Optimize GETU32 to use 4-byte memcpy (modern gcc will convert
such memcpy to single move instruction on i386).
Original GETU32 did four byte fetches, and shifted/XORed those.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Rename some macros to shorter names: CAMELLIA_RR8 -> ROR8,
making it easier to understand that it is just a right rotation,
nothing camellia-specific in it.
CAMELLIA_SUBKEY_L() -> SUBKEY_L() - just shorter.
Move be32 <-> cpu conversions out of en/decrypt128/256 and into
camellia_en/decrypt - no reason to have that code duplicated twice.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Move code blocks around so that related pieces are closer together:
e.g. CAMELLIA_ROUNDSM macro does not need to be separated
from the rest of the code by huge array of constants.
Remove unused macros (COPY4WORD, SWAP4WORD, XOR4WORD[2])
Drop SUBL(), SUBR() macros which only obscure things.
Same for CAMELLIA_SP1110() macro and KEY_TABLE_TYPE typedef.
Remove useless comments:
/* encryption */ -- well it's obvious enough already!
void camellia_encrypt128(...)
Combine swap with copying at the beginning/end of encrypt/decrypt.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Acked-by: Noriaki TAKAMIYA <takamiya@po.ntts.co.jp>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Currently twofish cipher key setup code
has unrolled loops - approximately 70-100
instructions are repeated 40 times.
As a result, twofish module is the biggest module
in crypto/*.
Unrolling produces x2.5 more code (+18k on i386), and speeds up key
setup by 7%:
unrolled: twofish_setkey/sec: 41128
loop: twofish_setkey/sec: 38148
CALC_K256: ~100 insns each
CALC_K192: ~90 insns
CALC_K: ~70 insns
Attached patch removes this unrolling.
$ size */twofish_common.o
text data bss dec hex filename
37920 0 0 37920 9420 crypto.org/twofish_common.o
13209 0 0 13209 3399 crypto/twofish_common.o
Run tested (modprobe tcrypt reports ok). Please apply.
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This three defines are used in all AES related hardware.
Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
HIFN driver update to use DES weak key checks (exported in this patch).
Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch creates include/crypto/des.h for common macros shared between
DES implementations.
Signed-off-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch implements CTR mode for IPsec.
It is based off of RFC 3686.
Please note:
1. CTR turns a block cipher into a stream cipher.
Encryption is done in blocks, however the last block
may be a partial block.
A "counter block" is encrypted, creating a keystream
that is xor'ed with the plaintext. The counter portion
of the counter block is incremented after each block
of plaintext is encrypted.
Decryption is performed in same manner.
2. The CTR counterblock is composed of,
nonce + IV + counter
The size of the counterblock is equivalent to the
blocksize of the cipher.
sizeof(nonce) + sizeof(IV) + sizeof(counter) = blocksize
The CTR template requires the name of the cipher
algorithm, the sizeof the nonce, and the sizeof the iv.
ctr(cipher,sizeof_nonce,sizeof_iv)
So for example,
ctr(aes,4,8)
specifies the counterblock will be composed of 4 bytes
from a nonce, 8 bytes from the iv, and 4 bytes for counter
since aes has a blocksize of 16 bytes.
3. The counter portion of the counter block is stored
in big endian for conformance to rfc 3686.
Signed-off-by: Joy Latten <latten@austin.ibm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As it is crypto_remove_spawn may try to unregister an instance which is
yet to be registered. This patch fixes this by checking whether the
instance has been registered before attempting to remove it.
It also removes a bogus cra_destroy check in crypto_register_instance as
1) it's outside the mutex;
2) we have a check in __crypto_register_alg already.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
It seems that newer versions of gcc have regressed in their abilities to
analyse initialisations. This patch moves the initialisations up to avoid
the warnings.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Not architecture specific code should not #include <asm/scatterlist.h>.
This patch therefore either replaces them with
#include <linux/scatterlist.h> or simply removes them if they were
unused.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
This patch moves the sg_init_table out of the timing loops for hash
algorithms so that it doesn't impact on the speed test results.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
hmac_setkey(), hmac_init(), and hmac_final() have
a singular on-stack scatterlist. Initialit is
using sg_init_one() instead of using sg_set_buf().
Signed-off-by: David S. Miller <davem@davemloft.net>
Crypto now uses SG helper functions. Fix hmac_digest to use those
functions correctly and fix the oops associated with it.
Signed-off-by: Vlad Yasevich <vladislav.yasevich@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Most drivers need to set length and offset as well, so may as well fold
those three lines into one.
