imx-drm: imx-drm-core: Fix circular locking dependency
authorFabio Estevam <fabio.estevam@freescale.com>
Wed, 9 Oct 2013 23:46:45 +0000 (20:46 -0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 11 Oct 2013 22:24:36 +0000 (15:24 -0700)
commitc35d6a37bdcb1dde42dd65b00676696e2dd78f2f
tree7de7c8d5b0a8fb63746a5a539eb06c8d8e591505
parent4ad0fdb0e23a4d31794f0818fb65eefdc8cd3306
imx-drm: imx-drm-core: Fix circular locking dependency

Booting a mx6 with CONFIG_PROVE_LOCKING we get:

======================================================
[ INFO: possible circular locking dependency detected ]
3.12.0-rc4-next-20131009+ #34 Not tainted
-------------------------------------------------------
swapper/0/1 is trying to acquire lock:
 (&imx_drm_device->mutex){+.+.+.}, at: [<804575a8>] imx_drm_encoder_get_mux_id+0x28/0x98

but task is already holding lock:
 (&crtc->mutex){+.+...}, at: [<802fe778>] drm_modeset_lock_all+0x40/0x54

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #2 (&crtc->mutex){+.+...}:
       [<800777d0>] __lock_acquire+0x18d4/0x1c24
       [<80077fec>] lock_acquire+0x68/0x7c
       [<805ead5c>] _mutex_lock_nest_lock+0x58/0x3a8
       [<802fec50>] drm_crtc_init+0x48/0xa8
       [<80457c88>] imx_drm_add_crtc+0xd4/0x144
       [<8045e2e8>] ipu_drm_probe+0x114/0x1fc
       [<80312278>] platform_drv_probe+0x20/0x50
       [<80310c68>] driver_probe_device+0x110/0x22c
       [<80310e20>] __driver_attach+0x9c/0xa0
       [<8030f218>] bus_for_each_dev+0x5c/0x90
       [<80310750>] driver_attach+0x20/0x28
       [<8031034c>] bus_add_driver+0xdc/0x1dc
       [<803114d8>] driver_register+0x80/0xfc
       [<80312198>] __platform_driver_register+0x50/0x64
       [<808172fc>] ipu_drm_driver_init+0x18/0x20
       [<800088c0>] do_one_initcall+0xfc/0x160
       [<807e7c5c>] kernel_init_freeable+0x104/0x1d4
       [<805e2930>] kernel_init+0x10/0xec
       [<8000ea68>] ret_from_fork+0x14/0x2c

-> #1 (&dev->mode_config.mutex){+.+.+.}:
       [<800777d0>] __lock_acquire+0x18d4/0x1c24
       [<80077fec>] lock_acquire+0x68/0x7c
       [<805eb100>] mutex_lock_nested+0x54/0x3a4
       [<802fe758>] drm_modeset_lock_all+0x20/0x54
       [<802fead4>] drm_encoder_init+0x20/0x7c
       [<80457ae4>] imx_drm_add_encoder+0x88/0xec
       [<80459838>] imx_ldb_probe+0x344/0x4fc
       [<80312278>] platform_drv_probe+0x20/0x50
       [<80310c68>] driver_probe_device+0x110/0x22c
       [<80310e20>] __driver_attach+0x9c/0xa0
       [<8030f218>] bus_for_each_dev+0x5c/0x90
       [<80310750>] driver_attach+0x20/0x28
       [<8031034c>] bus_add_driver+0xdc/0x1dc
       [<803114d8>] driver_register+0x80/0xfc
       [<80312198>] __platform_driver_register+0x50/0x64
       [<8081722c>] imx_ldb_driver_init+0x18/0x20
       [<800088c0>] do_one_initcall+0xfc/0x160
       [<807e7c5c>] kernel_init_freeable+0x104/0x1d4
       [<805e2930>] kernel_init+0x10/0xec
       [<8000ea68>] ret_from_fork+0x14/0x2c

-> #0 (&imx_drm_device->mutex){+.+.+.}:
       [<805e510c>] print_circular_bug+0x74/0x2e0
       [<80077ad0>] __lock_acquire+0x1bd4/0x1c24
       [<80077fec>] lock_acquire+0x68/0x7c
       [<805eb100>] mutex_lock_nested+0x54/0x3a4
       [<804575a8>] imx_drm_encoder_get_mux_id+0x28/0x98
       [<80459a98>] imx_ldb_encoder_prepare+0x34/0x114
       [<802ef724>] drm_crtc_helper_set_mode+0x1f0/0x4c0
       [<802f0344>] drm_crtc_helper_set_config+0x828/0x99c
       [<802ff270>] drm_mode_set_config_internal+0x5c/0xdc
       [<802eebe0>] drm_fb_helper_set_par+0x50/0xb4
       [<802af580>] fbcon_init+0x490/0x500
       [<802dd104>] visual_init+0xa8/0xf8
       [<802df414>] do_bind_con_driver+0x140/0x37c
       [<802df764>] do_take_over_console+0x114/0x1c4
       [<802af65c>] do_fbcon_takeover+0x6c/0xd4
       [<802b2b30>] fbcon_event_notify+0x7c8/0x818
       [<80049954>] notifier_call_chain+0x4c/0x8c
       [<80049cd8>] __blocking_notifier_call_chain+0x50/0x68
       [<80049d10>] blocking_notifier_call_chain+0x20/0x28
       [<802a75f0>] fb_notifier_call_chain+0x1c/0x24
       [<802a9224>] register_framebuffer+0x188/0x268
       [<802ee994>] drm_fb_helper_initial_config+0x2bc/0x4b8
       [<802f118c>] drm_fbdev_cma_init+0x7c/0xec
       [<80817288>] imx_fb_helper_init+0x54/0x90
       [<800088c0>] do_one_initcall+0xfc/0x160
       [<807e7c5c>] kernel_init_freeable+0x104/0x1d4
       [<805e2930>] kernel_init+0x10/0xec
       [<8000ea68>] ret_from_fork+0x14/0x2c

other info that might help us debug this:

Chain exists of:
  &imx_drm_device->mutex --> &dev->mode_config.mutex --> &crtc->mutex

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&crtc->mutex);
                               lock(&dev->mode_config.mutex);
                               lock(&crtc->mutex);
  lock(&imx_drm_device->mutex);

 *** DEADLOCK ***

6 locks held by swapper/0/1:
 #0:  (registration_lock){+.+.+.}, at: [<802a90bc>] register_framebuffer+0x20/0x268
 #1:  (&fb_info->lock){+.+.+.}, at: [<802a7a90>] lock_fb_info+0x20/0x44
 #2:  (console_lock){+.+.+.}, at: [<802a9218>] register_framebuffer+0x17c/0x268
 #3:  ((fb_notifier_list).rwsem){.+.+.+}, at: [<80049cbc>] __blocking_notifier_call_chain+0x34/0x68
 #4:  (&dev->mode_config.mutex){+.+.+.}, at: [<802fe758>] drm_modeset_lock_all+0x20/0x54
 #5:  (&crtc->mutex){+.+...}, at: [<802fe778>] drm_modeset_lock_all+0x40/0x54

In order to avoid this lockdep warning, remove the locking from
imx_drm_encoder_get_mux_id() and imx_drm_crtc_panel_format_pins().

Tested on a mx6sabrelite and mx53qsb.

Reported-by: Russell King <rmk+kernel@arm.linux.org.uk>
Tested-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/staging/imx-drm/imx-drm-core.c