drm: fb helper should avoid sleeping in panic context
authorRui Wang <rui.y.wang@intel.com>
Mon, 15 Dec 2014 19:28:26 +0000 (11:28 -0800)
committerDave Airlie <airlied@redhat.com>
Tue, 20 Jan 2015 23:26:15 +0000 (09:26 +1000)
commit9aa609e1a3846d3c17087b62579867bab0f488de
treef71359659439c5fe60588610c7963c74d9be09b4
parente902027a64589e53b6d6d7851438ffb4c65a572b
drm: fb helper should avoid sleeping in panic context

There are still some places in the fb helper that need to avoid
sleeping in panic context. Here's an example:

[   65.615496] bad: scheduling from the idle thread!
[   65.620747] CPU: 92 PID: 0 Comm: swapper/92 Tainted: G   M        E  3.18.0-rc4-7-default+ #20

[   65.630364] Hardware name: Intel Corporation BRICKLAND/BRICKLAND, BIOS
BRHSXSD1.86B.0056.R01.1409242327 09/24/2014
[   65.641923]  ffff88087f693d80 ffff88087f689878 ffffffff81566db9 0000000000000000
[   65.650226]  ffff88087f693d80 ffff88087f689898 ffffffff810871ff ffff88046eb3e0d0
[   65.658527]  ffff88087f693d80 ffff88087f6898c8 ffffffff8107c1fa 000000017f6898b8
[   65.666830] Call Trace:
[   65.669557]  <#MC>  [<ffffffff81566db9>] dump_stack+0x46/0x58
[   65.675994]  [<ffffffff810871ff>] dequeue_task_idle+0x2f/0x40
[   65.682412]  [<ffffffff8107c1fa>] dequeue_task+0x5a/0x80
[   65.688345]  [<ffffffff810804f3>] deactivate_task+0x23/0x30
[   65.694569]  [<ffffffff81569050>] __schedule+0x580/0x7f0
[   65.700502]  [<ffffffff81569739>] schedule_preempt_disabled+0x29/0x70
[   65.707696]  [<ffffffff8156abb6>] __ww_mutex_lock_slowpath+0xb8/0x162
[   65.714891]  [<ffffffff8156acb3>] __ww_mutex_lock+0x53/0x85
[   65.721125]  [<ffffffffa00b3a5d>] drm_modeset_lock+0x3d/0x110 [drm]
[   65.728132]  [<ffffffffa00b3c2a>] __drm_modeset_lock_all+0x8a/0x120 [drm]
[   65.735721]  [<ffffffffa00b3cd0>] drm_modeset_lock_all+0x10/0x30 [drm]
[   65.743015]  [<ffffffffa01af8bf>] drm_fb_helper_pan_display+0x2f/0xf0 [drm_kms_helper]
[   65.751857]  [<ffffffff8132bd21>] fb_pan_display+0xd1/0x1a0
[   65.758081]  [<ffffffff81326010>] bit_update_start+0x20/0x50
[   65.764400]  [<ffffffff813259f2>] fbcon_switch+0x3a2/0x550
[   65.770528]  [<ffffffff813a01c9>] redraw_screen+0x189/0x240
[   65.776750]  [<ffffffff81322f8a>] fbcon_blank+0x20a/0x2d0
[   65.782778]  [<ffffffff8137d359>] ? erst_writer+0x209/0x330
[   65.789002]  [<ffffffff810ba2f3>] ? internal_add_timer+0x63/0x80
[   65.795710]  [<ffffffff810bc137>] ? mod_timer+0x127/0x1e0
[   65.801740]  [<ffffffff813a0cd8>] do_unblank_screen+0xa8/0x1d0
[   65.808255]  [<ffffffff813a0e10>] unblank_screen+0x10/0x20
[   65.814381]  [<ffffffff812ca0d9>] bust_spinlocks+0x19/0x40
[   65.820508]  [<ffffffff81561ca7>] panic+0x106/0x1f5
[   65.825955]  [<ffffffff8102336c>] mce_panic+0x2ac/0x2e0
[   65.831789]  [<ffffffff812c796a>] ? delay_tsc+0x4a/0x80
[   65.837625]  [<ffffffff81024e1f>] do_machine_check+0xbaf/0xbf0
[   65.844138]  [<ffffffff813365d7>] ? intel_idle+0xc7/0x150
[   65.850166]  [<ffffffff8156f03f>] machine_check+0x1f/0x30
[   65.856195]  [<ffffffff813365d7>] ? intel_idle+0xc7/0x150
[   65.862222]  <<EOE>>  [<ffffffff814283d5>] cpuidle_enter_state+0x55/0x170
[   65.869823]  [<ffffffff814285a7>] cpuidle_enter+0x17/0x20
[   65.875852]  [<ffffffff81097b08>] cpu_startup_entry+0x2d8/0x370
[   65.882467]  [<ffffffff8102fe29>] start_secondary+0x159/0x180

There's __drm_modeset_lock_all() which Daniel Vetter introduced for this
purpose. We can leverage that without reinventing anything. This patch
works with the latest kernel.

Reviewed-by: Rob Clark <robdclark@gmail.com>
Tested-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Rui Wang <rui.y.wang@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
drivers/gpu/drm/drm_fb_helper.c