diff mbox

radeon causing sleeping function called from invalid context

Message ID 20130113113820.GA2503@swordfish (mailing list archive)
State New, archived
Headers show

Commit Message

Sergey Senozhatsky Jan. 13, 2013, 11:38 a.m. UTC 
On (01/12/13 20:27), Dave Jones wrote:
> BUG: sleeping function called from invalid context at mm/slub.c:925
> in_atomic(): 1, irqs_disabled(): 0, pid: 566, name: Xorg
> INFO: lockdep is turned off.
> Pid: 566, comm: Xorg Not tainted 3.8.0-rc3+ #49
> Call Trace:
>  [<ffffffff81087ad1>] __might_sleep+0x141/0x200
>  [<ffffffff8119eddb>] kmem_cache_alloc_trace+0x4b/0x2a0
>  [<ffffffffa0083fb3>] ttm_bo_move_accel_cleanup+0x1d3/0x330 [ttm]
>  [<ffffffffa00c5418>] radeon_move_blit.isra.4+0xf8/0x160 [radeon]
>  [<ffffffffa00c5db0>] radeon_bo_move+0xb0/0x1f0 [radeon]
>  [<ffffffffa00820ed>] ttm_bo_handle_move_mem+0x27d/0x5d0 [ttm]
>  [<ffffffff8108dd51>] ? get_parent_ip+0x11/0x50
>  [<ffffffffa0083307>] ttm_bo_move_buffer+0x127/0x140 [ttm]
>  [<ffffffffa008717d>] ? ttm_eu_list_ref_sub+0x3d/0x60 [ttm]
>  [<ffffffffa00833c2>] ttm_bo_validate+0xa2/0x120 [ttm]
>  [<ffffffffa00c6f45>] radeon_bo_list_validate+0x75/0x90 [radeon]
>  [<ffffffffa00dca22>] radeon_cs_ioctl+0x582/0x950 [radeon]
>  [<ffffffffa002e573>] drm_ioctl+0x4d3/0x580 [drm]
>  [<ffffffffa00dc4a0>] ? radeon_cs_finish_pages+0xf0/0xf0 [radeon]
>  [<ffffffff811c49f9>] do_vfs_ioctl+0x99/0x5a0
>  [<ffffffff812cd3d7>] ? file_has_perm+0x97/0xb0
>  [<ffffffff81103715>] ? rcu_eqs_exit+0x65/0xb0
>  [<ffffffff811c4f91>] sys_ioctl+0x91/0xb0
>  [<ffffffff81695ad4>] tracesys+0xdd/0xe2
> 

I see lots of these [mostly from page fault], the following one (quick and dirty) works for me.

----

Forbid ttm_buffer_object_transfer() sleeping in ISR context, do not pass 
__GFP_WAIT flag to kzalloc():

BUG: sleeping function called from invalid context at mm/slub.c:925
in_atomic(): 1, irqs_disabled(): 0, pid: 2179, name: X
3 locks held by X/2179:
 #0:  (&mm->mmap_sem){++++++}, at: [<ffffffff815fac53>] __do_page_fault+0xf3/0x590
 #1:  (&rdev->pm.mclk_lock){++++.+}, at: [<ffffffff81428e40>] radeon_ttm_fault+0x40/0x70
 #2:  (&(&bdev->fence_lock)->rlock){+.+.+.}, at: [<ffffffff813f3227>] ttm_bo_move_accel_cleanup+0x57/0x330
Pid: 2179, comm: X Not tainted 3.8.0-rc3-dbg-01259-gbdac2d4-dirty #1500
Call Trace:
 [<ffffffff8107adea>] __might_sleep+0x14a/0x200
 [<ffffffff811649c3>] kmem_cache_alloc_trace+0x153/0x240
 [<ffffffff813f33a3>] ttm_bo_move_accel_cleanup+0x1d3/0x330
 [<ffffffff814292e8>] radeon_move_blit.isra.6+0xf8/0x160
 [<ffffffff81429a90>] radeon_bo_move+0xb0/0x1f0
 [<ffffffff813f14b5>] ttm_bo_handle_move_mem+0x275/0x5c0
 [<ffffffff813f26f7>] ttm_bo_move_buffer+0x127/0x140
 [<ffffffff813f27b2>] ttm_bo_validate+0xa2/0x130
 [<ffffffff8142b0ed>] radeon_bo_fault_reserve_notify+0xad/0xe0
 [<ffffffff813f4600>] ttm_bo_vm_fault+0x60/0x370
 [<ffffffff81428e51>] radeon_ttm_fault+0x51/0x70
 [<ffffffff8114150a>] __do_fault+0x6a/0x540
 [<ffffffff81144771>] handle_pte_fault+0x91/0x810
 [<ffffffff815fac53>] ? __do_page_fault+0xf3/0x590
 [<ffffffff811464c1>] handle_mm_fault+0x231/0x330
 [<ffffffff815facbc>] __do_page_fault+0x15c/0x590
 [<ffffffff81184cd9>] ? do_vfs_ioctl+0x99/0x580
 [<ffffffff815ff7c9>] ? sysret_check+0x22/0x5d
 [<ffffffff815fb11b>] do_page_fault+0x2b/0x50
 [<ffffffff815f79a8>] page_fault+0x28/0x30


Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>

---

 drivers/gpu/drm/ttm/ttm_bo_util.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Andreas Bombe Feb. 7, 2013, 9:53 p.m. UTC | #1 
On Sun, Jan 13, 2013 at 02:38:20PM +0300, Sergey Senozhatsky wrote:
> On (01/12/13 20:27), Dave Jones wrote:
> > BUG: sleeping function called from invalid context at mm/slub.c:925
> > in_atomic(): 1, irqs_disabled(): 0, pid: 566, name: Xorg
> > INFO: lockdep is turned off.
> > Pid: 566, comm: Xorg Not tainted 3.8.0-rc3+ #49
> > Call Trace:
> >  [<ffffffff81087ad1>] __might_sleep+0x141/0x200
> >  [<ffffffff8119eddb>] kmem_cache_alloc_trace+0x4b/0x2a0
> >  [<ffffffffa0083fb3>] ttm_bo_move_accel_cleanup+0x1d3/0x330 [ttm]
> >  [<ffffffffa00c5418>] radeon_move_blit.isra.4+0xf8/0x160 [radeon]
> >  [<ffffffffa00c5db0>] radeon_bo_move+0xb0/0x1f0 [radeon]
> >  [<ffffffffa00820ed>] ttm_bo_handle_move_mem+0x27d/0x5d0 [ttm]
> >  [<ffffffff8108dd51>] ? get_parent_ip+0x11/0x50
> >  [<ffffffffa0083307>] ttm_bo_move_buffer+0x127/0x140 [ttm]
> >  [<ffffffffa008717d>] ? ttm_eu_list_ref_sub+0x3d/0x60 [ttm]
> >  [<ffffffffa00833c2>] ttm_bo_validate+0xa2/0x120 [ttm]
> >  [<ffffffffa00c6f45>] radeon_bo_list_validate+0x75/0x90 [radeon]
> >  [<ffffffffa00dca22>] radeon_cs_ioctl+0x582/0x950 [radeon]
> >  [<ffffffffa002e573>] drm_ioctl+0x4d3/0x580 [drm]
> >  [<ffffffffa00dc4a0>] ? radeon_cs_finish_pages+0xf0/0xf0 [radeon]
> >  [<ffffffff811c49f9>] do_vfs_ioctl+0x99/0x5a0
> >  [<ffffffff812cd3d7>] ? file_has_perm+0x97/0xb0
> >  [<ffffffff81103715>] ? rcu_eqs_exit+0x65/0xb0
> >  [<ffffffff811c4f91>] sys_ioctl+0x91/0xb0
> >  [<ffffffff81695ad4>] tracesys+0xdd/0xe2
> > 
> 
> I see lots of these [mostly from page fault], the following one (quick and dirty) works for me.

Is that patch or any other fix being picked up? It's over three weeks
now and I'm still seeing those BUGs with the latest 3.8-rc.
Sergey Senozhatsky Feb. 8, 2013, 7:47 a.m. UTC | #2 
On (02/07/13 22:53), Andreas Bombe wrote:
> On Sun, Jan 13, 2013 at 02:38:20PM +0300, Sergey Senozhatsky wrote:
> > On (01/12/13 20:27), Dave Jones wrote:
> > > BUG: sleeping function called from invalid context at mm/slub.c:925
> > > in_atomic(): 1, irqs_disabled(): 0, pid: 566, name: Xorg
> > > INFO: lockdep is turned off.
> > > Pid: 566, comm: Xorg Not tainted 3.8.0-rc3+ #49
> > > Call Trace:
> > >  [<ffffffff81087ad1>] __might_sleep+0x141/0x200
> > >  [<ffffffff8119eddb>] kmem_cache_alloc_trace+0x4b/0x2a0
> > >  [<ffffffffa0083fb3>] ttm_bo_move_accel_cleanup+0x1d3/0x330 [ttm]
> > >  [<ffffffffa00c5418>] radeon_move_blit.isra.4+0xf8/0x160 [radeon]
> > >  [<ffffffffa00c5db0>] radeon_bo_move+0xb0/0x1f0 [radeon]
> > >  [<ffffffffa00820ed>] ttm_bo_handle_move_mem+0x27d/0x5d0 [ttm]
> > >  [<ffffffff8108dd51>] ? get_parent_ip+0x11/0x50
> > >  [<ffffffffa0083307>] ttm_bo_move_buffer+0x127/0x140 [ttm]
> > >  [<ffffffffa008717d>] ? ttm_eu_list_ref_sub+0x3d/0x60 [ttm]
> > >  [<ffffffffa00833c2>] ttm_bo_validate+0xa2/0x120 [ttm]
> > >  [<ffffffffa00c6f45>] radeon_bo_list_validate+0x75/0x90 [radeon]
> > >  [<ffffffffa00dca22>] radeon_cs_ioctl+0x582/0x950 [radeon]
> > >  [<ffffffffa002e573>] drm_ioctl+0x4d3/0x580 [drm]
> > >  [<ffffffffa00dc4a0>] ? radeon_cs_finish_pages+0xf0/0xf0 [radeon]
> > >  [<ffffffff811c49f9>] do_vfs_ioctl+0x99/0x5a0
> > >  [<ffffffff812cd3d7>] ? file_has_perm+0x97/0xb0
> > >  [<ffffffff81103715>] ? rcu_eqs_exit+0x65/0xb0
> > >  [<ffffffff811c4f91>] sys_ioctl+0x91/0xb0
> > >  [<ffffffff81695ad4>] tracesys+0xdd/0xe2
> > > 
> > 
> > I see lots of these [mostly from page fault], the following one (quick and dirty) works for me.
> 
> Is that patch or any other fix being picked up? It's over three weeks
> now and I'm still seeing those BUGs with the latest 3.8-rc.
> 

