[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <d64fc362be63bb8540447f7df2c232eedd696edb.camel@linux.intel.com>
Date: Tue, 17 Jul 2018 15:19:08 +0300
From: Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To: Alexandre Belloni <alexandre.belloni@...tlin.com>,
Wolfram Sang <wsa@...-dreams.de>,
Jarkko Nikula <jarkko.nikula@...ux.intel.com>,
James Hogan <jhogan@...nel.org>
Cc: Paul Burton <paul.burton@...s.com>,
Mika Westerberg <mika.westerberg@...ux.intel.com>,
linux-i2c@...r.kernel.org, devicetree@...r.kernel.org,
linux-kernel@...r.kernel.org, linux-mips@...ux-mips.org,
Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
Allan Nielsen <allan.nielsen@...rosemi.com>,
Rob Herring <robh+dt@...nel.org>
Subject: Re: [PATCH 3/5] i2c: designware: add MSCC Ocelot support
On Tue, 2018-07-17 at 13:48 +0200, Alexandre Belloni wrote:
> The Microsemi Ocelot I2C controller is a designware IP. It also has a
> second set of registers to allow tweaking SDA hold time and spike
> filtering.
Can you elaborate a bit?
Are they platform specific? Are they shadow registers? Are they
something else? Datasheet link / excerpt would be also good to read.
> Optional properties :
> + - reg : for "mscc,ocelot-i2c", a second register set to configure
> the SDA hold
> + time, named ICPU_CFG:TWI_DELAY in the datasheet.
> +
Hmm... Is this registers unique to the SoC in question? Is address of
them fixed or may be configured on RTL level?
If former is right, why do we need a separate property?
>
> +#define MSCC_ICPU_CFG_TWI_DELAY 0x0
> +#define MSCC_ICPU_CFG_TWI_DELAY_ENABLE BIT(0)
> +#define MSCC_ICPU_CFG_TWI_SPIKE_FILTER 0x4
> +
> +static int mscc_twi_set_sda_hold_time(struct dw_i2c_dev *dev)
> +{
> + writel((dev->sda_hold_time << 1) |
> MSCC_ICPU_CFG_TWI_DELAY_ENABLE,
> + dev->base_ext + MSCC_ICPU_CFG_TWI_DELAY);
> +
> + return 0;
> +}
Hmm... And does how this make native DesignWare IP's registers obsolete?
> + if (of_device_is_compatible(pdev->dev.of_node, "mscc,ocelot-
> i2c"))
Can't you just ask for this unconditionally? Why not?
(It seems I might have known why not, but can we use named resource
instead in case this is not so SoC specific)
--
Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
Intel Finland Oy
Powered by blists - more mailing lists