All of lore.kernel.org
 help / color / mirror / Atom feed
From: Mike Strosaker <strosake@austin.ibm.com>
To: will_schmidt@vnet.ibm.com
Cc: Stephen Rothwell <sfr@canb.auug.org.au>,
	paulus@samba.org, linuxppc-dev@ozlabs.org
Subject: Re: [PATCH] powerpc: fix os-term usage on kernel panic
Date: Fri, 30 Nov 2007 13:26:50 -0600	[thread overview]
Message-ID: <475063FA.50008@austin.ibm.com> (raw)
In-Reply-To: <1196439093.11297.52.camel@farscape.rchland.ibm.com>

Will Schmidt wrote:
>>From my reading of the papr, I've got the impression that there are two
> possibilities.    
> 
> First, the os-term never returns, and it's up to the service processor
> to do whatever it's going to do.  (call home, dump, something else).
> Nothing we can do there. 
> 
> Second, os-term returns, and it's up to the OS to call into power-off or
> system-reboot.
> I think this is more likely.   I havn't followed the path back from
> machine_restart to see exactly how we got there, but probably means a
> bit more logic to decide whether to call into rtas_restart() or
> pSeries_power_off() after the call to machine_shutdown.

My understanding was that os-term is supposed to indicate an "abnormal" 
termination, which is why it was initially only associated with a panic.

The os-term behavior is kind of complex; there is both a standard behavior, and 
an "extended" behavior.  You can determine if the extended behavior will be used 
by looking for the ibm,extended-os-term property.  I think POWER5 largely uses 
the older behavior, and POWER6 uses the extended behavior.

I've never seen os-term return using the old behavior.  Whether the partition 
reboots after calling os-term depends on the setting of the 
partition_auto_restart RTAS parameter.  If partition_auto_restart is 1 when 
os-term is called, the hypervisor will restart the partition and reset 
partition_auto_restart to 0.  The OS is supposed to set partition_auto_restart 
back to 1 after it boots; that way, if the OS is failing to boot, the hypervisor 
won't continually try to restart the partition without any hope of success. 
That's why Linas sent a patch with a pSeries_auto_restart routine.

If the extended os-term behavior is used, the os-term call is supposed to return 
unless the ibm,configure-kernel-dump call was previously used (to prepare for a 
PHYP assisted dump).  Would that explain why Linas and Will are seeing different 
behaviors?  I may have missed it, but I don't see a check for 
ibm,extended-os-term anywhere.

- Mike

      reply	other threads:[~2007-11-30 19:26 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-11-20  1:28 [PATCH] powerpc: fix os-term usage on kernel panic Linas Vepstas
2007-11-28  0:15 ` Will Schmidt
2007-11-28 11:00   ` Olaf Hering
2007-11-28 20:09     ` Linas Vepstas
2007-11-29 10:41       ` Olaf Hering
2007-12-03 19:32         ` Linas Vepstas
2007-11-28 20:18   ` Linas Vepstas
2007-11-29 17:19     ` Will Schmidt
2007-11-30  5:56   ` Stephen Rothwell
2007-11-30 16:11     ` Will Schmidt
2007-11-30 19:26       ` Mike Strosaker [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=475063FA.50008@austin.ibm.com \
    --to=strosake@austin.ibm.com \
    --cc=linuxppc-dev@ozlabs.org \
    --cc=paulus@samba.org \
    --cc=sfr@canb.auug.org.au \
    --cc=will_schmidt@vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.