Re: opps 2.4.20-pre5-ac2

[Tabris]

This is an interaction of the rmap vm patch (included in -ac) and the nVidia 
binary driver. I have run into this myself, tho it doesn't usually cause a 
lockup for me. (instead, it puts kswapd in the Z state (it aparrently never 
respawns), and makes shutting down a PITA.))

Mind you, this is a bug in the nVidia driver, not the linux kernel, where it 
pulls a behaviour that never was really correct before, but did work. This 
bug truly needs to be reported to nVidia.

Not having that specific kernel tree on my system atm, I can't seem to find 
the line that is BUG()'ing. But iirc, it's in this set

        if (page->buffers)
                BUG();
        if (page->mapping)
                BUG();
        if (!VALID_PAGE(page))
                BUG();
        if (PageLocked(page))
                BUG();
        if (PageActive(page))
                BUG();
        page->flags &= ~((1<<PG_referenced) | (1<<PG_dirty));

and is the (page->mapping) test.

please, check your tree for the correct line (102 of page_alloc.c) to verify 
this, and then pass it along to nVidia.

--
tabris