Re: PROBLEM: kernel BUG at drivers/gpu/drm/drm_fops.c:146!

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

On Fri, 30 Jan 2009 11:06:47 +1000 Dave Airlie <airlied@gmail.com> wrote:

> On Fri, Jan 30, 2009 at 10:30 AM, Andrew Morton
> <akpm@linux-foundation.org> wrote:
> > (cc's added)
> >
> > On Wed, 21 Jan 2009 13:27:48 +0100
> > Sami Kerola <kerolasa@iki.fi> wrote:
> >
> >> I compiled the Torvalds git kernel 2.6.29-rc2-00013 and I got an oops.
> >> The oops happens when ever X starts. Initially I was booting with run
> >> level 5 and it hung. I tried to use run level to 3 and an operating
> >> system started just fine. When I type startx the hung happen again.
> >> Please let me know if you need some more information besides oops from
> >> messages file and lspci output.
> >>
> >>
> >> Jan 21 08:53:58 lelux kernel: ------------[ cut here ]------------
> >> Jan 21 08:53:58 lelux kernel: kernel BUG at drivers/gpu/drm/drm_fops.c:146!
> >
> > I assume that 2.6.28 didn't do this?
> 
> This is a userspace race between udev and libdrm, I'm not sure we can do
> anything in the kernel other than BUG, maybe we should just WARN instead.
> 
> Basically, libdrm creates devices nodes, the initial drm opening gets that, udev
> comes along when the module is loaded and re-creates the device node,
> when AIGLX opens the device
> it can't figure out wtf just happened, as the inode->i_mapping we use
> to store the GEM device mmap ranges is different.
> 
> I think building libdrm with --enable-udev is the correct answer, and
> maybe switching this to a WARN so it doesn't blow up.
> 
> maybe we shouldn't be storing the inode mapping like this? anyone any
> better idea?
> 

hm, I'm a bit surprised to see the drm code using `struct
address_space' and read_mapping_page() and unmap_mapping_range() and
such.  I thought those only worked with regular files and pagecache :)

Is it possible to briefly explain what's going on there?

What instance of address_space_operations does ->dev_mapping actually
point at?