Re: kernel BUG at mm/rmap.c:669 (during localegen)

[Hugh Dickins]

On Thu, 1 May 2008, PaweÅ wrote:
> 
> Every time I try to generate locales computer reboots or hangs. To ensure
> that's not matter of graphDrivers I checked also in runlevel 1 and
> "init=/bin/bash".
> 
> Linux debby 2.6.25 #1 PREEMPT Mon Apr 28 15:18:31 CEST 2008 x86_64 GNU/Linux
> gcc (GCC) 4.2.3 (Debian 4.2.3-3)
> 
> [kern.log:]
> //May  1 11:20:09 debby kernel: NVRM: bad caching on address
> 0xffff81003a910000: actual 0x173 != expected 0x17b
> 
> May  1 11:25:44 debby kernel: Eeek! page_mapcount(page) went negative! (-1)
> May  1 11:25:44 debby kernel:   page pfn = 257d2
> May  1 11:25:44 debby kernel:   page->flags = 2200000000001c
> May  1 11:25:44 debby kernel:   page->count = 0
> May  1 11:25:44 debby kernel:   page->mapping = 0000000000000000
> May  1 11:25:44 debby kernel:   vma->vm_ops = 0x0
> May  1 11:25:44 debby kernel: ------------[ cut here ]------------
> May  1 11:25:44 debby kernel: kernel BUG at mm/rmap.c:669!
> May  1 11:25:44 debby kernel: invalid opcode: 0000 [1] PREEMPT
> May  1 11:25:44 debby kernel: CPU 0
> May  1 11:25:44 debby kernel: Modules linked in: nvidia(P) ...
> May  1 11:25:44 debby kernel: Pid: 3007, comm: localedef Tainted: P
> 2.6.25 #1
> May  1 11:25:44 debby kernel: RIP: 0010:[<ffffffff80270c74>]
> [<ffffffff80270c74>] page_remove_rmap+0x164/0x180

Sorry, you'll have to ask Nvidia to help you with this: we don't
know what their driver is up to, and don't have any "NVRM" errors in
2.6.25 source; they should be grateful for your reproducible test case.

Hugh