From: BlaisorBlade <bla...@ya...> - 2004-04-11 16:42:14
|
Alle 16:50, sabato 10 aprile 2004, Michael Koehne ha scritto: > Moin Jeff Dike, > > > On Sat, Apr 10, 2004 at 01:06:51PM +0200, Sven K=F6hler wrote: > > > is there any way for the UML-kernel to allocate a non-swappable > > > memore-region in the host's memory? this could be especially important > > > for dedicated host-machines that are only to run UMLs. > > > > What I am going to do is make it possible to manage the host's memory > > so that it doesn't swap, and the UMLs in effect are mlocked. > > this reminds me on VM's V=3DR nucleus option. A V=3DR is a memory parti= tion > dedicated to 1 user, e.g. an MVS guest running under VM who is doing its > own demand paging. To translate this to Linux could be the following : > > e.g. a system with 1GB real ram could boot `linux mem=3D256m` with > only a quarter of ram for the hosting system. We now need two tricks: > > 1: add 768m of ram as V=3DR memory (dont use it for Linux processes, > buffers, ...) partition and offer it as /dev/v=3Dr. (this would > be a host kernel patch) > > 2: use this V=3DR region for UML `linux mem=3D256m@256m` (this would > be a user kernel patch) Well, if you mlock() the UML RAM, you already implement something such. The= =20 problem is that such a patch will never go inside the official UML kernel=20 tree, so someone should maintain it outside. Anyway, what you say, i.e. the= =20 fact that some system (S/390?) support it, could make it easier for the=20 "mlock" patch to be accepted. Bye =2D-=20 Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 |