On Thu, Sep 01, 2016 at 04:14:22PM +0100, Ard Biesheuvel wrote: > > The culprit appears to be that the .cra_name of the internal shash is > "ghash", (and not "__ghash" like in the x86 case) which causes the > test code to run the test on not only the public ahash, but also on > the internal cryptd() encapsulated shash, and also on the internal > shash itself. Hmm that's not good. The test code is not meant to process these internal algorithms. I recently added a check to automatically filter out internal algorithms when testing. But we should still change its name to __ghash just to avoid confusion. > However, that does not answer the question whether .init() must be > called before .import() [which the test code does not do]. If not, > then please disregard my v2, and I will followup with a patch that > renames ghash to __ghash (but .import() will still require the .init() > bits as well). Given that these internal shashes/ahashes are in fact > callable, and calling .import() will result in a crash, I suppose > duplicating some of the init() code in .import() makes sense > regardless. Right. This is definitely a bug and I will take your patches. Thanks, -- Email: Herbert Xu <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt -- To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html