Re: [RFC PATCH 00/18 v3] Signature verification of hibernate snapshot

[joeyli <jlee@suse.com>]

於 四，2013-08-29 於 23:32 +0200，Pavel Machek 提到：
> Hi!
> 
> > > >    - Bootloader store the public key to EFI boottime variable by itself
> > > >    - Bootloader put The private key to S4SignKey EFI variable for forward to
> > > >      kernel.
> > > 
> > > Is the UEFI NVRAM really suited for such regular updates?
> > > 
> > 
> > Yes, Matthew raised this concern at before. I modified patch to load
> > private key in efi stub kernel, before ExitBootServices(), that means we
> > don't need generate key-pair at every system boot. So, the above
> > procedure of efi bootloader will only run one time. 
> > 
> > User can enable SNAPSHOT_REGEN_KEYS kernel config to notify efi
> > booloader regenerate key-pair for every S4 to improve security if he
> > want. So, the key-pair re-generate procedure will only launched when S4
> > resume, not every system boot.
> 
> How many writes can UEFI NVRAM survive? (Is it NOR?)

Currently doesn't have enough information for normal. Yes, I don't know.

> 
> "every S4 resume" may be approximately "every boot" for some users...
> 									Pavel

Yes, it's possible.

So, this option will be disabled by default. Default will only generate
one key-pair for every hibernate.
If "re-generate key-pair for every S4" is still hurt lift of UEFI NVRAM
too much, then another thinking for re-generate key-pair are:
 
 + Re-generate key-pair after a number of hibernates.
   e.g. after 5, 10, 20... times
or
 + Random re-generate key-pair?


On the other hand...
In current design, GenS4Key EFI variable could be write by userland
hibernate tool, kernel will respect GenS4Key value from userland when
hibernate launch. So, userland can tell bootloader to lunch the key-pair
regeneration procedure.


Thanks a lot!
Joey LEe