This reverts commit 1d34b9757523c1ad547bd6d040381f62d74a3189.
RK3228 Only GPLL and CPLL, GPLL is a common clock, does not allow
dclk_vop to change its frequency, CPLL is used by GMAC,
if dclk_vop use CLK_SET_RATE_PARENT and CLK_SET_RATE_NO_REPARENT flags will
affect the GMAC function.
If the client application does not use GMAC and CPLL is free, make this
change on the local branch.
Signed-off-by: Elaine Zhang <zhangqing@rock-chips.com>
---
drivers/clk/rockchip/clk-rk3228.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/clk/rockchip/clk-rk3228.c b/drivers/clk/rockchip/clk-rk3228.c
index ed602c27b624..9c0284607766 100644
--- a/drivers/clk/rockchip/clk-rk3228.c
+++ b/drivers/clk/rockchip/clk-rk3228.c
@@ -409,7 +409,7 @@ static struct rockchip_clk_branch rk3228_clk_branches[] __initdata = {
RK2928_CLKSEL_CON(29), 0, 3, DFLAGS),
DIV(0, "sclk_vop_pre", "sclk_vop_src", 0,
RK2928_CLKSEL_CON(27), 8, 8, DFLAGS),
- MUX(DCLK_VOP, "dclk_vop", mux_dclk_vop_p, CLK_SET_RATE_PARENT | CLK_SET_RATE_NO_REPARENT,
+ MUX(DCLK_VOP, "dclk_vop", mux_dclk_vop_p, 0,
RK2928_CLKSEL_CON(27), 1, 1, MFLAGS),
FACTOR(0, "xin12m", "xin24m", 0, 1, 2),
--
2.17.1
Hi Elaine,
Am 25.01.25 um 02:15 schrieb Elaine Zhang:
> This reverts commit 1d34b9757523c1ad547bd6d040381f62d74a3189.
>
> RK3228 Only GPLL and CPLL, GPLL is a common clock, does not allow
> dclk_vop to change its frequency, CPLL is used by GMAC,
> if dclk_vop use CLK_SET_RATE_PARENT and CLK_SET_RATE_NO_REPARENT flags will
> affect the GMAC function.
>
how do you come to this conclusion?
On the RK3228, hdmiphy is the default parent of dclk_vop, a clock that is
not even generated by the CRU but is the output of an external HDMI PHY.
The CLK_SET_RATE_NO_REPARENT flag ensures that the parent of dclk_vop never
changes to sclk_vop_pre (and thus never uses CPLL or GPLL). With
CLK_SET_RATE_PARENT we only ensure that we can use all rates of [0] since
there is no divider between dclk_vop and hdmiphy. Overall it is pretty much
the same situation as for RK3328, which handles these clocks in the same
way (see dclk_lcdc for RK3328).
Regards,
Alex
[0] [1]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/rockchip/phy-rockchip-inno-hdmi.c?h=v6.13#n293
> If the client application does not use GMAC and CPLL is free, make this
> change on the local branch.
>
> Signed-off-by: Elaine Zhang <zhangqing@rock-chips.com>
> ---
> drivers/clk/rockchip/clk-rk3228.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/clk/rockchip/clk-rk3228.c b/drivers/clk/rockchip/clk-rk3228.c
> index ed602c27b624..9c0284607766 100644
> --- a/drivers/clk/rockchip/clk-rk3228.c
> +++ b/drivers/clk/rockchip/clk-rk3228.c
> @@ -409,7 +409,7 @@ static struct rockchip_clk_branch rk3228_clk_branches[] __initdata = {
> RK2928_CLKSEL_CON(29), 0, 3, DFLAGS),
> DIV(0, "sclk_vop_pre", "sclk_vop_src", 0,
> RK2928_CLKSEL_CON(27), 8, 8, DFLAGS),
> - MUX(DCLK_VOP, "dclk_vop", mux_dclk_vop_p, CLK_SET_RATE_PARENT | CLK_SET_RATE_NO_REPARENT,
> + MUX(DCLK_VOP, "dclk_vop", mux_dclk_vop_p, 0,
> RK2928_CLKSEL_CON(27), 1, 1, MFLAGS),
>
> FACTOR(0, "xin12m", "xin24m", 0, 1, 2),
Hi Elaine,
On 2025-01-25 06:38, Alex Bee wrote:
> Hi Elaine,
>
> Am 25.01.25 um 02:15 schrieb Elaine Zhang:
>> This reverts commit 1d34b9757523c1ad547bd6d040381f62d74a3189.
>>
>> RK3228 Only GPLL and CPLL, GPLL is a common clock, does not allow
>> dclk_vop to change its frequency, CPLL is used by GMAC,
>> if dclk_vop use CLK_SET_RATE_PARENT and CLK_SET_RATE_NO_REPARENT flags will
>> affect the GMAC function.
>>
> how do you come to this conclusion?
>
> On the RK3228, hdmiphy is the default parent of dclk_vop, a clock that is
> not even generated by the CRU but is the output of an external HDMI PHY.
> The CLK_SET_RATE_NO_REPARENT flag ensures that the parent of dclk_vop never
> changes to sclk_vop_pre (and thus never uses CPLL or GPLL). With
> CLK_SET_RATE_PARENT we only ensure that we can use all rates of [0] since
> there is no divider between dclk_vop and hdmiphy. Overall it is pretty much
> the same situation as for RK3328, which handles these clocks in the same
> way (see dclk_lcdc for RK3328).
I came to similar conclusion for v1 of this patch, see [2].
Maybe we should assign clk parent in DT similar to rk3328.dtsi to make
it extra clear that hdmiphy should be used as parent?
assigned-clocks = <&cru DCLK_VOP>;
assigned-clock-parents = <&cru SCLK_HDMI_PHY>;
Also for next revert patch you send, please include the patch author in
the recipient list :-)
[2] https://lore.kernel.org/all/d95d317c-5f6e-42bd-94a9-e1a6c7685e2f@kwiboo.se/
Regards,
Jonas
>
> Regards,
> Alex
>
> [0] [1]
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/phy/rockchip/phy-rockchip-inno-hdmi.c?h=v6.13#n293
>> If the client application does not use GMAC and CPLL is free, make this
>> change on the local branch.
>>
>> Signed-off-by: Elaine Zhang <zhangqing@rock-chips.com>
>> ---
>> drivers/clk/rockchip/clk-rk3228.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/clk/rockchip/clk-rk3228.c b/drivers/clk/rockchip/clk-rk3228.c
>> index ed602c27b624..9c0284607766 100644
>> --- a/drivers/clk/rockchip/clk-rk3228.c
>> +++ b/drivers/clk/rockchip/clk-rk3228.c
>> @@ -409,7 +409,7 @@ static struct rockchip_clk_branch rk3228_clk_branches[] __initdata = {
>> RK2928_CLKSEL_CON(29), 0, 3, DFLAGS),
>> DIV(0, "sclk_vop_pre", "sclk_vop_src", 0,
>> RK2928_CLKSEL_CON(27), 8, 8, DFLAGS),
>> - MUX(DCLK_VOP, "dclk_vop", mux_dclk_vop_p, CLK_SET_RATE_PARENT | CLK_SET_RATE_NO_REPARENT,
>> + MUX(DCLK_VOP, "dclk_vop", mux_dclk_vop_p, 0,
>> RK2928_CLKSEL_CON(27), 1, 1, MFLAGS),
>>
>> FACTOR(0, "xin12m", "xin24m", 0, 1, 2),
>
>
© 2016 - 2026 Red Hat, Inc.