Re: notifier chain problem? (was Re: 2.6.17-rc1 did break XFS)

[Herbert Poetzl <herbert@13thfloor.at>]

On Thu, Apr 13, 2006 at 04:40:54AM -0400, Jes Sorensen wrote:
> >>>>> "David" == David Chinner <dgc@sgi.com> writes:
> 
> David> On Thu, Apr 13, 2006 at 07:21:45AM +0200, Herbert Poetzl wrote:
> David> It looks like we landed on top of a a notifier call chain
> David> implementation change in -rc1. However, this should not matter
> David> to XFS because the interface to register_cpu_notifier() did not
> David> change and XFS is completely abstracted away from the notifier
> David> chain implementation. We do:
> 
> Dave,
> 
> Looks strange, the faulting address is in the same region as the
> eip. I am not that strong on x86 layouts, so I am not sure whether
> 0x78xxxxxx is the kernel's mapping or it's module space. Almost looks
> like something else had registered a notifier and then gone away
> without unregistering it.

sorry, the essential data I didn't provide here is
probably that I configured the 2G/2G split, which for
unknown reasons actually is a 2.125/1.875 split and
starts at 0x78000000 (instead of 0x80000000)

> Herbert, any chance you can make the complete boot log up to the point
> where it crashes, as well as a System.map and your .config available?
> (probably not posted to all the lists :)

sure, bootup is fine, as it boots on ext2/3 but once
it is up, and I mount the newly created xfs filesystem
the (virtual) machine (QEMU) panics ...

will provide all the data shortly via separated mail

best,
Herbert

> Cheers,
> Jes
>