Re: [PATCH 1/2] radeon: fix radeon kms framebuffer device

[Andrew Lutomirski <luto@mit.edu>]

On Tue, Jun 23, 2009 at 9:35 AM, Jerome Glisse<glisse@freedesktop.org> wrote:
> On Tue, 2009-06-23 at 09:12 -0400, Andy Lutomirski wrote:
>> Andy Lutomirski wrote:
>> > Jerome Glisse wrote:
>> >> smem.start is a physical address which kernel can remap to access
>> >> video memory of the fb buffer. We now pin the fb buffer into vram
>> >> by doing so we are loosing vram but fbdev need to be reworked to
>> >> allow change in framebuffer address.
>> >
>> > I tested this (and the corresponding 2/2 initialization order, but
>> > with radeon as a module), and plymouth seems to be fully functional
>> > (graphical boot, password prompt, etc).
>> >
>> > (The driver set the wrong mode, but that's a different issue.)
>> >
>> > Thanks!
>> >
>> > Tested-by: Andy Lutomirski <luto@mit.edu>
>> >
>> Got an oops after awhile:
>>
>> BUG: Bad page state in process gpg-agent  pfn:37dd5
>> page:ffffea0000c38698 flags:200000000050000c count:0 mapcount:0
>> mapping:(null) index:7fd35e311
>> Pid: 3131, comm: gpg-agent Not tainted 2.6.30 #5
>> Call Trace:
>>  [<ffffffff810bc99b>] bad_page+0x115/0x13e
>>  [<ffffffff810bcbd3>] free_pages_check+0x3c/0x6d
>>  [<ffffffff810bde8f>] free_hot_cold_page+0x4e/0x151
>>  [<ffffffff810bdfce>] __pagevec_free+0x3c/0x65
>>  [<ffffffff810c1c82>] release_pages+0x1a5/0x1cb
>>  [<ffffffff810de357>] free_pages_and_swap_cache+0x6d/0x9e
>>  [<ffffffff810d584b>] tlb_finish_mmu+0x41/0x63
>>  [<ffffffff810d5b3f>] exit_mmap+0xfb/0x138
>>  [<ffffffff8104beae>] mmput+0x55/0xc1
>>  [<ffffffff81050626>] exit_mm+0x10e/0x12f
>>  [<ffffffff8105226f>] do_exit+0x1b4/0x6a8
>>  [<ffffffff8106c6b0>] ? up_read+0x1c/0x32
>>  [<ffffffff810527e2>] do_group_exit+0x7f/0xac
>>  [<ffffffff81052834>] sys_exit_group+0x25/0x3d
>>  [<ffffffff8100beab>] system_call_fastpath+0x16/0x1b
>>
>> I don't know if this is related to the drm changes.
>>
>> Back to 2.6.29 for me :)  (I actually use this machine, so I'd rather
>> not run kernels that cause random corruption.)
>>
>> --Andy
>
> Could you get a trace using Linus pte check patch he posted earlier
> on the other thread ?

It didn't trigger Linus' warning.

I was also able to trigger this bug with modesetting turned off on the
kernel command line by switching VTs.

--Andy

>
> Jerome
>
>