On Wed, Jun 07, 2023 at 09:38:10AM -0700, Jakub Kicinski wrote:

On Wed, 7 Jun 2023 12:29:36 +0200 Piotr Gardocki wrote:

quoted

I need a piece of advice though:
1) Should I fix it in this patch set, or treat it as a separate thread?

Separate is probably better, you can post such a change directly 
to netdev, without going via the Intel tree.

That's what we like:D

quoted

2) I suppose the change is required only in dev_set_mac_address function, but
am I right assuming we should do it before call to dev_pre_changeaddr_notify
and return from function early? What about call to add_device_randomness?

I'd add the check right after the netif_device_present() check and not
worry about notifier or randomness. The address isn't changing so
nothing to notify about and no real randomness to be gained.

I find this as positive side effect - why would i want to notify anyone
that my addr has changed if it was not in fact changed? This is happening
in the current approach where you exit with success code from
ndo_set_mac_address().