From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tom Lendacky Subject: Is a crypto_ahash_init required before invoking crypto_ahash_import? Date: Thu, 25 Feb 2016 15:56:31 -0600 Message-ID: <56CF788F.5050209@amd.com> Mime-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit To: , Herbert Xu Return-path: Received: from mail-by2on0084.outbound.protection.outlook.com ([207.46.100.84]:59648 "EHLO na01-by2-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1750830AbcBYV4h (ORCPT ); Thu, 25 Feb 2016 16:56:37 -0500 Sender: linux-crypto-owner@vger.kernel.org List-ID: I'm seeing an issue on one system that I wasn't seeing on another system. It turns out that the testmgr sha testing exports an ahash request context, allocates a new ahash request context and then imports into that new ahash request context. Since crypto_ahash_init() is not performed the driver request context could have random data in it, which ends up causing an error. As part of the import/export support that I added for the ccp driver I reduced the amount of data that was exported, but I guess I always assumed that crypto_ahash_init() would have been called before doing a crypto_ahash_import(). I can fix this in the driver by doing a memset to zero of the request context area during the import. But I guess I'm also wondering if there is an expectation/requirement that crypto_ahash_init() be called before doing an import? If there is the I can add that to the testmgr code instead. Thanks, Tom