On Tue, 2006-06-06 at 19:12 -0700, Bryce Harrington wrote: > On Tue, Jun 06, 2006 at 05:54:07PM -0700, Dave Hansen wrote: > > On Tue, 2006-06-06 at 17:49 -0700, jnorman wrote: > > > patched 2.6.17-rc5 with: > > > 2.6.17-rc5-mm1 > > > patch-2.6.17-rc5-mm1-lxc2 > > > > Well, first of all, lxc patches aren't for memory hotplug development ;) > > Ah, could you recommend what we should be using? Jake was just > following the directions on http://lhms.sourceforge.net/howtouse.html > which suggested those were the patches to use. The document is wrong, or just nor specific enough. > > ia64 memory hotplug usually doesn't use probe files, but instead uses > > ACPI. The probe files are really only for debugging anyway, unless your > > firmware doesn't provide any real memory hotplug interfaces. > > Are there plans to make all systems use one approach or the other, or > will there always be several ways to access it? (So we can account for > them when designing the testsuite.) There will be several ways to access it. ppc64 doesn't provide any notifications to the kernel when a memory hotplug event occurs. It is all handled in userspace, so we need the probe file there. > Also, we haven't started looking at x386/x86_64 yet, but will they use > ACPI or probes? ACPI in production. But, the probes are useful for testing. -- Dave
