On 9/13/19 12:13 AM, Matthias Brugger wrote: > > > On 13/09/2019 03:20, Stefan Wahren wrote: >> Am 12.09.19 um 20:52 schrieb Eric Anholt: >>> Matthias Brugger <matthias.bgg@xxxxxxxxx> writes: >>> >>>> On 13/08/2019 18:20, Stefan Wahren wrote: >>>>> The new BCM2711 supports an additional clock for the emmc2 block. >>>>> So add a new compatible and register this clock only for BCM2711. >>>>> >>>>> Signed-off-by: Stefan Wahren <wahrenst@xxxxxxx> >>>>> Reviewed-by: Matthias Brugger <mbrugger@xxxxxxxx> >>>>> Acked-by: Eric Anholt <eric@xxxxxxxxxx> >>>>> --- >>>>> drivers/clk/bcm/clk-bcm2835.c | 20 +++++++++++++++++++- >>>>> 1 file changed, 19 insertions(+), 1 deletion(-) >>>>> >>>>> diff --git a/drivers/clk/bcm/clk-bcm2835.c b/drivers/clk/bcm/clk-bcm2835.c >>>>> index 21cd952..fdf672a 100644 >>>>> --- a/drivers/clk/bcm/clk-bcm2835.c >>>>> +++ b/drivers/clk/bcm/clk-bcm2835.c >>>>> @@ -114,6 +114,8 @@ >>>>> #define CM_AVEODIV 0x1bc >>>>> #define CM_EMMCCTL 0x1c0 >>>>> #define CM_EMMCDIV 0x1c4 >>>>> +#define CM_EMMC2CTL 0x1d0 >>>>> +#define CM_EMMC2DIV 0x1d4 >>>>> >>>>> /* General bits for the CM_*CTL regs */ >>>>> # define CM_ENABLE BIT(4) >>>>> @@ -290,7 +292,8 @@ >>>>> #define BCM2835_MAX_FB_RATE 1750000000u >>>>> >>>>> #define SOC_BCM2835 BIT(0) >>>>> -#define SOC_ALL (SOC_BCM2835) >>>>> +#define SOC_BCM2711 BIT(1) >>>>> +#define SOC_ALL (SOC_BCM2835 | SOC_BCM2711) >>>>> >>>>> /* >>>>> * Names of clocks used within the driver that need to be replaced >>>>> @@ -2003,6 +2006,16 @@ static const struct bcm2835_clk_desc clk_desc_array[] = { >>>>> .frac_bits = 8, >>>>> .tcnt_mux = 39), >>>>> >>>>> + /* EMMC2 clock (only available for BCM2711) */ >>>>> + [BCM2711_CLOCK_EMMC2] = REGISTER_PER_CLK( >>>>> + SOC_BCM2711, >>>>> + .name = "emmc2", >>>>> + .ctl_reg = CM_EMMC2CTL, >>>>> + .div_reg = CM_EMMC2DIV, >>>>> + .int_bits = 4, >>>>> + .frac_bits = 8, >>>>> + .tcnt_mux = 42), >>>>> + >>>>> /* General purpose (GPIO) clocks */ >>>>> [BCM2835_CLOCK_GP0] = REGISTER_PER_CLK( >>>>> SOC_ALL, >>>>> @@ -2238,8 +2251,13 @@ static const struct cprman_plat_data cprman_bcm2835_plat_data = { >>>>> .soc = SOC_BCM2835, >>>>> }; >>>>> >>>>> +static const struct cprman_plat_data cprman_bcm2711_plat_data = { >>>>> + .soc = SOC_BCM2711, >>>>> +}; >>>>> + >>>>> static const struct of_device_id bcm2835_clk_of_match[] = { >>>>> { .compatible = "brcm,bcm2835-cprman", .data = &cprman_bcm2835_plat_data }, >>>>> + { .compatible = "brcm,bcm2711-cprman", .data = &cprman_bcm2711_plat_data }, >>>> Because the RPi4 FW uses bcm2838-cprman as compatible, we will need to add this >>>> here as well. >>> Upstream has not committed to backwards compat with Pi's firmware. That >>> makes the ABI requirement we get held to for upstream's DT absurd, but >>> that's the state of things. >> >> We also learned from past, that's not possible to keep things downstream >> compatible. As soon as a binding is not accepted, this wont work >> anymore. A lot of the downstream stuff is hacky. >> >> For example yesterday, i learned that the thermal node is broken >> (register is part of ring oscillator block). So do we really want to be >> compatible with a hack? I would say: No >> > > There is always the possibility to fix this in the FW, which in many cases will > be trivial. In many cases, the firmware can certainly be changed to support both downstream and upstream properties, as long as the layout is not fundamentally incompatible obviously. If this is just a game of compatible strings, both can be provided and it should not cause any issues as long as you are not mixing downstream and upstream drivers for the same purpose. -- Florian