Re: [PATCH v4 4/7] x86/tlb: fall back to flush all when meet a THP large page

[Alex Shi <alex.shi@intel.com>]

On 05/12/2012 12:28 AM, Andrea Arcangeli wrote:

> Hi,
> 
> On Thu, May 10, 2012 at 01:00:10PM +0800, Alex Shi wrote:
>> +	for (addr = start; addr <= end; addr += HPAGE_SIZE) {
>> +		pgd = pgd_offset(mm, addr);
>> +		if (likely(!pgd_none(*pgd))) {
>> +			pud = pud_offset(pgd, addr);
>> +			if (likely(!pud_none(*pud))) {
>> +				pmd = pmd_offset(pud, addr);
>> +				if (likely(!pmd_none(*pmd)))
>> +					if (pmd_large(*pmd))
>> +						return 1;
> 
> Note with THP we've to set the pmd temporarily not present during
> split_huge_page just before it returns a regular pmd (to avoid erratas
> and mixing 4k and 2m tlb on the same physical page). So pmd_large
> would fail in that small time window and we would do the invlpg loop
> (which should still work safe so no big deal). But I believe you could
> use pmd_trans_huge above, maybe gcc can drop the whole block with
> CONFIG_TRANSPARENT_HUGEPAGE=n.


Thanks the explanation.
Sure, if THP was not enabled, we can return 0 directly. :)