From 7bd705045508a319acf9d9e01b5e5fc52f39201c Mon Sep 17 00:00:00 2001
From: Ben Hutchings <ben.hutchings@codethink.co.uk>
Date: Thu, 29 Mar 2018 15:12:41 +0100
Subject: [PATCH 3/3] netfilter: x_tables: Add note about how to free percpu
 counters
Git-commit: 9ba5c404bf1d6284f0269411b33394362b7ff405
Patch-mainline: v4.17-rc1
References: git-fixes

Due to the way percpu counters are allocated and freed in blocks,
it is not safe to free counters individually.  Currently all callers
do the right thing, but let's note this restriction.

Fixes: ae0ac0ed6fcf ("netfilter: x_tables: pack percpu counter allocations")
Signed-off-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Denis Kirjanov <denis.kirjanov@suse.com>
---
 net/netfilter/x_tables.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c
index 1c02e2d22824..4ec2f40d9521 100644
--- a/net/netfilter/x_tables.c
+++ b/net/netfilter/x_tables.c
@@ -1696,7 +1696,9 @@ EXPORT_SYMBOL_GPL(xt_proto_fini);
  * to fetch the real percpu counter.
  *
  * To speed up allocation and improve data locality, a 4kb block is
- * allocated.
+ * allocated.  Freeing any counter may free an entire block, so all
+ * counters allocated using the same state must be freed at the same
+ * time.
  *
  * xt_percpu_counter_alloc_state contains the base address of the
  * allocated page and the current sub-offset.
-- 
2.16.4