Re: [PATCH] net: fix race in the receive/select
From: Andi Kleen <hidden>
Date: 2009-07-01 06:27:47
Also in:
lkml
Possibly related (same subject, not in this thread)
- 2009-06-29 · Re: [PATCH] net: fix race in the receive/select · Jiri Olsa <hidden>
- 2009-06-29 · Re: [PATCH] net: fix race in the receive/select · Jiri Olsa <hidden>
- 2009-06-28 · Re: [PATCH] net: fix race in the receive/select · Jarek Poplawski <hidden>
- 2009-06-28 · Re: [PATCH] net: fix race in the receive/select · Oleg Nesterov <oleg@redhat.com>
- 2009-06-28 · Re: [PATCH] net: fix race in the receive/select · Jarek Poplawski <hidden>
On Wed, Jul 01, 2009 at 11:39:24AM +0800, Herbert Xu wrote:
Andi Kleen [off-list ref] wrote:quoted
At least on x86 I can't see how it happens. mb() is only a compile time barrier and the compiler doesn't optimize over indirect callbacks like __pollwait() anyways.mb() does more than just a compiler barrier, it also issues mfence.
mfence is not needed for normal C code (not using non temporal stores) in the Linux memory model on x86 and is a no-op. Only the compile time barrier matters. -Andi -- ak@linux.intel.com -- Speaking for myself only.