On Tue 27-03-18 21:52:17, Cyrill Gorcunov wrote: > On Tue, Mar 27, 2018 at 02:38:11PM -0400, Yang Shi wrote: > > > Why do we need to hold mmap_sem here and call find_vma, when only > > > PR_SET_MM_ENV_END: is consuming it? I guess we can replace it wit the > > > new lock and take the mmap_sem only for PR_SET_MM_ENV_END. > > > > Actually, I didn't think of why. It looks prctl_set_mm() checks if vma does > > exist when it tries to set stack_start, argv_* and env_*, btw not only > > env_end. > > > > Cyrill may be able to give us some hint since C/R is the main user of this > > API. > > First and most important it makes code smaller. This prctl call is really > rarely used. Of course we can optimize it, but as I said I would prefer > to simply deprecate this old interface (and I gonne to do so once time > permit). Ohh, it would be really great if we can remove this thingy altogether. I cannot say it has a wee bit of my sympathy. -- Michal Hocko SUSE Labs