Messages in this thread | | | Date | Fri, 10 Sep 2010 10:41:27 +0200 | From | Tejun Heo <> | Subject | Re: [RFC PATCH 1/2] perf_events: add support for per-cpu per-cgroup monitoring (v3) |
| |
Hello,
On 09/10/2010 10:16 AM, Peter Zijlstra wrote: > On Thu, 2010-09-09 at 23:41 +0200, Stephane Eranian wrote: >>> alloc_percpu() is zalloc_percpu() in fact, memory is already cleared. >>> >> I remember thinking about this and trying to trace to the code down >> to figure this out. But it is rather complicated. If alloc_percpu() always >> clears the memory, then I think that calling is zalloc_percpu() >> would be more helpful....
Maybe but at this point it might be a bit too late. The allocator has been that way since the beginning.
> pcpu_populate_chunk() in mm/percpu-vm.c does indeed do that memset, the > one in mm/percpu-km.c does not. > > It is not obviously clear to me the -km allocator does indeed result in > zero filled memory.
Nice catch. Fortunately, the -km allocator isn't currently being used in upstrea although it was enabled for linux-next a couple of days ago. I'll fix it up.
Thank you.
-- tejun
| |