On 22.01.2020 05:04, Florian Fainelli wrote: > > > On 1/21/2020 1:09 PM, Heiner Kallweit wrote: >> Convert suitable drivers to use new helper phy_do_ioctl_running. >> >> Signed-off-by: Heiner Kallweit <hkallweit1@xxxxxxxxx> > The vast majority of drivers that you are converting use the following > convention: > > - !netif_running -> return -EINVAL > - !dev->phydev -> return -ENODEV > > so it may make sense to change the helper to accommodate the majority > here, not that I believe this is going to make much practical > difference, but if there were test cases that were specifically looking > for such an error code, they could be failing after this changeset. > Right, I also stumbled across the different error codes, mainly as you say -EINVAL. However there is no "wrong value", if netdev isn't running, then typically the PHY is not attached, and from a netdev point of view it's not there. So ENODEV seems to be best suited. In kernel code the changed return code doesn't make a difference, but yes, in theory there could be userspace programs checking for -EINVAL. However such userspace programs should check for ENODEV too anyway to cover the second check that already returns -ENODEV. > For bgmac.c, bcmgenet.c and cpmac.c: > > Acked-by: Florian Fainelli <f.fainelli@xxxxxxxxx> > > Whether you decide to spin another version or not. > Heiner
