From: Colin Ian King <colin.king@xxxxxxxxxxxxx> This is a forward port of Thieu Le's patch from 2.6.39 to migrate to using the ablkcipher API for eCryptfs. Performance Improvements: I've instrumented this patch to measure TSC ticks per 4K encrypt and decrypt operations to see how this patch compares to the original code using the default AES generic crypto engine as well as the new Intel AES-NI instruction capable crypto engine on an Ivybridge i7-3770. Patched: TSC ticks for 4K TSC ticks per byte AES-Generic Read: 5843.5 1.42 AES-Generic Write: 19295.8 4.71 AES-NI Read: 5677.0 1.39 AES-NI Write: 19257.9 4.70 Unpatched: AES-Generic Read: 92861.5 22.67 AES-Generic Write: 93642 22.61 AES-NI Read: 91610.2 22.37 AES-NI Write: 93659.2 22.87 ..so at the crypto engine stage we see some considerable speed improvement with the patch. I've also run some simple benchmarking tests comparing this patch with the un-patched kernel on a variety of machines (Ivybridge, Sandybridge, Atom) and drives (HDD, SSD) to see how well it performs. A LibreOffice speadsheet of the test result data and a write-up are available: http://kernel.ubuntu.com/~cking/ecryptfs-async-testing/async-patch-results-1.ods http://kernel.ubuntu.com/~cking/ecryptfs-async-testing/async-patch-summary.txt Soak Testing: * many kernel builds using -j 64, on HDD and SSD * eCryptfs tests with lower filesystems: ext2, ext3, ext4, xfs, btrfs * exercised on a 4 CPU (+hyperthreaded) build machine * bonnie++ and tiobench tests Colin Ian King (1): ecryptfs: Migrate to ablkcipher API fs/ecryptfs/crypto.c | 678 +++++++++++++++++++++++++++++++---------- fs/ecryptfs/ecryptfs_kernel.h | 38 ++- fs/ecryptfs/main.c | 10 + fs/ecryptfs/mmap.c | 87 +++++- 4 files changed, 636 insertions(+), 177 deletions(-) -- 1.7.9.5 -- To unsubscribe from this list: send the line "unsubscribe ecryptfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
