Re: [PATCH mm] fix swapoff breakage; however...

[Balbir Singh]

Hugh Dickins wrote:
> On Tue, 18 Sep 2007, Balbir Singh wrote:
>> Hugh Dickins wrote:
>>> More fundamentally, it looks like any container brought over its limit in
>>> unuse_pte will abort swapoff: that doesn't doesn't seem "contained" to me.
>>> Maybe unuse_pte should just let containers go over their limits without
>>> error?  Or swap should be counted along with RSS?  Needs reconsideration.
>> Thanks, for the catching this. There are three possible solutions
>>
>> 1. Account each RSS page with a probable swap cache page, double
>>    the RSS accounting to ensure that swapoff will not fail.
>> 2. Account for the RSS page just once, do not account swap cache
>>    pages
> 
> Neither of those makes sense to me, but I may be misunderstanding.
> 
> What would make sense is (what I meant when I said swap counted
> along with RSS) not to count pages out and back in as they are
> go out to swap and back in, just keep count of instantiated pages
> 

I am not sure how you define instantiated pages. I suspect that
you mean RSS + pages swapped out (swap_pte)?

> I say "make sense" meaning that the numbers could be properly
> accounted; but it may well be unpalatable to treat fast RAM as
> equal to slow swap.
> 
>> 3. Follow your suggestion and let containers go over their limits
>>    without error
>>
>> With the current approach, a container over it's limit will not
>> be able to call swapoff successfully, is that bad?
> 
> That's not so bad.  What's bad is that anyone else with the
> CAP_SYS_ADMIN to swapoff is liable to be prevented by containers
> going over their limits.
> 

If a swapoff is going to push a container over it's limit, then
we break the container and the isolation it provides. Upon swapoff
failure, may be we could get the container to print a nice
little warning so that anyone else with CAP_SYS_ADMIN can fix the
container limit and retry swapoff.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL
-
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/