Re: [PATCH] hwmon: (cros-ec_hwmon) Fix access to restricted __le16

[Thomas Weißschuh]

On 2024-06-06 13:16:59+0000, Guenter Roeck wrote:
> On 6/6/24 12:53, Thomas Weißschuh wrote:
> > Thanks!
> > 
> > On 2024-06-06 11:05:07+0000, Guenter Roeck wrote:
> > > 0-day complains:
> > > 
> > > drivers-hwmon-cros_ec_hwmon.c:sparse:sparse:cast-to-restricted-__le16
> > > 
> > > Fix by using a __le16 typed variable as parameter to le16_to_cpu().
> > > 
> > > Fixes: bc3e45258096 ("hwmon: add ChromeOS EC driver")
> > > Cc: Thomas Weißschuh <linux@xxxxxxxxxxxxxx>
> > > Cc: Tzung-Bi Shih <tzungbi@xxxxxxxxxx>
> > > Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> > 
> > Acked-by: Thomas Weißschuh <linux@xxxxxxxxxxxxxx>
> > 
> > Guenter, does sparse work locally for you?
> > 
> 
> It does, but I use the version from git://repo.or.cz/smatch.git.

That does indeed look much better, thanks.


I have another question about the endianness conversion in general.
The only places I see doing a conversion are
cros_ec_sensors_cmd_read_u16() and the original cros_ec hwmon driver.

Also the documentation of the EC protocol does not specify anything in
that regard.
Instead there is the following comment in host_event_set_bit():

   /*
    * The overall host event implementation assumes it's running on and
    * communicating with little-endian architectures.
    */

Can the conversion be dropped?