From: chongli2 <chongli2@amd.com>
Date: Wed, 6 Nov 2024 03:43:09 +0000 (+0800)
Subject: drm/amdgpu: fix return random value when multiple threads read registers via mes.
X-Git-Tag: microblaze-v6.16~490^2~6^2~47
X-Git-Url: http://git.monstr.eu/?a=commitdiff_plain;h=f4a3246a2c7a595161f1ba11db53639b7f580104;p=linux-2.6-microblaze.git

drm/amdgpu: fix return random value when multiple threads read registers via mes.

The currect code use the address "adev->mes.read_val_ptr" to
store the value read from register via mes.
So when multiple threads read register,
multiple threads have to share the one address,
and overwrite the value each other.

Assign an address by "amdgpu_device_wb_get" to store register value.
each thread will has an address to store register value.

Signed-off-by: chongli2 <chongli2@amd.com>
Reviewed-by: Emily Deng <Emily.Deng@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
index b10383f83d73..f702b8b9c318 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c
@@ -192,17 +192,6 @@ int amdgpu_mes_init(struct amdgpu_device *adev)
 			(uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs[i]];
 	}
 
-	r = amdgpu_device_wb_get(adev, &adev->mes.read_val_offs);
-	if (r) {
-		dev_err(adev->dev,
-			"(%d) read_val_offs alloc failed\n", r);
-		goto error;
-	}
-	adev->mes.read_val_gpu_addr =
-		adev->wb.gpu_addr + (adev->mes.read_val_offs * 4);
-	adev->mes.read_val_ptr =
-		(uint32_t *)&adev->wb.wb[adev->mes.read_val_offs];
-
 	r = amdgpu_mes_doorbell_init(adev);
 	if (r)
 		goto error;
@@ -223,8 +212,6 @@ error:
 			amdgpu_device_wb_free(adev,
 				      adev->mes.query_status_fence_offs[i]);
 	}
-	if (adev->mes.read_val_ptr)
-		amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
 
 	idr_destroy(&adev->mes.pasid_idr);
 	idr_destroy(&adev->mes.gang_id_idr);
@@ -249,8 +236,6 @@ void amdgpu_mes_fini(struct amdgpu_device *adev)
 			amdgpu_device_wb_free(adev,
 				      adev->mes.query_status_fence_offs[i]);
 	}
-	if (adev->mes.read_val_ptr)
-		amdgpu_device_wb_free(adev, adev->mes.read_val_offs);
 
 	amdgpu_mes_doorbell_free(adev);
 
@@ -921,10 +906,19 @@ uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg)
 {
 	struct mes_misc_op_input op_input;
 	int r, val = 0;
+	uint32_t addr_offset = 0;
+	uint64_t read_val_gpu_addr;
+	uint32_t *read_val_ptr;
 
+	if (amdgpu_device_wb_get(adev, &addr_offset)) {
+		DRM_ERROR("critical bug! too many mes readers\n");
+		goto error;
+	}
+	read_val_gpu_addr = adev->wb.gpu_addr + (addr_offset * 4);
+	read_val_ptr = (uint32_t *)&adev->wb.wb[addr_offset];
 	op_input.op = MES_MISC_OP_READ_REG;
 	op_input.read_reg.reg_offset = reg;
-	op_input.read_reg.buffer_addr = adev->mes.read_val_gpu_addr;
+	op_input.read_reg.buffer_addr = read_val_gpu_addr;
 
 	if (!adev->mes.funcs->misc_op) {
 		DRM_ERROR("mes rreg is not supported!\n");
@@ -935,9 +929,11 @@ uint32_t amdgpu_mes_rreg(struct amdgpu_device *adev, uint32_t reg)
 	if (r)
 		DRM_ERROR("failed to read reg (0x%x)\n", reg);
 	else
-		val = *(adev->mes.read_val_ptr);
+		val = *(read_val_ptr);
 
 error:
+	if (addr_offset)
+		amdgpu_device_wb_free(adev, addr_offset);
 	return val;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
index 0684e482a204..129b8d1cacef 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h
@@ -120,9 +120,6 @@ struct amdgpu_mes {
 	uint32_t			query_status_fence_offs[AMDGPU_MAX_MES_PIPES];
 	uint64_t			query_status_fence_gpu_addr[AMDGPU_MAX_MES_PIPES];
 	uint64_t			*query_status_fence_ptr[AMDGPU_MAX_MES_PIPES];
-	uint32_t                        read_val_offs;
-	uint64_t			read_val_gpu_addr;
-	uint32_t			*read_val_ptr;
 
 	uint32_t			saved_flags;