On Wed, 2016-04-27 at 12:17 +0800, Herbert Xu wrote: > On Tue, Apr 26, 2016 at 01:42:56PM +0200, Ben Hutchings wrote: > > > > > > It looks like the bug was introduced in 3.10 by: > > > > d733ac90f9fe8ac284e523f9920b507555b12f6d > > Author: Jussi Kivilinna <jussi.kivilinna@xxxxxx> > > Date: Sun Apr 7 16:43:46 2013 +0300 > > > > crypto: gcm - fix rfc4543 to handle async crypto correctly > > > > So 3.2.y and 3.4.y don't need this fix - or should they get both > > fixes? > If that patch is not present then my fix can't be applied. However, > I think this change itself is probably needed in 3.2/3.4 as otherwise > GCM would be broken if the underlying cipher is async. It's not a > big deal on x86 because the main async AES provider also provides > GCM directly, but on other architectures it may be an issue. I've queued up both of these for 3.2. Ben. -- Ben Hutchings For every action, there is an equal and opposite criticism. - Harrison
Attachment:
signature.asc
Description: This is a digitally signed message part