From mboxrd@z Thu Jan 1 00:00:00 1970 From: "H. J. Lu" Date: Fri, 13 May 2005 14:10:23 +0000 Subject: Re: elf header bits for page migration support...... Message-Id: <20050513141023.GB24961@lucon.org> List-Id: References: <42847F6B.3080609@engr.sgi.com> In-Reply-To: <42847F6B.3080609@engr.sgi.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-ia64@vger.kernel.org On Fri, May 13, 2005 at 05:20:27AM -0500, Ray Bryant wrote: > Tony, > > As some of the readers of this list may know, I've been working on a > "manual page migration" facility for Altix. Basically this is intended > to allow a batch manager to move applications around on a large NUMA > system. Further details and discussion can be found on the linux-mm > lists, for example the thread starting at: > > http://marc.theaimsgroup.com/?l=linux-mm&m1578651020174&w=2 > > Now given the existence of the memory migration code in the memory > hotplug patch, the hard part of this project is not doing the migration, > but figuring out what not to migrate (e. g. one doesn't want to migrate > the shared pages of a shared library, for example). Currently, we are > flagging files as libraries by using a file system extended attribute. > But this is meeting with some resistance. > > The alternative being suggested is to mark the elf header of object files > that should only have shared pages migrated, or that should not be migrated > (e. g. you don't want to migrate /bin/csh all the time.) > > (We'll extend mmap() with a similar set of flags.) > > So the question has come up, which bits in the elf header are available > for such use. Apparently there is a bit of a muddle here and we need to > get the "community" to agree. Bits in elf header are very scare. Can you use an ELF segment like PT_GNU_EH_FRAME and PT_GNU_STACK? H.J.