Some engines have transfer size and address alignment restrictions. Add
a per-operation alignment property to struct dma_device that the async
routines and dmatest can use to check alignment capabilities.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Channel switching is problematic for some dmaengine drivers as the
architecture precludes separating the ->prep from ->submit. In these
cases the driver can select ASYNC_TX_DISABLE_CHANNEL_SWITCH to modify
the async_tx allocator to only return channels that support all of the
required asynchronous operations.
For example MD_RAID456=y selects support for asynchronous xor, xor
validate, pq, pq validate, and memcpy. When
ASYNC_TX_DISABLE_CHANNEL_SWITCH=y any channel with all these
capabilities is marked DMA_ASYNC_TX allowing async_tx_find_channel() to
quickly locate compatible channels with the guarantee that dependency
chains will remain on one channel. When
ASYNC_TX_DISABLE_CHANNEL_SWITCH=n async_tx_find_channel() may select
channels that lead to operation chains that need to cross channel
boundaries using the async_tx channel switch capability.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Some engines optimize operation by reading ahead in the descriptor chain
such that descriptor2 may start execution before descriptor1 completes.
If descriptor2 depends on the result from descriptor1 then a fence is
required (on descriptor2) to disable this optimization. The async_tx
api could implicitly identify dependencies via the 'depend_tx'
parameter, but that would constrain cases where the dependency chain
only specifies a completion order rather than a data dependency. So,
provide an ASYNC_TX_FENCE to explicitly identify data dependencies.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Port drivers/md/raid6test/test.c to use the async raid6 recovery
routines. This is meant as a unit test for raid6 acceleration drivers. In
addition to the 16-drive test case this implements tests for the 4-disk and
5-disk special cases (dma devices can not generically handle less than 2
sources), and adds a test for the D+Q case.
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
async_raid6_2data_recov() recovers two data disk failures
async_raid6_datap_recov() recovers a data disk and the P disk
These routines are a port of the synchronous versions found in
drivers/md/raid6recov.c. The primary difference is breaking out the xor
operations into separate calls to async_xor. Two helper routines are
introduced to perform scalar multiplication where needed.
async_sum_product() multiplies two sources by scalar coefficients and
then sums (xor) the result. async_mult() simply multiplies a single
source by a scalar.
This implemention also includes, in contrast to the original
synchronous-only code, special case handling for the 4-disk and 5-disk
array cases. In these situations the default N-disk algorithm will
present 0-source or 1-source operations to dma devices. To cover for
dma devices where the minimum source count is 2 we implement 4-disk and
5-disk handling in the recovery code.
[ Impact: asynchronous raid6 recovery routines for 2data and datap cases ]
Cc: Yuri Tikhonov <yur@emcraft.com>
Cc: Ilya Yanok <yanok@emcraft.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: David Woodhouse <David.Woodhouse@intel.com>
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
[ Based on an original patch by Yuri Tikhonov ]
This adds support for doing asynchronous GF multiplication by adding
two additional functions to the async_tx API:
async_gen_syndrome() does simultaneous XOR and Galois field
multiplication of sources.
async_syndrome_val() validates the given source buffers against known P
and Q values.
When a request is made to run async_pq against more than the hardware
maximum number of supported sources we need to reuse the previous
generated P and Q values as sources into the next operation. Care must
be taken to remove Q from P' and P from Q'. For example to perform a 5
source pq op with hardware that only supports 4 sources at a time the
following approach is taken:
p, q = PQ(src0, src1, src2, src3, COEF({01}, {02}, {04}, {08}))
p', q' = PQ(p, q, q, src4, COEF({00}, {01}, {00}, {10}))
p' = p + q + q + src4 = p + src4
q' = {00}*p + {01}*q + {00}*q + {10}*src4 = q + {10}*src4
Note: 4 is the minimum acceptable maxpq otherwise we punt to
synchronous-software path.
The DMA_PREP_CONTINUE flag indicates to the driver to reuse p and q as
sources (in the above manner) and fill the remaining slots up to maxpq
with the new sources/coefficients.
