Re: [PATCH 1/3] x86: Honour passed pgprot in track_pfn_insert() and track_pfn_remap()
From: Andy Lutomirski
Date: Tue Feb 09 2016 - 22:07:28 EST
On Feb 9, 2016 6:24 AM, "Ingo Molnar" <mingo@xxxxxxxxxx> wrote:
>
>
> * Matthew Wilcox <willy@xxxxxxxxxxxxxxx> wrote:
>
> > > I sure hope not. If vm_page_prot was writable, something was already broken,
> > > because this is the vvar mapping, and the vvar mapping is VM_READ (and not
> > > even VM_MAYREAD).
> >
> > I do beg yor pardon. I thought you were inserting a readonly page into the
> > middle of a writable mapping. Instead you're inserting a non-executable page
> > into the middle of a VM_READ | VM_EXEC mapping. Sorry for the confusion. I
> > should have written:
> >
> > "like your patch ends up mapping the HPET into userspace executable"
> >
> > which is far less exciting.
>
> Btw., a side note, an executable HPET page has its own dangers as well, for
> example because it always changes in value, it can probabilistically represent
> 'sensible' (and dangerous) executable x86 instructions that exploits can return
> to.
>
> So only mapping it readable (which Andy's patch attempts I think) is worthwile.
The whole vma is readable but not executable, so I don't think this
was a problem. It's also at a randomized address, which helps.
--Andy