]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
netfilter: x_tables: Use correct memory barriers.
authorMark Tomlinson <mark.tomlinson@alliedtelesis.co.nz>
Mon, 8 Mar 2021 01:24:13 +0000 (14:24 +1300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 30 Mar 2021 12:35:28 +0000 (14:35 +0200)
commit19a5fb4ceada903e692de96b8aa8494179abbf0b
tree4753e1f0353d63500eac75551cbe66245776ff59
parentc46cd29b89daf35f276faf62c69e8f90cfffecd0
netfilter: x_tables: Use correct memory barriers.

[ Upstream commit 175e476b8cdf2a4de7432583b49c871345e4f8a1 ]

When a new table value was assigned, it was followed by a write memory
barrier. This ensured that all writes before this point would complete
before any writes after this point. However, to determine whether the
rules are unused, the sequence counter is read. To ensure that all
writes have been done before these reads, a full memory barrier is
needed, not just a write memory barrier. The same argument applies when
incrementing the counter, before the rules are read.

Changing to using smp_mb() instead of smp_wmb() fixes the kernel panic
reported in cc00bcaa5899 (which is still present), while still
maintaining the same speed of replacing tables.

The smb_mb() barriers potentially slow the packet path, however testing
has shown no measurable change in performance on a 4-core MIPS64
platform.

Fixes: 7f5c6d4f665b ("netfilter: get rid of atomic ops in fast path")
Signed-off-by: Mark Tomlinson <mark.tomlinson@alliedtelesis.co.nz>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
include/linux/netfilter/x_tables.h
net/netfilter/x_tables.c