Note1: Some devices have native support for P+Q continuation and can skip
this extra work. Devices with this capability can advertise it with
dma_set_maxpq. It is up to each driver how to handle the
DMA_PREP_CONTINUE flag.
Note2: The api supports disabling the generation of P when generating Q,
this is ignored by the synchronous path but is implemented by some dma
devices to save unnecessary writes. In this case the continuation
algorithm is simplified to only reuse Q as a source.
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: David Woodhouse <David.Woodhouse@intel.com>
Signed-off-by: Yuri Tikhonov <yur@emcraft.com>
Signed-off-by: Ilya Yanok <yanok@emcraft.com>
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
We currently walk the parent chain when waiting for a given tx to
complete however this walk may race with the driver cleanup routine.
The routines in async_raid6_recov.c may fall back to the synchronous
path at any point so we need to be prepared to call async_tx_quiesce()
(which calls dma_wait_for_async_tx). To remove the ->parent walk we
guarantee that every time a dependency is attached ->issue_pending() is
invoked, then we can simply poll the initial descriptor until
completion.
This also allows for a lighter weight 'issue pending' implementation as
there is no longer a requirement to iterate through all the channels'
->issue_pending() routines as long as operations have been submitted in
an ordered chain. async_tx_issue_pending() is added for this case.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
If module_init and module_exit are nops then neither need to be defined.
[ Impact: pure cleanup ]
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Replace the flat zero_sum_result with a collection of flags to contain
the P (xor) zero-sum result, and the soon to be utilized Q (raid6 reed
solomon syndrome) zero-sum result. Use the SUM_CHECK_ namespace instead
of DMA_ since these flags will be used on non-dma-zero-sum enabled
platforms.
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
async_xor() needs space to perform dma and page address conversions. In
most cases the code can simply reuse the struct page * array because the
size of the native pointer matches the size of a dma/page address. In
order to support archs where sizeof(dma_addr_t) is larger than
sizeof(struct page *), or to preserve the input parameters, we utilize a
memory region passed in by the caller.
Since the code is now prepared to handle the case where it cannot
perform address conversions on the stack, we no longer need the
!HIGHMEM64G dependency in drivers/dma/Kconfig.
[ Impact: don't clobber input buffers for address conversions ]
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Prepare the api for the arrival of a new parameter, 'scribble'. This
will allow callers to identify scratchpad memory for dma address or page
address conversions. As this adds yet another parameter, take this
opportunity to convert the common submission parameters (flags,
dependency, callback, and callback argument) into an object that is
passed by reference.
Also, take this opportunity to fix up the kerneldoc and add notes about
the relevant ASYNC_TX_* flags for each routine.
[ Impact: moves api pass-by-value parameters to a pass-by-reference struct ]
Signed-off-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
In support of inter-channel chaining async_tx utilizes an ack flag to
gate whether a dependent operation can be chained to another. While the
flag is not set the chain can be considered open for appending. Setting
the ack flag closes the chain and flags the descriptor for garbage
collection. The ASYNC_TX_DEP_ACK flag essentially means "close the
chain after adding this dependency". Since each operation can only have
one child the api now implicitly sets the ack flag at dependency
submission time. This removes an unnecessary management burden from
clients of the api.
[ Impact: clean up and enforce one dependency per operation ]
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
A quirk that we've always supported is having an sg entry that's
bigger than a page, or more generally an sg entry that crosses
page boundaries. Even though it would be better to explicitly have
to sg entries for this, we need to support it for the existing users,
in particular, IPsec.
The new ahash sg walking code did try to handle this, but there was
a bug where we didn't increment the page so kept on walking on the
first page over an dover again.
This patch fixes it.
Tested-by: Martin Willi <martin@strongswan.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
crypto: padlock - Revert aes-all alias to aes
crypto: api - Fix algorithm module auto-loading
crypto: eseqiv - Fix IV generation for sync algorithms
crypto: ixp4xx - check firmware for crypto support
The commit a760a6656e (crypto:
api - Fix module load deadlock with fallback algorithms) broke
the auto-loading of algorithms that require fallbacks. The
problem is that the fallback mask check is missing an and which
cauess bits that should be considered to interfere with the
result.
