Re: [PATCH 2/2] x86/KASLR: Use old ident map page table if physical randomization failed

[Baoquan He <bhe@redhat.com>]

On 04/26/17 at 12:12pm, Kees Cook wrote:
> On Wed, Apr 26, 2017 at 3:39 AM, Baoquan He <bhe@redhat.com> wrote:
> > Dave found when kdump kernel will reset to bios immediately if kaslr
> > is enabled and physical randomization failed to faind a new position
> > for kernel. But nokaslr works in this case.
> >
> > The reason is kaslr will install a new page table for ident mapping,
> > while it missed to consider building ident mapping for original area
> > of kernel if kaslr failed on physical randomization.
> >
> > In fact bootloaders including kexec/kdump have built ident mapping
> > for original place of kernel. We can only install new ident mapping
> > page table when physical kaslr succeeds. Otherwise we just keep the
> > old page table unchanged just like nokaslr does.
> >
> > Signed-off-by: Baoquan He <bhe@redhat.com>
> > Signed-off-by: Dave Young <dyoung@redhat.com>
> > Cc: "H. Peter Anvin" <hpa@zytor.com>
> > Cc: Thomas Gleixner <tglx@linutronix.de>
> > Cc: Ingo Molnar <mingo@redhat.com>
> > Cc: x86@kernel.org
> > Cc: Kees Cook <keescook@chromium.org>
> > Cc: Yinghai Lu <yinghai@kernel.org>
> > Cc: Borislav Petkov <bp@suse.de>
> > Cc: Dave Jiang <dave.jiang@intel.com>
> > Cc: Thomas Garnier <thgarnie@google.com>
> 
> Nice catch!
> 
> Acked-by: Kees Cook <keescook@chromium.org>

Thanks, Kees.

Seems I forget telling this only happens in kexec/kdump kernel. Since
the ident mapping has been built for kexec/kdump in 1st kernel for the
whole memory by calling init_pgtable(). Here if physical randomizaiton
failed, it won't build ident mapping for the original area of kernel but
change to new page table '_pgtable'. Then kernel will reset to bios
immediately caused by no ident mapping.

While normal kernel won't be impacted because it comes here via
startup_32() and cr3 will be _pgtable already. In startup_32() ident
mapping is built for 0~4G area. In kaslr We just append to the existing
area instead of entirely overwriting it for on-demand ident mapping
building. So ident mapping for the original area of kernel is still
there.

I will post v2 with a improved patch log, and with your Acked-by.

> 
> -Kees
> 
> > ---
> >  arch/x86/boot/compressed/kaslr.c | 10 ++++++++--
> >  1 file changed, 8 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/x86/boot/compressed/kaslr.c b/arch/x86/boot/compressed/kaslr.c
> > index e5eb0c3..7a8b443 100644
> > --- a/arch/x86/boot/compressed/kaslr.c
> > +++ b/arch/x86/boot/compressed/kaslr.c
> > @@ -650,10 +650,16 @@ void choose_random_location(unsigned long input,
> >                         add_identity_map(random_addr, output_size);
> >                         *output = random_addr;
> >                 }
> > +
> > +               /*
> > +                * This actually loads the identity pagetable on x86_64.
> > +                * And this should only be done only if a new position
> > +                * is found. Otherwise we should keep the old page table
> > +                * to make it be like nokaslr case.
> > +                */
> > +               finalize_identity_maps();
> >         }
> >
> > -       /* This actually loads the identity pagetable on x86_64. */
> > -       finalize_identity_maps();
> >
> >         /* Pick random virtual address starting from LOAD_PHYSICAL_ADDR. */
> >         if (IS_ENABLED(CONFIG_X86_64))
> > --
> > 2.5.5
> >
> 
> 
> 
> -- 
> Kees Cook
> Pixel Security