Add sg_assign_page() for those two locations that only needed to set
the page, where the offset/length is set outside of the function context.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Convert the subdirectory "crypto" to UTF-8. The files changed are
<crypto/fcrypt.c> and <crypto/api.c>.
Signed-off-by: John Anthony Kazos Jr. <jakj@j-a-k-j.com>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
There are currently several SHA implementations that all define their own
initialization vectors and size values. Since this values are idential
move them to a header file under include/crypto.
Signed-off-by: Jan Glauber <jang@de.ibm.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Loading the crypto algorithm by the alias instead of by module directly
has the advantage that all possible implementations of this algorithm
are loaded automatically and the crypto API can choose the best one
depending on its priority.
Additionally it ensures that the generic implementation as well as the
HW driver (if available) is loaded in case the HW driver needs the
generic version as fallback in corner cases.
Also remove the probe for sha1 in padlock's init code.
Quote from Herbert:
The probe is actually pointless since we can always probe when
the algorithm is actually used which does not lead to dead-locks
like this.
Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Loading the crypto algorithm by the alias instead of by module directly
has the advantage that all possible implementations of this algorithm
are loaded automatically and the crypto API can choose the best one
depending on its priority.
Additionally it ensures that the generic implementation as well as the
HW driver (if available) is loaded in case the HW driver needs the
generic version as fallback in corner cases.
Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Loading the crypto algorithm by the alias instead of by module directly
has the advantage that all possible implementations of this algorithm
are loaded automatically and the crypto API can choose the best one
depending on its priority.
Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the helper blkcipher_walk_virt_block which is similar to
blkcipher_walk_virt but uses a supplied block size instead of the block
size of the block cipher. This is useful for CTR where the block size is
1 but we still want to walk by the block size of the underlying cipher.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Now that the block size is no longer a multiple of the alignment, we need to
increase the kmalloc amount in blkcipher_next_slow to use the aligned block
size.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds a comment to explain why we compare the cra_driver_name of
the algorithm being registered against the cra_name of a larval as opposed
to the cra_driver_name of the larval.
In fact larvals have only one name, cra_name which is the name that was
requested by the user. The test here is simply trying to find out whether
the algorithm being registered can or can not satisfy the larval.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Previously we assumed for convenience that the block size is a multiple of
the algorithm's required alignment. With the pending addition of CTR this
will no longer be the case as the block size will be 1 due to it being a
stream cipher. However, the alignment requirement will be that of the
underlying implementation which will most likely be greater than 1.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
We do not allow spaces in algorithm names or parameters. Thanks to Joy Latten
for pointing this out.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As Joy Latten points out, inner algorithm parameters will miss the closing
bracket which will also cause the outer algorithm to terminate prematurely.
This patch fixes that also kills the WARN_ON if the number of parameters
exceed the maximum as that is a user error.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
XTS currently considered to be the successor of the LRW mode by the IEEE1619
workgroup. LRW was discarded, because it was not secure if the encyption key
itself is encrypted with LRW.
XTS does not have this problem. The implementation is pretty straightforward,
a new function was added to gf128mul to handle GF(128) elements in ble format.
Four testvectors from the specification
http://grouper.ieee.org/groups/1619/email/pdf00086.pdf
were added, and they verify on my system.
Signed-off-by: Rik Snel <rsnel@cube.dyndns.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Use max in blkcipher_get_spot() instead of open coding it.
Signed-off-by: Ingo Oeser <ioe-lkml@rameria.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When scatterwalk is built as a module digest.c was broken because it
requires the crypto_km_types structure which is in scatterwalk. This
patch removes the crypto_km_types structure by encoding the logic into
crypto_kmap_type directly.
In fact, this even saves a few bytes of code (not to mention the data
structure itself) on i386 which is about the only place where it's
needed.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the authenc algorithm which constructs an AEAD algorithm
from an asynchronous block cipher and a hash. The construction is done
by concatenating the encrypted result from the cipher with the output
from the hash, as is used by the IPsec ESP protocol.
The authenc algorithm exists as a template with four parameters:
authenc(auth, authsize, enc, enckeylen).
The authentication algorithm, the authentication size (i.e., truncating
the output of the authentication algorithm), the encryption algorithm,
and the encryption key length. Both the size field and the key length
field are in bytes. For example, AES-128 with SHA1-HMAC would be
represented by
authenc(hmac(sha1), 12, cbc(aes), 16)
The key for the authenc algorithm is the concatenation of the keys for
the authentication algorithm with the encryption algorithm. For the
above example, if a key of length 36 bytes is given, then hmac(sha1)
would receive the first 20 bytes while the last 16 would be given to
cbc(aes).
