On Thu, 2013-06-20 at 19:47 +0000, Nelson, Shannon wrote:
> > From: Ben Hutchings [mailto:[email protected]]
> > Sent: Wednesday, June 19, 2013 11:27 AM
> >
> > On Thu, 2013-06-13 at 20:55 -0700, Jeff Kirsher wrote:
> > > From: Jesse Brandeburg <[email protected]>
[...]
> > > +static struct i40e_stats i40e_gstrings_stats[] = {
> > > + I40E_PF_STAT("rx_bytes", stats.eth.rx_bytes),
> > > + I40E_PF_STAT("tx_bytes", stats.eth.tx_bytes),
> > > + I40E_PF_STAT("rx_errors", stats.eth.rx_errors),
> > > + I40E_PF_STAT("tx_errors", stats.eth.tx_errors),
> >
> > Duplicates.
>
> Well, not really are they duplicates - the NETDEV_STAT numbers are
> specifically for each netdev that is attached to this device. Because
> the XL710 supports VMDq, SR-IOV, and some other device multiplexing
> schemes, each netdev's NETDEV_STAT struct is the subset of the traffic
> going through its portion of the device.
>
> The PF_STAT struct holds the stats for what actually goes through the
> physical port. This is not necessarily a summation of the various
> NETDEV_STAT structs, because some connections may be talking to each
> other and not out the physical port.
>
> When ethtool -S is called on the PF's netdev, we see both the netdev
> and the ports stats printed. When called on non-PF connection we only
> see the netdev stats, none of the port stats.
Good point. And this is something I've been meaning to bring up for
discussion on netdev myself - we should have some naming convention to
distinguish per-port and per-function/per-netdev statistics.
> > [...]
> > > +#define I40E_QUEUE_STATS_LEN \
> > > + ((((struct i40e_netdev_priv *)netdev_priv(netdev))->vsi-
> > >num_queue_pairs + \
> > > + ((struct i40e_netdev_priv *)netdev_priv(netdev))->vsi-
> > >num_queue_pairs) * 2)
> >
> > netdev should be a macro parameter. Actually this could just as well be
> > a function.
>
> Sure.
> Can you tell we inherited a lot from our previous drivers?
Just a little. :-)
[...]
> > > +static int i40e_set_phys_id(struct net_device *netdev,
> > > + enum ethtool_phys_id_state state)
> > > +{
> > > + struct i40e_netdev_priv *np = netdev_priv(netdev);
> > > + struct i40e_pf *pf = np->vsi->back;
> > > + struct i40e_hw *hw = &pf->hw;
> > > + int blink_freq = 2;
> > > + static u32 led_status;
> >
> > Ugh, no.
> >
> > Since ETHTOOL_PHYSID drops the RTNL between calls to set_phys_id, it's
> > possible to invoke it on multiple devices at the same time...
>
> Yep, that should become PF netdev specific.
[...]
My point was that you have a static variable here, whose use is not
serialised.
Ben.
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:
Build for Windows Store.
http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
E1000-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel® Ethernet, visit
http://communities.intel.com/community/wired
