[PATCH] Loongarch:Fixed up panic cause by a NULL-pmd

[PATCH] Loongarch:Fixed up panic cause by a NULL-pmd

[Tianyang Zhang]

From: zhangtianyang <zhangtianyang@loongson.cn>

Fixes: bd51834d1cf6 ("LoongArch: Return NULL from huge_pte_offset() for invalid PMD")
ERROR INFO:

CPU 25 Unable to handle kernel paging request at virtual address 0x0
         ...
 Call Trace:
 [<900000000023c30c>] huge_pte_offset+0x3c/0x58
 [<900000000057fd4c>] hugetlb_follow_page_mask+0x74/0x438
 [<900000000051fee8>] __get_user_pages+0xe0/0x4c8
 [<9000000000522414>] faultin_page_range+0x84/0x380
 [<9000000000564e8c>] madvise_vma_behavior+0x534/0xa48
 [<900000000056689c>] do_madvise+0x1bc/0x3e8
 [<9000000000566df4>] sys_madvise+0x24/0x38
 [<90000000015b9e88>] do_syscall+0x78/0x98
 [<9000000000221f18>] handle_syscall+0xb8/0x158

In some cases, pmd may be NULL and rely on NULL as the return value
for processing, so it is necessary to determine this situation here
Signed-off-by: zhangtianyang <zhangtianyang@loongson.cn>
---
 arch/loongarch/mm/hugetlbpage.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/arch/loongarch/mm/hugetlbpage.c b/arch/loongarch/mm/hugetlbpage.c
index cea84d7f2b91..02dad4624fe3 100644
--- a/arch/loongarch/mm/hugetlbpage.c
+++ b/arch/loongarch/mm/hugetlbpage.c
@@ -47,7 +47,8 @@ pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr,
 				pmd = pmd_offset(pud, addr);
 		}
 	}
-	return pmd_none(pmdp_get(pmd)) ? NULL : (pte_t *) pmd;
+
+	return (!pmd || pmd_none(pmdp_get(pmd))) ? NULL : (pte_t *) pmd;
 }
 
 uint64_t pmd_to_entrylo(unsigned long pmd_val)
-- 
2.20.1

Re: [PATCH] Loongarch:Fixed up panic cause by a NULL-pmd

[Yanteng Si]

在 5/26/25 2:55 PM, Tianyang Zhang 写道:
> From: zhangtianyang <zhangtianyang@loongson.cn>
> 
> Fixes: bd51834d1cf6 ("LoongArch: Return NULL from huge_pte_offset() for invalid PMD")
> ERROR INFO:
> 
> CPU 25 Unable to handle kernel paging request at virtual address 0x0
>           ...
>   Call Trace:
>   [<900000000023c30c>] huge_pte_offset+0x3c/0x58
>   [<900000000057fd4c>] hugetlb_follow_page_mask+0x74/0x438
>   [<900000000051fee8>] __get_user_pages+0xe0/0x4c8
>   [<9000000000522414>] faultin_page_range+0x84/0x380
>   [<9000000000564e8c>] madvise_vma_behavior+0x534/0xa48
>   [<900000000056689c>] do_madvise+0x1bc/0x3e8
>   [<9000000000566df4>] sys_madvise+0x24/0x38
>   [<90000000015b9e88>] do_syscall+0x78/0x98
>   [<9000000000221f18>] handle_syscall+0xb8/0x158
> 

> In some cases, pmd may be NULL and rely on NULL as the return value
> for processing, so it is necessary to determine this situation here

Your description is a bit vague. Could you please specify
on which machines and in what scenarios this bug can be
reproduced? I believe such information should be included
in the commit message.

Thanks,
Yanteng

Re: [PATCH] Loongarch:Fixed up panic cause by a NULL-pmd

[Tianyang Zhang]

Hi, Yanteng

在 2025/5/27 下午1:47, Yanteng Si 写道:
> 在 5/26/25 2:55 PM, Tianyang Zhang 写道:
>> From: zhangtianyang <zhangtianyang@loongson.cn>
>>
>> Fixes: bd51834d1cf6 ("LoongArch: Return NULL from huge_pte_offset() 
>> for invalid PMD")
>> ERROR INFO:
>>
>> CPU 25 Unable to handle kernel paging request at virtual address 0x0
>>           ...
>>   Call Trace:
>>   [<900000000023c30c>] huge_pte_offset+0x3c/0x58
>>   [<900000000057fd4c>] hugetlb_follow_page_mask+0x74/0x438
>>   [<900000000051fee8>] __get_user_pages+0xe0/0x4c8
>>   [<9000000000522414>] faultin_page_range+0x84/0x380
>>   [<9000000000564e8c>] madvise_vma_behavior+0x534/0xa48
>>   [<900000000056689c>] do_madvise+0x1bc/0x3e8
>>   [<9000000000566df4>] sys_madvise+0x24/0x38
>>   [<90000000015b9e88>] do_syscall+0x78/0x98
>>   [<9000000000221f18>] handle_syscall+0xb8/0x158
>>
>
>> In some cases, pmd may be NULL and rely on NULL as the return value
>> for processing, so it is necessary to determine this situation here
>
> Your description is a bit vague. Could you please specify
> on which machines and in what scenarios this bug can be
> reproduced? I believe such information should be included
> in the commit message.
>
> Thanks,
> Yanteng

Sorry for the late reply.

I will provide more detailed information in the commit of the next patch

Thanks

Tianyang

Re: [PATCH] Loongarch:Fixed up panic cause by a NULL-pmd

[Huacai Chen]

On Thu, Jun 5, 2025 at 9:17 AM Tianyang Zhang <zhangtianyang@loongson.cn> wrote:
>
> Hi, Yanteng
>
> 在 2025/5/27 下午1:47, Yanteng Si 写道:
> > 在 5/26/25 2:55 PM, Tianyang Zhang 写道:
> >> From: zhangtianyang <zhangtianyang@loongson.cn>
> >>
> >> Fixes: bd51834d1cf6 ("LoongArch: Return NULL from huge_pte_offset()
> >> for invalid PMD")
> >> ERROR INFO:
> >>
> >> CPU 25 Unable to handle kernel paging request at virtual address 0x0
> >>           ...
> >>   Call Trace:
> >>   [<900000000023c30c>] huge_pte_offset+0x3c/0x58
> >>   [<900000000057fd4c>] hugetlb_follow_page_mask+0x74/0x438
> >>   [<900000000051fee8>] __get_user_pages+0xe0/0x4c8
> >>   [<9000000000522414>] faultin_page_range+0x84/0x380
> >>   [<9000000000564e8c>] madvise_vma_behavior+0x534/0xa48
> >>   [<900000000056689c>] do_madvise+0x1bc/0x3e8
> >>   [<9000000000566df4>] sys_madvise+0x24/0x38
> >>   [<90000000015b9e88>] do_syscall+0x78/0x98
> >>   [<9000000000221f18>] handle_syscall+0xb8/0x158
> >>
> >
> >> In some cases, pmd may be NULL and rely on NULL as the return value
> >> for processing, so it is necessary to determine this situation here
> >
> > Your description is a bit vague. Could you please specify
> > on which machines and in what scenarios this bug can be
> > reproduced? I believe such information should be included
> > in the commit message.
> >
> > Thanks,
> > Yanteng
>
> Sorry for the late reply.
>
> I will provide more detailed information in the commit of the next patch
I think this bug is not machine-specific, just provide something about
the test program.

Huacai

>
> Thanks
>
> Tianyang
>