Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Tobias Waldekranz <tobias@waldekranz.com>
To:     Andrew Lunn <andrew@lunn.ch>
Cc:     davem@davemloft.net, kuba@kernel.org, netdev@vger.kernel.org,
        Marcin Wojtas <mw@semihalf.com>,
        Calvin Johnson <calvin.johnson@oss.nxp.com>,
        Markus Koch <markus@notsyncing.net>,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH net-next 4/5] net/fsl: xgmac_mdio: Support setting the
 MDC frequency
In-Reply-To: <YfFHmkFXtlVus9IW@lunn.ch>
References: <20220126101432.822818-1-tobias@waldekranz.com>
 <20220126101432.822818-5-tobias@waldekranz.com> <YfFHmkFXtlVus9IW@lunn.ch>
Date:   Wed, 26 Jan 2022 14:30:28 +0100
Message-ID: <87r18ubp57.fsf@waldekranz.com>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

On Wed, Jan 26, 2022 at 14:07, Andrew Lunn <andrew@lunn.ch> wrote:
> Hi Tobias
>
>>  struct mdio_fsl_priv {
>>  	struct	tgec_mdio_controller __iomem *mdio_base;
>> +	struct clk *enet_clk;
>
> It looks like there is a whitespace issue here?
>

Ahh, sorry about that!

>> +	u32	mdc_freq;
>>  	bool	is_little_endian;
>>  	bool	has_a009885;
>>  	bool	has_a011043;
>> @@ -255,6 +258,34 @@ static int xgmac_mdio_read(struct mii_bus *bus, int phy_id, int regnum)
>>  	return ret;
>>  }
>>  
>> +static void xgmac_mdio_set_mdc_freq(struct mii_bus *bus)
>> +{
>> +	struct mdio_fsl_priv *priv = (struct mdio_fsl_priv *)bus->priv;
>> +	struct tgec_mdio_controller __iomem *regs = priv->mdio_base;
>> +	struct device *dev = bus->parent;
>> +	u32 mdio_stat, div;
>> +
>> +	if (device_property_read_u32(dev, "clock-frequency", &priv->mdc_freq))
>> +		return;
>> +
>> +	priv->enet_clk = devm_clk_get(dev, NULL);
>> +	if (IS_ERR(priv->enet_clk)) {
>> +		dev_err(dev, "Input clock unknown, not changing MDC frequency");
>
> Is the clock optional or mandatory?

As the code is now, it is mandatory. I could add some default frequency,
but I fear that could do more harm than good?

> If mandatory, then i would prefer -ENODEV and fail the probe.
>
>> +		return;
>> +	}
>> +
>> +	div = ((clk_get_rate(priv->enet_clk) / priv->mdc_freq) - 1) / 2;
>> +	if (div < 5 || div > 0x1ff) {
>> +		dev_err(dev, "Requested MDC frequecy is out of range, ignoring");
>
> and here return -EINVAL.
>
> I always think it is best to make it obvious something is broken. One
> additional line on the console can be missed for a while. All the
> Ethernet devices missing because the PHYs are missing, because the
> MDIO bus is missing is going to get noticed very quickly.

Darn, the last thing I did was to change this from a hard to a soft
error :)

Ok, no worries about regressions for deployed stuff already out there?