On 09/12/2021 21:40, Bjorn Andersson wrote:
On Tue 07 Dec 18:22 PST 2021, Dmitry Baryshkov wrote:
To stop disp_cc_mdss_mdp_clk_src from getting stuck during boot if it
was enabled by the bootloader, part it to the TCXO clock source.
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxx>
---
drivers/clk/qcom/dispcc-sdm845.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/clk/qcom/dispcc-sdm845.c b/drivers/clk/qcom/dispcc-sdm845.c
index 735adfefc379..f2afbba7bc72 100644
--- a/drivers/clk/qcom/dispcc-sdm845.c
+++ b/drivers/clk/qcom/dispcc-sdm845.c
@@ -858,6 +858,9 @@ static int disp_cc_sdm845_probe(struct platform_device *pdev)
clk_fabia_pll_configure(&disp_cc_pll0, regmap, &disp_cc_pll0_config);
+ /* Park disp_cc_mdss_mdp_clk_src */
+ clk_rcg2_park_safely(regmap, 0x2088, 0);
Today booting the system with "clk_ignore_unused" will give you a
working efifb up until the point where the display driver kicks in and
reinitializes the hardware state - which during development might be
indefinite.
If we blindly cut the mdp_clk_src here that will no longer be possible.
Yep. There is an opposite issue. I was getting the rcg2 stuck messages
_before_ the clk_disable_unused kicks in. So we definitely need
something more complex than both our proposals. Not to mention that
Steev tested that this does not fix the issue on C630.
Regards,
Bjorn
+
/* Enable hardware clock gating for DSI and MDP clocks */
regmap_update_bits(regmap, 0x8000, 0x7f0, 0x7f0);
--
2.33.0
--
With best wishes
Dmitry