Reported-by: Chuck Ebbert <cebbert@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
If crypto_ablkcipher_encrypt() returns synchronous,
eseqiv_complete2() is called even if req->giv is already the
pointer to the generated IV. The generated IV is overwritten
with some random data in this case. This patch fixes this by
calling eseqiv_complete2() just if the generated IV has to be
copied to req->giv.
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
'zero_sum' does not properly describe the operation of generating parity
and checking that it validates against an existing buffer. Change the
name of the operation to 'val' (for 'validate'). This is in
anticipation of the p+q case where it is a requirement to identify the
target parity buffers separately from the source buffers, because the
target parity buffers will not have corresponding pq coefficients.
Reviewed-by: Andre Noll <maan@systemlinux.org>
Acked-by: Maciej Sosnowski <maciej.sosnowski@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
dma: Add SoF and EoF debugging to ipu_idmac.c, minor cleanup
dw_dmac: add cyclic API to DW DMA driver
dmaengine: Add privatecnt to revert DMA_PRIVATE property
dmatest: add dma interrupts and callbacks
dmatest: add xor test
dmaengine: allow dma support for async_tx to be toggled
async_tx: provide __async_inline for HAS_DMA=n archs
dmaengine: kill some unused headers
dmaengine: initialize tx_list in dma_async_tx_descriptor_init
dma: i.MX31 IPU DMA robustness improvements
dma: improve section assignment in i.MX31 IPU DMA driver
dma: ipu_idmac driver cosmetic clean-up
dmaengine: fail device registration if channel registration fails
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
crypto: ixp4xx - Fix handling of chained sg buffers
crypto: shash - Fix unaligned calculation with short length
hwrng: timeriomem - Use phys address rather than virt
* 'for-linus' of git://neil.brown.name/md: (53 commits)
md/raid5 revise rules for when to update metadata during reshape
md/raid5: minor code cleanups in make_request.
md: remove CONFIG_MD_RAID_RESHAPE config option.
md/raid5: be more careful about write ordering when reshaping.
md: don't display meaningless values in sysfs files resync_start and sync_speed
md/raid5: allow layout and chunksize to be changed on active array.
md/raid5: reshape using largest of old and new chunk size
md/raid5: prepare for allowing reshape to change layout
md/raid5: prepare for allowing reshape to change chunksize.
md/raid5: clearly differentiate 'before' and 'after' stripes during reshape.
Documentation/md.txt update
md: allow number of drives in raid5 to be reduced
md/raid5: change reshape-progress measurement to cope with reshaping backwards.
md: add explicit method to signal the end of a reshape.
md/raid5: enhance raid5_size to work correctly with negative delta_disks
md/raid5: drop qd_idx from r6_state
md/raid6: move raid6 data processing to raid6_pq.ko
md: raid5 run(): Fix max_degraded for raid level 4.
md: 'array_size' sysfs attribute
md: centralize ->array_sectors modifications
...
This makes the includes more explicit, and is preparation for moving
md_k.h to drivers/md/md.h
Remove include/raid/md.h as its only remaining use was to #include
other files.
Signed-off-by: NeilBrown <neilb@suse.de>
When the total length is shorter than the calculated number of unaligned bytes, the call to shash->update breaks. For example, calling crc32c on unaligned buffer with length of 1 can result in a system crash.
Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (29 commits)
crypto: sha512-s390 - Add missing block size
hwrng: timeriomem - Breaks an allyesconfig build on s390:
nlattr: Fix build error with NET off
crypto: testmgr - add zlib test
crypto: zlib - New zlib crypto module, using pcomp
crypto: testmgr - Add support for the pcomp interface
crypto: compress - Add pcomp interface
netlink: Move netlink attribute parsing support to lib
crypto: Fix dead links
hwrng: timeriomem - New driver
crypto: chainiv - Use kcrypto_wq instead of keventd_wq
crypto: cryptd - Per-CPU thread implementation based on kcrypto_wq
crypto: api - Use dedicated workqueue for crypto subsystem
crypto: testmgr - Test skciphers with no IVs
crypto: aead - Avoid infinite loop when nivaead fails selftest
crypto: skcipher - Avoid infinite loop when cipher fails selftest
crypto: api - Fix crypto_alloc_tfm/create_create_tfm return convention
crypto: api - crypto_alg_mod_lookup either tested or untested
crypto: amcc - Add crypt4xx driver
crypto: ansi_cprng - Add maintainer
...
