ip_set_init_comment alters set->ext_size, this is racy. bitmap and list set types serialize on set->lock, but the hash versions don't do that. Switch hash to new ip_set_init_comment_slow. Alternatives would be to avoid set->ext_size altogether or use atomic_add/sub for this. Prefer simpler version. Also add lockdep annotations to document this. ip_set_ext_destroy() has the same issue, but this is harder to fix because there is at least one case where this gets called without lock where its safe to do so: set is destroyed. Signed-off-by: Florian Westphal --- include/linux/netfilter/ipset/ip_set.h | 9 +++++++++ net/netfilter/ipset/ip_set_bitmap_gen.h | 2 ++ net/netfilter/ipset/ip_set_core.c | 3 ++- net/netfilter/ipset/ip_set_hash_gen.h | 2 +- net/netfilter/ipset/ip_set_list_set.c | 4 ++++ 5 files changed, 18 insertions(+), 2 deletions(-) diff --git a/include/linux/netfilter/ipset/ip_set.h b/include/linux/netfilter/ipset/ip_set.h index b98331572ad2..50cd719bc270 100644 --- a/include/linux/netfilter/ipset/ip_set.h +++ b/include/linux/netfilter/ipset/ip_set.h @@ -502,6 +502,15 @@ ip_set_timeout_set(unsigned long *timeout, u32 value) void ip_set_init_comment(struct ip_set *set, struct ip_set_comment *comment, const struct ip_set_ext *ext); +static inline void +ip_set_init_comment_slow(struct ip_set *set, struct ip_set_comment *comment, + const struct ip_set_ext *ext) +{ + spin_lock_bh(&set->lock); + ip_set_init_comment(set, comment, ext); + spin_unlock_bh(&set->lock); +} + static inline void ip_set_init_counter(struct ip_set_counter *counter, const struct ip_set_ext *ext) diff --git a/net/netfilter/ipset/ip_set_bitmap_gen.h b/net/netfilter/ipset/ip_set_bitmap_gen.h index bb9b5bed10e1..b13cde902c17 100644 --- a/net/netfilter/ipset/ip_set_bitmap_gen.h +++ b/net/netfilter/ipset/ip_set_bitmap_gen.h @@ -135,6 +135,8 @@ mtype_add(struct ip_set *set, void *value, const struct ip_set_ext *ext, void *x = get_ext(set, map, e->id); int ret = mtype_do_add(e, map, flags, set->dsize); + lockdep_assert_held(&set->lock); + if (ret == IPSET_ADD_FAILED) { if (SET_WITH_TIMEOUT(set) && ip_set_timeout_expired(ext_timeout(x, set))) { diff --git a/net/netfilter/ipset/ip_set_core.c b/net/netfilter/ipset/ip_set_core.c index 6cfad152d7d1..3d6a78ad93f5 100644 --- a/net/netfilter/ipset/ip_set_core.c +++ b/net/netfilter/ipset/ip_set_core.c @@ -346,7 +346,8 @@ void ip_set_init_comment(struct ip_set *set, struct ip_set_comment *comment, const struct ip_set_ext *ext) { - struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, 1); + struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, + lockdep_is_held(&set->lock)); size_t len = ext->comment ? strlen(ext->comment) : 0; if (unlikely(c)) { diff --git a/net/netfilter/ipset/ip_set_hash_gen.h b/net/netfilter/ipset/ip_set_hash_gen.h index 129f06ed85f8..c3dda56d786c 100644 --- a/net/netfilter/ipset/ip_set_hash_gen.h +++ b/net/netfilter/ipset/ip_set_hash_gen.h @@ -1065,7 +1065,7 @@ mtype_add(struct ip_set *set, void *value, const struct ip_set_ext *ext, if (SET_WITH_COUNTER(set)) ip_set_init_counter(ext_counter(data, set), ext); if (SET_WITH_COMMENT(set)) - ip_set_init_comment(set, ext_comment(data, set), ext); + ip_set_init_comment_slow(set, ext_comment(data, set), ext); if (SET_WITH_SKBINFO(set)) ip_set_init_skbinfo(ext_skbinfo(data, set), ext); /* Must come last for the case when timed out entry is reused */ diff --git a/net/netfilter/ipset/ip_set_list_set.c b/net/netfilter/ipset/ip_set_list_set.c index 1cef84f15e8c..d7ddc57a4eca 100644 --- a/net/netfilter/ipset/ip_set_list_set.c +++ b/net/netfilter/ipset/ip_set_list_set.c @@ -221,6 +221,8 @@ static void list_set_init_extensions(struct ip_set *set, const struct ip_set_ext *ext, struct set_elem *e) { + lockdep_assert_held(&set->lock); + if (SET_WITH_COUNTER(set)) ip_set_init_counter(ext_counter(e, set), ext); if (SET_WITH_COMMENT(set)) @@ -241,6 +243,8 @@ list_set_uadd(struct ip_set *set, void *value, const struct ip_set_ext *ext, struct set_elem *e, *n, *prev, *next; bool flag_exist = flags & IPSET_FLAG_EXIST; + lockdep_assert_held(&set->lock); + /* Find where to add the new entry */ n = prev = next = NULL; list_for_each_entry_rcu(e, &map->members, list) { -- 2.54.0