[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <2474f02c-c835-4691-b531-62c8c747aede@linux.ibm.com>
Date: Tue, 16 Apr 2024 13:09:04 -0500
From: Eddie James <eajames@...ux.ibm.com>
To: Andi Shyti <andi.shyti@...nel.org>
Cc: linux-fsi@...ts.ozlabs.org, linux-kernel@...r.kernel.org,
linux-i2c@...r.kernel.org, devicetree@...r.kernel.org,
alistair@...ple.id.au, joel@....id.au, jk@...abs.org, robh@...nel.org,
krzysztof.kozlowski+dt@...aro.org, conor+dt@...nel.org,
lakshmiy@...ibmcom
Subject: Re: [PATCH v2 25/31] i2c: fsi: Calculate clock divider from local bus
frequency
On 4/15/24 17:11, Andi Shyti wrote:
> Hi Eddie,
>
>> @@ -689,6 +692,20 @@ static int fsi_i2c_probe(struct device *dev)
>> mutex_init(&i2c->lock);
>> i2c->fsi = to_fsi_dev(dev);
>> INIT_LIST_HEAD(&i2c->ports);
>> + i2c->clock_div = I2C_DEFAULT_CLK_DIV;
>> +
>> + lbus = fsi_device_local_bus_frequency(i2c->fsi);
>> + if (lbus) {
>> + u32 clock = I2C_DEFAULT_CLK_RATE;
> I don't see the need for initialization.
Does device_property_read_u32 set clock if the property isn't found? If
not, it needs to be initialized here. Or I can set it in an else
statement from device_property_read_u32.
>
>> +
>> + if (!device_property_read_u32(dev, "clock-frequency", &clock)) {
>> + if (!clock)
>> + clock = I2C_DEFAULT_CLK_RATE;
>> + }
> no need for brackets.
Perhaps not, but checkpatch didn't complain, and I personally like
brackets if multiple lines are included.
>
>> +
>> + // i2c clock rate = local bus clock / (4 * (i2c clock div + 1))
> You forgot to remove this.
I actually meant to leave that comment to explain how the clock rate is
calculated, as the reverse calculation in the code is a little more
confusing.
>
> Andi
>
>> + i2c->clock_div = (((lbus + (clock - 1)) / clock) / 4) - 1;
>> + }
>>
>> rc = fsi_i2c_dev_init(i2c);
>> if (rc)
>> --
>> 2.39.3
>>
Powered by blists - more mailing lists