Re: [PATCH] x86/mm: fix regression with huge pages on PAE

From: "Kirill A. Shutemov" <kirill@shutemov.name>
To: Borislav Petkov <bp@alien8.de>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
	hpa@zytor.com, tglx@linutronix.de, mingo@redhat.com,
	akpm@linux-foundation.org, linux-mm@kvack.org,
	linux-kernel@vger.kernel.org, x86@kernel.org, jgross@suse.com,
	konrad.wilk@oracle.com, elliott@hpe.com,
	boris.ostrovsky@oracle.com, Toshi Kani <toshi.kani@hpe.com>
Subject: Re: [PATCH] x86/mm: fix regression with huge pages on PAE
Date: Tue, 10 Nov 2015 15:53:03 +0200	[thread overview]
Message-ID: <20151110135303.GA11246@node.shutemov.name> (raw)
In-Reply-To: <20151110123429.GE19187@pd.tnic>

On Tue, Nov 10, 2015 at 01:34:29PM +0100, Borislav Petkov wrote:
> On Tue, Nov 10, 2015 at 01:18:10AM +0200, Kirill A. Shutemov wrote:
> > diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
> > index dd5b0aa9dd2f..c1e797266ce9 100644
> > --- a/arch/x86/include/asm/pgtable_types.h
> > +++ b/arch/x86/include/asm/pgtable_types.h
> > @@ -279,17 +279,14 @@ static inline pmdval_t native_pmd_val(pmd_t pmd)
> >  static inline pudval_t pud_pfn_mask(pud_t pud)
> >  {
> >  	if (native_pud_val(pud) & _PAGE_PSE)
> > -		return PUD_PAGE_MASK & PHYSICAL_PAGE_MASK;
> > +		return ~((1ULL << PUD_SHIFT) - 1) & PHYSICAL_PAGE_MASK;
> 
> In file included from ./arch/x86/include/asm/boot.h:5:0,
>                  from ./arch/x86/boot/boot.h:26,
>                  from arch/x86/realmode/rm/wakemain.c:2:
> ./arch/x86/include/asm/pgtable_types.h: In function ‘pud_pfn_mask’:
> ./arch/x86/include/asm/pgtable_types.h:282:10: warning: large integer implicitly truncated to unsigned type [-Woverflow]
>    return ~((1ULL << PUD_SHIFT) - 1) & PHYSICAL_PAGE_MASK;
>           ^
> ./arch/x86/include/asm/pgtable_types.h: In function ‘pmd_pfn_mask’:
> ./arch/x86/include/asm/pgtable_types.h:300:10: warning: large integer implicitly truncated to unsigned type [-Woverflow]
>    return ~((1ULL << PMD_SHIFT) - 1) & PHYSICAL_PAGE_MASK;
>           ^
> In file included from ./arch/x86/include/asm/boot.h:5:0,
>                  from arch/x86/realmode/rm/../../boot/boot.h:26,
>                  from arch/x86/realmode/rm/../../boot/video-mode.c:18,
>                  from arch/x86/realmode/rm/video-mode.c:1:
> ./arch/x86/include/asm/pgtable_types.h: In function ‘pud_pfn_mask’:
> ./arch/x86/include/asm/pgtable_types.h:282:10: warning: large integer implicitly truncated to unsigned type [-Woverflow]
>    return ~((1ULL << PUD_SHIFT) - 1) & PHYSICAL_PAGE_MASK;
>           ^
> ...
> 
> That's a 64-bit config.

Oh.. pmdval_t/pudval_t is 'unsinged long' on 64 bit. But realmode code
uses -m16 which makes 'unsigned long' 32-bit therefore truncation warning.

These helpers not really used in realmode code. I see few ways out:

 - convert helpers to macros to avoid their translation;

 - wrap the code into not-for-realmode ifdef. (Do we have any?);

 - convert pudval_t/pmdval_t to u64 for 64-bit. I'm not sure what side
   effects would it have.

Any opinions?

-- 
 Kirill A. Shutemov