ptep_test_and_clear_young() i386 vs. x86_64

ptep_test_and_clear_young() i386 vs. x86_64

[Slava Abramov]

Hi All!

Is there any special reason why ptep_test_and_clear_young()
implementation for x86_64 does not call pte_update()?

Compare the following:

include/asm-x86_64/pgtable.h:

static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, 
unsigned long addr, pte_t *ptep)
{
        if (!pte_young(*ptep))
               return 0;
        return test_and_clear_bit(_PAGE_BIT_ACCESSED, &ptep->pte);

}

include/asm-i386/pgtable.h:

#define ptep_test_and_clear_young(vma, addr, ptep) ({                   \
        int __ret = 0;                                                  \
        if (pte_young(*(ptep)))                                         \
                __ret = test_and_clear_bit(_PAGE_BIT_ACCESSED,          \
                                                &(ptep)->pte_low);       \
        if (__ret)                                                      \
                pte_update((vma)->vm_mm, addr, ptep);                   
     \
        __ret;                                                          \

})


Slava

Re: ptep_test_and_clear_young() i386 vs. x86_64

[Hugh Dickins]

On Fri, 21 Sep 2007, Slava Abramov wrote:
> 
> Is there any special reason why ptep_test_and_clear_young()
> implementation for x86_64 does not call pte_update()?

Because the i386 pte_update is there for CONFIG_PARAVIRT (and
a no-op without it), and CONFIG_PARAVIRT is so far peculiar to
i386: x86_64 doesn't even define a pte_update to call there.

Hugh