Messages in this thread | | | From | KOSAKI Motohiro <> | Subject | Re: [PATCH 0/7] Kill PF_MEMALLOC abuse | Date | Wed, 18 Nov 2009 14:55:51 +0900 (JST) |
| |
> On Tue, 2009-11-17 at 17:33 +0900, KOSAKI Motohiro wrote: > > > > if there is so such reason. we might need to implement another MM trick. > > but keeping this strage usage is not a option. All memory freeing activity > > (e.g. page out, task killing) need some memory. we need to protect its > > emergency memory. otherwise linux reliability decrease dramatically when > > the system face to memory stress. > > In general PF_MEMALLOC is a particularly bad idea, even for the VM when > not coupled with limiting the consumption. That is one should make an > upper-bound estimation of the memory needed for a writeout-path per > page, and reserve a small multiple thereof, and limit the number of > pages written out so as to never exceed this estimate. > > If the current mempool interface isn't sufficient (not hard to imagine), > look at the swap over NFS patch-set, that includes a much more able > reservation scheme, and accounting framework.
Yes, I agree.
In this discussion, some people explained why their subsystem need emergency memory, but nobody claim sharing memory pool against VM and surely want to stop reclaim (PF_MEMALLOC's big side effect).
OK. I try to review your patch carefully and remake this patch series on top your reservation framework in swap-over-nfs patch series.
| |