linux-kernel - Re: [PATCH] Loongarch:Make pte/pmd_modify can set _PAGE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <b839a6a6-3791-ba73-baff-e860aa879bbc@loongson.cn>
Date: Wed, 5 Nov 2025 09:18:39 +0800
From: Bibo Mao <maobibo@...ngson.cn>
To: Huacai Chen <chenhuacai@...nel.org>,
 Tianyang Zhang <zhangtianyang@...ngson.cn>
Cc: kernel@...0n.name, akpm@...ux-foundation.org, willy@...radead.org,
 david@...hat.com, linmag7@...il.com, thuth@...hat.com, apopple@...dia.com,
 loongarch@...ts.linux.dev, linux-kernel@...r.kernel.org,
 Liupu Wang <wangliupu@...ngson.cn>
Subject: Re: [PATCH] Loongarch:Make pte/pmd_modify can set _PAGE_MODIFIED



On 2025/11/5 上午9:07, Huacai Chen wrote:
> On Wed, Nov 5, 2025 at 8:57 AM Tianyang Zhang <zhangtianyang@...ngson.cn> wrote:
>>
>> Hi, Huacai
>>
>> 在 2025/11/4 下午4:00, Huacai Chen 写道:
>>> Hi, Tianyang,
>>>
>>> The subject line can be:
>>> LoongArch: Let {pte,pmd}_modify() record the status of _PAGE_DIRTY (If
>>> I'm right in the later comments).
>> Ok. I got it
>>>
>>> On Tue, Nov 4, 2025 at 3:30 PM Tianyang Zhang <zhangtianyang@...ngson.cn> wrote:
>>>> In the current pte_modify operation, _PAGE_DIRTY might be cleared. Since
>>>> the hardware-page-walk does not have a predefined _PAGE_MODIFIED flag,
>>>> this could lead to loss of valid data in certain scenarios.
>>>>
>>>> The new modification involves checking whether the original PTE has the
>>>> _PAGE_DIRTY flag. If it exists, the _PAGE_MODIFIED bit is set, ensuring
>>>> that the pte_dirty interface can return accurate information.
>>> The description may be wrong here. Because pte_dirty() returns
>>> pte_val(pte) & (_PAGE_DIRTY | _PAGE_MODIFIED).
>>> If _PAGE_DIRTY isn't lost, pte_dirty() is always right, no matter
>>> whether there is or isn't _PAGE_MODIFIED.
>>>
>>> I think the real reason is we need to set _PAGE_MODIFIED in
>>> pte/pmd_modify to record the status of _PAGE_DIRTY, so that we can
>>> recover _PAGE_DIRTY afterwards, such as in pte/pmd_mkwrite().
>> Ok, I will adjust the description
> After some thinking, your original description may be right. Without
> this patch the scenario maybe like this:
> The pte is dirty _PAGE_DIRTY but without _PAGE_MODIFIED, after
> pte_modify() we lose _PAGE_DIRTY, then pte_dirty() returns false. So
> we need _PAGE_MODIFIED to record _PAGE_DIRTY here.
In theory pte_modify() is to modify RWX attribute. I think that it is a 
tricky to remove _PAGE_DIRTY and add _PAGE_MODIFIED with HW PTW system.

Also _PAGE_ACCESSED is lost with pte_modify() API, is there any 
influence with HW PTW system, or wait until possible problems coming out.

