2023-07-19 07:01:11

by Mika Penttilä

Subject: Re: [PATCH] mm: hugetlb_vmemmap: use PageCompound() instead of PageReserved()

Hi,


On 19.7.2023 9.31, Muchun Song wrote:
> The ckeck of PageReserved() is easy to be broken in the future, PageCompound()
> is more stable to check if the page should be split.
>
> Signed-off-by: Muchun Song <[email protected]>
> ---
> mm/hugetlb_vmemmap.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
> index 4b9734777f69..8068fe890f52 100644
> --- a/mm/hugetlb_vmemmap.c
> +++ b/mm/hugetlb_vmemmap.c
> @@ -73,8 +73,10 @@ static int split_vmemmap_huge_pmd(pmd_t *pmd, unsigned long start)
> * be treated as indepdenent small pages (as they can be freed
> * individually).
> */
> - if (!PageReserved(head))
> + if (PageCompound(head)) {
> + VM_BUG_ON(compound_order(head) != get_order(PMD_SIZE));
> split_page(head, get_order(PMD_SIZE));

I think vmemmap pages are not compound pages (even order > 0).
They would bug on here trying to split_page() :


void split_page(struct page *page, unsigned int order)
{
int i;

VM_BUG_ON_PAGE(PageCompound(page), page);


> + }
>
> /* Make pte visible before pmd. See comment in pmd_install(). */
> smp_wmb();

--Mika

2023-07-19 07:13:32

by Muchun Song

Subject: Re: [PATCH] mm: hugetlb_vmemmap: use PageCompound() instead of PageReserved()

> On Jul 19, 2023, at 14:51, Mika Penttilä <[email protected]> wrote:
>
> Hi,
>
>
> On 19.7.2023 9.31, Muchun Song wrote:
>> The ckeck of PageReserved() is easy to be broken in the future, PageCompound()
>> is more stable to check if the page should be split.
>> Signed-off-by: Muchun Song <[email protected]>
>> ---
>> mm/hugetlb_vmemmap.c | 4 +++-
>> 1 file changed, 3 insertions(+), 1 deletion(-)
>> diff --git a/mm/hugetlb_vmemmap.c b/mm/hugetlb_vmemmap.c
>> index 4b9734777f69..8068fe890f52 100644
>> --- a/mm/hugetlb_vmemmap.c
>> +++ b/mm/hugetlb_vmemmap.c
>> @@ -73,8 +73,10 @@ static int split_vmemmap_huge_pmd(pmd_t *pmd, unsigned long start)
>> * be treated as indepdenent small pages (as they can be freed
>> * individually).
>> */
>> - if (!PageReserved(head))
>> + if (PageCompound(head)) {
>> + VM_BUG_ON(compound_order(head) != get_order(PMD_SIZE));
>> split_page(head, get_order(PMD_SIZE));
>
> I think vmemmap pages are not compound pages (even order > 0).
> They would bug on here trying to split_page() :

You are right. I have missed this. Now I remember why I use
PageReserved() instead of PageCompound() when I fist submit
the commit 39d35edee453. Thanks for your reminder.

Sorry for the noise. Please ignore this patch.

Thanks.

>
>
> void split_page(struct page *page, unsigned int order)
> {
> int i;
>
> VM_BUG_ON_PAGE(PageCompound(page), page);
>
>
>> + }
>> /* Make pte visible before pmd. See comment in pmd_install(). */
>> smp_wmb();
>
> --Mika
>