drm/amdgpu: allocate doorbell index for multi-die case

Allocate different doorbell index for kiq/kcq rings
on each die
Signed-off-by: Le Ma <le.ma@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/amd/amdgpu/amdgpu_doorbell.h

@@ -83,6 +83,8 @@ struct amdgpu_doorbell_index {
 	};
 	uint32_t first_non_cp;
 	uint32_t last_non_cp;
+	uint32_t xcc1_kiq_start;
+	uint32_t xcc1_mec_ring0_start;
 	uint32_t max_assignment;
 	/* Per engine SDMA doorbell size in dword */
 	uint32_t sdma_doorbell_range;
@@ -164,7 +166,12 @@ typedef enum _AMDGPU_VEGA20_DOORBELL_ASSIGNMENT
 	AMDGPU_VEGA20_DOORBELL64_FIRST_NON_CP            = AMDGPU_VEGA20_DOORBELL_sDMA_ENGINE0,
 	AMDGPU_VEGA20_DOORBELL64_LAST_NON_CP             = AMDGPU_VEGA20_DOORBELL64_VCE_RING6_7,
 
-	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x18F,
+	/* kiq/kcq from second XCD. Max 8 XCDs */
+	AMDGPU_VEGA20_DOORBELL_XCC1_KIQ_START             = 0x190,
+	/* 8 compute rings per GC. Max to 0x1CE */
+	AMDGPU_VEGA20_DOORBELL_XCC1_MEC_RING0_START       = 0x197,
+
+	AMDGPU_VEGA20_DOORBELL_MAX_ASSIGNMENT            = 0x1CE,
 	AMDGPU_VEGA20_DOORBELL_INVALID                   = 0xFFFF
 } AMDGPU_VEGA20_DOORBELL_ASSIGNMENT;
 

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

@@ -316,6 +316,11 @@ int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev,
 	ring->doorbell_index = adev->doorbell_index.kiq;
 	ring->xcc_id = xcc_id;
 	ring->vm_hub = AMDGPU_GFXHUB_0;
+	if (xcc_id >= 1)
+		ring->doorbell_index = adev->doorbell_index.xcc1_kiq_start +
+					xcc_id - 1;
+	else
+		ring->doorbell_index = adev->doorbell_index.kiq;
 
 	r = amdgpu_gfx_kiq_acquire(adev, ring, xcc_id);
 	if (r)

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

@@ -270,6 +270,7 @@ void amdgpu_ih_decode_iv_helper(struct amdgpu_device *adev,
 	entry->timestamp = dw[1] | ((u64)(dw[2] & 0xffff) << 32);
 	entry->timestamp_src = dw[2] >> 31;
 	entry->pasid = dw[3] & 0xffff;
+	entry->node_id = (dw[3] >> 16) & 0xff;
 	entry->pasid_src = dw[3] >> 31;
 	entry->src_data[0] = dw[4];
 	entry->src_data[1] = dw[5];

drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h

@@ -53,6 +53,7 @@ struct amdgpu_iv_entry {
 	uint64_t timestamp;
 	unsigned timestamp_src;
 	unsigned pasid;
+	unsigned node_id;
 	unsigned pasid_src;
 	unsigned src_data[AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW];
 	const uint32_t *iv_entry;

drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c

@@ -752,7 +752,13 @@ static int gfx_v9_4_3_compute_ring_init(struct amdgpu_device *adev, int ring_id,
 
 	ring->ring_obj = NULL;
 	ring->use_doorbell = true;
-	ring->doorbell_index = (adev->doorbell_index.mec_ring0 + ring_id) << 1;
+	if (xcc_id >= 1)
+		ring->doorbell_index =
+				(adev->doorbell_index.xcc1_mec_ring0_start +
+				ring_id - adev->gfx.num_compute_rings) << 1;
+	else
+		ring->doorbell_index =
+				(adev->doorbell_index.mec_ring0 + ring_id) << 1;
 	ring->eop_gpu_addr = adev->gfx.mec.hpd_eop_gpu_addr
 				+ (ring_id * GFX9_MEC_HPD_SIZE);
 	ring->vm_hub = AMDGPU_GFXHUB_0;