[PATCH 4/7] Store huge page linux pte in mm_struct

[bill4carson]

On 2012年02月01日 13:45, bill4carson wrote:
>
>
> On 2012年01月31日 18:01, Russell King - ARM Linux wrote:
>> On Mon, Jan 30, 2012 at 03:57:15PM +0800, bill4carson at gmail.com wrote:
>>> From: Bill Carson<bill4carson at gmail.com>
>>>
>>> One easy way to store huge page linux pte is mm_struct instead of
>>> thread_info
>>> that's because when parent task with huge page VMA calls fork, parent
>>> huge page
>>> pagetable entries are copied into child pagetable. This is done in
>>>
>>> int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct
>>> *src,
>>> struct vm_area_struct *vma)
>>>
>>> We cannot derive child's thread_info just using struct mm_struct *dst.
>>> if we have struct mm_struct **dst, then it's easy to find the
>>> corresponding
>>> task_struct as well as thread_info, but we only get struct mm_struct
>>> *dst.
>>> It's possible to find the desired task_struct by iterating the global
>>> task list
>>> by comparing task_struct->mm with dst.
>>> So mm_struct is used for huge page linux pte for faster lookup and
>>> efficient.
>> I really do not understand this description, and it doesn't seem to tie
>> up with the code. What problem are you trying to solve here?
>>
>> Note that a mm_struct can be shared between multiple task_structs, so
>> if your thinking is that something in the mm_struct or page table needs
>> to know about a task_struct, you're ideas are wrong.
>>
> Normal page based pte has hardware version and linux version, these two
> kinds of
> ptes occupy half of 4K page, and each of two pmd level entry pointer to
> this half 4K
> page.
>
> For huge page, mappings only exist in pmd level from hardware
> perspective, however
> mm subsystem also needs to know the linux version *pte* of this huge
> page based
> mapping, problems is where to store these huge page linux pte?
>
> A:
> Store huge page linux pte in mm_struct
>
> B:
> Store huge page linux pte in mm_context_t suggested by Catalin.
> This is almost like option A, while it's nice to modify arm code only
> and take small effort to make it happen.
>
> C:
> Modify pgd_alloc to allocate 2048 pgd_t entries plus two void pointers
> Use these two additional pointers to store huge page linux pte.
> It's feasible, but I don't know whether this is a good idea.
>
> Russelll and Catalin:
>
> Could you please give your advice on this?
>
>

Hi, Russell

I didn't mean to derive task_struct from a mm_struct, as you said,
multiple tasks could possible share one mm_struct.

What I did is to store huge page linux pte on mm_struct(or
mm_context_t), cause all hugetlb up layer hooks have mm_struct in
parameters. IMHO, that's what I could figure out right now, if this
idea sounds wrong or silly, please let me know.

And I'm wondering could you please review the other parts of this patch?


thanks


-- 
I am a slow learner
but I will keep trying to fight for my dreams!

--bill