drm/nvc0/gr: fix TP init for transform feedback offset queries

Without this, they return bytes written since the last update of
the offset, but we want the full offset.

Trace shows setting this on GPC[0]/TP[0] is enough.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
diff --git a/drivers/gpu/drm/nouveau/nvc0_graph.c b/drivers/gpu/drm/nouveau/nvc0_graph.c
index a74e501..ecfafd7 100644
--- a/drivers/gpu/drm/nouveau/nvc0_graph.c
+++ b/drivers/gpu/drm/nouveau/nvc0_graph.c
@@ -381,6 +381,8 @@
 	u8  tpnr[GPC_MAX];
 	int i, gpc, tpc;
 
+	nv_wr32(dev, TP_UNIT(0, 0, 0x5c), 1); /* affects TFB offset queries */
+
 	/*
 	 *      TP      ROP UNKVAL(magic_not_rop_nr)
 	 * 450: 4/0/0/0 2        3