kern/94939: [acpi] [patch] reboot(8) fails on IBM / Intel blades

[John Baldwin]

The following reply was made to PR kern/94939; it has been noted by GNATS.

From: John Baldwin <jhb at freebsd.org>
To: Nate Lawson <nate at root.org>
Cc: "Devon H. O'Dell" <dodell at ixsystems.com>, bug-followup at freebsd.org
Subject: Re: kern/94939: [acpi] [patch] reboot(8) fails on IBM / Intel blades
Date: Thu, 30 Mar 2006 14:03:57 -0500

 On Thursday 30 March 2006 13:23, Nate Lawson wrote:
 > John Baldwin wrote:
 > > On Tuesday 28 March 2006 02:22 pm, Devon H. O'Dell wrote:
 > >> On Tue, Mar 28, 2006 at 11:08:02AM -0800, Nate Lawson wrote:
 > >>> The system must reset immediately following the write to this register.
 > >>> OSPM assumes that the processor will not execute beyond the write
 > >>> instruction. OSPM should execute spin loops on the CPUs in the system
 > >>> following a write to this register.
 > >> My interpretation of this is ``don't do anything else after
 > >> the write to the register, because you can't expect to do
 > >> it.'' Since they say that the system ``must reset immediately
 > >> following the write'', it seems that this is implemented in
 > >> hardware, and we can't assume that we will be able to do
 > >> anything afterwards, anyway.
 > >>
 > >>> So I'm ok with the patch being committed if no other tasks need to
 > >>> happen after this shutdown handler is called.  Also, all APs should be
 > >>> stopped before this happens and it should only occur once on the BSP.
 > >> I was curious if anything happens after this handler is
 > >> called -- if there is, we definitely need to move it back
 > >> to later in the process. Again, I put the code here because it
 > >> looked to me like the procedure already assumed nothing else
 > >> is happening, but it sounds like there are other procedures
 > >> that are in the call queue after this one.
 > > 
 > > It really should be much later I think: in cpu_reset_real() as that
 > > is the only place that you know that the APs are stopped.
 > 
 > I'm not near a BSD box today.  Is there a simple, MI way of hooking 
 > there that doesn't require ACPI compiled into the kernel?  If it's a 
 > simple matter of moving it to a different shutdown handler or adding a 
 > way for acpi to conditionally override cpu_reset_real, that's ok with 
 > me.  I don't want acpi being partially merged into the main kernel.
 
 Not currently.
 
 -- 
 John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
 "Power Users Use the Power to Serve"  =  http://www.FreeBSD.org