Thread (17 messages) flat view 17 messages, 2 authors, 2d ago

Re: [PATCH v4 net-next 00/15] neighbour: Namespacify arp_tbl and nd_tbl.

From: Nikolay Aleksandrov <razor@blackwall.org>
Date: 2026-08-13 15:32:42

On 13/08/2026 11:02, Kuniyuki Iwashima wrote:
The neighbour subsystem is almost ready to drop RTNL.

However, the control paths are serialised by the global
per-table lock.

This series converts arp_tbl and nd_tbl to per-netns table.

With the series, /proc/sys/net/ipv{4,6}/neigh/default/ can
be configured per netns, which was only configurable in
init_net.

To avoid potential regression, all the settings are inherited
from init_net by default, and this behaviour is controlled by
a new sysctl knob, net.core.neigh_inherit_init_net:

   # sysctl net.core.neigh_inherit_init_net
   net.core.neigh_inherit_init_net = 1
   # sysctl net.ipv4.neigh.default.gc_thresh1
   net.ipv4.neigh.default.gc_thresh1 = 128
   # sysctl net.ipv4.neigh.default.gc_thresh1=129
   net.ipv4.neigh.default.gc_thresh1 = 129
   # unshare -n sysctl net.ipv4.neigh.default.gc_thresh1
   net.ipv4.neigh.default.gc_thresh1 = 129

If it is turned off, all settings are reset in the new netns:

   # sysctl net.core.neigh_inherit_init_net=0
   net.core.neigh_inherit_init_net = 0
   # unshare -n sysctl net.ipv4.neigh.default.gc_thresh1
   net.ipv4.neigh.default.gc_thresh1 = 128

Series overview:

   Patch 1 deflakes test_neigh.sh.

   Patch 2 ~ 3 are misc cleanup.

   Patch 4 ~ 7 store arp_tbl/nd_tbl to net->neigh_tables[] and
   remove the global neigh_tables[].

   Patch 8 ~ 9 replace the direct access to arp_tbl/nd_tbl to
   net->neigh_tables[] using new helpers.

   Patch 10 ~ 12 finally replace the global table with per-netns
   table.

   Patch 13 ~ 14 clean up unnecessary net_eq().

   Patch 15 updates test_neigh.sh.

Note that some buggy drivers access nd_tbl without checking
disable_ipv6_mod, so nd_tbl's extern definition is still left.


Changes:
   v4:
     * Patch 12
       * Add sysctl knob, net.core.neigh_inherit_init_net
       * Inherit all neigh parms by default

   v3: https://lore.kernel.org/netdev/20260811022448.116235-1-kuniyu@google.com/ (local)
     * Add Patch 1 & 11
     * Patch 12
       * Remove timer_shutdown_sync() in neigh_flush_one() and
         rely on tbl->entries (Patch 11) to free it in neigh_table_free().
     * Patch 15
       * Remove stale comments

   v2: https://lore.kernel.org/netdev/20260807232932.3986667-1-kuniyu@google.com/ (local)
     * Add Patch 11 & 13
     * Patch 7
       * Add note about mlx5e_tc_update_neigh_used_value()
     * Patch 9
       * Add __maybe_unused to net in neigh_table_clear()
     * Patch 10
       * panic() when register_pernet_subsys(&arp_net_ops) fails
       * Add timer_shutdown_sync() in neigh_flush_one()
     * Patch 11
       * Split from the next patch
       * Remove net comparison in pneigh_dump_table()
       * Remove net arg of pneigh_create(), pneigh_delete(), and
         pneigh_lookup()

   v1: https://lore.kernel.org/netdev/20260806011152.1170012-1-kuniyu@google.com/ (local)


Kuniyuki Iwashima (15):
   selftest: net: Deflake Periodic GC test in test_neigh.sh.
   neighbour: Remove __neigh_for_each_release().
   neighbour: Remove lock dance for neigh_update_{gc,managed}_list().
   neighbour: Remove unnecessary EXPORT_SYMBOL().
   neighbour: Remove __rcu from neigh_tables[].
   neighbour: Store arp_tbl and nd_tbl in net->neigh_tables[].
   neighbour: Remove neigh_tables[].
   ipv4: Replace &arp_tbl with arp_table(net).
   ipv6: Replace &nd_tbl with nd_table(net).
   neighbour: Clean up neigh_table_init() and neigh_table_clear().
   neighbour: Convert neigh_table.entries to refcount_t.
   neighbour: Namespacify neigh_tables.
   neighbour: Don't store net in struct pneigh_entry.
   neighbour: Remove unnecessary net_eq().
   selftest: net: Specify netns for ip ntable in test_neigh.sh.

  Documentation/admin-guide/sysctl/net.rst      |  14 +
  drivers/infiniband/ulp/ipoib/ipoib_main.c     |  27 +-
  .../marvell/prestera/prestera_router.c        |  10 +-
  .../mellanox/mlx5/core/en/rep/neigh.c         |  29 +-
  .../mellanox/mlx5/core/en/tc_tun_encap.c      |  23 +-
  .../mellanox/mlx5/core/en_accel/ipsec.c       |   6 +-
  .../ethernet/mellanox/mlxsw/spectrum_router.c |  31 +-
  .../ethernet/mellanox/mlxsw/spectrum_span.c   |  10 +-
  .../netronome/nfp/flower/tunnel_conf.c        |  14 +-
  drivers/net/ethernet/rocker/rocker_main.c     |   2 +-
  drivers/net/ethernet/rocker/rocker_ofdpa.c    |   2 +-
  drivers/net/ethernet/sfc/tc_counters.c        |   8 +-
  drivers/net/ethernet/sfc/tc_encap_actions.c   |   4 +-
  drivers/net/vrf.c                             |   2 +-
  drivers/net/vxlan/vxlan_core.c                |  16 +-
  include/net/arp.h                             |  10 +-
  include/net/ndisc.h                           |  20 +-
  include/net/neighbour.h                       |  22 +-
  include/net/net_namespace.h                   |   4 +
  include/net/route.h                           |   7 +-
  net/bridge/br_arp_nd_proxy.c                  |   4 +-
  net/core/neighbour.c                          | 400 +++++++++---------
  net/core/sysctl_net_core.c                    |  12 +
  net/ieee802154/6lowpan/tx.c                   |   3 +-
  net/ipv4/arp.c                                | 131 ++++--
  net/ipv4/devinet.c                            |  18 +-
  net/ipv4/fib_semantics.c                      |   7 +-
  net/ipv4/route.c                              |   2 +-
  net/ipv6/addrconf.c                           |  17 +-
  net/ipv6/ip6_output.c                         |   4 +-
  net/ipv6/ndisc.c                              | 146 ++++---
  net/ipv6/route.c                              |  18 +-
  tools/testing/selftests/net/test_neigh.sh     |  66 +--
  33 files changed, 620 insertions(+), 469 deletions(-)
Nice work! With init_net inherit as default, for the set:
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>

Cheers,
  Nik
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help