[PATCH batadv v5 0/6] batman-adv: drop global hard interface list

Cover letter from Sven's v4:

The global hard interface list was used in the past to provide and sysfs
(debugfs, procfs) based configuration interface. This requirement is gone
after it was switched to generic netlink and NETLINK_ROUTE. And after the
wifi-flags cache was introduced, it is also no longer used to get
non-batman-adv attached interface information (for ap_isolation and
re-broadcast configuration).

But this odd net_devices list (batadv_hardif_list) stayed and caused some
headaches:

* memory requirement increased (useless) for each network interface of the
  system
* massive increase of various ethernet operations due to the O(n) nature of
  this global list
* the code had to handle switch of a batadv_hard_iface from one mesh_iface
  to either NULL or a different mesh_iface

Just get rid of it now and start to simplify the code around it.

---

@Sven: Feel free to update the author of "make hard_iface->mesh_iface immutable"
to yourself, as I only pulled your change out into a separate patch.

---

Changes in v5:
- Rebase
- Improve commit description of patch 1
- Further simplify refcounting in enable/disable (holding only 1 ref instead of 3)
- Move immutable hard_iface->mesh_iface change to a separate change
- One new patch with a minor code cleanup
- Sven's follow-up cleanups are not included in this series


Revisions by Sven Eckelmann:

Changes in v4:
- rebase
- don't double-netdev_put in batadv_hardif_enable_interface() error path
  (just let batadv_hardif_put()/batadv_hardif_release() deal with it)
- add patches which cleanup the code after batadv_hard_iface->mesh_iface
  became immutable
- Link to v3: https://patch.msgid.link/[email protected]

Changes in v3:
- fix return kernel-doc for batadv_hardif_enable_interface
- drop merged first patch
- really switch to RFC
- in merged batadv_hardif_enable_interface (as intermediate step), only add
  the hard_iface to the batadv_hardif_list when hard_iface is really
  initialized
  + this list is dropped anyway with the patch "remove global hardif list"
- stop setting "mesh_iface" to NULL on error. it is now only important that
  the hard_iface gets unlinked from the mesh_iface
- first drop the batadv_hardif_list before removing the "safety" state
  BATADV_IF_NOT_IN_USE
- Link to v2: https://patch.msgid.link/[email protected]

Changes in v2:
- rebased
- submit as RFC to get the discussion started again
- drop already merged "batman-adv: store hard_iface as iflink private data"
- switch from kzalloc to kzalloc_obj
- update author's mail and name
- fix reference counting for batman_adv_ptype
- Link to v1: https://patch.msgid.link/0b26554afea5203820faef1dfb498af7533a9b5d.1747687504.git.mschiffer@universe-factory.net


Nora Schiffer (6):
  batman-adv: create hardif only for netdevs that are part of a mesh
  batman-adv: remove global hardif list
  batman-adv: make hard_iface->mesh_iface immutable
  batman-adv: remove BATADV_IF_NOT_IN_USE hardif state
  batman-adv: move hardif generation counter into batadv_priv
  batman-adv: drop unneeded goto and initialization from
    batadv_hardif_disable_interface()

 net/batman-adv/bat_iv_ogm.c     |   3 +-
 net/batman-adv/bat_v_elp.c      |   3 +-
 net/batman-adv/hard-interface.c | 157 +++++++++++---------------------
 net/batman-adv/hard-interface.h |  10 +-
 net/batman-adv/main.c           |   6 --
 net/batman-adv/main.h           |   3 -
 net/batman-adv/mesh-interface.c |  13 +--
 net/batman-adv/netlink.c        |   4 +-
 net/batman-adv/originator.c     |   4 -
 net/batman-adv/types.h          |   6 +-
 10 files changed, 62 insertions(+), 147 deletions(-)

-- 
2.54.0