Re: 2.6.23-rc1 regression: mm: fix fault vs invalidate race for linear mappings

[Nick Piggin <nickpiggin@yahoo.com.au>]

Bret Towe wrote:
> for a while in -git I've had an issue that on boot when gdm loads the
> screen stays black
> using ctrl-f1 doesn't return to a console and killing X doesn't help any
> ssh'ing into the box does work top only shows 100% io-wait
> dmesg shows nothing odd
> 
> the work around I have is at the moment is to just move the radeon drm 
> module
> out of the way so it doesn't load on boot and X works just fine like that
> I did some bisecting which took a few days and tracked it down to
> commit d00806b183152af6d24f46f0c33f14162ca1262a
> 
> its way to complex for me to revert it on top of -rc1 to verify that's
> the issue tho
> I keep forgetting to get a trace of what its waiting on when I'm in that 
> kernel
> I assume that would be of use and Ill get that later
> 
> the box this is happening on is a g4 mac mini the built in card is a 
> radeon 9200
> I'm not seeing any issues on an amd64 box with radeon card it's also a 
> 9600 tho

Is this with a binary-only module? We saw an issue with that in SLES9
where the module is returning a locked page from its nopage handler
when it isn't really supposed to. It might be fixed in latest drivers,
have you tried them?

Thanks,
Nick

-- 
SUSE Labs, Novell Inc.