To allow an async_tx routine to be compiled away on HAS_DMA=n arch it
needs to be declared __always_inline otherwise the compiler may emit
code and cause a link error.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
The current "comp" crypto interface supports one-shot (de)compression only,
i.e. the whole data buffer to be (de)compressed must be passed at once, and
the whole (de)compressed data buffer will be received at once.
In several use-cases (e.g. compressed file systems that store files in big
compressed blocks), this workflow is not suitable.
Furthermore, the "comp" type doesn't provide for the configuration of
(de)compression parameters, and always allocates workspace memory for both
compression and decompression, which may waste memory.
To solve this, add a "pcomp" partial (de)compression interface that provides
the following operations:
- crypto_compress_{init,update,final}() for compression,
- crypto_decompress_{init,update,final}() for decompression,
- crypto_{,de}compress_setup(), to configure (de)compression parameters
(incl. allocating workspace memory).
The (de)compression methods take a struct comp_request, which was mimicked
after the z_stream object in zlib, and contains buffer pointer and length
pairs for input and output.
The setup methods take an opaque parameter pointer and length pair. Parameters
are supposed to be encoded using netlink attributes, whose meanings depend on
the actual (name of the) (de)compression algorithm.
Signed-off-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
With the mandatory algorithm testing at registration, we have
now created a deadlock with algorithms requiring fallbacks.
This can happen if the module containing the algorithm requiring
fallback is loaded first, without the fallback module being loaded
first. The system will then try to test the new algorithm, find
that it needs to load a fallback, and then try to load that.
As both algorithms share the same module alias, it can attempt
to load the original algorithm again and block indefinitely.
As algorithms requiring fallbacks are a special case, we can fix
this by giving them a different module alias than the rest. Then
it's just a matter of using the right aliases according to what
algorithms we're trying to find.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
crypto_ahash_show changed to use cra_ahash for digestsize reference.
Signed-off-by: Lee Nipper <lee.nipper@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
keventd_wq has potential starvation problem, so use dedicated
kcrypto_wq instead.
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Use dedicated workqueue for crypto subsystem
A dedicated workqueue named kcrypto_wq is created to be used by crypto
subsystem. The system shared keventd_wq is not suitable for
encryption/decryption, because of potential starvation problem.
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As it is an skcipher with no IV escapes testing altogether because
we only test givcipher objects. This patch fixes the bypass logic
to test these algorithms.
Conversely, we're currently testing nivaead algorithms with IVs,
which would have deadlocked had it not been for the fact that no
nivaead algorithms have any test vectors. This patch also fixes
that case.
Both fixes are ugly as hell, but this ugliness should hopefully
disappear once we move them into the per-type code (i.e., the
AEAD test would live in aead.c and the skcipher stuff in ablkcipher.c).
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When an aead constructed through crypto_nivaead_default fails
its selftest, we'll loop forever trying to construct new aead
objects but failing because it already exists.
The crux of the issue is that once an aead fails the selftest,
we'll ignore it on the next run through crypto_aead_lookup and
attempt to construct a new aead.
We should instead return an error to the caller if we find an
an that has failed the test.
This bug hasn't manifested itself yet because we don't have any
test vectors for the existing nivaead algorithms. They're tested
through the underlying algorithms only.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
When an skcipher constructed through crypto_givcipher_default fails
its selftest, we'll loop forever trying to construct new skcipher
objects but failing because it already exists.
The crux of the issue is that once a givcipher fails the selftest,
we'll ignore it on the next run through crypto_skcipher_lookup and
attempt to construct a new givcipher.
