On Thu, Nov 06, 2014 at 05:27:14PM +0000, Luck, Tony wrote:
> >> +int mce_severity(struct mce *m, int tolerant, char **msg, bool is_excp)
> >
> > You're adding a function argument which is carrying redundant info which
> > is already present in *m...
> >
> >> {
> >> + enum exception excp = (is_excp ? EXCP_CONTEXT : NO_EXCP);
> >
> > ... and so this should be:
> >
> > excp = ((m->mcg_status & MCG_STATUS_MCIP) ? EXCP_CONTEXT : NO_EXCP);
>
> That only works if you trust that MCG_STATUS.MCIP is correctly set to
> indicate whether
> we are in MCE or CMCI context. The current code doesn't do that - we check
> for, and flag
> it as a fatal error if we find ourselves in the MCE handler with MCIP==0. If
> you add the
> code you suggest, then it completely neuters the severity check:
>
> MCESEV(
> PANIC, "MCIP not set in MCA handler",
> MCGMASK(MCG_STATUS_MCIP, 0)
> ),
I was looking at the version Chen did:
MCESEV(
PANIC, "MCIP not set in MCA handler",
EXCP, MCGMASK(MCG_STATUS_MCIP, 0)
),
and then
if (s->excp && excp != s->excp)
continue;
Basically, this check is being done only for machine check exceptions
only.
> I'm also a bit worried about the check for DEFERRED errors in
> the severity table. That isn't conditional on an:
> if (intel) do_onething(); else /*amd/ do_anotherthing();
> So if we can misinterpret some bits on an Intel cpu as if
> we had a deferred error.
>
> Overall, this might have seemed like a good idea to begin with,
> but we are piling more complexity into mce_severity() [a routine
> which everyone agrees is already tough to understand].
>
> It doesn't even buy us some simple code in the polling path.
> We still have to do more checks on MCi_STATUS.MCACOD above
> and beyond what we get back from mce_severity()
>
> Boris: Do you still want to keep pushing this way? Or should
> we look back fondly at version 1 of this patch?
You mean the one which doesn't touch mce_severity() at all and decides
on deferred errors in a separate, completely unrelated function? Yeah,
that might be cleaner after all.
--
Regards/Gruss,
Boris.
Sent from a fat crate under my desk. Formatting is fine.
--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
