On Monday 29 of August 2011 17:39:46 Keith Packard wrote:
> On Mon, 29 Aug 2011 15:22:10 +0200, Pawel Sikora <pluto@agmk.net> wrote:
> > Aug 29 15:10:55 vmx kernel: [22761.103401] BUG: unable to handle kernel paging request at 000000104a69c2f8
> > Aug 29 15:10:55 vmx kernel: [22761.103403] IP: [<ffffffff81265d02>] fb_flashcursor+0x42/0x150
> 
> I'm assuming this is a regression; can you find a version that it worked
> on without any other changes in the environment? I'm concerned that
> you're loading both vesafb and then switching to the drm fb, which
> always seems fraught with peril to me.
> 
> And, of course, a bisect would be most helpful if possible. I can't
> think of any i915-specific changes that would cause fb mappings to go wrong.

git bisect shows the 695434e1cbd57f404110bf4ab187a5127ffd79bb as a first commit
which prints fb_flashcursor ooops but this commit is imho completely unrelated,
so probably there's a hidden bug exposed by minimal binary code movement.

the testcease is:

$login_as_root_locally
run locally 'service fbset start'
$login_as_root_via_ssh_to_grab_dmesg
run locally 'Xorg -configure'

BR,
Paweł.