ARM: dts: rockchip: use non-secure timer as broadcast on rk3288
Subsystems (2)
Heiko Stuebner
maintainerPATCH 1/1 ARM: dts: rockchip: use non-secure timer as broadcast on rk3288
The rk3288 soc has 2 separate timer blocks one with 6 channels and another one with 2 channels that can be set as secure.
Firmware variants using psci might want to make use of that feature making both channels of the secure timer inaccessible for a kernel running in non-secure mode, which affects the timer currently used as broadcast.
Therefore move over to using the other (still unused) timer block for that. The timer clock in the original node was also wrong, as it was pointing to the 6-channel block already, while the clock of the 2ch block gets controlled from the secure-grf.
--- arch/arm/boot/dts/rk3288.dtsi | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
arch/arm/boot/dts/rk3288.dtsi +8 -8
@@ -229,14 +229,6 @@ clock-frequency = <24000000>; }; - timer: timer@ff810000 { - compatible = "rockchip,rk3288-timer"; - reg = <0x0 0xff810000 0x0 0x20>; - interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_HIGH>; - clocks = <&xin24m>, <&cru PCLK_TIMER>; - clock-names = "timer", "pclk"; - }; - display-subsystem { compatible = "rockchip,display-subsystem"; ports = <&vopl_out>, <&vopb_out>;
@@ -714,6 +706,14 @@ status = "disabled"; }; + timer: timer@ff6b0000 { + compatible = "rockchip,rk3288-timer"; + reg = <0x0 0xff6b0000 0x0 0x20>; + interrupts = <GIC_SPI 66 IRQ_TYPE_LEVEL_HIGH>; + clocks = <&xin24m>, <&cru PCLK_TIMER>; + clock-names = "timer", "pclk"; + }; + bus_intmem@ff700000 { compatible = "mmio-sram"; reg = <0x0 0xff700000 0x0 0x18000>;
Signature · 1 line
2.20.1
Heiko Stübner
maintainerAm Freitag, 12. April 2019, 18:00:12 CEST schrieb Heiko Stuebner:
Quoted 8 lines
The rk3288 soc has 2 separate timer blocks one with 6 channels and another one with 2 channels that can be set as secure.
Firmware variants using psci might want to make use of that feature making both channels of the secure timer inaccessible for a kernel running in non-secure mode, which affects the timer currently used as broadcast.
Quoted 6 lines
Therefore move over to using the other (still unused) timer block for that. The timer clock in the original node was also wrong, as it was pointing to the 6-channel block already, while the clock of the 2ch block gets controlled from the secure-grf.
Signed-off-by Heiko Stuebner <[email protected]>
And I just realized, that this is probably dumb, as the timers0-5 are in cpu powerdomain, while 6+7 are in the alive pd, so probably only the secure timers are always on, while 0-5 are likely shutdown during suspend.
Quoted 39 lines
--- arch/arm/boot/dts/rk3288.dtsi | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)arch/arm/boot/dts/rk3288.dtsi +8
-8@@ -229,14 +229,6 @@ clock-frequency = <24000000>; }; - timer: timer@ff810000 { - compatible = "rockchip,rk3288-timer"; - reg = <0x0 0xff810000 0x0 0x20>; - interrupts = <GIC_SPI 72 IRQ_TYPE_LEVEL_HIGH>; - clocks = <&xin24m>, <&cru PCLK_TIMER>; - clock-names = "timer", "pclk"; - }; - display-subsystem { compatible = "rockchip,display-subsystem"; ports = <&vopl_out>, <&vopb_out>;@@ -714,6 +706,14 @@ status = "disabled"; }; + timer: timer@ff6b0000 { + compatible = "rockchip,rk3288-timer"; + reg = <0x0 0xff6b0000 0x0 0x20>; + interrupts = <GIC_SPI 66 IRQ_TYPE_LEVEL_HIGH>; + clocks = <&xin24m>, <&cru PCLK_TIMER>; + clock-names = "timer", "pclk"; + }; + bus_intmem@ff700000 { compatible = "mmio-sram"; reg = <0x0 0xff700000 0x0 0x18000>;