Likely Duplicate Bugs

[Brian Murray]

On Mon, Feb 08, 2010 at 03:15:57PM +0000, Colin Ian King wrote:
> Hi Brian,
> On Sun, 2010-02-07 at 18:25 -0800, Brian Murray wrote:
> > On Mon, Feb 01, 2010 at 11:56:00AM +0000, Colin Ian King wrote:
> > > Hi Brian,
> > > 
> > > This kernel message are dups of the same BIOS corruption message, but
> > > occurs on a wide range of machines. I had a look at the first 15 or so
> > > of the dups and saw that there was a wide spread of Aspire, HP Compaqs
> > > and Pavilions and quite a number of unknown systems too.
> > 
> > Should these duplicate bug reports be consolidated into one bug or
> > should there be multiple bug reports for each system model?
> 
> Well, the warning is the same manifestation of possibly different BIOS
> issues from many different BIOSes (possibly even different versions of
> the BIOS on the *same* system model), so I'd count these as separate
> reports even if there are a lot of them.   I think I would only attempt
> to consolidate bug reports based on the same system model or the
> "Hardware name" field as reported in the OopsText.txt, e.g.:
> 
> Hardware name: HP Compaq nc6120 (PY390ES#ABZ)
> 
> ..not sure how easy that is to do automatically.

It wasn't that hard so I've consolidated the bug reports based on the
hardware models per your suggestion.  There were quite a few with a
Hardware name of INVALID which I left alone.  Maybe it would be a good
idea to also modify the bug summary / title with the hardware name?

> > > Specifically, the kernel fills known regions of the low 64K of memory
> > > with a known pattern and periodically monitors them.  Any buggy BIOS
> > > that writes to these regions gets detected and the warning is issued.
> > > 
> > > BIOS corruption of these regions can occur when doing suspend/resume or
> > > HDMI cable unplugging.
> > > 
> > > The error message is a warning - the system's stability is not
> > > compromised as the pages being monitored are already reserved for the
> > > purpose of being monitored for corruption in the first place.
> > > 
> > > This check can be disabled by setting the kernel boot parameter
> > > memory_corruption_check=0
> > > 
> > > Since this is intended as a BIOS corruption detection tool perhaps it
> > > should be disabled as a compile time option to stop getting these
> > > messages. However, it does have some value in showing that the BIOS may
> > > be dodgy. 
> > 
> > Perhaps the ideal solution is to keep it enabled until the final version
> > of the release.  This will still provide us with useful information
> > about buggy BIOSes but by disabling it for the final release we will
> > reduce the quantity of redundant bugs.
> > 
> Information about buggy BIOSes is useful in one respect, but generally
> requires a BIOS fix which means we cannot do much apart from perhaps
> recommend a BIOS upgrade - but that's a risky operation which may
> provide little or no benefit.
> 
> Since the kernel isn't using the memory for anything important then
> these kind of warnings are of a small amount of use. However they
> possibly give users alarming messages when in fact they are just helpful
> for some classes of debugging. Hence they should be totally disabling on
> the final release IMHO.
> 
> Any comments anyone?

It shouldn't be hard to write bug patterns for the check_bios_corruption
messages and the specific hardware.  Does that seem like a good idea to
anyone else?
 
-- 
Brian Murray                                                 @ubuntu.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <https://lists.ubuntu.com/archives/kernel-team/attachments/20100211/53ea22d1/attachment.sig>