Re: Splat during resume

[Borislav Petkov]

On Sat, Mar 25, 2017 at 10:46:15PM +0100, Borislav Petkov wrote:
> On Sat, Mar 25, 2017 at 07:58:55PM +0100, Borislav Petkov wrote:
> > Hey Rafael,
> > 
> > have you seen this already (partial splat photo attached)? Happens
> > during resume from s2d. Judging by the timestamps, this looks like the
> > resume kernel before we switch to the original, boot one but I could be
> > mistaken.
> > 
> > This is -rc3+tip/master.
> > 
> > I can't catch a full splat because this is a laptop and it doesn't have
> > serial. netconsole is helping me for shit so we'd need some guess work.
> > 
> > So I'm open to suggestions.
> > 
> > Please don't say "bisect" yet ;-)))
> 
> No need, I found it. Reverting
> 
>   ea3b5e60ce80 ("x86/mm/ident_map: Add 5-level paging support")
> 
> makes the machine suspend and resume just fine again. Lemme add people to CC.

So I see rIP pointing to ident_pmd_init() and the stack trace has
load_image_and_restore() so if I try to connect the dots, I get:

load_image_and_restore
|-> hibernation_restore
 |-> resume_target_kernel
  |-> swsusp_arch_resume
   |-> set_up_temporary_mappings
    |-> kernel_ident_mapping_init
     |-> ... ident_pmd_init

I'll let you folks make sense of what's going on.

Thanks.

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.