On Wed, Sep 24, 2025 at 11:26:58PM +0100, Andre Carvalho wrote: > Hi Breno, > > On Wed, Sep 24, 2025 at 01:26:16AM -0700, Breno Leitao wrote: > > The other option is to always populate the mac during netpoll setup and > > then always resume based on mac. This seems a more precise resume. > > > > In this case, if the device goes to DEACTIVATED, then np.dev_mac will be > > populated, and you only compare it to check if you want to resume it. > > This sounds good to me. I've done some initial testing patching > __netpoll_setup > to always set np->dev_mac, changing maybe_resume_target to simply compare the > mac as you suggested and seems like this approach works.
Thanks. You probably want to clean the dev_mac once the is disabled for such case. in other words, if user configured a target to be dev_name bound, dev_mac might be NULL once the interface got disbled. So, if user disable the interface, it should unbound from the mac. In case the user re-enable it later, it needs to bind by dev_name instead of dev_mac.
