Re: [PATCH 4.4 41/43] x86/mm/pat: Make set_memory_np() L1TF safe

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Sun, Sep 09, 2018 at 10:06:06AM -0700, Guenter Roeck wrote:
> On 09/09/2018 09:46 AM, Ben Hutchings wrote:
> > On Tue, 2018-08-14 at 19:18 +0200, Greg Kroah-Hartman wrote:
> > > 4.4-stable review patch.  If anyone has any objections, please let me know.
> > > 
> > > ------------------
> > > 
> > > From: Andi Kleen <ak@linux.intel.com>
> > > 
> > > commit 958f79b9ee55dfaf00c8106ed1c22a2919e0028b upstream
> > [...]
> > > --- a/arch/x86/include/asm/pgtable.h
> > > +++ b/arch/x86/include/asm/pgtable.h
> > > @@ -378,12 +378,39 @@ static inline pmd_t pfn_pmd(unsigned lon
> > > >   	return __pmd(pfn | massage_pgprot(pgprot));
> > >   }
> > > +static inline pud_t pfn_pud(unsigned long page_nr, pgprot_t pgprot)
> > > +{
> > > +	phys_addr_t pfn = page_nr << PAGE_SHIFT;
> > > +	pfn ^= protnone_mask(pgprot_val(pgprot));
> > > +	pfn &= PHYSICAL_PUD_PAGE_MASK;
> > > +	return __pud(pfn | massage_pgprot(pgprot));
> > > +}
> > [...]
> > 
> > This (and the backport to 4.9) are missing the fix from commit
> > e14d7dfb41f5 "x86/speculation/l1tf: Fix up pte->pfn conversion for
> > PAE", as that was applied earlier in the series.  But since PAE implies
> > only 3-level paging I don't know how the PUD functions get used or
> > whether this actually matters.
> > 
> Excellent find.
> 
> e14d7dfb41f5 (re-)applies cleanly to both 4.4.y and 4.9.y. Since most of its
> changes are already applied, the only remaining change is
> 
> diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
> index c535012bdb56..5736306bdaab 100644
> --- a/arch/x86/include/asm/pgtable.h
> +++ b/arch/x86/include/asm/pgtable.h
> @@ -420,7 +420,7 @@ static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
> 
>  static inline pud_t pfn_pud(unsigned long page_nr, pgprot_t pgprot)
>  {
> -       phys_addr_t pfn = page_nr << PAGE_SHIFT;
> +       phys_addr_t pfn = (phys_addr_t)page_nr << PAGE_SHIFT;
>         pfn ^= protnone_mask(pgprot_val(pgprot));
>         pfn &= PHYSICAL_PUD_PAGE_MASK;
>         return __pud(pfn | massage_pgprot(pgprot));
> 
> after cherry-picking it into both branches.
> 
> I think we should re-apply it to both 4.4.y and 4.9.y be on the safe side.

Thanks for finding and figuring out the fix.  Now queued up to both
trees.

greg k-h