[PATCH] drm/amdgpu: use spin_lock_irqsave to avoid deadlock by local interrupt

Zytaruk, Kelly Kelly.Zytaruk at amd.com
Mon Jan 10 15:41:47 UTC 2022


I can confirm that the fault no longer occurs when the patch is implemented.

Thanks,
Kelly

-----Original Message-----
From: Chen, Guchun <Guchun.Chen at amd.com> 
Sent: January 10, 2022 12:14 AM
To: amd-gfx at lists.freedesktop.org; Deucher, Alexander <Alexander.Deucher at amd.com>; Koenig, Christian <Christian.Koenig at amd.com>; Pan, Xinhui <Xinhui.Pan at amd.com>; Zytaruk, Kelly <Kelly.Zytaruk at amd.com>; Huang, Qiliang (Warden) <Qiliang.Huang at amd.com>; Zhang, Bokun <Bokun.Zhang at amd.com>
Cc: Chen, Guchun <Guchun.Chen at amd.com>
Subject: [PATCH] drm/amdgpu: use spin_lock_irqsave to avoid deadlock by local interrupt

This is observed in SRIOV case with virtual KMS as display.

_raw_spin_lock_irqsave+0x37/0x40
drm_handle_vblank+0x69/0x350 [drm]
? try_to_wake_up+0x432/0x5c0
? amdgpu_vkms_prepare_fb+0x1c0/0x1c0 [amdgpu]
drm_crtc_handle_vblank+0x17/0x20 [drm]
amdgpu_vkms_vblank_simulate+0x4d/0x80 [amdgpu]
__hrtimer_run_queues+0xfb/0x230
hrtimer_interrupt+0x109/0x220
__sysvec_apic_timer_interrupt+0x64/0xe0
asm_call_irq_on_stack+0x12/0x20

Fixes: ba5317109d0c("drm/amdgpu: create amdgpu_vkms (v4)")
Signed-off-by: Guchun Chen <guchun.chen at amd.com>
---
 drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
index 2dcc68e04e84..d99c8779b51e 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vkms.c
@@ -144,15 +144,16 @@ static void amdgpu_vkms_crtc_atomic_disable(struct drm_crtc *crtc,  static void amdgpu_vkms_crtc_atomic_flush(struct drm_crtc *crtc,
 					  struct drm_atomic_state *state)
 {
+	unsigned long flags;
 	if (crtc->state->event) {
-		spin_lock(&crtc->dev->event_lock);
+		spin_lock_irqsave(&crtc->dev->event_lock, flags);
 
 		if (drm_crtc_vblank_get(crtc) != 0)
 			drm_crtc_send_vblank_event(crtc, crtc->state->event);
 		else
 			drm_crtc_arm_vblank_event(crtc, crtc->state->event);
 
-		spin_unlock(&crtc->dev->event_lock);
+		spin_unlock_irqrestore(&crtc->dev->event_lock, flags);
 
 		crtc->state->event = NULL;
 	}
--
2.17.1



More information about the amd-gfx mailing list