Re: [PATCH 1/1] sysfs: add more info to the oops dump

[Greg KH <gregkh@suse.de>]

On Thu, Mar 10, 2011 at 08:32:40PM +0200, Phil Carmody wrote:
> On 10/03/11 08:25 -0800, ext Greg KH wrote:
> > On Thu, Mar 10, 2011 at 05:53:08PM +0200, Phil Carmody wrote:
> > > If we're going to remember which file we accessed, we might as well
> > > also remember whether it was a read or a write, and if the latter,
> > > some indication of what was written.
> > > 
> > > e.g.
> > > $ echo 1 > /sys/kernel/slab/:at-0000064/sanity_checks
> > > $ echo c > /proc/sysrq-trigger
> > > ...
> > > [  112.457580] last sysfs file (w): /sys/kernel/slab/:at-0000064/sanity_checks
> > > [  112.464569]   written: 1
> > 
> > Has this actually helped you out?  I've been thinking of removing this
> > line in the dmesg entirely as I haven't seen it help in a very long time
> > to track anything down.
> 
> I'm glad I started my commit message with that 'if', that tells you
> something. I have considered it, in its original form, not particularly
> helpful too. But I blamed that on it not telling me enough. Today I saw
> an oops related to a sysfs write, and this just patch basically wrote
> itself instantly. But it's not stood the test of time, that's for sure.

But would the trace-back also have shown you the problem?  If not, then
that's a good reason to keep this feature, and to add your patch.

But, I would suggest keeping it on one line, adding more lines to oops
messages is not good if at all possible.

thanks,

greg k-h