Date: Wed, 25 Apr 2007 14:10:04 -0400 From: John Baldwin <jhb@freebsd.org> To: Ariff Abdullah <ariff@freebsd.org> Cc: freebsd-i386@freebsd.org, freebsd-amd64@freebsd.org Subject: Re: AMD64 K8 Family C1E dead apic timer fixup Message-ID: <200704251410.04512.jhb@freebsd.org> In-Reply-To: <20070425011720.5725e654.ariff@FreeBSD.org> References: <20070420195444.20bc4d3a.ariff@FreeBSD.org> <200704241147.57194.jhb@freebsd.org> <20070425011720.5725e654.ariff@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday 24 April 2007 01:17:20 pm Ariff Abdullah wrote: > On Tue, 24 Apr 2007 11:47:56 -0400 > John Baldwin <jhb@freebsd.org> wrote: > > On Monday 23 April 2007 09:39:59 pm Ariff Abdullah wrote: > > > On Mon, 23 Apr 2007 14:46:53 -0400 > > > John Baldwin <jhb@freebsd.org> wrote: > > > > On Friday 20 April 2007 07:54:44 am Ariff Abdullah wrote: > > > > > Will anybody object if I (or somebody else) commit this patch? > > > > > > > > > > http://people.freebsd.org/~ariff/test/k8_c1e.diff > > > > > > > > > > This should "fix" dead apic timer issues on latest AMD64 > > > > > (Turion64 X2 mostly, or later revision) where C1 Enhanced > > > > > (C1E) mode enabled. > > > > > > > > > > Without this, the only way to make the timer work is to > > > > > disable APIC or even worst, to avoid executing HLT instruction > > > > > (machdep.cpu_idle_hlt=0), smp or non-smp. It affect both i386 > > > > > and amd64. > > > > > > > > > > See PR: i386/104678 for details. > > > > > > > > Does this fix all of the Turion lapic timer hangs? > > > > > > Yes, and hopefully to the later revisions (if any) as well. > > > > Ok, cool. > > > > > > Some small tweaks to the comment would be: > > > > > > > > - s/present/presence/ > > > > - s/render/renders/ > > > > - s/apic timer/local APIC timer/ > > > > - s/dead, and/dead, so/ > > > > - s/reading Interrupt/reading the Interrupt/ > > > > - s/unset/clearing/ > > > > > > > > > > Thanks. > > > > > > > Also, is it ok to always mask off the upper 31 bits of the MSR > > > > instead of preserving their current state? > > > > > > > It should work either way, since the upper bits are read only. > > > > Why bother masking it off then? > > > > Except of my personal tendency to kill it, I don't think we need to > mask it off. See latest patch (same link). Go for it. -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200704251410.04512.jhb>