Re: [PATCH 6/7] clk: qcom: gcc-sc8180x: Add runtime PM

[Konrad Dybcio]

On 3/9/26 2:06 AM, Val Packett wrote:
> The GCC block on SC8180X is powered by the CX rail. We need to ensure
> that it's enabled to prevent unwanted power collapse.
> 
> Enable runtime PM to keep the power flowing only when necessary.
> 
> Signed-off-by: Val Packett <val@xxxxxxxxxxxx>
> ---

I was always skeptical whether this is useful for GCC - here's an
excerpt from /sys/kernel/debug/pm_genpd/pm_genpd_summary:

cx                              on                              256
                                                gcc_pcie_0_tunnel_gdsc, gcc_pcie_1_tunnel_gdsc, gcc_pcie_2_tunnel_gdsc, gcc_pcie_3_gdsc, gcc_pcie_3_phy_gdsc, gcc_pcie_4_gdsc, gcc_pcie_4_phy_gdsc, gcc_pcie_5_gdsc, gcc_pcie_5_phy_gdsc, gcc_pcie_6_phy_gdsc, gcc_pcie_6a_gdsc, gcc_pcie_6b_gdsc, gcc_ufs_mem_phy_gdsc, gcc_ufs_phy_gdsc, gcc_usb20_prim_gdsc, gcc_usb30_mp_gdsc, gcc_usb30_prim_gdsc, gcc_usb30_sec_gdsc, gcc_usb30_tert_gdsc, gcc_usb3_mp_ss0_phy_gdsc, gcc_usb3_mp_ss1_phy_gdsc, gcc_usb4_0_gdsc, gcc_usb4_1_gdsc, gcc_usb4_2_gdsc, gcc_usb_0_phy_gdsc, gcc_usb_1_phy_gdsc, gcc_usb_2_phy_gdsc
    100000.clock-controller         unsupported                 0           SW
    genpd:0:32300000.remoteproc     suspended                   0           SW
    894000.serial                   active                      64          SW
    a80000.i2c                      suspended                   0           SW
    b80000.i2c                      suspended                   0           SW
    b84000.i2c                      suspended                   0           SW
    b8c000.i2c                      suspended                   0           SW
    b94000.i2c                      suspended                   0           SW
    b9c000.i2c                      suspended                   0           SW

(this is on Hamoa but it's not much different)

You'll notice that the GDSCs are counter-intuitively **not** children
of the clock controller (perhaps "anymore"? maybe that used to be a thing
in the past? IDR)

This means that the GDSCs (and therefore their consumers) have their own
impact on the enable state. IIRC (which may be wrong), the clock controller
would be runtime-active if any of the clocks it provides is, but for that
case, we already (should) have clients voting through OPP.

GCC also has no 'required-opps' (which would make it hold a permanent
nonzero vote like some multimedia clock controllers do, for PLL stability)

I was curious whether 'unsupported' (i.e. not RunPM-enabled) causes the power
to be kept on, and it certainly seems that way:

gcc_pcie_6_phy_gdsc             on                              0
    1bfc000.phy                     unsupported                 0           SW

(note this is *without* pd_ignore_unused)

A zero-but-on vote will be translated into "lowest active state" by the
RPMHPD driver


So perhaps we should do that after all, as even with an aggregated vote of
0, CX may be kept on, but as Dmitry mentioned, .use_rpm is the correct tool
to achieve this. I would appreciate if someone could (n)ack my thoughts..

Konrad