netfilter: ipset: fix race condition in ipset save, swap and delete
authorVishwanath Pai <vpai@akamai.com>
Wed, 16 Mar 2016 20:49:00 +0000 (21:49 +0100)
committerPablo Neira Ayuso <pablo@netfilter.org>
Mon, 28 Mar 2016 15:57:45 +0000 (17:57 +0200)
commit596cf3fe5854fe2b1703b0466ed6bf9cfb83c91e
treefe02169500b0046f42b7f85e91c92b3cfb4fa342
parentd7be81a5916bdb1d904803958e5991a16f7ae4b2
netfilter: ipset: fix race condition in ipset save, swap and delete

This fix adds a new reference counter (ref_netlink) for the struct ip_set.
The other reference counter (ref) can be swapped out by ip_set_swap and we
need a separate counter to keep track of references for netlink events
like dump. Using the same ref counter for dump causes a race condition
which can be demonstrated by the following script:

ipset create hash_ip1 hash:ip family inet hashsize 1024 maxelem 500000 \
counters
ipset create hash_ip2 hash:ip family inet hashsize 300000 maxelem 500000 \
counters
ipset create hash_ip3 hash:ip family inet hashsize 1024 maxelem 500000 \
counters

ipset save &

ipset swap hash_ip3 hash_ip2
ipset destroy hash_ip3 /* will crash the machine */

Swap will exchange the values of ref so destroy will see ref = 0 instead of
ref = 1. With this fix in place swap will not succeed because ipset save
still has ref_netlink on the set (ip_set_swap doesn't swap ref_netlink).

Both delete and swap will error out if ref_netlink != 0 on the set.

Note: The changes to *_head functions is because previously we would
increment ref whenever we called these functions, we don't do that
anymore.

Reviewed-by: Joshua Hunt <johunt@akamai.com>
Signed-off-by: Vishwanath Pai <vpai@akamai.com>
Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
include/linux/netfilter/ipset/ip_set.h
net/netfilter/ipset/ip_set_bitmap_gen.h
net/netfilter/ipset/ip_set_core.c
net/netfilter/ipset/ip_set_hash_gen.h
net/netfilter/ipset/ip_set_list_set.c