Fix generic hugetlb pte check return type.

The include/asm-generic/hugetlb.h stubs that just vector huge_pte_*()
calls to the pte_*() implementations won't work in certain situations.

x86 and sparc, for example, return "unsigned long" from the bit
checks, and just go "return pte_val(pte) & PTE_BIT_FOO;"

But since huge_pte_*() returns 'int', if any high bits on 64-bit are
relevant, they get chopped off.

The net effect is that we can loop forever trying to COW a huge page,
because the huge_pte_write() check signals false all the time.

Reported-by: Gurudas Pai <gurudas.pai@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
---

Anyone mind if I push this through the sparc tree?

--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

On Wed, 2 Oct 2013, David Miller wrote:

> The include/asm-generic/hugetlb.h stubs that just vector huge_pte_*()
> calls to the pte_*() implementations won't work in certain situations.
> 
> x86 and sparc, for example, return "unsigned long" from the bit
> checks, and just go "return pte_val(pte) & PTE_BIT_FOO;"
> 

x86 should be safe since it's known that both _PAGE_BIT_RW and 
_PAGE_BIT_DIRTY are <= 32.

> But since huge_pte_*() returns 'int', if any high bits on 64-bit are
> relevant, they get chopped off.
> 
> The net effect is that we can loop forever trying to COW a huge page,
> because the huge_pte_write() check signals false all the time.
> 
> Reported-by: Gurudas Pai <gurudas.pai@oracle.com>
> Signed-off-by: David S. Miller <davem@davemloft.net>

Acked-by: David Rientjes <rientjes@google.com>

> ---
> 
> Anyone mind if I push this through the sparc tree?
> 

Makes sense since it's the only arch that appears affected.

> diff --git a/include/asm-generic/hugetlb.h b/include/asm-generic/hugetlb.h
> index d06079c..99b490b 100644
> --- a/include/asm-generic/hugetlb.h
> +++ b/include/asm-generic/hugetlb.h
> @@ -6,12 +6,12 @@ static inline pte_t mk_huge_pte(struct page *page, pgprot_t pgprot)
>  	return mk_pte(page, pgprot);
>  }
>  
> -static inline int huge_pte_write(pte_t pte)
> +static inline unsigned long huge_pte_write(pte_t pte)
>  {
>  	return pte_write(pte);
>  }
>  
> -static inline int huge_pte_dirty(pte_t pte)
> +static inline unsigned long huge_pte_dirty(pte_t pte)
>  {
>  	return pte_dirty(pte);
>  }
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

From: David Rientjes <rientjes@google.com>
Date: Wed, 2 Oct 2013 15:34:48 -0700 (PDT)

> On Wed, 2 Oct 2013, David Miller wrote:
> 
> Acked-by: David Rientjes <rientjes@google.com>
> 
>> ---
>> 
>> Anyone mind if I push this through the sparc tree?
>> 
> 
> Makes sense since it's the only arch that appears affected.

Ok, thanks for reviewing.
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

David Miller <davem@davemloft.net> wrote:

> The include/asm-generic/hugetlb.h stubs that just vector huge_pte_*()
> calls to the pte_*() implementations won't work in certain situations.
> 
> x86 and sparc, for example, return "unsigned long" from the bit
> checks, and just go "return pte_val(pte) & PTE_BIT_FOO;"
> 
> But since huge_pte_*() returns 'int', if any high bits on 64-bit are
> relevant, they get chopped off.
> 
> The net effect is that we can loop forever trying to COW a huge page,
> because the huge_pte_write() check signals false all the time.
> 
> Reported-by: Gurudas Pai <gurudas.pai@oracle.com>
> Signed-off-by: David S. Miller <davem@davemloft.net>

Acked-by: David Howells <dhowells@redhat.com>
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html