> Any specific reason you're not reusing sock_setbindtodevice or at least > sock_setbindtodevice_locked here? I think, historically, we've > reimplemented some of the sockopts because they were 'easy' (i.e. > were just setting a flag in the socket), this one looks more involved. Yes, there is a copy_from_user in the sock_setbindtodevice for copying the ioctl netdev name from the user which (I think) not necessary here. However sock_setbindtodevice_locked is the way to go but I was afraid to forward declare it in sock.h, change the linkage and export it in sock.c (I find that a little bit too intrusive). > I'd suggest, add an optional 'lock_sk' argument to sock_setbindtodevice, > call it with 'true' from real setsockopt, and call it with 'false' > here. Thanks for the advice. However I think I'll wait what happens with this patch: https://lore.kernel.org/netdev/20200520195509.2215098-8-hch@xxxxxx/T/#u Very strange coincidence that patch was submitted a few hours before mine (but I noticed just now) and refactor the sock_setbindtodevice in a way that will useful in my case (also define it in sock.h). > And, as Andrii pointed out, it would be nice to have a selftest > that exercises this new option. Thanks, I will implement them in the next iteration.