None that I'm aware of.


	-ss
Daniel Vetter Feb. 8, 2013, 8:44 a.m. UTC | #3 
On Fri, Feb 08, 2013 at 10:47:42AM +0300, Sergey Senozhatsky wrote:
> On (02/07/13 22:53), Andreas Bombe wrote:
> > On Sun, Jan 13, 2013 at 02:38:20PM +0300, Sergey Senozhatsky wrote:
> > > On (01/12/13 20:27), Dave Jones wrote:
> > > > BUG: sleeping function called from invalid context at mm/slub.c:925
> > > > in_atomic(): 1, irqs_disabled(): 0, pid: 566, name: Xorg
> > > > INFO: lockdep is turned off.
> > > > Pid: 566, comm: Xorg Not tainted 3.8.0-rc3+ #49
> > > > Call Trace:
> > > >  [<ffffffff81087ad1>] __might_sleep+0x141/0x200
> > > >  [<ffffffff8119eddb>] kmem_cache_alloc_trace+0x4b/0x2a0
> > > >  [<ffffffffa0083fb3>] ttm_bo_move_accel_cleanup+0x1d3/0x330 [ttm]
> > > >  [<ffffffffa00c5418>] radeon_move_blit.isra.4+0xf8/0x160 [radeon]
> > > >  [<ffffffffa00c5db0>] radeon_bo_move+0xb0/0x1f0 [radeon]
> > > >  [<ffffffffa00820ed>] ttm_bo_handle_move_mem+0x27d/0x5d0 [ttm]
> > > >  [<ffffffff8108dd51>] ? get_parent_ip+0x11/0x50
> > > >  [<ffffffffa0083307>] ttm_bo_move_buffer+0x127/0x140 [ttm]
> > > >  [<ffffffffa008717d>] ? ttm_eu_list_ref_sub+0x3d/0x60 [ttm]
> > > >  [<ffffffffa00833c2>] ttm_bo_validate+0xa2/0x120 [ttm]
> > > >  [<ffffffffa00c6f45>] radeon_bo_list_validate+0x75/0x90 [radeon]
> > > >  [<ffffffffa00dca22>] radeon_cs_ioctl+0x582/0x950 [radeon]
> > > >  [<ffffffffa002e573>] drm_ioctl+0x4d3/0x580 [drm]
> > > >  [<ffffffffa00dc4a0>] ? radeon_cs_finish_pages+0xf0/0xf0 [radeon]
> > > >  [<ffffffff811c49f9>] do_vfs_ioctl+0x99/0x5a0
> > > >  [<ffffffff812cd3d7>] ? file_has_perm+0x97/0xb0
> > > >  [<ffffffff81103715>] ? rcu_eqs_exit+0x65/0xb0
> > > >  [<ffffffff811c4f91>] sys_ioctl+0x91/0xb0
> > > >  [<ffffffff81695ad4>] tracesys+0xdd/0xe2
> > > > 
> > > 
> > > I see lots of these [mostly from page fault], the following one (quick and dirty) works for me.
> > 
> > Is that patch or any other fix being picked up? It's over three weeks
> > now and I'm still seeing those BUGs with the latest 3.8-rc.
> > 
> 
> None that I'm aware of.

Either this one

https://patchwork.kernel.org/patch/2094501/

or a bit an older approach here

https://patchwork.kernel.org/patch/1972071/

should fix this.

/me looks at Dave

Cheers, Daniel
diff mbox

Patch

diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index d73d6e3..92e0dea 100644
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -422,7 +422,7 @@  static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
 	struct ttm_bo_device *bdev = bo->bdev;
 	struct ttm_bo_driver *driver = bdev->driver;
 
-	fbo = kzalloc(sizeof(*fbo), GFP_KERNEL);
+	fbo = kzalloc(sizeof(*fbo), __GFP_IO | __GFP_FS);
 	if (!fbo)
 		return -ENOMEM;