[RFC,0/1] Move IFF_LIVE_ADDR_CHANGE to public flag

Message

James Prestwood Aug. 4, 2022, 5:43 p.m. UTC

Most of the motivation behind this is in the commit description, but
I'm sending this as an RFC since I'm not fully sure messing with these
flags is ok in terms of backwards compatibility. There is also a bit
more info thats out of scope of a commit description.

Originally I tried to achieve this same behavior but only touching
the wireless subsystem. This was easy enough but the critical piece
was exposing some userspace flag. Without this there isn't much
point in allowing a live address change from userspace since there
is no way of knowing if its possible ahead of time. The original
motivation was also time savings, not related to address randomization.

The wireless maintainers weren't keen on adding a flag in nl80211
since this functionality isn't wireless specific.

Now with the addition of 6GHz we ran into this issue with address
randomization and I'm now reviving the old patches, but taking
the comments from the wireless maintainers and putting this
userspace flag into RTNL rather than nl80211.

Here are the archives to the original thread:
https://lore.kernel.org/linux-wireless/[e-mail]/

Before I CC a ton of maintainers (since I'm touching so many drivers)
I wanted to get a general answer to whether or not this patch would
be accepted, or if we need to go another route like push back on
wireless or maybe another suggestion.

James Prestwood (1):
  net: move IFF_LIVE_ADDR_CHANGE to public flag

 drivers/infiniband/ulp/ipoib/ipoib_main.c     |  2 +-
 .../infiniband/ulp/opa_vnic/opa_vnic_netdev.c |  2 +-
 drivers/net/dummy.c                           |  4 +--
 .../net/ethernet/aquantia/atlantic/aq_nic.c   |  2 +-
 .../net/ethernet/freescale/dpaa/dpaa_eth.c    |  2 +-
 .../net/ethernet/freescale/dpaa2/dpaa2-eth.c  |  3 +-
 drivers/net/ethernet/freescale/gianfar.c      |  2 +-
 drivers/net/ethernet/marvell/mvneta.c         |  2 +-
 .../ethernet/netronome/nfp/nfp_net_common.c   |  2 +-
 .../net/ethernet/netronome/nfp/nfp_net_repr.c |  2 +-
 .../net/ethernet/pensando/ionic/ionic_lif.c   |  5 +--
 drivers/net/ethernet/realtek/r8169_main.c     |  2 +-
 drivers/net/geneve.c                          |  3 +-
 drivers/net/loopback.c                        |  4 +--
 drivers/net/netdevsim/netdev.c                |  5 ++-
 drivers/net/ntb_netdev.c                      |  2 +-
 drivers/net/team/team.c                       |  3 +-
 drivers/net/tun.c                             |  2 +-
 drivers/net/veth.c                            |  2 +-
 drivers/net/virtio_net.c                      |  4 +--
 drivers/net/vrf.c                             |  2 +-
 drivers/net/vxlan/vxlan_core.c                |  2 +-
 include/linux/netdevice.h                     | 36 +++++++++----------
 include/uapi/linux/if.h                       |  7 +++-
 net/ethernet/eth.c                            |  2 +-
 net/ipv4/ip_gre.c                             |  8 ++---
 net/ipv6/ip6_gre.c                            |  8 ++---
 net/ncsi/ncsi-rsp.c                           |  6 ++--
 net/openvswitch/vport-internal_dev.c          |  4 +--
 29 files changed, 67 insertions(+), 63 deletions(-)

Comments

Andrew Lunn Aug. 4, 2022, 8:49 p.m. UTC | #1

>   net: move IFF_LIVE_ADDR_CHANGE to public flag

If this concept is accepted, i would actually change the flag name a
little. That will help developers know the semantics of the flag have
changed when their code fails to compile. Maybe make the public flag
IFF_ADDR_CHANGE_LIVE?

	Andrew