Re: [PATCH net-next v3 2/3] net: phy: air_en8811h: add Airoha AN8811HB support

[Bjørn Mork]

Jakub Kicinski <kuba@xxxxxxxxxx> writes:
> On Mon, 26 Jan 2026 07:57:48 +0100 Bjørn Mork wrote:

>> +		mdelay(300);
>
> Did you mean msleep()? mdelay(300) is a lot of spinning.

Doh!  Thanks.  I will write that 300 times on the chalkboard.

>
>> +		air_buckpbus_reg_read(phydev, mon2, &pbus_value);
>> +
>> +		if (pbus_value & AN8811HB_CRC_ST) {
>> +			air_buckpbus_reg_read(phydev, mon3, &pbus_value);
>> +			phydev_dbg(phydev, "CRC Check %s!\n",
>> +				   pbus_value & AN8811HB_CRC_CHECK_PASS ?
>> +					"PASS" : "FAIL");
>
> AI code review points out on failure you just print a FAIL and carry on.
> Is this because this is what the vendor driver does? Or we know bad CRC
> FWs exist in the wild? A comment would be useful here..

Spot on. Explaining this in a comment for you, me and the AI is a good
idea.

> Please name labels after what they jump to. Per CodingStyle..

will do if there are any labels left after refactoring.

>> +	ret = air_buckpbus_reg_write(phydev, EN8811H_FW_CTRL_1,
>> +				     EN8811H_FW_CTRL_1_START);
>> +	if (ret < 0)
>> +		goto an8811hb_load_firmware_out;
>> +
>> +	ret = air_write_buf(phydev, AIR_FW_ADDR_DM,  fw1);
>> +	if (ret < 0)
>> +		goto an8811hb_load_firmware_out;
>> +
>> +	ret = an8811hb_check_crc(phydev, AN8811HB_CRC_DM_SET1,
>> +				 AN8811HB_CRC_DM_MON2,
>> +				 AN8811HB_CRC_DM_MON3);
>> +	if (ret < 0)
>> +		goto an8811hb_load_firmware_out;
>> +
>> +	ret = air_write_buf(phydev, AIR_FW_ADDR_DSP, fw2);
>> +	if (ret < 0)
>> +		goto an8811hb_load_firmware_out;
>> +
>> +	ret = an8811hb_check_crc(phydev, AN8811HB_CRC_PM_SET1,
>> +				 AN8811HB_CRC_PM_MON2,
>> +				 AN8811HB_CRC_PM_MON3);
>> +	if (ret < 0)
>> +		goto an8811hb_load_firmware_out;
>> +
>> +	ret = en8811h_wait_mcu_ready(phydev);
>> +	if (ret < 0)
>> +		goto an8811hb_load_firmware_out;
>
> TBH the gotos are a bit hard to read, maybe factor out the logic that
> can fail to a separate function? Then we can just capture ret here and
> fall thru; and the helper itself can return ret; directly.

Yes, this was almost unreadable.

But fixing it turned out harder than I thought. There are so many steps
and we want to skip everything after the first failure.  Could be just
me being daft.  Probably is

Anyway, I have made an attempt.  At least got rid of the gotos.  Will
send a new version after some testing.


Bjørn