Messages in this thread | | | Date | Fri, 21 Feb 2014 14:44:05 -0800 (PST) | From | David Rientjes <> | Subject | Re: [PATCH 4/4] hugetlb: add hugepages_node= command-line option |
| |
On Fri, 21 Feb 2014, Andi Kleen wrote:
> > > 2) it improves the kernel command line interface from incomplete > > > (lacking the ability to specify node<->page correlation), to > > > a complete interface. > > > > > > > If GB hugepages can be allocated dynamically, I really think we should be > > able to remove hugepagesz= entirely for x86 after a few years of > > supporting it for backwards compatibility, even though Linus has insisted > > That doesn't make any sense. Why break a perfectly fine interface? >
I think doing hugepagesz= and not default_hugepagesz= is more of a hack just because we lack support for dynamically allocating some class of hugepage sizes and this is the only way to currently do it; if we had support for doing it at runtime then that hack probably isn't needed. You would still be able to do default_hugepagesz=1G and allocate a ton of them when fragmentation is a concern and it can only truly be done at boot. Even then, with such a large size it doesn't seem absolutely necessary since you'd either be (a) oom as a result of all those hugepages or (b) there would be enough memory for initscripts to do this at runtime, this isn't the case with 2MB.
But, like I said, I'm not sure we'd ever be able to totally remove it because of backwards compatibility, but the point is that nobody would have to use it anymore as a hack for 1GB.
| |