Re: [PATCHv2 1/2] clk: rockchip: rk3588: make gate linked clocks ignore unused

[Jagan Teki]

Hi Sebastian,

On Tue, 4 Apr 2023 at 01:03, Sebastian Reichel
<sebastian.reichel@xxxxxxxxxxxxx> wrote:
>
> RK3588 has a couple of hardware blocks called Native Interface Unit
> (NIU) that gate the clocks to devices behind them. Effectively this
> means that some clocks require two parent clocks being enabled.
> Downstream implemented this by using a separate clock driver
> ("clk-link") for them, which enables the second clock using PM
> framework.
>
> In the upstream kernel we are currently missing support for the second
> parent. The information about it is in the GATE_LINK() macro as
> linkname, but that is not used. Thus the second parent clock is not
> properly enabled. So far this did not really matter, since these clocks
> are mostly required for the more advanced IP blocks, that are not yet
> supported upstream. As this is about to change we need a fix. There
> are three options available:
>
> 1. Properly implement support for having two parent clocks in the
>    clock framework.
> 2. Mark the affected clocks CLK_IGNORE_UNUSED, so that they are not
>    disabled. This wastes some power, but keeps the hack contained
>    within the clock driver. Going from this to the first solution
>    is easy once that has been implemented.
> 3. Enabling the extra clock in the consumer driver. This leaks some
>    implementation details into DT.
>
> This patch implements the second option as an intermediate solution
> until the first one is available. I used an alias for CLK_IS_CRITICAL,
> so that it's easy to see which clocks are not really critical once
> the clock framework supports a better way to implement this.
>
> Tested-by: Vincent Legoll <vincent.legoll@xxxxxxxxx>
> Signed-off-by: Sebastian Reichel <sebastian.reichel@xxxxxxxxxxxxx>
> ---
>  drivers/clk/rockchip/clk-rk3588.c | 42 +++++++++++++++++++------------
>  1 file changed, 26 insertions(+), 16 deletions(-)
>
> diff --git a/drivers/clk/rockchip/clk-rk3588.c b/drivers/clk/rockchip/clk-rk3588.c
> index b7ce3fbd6fa6..6994165e0395 100644
> --- a/drivers/clk/rockchip/clk-rk3588.c
> +++ b/drivers/clk/rockchip/clk-rk3588.c
> @@ -13,15 +13,25 @@
>  #include "clk.h"
>
>  /*
> - * GATE with additional linked clock. Downstream enables the linked clock
> - * (via runtime PM) whenever the gate is enabled. The downstream implementation
> - * does this via separate clock nodes for each of the linked gate clocks,
> - * which leaks parts of the clock tree into DT. It is unclear why this is
> - * actually needed and things work without it for simple use cases. Thus
> - * the linked clock is ignored for now.
> + * Recent Rockchip SoCs have a new hardware block called Native Interface
> + * Unit (NIU), which gates clocks to devices behind them. These effectively
> + * need two parent clocks.
> + *
> + * Downstream enables the linked clock via runtime PM whenever the gate is
> + * enabled. This implementation uses separate clock nodes for each of the
> + * linked gate clocks, which leaks parts of the clock tree into DT.
> + *
> + * The GATE_LINK macro instead takes the second parent via 'linkname', but
> + * ignores the information. Once the clock framework is ready to handle it, the
> + * information should be passed on here. But since these clocks are required to
> + * access multiple relevant IP blocks, such as PCIe or USB, we mark all linked
> + * clocks critical until a better solution is available. This will waste some
> + * power, but avoids leaking implementation details into DT or hanging the
> + * system.
>   */

Does it mean the clk-link topology in the downstream kernel can be
reused the same as normal clock notation?

For example, I'm trying to add HCLK_VO1 directly to VO1 syscon instead
of routing to pclk_vo1_grf(done downstream)
      vo1_grf: syscon@fd5a8000 {
               compatible = "rockchip,rk3588-vo-grf", "syscon";
               reg = <0x0 0xfd5a8000 0x0 0x100>;
             clocks = <&cru HCLK_VO1>;
      };

This seems breaking syscon for vo1_grf and observed a bus error while
accessing regmap. I remember in one of the RKDC discussion that the
double parenting of these clocks is mandatory while accessing
associated IP blocks. Any thoughts?

Thanks,
Jagan.