[Intel-gfx] [PATCH] drm/i915: Release the forgotten connector reference

[Chris Wilson]

On Thu, Feb 02, 2017 at 12:35:26PM +0100, Daniel Vetter wrote:
> On Thu, Feb 02, 2017 at 08:16:23AM +0000, Chris Wilson wrote:
> > On Thu, Feb 02, 2017 at 10:08:57AM +0200, Joonas Lahtinen wrote:
> > > On ke, 2017-02-01 at 19:58 +0000, Chris Wilson wrote:
> > > > The reference was gained in
> > > > intel_modeset_update_connector_atomic_state() [called from
> > > > intel_modeset_setup_hw_state()] and is never lost if no client ever
> > > > performs a modeset.
> > > > 
> > > > [  649.836069] WARNING: CPU: 6 PID: 8865 at drivers/gpu/drm/drm_mode_config.c:424 drm_mode_config_cleanup+0x21b/0x290 [drm]
> > > > [  649.836078] Modules linked in: i915(-) intel_gtt drm_kms_helper cfbfillrect syscopyarea cfbimgblt sysfillrect sysimgblt fb_sys_fops cfbcopyarea drm
> > > > [  649.836099] CPU: 6 PID: 8865 Comm: drv_selftest Not tainted 4.10.0-rc6+ #908
> > > > [  649.836106] Hardware name: Intel Corporation 2012 Client Platform/Emerald Lake 2, BIOS ACRVMBY1.86C.0078.P00.1201161002 01/16/2012
> > > > [  649.836114] Call Trace:
> > > > [  649.836125]  dump_stack+0x4d/0x6f
> > > > [  649.836136]  __warn+0xc1/0xe0
> > > > [  649.836144]  warn_slowpath_null+0x18/0x20
> > > > [  649.836163]  drm_mode_config_cleanup+0x21b/0x290 [drm]
> > > > [  649.836213]  intel_modeset_cleanup+0x59/0xa0 [i915]
> > > > [  649.836242]  i915_driver_unload+0x84/0x170 [i915]
> > > > [  649.836277]  i915_pci_remove+0x14/0x20 [i915]
> > > > [  649.836287]  pci_device_remove+0x28/0x60
> > > > [  649.836301]  device_release_driver_internal+0x132/0x1d0
> > > > [  649.836313]  driver_detach+0x3a/0x80
> > > > [  649.836324]  bus_remove_driver+0x47/0xa0
> > > > [  649.836335]  driver_unregister+0x27/0x50
> > > > [  649.836344]  pci_unregister_driver+0x34/0xa0
> > > > [  649.836387]  i915_exit+0x1a/0x71 [i915]
> > > > [  649.836401]  SyS_delete_module+0x173/0x1c0
> > > > [  649.836413]  entry_SYSCALL_64_fastpath+0x17/0x98
> > > > [  649.836422] RIP: 0033:0x7f8d5a841ee7
> > > > [  649.836432] RSP: 002b:00007fff89161a28 EFLAGS: 00000206 ORIG_RAX: 00000000000000b0
> > > > [  649.836448] RAX: ffffffffffffffda RBX: 000055f6cd1db5c0 RCX: 00007f8d5a841ee7
> > > > [  649.836458] RDX: 0000000000000000 RSI: 0000000000000800 RDI: 000055f6cd1da7b8
> > > > [  649.836467] RBP: 00007f8d5aaee440 R08: 0000000000000000 R09: 00007fff89161a58
> > > > [  649.836476] R10: 0000000000000062 R11: 0000000000000206 R12: 0000000000000000
> > > > [  649.836486] R13: 000055f6cd1d9010 R14: 000000000000003a R15: 00007fff891609f0
> > > > [  649.836514] ---[ end trace 0e529da316e2a3d1 ]---
> > > > [  649.836536] [drm:drm_mode_config_cleanup [drm]] *ERROR* connector VGA-1 leaked!
> > > > 
> > > > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > > 
> > > Reviewed-by: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
> > 
> > I'm hoping that this won't be required after Maarten's force disable all
> > CRTC on unload. We will see.
> 
> Yeah, I think landing that would be better, but if this is annoying CI or
> something like that, I'm ok with merging already with a FIXME comment
> added.

By itself the first force disable CRTCs on unload wasn't enough, but
Maarten has a second patch to apply more force that does. This patch is
not required.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre