linux-kernel - Re: [PATCH 2/2] hwmon: ina3221: Get channel names from DT node

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20180921012038.GA23354@Asurada-Nvidia.nvidia.com>
Date:   Thu, 20 Sep 2018 18:20:39 -0700
From:   Nicolin Chen <nicoleotsuka@...il.com>
To:     Guenter Roeck <linux@...ck-us.net>
Cc:     jdelvare@...e.com, robh+dt@...nel.org, mark.rutland@....com,
        afd@...com, linux-hwmon@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/2] hwmon: ina3221: Get channel names from DT node

Hi Guenter,

Thanks for the review.

On Thu, Sep 20, 2018 at 05:41:48PM -0700, Guenter Roeck wrote:

> > Meanwhile, channels could be left unconnected based on
> > the hardware design. So the channel name should support
> > NC so the driver could disable the channel accordingly.
> > 
> 
> I am not in favor of such indirect settings. If a channel is
> to be disconnected, define a property for it.

OK. I can add a bool property for it instead.

> I am personally also not in favor of using devicetree to define
> channel names like this, much less so for a single driver.

As DT is used to describe hardware, I felt plausible to put
them in since the names are mentioned in the schematics. Do
you have any advise to handle this better?

> > +	/* Fetch hardware information from Device Tree */
> > +	for (i = 0, g = 0; i < INA3221_NUM_CHANNELS; i++) {
> > +		/* Fetch the channel name */
> > +		sprintf(prop, "ti,channel%d-name", i + 1);
> > +		/* Set a default name on failure */
> > +		if (of_property_read_string(np, prop, &str))
> > +			str = "unknown";
> 
> So if there are no devicetree entries, the user now gets a sequence of
> "unknown" sensors ? I don't think so. Please keep in mind that there are
> users of this chip who don't have devicetree systems, and other users
> may not want to specify any specific name properties (or they use sensors3.conf
> to do it).

Being enlightened by your comments below, maybe adding a
separate group for channel names by attaching is_visible
to it could be acceptable? Then, name nodes can hide from
those who don't want to specify.

> > +		/* Ignore unconnected channels */
> > +		if (!strcmp(str, INA3221_NOT_CONNECTED))
> > +			continue;
> 
> Sorry, I won't accept this. I am sure we can come up with some useful means
> to define in devicetree if individual channels of a hardware monitoring chip
> are enabled or not, but a channel name of "NC" won't be it.

I will try the bool property as you mentioned earlier.

> > +		/* Log connected channels */
> > +		ina->attr_group[g++] = &ina3221_group[i];
> > +		ina->channel_name[i] = str;
> > +		ina->enable[i] = true;
> 
> I also don't see the need for defining the group dynamically. The
> is_visible callback is very well suited for handling optional sysfs
> attributes.

I will add an is_visible callback.

Thanks
Nicolin