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds the function scatterwalk_map_and_copy which reads or
writes a chunk of data from a scatterlist at a given offset. It will
be used by authenc which would read/write the authentication data at
the end of the cipher/plain text.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The scatterwalk code is only used by algorithms that can be built as
a module. Therefore we can move it into algapi.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Since not everyone needs a queue pointer and those who need it can
always get it from the context anyway the queue pointer in the
common alg object is redundant.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch ensures that kernel.h and slab.h are included for
the setkey_unaligned function. It also breaks a couple of
long lines.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds support for having multiple parameters to
a template, separated by a comma. It also adds support
for integer parameters in addition to the current algorithm
parameter type.
This will be used by the authenc template which will have
four parameters: the authentication algorithm, the encryption
algorithm, the authentication size and the encryption key
length.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds crypto_aead which is the interface for AEAD
(Authenticated Encryption with Associated Data) algorithms.
AEAD algorithms perform authentication and encryption in one
step. Traditionally users (such as IPsec) would use two
different crypto algorithms to perform these. With AEAD
this comes down to one algorithm and one operation.
Of course if traditional algorithms were used we'd still
be doing two operations underneath. However, real AEAD
algorithms may allow the underlying operations to be
optimised as well.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds support for the SEED cipher (RFC4269).
This patch have been used in few VPN appliance vendors in Korea for
several years. And it was verified by KISA, who developed the
algorithm itself.
As its importance in Korean banking industry, it would be great
if linux incorporates the support.
Signed-off-by: Hye-Shik Chang <perky@FreeBSD.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Other options requiring specific block cipher algorithms already have
the appropriate select's.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Fix dma_wait_for_async_tx to not loop forever in the case where a
dependency chain is longer than two entries. This condition will not
happen with current in-kernel drivers, but fix it for future drivers.
Found-by: Saeed Bishara <saeed.bishara@gmail.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The previous patch had the conditional inverted. This patch fixes it
so that we return the original position if it does not straddle a page.
Thanks to Bob Gilligan for spotting this.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The function blkcipher_get_spot tries to return a buffer of
the specified length that does not straddle a page. It has
an off-by-one bug so it may advance a page unnecessarily.
What's worse, one of its callers doesn't provide a buffer
that's sufficiently long for this operation.
This patch fixes both problems. Thanks to Bob Gilligan for
diagnosing this problem and providing a fix.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
setkey_unaligned() commited in ca7c39385c
overwrites unallocated memory in the following memset() because
I used the wrong buffer length.
Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Andrew Morton:
[async_memcpy] is very wrong if both ASYNC_TX_KMAP_DST and
ASYNC_TX_KMAP_SRC can ever be set. We'll end up using the same kmap
slot for both src add dest and we get either corrupted data or a BUG.
Evgeniy Polyakov:
Btw, shouldn't it always be kmap_atomic() even if flag is not set.
That pages are usual one returned by alloc_page().
So fix the usage of kmap_atomic and kill the ASYNC_TX_KMAP_DST and
ASYNC_TX_KMAP_SRC flags.
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Simple and stupid - just use the same code from another place in the kernel.
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Acked-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The async_tx api provides methods for describing a chain of asynchronous
bulk memory transfers/transforms with support for inter-transactional
dependencies. It is implemented as a dmaengine client that smooths over
the details of different hardware offload engine implementations. Code
that is written to the api can optimize for asynchronous operation and the
api will fit the chain of operations to the available offload resources.
I imagine that any piece of ADMA hardware would register with the
'async_*' subsystem, and a call to async_X would be routed as
appropriate, or be run in-line. - Neil Brown
async_tx exploits the capabilities of struct dma_async_tx_descriptor to
provide an api of the following general format:
struct dma_async_tx_descriptor *
async_<operation>(..., struct dma_async_tx_descriptor *depend_tx,
dma_async_tx_callback cb_fn, void *cb_param)
{
struct dma_chan *chan = async_tx_find_channel(depend_tx, <operation>);
struct dma_device *device = chan ? chan->device : NULL;
int int_en = cb_fn ? 1 : 0;
struct dma_async_tx_descriptor *tx = device ?
device->device_prep_dma_<operation>(chan, len, int_en) : NULL;
if (tx) { /* run <operation> asynchronously */
...
tx->tx_set_dest(addr, tx, index);
...
tx->tx_set_src(addr, tx, index);
...
async_tx_submit(chan, tx, flags, depend_tx, cb_fn, cb_param);
} else { /* run <operation> synchronously */
...
