[2.6.32 regression] lzo fails crypto self-tests

[2.6.32 regression] lzo fails crypto self-tests

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 524 bytes --]

Linux 2.6.32.64 included an update to the lzo compression code.  So far
as I know, there is nothing wrong with it, but it no longer matches the
test vectors in crypto/testmgr.  Please cherry-pick this fix:

commit 0ec7382036922be063b515b2a3f1d6f7a607392c
Author: Markus F.X.J. Oberhumer <markus@oberhumer.com>
Date:   Sun Oct 14 15:39:04 2012 +0200

    crypto: testmgr - update LZO compression test vectors

Ben.

-- 
Ben Hutchings
One of the nice things about standards is that there are so many of them.


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]

Re: [2.6.32 regression] lzo fails crypto self-tests

[Willy Tarreau]

Hi Ben,

On Sat, Aug 01, 2015 at 06:33:34PM +0100, Ben Hutchings wrote:
> Linux 2.6.32.64 included an update to the lzo compression code.  So far
> as I know, there is nothing wrong with it, but it no longer matches the
> test vectors in crypto/testmgr.  Please cherry-pick this fix:
> 
> commit 0ec7382036922be063b515b2a3f1d6f7a607392c
> Author: Markus F.X.J. Oberhumer <markus@oberhumer.com>
> Date:   Sun Oct 14 15:39:04 2012 +0200
> 
>     crypto: testmgr - update LZO compression test vectors

Noted, thank you for the information. I don't understand much the problem
you spotted since the change only applies to decompression while the patch
above seems to only adjust the compression patterns (I don't know how the
compression tests work though). Also, for the fix to behave differently,
it requires slightly more than 16 million consecutive 255 bytes to trigger
the error path, which the commit above doesn't seem to perform.

But I'm queuing this patch anyway.

Thanks!
Willy

Re: [2.6.32 regression] lzo fails crypto self-tests

[Ben Hutchings]

[-- Attachment #1: Type: text/plain, Size: 1581 bytes --]

On Sat, 2015-08-01 at 22:26 +0200, Willy Tarreau wrote:
> Hi Ben,
> 
> On Sat, Aug 01, 2015 at 06:33:34PM +0100, Ben Hutchings wrote:
> > Linux 2.6.32.64 included an update to the lzo compression code.  So far
> > as I know, there is nothing wrong with it, but it no longer matches the
> > test vectors in crypto/testmgr.  Please cherry-pick this fix:
> > 
> > commit 0ec7382036922be063b515b2a3f1d6f7a607392c
> > Author: Markus F.X.J. Oberhumer <markus@oberhumer.com>
> > Date:   Sun Oct 14 15:39:04 2012 +0200
> > 
> >     crypto: testmgr - update LZO compression test vectors
> 
> Noted, thank you for the information. I don't understand much the problem
> you spotted since the change only applies to decompression while the patch
> above seems to only adjust the compression patterns (I don't know how the
> compression tests work though). Also, for the fix to behave differently,
> it requires slightly more than 16 million consecutive 255 bytes to trigger
> the error path, which the commit above doesn't seem to perform.
> 
> But I'm queuing this patch anyway.

You applied this before the fix:

commit bf1d894fed5852708a989a0a3f5df1a169a06afc
Author: Markus F.X.J. Oberhumer <markus@oberhumer.com>
Date:   Mon Aug 13 17:25:44 2012 +0200

    lib/lzo: Update LZO compression to current upstream version

and that's what breaks the self-tests.

You can see this by trying to load the ubifs module, which uses lzo via
the crypto API.

Ben.

-- 
Ben Hutchings
One of the nice things about standards is that there are so many of them.


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 811 bytes --]

Re: [2.6.32 regression] lzo fails crypto self-tests

[Willy Tarreau]

On Sat, Aug 01, 2015 at 09:40:43PM +0100, Ben Hutchings wrote:
> On Sat, 2015-08-01 at 22:26 +0200, Willy Tarreau wrote:
> > Hi Ben,
> > 
> > On Sat, Aug 01, 2015 at 06:33:34PM +0100, Ben Hutchings wrote:
> > > Linux 2.6.32.64 included an update to the lzo compression code.  So far
> > > as I know, there is nothing wrong with it, but it no longer matches the
> > > test vectors in crypto/testmgr.  Please cherry-pick this fix:
> > > 
> > > commit 0ec7382036922be063b515b2a3f1d6f7a607392c
> > > Author: Markus F.X.J. Oberhumer <markus@oberhumer.com>
> > > Date:   Sun Oct 14 15:39:04 2012 +0200
> > > 
> > >     crypto: testmgr - update LZO compression test vectors
> > 
> > Noted, thank you for the information. I don't understand much the problem
> > you spotted since the change only applies to decompression while the patch
> > above seems to only adjust the compression patterns (I don't know how the
> > compression tests work though). Also, for the fix to behave differently,
> > it requires slightly more than 16 million consecutive 255 bytes to trigger
> > the error path, which the commit above doesn't seem to perform.
> > 
> > But I'm queuing this patch anyway.
> 
> You applied this before the fix:
> 
> commit bf1d894fed5852708a989a0a3f5df1a169a06afc
> Author: Markus F.X.J. Oberhumer <markus@oberhumer.com>
> Date:   Mon Aug 13 17:25:44 2012 +0200
> 
>     lib/lzo: Update LZO compression to current upstream version
> 
> and that's what breaks the self-tests.

Ah OK I thought it was just the fix which did it, now I understand.

> You can see this by trying to load the ubifs module, which uses lzo via
> the crypto API.

Thanks for the info.

Willy