Messages in this thread | | | Date | Fri, 19 Feb 2010 01:42:27 -0800 (PST) | From | David Rientjes <> | Subject | Re: [regression] cpuset,mm: update tasks' mems_allowed in time (58568d2) |
| |
On Fri, 19 Feb 2010, KOSAKI Motohiro wrote:
> Personally, I like just revert at once than bandaid. 58568d2 didn't > introduce any new feature, then we can revet it without abi breakage. >
Revert a commit from more than six months ago when the fix is probably a small patch in cpuset_attach()? I think we can do better than that.
This may not have introduced a new feature, but it was a worthwhile change to avoid the old cpuset_update_task_memory_state() hooks in mempolicy, page allocator, etc. code that could block on callback_mutex for iterating the hierarchy.
| |