On Thu, Oct 08, 2020 at 08:36:23AM +0200, Corentin Labbe wrote: > On Thu, Oct 08, 2020 at 04:52:38PM +1100, Herbert Xu wrote: > > On Thu, Sep 24, 2020 at 03:27:38PM +0200, Corentin Labbe wrote: > > > > > > This is an example on next-20200923+BigEndian > > > alg: ahash: sha1 test failed (wrong result) on test vector \"random: psize=194 ksize=0\", cfg=\"random: inplace may_sleep use_finup src_divs=[98.25%@+1124, <flush>1.75%@+5] iv_offset=18\" This failure is in one of the randomly generated test cases. If it doesn't reproduce reliably, you can set cryptomgr.fuzz_iterations=1000 on the kernel command line (increased from the default 100). It is confusing that it says just "sha1". This seems to be a quirk specific to how tcrypt calls alg_test(). It's probably really testing "sha1-sun4i-ss". I guess that testmgr.c should be using the actual cra_driver_name in the log messages, not the 'driver' string that was passed into alg_test(). - Eric