Re: [PATCH] proc: mm: export PTE sizes directly in smaps (v2)

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 11/24/2016 08:00 PM, Anshuman Khandual wrote:
...
>> The current mechanisms work fine when we have one or two page sizes.
>> But, they start to get a bit muddled when we mix page sizes inside
>> one VMA.  For instance, the DAX folks were proposing adding a set of
>> fields like:
> 
> So DAX is only case which creates this scenario of multi page sizes in
> the same VMA ? Is there any cases other than DAX mapping ?

Both file and anonymous huge pages.  No other ones in the core VM that I
can think of.

>> 	DevicePages:
>> 	DeviceHugePages:
>> 	DeviceGiganticPages:
>> 	DeviceGinormousPages:
> 
> I guess these are the page sizes supported at PTE, PMD, PUD, PGD level.
> Are all these page sizes supported right now or we are just creating
> place holder for future.

I know there are patches for PUD level support in DAX, but I don't think
they're merged yet.  There is definitely *not* support for PGD level
since we don't have such support in hardware on x86 as far as I know.

>> SwapPss:               0 kB
>> KernelPageSize:        4 kB
>> MMUPageSize:           4 kB
>> Locked:                0 kB
>> Ptes@4kB:	      32 kB
>> Ptes@2MB:	    2048 kB
> 
> So in the left column we are explicitly indicating the size of the PTE
> and expect the user to figure out where it can really be either at PTE,
> PMD, PUD etc. Thats little bit different that 'AnonHugePages' or the
> Shared_HugeTLB/Private_HugeTLB pages which we know are the the PMD/PUD
> level.

Yeah, it's a little different from what we have.

>> The format I used here should be unlikely to break smaps parsers
>> unless they're looking for "kB" and now match the 'Ptes@4kB' instead
>> of the one at the end of the line.
> 
> Right. So you are dropping the idea to introduce these fields as you
> mentioned before for DAX mappings.
> 
>  	DevicePages:
>  	DeviceHugePages:
>  	DeviceGiganticPages:
>  	DeviceGinormousPages:

Right.  We don't need those if we have this patch.

>>  	if (page) {
>>  		int mapcount = page_mapcount(page);
>> +		unsigned long hpage_size = huge_page_size(hstate_vma(vma));
>>
>> +		mss->rss_pud += hpage_size;
>>  		if (mapcount >= 2)
>> -			mss->shared_hugetlb += huge_page_size(hstate_vma(vma));
>> +			mss->shared_hugetlb += hpage_size;
>>  		else
>> -			mss->private_hugetlb += huge_page_size(hstate_vma(vma));
>> +			mss->private_hugetlb += hpage_size;
>>  	}
>>  	return 0;
> 
> Hmm, is this related to these new changes ? The replacement of 'hpage_size'
> instead of huge_page_size(hstate_vma(vma)) can be done in a separate patch.

Yes, this is theoretically unrelated, but I'm not breaking this 3-line
change up into a different patch unless there's a pretty good reason reason.

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>



[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]