:37:03 +0000 (19:37 +0200)]
crypto: sha3 - Add missing ULL suffixes for 64-bit constants

On 32-bit (e.g. with m68k-linux-gnu-gcc-4.1):

    crypto/sha3_generic.c:27: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:28: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:29: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:29: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:31: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:31: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:32: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:32: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:32: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:33: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:33: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:34: warning: integer constant is too large for ‘long’ type
    crypto/sha3_generic.c:34: warning: integer constant is too large for ‘long’ type

Fixes: 53964b9ee63b7075 ("crypto: sha3 - Add SHA-3 hash algorithm")
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
7 years agoLinux 4.8-rc1 v4.8-rc1
Linus Torvalds [Mon, 8 Aug 2016 01:18:00 +0000 (18:18 -0700)]
Linux 4.8-rc1

7 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Sun, 7 Aug 2016 23:38:45 +0000 (16:38 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull more block fixes from Jens Axboe:
 "As mentioned in the pull the other day, a few more fixes for this
  round, all related to the bio op changes in this series.

  Two fixes, and then a cleanup, renaming bio->bi_rw to bio->bi_opf.  I
  wanted to do that change right after or right before -rc1, so that
  risk of conflict was reduced.  I just rebased the series on top of
  current master, and no new ->bi_rw usage has snuck in"

* 'for-linus' of git://git.kernel.dk/linux-block:
  block: rename bio bi_rw to bi_opf
  target: iblock_execute_sync_cache() should use bio_set_op_attrs()
  mm: make __swap_writepage() use bio_set_op_attrs()
  block/mm: make bdev_ops->rw_page() take a bool for read/write

7 years agoMerge tag 'drm-for-v4.8-zpos' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Sun, 7 Aug 2016 23:35:08 +0000 (16:35 -0700)]
Merge tag 'drm-for-v4.8-zpos' of git://people.freedesktop.org/~airlied/linux

Pull drm zpos property support from Dave Airlie:
 "This tree was waiting on some media stuff I hadn't had time to get a
  stable branchpoint off, so I just waited until it was all in your tree
  first.

  It's been around a bit on the list and shouldn't affect anything
  outside adding the generic API and moving some ARM drivers to using
  it"

* tag 'drm-for-v4.8-zpos' of git://people.freedesktop.org/~airlied/linux:
  drm: rcar: use generic code for managing zpos plane property
  drm/exynos: use generic code for managing zpos plane property
  drm: sti: use generic zpos for plane
  drm: add generic zpos property

7 years agoblock: rename bio bi_rw to bi_opf
Jens Axboe [Fri, 5 Aug 2016 21:35:16 +0000 (15:35 -0600)]
block: rename bio bi_rw to bi_opf

Since commit 63a4cc24867d, bio->bi_rw contains flags in the lower
portion and the op code in the higher portions. This means that
old code that relies on manually setting bi_rw is most likely
going to be broken. Instead of letting that brokeness linger,
rename the member, to force old and out-of-tree code to break
at compile time instead of at runtime.

No intended functional changes in this commit.

Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agotarget: iblock_execute_sync_cache() should use bio_set_op_attrs()
Jens Axboe [Mon, 1 Aug 2016 15:39:23 +0000 (09:39 -0600)]
target: iblock_execute_sync_cache() should use bio_set_op_attrs()

The original commit missed this function, it needs to mark it a
write flush.

Cc: Mike Christie <mchristi@redhat.com>
Fixes: e742fc32fcb4 ("target: use bio op accessors")
Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agomm: make __swap_writepage() use bio_set_op_attrs()
Jens Axboe [Mon, 1 Aug 2016 15:38:44 +0000 (09:38 -0600)]
mm: make __swap_writepage() use bio_set_op_attrs()

Cleaner than manipulating bio->bi_rw flags directly.

Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agoblock/mm: make bdev_ops->rw_page() take a bool for read/write
Jens Axboe [Fri, 5 Aug 2016 14:11:04 +0000 (08:11 -0600)]
block/mm: make bdev_ops->rw_page() take a bool for read/write

Commit abf545484d31 changed it from an 'rw' flags type to the
newer ops based interface, but now we're effectively leaking
some bdev internals to the rest of the kernel. Since we only
care about whether it's a read or a write at that level, just
pass in a bool 'is_write' parameter instead.

Then we can also move op_is_write() and friends back under
CONFIG_BLOCK protection.

Reviewed-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
7 years agoMerge tag 'doc-4.8-fixes' of git://git.lwn.net/linux
Linus Torvalds [Sun, 7 Aug 2016 14:23:17 +0000 (10:23 -0400)]
Merge tag 'doc-4.8-fixes' of git://git.lwn.net/linux

Pull documentation fixes from Jonathan Corbet:
 "Three fixes for the docs build, including removing an annoying warning
  on 'make help' if sphinx isn't present"

* tag 'doc-4.8-fixes' of git://git.lwn.net/linux:
  DocBook: use DOCBOOKS="" to ignore DocBooks instead of IGNORE_DOCBOOKS=1
  Documenation: update cgroup's document path
  Documentation/sphinx: do not warn about missing tools in 'make help'