drm/radeon: re-enable PTE/PDE packet for set_page on cayman/TN

PTE/PDE doesn't support a single update (count = 1).  We had
previously disabled it since it we were hitting that case which
let to hangs.  The PTE/PDE packet is much more efficient for VM
updates where it can be used.
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

drivers/gpu/drm/radeon/ni.c

@@ -2079,6 +2079,8 @@ void cayman_vm_set_page(struct radeon_device *rdev,
 			}
 		}
 	} else {
+		if ((flags & RADEON_VM_PAGE_SYSTEM) ||
+		    (count == 1)) {
 			while (count) {
 				ndw = count * 2;
 				if (ndw > 0xFFFFE)
@@ -2105,6 +2107,33 @@ void cayman_vm_set_page(struct radeon_device *rdev,
 			}
 			while (ib->length_dw & 0x7)
 				ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0);
+		} else {
+			while (count) {
+				ndw = count * 2;
+				if (ndw > 0xFFFFE)
+					ndw = 0xFFFFE;
+
+				if (flags & RADEON_VM_PAGE_VALID)
+					value = addr;
+				else
+					value = 0;
+				/* for physically contiguous pages (vram) */
+				ib->ptr[ib->length_dw++] = DMA_PTE_PDE_PACKET(ndw);
+				ib->ptr[ib->length_dw++] = pe; /* dst addr */
+				ib->ptr[ib->length_dw++] = upper_32_bits(pe) & 0xff;
+				ib->ptr[ib->length_dw++] = r600_flags; /* mask */
+				ib->ptr[ib->length_dw++] = 0;
+				ib->ptr[ib->length_dw++] = value; /* value */
+				ib->ptr[ib->length_dw++] = upper_32_bits(value);
+				ib->ptr[ib->length_dw++] = incr; /* increment size */
+				ib->ptr[ib->length_dw++] = 0;
+				pe += ndw * 4;
+				addr += (ndw / 2) * incr;
+				count -= ndw / 2;
+			}
+		}
+		while (ib->length_dw & 0x7)
+			ib->ptr[ib->length_dw++] = DMA_PACKET(DMA_PACKET_NOP, 0, 0, 0);
 	}
 }
 

drivers/gpu/drm/radeon/nid.h

@@ -684,6 +684,11 @@
 					 (((vmid) & 0xF) << 20) |	\
 					 (((n) & 0xFFFFF) << 0))
 
+#define DMA_PTE_PDE_PACKET(n)		((2 << 28) |			\
+					 (1 << 26) |			\
+					 (1 << 21) |			\
+					 (((n) & 0xFFFFF) << 0))
+
 /* async DMA Packet types */
 #define	DMA_PACKET_WRITE				  0x2
 #define	DMA_PACKET_COPY					  0x3