Re: [RESEND PATCH v3 1/2] mm: Introducing arch_remap hook

[Laurent Dufour]

On 13/04/2015 16:02, Kirill A. Shutemov wrote:
> On Mon, Apr 13, 2015 at 04:35:21PM +0300, Pavel Emelyanov wrote:
>> On 04/13/2015 04:21 PM, Laurent Dufour wrote:
>>> On 13/04/2015 15:13, Kirill A. Shutemov wrote:
>>>> On Mon, Apr 13, 2015 at 02:41:22PM +0200, Laurent Dufour wrote:
>>>>> On 13/04/2015 13:58, Kirill A. Shutemov wrote:
>>>>>> On Mon, Apr 13, 2015 at 11:56:27AM +0200, Laurent Dufour wrote:
>>>>>>> Some architecture would like to be triggered when a memory area is moved
>>>>>>> through the mremap system call.
>>>>>>>
>>>>>>> This patch is introducing a new arch_remap mm hook which is placed in the
>>>>>>> path of mremap, and is called before the old area is unmapped (and the
>>>>>>> arch_unmap hook is called).
>>>>>>>
>>>>>>> The architectures which need to call this hook should define
>>>>>>> __HAVE_ARCH_REMAP in their asm/mmu_context.h and provide the arch_remap
>>>>>>> service with the following prototype:
>>>>>>> void arch_remap(struct mm_struct *mm,
>>>>>>>                 unsigned long old_start, unsigned long old_end,
>>>>>>>                 unsigned long new_start, unsigned long new_end);
>>>>>>>
>>>>>>> Signed-off-by: Laurent Dufour <ldufour@xxxxxxxxxxxxxxxxxx>
>>>>>>> Reviewed-by: Ingo Molnar <mingo@xxxxxxxxxx>
>>>>>>> ---
>>>>>>>  mm/mremap.c | 19 +++++++++++++------
>>>>>>>  1 file changed, 13 insertions(+), 6 deletions(-)
>>>>>>>
>>>>>>> diff --git a/mm/mremap.c b/mm/mremap.c
>>>>>>> index 2dc44b1cb1df..009db5565893 100644
>>>>>>> --- a/mm/mremap.c
>>>>>>> +++ b/mm/mremap.c
>>>>>>> @@ -25,6 +25,7 @@
>>>>>>>  
>>>>>>>  #include <asm/cacheflush.h>
>>>>>>>  #include <asm/tlbflush.h>
>>>>>>> +#include <asm/mmu_context.h>
>>>>>>>  
>>>>>>>  #include "internal.h"
>>>>>>>  
>>>>>>> @@ -286,13 +287,19 @@ static unsigned long move_vma(struct vm_area_struct *vma,
>>>>>>>  		old_len = new_len;
>>>>>>>  		old_addr = new_addr;
>>>>>>>  		new_addr = -ENOMEM;
>>>>>>> -	} else if (vma->vm_file && vma->vm_file->f_op->mremap) {
>>>>>>> -		err = vma->vm_file->f_op->mremap(vma->vm_file, new_vma);
>>>>>>> -		if (err < 0) {
>>>>>>> -			move_page_tables(new_vma, new_addr, vma, old_addr,
>>>>>>> -					 moved_len, true);
>>>>>>> -			return err;
>>>>>>> +	} else {
>>>>>>> +		if (vma->vm_file && vma->vm_file->f_op->mremap) {
>>>>>>> +			err = vma->vm_file->f_op->mremap(vma->vm_file, new_vma);
>>>>>>> +			if (err < 0) {
>>>>>>> +				move_page_tables(new_vma, new_addr, vma,
>>>>>>> +						  old_addr, moved_len, true);
>>>>>>> +				return err;
>>>>>>> +			}
>>>>>>>  		}
>>>>>>> +#ifdef __HAVE_ARCH_REMAP
>>>>>>
>>>>>> It would be cleaner to provide dummy arch_remap() for !__HAVE_ARCH_REMAP
>>>>>> in some generic header.
>>>>>
>>>>> The idea was to not impact all the architectures as arch_unmap(),
>>>>> arch_dup_mmap() or arch_exit_mmap() implies.
>>>>>
>>>>> I look at the headers where such a dummy arch_remap could be put but I
>>>>> can't figure out one which will not impact all the architecture.
>>>>> What about defining a dummy service earlier in mm/remap.c in the case
>>>>> __HAVE_ARCH_REMAP is not defined ?
>>>>> Something like :
>>>>> #ifndef __HAVE_ARCH_REMAP
>>>>> static inline void void arch_remap(struct mm_struct *mm,
>>>>>                                    unsigned long old_start,
>>>>>                                    unsigned long old_end,
>>>>>                                    unsigned long new_start,
>>>>>                                    unsigned long new_end)
>>>>> {
>>>>> }
>>>>> #endif
>>>>
>>>> Or just #define arch_remap(...) do { } while (0)
>>>>
>>>
>>> I guessed you wanted the arch_remap() prototype to be exposed somewhere
>>> in the code.
>>>
>>> To be honest, I can't find the benefit of defining a dummy arch_remap()
>>> in mm/remap.c if __HAVE_ARCH_REMAP is not defined instead of calling it
>>> in move_vma if __HAVE_ARCH_REMAP is defined.
>>> Is it really what you want ?
>>
>> I think Kirill meant something like e.g. the arch_enter_lazy_mmu_mode()
>> is implemented and called in mm/mremap.c -- the "generic" part is in the
>> include/asm-generic/pgtable.h and those architectures willing to have
>> their own implementation are in arch/$arch/...
>>
>> Kirill, if I'm right with it, can you suggest the header where to put
>> the "generic" mremap hook's (empty) body?
> 
> I initially thought it would be enough to put it into
> <asm-generic/mmu_context.h>, expecting it works as
> <asm-generic/pgtable.h>. But that's not the case.
> 
> It probably worth at some point rework all <asm/mmu_context.h> to include
> <asm-generic/mmu_context.h> at the end as we do for <asm/pgtable.h>.
> But that's outside the scope of the patchset, I guess.
> 
> I don't see any better candidate for such dummy header. :-/

Clearly, I'm not confortable with a rewrite of <asm/mmu_context.h> :(

So what about this patch, is this v3 acceptable ?

Cheers,
Laurent.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/