[v2] mm/hugetlb: fix copy_hugetlb_page_range() to use ->pt_share_count

[PATCH v2] mm/hugetlb: fix copy_hugetlb_page_range() to use ->pt_share_count

Posted by Jane Chu 3 weeks, 1 day ago

commit 59d9094df3d79 ("mm: hugetlb: independent PMD page table shared count")
introduced ->pt_share_count dedicated to hugetlb PMD share count tracking,
but omitted fixing copy_hugetlb_page_range(), leaving the function relying on
page_count() for tracking that no longer works.

When lazy page table copy for hugetlb is disabled, that is, revert
commit bcd51a3c679d ("hugetlb: lazy page table copies in fork()")
fork()'ing with hugetlb PMD sharing quickly lockup -

[  239.446559] watchdog: BUG: soft lockup - CPU#75 stuck for 27s!
[  239.446611] RIP: 0010:native_queued_spin_lock_slowpath+0x7e/0x2e0
[  239.446631] Call Trace:
[  239.446633]  <TASK>
[  239.446636]  _raw_spin_lock+0x3f/0x60
[  239.446639]  copy_hugetlb_page_range+0x258/0xb50
[  239.446645]  copy_page_range+0x22b/0x2c0
[  239.446651]  dup_mmap+0x3e2/0x770
[  239.446654]  dup_mm.constprop.0+0x5e/0x230
[  239.446657]  copy_process+0xd17/0x1760
[  239.446660]  kernel_clone+0xc0/0x3e0
[  239.446661]  __do_sys_clone+0x65/0xa0
[  239.446664]  do_syscall_64+0x82/0x930
[  239.446668]  ? count_memcg_events+0xd2/0x190
[  239.446671]  ? syscall_trace_enter+0x14e/0x1f0
[  239.446676]  ? syscall_exit_work+0x118/0x150
[  239.446677]  ? arch_exit_to_user_mode_prepare.constprop.0+0x9/0xb0
[  239.446681]  ? clear_bhb_loop+0x30/0x80
[  239.446684]  ? clear_bhb_loop+0x30/0x80
[  239.446686]  entry_SYSCALL_64_after_hwframe+0x76/0x7e

There are two options to resolve the potential latent issue:
  1. warn against PMD sharing in copy_hugetlb_page_range(),
  2. fix it.
This patch opts for the second option.

Fixes: 59d9094df3d79 ("mm: hugetlb: independent PMD page table shared count")
Signed-off-by: Jane Chu <jane.chu@oracle.com>
Reviewed-by: Harry Yoo <harry.yoo@oracle.com>
---
 mm/hugetlb.c | 13 ++++---------
 1 file changed, 4 insertions(+), 9 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 753f99b4c718..8ca5b4f7805f 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -5594,18 +5594,13 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
 			break;
 		}
 
-		/*
-		 * If the pagetables are shared don't copy or take references.
-		 *
-		 * dst_pte == src_pte is the common case of src/dest sharing.
-		 * However, src could have 'unshared' and dst shares with
-		 * another vma. So page_count of ptep page is checked instead
-		 * to reliably determine whether pte is shared.
-		 */
-		if (page_count(virt_to_page(dst_pte)) > 1) {
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
+		/* If the pagetables are shared don't copy or take references. */
+		if (ptdesc_pmd_pts_count(virt_to_ptdesc(dst_pte)) > 0) {
 			addr |= last_addr_mask;
 			continue;
 		}
+#endif
 
 		dst_ptl = huge_pte_lock(h, dst, dst_pte);
 		src_ptl = huge_pte_lockptr(h, src, src_pte);
-- 
2.43.5

Re: [PATCH v2] mm/hugetlb: fix copy_hugetlb_page_range() to use ->pt_share_count

Posted by David Hildenbrand 3 weeks ago

On 10.09.25 21:27, Jane Chu wrote:
> commit 59d9094df3d79 ("mm: hugetlb: independent PMD page table shared count")
> introduced ->pt_share_count dedicated to hugetlb PMD share count tracking,
> but omitted fixing copy_hugetlb_page_range(), leaving the function relying on
> page_count() for tracking that no longer works.
> 
> When lazy page table copy for hugetlb is disabled, that is, revert
> commit bcd51a3c679d ("hugetlb: lazy page table copies in fork()")
> fork()'ing with hugetlb PMD sharing quickly lockup -
> 
> [  239.446559] watchdog: BUG: soft lockup - CPU#75 stuck for 27s!
> [  239.446611] RIP: 0010:native_queued_spin_lock_slowpath+0x7e/0x2e0
> [  239.446631] Call Trace:
> [  239.446633]  <TASK>
> [  239.446636]  _raw_spin_lock+0x3f/0x60
> [  239.446639]  copy_hugetlb_page_range+0x258/0xb50
> [  239.446645]  copy_page_range+0x22b/0x2c0
> [  239.446651]  dup_mmap+0x3e2/0x770
> [  239.446654]  dup_mm.constprop.0+0x5e/0x230
> [  239.446657]  copy_process+0xd17/0x1760
> [  239.446660]  kernel_clone+0xc0/0x3e0
> [  239.446661]  __do_sys_clone+0x65/0xa0
> [  239.446664]  do_syscall_64+0x82/0x930
> [  239.446668]  ? count_memcg_events+0xd2/0x190
> [  239.446671]  ? syscall_trace_enter+0x14e/0x1f0
> [  239.446676]  ? syscall_exit_work+0x118/0x150
> [  239.446677]  ? arch_exit_to_user_mode_prepare.constprop.0+0x9/0xb0
> [  239.446681]  ? clear_bhb_loop+0x30/0x80
> [  239.446684]  ? clear_bhb_loop+0x30/0x80
> [  239.446686]  entry_SYSCALL_64_after_hwframe+0x76/0x7e
> 
> There are two options to resolve the potential latent issue:
>    1. warn against PMD sharing in copy_hugetlb_page_range(),
>    2. fix it.
> This patch opts for the second option.
> 
> Fixes: 59d9094df3d79 ("mm: hugetlb: independent PMD page table shared count")
> Signed-off-by: Jane Chu <jane.chu@oracle.com>
> Reviewed-by: Harry Yoo <harry.yoo@oracle.com>
> ---
>   mm/hugetlb.c | 13 ++++---------
>   1 file changed, 4 insertions(+), 9 deletions(-)
> 
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 753f99b4c718..8ca5b4f7805f 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -5594,18 +5594,13 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
>   			break;
>   		}
>   
> -		/*
> -		 * If the pagetables are shared don't copy or take references.
> -		 *
> -		 * dst_pte == src_pte is the common case of src/dest sharing.
> -		 * However, src could have 'unshared' and dst shares with
> -		 * another vma. So page_count of ptep page is checked instead
> -		 * to reliably determine whether pte is shared.
> -		 */