<operation>
...
async_tx_sync_epilog(flags, depend_tx, cb_fn, cb_param);
}
return tx;
}
async_tx_find_channel() returns a capable channel from its pool. The
channel pool is organized as a per-cpu array of channel pointers. The
async_tx_rebalance() routine is tasked with managing these arrays. In the
uniprocessor case async_tx_rebalance() tries to spread responsibility
evenly over channels of similar capabilities. For example if there are two
copy+xor channels, one will handle copy operations and the other will
handle xor. In the SMP case async_tx_rebalance() attempts to spread the
operations evenly over the cpus, e.g. cpu0 gets copy channel0 and xor
channel0 while cpu1 gets copy channel 1 and xor channel 1. When a
dependency is specified async_tx_find_channel defaults to keeping the
operation on the same channel. A xor->copy->xor chain will stay on one
channel if it supports both operation types, otherwise the transaction will
transition between a copy and a xor resource.
Currently the raid5 implementation in the MD raid456 driver has been
converted to the async_tx api. A driver for the offload engines on the
Intel Xscale series of I/O processors, iop-adma, is provided in a later
commit. With the iop-adma driver and async_tx, raid456 is able to offload
copy, xor, and xor-zero-sum operations to hardware engines.
On iop342 tiobench showed higher throughput for sequential writes (20 - 30%
improvement) and sequential reads to a degraded array (40 - 55%
improvement). For the other cases performance was roughly equal, +/- a few
percentage points. On a x86-smp platform the performance of the async_tx
implementation (in synchronous mode) was also +/- a few percentage points
of the original implementation. According to 'top' on iop342 CPU
utilization drops from ~50% to ~15% during a 'resync' while the speed
according to /proc/mdstat doubles from ~25 MB/s to ~50 MB/s.
The tiobench command line used for testing was: tiobench --size 2048
--block 4096 --block 131072 --dir /mnt/raid --numruns 5
* iop342 had 1GB of memory available
Details:
* if CONFIG_DMA_ENGINE=n the asynchronous path is compiled away by making
async_tx_find_channel a static inline routine that always returns NULL
* when a callback is specified for a given transaction an interrupt will
fire at operation completion time and the callback will occur in a
tasklet. if the the channel does not support interrupts then a live
polling wait will be performed
* the api is written as a dmaengine client that requests all available
channels
* In support of dependencies the api implicitly schedules channel-switch
interrupts. The interrupt triggers the cleanup tasklet which causes
pending operations to be scheduled on the next channel
* Xor engines treat an xor destination address differently than a software
xor routine. To the software routine the destination address is an implied
source, whereas engines treat it as a write-only destination. This patch
modifies the xor_blocks routine to take a an explicit destination address
to mirror the hardware.
Changelog:
* fixed a leftover debug print
* don't allow callbacks in async_interrupt_cond
* fixed xor_block changes
* fixed usage of ASYNC_TX_XOR_DROP_DEST
* drop dma mapping methods, suggested by Chris Leech
* printk warning fixups from Andrew Morton
* don't use inline in C files, Adrian Bunk
* select the API when MD is enabled
* BUG_ON xor source counts <= 1
* implicitly handle hardware concerns like channel switching and
interrupts, Neil Brown
* remove the per operation type list, and distribute operation capabilities
evenly amongst the available channels
* simplify async_tx_find_channel to optimize the fast path
* introduce the channel_table_initialized flag to prevent early calls to
the api
* reorganize the code to mimic crypto
* include mm.h as not all archs include it in dma-mapping.h
* make the Kconfig options non-user visible, Adrian Bunk
* move async_tx under crypto since it is meant as 'core' functionality, and
the two may share algorithms in the future
* move large inline functions into c files
* checkpatch.pl fixes
* gpl v2 only correction
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-By: NeilBrown <neilb@suse.de>
The async_tx api tries to use a dma engine for an operation, but will fall
back to an optimized software routine otherwise. Xor support is
implemented using the raid5 xor routines. For organizational purposes this
routine is moved to a common area.
The following fixes are also made:
* rename xor_block => xor_blocks, suggested by Adrian Bunk
* ensure that xor.o initializes before md.o in the built-in case
* checkpatch.pl fixes
* mark calibrate_xor_blocks __init, Adrian Bunk
Cc: Adrian Bunk <bunk@stusta.de>
Cc: NeilBrown <neilb@suse.de>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
