[PATCH 1/2] rtc: pch-rtc: add RTC driver for Intel Series PCH

Alexandre Belloni alexandre.belloni at bootlin.com
Sun Sep 26 08:24:13 AEST 2021


On 20/08/2021 15:34:10+0300, Ivan Mikhaylov wrote:
> On Tue, Aug 15, 2021 at 01:52:42PM +0300, Paul Fertser wrote:
> > On Tue, Aug 10, 2021 at 06:44:35PM +0300, Ivan Mikhaylov wrote:
> > > +config RTC_DRV_PCH
> > > +	tristate "PCH RTC driver"
> > > +	help
> > > +	  If you say yes here you get support for the Intel Series PCH
> >
> > I'm afraid this is really lacking some specification of devices that
> > are supported. Is it really everything that Intel currently calls PCH?
> 
> Yes, from infromation that I know.
> 
> > > +static int pch_rtc_read_time(struct device *dev, struct rtc_time *tm)
> > > +{
> > > +	struct i2c_client *client = to_i2c_client(dev);
> > > +	struct pch *pch = i2c_get_clientdata(client);
> > > +	unsigned char rtc_data[NUM_TIME_REGS] = {0};
> > > +	int rc;
> > > +
> > > +	rc = regmap_bulk_read(pch->regmap, PCH_REG_SC, rtc_data, NUM_TIME_REGS);
> > > +	if (rc < 0) {
> > > +		dev_err(dev, "fail to read time reg(%d)\n", rc);
> > > +		return rc;
> > > +	}
> >
> > Citing 26.7.2.3 from C620 (Lewisburg/Purley) datasheet:
> > 
> > "The PCH SMBus slave interface only supports Byte Read operation. The
> > external SMBus master will read the RTC time bytes one after
> > another. It is software’s responsibility to check and manage the
> > possible time rollover when subsequent time bytes are read.
> > 
> > For example, assuming the RTC time is 11 hours: 59 minutes: 59
> > seconds. When the external SMBus master reads the hour as 11, then
> > proceeds to read the minute, it is possible that the rollover happens
> > between the reads and the minute is read as 0. This results in 11
> > hours: 0 minutes instead of the correct time of 12 hours: 0 minutes.
> > Unless it is certain that rollover will not occur, software is
> > required to detect the possible time rollover by reading multiple
> > times such that the read time bytes can be adjusted accordingly if
> > needed."
> > 
> > Should this be taken additional care of somehow?
> 
> 1. .use_single_read in regmap_config.
> 2. Maybe that is the right place for rollover check.
> 
> > > +static ssize_t force_off_store(struct device *dev,
> > > +			       struct device_attribute *attr,
> > > +			       const char *buf, size_t count)
> > > +{
> > > +	struct i2c_client *client = to_i2c_client(dev);
> > > +	struct pch *pch = i2c_get_clientdata(client);
> > > +	unsigned long val;
> > > +	int rc;
> > > +
> > > +	if (kstrtoul(buf, 10, &val))
> > > +		return -EINVAL;
> > > +
> > > +	if (val) {
> > > +		/* 0x02 host force off */
> > 
> > I wonder why you write "host force off" while the C620 datasheet calls
> > it "Unconditional Power Down", does your PCH manual use different
> > naming?
> 
> It just a synonym, does the same. I can change it but first it's need to
> be decided if attribute would go or not.
> 

I would prefer if this was added later on in a later patch, with the
proper ABI documentation so that the ABI can be discussed.


-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com


More information about the openbmc mailing list