lkml.org 
[lkml]   [2018]   [Jul]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 2/3][RFC] PM / Hibernate: Encrypt the snapshot pages before submitted to the block device
On Wed, Jul 18, 2018 at 11:48:19PM +0800, joeyli wrote:
> On Fri, Jul 13, 2018 at 03:34:25PM +0800, Yu Chen wrote:
> > Hi,
> > On Thu, Jul 12, 2018 at 06:10:37PM +0800, joeyli wrote:
> > > Hi Yu Chen,
> > >
> > > Sorry for my delay...
> > >
> > > On Fri, Jul 06, 2018 at 11:28:56PM +0800, Yu Chen wrote:
> [...snip]
> > > > > Why the cryption code must be indepent of snapshot code?
> > > > >
> > > > Modules can be easier to be maintained and customized/improved in the future IMO..
> > >
> > > hm... currently I didn't see apparent benefit on this...
> > >
> > > About modularization, is it possible to separate the key handler code
> > > from crypto_hibernation.c? Otherwise the snapshot signature needs
> > > to depend on encrypt function.
> > >
> > I understand.
> > It seems that we can reuse crypto_data() for the signature logic.
> > For example, add one parameter for crypto_data(..., crypto_mode)
> > the crypto_mode is a combination of ENCRYPT/DECRYPT/SIGNATURE_START/SIGNATURE_END,
> > and can be controled by sysfs. If SIGNATURE_START is enabled, the crypto_data()
> > invokes crypto_shash_update() to get "hmac(sha512)" hash, and if SIGNATURE_END
> > is enabled, crypto_shash_final() stores the final result in global buffer
>
> I agree that the swsusp_prepare and hmac-hash codes can be extract to
> crypto_hibernation.
>
> > struct hibernation_crypto.keys.digest[SNAPSHOT_DIGEST_SIZE] which can be
> > passed to the restore kernel, the same as the salt. Besides,
>
> The salt is stored in the swsusp_header and put in swap. What I want
> is that the signature of snapshot should be keep in the snapshot header
> as my original design in patch. The benefit is that the snapshot with
> signature can be stored to any place but not just swap. The user space
> can take snapshot image with signature to anywhere.
>
Okay, got it.
> > the swsusp_prepare_hash() in your code could be moved into
> > init_crypto_helper(), thus the signature key could also reuse
> > the same key passed from user space or via get_efi_secret_key().
> > In this way, the change in snapshot.c is minimal and the crypto
> > facilities could be maintained in one file.
>
> I agree. But as I mentioned in that mail, the key handler codes
> in hibernation_crypto() should be extract to another C file to avoid
> the logic is mixed with the crypto code. The benefit is that we can
> add more key sources like encrypted key or EFI key in the future.
>
Ok, will extract the key handler code from this file.
> > > > > > 2. There's no need to traverse the snapshot image twice, if the
> > > > > > image is large(there's requirement on servers now) we can
> > > > > > simplyly do the encryption before the disk IO, and this is
> > > > > > why PATCH[2/3] looks like this.
> > > > >
> > > > > If the encryption solution is only for block device, then the uswsusp
> > > > > interface must be locked-down when kernel is in locked mode:
> > > > >
> > > > > uswsusp: Disable when the kernel is locked down
> > > > > https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=lockdown-20180410&id=8732c1663d7c0305ae01ba5a1ee4d2299b7b4612
> > > > >
> > > > > I still suggest to keep the solution to direct encript the snapshot
> > > > > image for uswsusp because the snapshot can be encrypted by kernel
> > > > > before user space get it.
> > > > >
> > > > > I mean that if the uswsusp be used, then kernel direct encrypts the
> > > > > snapshot image, otherwise kernel encrypts pages before block io.
> > > > >
> > > > I did not quite get the point, if the kernel has been locked down,
> > > > then the uswsusp is disabled, why the kernel encrypts the snapshot
> > > > for uswsusp?
> > >
> > > There have some functions be locked-down because there have no
> > > appropriate mechanisms to check the integrity of writing data. If
> > > the snapshot image can be encrypted and authentication, then the
> > > uswsusp interface is avaiable for userspace. We do not need to lock
> > > it down.
> > Ok, I got your point. To be honest, I like your implementation to treat
> > the snapshot data seperately except that it might cause small overhead
> > when traversing large number of snapshot and make snapshot logic a little
> > coupling with crypto logic. Let me send our v2 using the crypto-before-blockio
> > for review and maybe change to your solution using the encapsulated APIs in
> > crypto_hibernate.c.
>
> Because sometimes I stick on other topics...
>
> If you are urgent for pushing your encryption solution. Please do not
> consider too much on signature check. Just complete your patches and push
> to kernel mainline. I will follow your result to respin my signature work.
>
Thanks, I've just sent out another version before I noticed your
comment yesterday, let me address your concern in v3,
but please feel free to comment on v2.
Thanks,
Yu

> Thanks
> Joey Lee

\
 
 \ /
  Last update: 2018-07-19 11:11    [W:0.113 / U:0.124 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site