drm/amdgpu: add an auto setting to the noretry parameter

This allows us to set different defaults on a per asic basis.  This
way we can enable noretry on dGPUs where it can increase performance
in certain cases and disable it on chips where it can be problematic.

For now the default is 0 for all asics, but we may want to try and
enable it again for newer dGPUs.

Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Luben Tuikov <luben.tuikov@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

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

@@ -147,7 +147,7 @@ int amdgpu_async_gfx_ring = 1;
 int amdgpu_mcbp = 0;
 int amdgpu_discovery = -1;
 int amdgpu_mes = 0;
-int amdgpu_noretry;
+int amdgpu_noretry = -1;
 int amdgpu_force_asic_type = -1;
 int amdgpu_tmz = 0;
 int amdgpu_reset_method = -1; /* auto */
@@ -596,8 +596,13 @@ MODULE_PARM_DESC(mes,
 	"Enable Micro Engine Scheduler (0 = disabled (default), 1 = enabled)");
 module_param_named(mes, amdgpu_mes, int, 0444);
 
+/**
+ * DOC: noretry (int)
+ * Disable retry faults in the GPU memory controller.
+ * (0 = retry enabled, 1 = retry disabled, -1 auto (default))
+ */
 MODULE_PARM_DESC(noretry,
-	"Disable retry faults (0 = retry enabled (default), 1 = retry disabled)");
+	"Disable retry faults (0 = retry enabled, 1 = retry disabled, -1 auto (default))");
 module_param_named(noretry, amdgpu_noretry, int, 0644);
 
 /**

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

@@ -424,7 +424,31 @@ void amdgpu_gmc_noretry_set(struct amdgpu_device *adev)
 {
 	struct amdgpu_gmc *gmc = &adev->gmc;
 
-	gmc->noretry = amdgpu_noretry;
+	switch (adev->asic_type) {
+	case CHIP_RAVEN:
+		/* Raven currently has issues with noretry
+		 * regardless of what we decide for other
+		 * asics, we should leave raven with
+		 * noretry = 0 until we root cause the
+		 * issues.
+		 */
+		if (amdgpu_noretry == -1)
+			gmc->noretry = 0;
+		else
+			gmc->noretry = amdgpu_noretry;
+		break;
+	default:
+		/* default this to 0 for now, but we may want
+		 * to change this in the future for certain
+		 * GPUs as it can increase performance in
+		 * certain cases.
+		 */
+		if (amdgpu_noretry == -1)
+			gmc->noretry = 0;
+		else
+			gmc->noretry = amdgpu_noretry;
+		break;
+	}
 }
 
 void amdgpu_gmc_set_vm_fault_masks(struct amdgpu_device *adev, int hub_type,