Fixing userspace crypto in 3.14.x?

Fixing userspace crypto in 3.14.x?

[J. Paul Reed]

Userspace crypto has been broken in the 3.14 series since 3.14.61; see:

http://marc.info/?t=145571848500001&r=1&w=2

Specifically: 

http://marc.info/?l=linux-crypto-vger&m=145597880120294&w=2

and

http://marc.info/?l=linux-crypto-vger&m=145579378307060&w=2

I waited a few 3.14 maintenance releases to see if this would get fixed,
but as of 3.14.64, it remains broken.

Are there any plans plans to fix this? (For those of us with luks-encrypted
partitions, the 3.14 tree is now unusable past 3.14.60.)

thanks,
preed
-- 
J. Paul Reed         PGP: 0xDF8708F8 -- http://www.soberbuildengineer.com
=========================================================================
I've never seen an airplane yet that can read the type ratings on your
pilot's license.                                       -- Chuck Boedecker

Re: Fixing userspace crypto in 3.14.x?

[Greg KH]

On Sat, Mar 12, 2016 at 04:26:54PM -0800, J. Paul Reed wrote:
> 
> Userspace crypto has been broken in the 3.14 series since 3.14.61; see:
> 
> http://marc.info/?t=145571848500001&r=1&w=2
> 
> Specifically: 
> 
> http://marc.info/?l=linux-crypto-vger&m=145597880120294&w=2
> 
> and
> 
> http://marc.info/?l=linux-crypto-vger&m=145579378307060&w=2
> 
> I waited a few 3.14 maintenance releases to see if this would get fixed,
> but as of 3.14.64, it remains broken.
> 
> Are there any plans plans to fix this? (For those of us with luks-encrypted
> partitions, the 3.14 tree is now unusable past 3.14.60.)

So, what patch should I revert from 3.10.y and 3.14.y to "fix" this?  I
thought only 4.1.y was affected.  Or better yet, I asked for backports
to resolve this, if someone could come up with them, that would be even
better.

thanks,

greg k-h

Re: Fixing userspace crypto in 3.14.x?

[J. Paul Reed]

On 12 Mar 2016 at 16:45:10, Greg KH arranged the bits on my disk to say:

> So, what patch should I revert from 3.10.y and 3.14.y to "fix" this?  I
> thought only 4.1.y was affected.  Or better yet, I asked for backports to
> resolve this, if someone could come up with them, that would be even
> better.

The following message in that thread has assumed-bad commit (and the
associated bisect log):

http://marc.info/?l=linux-crypto-vger&m=145575300630247&w=2

As for backports, I'm merely a lowly kernel user.

later,
preed
-- 
J. Paul Reed         PGP: 0xDF8708F8 -- http://www.soberbuildengineer.com
=========================================================================
I've never seen an airplane yet that can read the type ratings on your
pilot's license.                                       -- Chuck Boedecker

Re: Fixing userspace crypto in 3.14.x?

[Greg KH]

On Sat, Mar 12, 2016 at 05:10:53PM -0800, J. Paul Reed wrote:
> 
> On 12 Mar 2016 at 16:45:10, Greg KH arranged the bits on my disk to say:
> 
> > So, what patch should I revert from 3.10.y and 3.14.y to "fix" this?  I
> > thought only 4.1.y was affected.  Or better yet, I asked for backports to
> > resolve this, if someone could come up with them, that would be even
> > better.
> 
> The following message in that thread has assumed-bad commit (and the
> associated bisect log):
> 
> http://marc.info/?l=linux-crypto-vger&m=145575300630247&w=2
> 
> As for backports, I'm merely a lowly kernel user.

I don't see how that breaks anything in userspace, can you please bring
this up with the developers of that patch and the crypto mailing list?

thanks,

greg k-h

Re: Fixing userspace crypto in 3.14.x?

[J. Paul Reed]

On 12 Mar 2016 at 19:02:46, Greg KH arranged the bits on my disk to say:

> On Sat, Mar 12, 2016 at 05:10:53PM -0800, J. Paul Reed wrote:
> > 
> > On 12 Mar 2016 at 16:45:10, Greg KH arranged the bits on my disk to say:
> > 
> > > So, what patch should I revert from 3.10.y and 3.14.y to "fix" this?  I
> > > thought only 4.1.y was affected.  Or better yet, I asked for backports to
> > > resolve this, if someone could come up with them, that would be even
> > > better.
> > 
> > The following message in that thread has assumed-bad commit (and the
> > associated bisect log):
> > 
> > http://marc.info/?l=linux-crypto-vger&m=145575300630247&w=2
> > 
> > As for backports, I'm merely a lowly kernel user.
> 
> I don't see how that breaks anything in userspace, can you please bring
> this up with the developers of that patch and the crypto mailing list?

