summaryrefslogtreecommitdiff
path: root/sys/dev
diff options
context:
space:
mode:
authorMark Kettenis <kettenis@cvs.openbsd.org>2018-08-03 16:45:18 +0000
committerMark Kettenis <kettenis@cvs.openbsd.org>2018-08-03 16:45:18 +0000
commitd613fba149a41438d558f92ccd4bd71951d001bf (patch)
tree8219af4ea5a7203a865df8fc081122c65ca32a48 /sys/dev
parent2c4b8de745a0ff2a6a994d7889f98dd2a952d231 (diff)
Implement DVFS support.
ok patrick@
Diffstat (limited to 'sys/dev')
-rw-r--r--sys/dev/fdt/rkclock.c28
1 files changed, 1 insertions, 27 deletions
diff --git a/sys/dev/fdt/rkclock.c b/sys/dev/fdt/rkclock.c
index 28bf5091e30..6161ea5ffa8 100644
--- a/sys/dev/fdt/rkclock.c
+++ b/sys/dev/fdt/rkclock.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: rkclock.c,v 1.27 2018/08/01 15:55:50 kettenis Exp $ */
+/* $OpenBSD: rkclock.c,v 1.28 2018/08/03 16:45:17 kettenis Exp $ */
/*
* Copyright (c) 2017, 2018 Mark Kettenis <kettenis@openbsd.org>
*
@@ -762,38 +762,12 @@ struct rkclock_softc *rk3399_pmucru_sc;
void
rk3399_init(struct rkclock_softc *sc)
{
- int node;
int i;
/* PMUCRU instance should attach before us. */
KASSERT(rk3399_pmucru_sc != NULL);
/*
- * Since the hardware comes up with a really conservative CPU
- * clock frequency, and U-Boot doesn't set it to a more
- * reasonable default, try to do so here. These defaults were
- * chosen assuming that the voltage for both clusters is at
- * least 1.0 V. Only do this on the Firefly-RK3399 for now
- * where this is likely to be true given the default voltages
- * for the regulators on that board.
- */
- node = OF_finddevice("/");
- if (OF_is_compatible(node, "firefly,firefly-rk3399")) {
- uint32_t idx;
-
- /* Run the "LITTLE" cluster at 1.2 GHz. */
- idx = RK3399_ARMCLKL;
- rk3399_set_frequency(sc, &idx, 1200000000);
-
-#ifdef MULTIPROCESSOR
- /* Switch PLL of the "big" cluster into normal mode. */
- HWRITE4(sc, RK3399_CRU_BPLL_CON(3),
- RK3399_CRU_PLL_PLL_WORK_MODE_MASK << 16 |
- RK3399_CRU_PLL_PLL_WORK_MODE_NORMAL);
-#endif
- }
-
- /*
* The U-Boot shipped on the Theobroma Systems RK3399-Q7
* module is buggy and sets the parent of the clock for the
* "big" cluster to LPLL. Undo that mistake here such that