On Tue 10-04-18 14:02:42, Cyrill Gorcunov wrote: > On Tue, Apr 10, 2018 at 12:42:15PM +0200, Michal Hocko wrote: > > On Tue 10-04-18 12:40:47, Cyrill Gorcunov wrote: > > > On Tue, Apr 10, 2018 at 11:09:17AM +0200, Michal Hocko wrote: > > > > On Tue 10-04-18 05:52:54, Yang Shi wrote: > > > > [...] > > > > > So, introduce a new spinlock in mm_struct to protect the concurrent > > > > > access to arg_start|end, env_start|end and others except start_brk and > > > > > brk, which are still protected by mmap_sem to avoid concurrent access > > > > > from do_brk(). > > > > > > > > Is there any fundamental problem with brk using the same lock? > > > > > > Seems so. Look into mm/mmap.c:brk syscall which reads and writes > > > brk value under mmap_sem ('cause of do_brk called inside). > > > > Why cannot we simply use the lock when the value is updated? > > Because do_brk does vma manipulations, for this reason it's > running under down_write_killable(&mm->mmap_sem). Or you > mean something else?
Yes, all we need the new lock for is to get a consistent view on brk values. I am simply asking whether there is something fundamentally wrong by doing the update inside the new lock while keeping the original mmap_sem locking in the brk path. That would allow us to drop the mmap_sem lock in the proc path when looking at brk values. -- Michal Hocko SUSE Labs