On Thu, Jun 16, 2016 at 07:06:52PM +0100, Ben Hutchings wrote:
> On Thu, 2016-06-16 at 10:55 -0700, Benjamin Poirier wrote:
> > On 2016/06/13 11:46, Netanel Belgazal wrote:
> [...]
> > > +static ssize_t ena_show_small_copy_len(struct device *dev,
> > > + struct device_attribute *attr, char *buf)
> > > +{
> > > + struct ena_adapter *adapter = dev_get_drvdata(dev);
> > > +
> > > + return sprintf(buf, "%d\n", adapter->small_copy_len);
> > > +}
> > > +
> > > +static DEVICE_ATTR(small_copy_len, S_IRUGO | S_IWUSR,
> > > ena_show_small_copy_len,
> > > + ena_store_small_copy_len);
> >
> > This is what many other drivers call (rx_)copybreak. Perhaps it's time
> > to add it to ethtool as well?
> [...]
>
> There is the 'tunable' ethtool API for random parameters like
> rx_copybreak. The ethtool utility doesn't support it yet, though.
So I started implementing proper support for this in ethtool in a
generic way (retrieving the string set, parsing generically, etc), but
it seems that the implementation isn't quite as complete as one would
like. You said [1] that the kernel returns the type of each tunable,
but I don't see this implemented anywhere.
The string set also includes "Unspec", which is of dubious value.
I think that something tunable specific is going to need to be added
apart from ETHTOOL_GSTRINGS to provide ethtool the needed type
information.
--msw
[1] http://thread.gmane.org/gmane.linux.network/367058/focus=376701