We should instead return an error to the caller if we find a
givcipher that has failed the test.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This is based on a report and patch by Geert Uytterhoeven.
The functions crypto_alloc_tfm and create_create_tfm return a
pointer that needs to be adjusted by the caller when successful
and otherwise an error value. This means that the caller has
to check for the error and only perform the adjustment if the
pointer returned is valid.
Since all callers want to make the adjustment and we know how
to adjust it ourselves, it's much easier to just return adjusted
pointer directly.
The only caveat is that we have to return a void * instead of
struct crypto_tfm *. However, this isn't that bad because both
of these functions are for internal use only (by types code like
shash.c, not even algorithms code).
This patch also moves crypto_alloc_tfm into crypto/internal.h
(crypto_create_tfm is already there) to reflect this.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
As it stands crypto_alg_mod_lookup will search either tested or
untested algorithms, but never both at the same time. However,
we need exactly that when constructing givcipher and aead so
this patch adds support for that by setting the tested bit in
type but clearing it in mask. This combination is currently
unused.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
FIPS 140-2 specifies that all access to various cryptographic modules be
prevented in the event that any of the provided self tests fail on the various
implemented algorithms. We already panic when any of the test in testmgr.c
fail when we are operating in fips mode. The continuous test in the cprng here
was missed when that was implmented. This code simply checks for the
fips_enabled flag if the test fails, and warns us via syslog or panics the box
accordingly.
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Pseudo RNGs provide predictable outputs based on input parateters {key, V, DT},
the idea behind them is that only the user should know what the inputs are.
While its nice to have default known values for testing purposes, it seems
dangerous to allow the use of those default values without some sort of safety
measure in place, lest an attacker easily guess the output of the cprng. This
patch forces the NEED_RESET flag on when allocating a cprng context, so that any
user is forced to reseed it before use. The defaults can still be used for
testing, but this will prevent their inadvertent use, and be more secure.
Signed-off-by: Neil Horman <nhorman@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Intel AES-NI is a new set of Single Instruction Multiple Data (SIMD)
instructions that are going to be introduced in the next generation of
Intel processor, as of 2009. These instructions enable fast and secure
data encryption and decryption, using the Advanced Encryption Standard
(AES), defined by FIPS Publication number 197. The architecture
introduces six instructions that offer full hardware support for
AES. Four of them support high performance data encryption and
decryption, and the other two instructions support the AES key
expansion procedure.
The white paper can be downloaded from:
http://softwarecommunity.intel.com/isn/downloads/intelavx/AES-Instructions-Set_WP.pdf
AES may be used in soft_irq context, but MMX/SSE context can not be
touched safely in soft_irq context. So in_interrupt() is checked, if
in IRQ or soft_irq context, the general x86_64 implementation are used
instead.
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
cryptd_alloc_ablkcipher() will allocate a cryptd-ed ablkcipher for
specified algorithm name. The new allocated one is guaranteed to be
cryptd-ed ablkcipher, so the blkcipher underlying can be gotten via
cryptd_ablkcipher_child().
Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
We're currently checking the frontend type in init_tfm. This is
completely pointless because the fact that we're called at all
means that the frontend is ours so the type must match as well.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
It turns out that LRW has never worked properly on big endian.
This was never discussed because nobody actually used it that
way. In fact, it was only discovered when Geert Uytterhoeven
loaded it through tcrypt which failed the test on it.
The fix is straightforward, on big endian the to find the nth
bit we should be grouping them by words instead of bytes. So
setbit128_bbe should xor with 128 - BITS_PER_LONG instead of
128 - BITS_PER_BYTE == 0x78.
Tested-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
It's illegal to call flush_dcache_page on slab pages on a number
of architectures. So this patch avoids doing so if PageSlab is
true.
In future we can move the flush_dcache_page call to those page
cache users that actually need it.
Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Geert Uytterhoeven pointed out that we're not zeroing all the
memory when freeing a transform. This patch fixes it by calling
ksize to ensure that we zero everything in sight.
Reported-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
