Re: [PATCH v4 4/4] nvmem: eeprom: at25: export FRAM serial num

[Greg Kroah-Hartman]

On Thu, May 20, 2021 at 08:56:39AM +0200, Jiří Prchal wrote:
> Here I'm completlly lost:
> 
> On 20. 05. 21 7:52, Greg Kroah-Hartman wrote:
> > On Thu, May 20, 2021 at 07:47:14AM +0200, Jiri Prchal wrote:
> > > This exports serial number of FRAM in sysfs file named "sernum".
> > > Formatted in hex, each byte separated by space.
> > > Example:
> > > $ cat /sys/class/spi_master/spi0/spi0.0/sernum
> > 
> > No new Documentation/ABI/ entry for this?
> No, should I do and how / where?

All sysfs files need a Documentation/ABI/ entry.

> > > +static ssize_t sernum_show(struct device *dev, struct device_attribute *attr, char *buf)
> > > +{
> > > +	struct at25_data *at25;
> > > +	int i;
> > > +
> > > +	at25 = dev_get_drvdata(dev);
> > > +	for (i = 0; i < FM25_SN_LEN; i++)
> > > +		buf += sprintf(buf, "%02x ", at25->sernum[i]);
> > > +	sprintf(--buf, "\n");
> > > +	return (3 * i);
> > 
> > No, that is not how sysfs files work, sorry.  They are "one value per
> > file".  This looks like multiple values in the same file, why not just
> > one file per "sernum"?
> It's formatted by spaces. It's one long number like MAC addr, so is better
> to expose it as hex string without spaces? Or like MAC separated by colon?

Why format it at all?  Just dump out the whole thing as a string.

But who is going to care about this?  What tool will be reading it and
what will it be for?  The Documentation/ABI/ entry would help explain
all of this.

> > > +	/* Export the FM25 serial number */
> > > +	if (at25->has_sernum) {
> > > +		err = device_create_file(&spi->dev, &dev_attr_sernum);
> > 
> > You just raced with userspace and lost :(
> ?
> > 
> > Please do this correctly, by setting the driver group if you need a file
> > like this.
> Any example, please?

Loads, see any platform driver that sets the dev_groups pointer.

thanks,

greg k-h