drm/amdgpu: Rearm IRQ in Vega10 SR-IOV if IRQ lost

In Multi-VFs stress test, sometimes we see IRQ lost when running
benchmark, just rearm it.

Signed-off-by: Trigger Huang <[e-mail]>
Acked-by: Christian König <[e-mail]>
Signed-off-by: Alex Deucher <[e-mail]>

drivers/gpu/drm/amd/amdgpu/vega10_ih.c

···
 #include "soc15_common.h"
 #include "vega10_ih.h"
 
-
+#define MAX_REARM_RETRY 10
 
 static void vega10_ih_set_interrupt_funcs(struct amdgpu_device *adev);
 
···
 }
 
 /**
+ * vega10_ih_irq_rearm - rearm IRQ if lost
+ *
+ * @adev: amdgpu_device pointer
+ *
+ */
+static void vega10_ih_irq_rearm(struct amdgpu_device *adev,
+			       struct amdgpu_ih_ring *ih)
+{
+	uint32_t reg_rptr = 0;
+	uint32_t v = 0;
+	uint32_t i = 0;
+
+	if (ih == &adev->irq.ih)
+		reg_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR);
+	else if (ih == &adev->irq.ih1)
+		reg_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR_RING1);
+	else if (ih == &adev->irq.ih2)
+		reg_rptr = SOC15_REG_OFFSET(OSSSYS, 0, mmIH_RB_RPTR_RING2);
+	else
+		return;
+
+	/* Rearm IRQ / re-wwrite doorbell if doorbell write is lost */
+	for (i = 0; i < MAX_REARM_RETRY; i++) {
+		v = RREG32_NO_KIQ(reg_rptr);
+		if ((v < ih->ring_size) && (v != ih->rptr))
+			WDOORBELL32(ih->doorbell_index, ih->rptr);
+		else
+			break;
+	}
+}
+
+/**
  * vega10_ih_set_rptr - set the IH ring buffer rptr
  *
  * @adev: amdgpu_device pointer
···
 		/* XXX check if swapping is necessary on BE */
 		*ih->rptr_cpu = ih->rptr;
 		WDOORBELL32(ih->doorbell_index, ih->rptr);
+
+		if (amdgpu_sriov_vf(adev))
+			vega10_ih_irq_rearm(adev, ih);
 	} else if (ih == &adev->irq.ih) {
 		WREG32_SOC15(OSSSYS, 0, mmIH_RB_RPTR, ih->rptr);
 	} else if (ih == &adev->irq.ih1) {