Hi Correntin, On Mon, Feb 22, 2016 at 11:24 AM, LABBE Corentin <clabbe.montj...@gmail.com> wrote: > On Sun, Feb 21, 2016 at 07:24:36AM -0800, m.silentcr...@gmail.com wrote: >> Hi, >> >> Am Mittwoch, 17. Februar 2016 18:16:09 UTC+1 schrieb txsa...@gmail.com: >> > Hi >> > >> > I have a Cubieboard2 device, is Allwinner A20 SOC, running and Armbian >> > 5.00 with kernel 4.4.1. >> > >> > As in the latest version the sun4i-ss module is added, i tried the >> > hardware crypto acceleration with my 2 TB LUKS encrypted HDD, with 256bit >> > AES-CBC-PLAIN64 encryption. >> > >> > It was fast, but i noticed that when playing videos from that HDD, there >> > are random errors coming from somewhere. As the same happened when >> > streaming the video on DLNA, and SAMBA, i did some tests. >> > >> > As the HDD is also used for a backup storage, i compared the original >> > files with the backup on the encrypted HDD, and the most of the bigger >> > files were different. >> > >> > So i've unloaded sun4i-ss module with "rmmod sun4i_ss" and everything >> > become OK, no video errors, the backup and the original files are >> > identical. >> > >> > So there is something wrong with this module, as i see only the decryption >> > side is faulty, the encryption seems to be OK. >> > >> > Any suggestions? >> >> If there are any easy-to-setup tests I could run to help identify this >> issue, I'd be happy to run some tests on a spare Banana Pi I have. If there >> really is a bug in the sun4i-ss driver or in the accelerator itself that >> causes data corruption, I'd rather have it found as soon as possible. >> >> @Correntin: Is it possible to let your cryptotest testsuite run tests >> specifically for large data blocks? >> >> Regards, >> >> Timo > > Hello > > Changing the cryptotest is useless since cryptsetup/dm-crypt cipher data by > blocks of 512. > So the length of data changes only the statistic to hit the bug. > So I have writed a test (lukstest) which use cryptsetup, create files on a > cipherered partiton and checked it. > If you could test it for knowing if it find the issue. (Just uploaded it on > github). > > Actualy all my works is to try to reproduce this issue on my setup. > I also try to know if this thread > https://www.mail-archive.com/linux-crypto@vger.kernel.org/msg17926.html is > related. > > Regards >
Thanks, I installed your cryptotest suite including your new lukstest tool and ran some tests on a Banana Pi tonight. The results were all good. After a first successful run of lukstest, I modified the script and increased the sample size by a factor of 100 (and the image file, too). Even with the increased sample size, I didn't see any errors (because of the increased test duration, I ran only 5 iterations, though). Tests were done on an Armbian installation with Kernel 4.4.1 (with your patch "Add missing state size" on top of it) in order to resemble the original reporters conditions more closely. I was planning to do tests with 4.4.2 as well, but since all tests succeeded so far, I don't think that would change anything. Regards, Timo -- You received this message because you are subscribed to the Google Groups "linux-sunxi" group. To unsubscribe from this group and stop receiving emails from it, send an email to linux-sunxi+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
