Re: [mmotm] Backtrace __change_page_attr_set_clr

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 27 Jan 2009 16:38:23 +0530 Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> Hi, Andrew,
> 
> I see the following backtrace with 2009-01-16-16-18
> 
> ------------[ cut here ]------------
> WARNING: at arch/x86/mm/pageattr.c:560
> __change_page_attr_set_clr+0xdc/0x891()
> ...
> CPA: called for zero pte. vaddr = ffff8800cfee2000 cpa->vaddr =
> ffff8800cfee2000
> Modules linked in: coretemp hwmon kvm_intel kvm rtc_cmos rtc_core
> rtc_lib mptsas mptscsih mptbase scsi_transport_sas
> Pid: 3486, comm: dmidecode Not tainted 2.6.29-rc1 #2
> Call Trace:
>  [<ffffffff8023e9da>] warn_slowpath+0xd8/0xf5
>  [<ffffffff8024f00b>] __kernel_text_address+0x1a/0x26
>  [<ffffffff8020f307>] print_context_stack+0x90/0xa6
>  [<ffffffff8025cc63>] find_usage_backwards+0xe/0xfc
>  [<ffffffff8025f500>] __lock_acquire+0x763/0x7ce
>  [<ffffffff80227237>] __change_page_attr_set_clr+0x5b/0x891
>  [<ffffffff80227237>] __change_page_attr_set_clr+0x5b/0x891
>  [<ffffffff802272b8>] __change_page_attr_set_clr+0xdc/0x891
>  [<ffffffff802a2a5c>] __purge_vmap_area_lazy+0x195/0x1a2
>  [<ffffffff802a2912>] __purge_vmap_area_lazy+0x4b/0x1a2
>  [<ffffffff802a3b49>] vm_unmap_aliases+0x1b9/0x1c8
>  [<ffffffff802a39cd>] vm_unmap_aliases+0x3d/0x1c8
>  [<ffffffff80227bac>] change_page_attr_set_clr+0x13f/0x2f1
>  [<ffffffff80597355>] _spin_unlock+0x17/0x20
>  [<ffffffff80227e56>] _set_memory_wb+0x1d/0x20
>  [<ffffffff80229189>] reserve_pfn_range+0x1a1/0x23a
>  [<ffffffff8029ab59>] remap_pfn_range+0x90/0x33a
>  [<ffffffff802294a2>] phys_mem_access_prot_allowed+0x98/0x1b9
>  [<ffffffff802b0e3f>] kmem_cache_alloc+0x94/0xbc
>  [<ffffffff803d8d48>] mmap_mem+0x6c/0x91
>  [<ffffffff8029ee18>] mmap_region+0x262/0x4a9
>  [<ffffffff8029f5c4>] do_mmap_pgoff+0x2a6/0x30d
>  [<ffffffff8020fa95>] sys_mmap+0x8b/0x110
>  [<ffffffff8020bedb>] system_call_fastpath+0x16/0x1b
> ---[ end trace 21993f860dedca28 ]---
> dmidecode:3486 reserve_pfn_range ioremap_change_attr failed write-back
> for cfee2000-cfee3000
> dmidecode:3486 freeing invalid memtype cfee2000-cfee3000
> 
> I am yet to fully investigate the problem, but did not want to hold of
> reporting it.

OK, thanks.  It looks like you cc'ed the right person ;)

The code in current linux-next has changed a bit since 16 Jan. 
Hopefully this warning from dmidecode got fixed.

I'm still quite backlogged from the week overseas, and I've been
reluctant to do another mmotm because it would probably be an
uncompileable pile of junk.