extreme RTAS printks

Mike Strosaker strosake at austin.ibm.com
Wed Feb 11 09:30:50 EST 2004


Olof Johansson wrote:
> Hmm, there's a userspace package that's used to parse all that info. I
> don't know enough about the internal binary format, but I'd think that
> there's severity levels to it. Maybe we can have a threshold for what
> gets printed at boot and not?

It would require a fairly significant amount of parsing inside the
kernel to determine if the message is worthy of being printed.  There's
no one severity field that makes thresholding easy.  Also, the parsing
code would need to be updated to reflect any new error log formats in
the future, which is why it's better done in userspace.

Nathan Lynch wrote:
> How about making it a config option? Patch attached. I'm not sure the
> help text is 100% correct - do RAS tools depend on the messages being
> in dmesg output or do they look at the /proc files?
>
> Another idea is changing the log level of these messages from KERN_ERR
> to something lower priority like KERN_INFO or KERN_DEBUG. At least
> that way the console doesn't get spammed in default configurations.

The userspace RAS tools look at the /proc file; there's no harm from
that perspective in either of the above solutions.  I'm all for doing
either, but the latter (KERN_INFO or KERN_DEBUG) sounds like a
particularly good compromise to me...  that way the messages are still
there in case of emergency.

Thanks,
Mike Strosaker

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





More information about the Linuxppc64-dev mailing list