Regards
Bibo Mao
> 
> But the description also needs to be updated.
> 
>>>
>>>> Co-developed-by: Liupu Wang <wangliupu@...ngson.cn>
>>>> Signed-off-by: Liupu Wang <wangliupu@...ngson.cn>
>>>> Signed-off-by: Tianyang Zhang <zhangtianyang@...ngson.cn>
>>>> ---
>>>>    arch/loongarch/include/asm/pgtable.h | 17 +++++++++++++----
>>>>    1 file changed, 13 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/arch/loongarch/include/asm/pgtable.h b/arch/loongarch/include/asm/pgtable.h
>>>> index bd128696e96d..106abfa5183b 100644
>>>> --- a/arch/loongarch/include/asm/pgtable.h
>>>> +++ b/arch/loongarch/include/asm/pgtable.h
>>>> @@ -424,8 +424,13 @@ static inline unsigned long pte_accessible(struct mm_struct *mm, pte_t a)
>>>>
>>>>    static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
>>>>    {
>>>> -       return __pte((pte_val(pte) & _PAGE_CHG_MASK) |
>>>> -                    (pgprot_val(newprot) & ~_PAGE_CHG_MASK));
>>>> +       unsigned long val = (pte_val(pte) & _PAGE_CHG_MASK) |
>>>> +                    (pgprot_val(newprot) & ~_PAGE_CHG_MASK);
>>>> +
>>>> +       if (pte_val(pte) & _PAGE_DIRTY)
>>>> +               val |= _PAGE_MODIFIED;
>>>> +
>>>> +       return __pte(val);
>>>>    }
>>>>
>>>>    extern void __update_tlb(struct vm_area_struct *vma,
>>>> @@ -547,9 +552,13 @@ static inline struct page *pmd_page(pmd_t pmd)
>>>>
>>>>    static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
>>>>    {
>>>> -       pmd_val(pmd) = (pmd_val(pmd) & _HPAGE_CHG_MASK) |
>>>> +       unsigned long val = (pmd_val(pmd) & _HPAGE_CHG_MASK) |
>>>>                                   (pgprot_val(newprot) & ~_HPAGE_CHG_MASK);
>>>> -       return pmd;
>>>> +
>>>> +       if (pmd_val(pmd) & _PAGE_DIRTY)
>>>> +               val |= _PAGE_MODIFIED;
>>>> +
>>>> +       return __pmd(val);
>>>>    }
>>> A minimal modification can be:
>>> diff --git a/arch/loongarch/include/asm/pgtable.h
>>> b/arch/loongarch/include/asm/pgtable.h
>>> index 1f20e9280062..907ece0199e0 100644
>>> --- a/arch/loongarch/include/asm/pgtable.h
>>> +++ b/arch/loongarch/include/asm/pgtable.h
>>> @@ -448,8 +448,13 @@ static inline unsigned long pte_accessible(struct
>>> mm_struct *mm, pte_t a)
>>>
>>>    static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
>>>    {
>>> -       return __pte((pte_val(pte) & _PAGE_CHG_MASK) |
>>> -                    (pgprot_val(newprot) & ~_PAGE_CHG_MASK));
>>> +       pte_val(pte) = (pte_val(pte) & _PAGE_CHG_MASK) |
>>> +                       (pgprot_val(newprot) & ~_PAGE_CHG_MASK);
>>> +
>>> +       if (pte_val(pte) & _PAGE_DIRTY)
>>> +               pte_val(pte) |= _PAGE_MODIFIED;
>>> +
>>> +       return pte;
>>>    }
>>
>> +       pte_val(pte) = (pte_val(pte) & _PAGE_CHG_MASK) |
>> +                       (pgprot_val(newprot) & ~_PAGE_CHG_MASK);
>>
>> After this step, _PAGE_DIRTY may have already disappeared,
>> If no new variables are added, they can be modified in follow way:
>>
>>    static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
>>    {
>> +       if (pte_val(pte) & _PAGE_DIRTY)
>> +               pte_val(pte) |= _PAGE_MODIFIED;
>> +
>>          return __pte((pte_val(pte) & _PAGE_CHG_MASK) |
>>           (pgprot_val(newprot) & ~_PAGE_CHG_MASK));
>>
>>    }
> OK, it makes sense.
> 
> Huacai
>>
>>>
>>>    extern void __update_tlb(struct vm_area_struct *vma,
>>> @@ -583,7 +588,11 @@ static inline struct page *pmd_page(pmd_t pmd)
>>>    static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
>>>    {
>>>           pmd_val(pmd) = (pmd_val(pmd) & _HPAGE_CHG_MASK) |
>>> -                               (pgprot_val(newprot) & ~_HPAGE_CHG_MASK);
>>> +                       (pgprot_val(newprot) & ~_HPAGE_CHG_MASK);
>>> +
>>> +       if (pmd_val(pmd) & _PAGE_DIRTY)
>>> +               pmd_val(pmd) |= _PAGE_MODIFIED;
>>> +
>>>           return pmd;
>>>    }
>>>
>>> You needn't define a new variable.
>>>
>>>
>>> Huacai
>>>
>>>>    static inline pmd_t pmd_mkinvalid(pmd_t pmd)
>>>> --
>>>> 2.41.0
>>>>
>>>>
>> Thanks
>>
>> Tianyang
>>