[Nottingham] Hibernate issues

[Jason Irwin]

On 25/11/13 13:51, Mike Cardwell wrote:
> Ah, that makes sense. FWIW, I stopped using swap partitions a couple of years
> ago when I started using SSDs. Neither my laptop nor my desktop have a swap
> partition; I just made sure I had more than enough RAM. 8GB in the laptop,
> 16GB in the desktop. I can't use hibernate because of this, but that doesn't
> bother me.
Yeah, the way GNU/Linux does hibernate bothers me. If the laptop is
stolen, I *want* someone to be able to boot it. I want the tracking
software to activate and speak to mother. Which won't happen in this
scenario.

What I am thinking (and I be no kernel dev) is give up on using the swap
for hibernate. Dump a file encrypted with the user's key (or drop it
into their home folder) that holds the state. Set the various flags to
the kernel knows a resume is coming. When the machine next powers on (by
some magical means) let the user choose to log into their previous
session, dump their previous session or login as someone else. Clearly
the first two require the correct password. If the user choose resume,
use that file to re-populate memory as it does now - maybe what it does
is repopulate swap and then do a mini-reboot or something? I get a
feeling there's a chicken/egg issue.

This  means that swap can go back to using a random key and you don't
need to have two passwords to log in (one shared with all users).

Obviously you need enough space *somewhere* to store the info, but a bit
of compression should reduce that.

I am sure there's serious issues with the above, but the current method
appears to be a bit clunky. Certainly when compared to Windows and OS X
(although maybe they don't encrypt swap or temp and have heaps of other
issues).

-- 
╔═════════════╦══════════════════════════════════════════╗
║ Jason Irwin ║ OpenPGP (GPG/PGP) Public Key: 0xD0C592B1 ║
║             ║ Import from hkp://subkeys.pgp.net        ║
╚═════════════╩══════════════════════════════════════════╝