From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: In-Reply-To: <1171914157.18571.38.camel@localhost.localdomain> References: <1171867418.18571.3.camel@localhost.localdomain> <934E7F45-153D-4558-A774-FA5B6AB69E3C@watson.ibm.com> <1171914157.18571.38.camel@localhost.localdomain> Mime-Version: 1.0 (Apple Message framework v752.2) Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed Message-Id: <275B5FDA-C250-4506-B79A-9F99107BF06C@watson.ibm.com> From: Jimi Xenidis Subject: Re: [PATCH] powerpc: Introduce address space "slices" Date: Mon, 19 Feb 2007 14:57:37 -0500 To: Benjamin Herrenschmidt Cc: linuxppc-dev list , "cbe-oss-dev@ozlabs.org" , "Hollis R. Blanchard" List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Feb 19, 2007, at 2:42 PM, Benjamin Herrenschmidt wrote: > On Mon, 2007-02-19 at 08:23 -0500, Jimi Xenidis wrote: >> Hey Ben, >> is it your intention to eventually support "/memory@x/ >> ibm,expected#pages" and will these rules effect the kernel's >> linear map? >> My interest is with hypervisors that would like to restrict certain >> LMBs to 4k so we can do our nasty memory tricks, but still have most >> of the memory use large pages. > > The slices mecanism is strictly a mecanism for dealing with segment > constraints on userland VMA layout, so I don't think it will have any > impact either negative or positive on you. > > Limiting physical memory to different page sizes is a whole different > problem that I haven't though about (damn, that would be hard). Damn, I know, I was hoping you'd be taking care of it. :) -JX