Messages in this thread | | | Date | Mon, 19 Apr 2010 17:37:46 +0300 | From | Avi Kivity <> | Subject | Re: [PATCH 1/5] Add a global synchronization point for pvclock |
| |
On 04/19/2010 05:32 PM, Glauber Costa wrote: > >> Right, another option is to put the initial read outside of the loop, >> that way you'll have the best of all cases, a single LOCK'ed op in the >> loop, and only a single LOCK'ed op for the fast path on sensible >> architectures ;-) >> >> last = atomic64_read(&last_value); >> > isn't a barrier enough here? > >
No. On i386, the statement
last = last_value;
will be split by the compiler into two 32-bit loads. If a write (atomic, using cmpxchg) on another cpu happens between those two loads, then the variable last will have a corrupted value.
-- Do not meddle in the internals of kernels, for they are subtle and quick to panic.
| |