Re: [PATCH v5] hugetlbfs: Get unmapped area below TASK_UNMAPPED_BASE for hugetlbfs

[Hushijie <hushijie3@huawei.com>]

>On 5/16/20 12:47 AM, Hushijie wrote:
>>> On 5/14/20 7:31 AM, Shijie Hu wrote:
>>>> +	if (mm->get_unmapped_area == arch_get_unmapped_area)
>>>> +		return hugetlb_get_unmapped_area_bottomup(file, addr, len,
>>>> +				pgoff, flags);
>>>> +	return hugetlb_get_unmapped_area_topdown(file, addr, len,
>>>> +			pgoff, flags);
>>>
>>> I like this code using the value of mm->get_unmapped_area to determine
>>> which routine to call.  It is used by a few architectures.   However, I
>>> noticed that on at least one architecture (powerpc) mm->get_unmapped_area
>>> may be assigned to routines other than arch_get_unmapped_area or
>>> arch_get_unmapped_area_topdown.  In such a case, we would call the 'new'
>>> topdown routine.  I would prefer that we call the bottomup routine in this
>>> default case.
>>>
>>> In reality, this does not impact powerpc as that architecture has it's
>>> own hugetlb_get_unmapped_area routine.
>>>
>> 
>> Yes, I also noticed this before, powerpc uses radix__arch_get_unmapped_area*() 
>> when CONFIG_PPC_RADIX_MMU opened as 'y' and radix_enabled() returns 
>> true. However, powerpc implemented its own hugetlb_get_unmapped_area(). This
>> patch actually has no effect on powerpc.
>> 
>>> Because of this, I suggest we add a comment above this code and switch
>>> the if/else order.  For example,
>>>
>>> +       /*
>>> +        * Use mm->get_unmapped_area value as a hint to use topdown routine.
>>> +        * If architectures have special needs, they should define their own
>>> +        * version of hugetlb_get_unmapped_area.
>>> +        */
>>> +       if (mm->get_unmapped_area == arch_get_unmapped_area_topdown)
>>> +               return hugetlb_get_unmapped_area_topdown(file, addr, len,
>>> +                               pgoff, flags);
>>> +       return hugetlb_get_unmapped_area_bottomup(file, addr, len,
>>> +                       pgoff, flags);
>>>
>>> Thoughts?
>>> -- 
>>> Mike Kravetz
>>>
>> I agree with you. It's clever to switch the if/else order. If there is such
>> a case, mm->get_unmapped_area() is neihter arch_get_unmapped_area() nor
>> arch_get_unmapped_area_topdown(), it is indeed more appropriate to make the
>> bottomup routine as the default behavior.
>> 
>> May I put this code and comment you show above into patch v6 and add 
>> "Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>" to it?
>
>Feel free to add this code and my Signed-off-by.
>
>I assume this still works for your use case.  Correct?
>-- 
>Mike Kravetz
>

Yes, It still works for our use case.

Thanks for your replies and suggestions, I will submit patch v6 later.
--
Shijie Hu