Re: [PATCH] hugetlb: simplify hugetlb handling in follow_page_mask

[Baolin Wang <baolin.wang@linux.alibaba.com>]

On 9/1/2022 2:59 PM, David Hildenbrand wrote:
> On 01.09.22 03:24, Baolin Wang wrote:
>>
>>
>> On 9/1/2022 8:00 AM, Mike Kravetz wrote:
>>> On 08/31/22 09:07, Baolin Wang wrote:
>>>>
>>>>
>>>> On 8/31/2022 2:39 AM, Mike Kravetz wrote:
>>>>> On 08/30/22 09:44, Mike Kravetz wrote:
>>>>>> On 08/30/22 09:06, Baolin Wang wrote:
>>>>>>> Hi Mike,
>>>>>>>
>>>>>>> On 8/30/2022 7:40 AM, Mike Kravetz wrote:
>>>>>>>> During discussions of this series [1], it was suggested that hugetlb
>>>>>>>> handling code in follow_page_mask could be simplified.  At the beginning
>>>>>>>> of follow_page_mask, there currently is a call to follow_huge_addr which
>>>>>>>> 'may' handle hugetlb pages.  ia64 is the only architecture which provides
>>>>>>>> a follow_huge_addr routine that does not return error.  Instead, at each
>>>>>>>> level of the page table a check is made for a hugetlb entry.  If a hugetlb
>>>>>>>> entry is found, a call to a routine associated with that entry is made.
>>>>>>>>
>>>>>>>> Currently, there are two checks for hugetlb entries at each page table
>>>>>>>> level.  The first check is of the form:
>>>>>>>> 	if (p?d_huge())
>>>>>>>> 		page = follow_huge_p?d();
>>>>>>>> the second check is of the form:
>>>>>>>> 	if (is_hugepd())
>>>>>>>> 		page = follow_huge_pd().
>>>>>>>>
>>>>>>>> We can replace these checks, as well as the special handling routines
>>>>>>>> such as follow_huge_p?d() and follow_huge_pd() with a single routine to
>>>>>>>> handle hugetlb vmas.
>>>>>>>>
>>>>>>>> A new routine hugetlb_follow_page_mask is called for hugetlb vmas at the
>>>>>>>> beginning of follow_page_mask.  hugetlb_follow_page_mask will use the
>>>>>>>> existing routine huge_pte_offset to walk page tables looking for hugetlb
>>>>>>>> entries.  huge_pte_offset can be overwritten by architectures, and already
>>>>>>>> handles special cases such as hugepd entries.
>>>>>>>
>>>>>>> Could you also mention that this patch will fix the lock issue for
>>>>>>> CONT-PTE/PMD hugetlb by changing to use huge_pte_lock()? which will help
>>>>>>> people to understand the issue.
>>>>>>
>>>>>> Will update message in v2.  Thanks for taking a look!
>>>>>>
>>>>>
>>>>> One additional thought, we 'may' need a separate patch to fix the locking
>>>>> issues that can be easily backported.  Not sure this 'simplification' is
>>>>> a good backport candidate.
>>>>
>>>> Yes, that was my thought before, but David did not like adding more
>>>> make-legacy-cruft-happy code.
>>>>
>>>> So how about creating a series that contains 3 patches: picking up patch 1
>>>> and patch 3 of my previous series [1], and your current patch? That means
>>>> patch 1 and patch 2 in this series can fix the lock issue explicitly and be
>>>> suitable to backport, meanwhile patch 3 (which is your current patch) will
>>>> cleanup the legacy code.
>>>>
>>>
>>> When I looked at patch 3, I was thinking the update follow_huge_pmd routine
>>> would work for the PTE level with a few more modifications.  Perhaps, this is
>>> too ugly but it is a smaller set of changes for backport.
>>>
>>> Of course, this would be followed up with the simplification patch which
>>> removes all this code.
>>
>> Yes, looks more simple. I can send you a formal patch with your
>> suggestion, which can be added into your cleanup series. Thanks.
> 
> As an alternative, we can have a stable-only version that does that.

But from stable-kernel-rules, we should follow "It or an equivalent fix 
must already exist in Linus' tree (upstream)."