On Tue 11-09-18 19:29:21, Matthew Wilcox wrote: > On Tue, Sep 11, 2018 at 04:35:03PM -0700, Yang Shi wrote: [...] I didn't get to read the patch yet. > > And, Michal prefers have VM_HUGETLB and VM_PFNMAP handled separately for > > safe and bisectable sake, which needs call the regular do_munmap(). > > That can be introduced and then taken out ... indeed, you can split this into > many patches, starting with this: > > + if (tmp->vm_file) > + downgrade = false; > > to only allow this optimisation for anonymous mappings at first. or add a helper function to check for special cases and make the downgrade behavior conditional on it. -- Michal Hocko SUSE Labs
