On Tue, Aug 26, 2014 at 05:56:12PM +0300, Kirill A. Shutemov wrote: > > > > > > It seems safe in vma-softdirty context. But if somebody else will decide that > > > it's fine to modify vm_flags without down_write (in their context), we > > > will get trouble. Sasha will come with weird bug report one day ;) > > > > > > At least vm_flags must be updated atomically to avoid race in middle of > > > load-modify-store. > > > > Which race you mean here? Two concurrent clear-refs? > > Two concurent clear-refs is fine. But if somebody else will exploit the > same approch to set/clear other VM_FOO and it will race with clear-refs > we get trouble: some modifications can be lost. yup, i see > Basically, it's safe if only soft-dirty is allowed to modify vm_flags > without down_write(). But why is soft-dirty so special? because how we use this bit, i mean in normal workload this bit won't be used intensively i think so it's not widespread in kernel code > Should we consider moving protection of some vma fields under per-vma lock > rather use over-loaded mmap_sem? Hard to say, if vma-softdirty bit is the reason then I guess no, probably it worth to estimate how much profit we would have if using per-vma lock. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>