extreme RTAS printks

Jake Moilanen moilanen at austin.ibm.com
Thu Feb 12 01:22:35 EST 2004


On Tue, 2004-02-10 at 17:30, Olof Johansson wrote:
> David Boutcher wrote:
>
> > So we are going to document the format of the hex dump so that it is useful
> > to people?  If not, I'm back to wondering exactly WHO the large kernel
> > messages are useful for.
>
>
> Doesn't look like it. I was of the impression that the RAS tools used
> the /var/log/messages dump. If they don't, then there's no use in
> printing it at all (_as long as_ there's a way to get to it from a
> debugger, and/or as long as the last ones are dumped right before a panic).

Putting them both in /proc and in /var/log/messages was a interim
solution until all boxes have rtas_errd and diagela.  These errors need
to saved for a CE to diagnose the problem in the field, otherwise we
have no first failure data to analyze.

These messages should only be seen in two cases, either there is a error
that needs to be reported from a real failure, or NVRAM is not being
cleared of the error because rtas_errd is not installed on the machine
and it's showing up on every boot.

The error logs are going to 2k, so there could be a lot of messages
printed in the future.  I think it's a good compromise to move the log
level to KERN_INFO.  That way the data is still stored for a CE, and the
messages won't annoy Hollis. :)

Thanks,
Jake


** Sent via the linuxppc64-dev mail list. See http://lists.linuxppc.org/





More information about the Linuxppc64-dev mailing list