Re: [PATCH encrypted swsusp 1/3] core functionality

[Matt Mackall <mpm@selenic.com>]

On Thu, Apr 14, 2005 at 10:18:12PM +0200, Pavel Machek wrote:
> Hi!
> 
> > > So we would need to zero out the suspend image in swap to prevent the
> > > retrieval of this data from the running machine (imagine a
> > > remote-root-hole).
> > >
> > > Zeroing out the suspend image means "write lots of megabytes to the
> > > disk" which takes a lot of time.
> > 
> > Zero only the mlocked regions. This should take essentially no time at
> > all. Swsusp knows which these are because they have to be mlocked
> 
> I believe this is tricky to implement. You are free to produce patch,
> and if that patch is nicer/simpler than Anreas's code, I may consider
> it.

If I understand swsusp correctly, we can simply set a bit in the pbe
struct to indicate that it's a locked page. 

This can be done by walking the vma list attached to the page's
address space with vma_prio_tree_foreach() and checking the
vma->vm_flags with VM_LOCKED. Analogous to what the swapout code does.

We can either do this in data_write() or preferably higher
(copy_data?) when we have the pfn handy. The lock bit can be stashed
in bit 0 of pbe->address, among other places. Then in data_read, we
check the bit and zero the source.

As I'm not about to actually use swsusp any time soon, someone else is
invited to implement the above. Should take about 10-20 lines.

-- 
Mathematics is the supreme nostalgia of our time.