Re: [PATCH mm-new v2 1/1] mm/khugepaged: guard is_zero_pfn() calls with pte_present()

[Lance Yang <lance.yang@linux.dev>]

On 2025/10/17 22:51, David Hildenbrand wrote:
> On 17.10.25 11:38, Lance Yang wrote:
>> From: Lance Yang <lance.yang@linux.dev>
>>
>> A non-present entry, like a swap PTE, contains completely different data
>> (swap type and offset). pte_pfn() doesn't know this, so if we feed it a
>> non-present entry, it will spit out a junk PFN.
>>
>> What if that junk PFN happens to match the zeropage's PFN by sheer
>> chance? While really unlikely, this would be really bad if it did.
>>
>> So, let's fix this potential bug by ensuring all calls to is_zero_pfn()
>> in khugepaged.c are properly guarded by a pte_present() check.
>>
>> Suggested-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
>> Reviewed-by: Dev Jain <dev.jain@arm.com>
>> Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
>> Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
>> Signed-off-by: Lance Yang <lance.yang@linux.dev>
>> ---
>> Applies against commit 0f22abd9096e in mm-new.
>>
>> v1 -> v2:
>>   - Collect Reviewed-by from Dev, Wei and Baolin - thanks!
>>   - Reduce a level of indentation (per Dev)
>>   - https://lore.kernel.org/linux-mm/20251016033643.10848-1- 
>> lance.yang@linux.dev/
>>
>>   mm/khugepaged.c | 29 ++++++++++++++++-------------
>>   1 file changed, 16 insertions(+), 13 deletions(-)
>>
>> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
>> index d635d821f611..648d9335de00 100644
>> --- a/mm/khugepaged.c
>> +++ b/mm/khugepaged.c
>> @@ -516,7 +516,7 @@ static void release_pte_pages(pte_t *pte, pte_t 
>> *_pte,
>>           pte_t pteval = ptep_get(_pte);
>>           unsigned long pfn;
>> -        if (pte_none(pteval))
>> +        if (!pte_present(pteval))
>>               continue;
> 
> 
> Isn't it rather that if we would ever get a !pte_none() && ! 
> pte_present() here, something would be deeply flawed?
> 
> I'd much rather spell that out and do here
> 
> VM_WARN_ON_ONCE(!pte_present(pteval));
> 
> keeping the original check.

Right, it's much better to be loud with a VM_WARN if we see
a weird PTE, as Dev also suggested :)

> 
> 
>>           pfn = pte_pfn(pteval);
>>           if (is_zero_pfn(pfn))
>> @@ -690,17 +690,18 @@ static void 
>> __collapse_huge_page_copy_succeeded(pte_t *pte,
>>            address += nr_ptes * PAGE_SIZE) {
>>           nr_ptes = 1;
>>           pteval = ptep_get(_pte);
>> -        if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
>> +        if (pte_none(pteval) ||
>> +            (pte_present(pteval) && is_zero_pfn(pte_pfn(pteval)))) {
> 
> This now seems to be a common pattern now :)
> 
> 
> Should we have a simple helper
> 
> static inline void pte_none_or_zero(pte_t pte)
> {
>      if (pte_none(pte))
>          return true;
>      return pte_present(pte) && is_zero_pfn(pte_pfn(pte)
> }
> 
> initially maybe local to this file?

And yeah, that logic is crying out for a new helper.

Thanks!