On Mon, 2018-02-12 at 14:15 -0800, Eric Dumazet wrote:
> On Mon, 2018-02-12 at 13:58 -0800, Yonghong Song wrote:
> > There is a memory leak happening in lpm_trie map_free callback
> > function trie_free. The trie structure itself does not get freed.
> >
> > Also, trie_free function did not do synchronize_rcu before freeing
> > various data structures. This is incorrect as some rcu_read_lock
> > region(s) for lookup, update, delete or get_next_key may not complete yet.
> > The fix is to add synchronize_rcu in the beginning of trie_free.
> > The useless spin_lock is removed from this function as well.
> >
> > Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map
> > implementation")
> > Reported-by: Mathieu Malaterre <ma...@debian.org>
> > Reported-by: Alexei Starovoitov <a...@kernel.org>
> > Tested-by: Mathieu Malaterre <ma...@debian.org>
> > Signed-off-by: Yonghong Song <y...@fb.com>
> > ---
> > kernel/bpf/lpm_trie.c | 9 +++++++--
> > 1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/bpf/lpm_trie.c b/kernel/bpf/lpm_trie.c
> > index 7b469d1..9b41ea4 100644
> > --- a/kernel/bpf/lpm_trie.c
> > +++ b/kernel/bpf/lpm_trie.c
> > @@ -555,7 +555,12 @@ static void trie_free(struct bpf_map *map)
> > struct lpm_trie_node __rcu **slot;
> > struct lpm_trie_node *node;
> >
> > - raw_spin_lock(&trie->lock);
> > + /* at this point bpf_prog->aux->refcnt == 0 and this map->refcnt == 0,
> > + * so the programs (can be more than one that used this map) were
> > + * disconnected from events. Wait for outstanding programs to complete
> > + * update/lookup/delete/get_next_key and free the trie.
> > + */
> > + synchronize_rcu();
> >
>
> Please do not do that.
>
> Use kfree_rcu() instead (adding one struct rcu_head in struct lpm_trie)
Oh well, I take this back. It looks we heavily use synchronize_rcu()
all over the places for ->map_free() already.
