ioat3.2 does not support asynchronous error notifications which makes
the driver experience latencies when non-zero pq validate results are
expected.  Provide a mechanism for turning off async_xor_val and
async_syndrome_val via Kconfig.  This approach is generally useful for
any driver that specifies ASYNC_TX_DISABLE_CHANNEL_SWITCH and would like
to force the async_tx api to fall back to the synchronous path for
certain operations.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

The flow of the complete function (xxx_done) in gcm.c is as follow:

void complete(struct crypto_async_request *areq, int err)
{
	struct aead_request *req = areq->data;

	if (!err) {
		err = async_next_step();
		if (err == -EINPROGRESS || err == -EBUSY)
			return;
	}

	complete_for_next_step(areq, err);
}

But *areq may be destroyed in async_next_step(), this makes
complete_for_next_step() can not work properly. To fix this, one of
following methods is used for each complete function.

- Add a __complete() for each complete(), which accept struct
  aead_request *req instead of areq, so avoid using areq after it is
  destroyed.

- Expand complete_for_next_step().

The fixing method is based on the idea of Herbert Xu.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The raid6 recovery code currently requires special handling of the
4-disk and 5-disk recovery scenarios for the native layout.  Quoting
from commit 0a82a623:

     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.

The ddf layout presents disks=6 and disks=7 to the recovery code in
these situations.  Instead of looking at the number of disks count the
number of non-zero sources in the list and call the special case code
when the number of non-failed sources is 0 or 1.

[neilb@suse.de: replace 'ddf' flag with counting good sources]
Signed-off-by: Dan Williams <dan.j.williams@intel.com>

The global scribble page is used as a temporary destination buffer when
disabling the P or Q result is requested.  The local scribble buffer
contains memory for performing address conversions.  Rename the global
variable to avoid confusion.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

- update the kernel doc for async_syndrome to indicate what NULL in the
  source list means
- whitespace fixups

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

async_syndrome_val check the P and Q blocks used for RAID6
calculations.
With DDF raid6, some of the data blocks might be NULL, so
this needs to be handled in the same way that async_gen_syndrome
handles it.

As async_syndrome_val calls async_xor, also enhance async_xor
to detect and skip NULL blocks in the list.

Signed-off-by: NeilBrown <neilb@suse.de>

md/raid6 passes a list of 'struct page *' to the async_tx routines,
which then either DMA map them for offload, or take the page_address
for CPU based calculations.

For RAID6 we sometime leave 'blanks' in the list of pages.
For CPU based calcs, we want to treat theses as a page of zeros.
For offloaded calculations, we simply don't pass a page to the
hardware.

Currently the 'blanks' are encoded as a pointer to
raid6_empty_zero_page.  This is a 4096 byte memory region, not a
'struct page'.  This is mostly handled correctly but is rather ugly.

So change the code to pass and expect a NULL pointer for the blanks.
When taking page_address of a page, we need to check for a NULL and
in that case use raid6_empty_zero_page.

Signed-off-by: NeilBrown <neilb@suse.de>

After m68k's task_thread_info() doesn't refer to current,
it's possible to remove sched.h from interrupt.h and not break m68k!
Many thanks to Heiko Carstens for allowing this.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>

If we are unable to offload async_mult() or async_sum_product(), then
unmap the buffers before falling through to the synchronous path.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

Testing on x86_64 with NDISKS=255 yields:

   do_IRQ: modprobe near stack overflow (cur:ffff88007d19c000,sp:ffff88007d19c128)

...and eventually

   general protection fault: 0000 [#1]

Moving the scribble buffers off the stack allows the test to complete
successfully.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

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>

This patch adds VMAC (a fast MAC) support into crypto framework.

Signed-off-by: Shane Wang <shane.wang@intel.com>
Signed-off-by: Joseph Cihula <joseph.cihula@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

We have a mechanism where newly registered algorithms of a higher
priority can displace existing instances that use a different
implementation of the same algorithm with a lower priority.

Unfortunately the same mechanism can cause a newly registered
algorithm to displace itself if it depends on an existing version
of the same algorithm.

This patch fixes this by keeping all algorithms that the newly
reigstered algorithm depends on, thus protecting them from being
removed.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

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>

As struct skcipher_givcrypt_request includes struct crypto_request
at a non-zero offset, testing for NULL after converting the pointer
returned by crypto_dequeue_request does not work.  This can result
in IPsec crashes when the queue is depleted.

