On Wed, 03 November 2010 Joe Perches wrote:

On Wed, 2010-11-03 at 11:50 -0700, Randy Dunlap wrote:

quoted

On Wed, 3 Nov 2010 11:16:34 -0700 Greg KH wrote:

quoted

On Tue, Nov 02, 2010 at 08:37:42PM -0700, Mike Waychison wrote:

quoted

On Tue, Nov 2, 2010 at 7:34 PM, Greg KH wrote:

As for the user/kernel interface, perhaps exporting the data in a text
format that is "tagged" would be best?  Then the whole world can parse
it easily.

I have been (occasionally) looking at critical kernel messages.
IMO we really need an easy way to find them.
They can begin with any of these strings (and others can be added
too easily):
BUG|panic|MCE|NMI|error:|Oops|Bad|Fatal|Unrecoverable|Unhandled|Weird
We need a simple (single?) tagging method to identify any/all of these,
/methinks.

Simply marking these messages KERN_CRIT or higher would seem to work.

That's an idea (probably better higher that can be enabled on a console with
a specific flag), though it also invites for filter levels on a per-console
base (e.g. let uart only show KERN_WARN..KERN_CRIT but having more verbosity
on termial or netconsole [per target]).

I've been playing with this last winter, need to dig out that code again and
clean it up for sharing.

Bruno