Messages in this thread |  | | From | Andy Lutomirski <> | Date | Thu, 12 Dec 2013 12:14:53 -0800 | Subject | Re: Clock control algorithms (Re: [RFC][PATCH 5/7] x86: Use latch data structure for cyc2ns) |
| |
On Sat, Nov 30, 2013 at 9:34 AM, H. Peter Anvin <hpa@zytor.com> wrote: > There is a huge difference between something that breaks after 2^32 and 2^64 events. Very few computers will ever be able to have 2^64 events of any kind in their lifetime, never mind a single boot.
Given that struct seqcount contains an "unsigned" counter, the 32-bit wraparound thing could be a problem in practice. I hope there aren't security vulnerabilities in which userspace overflows a kernel refcount, seqcount, or other similar structure.
--Andy
|  |