This patch fixes it by doing the pointer conversion only when the
return value is non-NULL.  In particular, we create a new function
__crypto_dequeue_request that does the pointer conversion.

Reported-by: Brad Bosch <bradbosch@comcast.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Return the value we got from crypto_register_alg() instead of
returning 0 in any case.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

The alignment calculation of xcbc_tfm_ctx uses alg->cra_alignmask
and not alg->cra_alignmask + 1 as it should. This led to frequent
crashes during the selftest of xcbc(aes-asm) on x86_64
machines. This patch fixes this. Also we use the alignmask
of xcbc and not the alignmask of the underlying algorithm
for the alignmnent calculation in xcbc_create now.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

What about something like this?  It defaults the CPRNG to m and makes FIPS
dependent on the CPRNG.  That way you get a module build by default, but you can
change it to y manually during config and still satisfy the dependency, and if
you select N it disables FIPS as well.  I rather like that better than making
FIPS a tristate.  I just tested it out here and it seems to work well.  Let me
know what you think

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Recently we switched to using eseqiv on SMP machines in preference
over chainiv.  However, eseqiv does not support stream ciphers so
they should still default to chainiv.

This patch applies the same check as done by eseqiv to weed out
the stream ciphers.  In particular, all algorithms where the IV
size is not equal to the block size will now default to chainiv.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Raw counter mode only works with chainiv, which is no longer
the default IV generator on SMP machines.  This broke raw counter
mode as it can no longer instantiate as a givcipher.

This patch fixes it by always picking chainiv on raw counter
mode.  This is based on the diagnosis and a patch by Huang
Ying.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This reverts commit 215ccd6f.

It causes CPRNG and everything selected by it to be built-in
whenever FIPS is enabled.  The problem is that it is selecting
a tristate from a bool, which is usually not what is intended.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Correct a typo in crypto/rng.c

Signed-off-by: Christian Kujau <lists@nerdbynature.de>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

cryptd_alloc_ahash() will allocate a cryptd-ed ahash for specified
algorithm name. The new allocated one is guaranteed to be cryptd-ed
ahash, so the shash underlying can be gotten via cryptd_ahash_child().

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

Remove the dedicated GHASH implementation in GCM, and uses the GHASH
digest algorithm instead. This will make GCM uses hardware accelerated
GHASH implementation automatically if available.

ahash instead of shash interface is used, because some hardware
accelerated GHASH implementation needs asynchronous interface.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

GHASH is implemented as a shash algorithm. The actual implementation
is copied from gcm.c. This makes it possible to add
architecture/hardware accelerated GHASH implementation.

Signed-off-by: Huang Ying <ying.huang@intel.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch converts authenc to the new ahash interface.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch uses crypto_shash_export/crypto_shash_import to prehash
ipad/opad to speed up hmac.  This is partly based on a similar patch
by Steffen Klassert.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

It's undefined behaviour in C to write outside the bounds of an array.
The key expansion routine takes a shortcut of creating 8 words at a
time, but this creates 4 additional words which don't fit in the array.

As everyone is hopefully now aware, GCC is at liberty to make any
assumptions and optimisations it likes in situations where it can
detect that UB has occured, up to and including nasal demons, and
as the indices being accessed in the array are trivially calculable,
it's rash to invite gcc to do take any liberties at all.

Signed-off-by: Phil Carmody <ext-phil.2.carmody@nokia.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

crypto_init_shash_ops_async() tests for setkey and not for import
before exporting the algorithms import function to ahash.
This patch fixes this.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

ahash_op_unaligned() and ahash_def_finup() allocate memory atomically,
regardless whether the request can sleep or not. This patch changes
this to use GFP_KERNEL if the request can sleep.

Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch provides a default export/import function for all
shash algorithms.  It simply copies the descriptor context as
is done by sha1_generic.

This in essence means that all existing shash algorithms now
support export/import.  This is something that will be depended
upon in implementations such as hmac.  Therefore all new shash
and ahash implementations must support export/import.

For those that cannot obtain a partial result, padlock-sha's
fallback model should be used so that a partial result is always
available.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch replaces the 32-bit counters in sha512_generic with
64-bit counters.  It also switches the bit count to the simpler
byte count.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

This patch renames struct sha512_ctx and exports it as struct
sha512_state so that other sha512 implementations can use it
as the reference structure for exporting their state.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>