linux-kernel - Re: [PATCH v3 4/9] mm: introduce skip_none

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] [thread-next>] [day] [month] [year] [list]

Message-ID: <253e5fd0-7e98-43fd-b0d7-8a5b739ae4aa@bytedance.com>
Date: Fri, 15 Nov 2024 11:03:42 +0800
From: Qi Zheng <zhengqi.arch@...edance.com>
To: David Hildenbrand <david@...hat.com>
Cc: jannh@...gle.com, hughd@...gle.com, willy@...radead.org,
 muchun.song@...ux.dev, vbabka@...nel.org, akpm@...ux-foundation.org,
 peterx@...hat.com, mgorman@...e.de, catalin.marinas@....com,
 will@...nel.org, dave.hansen@...ux.intel.com, luto@...nel.org,
 peterz@...radead.org, x86@...nel.org, lorenzo.stoakes@...cle.com,
 linux-mm@...ck.org, linux-kernel@...r.kernel.org, zokeefe@...gle.com,
 rientjes@...gle.com
Subject: Re: [PATCH v3 4/9] mm: introduce skip_none_ptes()



On 2024/11/15 05:19, David Hildenbrand wrote:
> On 14.11.24 13:51, Qi Zheng wrote:
>>
>>
>> On 2024/11/14 20:32, David Hildenbrand wrote:
>>> On 14.11.24 10:20, Qi Zheng wrote:
>>>>
>>>>
>>>> On 2024/11/14 16:04, David Hildenbrand wrote:
>>>>>
>>>>>>     static unsigned long zap_pte_range(struct mmu_gather *tlb,
>>>>>>                     struct vm_area_struct *vma, pmd_t *pmd,
>>>>>>                     unsigned long addr, unsigned long end,
>>>>>> @@ -1682,13 +1704,17 @@ static unsigned long zap_pte_range(struct
>>>>>> mmu_gather *tlb,
>>>>>>             pte_t ptent = ptep_get(pte);
>>>>>>             int max_nr;
>>>>>> -        nr = 1;
>>>>>> -        if (pte_none(ptent))
>>>>>> -            continue;
>>>>>> -
>>>>>>             if (need_resched())
>>>>>>                 break;
>>>>>> +        nr = skip_none_ptes(pte, addr, end);
>>>>>> +        if (nr) {
>>>>>> +            addr += PAGE_SIZE * nr;
>>>>>> +            if (addr == end)
>>>>>> +                break;
>>>>>> +            pte += nr;
>>>>>> +        }
>>>>>> +
>>>>>>             max_nr = (end - addr) / PAGE_SIZE;
>>>>>
>>>>> I dislike calculating max_nr twice, once here and once in 
>>>>> skip_non_ptes.
>>>>>
>>>>> Further, you're missing to update ptent here.
>>>>
>>>> Oh, my bad. However, with [PATCH v3 5/9], there will be no problem, but
>>>> there are still two ptep_get() and max_nr calculation.
>>>>
>>>> If you inline it you can
>>>>> avoid another ptep_get().
>>>>
>>>> Do you mean to inline the skip_none_ptes() into do_zap_pte_range()?
>>>
>>> Effectively moving this patch after #5, and have it be something like:
>>>
>>> diff --git a/mm/memory.c b/mm/memory.c
>>> index 1949f5e0fece5..4f5d1e4c6688e 100644
>>> --- a/mm/memory.c
>>> +++ b/mm/memory.c
>>> @@ -1667,8 +1667,21 @@ static inline int do_zap_pte_range(struct
>>> mmu_gather *tlb,
>>>           pte_t ptent = ptep_get(pte);
>>>           int max_nr = (end - addr) / PAGE_SIZE;
>>>
>>> -       if (pte_none(ptent))
>>> -               return 1;
>>> +       /* Skip all consecutive pte_none(). */
>>> +       if (pte_none(ptent)) {
>>> +               int nr;
>>> +
>>> +               for (nr = 1; nr < max_nr; nr++) {
>>> +                       ptent = ptep_get(pte + nr);
>>> +                       if (!pte_none(ptent))
>>> +                               break;
>>> +               }
>>> +               max_nr -= nr;
>>> +               if (!max_nr)
>>> +                       return nr;
>>> +               pte += nr;
>>> +               addr += nr * PAGE_SIZE;
>>> +       }
>>>
>>>           if (pte_present(ptent))
>>>                   return zap_present_ptes(tlb, vma, pte, ptent, max_nr,
>>>
>>>
>>> In the context of this patch this makes most sense.
>>>
>>> Regarding "count_pte_none" comment, I assume you talk about patch #7.
>>
>> Yes.
>>
>>>
>>> Can't you simply return the number of pte_none that you skipped here
>>> using another
>>> input variable, if really required?
>>
>> Suppose we add an input variable nr_skip to do_zap_pte_range(), you mean
>> to return the above nr to zap_pte_range() through:
> 
> Maybe "cur_none_nr" or something similar.

OK.

> 
>>
>> *nr_skip = nr;
>>
>> and then:
>>
>> zap_pte_range
>> --> nr = do_zap_pte_range(tlb, vma, pte, addr, end, details, &skip_nr,
>>                          rss, &force_flush, &force_break);
>>       if (can_reclaim_pt) {
>>           none_nr += count_pte_none(pte, nr);
>>           none_nr += nr_skip;
>>       }
>>
>> Right?
> 
> Yes. I did not look closely at the patch that adds the counting of 

Got it.

> pte_none though (to digest why it is required :) ).

Because 'none_nr == PTRS_PER_PTE' is used in patch #7 to detect
empty PTE page.

Looking forward to your more review feedback on this series.

Thanks!

>