[RFC v1 7/7] usb: onboard-hub: Add conditional compilation for I2C initialization

[Michal Simek]

On 5/21/25 17:18, Quentin Schulz wrote:
> Hi Michal,
> 
> On 5/21/25 4:34 PM, Michal Simek wrote:
>>
>>
>> On 5/12/25 09:06, Quentin Schulz wrote:
>>> Hi Anand,
>>>
>>> On 5/9/25 7:45 PM, Anand Moon wrote:
>>>> Hi Quentin,
>>>>
>>>> On Fri, 9 May 2025 at 17:21, Quentin Schulz <quentin.schulz at cherry.de> wrote:
>>>>>
>>>>> Hi Neil, Anand,
>>>>>
>>>>> On 5/9/25 9:57 AM, Neil Armstrong wrote:
>>>>>> On 09/05/2025 09:02, Anand Moon wrote:
>>>>>>> Add conditional compilation for the usb5744_i2c_init() function based
>>>>>>> on the CONFIG_DM_I2C configuration, to avoid compilation failure.
>>>>>>>
>>>>>>>     CC      net/net-common.o
>>>>>>>     AR      net/built-in.o
>>>>>>>     LDS     u-boot.lds
>>>>>>>     LD      u-boot
>>>>>>> aarch64-linux-gnu-ld.bfd: common/usb_onboard_hub.o: in function
>>>>>>> `usb5744_i2c_init':
>>>>>>> /home/amoon/mainline/u-boot/amlogic/u-boot-maineline/common/
>>>>>>> usb_onboard_hub.c:74:(.text.usb5744_i2c_init+0xfc): undefined
>>>>>>> reference to `i2c_get_chip'
>>>>>>> aarch64-linux-gnu-ld.bfd: /home/amoon/mainline/u-boot/amlogic/u-boot-
>>>>>>> maineline/common/usb_onboard_hub.c:89:(.text.usb5744_i2c_init+0x1a0):
>>>>>>> undefined reference to `dm_i2c_write'
>>>>>>> aarch64-linux-gnu-ld.bfd: /home/amoon/mainline/u-boot/amlogic/u-boot-
>>>>>>> maineline/common/usb_onboard_hub.c:96:(.text.usb5744_i2c_init+0x1c8):
>>>>>>> undefined reference to `dm_i2c_write'
>>>>>>> aarch64-linux-gnu-ld.bfd: /home/amoon/mainline/u-boot/amlogic/u-boot-
>>>>>>> maineline/common/usb_onboard_hub.c:104: (.text.usb5744_i2c_init+0x1f0):
>>>>>>> undefined reference to `dm_i2c_write'
>>>>>>> Segmentation fault (core dumped)
>>>>>>> make: *** [Makefile:1824: u-boot] Error 139
>>>>>>> make: *** Deleting file 'u-boot'
>>>>>>>
>>>>>>> Signed-off-by: Anand Moon <linux.amoon at gmail.com>
>>>>>>> ---
>>>>>>>    common/usb_onboard_hub.c | 4 ++++
>>>>>>>    1 file changed, 4 insertions(+)
>>>>>>>
>>>>>>> diff --git a/common/usb_onboard_hub.c b/common/usb_onboard_hub.c
>>>>>>> index 1242513c631..92e3f3a92c9 100644
>>>>>>> --- a/common/usb_onboard_hub.c
>>>>>>> +++ b/common/usb_onboard_hub.c
>>>>>>> @@ -30,6 +30,7 @@ struct onboard_hub_data {
>>>>>>>        int (*init)(struct udevice *dev);
>>>>>>>    };
>>>>>>> +#if CONFIG_IS_ENABLED(DM_I2C)
>>>>>>>    static int usb5744_i2c_init(struct udevice *dev)
>>>>>>>    {
>>>>>>>        /*
>>>>>>> @@ -109,6 +110,7 @@ static int usb5744_i2c_init(struct udevice *dev)
>>>>>>>        return 0;
>>>>>>>    }
>>>>>>> +#endif
>>>>>>>    int usb_onboard_hub_reset(struct udevice *dev)
>>>>>>>    {
>>>>>>> @@ -222,7 +224,9 @@ static const struct onboard_hub_data usb2514_data = {
>>>>>>>    };
>>>>>>>    static const struct onboard_hub_data usb5744_data = {
>>>>>>> +#if CONFIG_IS_ENABLED(DM_I2C)
>>>>>>>        .init = usb5744_i2c_init,
>>>>>>> +#endif
>>>>>>>        .power_on_delay_us = 1000,
>>>>>>>        .reset_us = 5,
>>>>>>>    };
>>>>>>
>>>>>> I think you should completely disable usb5744_data in this case
>>>>>>
>>>>>
>>>>> I would recommend to go the same route as the Linux kernel which has a
>>>>> special Kconfig option for it, c.f. USB_ONBOARD_DEV_USB5744 which adds
>>>>> the necessary dependencies.
>>>>>
>>>> On Linux kernel, it needs to parse the device node with i2c-bus.
>>>>
>>>
>>> Seems like it does in U-Boot too? c.f. usb5744_i2c_init() with 
>>> dev_read_phandle_with_args with i2c-bus?
>>>
>>>>> We probably should be careful and figure out which boards are actually
>>>>> using the usb5744 and add this symbol to their defconfig so they don't
>>>>> regress.
>>>>>
>>>>  From my initial analysis configs/xilinx_zynqmp_kria_defconfig
>>>> which use CONFIG_USB_ONBOARD_HUB
>>>> I believe using DM_I2C to guard the code is the best approach.
>>>> it should be around .init() callback to guard this to work.
>>>>
>>>
>>> I disagree. The driver clearly supports the USB5744 and the driver is 
>>> compiled regardless of DM_I2C support which is required (as far as I 
>>> understood, maybe that's incorrect?) for being able to use USB5744? I don't 
>>> like hidden dependencies.
>>>
>>> Also, I believe Neil's point is we shouldn't try to probe the USB5744 if 
>>> there is no DM_I2C support, simply excluding .init() callback will probably 
>>> just render this "support" of USB5744 non-functional?
>>
>> USB5744 can be configured differently. We are using two configurations.
>> 1. no initialization over i2c
>> (but still you can have a need to have driver for reset or power)
> 
> Fair enough, some hubs are configured "properly" by default so don't necessarily 
> need an interface. Such is the case for the Cypress HX3 we use on RK3399 Puma 
> for example. The default is enough so we don't need the i2c interface even 
> though it's exposed by the IC. Thanks for the reminder.
> 
>> 2. initialization over i2c to start the hub
>>
>> And there is also one more configuration via SPI.
>>
>> It means I don't think DM_I2C should be real dependency on getting usb5744 to 
>> operate.
>>
> 
> Would diverge from how it's handled in the kernel (today) though.
> 
> Can we please follow the same logic as in the kernel though? Meaning we should 
> report some message if the i2c-bus property is present but DM_I2C is not 
> enabled, in which case it's likely a misconfiguration of the bootloader.

Make sense.

M