Re: 2.6.15-rc4 panic in __nr_to_section() with CONFIG_SPARSEMEM

[Dave Hansen <haveblue@us.ibm.com>]

On Wed, 2005-12-07 at 15:22 -0800, Badari Pulavarty wrote:
> On Wed, 2005-12-07 at 23:05 +0000, Andy Whitcroft wrote:
> > Badari Pulavarty wrote:
> > > Hi Andy,
> > > 
> > > I getting a panic while doing "cat /proc/<pid>/smaps" on
> > > a process. I debugged a little to find out that faulting
> > > IP is in _nr_to_section() - seems to be getting somehow
> > > called by  pte_offset_map_lock() from smaps_pte_range
> > > (which show_smaps) calls.
> > > 
> > > Any ideas on why or how to debug further ? 
> > 
> > From dave's call graph I'd ask the question whether we should be calling
> > pfn_valid() before pfn_to_page().  When I reviewed the proposed
> > pfn_to_page() implementation I only recall one use and that already had
> > the pfn_valid() in it.  I'll review -rc4 in the morning.
> 
> BTW, the problem seems to be while dealing with shared memory areas
> that are backed by largepages.

Should you even be making it into the pte function with large pages?
Don't they just stop at the pmd level?

Maybe smaps_pmd_range() needs a pmd_huge() check.

-- Dave