]> git.ipfire.org Git - thirdparty/linux.git/commit
neighbour: Free pneigh_entry after RCU grace period.
authorKuniyuki Iwashima <kuniyu@google.com>
Wed, 16 Jul 2025 22:08:12 +0000 (22:08 +0000)
committerJakub Kicinski <kuba@kernel.org>
Thu, 17 Jul 2025 23:25:20 +0000 (16:25 -0700)
commitd539d8fbd8fcf64a1492c51f5ee99aaa8a8dc9ab
treedda874417e9d21e8e3bc0358a02b31da5401c9ea
parentd63382aea70aa4ecb516126e00930bc8ab5e55ef
neighbour: Free pneigh_entry after RCU grace period.

We will convert RTM_GETNEIGH to RCU.

neigh_get() looks up pneigh_entry by pneigh_lookup() and passes
it to pneigh_fill_info().

Then, we must ensure that the entry is alive till pneigh_fill_info()
completes, but read_lock_bh(&tbl->lock) in pneigh_lookup() does not
guarantee that.

Also, we will convert all readers of tbl->phash_buckets[] to RCU.

Let's use call_rcu() to free pneigh_entry and update phash_buckets[]
and ->next by rcu_assign_pointer().

pneigh_ifdown_and_unlock() uses list_head to avoid overwriting
->next and moving RCU iterators to another list.

pndisc_destructor() (only IPv6 ndisc uses this) uses a mutex, so it
is not delayed to call_rcu(), where we cannot sleep.  This is fine
because the mcast code works with RCU and ipv6_dev_mc_dec() frees
mcast objects after RCU grace period.

While at it, we change the return type of pneigh_ifdown_and_unlock()
to void.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20250716221221.442239-8-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
include/net/neighbour.h
net/core/neighbour.c