Re: [PATCH V8 net-next 05/11] net: hibmcge: Implement some .ndo functions

[Jijie Shao]

on 2024/9/9 22:48, Andrew Lunn wrote:
> > No, HBG_NIC_STATE_OPEN is not intended to ensure that hbg_net_open() and
> > hbg_net_stop() are mutually exclusive.
> >
> > Actually, when the driver do reset or self-test(ethtool -t or ethtool --reset or FLR).
> > We hope that no other data is transmitted or received at this time.
> That is an invalid assumption. You could be receiving line rate
> broadcast traffic for example, because there is a broadcast storm
> happening.
>
> I assume for testing, you are configuring a loopback somewhere? PHY
> loopback or PCS loopback? I've seen some PHYs do 'broken' loopback
> where egress traffic is looped back, but ingress traffic is also still
> received. Is this true for your hardware? Is this why you make this
> assumption?
>
> What is your use case for ethtool --reset? Are you working around
> hardware bugs? Why not simply return -EBUSY if the user tries to use
> --reset when the interface is admin up. You then know the interface is
> down, you don't need open/close to be re-entrant safe.
>
> Same for testing. Return -ENETDOWN if the user tries to do self test
> on an interface which is admin down.
>
> 	Andrew

Good idea. I'll remove priv->state first in v9. We'll discuss this 
internally. Thank you very much. Jiji Shao