Hi Simon, On 8/8/2026 7:30 PM, Simon Glass wrote: > Hi Jonas, > > On 2026-08-04T00:26:39, Jonas Karlman <[email protected]> wrote: >> clk: rockchip: rk3568: Always init clocks in SPL build >> >> The ACLK_BUS clock is initialized to 150 MHz in SPL on RK3568. > >> >> Ensure clock driver is always probed after bind for SPL builds to >> initialize core PLLs and clocks in SPL phase on RK3568. > > Just to check: "is initialized to 150 MHz" reads as if this already > happens. My understanding is that probe() contains the ACLK_BUS setup, > but nothing forces the CRU to be probed in SPL, so it is silently > skipped. Please spell that out - something like "the clock driver is > currently only probed on demand, so the ACLK_BUS init in probe() is > skipped when no SPL consumer references it" - so the motivation > matches the speedup Alexey reported on the rk3576 patch.
Correct, I will reword the commit message. Unless the clock driver is probed in SPL, the ACLK_BUS never gets initialized to 150 MHz. > >> >> Ensure clock driver is always probed after bind for SPL builds to >> initialize core PLLs and clocks in SPL phase on RK3568. >> >> Signed-off-by: Jonas Karlman <[email protected]> >> >> drivers/clk/rockchip/clk_rk3568.c | 4 ++++ >> 1 file changed, 4 insertions(+) > >> diff --git a/drivers/clk/rockchip/clk_rk3568.c >> b/drivers/clk/rockchip/clk_rk3568.c >> @@ -2938,6 +2938,10 @@ static int rk3568_clk_bind(struct udevice *dev) >> struct udevice *sys_child; >> struct sysreset_reg *priv; >> >> + /* Ensure clocks are initialized after bind in SPL build */ >> + if (IS_ENABLED(CONFIG_SPL_BUILD)) >> + dev_or_flags(dev, DM_FLAG_PROBE_AFTER_BIND); >> + > > Have you considered marking the CRU node bootph-all (or equivalent) in > the DT instead? That is the more usual DM way to force a device up > early, and it avoids sprinkling a phase check into every SoC driver. > If PROBE_AFTER_BIND is genuinely the right hammer here, please add a > note in the commit message explaining why. Use of bootph-all does not automatically triggers probe of the device. The main issue is the on demand driver probing, and depending on what boot source is used the clock driver may never be probed in SPL, e.g. when booting from RAM/USB. We should possible move out some of these one-time clock initialization into e.g. arch_cpu_init() or similar in future to avoid having to use probe after bind. Regards, Jonas > > Regards, > Simon
