Messages in this thread | | | Date | Wed, 6 May 2009 17:36:51 +0100 (BST) | From | Hugh Dickins <> | Subject | Re: [PATCH 3/6] ksm: change the KSM_REMOVE_MEMORY_REGION ioctl. |
| |
On Wed, 6 May 2009, Chris Wright wrote: > > There's already locking issues w/ using madvise and ksm, so yes, > changes would need to be made. Some question of how (whether) to handle > registration of unmapped ranges, closest to say ->mm->def_flags=VM_MERGE. > My hunch is there's 2 cases users might care about, a specific range > (qemu-kvm, CERN app, etc) or the entire vma space of a process.
Good food for thought there, but not on my mind at this moment.
> Another > question of what to do w/ VM_LOCKED, should that exclude VM_MERGE or > let user get what asked for?
What's the issue with VM_LOCKED? We wouldn't want to merge a page while it was under get_user_pages (unless KSM's own, but ignore that), but what's the deal with VM_LOCKED?
Is the phrase "covert channel" going to come up somehow?
Hugh
| |