I think you ignored my question to v1 regarding the change of comment.

-- 
Cheers

David / dhildenb

Re: [PATCH v2] mm/hugetlb: fix copy_hugetlb_page_range() to use ->pt_share_count

Posted by jane.chu@oracle.com 3 weeks ago


On 9/11/2025 1:17 AM, David Hildenbrand wrote:
[..]
>>
>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>> index 753f99b4c718..8ca5b4f7805f 100644
>> --- a/mm/hugetlb.c
>> +++ b/mm/hugetlb.c
>> @@ -5594,18 +5594,13 @@ int copy_hugetlb_page_range(struct mm_struct 
>> *dst, struct mm_struct *src,
>>               break;
>>           }
>> -        /*
>> -         * If the pagetables are shared don't copy or take references.
>> -         *
>> -         * dst_pte == src_pte is the common case of src/dest sharing.
>> -         * However, src could have 'unshared' and dst shares with
>> -         * another vma. So page_count of ptep page is checked instead
>> -         * to reliably determine whether pte is shared.
>> -         */
> 
> I think you ignored my question to v1 regarding the change of comment.
> 

Sorry David, didn't mean disrespect, I missed your earlier comments.
I just replied.

Thanks,
-jane

Re: [PATCH v2] mm/hugetlb: fix copy_hugetlb_page_range() to use ->pt_share_count

Posted by David Hildenbrand 2 weeks, 6 days ago

On 11.09.25 21:58, jane.chu@oracle.com wrote:
> 
> 
> On 9/11/2025 1:17 AM, David Hildenbrand wrote:
> [..]
>>>
>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>> index 753f99b4c718..8ca5b4f7805f 100644
>>> --- a/mm/hugetlb.c
>>> +++ b/mm/hugetlb.c
>>> @@ -5594,18 +5594,13 @@ int copy_hugetlb_page_range(struct mm_struct
>>> *dst, struct mm_struct *src,
>>>                break;
>>>            }
>>> -        /*
>>> -         * If the pagetables are shared don't copy or take references.
>>> -         *
>>> -         * dst_pte == src_pte is the common case of src/dest sharing.
>>> -         * However, src could have 'unshared' and dst shares with
>>> -         * another vma. So page_count of ptep page is checked instead
>>> -         * to reliably determine whether pte is shared.
>>> -         */
>>
>> I think you ignored my question to v1 regarding the change of comment.
>>
> 
> Sorry David, didn't mean disrespect, I missed your earlier comments.

No worries. Replied there about possibly simplifying the comment further.

BTW, I think the code would be even clearer if we would have

ptdesc_pmd_is_shared()

that is simply a wrapper around the ptdesc_pmd_pts_count() check.

-- 
Cheers

David / dhildenb

Re: [PATCH v2] mm/hugetlb: fix copy_hugetlb_page_range() to use ->pt_share_count

Posted by jane.chu@oracle.com 2 weeks, 3 days ago


On 9/12/2025 12:32 AM, David Hildenbrand wrote:
> On 11.09.25 21:58, jane.chu@oracle.com wrote:
>>
>>
>> On 9/11/2025 1:17 AM, David Hildenbrand wrote:
>> [..]
>>>>
>>>> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
>>>> index 753f99b4c718..8ca5b4f7805f 100644
>>>> --- a/mm/hugetlb.c
>>>> +++ b/mm/hugetlb.c
>>>> @@ -5594,18 +5594,13 @@ int copy_hugetlb_page_range(struct mm_struct
>>>> *dst, struct mm_struct *src,
>>>>                break;
>>>>            }
>>>> -        /*
>>>> -         * If the pagetables are shared don't copy or take references.
>>>> -         *
>>>> -         * dst_pte == src_pte is the common case of src/dest sharing.
>>>> -         * However, src could have 'unshared' and dst shares with
>>>> -         * another vma. So page_count of ptep page is checked instead
>>>> -         * to reliably determine whether pte is shared.
>>>> -         */
>>>
>>> I think you ignored my question to v1 regarding the change of comment.
>>>
>>
>> Sorry David, didn't mean disrespect, I missed your earlier comments.
> 
> No worries. Replied there about possibly simplifying the comment further.
> 
> BTW, I think the code would be even clearer if we would have
> 
> ptdesc_pmd_is_shared()
> 
> that is simply a wrapper around the ptdesc_pmd_pts_count() check.

Agreed, will add ptdesc_pmd_is_shared().

thanks a lot!

-jane>