RE: [Xen-devel] Sixth (and final?) release candidate for Xen 3.4.3

["Langsdorf, Mark" <mark.langsdorf@xxxxxxx>]

Keir -

AMD is seeing an issue with Xen-3.4.3-rc7-pre and
Xen-4.0-rc1 that makes me think there's an unknown
race condition lurking in on_selected_cpus().

Sometimes, on the new 6-core processors, the powernow.c
driver call:

    on_selected_cpus(&cmd.mask, transition_pstate, &cmd, 0);

results in a corrupted cmd structure in transition_pstate().
I wrote some test code that checks that cmd->val is the
desired value before the on_selected_cpus() call, and it
is, but sometimes (very irregularly), cmd->val is grossly
out of range inside transition_pstate().

We're mostly seeing this on our new 6 core systems, but
they're getting the most attention in our testing right 
now.

Any help in debugging why this is occuring would be 
appreciated.  I'm requesting that Xen-3.4.3 be held until
this is resolved.

--Mark Langsdorf
Operating System Research Center
AMD


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel