Recent Linux kernel releases have seen a number of crypto performance optimizations
for this kernel subsystem by taking advantage of newer CPU instruction set extensions for accelerating various cryptographic workloads. This theme has continued for Linux 3.11.
Herbert Xu's crypto update for the Linux 3.11 kernel has been merged to mainline. Optiomizations this time around include more work on the AVX2 code-path for Camellia AES-NI acceleration. AVX2 is present in new Intel Haswell CPUs and will allow for faster performance for encrypted disks taking advantage of these new Crypto code-paths.
There's also been new SHA224 and SHA384 shims added for the widely-present SSSE3 instruction set extension. Meanwhile though, defunct Blowfish and Twofish AVX2 implementations have been dropped.
Other Crypto changes include not idling OMAP SoC devices between crypto operations happening in one session, unaligned buffer self-tests, a PCLMULQDQ optimization for CRCT10DIF. and support for FreeScale's DCP co-processor.
More details on these changes can be found via the Linux 3.11 Crypto pull request