drm/radeon: stop page faults from hanging the system (v2)

Redirect invalid memory accesses to the default page
instead of locking up the memory controller. Also
enable the invalid memory access interrupts and
start spamming system log with it.

v2 (agd5f): fix up against 2 level PT changes
Signed-off-by: Christian König <deathsimple@vodafone.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org

drivers/gpu/drm/radeon/evergreen.c

@@ -3093,6 +3093,16 @@ int evergreen_irq_process(struct radeon_device *rdev)
 				break;
 			}
 			break;
+		case 146:
+		case 147:
+			dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data);
+			dev_err(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+				RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR));
+			dev_err(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+				RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS));
+			/* reset addr and status */
+			WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1);
+			break;
 		case 176: /* CP_INT in ring buffer */
 		case 177: /* CP_INT in IB1 */
 		case 178: /* CP_INT in IB2 */

drivers/gpu/drm/radeon/evergreend.h

@@ -699,6 +699,7 @@
 #define		PAGE_TABLE_DEPTH(x)				(((x) & 3) << 1)
 #define		RANGE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 4)
 #define VM_CONTEXT1_CNTL				0x1414
+#define VM_CONTEXT1_CNTL2				0x1434
 #define	VM_CONTEXT0_PAGE_TABLE_BASE_ADDR		0x153C
 #define	VM_CONTEXT0_PAGE_TABLE_END_ADDR			0x157C
 #define	VM_CONTEXT0_PAGE_TABLE_START_ADDR		0x155C
@@ -720,6 +721,8 @@
 #define		CACHE_UPDATE_MODE(x)				((x) << 6)
 #define	VM_L2_STATUS					0x140C
 #define		L2_BUSY						(1 << 0)
+#define	VM_CONTEXT1_PROTECTION_FAULT_ADDR		0x14FC
+#define	VM_CONTEXT1_PROTECTION_FAULT_STATUS		0x14DC
 
 #define	WAIT_UNTIL					0x8040
 

drivers/gpu/drm/radeon/ni.c

@@ -784,10 +784,20 @@ static int cayman_pcie_gart_enable(struct radeon_device *rdev)
 	/* enable context1-7 */
 	WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
 	       (u32)(rdev->dummy_page.addr >> 12));
-	WREG32(VM_CONTEXT1_CNTL2, 0);
-	WREG32(VM_CONTEXT1_CNTL, 0);
+	WREG32(VM_CONTEXT1_CNTL2, 4);
 	WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(1) |
-				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
+				RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT |
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT |
+				PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT |
+				VALID_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT |
+				READ_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT |
+				WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT);
 
 	cayman_pcie_gart_tlb_flush(rdev);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",

drivers/gpu/drm/radeon/nid.h

@@ -80,7 +80,18 @@
 #define VM_CONTEXT0_CNTL				0x1410
 #define		ENABLE_CONTEXT					(1 << 0)
 #define		PAGE_TABLE_DEPTH(x)				(((x) & 3) << 1)
+#define		RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 3)
 #define		RANGE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 4)
+#define		DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT	(1 << 6)
+#define		DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT	(1 << 7)
+#define		PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 9)
+#define		PDE0_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 10)
+#define		VALID_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 12)
+#define		VALID_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 13)
+#define		READ_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 15)
+#define		READ_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 16)
+#define		WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 18)
+#define		WRITE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 19)
 #define VM_CONTEXT1_CNTL				0x1414
 #define VM_CONTEXT0_CNTL2				0x1430
 #define VM_CONTEXT1_CNTL2				0x1434

drivers/gpu/drm/radeon/si.c

@@ -2426,9 +2426,20 @@ static int si_pcie_gart_enable(struct radeon_device *rdev)
 	/* enable context1-15 */
 	WREG32(VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR,
 	       (u32)(rdev->dummy_page.addr >> 12));
-	WREG32(VM_CONTEXT1_CNTL2, 0);
+	WREG32(VM_CONTEXT1_CNTL2, 4);
 	WREG32(VM_CONTEXT1_CNTL, ENABLE_CONTEXT | PAGE_TABLE_DEPTH(1) |
-				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT);
+				RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				RANGE_PROTECTION_FAULT_ENABLE_DEFAULT |
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT |
+				PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				PDE0_PROTECTION_FAULT_ENABLE_DEFAULT |
+				VALID_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				VALID_PROTECTION_FAULT_ENABLE_DEFAULT |
+				READ_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				READ_PROTECTION_FAULT_ENABLE_DEFAULT |
+				WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT |
+				WRITE_PROTECTION_FAULT_ENABLE_DEFAULT);
 
 	si_pcie_gart_tlb_flush(rdev);
 	DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n",
@@ -3684,6 +3695,16 @@ int si_irq_process(struct radeon_device *rdev)
 				break;
 			}
 			break;
+		case 146:
+		case 147:
+			dev_err(rdev->dev, "GPU fault detected: %d 0x%08x\n", src_id, src_data);
+			dev_err(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_ADDR   0x%08X\n",
+				RREG32(VM_CONTEXT1_PROTECTION_FAULT_ADDR));
+			dev_err(rdev->dev, "  VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n",
+				RREG32(VM_CONTEXT1_PROTECTION_FAULT_STATUS));
+			/* reset addr and status */
+			WREG32_P(VM_CONTEXT1_CNTL2, 1, ~1);
+			break;
 		case 176: /* RINGID0 CP_INT */
 			radeon_fence_process(rdev, RADEON_RING_TYPE_GFX_INDEX);
 			break;

drivers/gpu/drm/radeon/sid.h

@@ -91,7 +91,18 @@
 #define VM_CONTEXT0_CNTL				0x1410
 #define		ENABLE_CONTEXT					(1 << 0)
 #define		PAGE_TABLE_DEPTH(x)				(((x) & 3) << 1)
+#define		RANGE_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 3)
 #define		RANGE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 4)
+#define		DUMMY_PAGE_PROTECTION_FAULT_ENABLE_INTERRUPT	(1 << 6)
+#define		DUMMY_PAGE_PROTECTION_FAULT_ENABLE_DEFAULT	(1 << 7)
+#define		PDE0_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 9)
+#define		PDE0_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 10)
+#define		VALID_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 12)
+#define		VALID_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 13)
+#define		READ_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 15)
+#define		READ_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 16)
+#define		WRITE_PROTECTION_FAULT_ENABLE_INTERRUPT		(1 << 18)
+#define		WRITE_PROTECTION_FAULT_ENABLE_DEFAULT		(1 << 19)
 #define VM_CONTEXT1_CNTL				0x1414
 #define VM_CONTEXT0_CNTL2				0x1430
 #define VM_CONTEXT1_CNTL2				0x1434
@@ -104,6 +115,9 @@
 #define	VM_CONTEXT14_PAGE_TABLE_BASE_ADDR		0x1450
 #define	VM_CONTEXT15_PAGE_TABLE_BASE_ADDR		0x1454
 
+#define	VM_CONTEXT1_PROTECTION_FAULT_ADDR		0x14FC
+#define	VM_CONTEXT1_PROTECTION_FAULT_STATUS		0x14DC
+
 #define VM_INVALIDATE_REQUEST				0x1478
 #define VM_INVALIDATE_RESPONSE				0x147c