checkstack reports report_deco_status(), report_ccb_status() as
particularly excessive stack users. Move their lookup tables
off the stack and put them in .rodata.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This change is required for post SEC-5.0 devices which have RNG4.
Setting RDB in security configuration register allows CAAM to use the
"Random Data Buffer" to be filled by a single request. The Random Data
Buffer is large enough for ten packets to get their IVs from a single
request. If the Random Data Buffer is not enabled, then each IV causes a
separate request, and RNG4 hardware cannot keep up resulting in lower
IPSEC throughput if random IVs are used.
Signed-off-by: Vakul Garg <vakul@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The Kconfig entry for CAAM's hash algorithm implementations has always
selected CRYPTO_AHASH. But there's no corresponding Kconfig symbol.
It seems it was intended to select CRYPTO_HASH, like other crypto
drivers do. That would apparently (indirectly) select CRYPTO_HASH2,
which would enable the ahash functionality this driver uses.
Signed-off-by: Paul Bolle <pebolle@tiscali.nl>
Reviewed-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Pull crypto update from Herbert Xu:
- Optimised AES/SHA1 for ARM.
- IPsec ESN support in talitos and caam.
- x86_64/avx implementation of cast5/cast6.
- Add/use multi-algorithm registration helpers where possible.
- Added IBM Power7+ in-Nest support.
- Misc fixes.
Fix up trivial conflicts in crypto/Kconfig due to the sparc64 crypto
config options being added next to the new ARM ones.
[ Side note: cut-and-paste duplicate help texts make those conflicts
harder to read than necessary, thanks to git being smart about
minimizing conflicts and maximizing the common parts... ]
* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (71 commits)
crypto: x86/glue_helper - fix storing of new IV in CBC encryption
crypto: cast5/avx - fix storing of new IV in CBC encryption
crypto: tcrypt - add missing tests for camellia and ghash
crypto: testmgr - make test_aead also test 'dst != src' code paths
crypto: testmgr - make test_skcipher also test 'dst != src' code paths
crypto: testmgr - add test vectors for CTR mode IV increasement
crypto: testmgr - add test vectors for partial ctr(cast5) and ctr(cast6)
crypto: testmgr - allow non-multi page and multi page skcipher tests from same test template
crypto: caam - increase TRNG clocks per sample
crypto, tcrypt: remove local_bh_disable/enable() around local_irq_disable/enable()
crypto: tegra-aes - fix error return code
crypto: crypto4xx - fix error return code
crypto: hifn_795x - fix error return code
crypto: ux500 - fix error return code
crypto: caam - fix error IDs for SEC v5.x RNG4
hwrng: mxc-rnga - Access data via structure
hwrng: mxc-rnga - Adapt clocks to new i.mx clock framework
crypto: caam - add IPsec ESN support
crypto: 842 - remove .cra_list initialization
Revert "[CRYPTO] cast6: inline bloat--"
...
we need to configure the TRNG to use more clocks per sample
to handle the two back-to-back 64KiB random descriptor requests
on higher frequency P5040s.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Support for ESNs (extended sequence numbers).
Tested with strongswan by connecting back-to-back P1010RDB with P2020RDB.
Signed-off-by: Horia Geanta <horia.geanta@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In 3.6-rc3, without this patch, the following error occurs with a modular build:
ERROR: "gen_split_key" [drivers/crypto/caam/caamhash.ko] undefined!
ERROR: "gen_split_key" [drivers/crypto/caam/caamalg.ko] undefined!
Signed-off-by: Ben Collins <ben.c@servergy.com>
Cc: Yuan Kang <Yuan.Kang@freescale.com>
Acked-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
SHARE_WAIT, whilst more optimal for association-less crypto,
has the ability to start thrashing the CCB descriptor/key
caches, given high levels of traffic across multiple security
associations (and thus keys).
Switch to using the SERIAL sharing type, which prefers
the last used CCB for the SA. On a 2-DECO platform
such as the P3041, this can improve performance by
about 3.7%.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
In some device trees of previous version, there were string "fsl,sec4.0".
To be backward compatible with device trees, we first check "fsl,sec-v4.0",
if it fails, then check for "fsl,sec4.0".
Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
extended to include new hash and rng code, which was omitted from
the previous version of this patch during a rebase of the SDK
version.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
This patch adds support for retrieving and printing of
SEC ERA information. It is useful for knowing beforehand
what features exist from the SEC point of view on a
certain SoC. Only era-s 1 to 4 are currently supported;
other eras will appear as unknown.
Signed-off-by: Alex Porosanu <alexandru.porosanu@freescale.com>
- rebased onto current cryptodev master
- made caam_eras static
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The caam job rings (input/output job ring) are allocated using
dma_map_single(). These job rings can be visualized as the ring
buffers in which the jobs are en-queued/de-queued. The s/w enqueues
the jobs in input job ring which h/w dequeues and after processing
it copies the jobs in output job ring. Software then de-queues the
job from output ring. Using dma_map/unmap_single() is not preferred
way to allocate memory for this type of requirements because this
adds un-necessary complexity.
Example, if bounce buffer (SWIOTLB) will get used then to make any
change visible in this memory to other processing unit requires
dmap_unmap_single() or dma_sync_single_for_cpu/device(). The
dma_unmap_single() can not be used as this will free the bounce
buffer, this will require changing the job rings on running system
and I seriously doubt that it will be not possible or very complex
to implement. Also using dma_sync_single_for_cpu/device() will also
add unnecessary complexity.
The simple and preferred way is using dma_alloc_coherent() for these
type of memory requirements.
This resolves the Linux boot crash issue when "swiotlb=force" is set
in bootargs on systems which have memory more than 4G.
Signed-off-by: Bharat Bhushan <bharat.bhushan@freescale.com>
Acked-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
there is no noticeable benefit for multiple cores to process one
job ring's output ring: in fact, we can benefit from cache effects
of having the back-half stay on the core that receives a particular
ring's interrupts, and further relax general contention and the
locking involved with reading outring_used, since tasklets run
atomically.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Memory barriers are implied by the i/o register write implementation
(at least on Power). So we can remove the redundant wmb() in
caam_jr_enqueue, and, in dequeue(), hoist the h/w done notification
write up to before we need to increment the head of the ring, and
save an smp_mb.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Code was needlessly checking the s/w job ring when there
would be nothing to process if the h/w's output completion
ring were empty anyway.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The enqueue lock isn't used in any interrupt context, and
the dequeue lock isn't used in the h/w interrupt context,
only in bh context.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
It has been observed that in zero-loss benchmarks, when a
slow traffic rate is being tested, the IRQ timer coalescing
parameter was set too high, and the ethernet controller
would start dropping packets because the job ring back half
wouldn't be executed in time before the ethernet controller
would fill its buffers, thereby significantly reducing the
zero-loss performance figures.
Empirical testing has shown that the best zero-loss performance
is achieved when IRQ coalescing is set to minimum values and/or
turned off, since apparently the job ring driver already implements
an adequately-performing general-purpose IRQ mitigation strategy
in software.
Whilst we could go with minimal count (2-8) and timing settings
(192-256), we prefer just turning h/w coalescing altogether off
to minimize setkey latency (due to split key generation), and
for consistent cross-SoC performance (the SEC vs. core clock
ratio changes).
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
The SEC v4.x' RNGB h/w block self-initialized. RNG4, available
on SEC versions 5 and beyond, is based on a different standard
that requires manual initialization.
Also update any new errors From the SEC v5.2 reference manual:
The SEC v5.2's RNG4 unit reuses some error IDs, thus the addition
of rng_err_id_list over the CHA-independent err_id_list.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
SEC v4.x were only 36-bit, SEC v5+ are 40-bit capable.
Also set a DMA mask for any job ring devices created.
Signed-off-by: Kim Phillips <kim.phillips@freescale.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Kim Phillips
Vakul Garg
Paul Bolle
Greg Kroah-Hartman
Linus Torvalds
Horia Geanta
Ben Collins
Shengzhou Liu
Alex Porosanu
Bharat Bhushan