On Sat, Mar 26, 2011 at 12:18 PM, Christoph Lameter <cl@xxxxxxxxx> wrote:
>
> Right. RSI should be in the range of the values you printed.
> However, the determination RSI is independent of the emulation of the
> instruction. If RSI is set wrong then we should see these failures on
> machines that do not do the instruction emulation. A kvm run with Ingo's
> config should show the same issues. Will do that now.
I bet it's timing-dependent and/or dependent on some code layout
issue. Why else would the totally pointless previous patch have made
any difference for Ingo (the "SLUB: Write to per cpu data when
allocating it" thing seems to be pure voodoo programming)?
That said, this early in the boot I don't think we should have any
parallelism going on, so I don't see what could make those kinds of
random effects.
Linus
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxxx For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>