Re: PROBLEM: Resume form hibernate broken by setting NX on gap

[Rafael J. Wysocki]

On Friday, June 10, 2016 11:27:29 PM Rafael J. Wysocki wrote:
> On Friday, June 10, 2016 11:09:22 AM Kees Cook wrote:
> > On Thu, Jun 9, 2016 at 9:14 PM, Logan Gunthorpe <logang@xxxxxxxxxxxx> wrote:
> > > Hey,
> > >
> > > I've still be trying to figure this out as I have time.
> > >
> > > I tried printing a couple restore addresses and nothing I can find seems
> > > anywhere near the rodata/ex_table boundary.
> > >
> > > I tried with the (badly formatted) below and got the following. Nothing too
> > > surprising. I've attached a kallsyms that matches the kernel for reference.
> > >
> > > restore_code: ffff880157c3b000
> > > jump_addr: ffffffff81446be0
> > >
> > >
> > > diff --git a/arch/x86/power/hibernate_64.c b/arch/x86/power/hibernate_64.c
> > > index 009947d..6efedb7 100644
> > > --- a/arch/x86/power/hibernate_64.c
> > > +++ b/arch/x86/power/hibernate_64.c
> > > @@ -92,6 +92,9 @@ int swsusp_arch_resume(void)
> > >         memcpy(relocated_restore_code, &core_restore_code,
> > >                &restore_registers - &core_restore_code);
> > >
> > > +       pr_info("restore_code: %p\n", relocated_restore_code);
> > > +       pr_info("jump_addr: %lx\n", restore_jump_address);
> > > +
> > 
> > Also interesting would be the "relocated_restore_code" address, as
> > well as a dump of /sys/kernel/debug/kernel_page_tables (from
> > CONFIG_X86_PTDUMP).
> > 
> > I'm baffled by the problem, but the best I can understand is the the
> > relocated_restore_code range isn't executable (which should be visible
> > from finding it in /sys/kernel/debug/kernel_page_tables), but I don't
> > see how to solve that since my original patch didn't work.
> > 
> > Rafael, is this something you have time to look at quickly?
> 
> Well, not really, but I'll do my best to look at it in the next few days.

OK, I have a theory, but I need a bit of help.

This may be a dumb question, but I don't quite remember the answer readily.

Given a physical address, how do I get the corresponding virtual one under
the kernel identity mapping?

Thanks,
Rafael