I already did:

http://marc.info/?l=linux-crypto-vger&m=145620828015563&w=2

The thread I pointed you to was what I was told by the crypto list:

http://marc.info/?l=linux-crypto-vger&m=145620853015626&w=2

In that thread, the developer was also apparently made aware:

http://www.spinics.net/lists/linux-crypto/msg17931.html

I think bringing it up to the developer and the crypto mailing list again is
unlikely to get this issue addressed, as they are both aware of it.

later,
preed
-- 
J. Paul Reed         PGP: 0xDF8708F8 -- http://www.soberbuildengineer.com
=========================================================================
I've never seen an airplane yet that can read the type ratings on your
pilot's license.                                       -- Chuck Boedecker

Re: Fixing userspace crypto in 3.14.x?

[Greg KH]

On Sat, Mar 12, 2016 at 07:55:57PM -0800, J. Paul Reed wrote:
> On 12 Mar 2016 at 19:02:46, Greg KH arranged the bits on my disk to say:
> 
> > On Sat, Mar 12, 2016 at 05:10:53PM -0800, J. Paul Reed wrote:
> > > 
> > > On 12 Mar 2016 at 16:45:10, Greg KH arranged the bits on my disk to say:
> > > 
> > > > So, what patch should I revert from 3.10.y and 3.14.y to "fix" this?  I
> > > > thought only 4.1.y was affected.  Or better yet, I asked for backports to
> > > > resolve this, if someone could come up with them, that would be even
> > > > better.
> > > 
> > > The following message in that thread has assumed-bad commit (and the
> > > associated bisect log):
> > > 
> > > http://marc.info/?l=linux-crypto-vger&m=145575300630247&w=2
> > > 
> > > As for backports, I'm merely a lowly kernel user.
> > 
> > I don't see how that breaks anything in userspace, can you please bring
> > this up with the developers of that patch and the crypto mailing list?
> 
> I already did:
> 
> http://marc.info/?l=linux-crypto-vger&m=145620828015563&w=2
> 
> The thread I pointed you to was what I was told by the crypto list:
> 
> http://marc.info/?l=linux-crypto-vger&m=145620853015626&w=2
> 
> In that thread, the developer was also apparently made aware:
> 
> http://www.spinics.net/lists/linux-crypto/msg17931.html
> 
> I think bringing it up to the developer and the crypto mailing list again is
> unlikely to get this issue addressed, as they are both aware of it.

Ok, you are right, and have been very patient, thanks for bringing this
up again.  I've reverted this patch, hopefully that resolves the issue.

thanks,

greg k-h

Re: Fixing userspace crypto in 3.14.x?

[J. Paul Reed]

On 12 Mar 2016 at 21:39:16, Greg KH arranged the bits on my disk to say:

> Ok, you are right, and have been very patient, thanks for bringing this
> up again.  I've reverted this patch, hopefully that resolves the issue.

I will test 3.14.65 when it ships, and let you know if, it turns out, that
wasn't the commit that broke it.

thanks again,
preed
-- 
J. Paul Reed         PGP: 0xDF8708F8 -- http://www.soberbuildengineer.com
=========================================================================
I've never seen an airplane yet that can read the type ratings on your
pilot's license.                                       -- Chuck Boedecker

Re: Fixing userspace crypto in 3.14.x?

[J. Paul Reed]

On 12 Mar 2016 at 21:39:16, Greg KH arranged the bits on my disk to say:

> Ok, you are right, and have been very patient, thanks for bringing this
> up again.  I've reverted this patch, hopefully that resolves the issue.

Greg,

I was finally able to get booted into a new kernel: reverting this patch
makes my luks-encrypted partitions work again. I was only able to test
3.14, not 3.10.

Thanks for working to get it into 3.14.65.

best,
preed
-- 
J. Paul Reed         PGP: 0xDF8708F8 -- http://www.soberbuildengineer.com
=========================================================================
I've never seen an airplane yet that can read the type ratings on your
pilot's license.                                       -- Chuck Boedecker
> 
> thanks,
> 
> greg k-h