On Fri, Apr 19, 2019 at 04:41:24PM +0800, Li Wang wrote: > > Avoiding the scenario is pointless because it's not wrong. The check was > > initially meant to catch serious programming errors such as using a > > stale page pointer so I think the right patch is below. Li Wang, how > > reproducible is this and would you be willing to test it? > > > > It's not easy to reproduce that again. I just saw only once during the OOM > phase that occurred on my s390x platform. > > Sure, I run the stress test against a new kernel(build with this patch > applied) for many rounds, so far so good. > I think the patch is safe enough and have sent it on to Andrew. Thanks for reporting and testing this, it's appreciated. -- Mel Gorman SUSE Labs
