Re: [PATCH] AMD Opteron Rev. E hack

From: Maxim Dounin <mdounin_at_mdounin.ru>
Date: Thu, 5 Nov 2009 15:30:29 +0300
Hello!

On Thu, Nov 05, 2009 at 02:00:34PM +0200, Kostik Belousov wrote:

> On Thu, Nov 05, 2009 at 12:52:00PM +0100, Dag-Erling Sm??rgrav wrote:
> > Kostik Belousov <kostikbel_at_gmail.com> writes:
> > > I think there is no much sense in printing that hack in unused;
> > > instead, you should print info when option is enabled and vulnerable
> > > CPU is detected.
> > 
> > We should *definitely* print a warninhg when a vulnerable CPU is
> > detected and the option is *not* enabled.  How do you justify not
> > telling the user that you know the machine will crash as soon as he runs
> > 'make buildworld' with a high -j value?
> 
> We do not do this for other cpu bugs workarounds, why this should be
> different.

Well, probably is't a good idea to do so?  Something like NetBSD's 
sys/arch/x86/x86/errata.c seems to be right way to go.

> Besides, there were no confirmed reports of this happening
> in field (I mean the bug manifestation, not make -j panicing or hanging
> machine :).

http://bugs.mysql.com/bug.php?id=26081

Maxim Dounin
Received on Thu Nov 05 2009 - 11:30:31 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:57 UTC