diff options
author | John Fastabend <john.fastabend@gmail.com> | 2017-10-18 07:10:36 -0700 |
---|---|---|
committer | David S. Miller <davem@davemloft.net> | 2017-10-20 13:01:29 +0100 |
commit | 34f79502bbcfab659b8729da68b5e387f96eb4c1 (patch) | |
tree | 27084757c315fdf13c682c3d6c57922ed5f8447c /kernel | |
parent | 435bf0d3f99a164df7e8c30428cef266b91d1d3b (diff) | |
download | linux-34f79502bbcfab659b8729da68b5e387f96eb4c1.tar.gz linux-34f79502bbcfab659b8729da68b5e387f96eb4c1.tar.bz2 linux-34f79502bbcfab659b8729da68b5e387f96eb4c1.zip |
bpf: avoid preempt enable/disable in sockmap using tcp_skb_cb region
SK_SKB BPF programs are run from the socket/tcp context but early in
the stack before much of the TCP metadata is needed in tcp_skb_cb. So
we can use some unused fields to place BPF metadata needed for SK_SKB
programs when implementing the redirect function.
This allows us to drop the preempt disable logic. It does however
require an API change so sk_redirect_map() has been updated to
additionally provide ctx_ptr to skb. Note, we do however continue to
disable/enable preemption around actual BPF program running to account
for map updates.
Signed-off-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'kernel')
-rw-r--r-- | kernel/bpf/sockmap.c | 19 |
1 files changed, 9 insertions, 10 deletions
diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index c68899d5b246..beaabb21c3a3 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -39,6 +39,7 @@ #include <linux/workqueue.h> #include <linux/list.h> #include <net/strparser.h> +#include <net/tcp.h> struct bpf_stab { struct bpf_map map; @@ -101,9 +102,16 @@ static int smap_verdict_func(struct smap_psock *psock, struct sk_buff *skb) return SK_DROP; skb_orphan(skb); + /* We need to ensure that BPF metadata for maps is also cleared + * when we orphan the skb so that we don't have the possibility + * to reference a stale map. + */ + TCP_SKB_CB(skb)->bpf.map = NULL; skb->sk = psock->sock; bpf_compute_data_end(skb); + preempt_disable(); rc = (*prog->bpf_func)(skb, prog->insnsi); + preempt_enable(); skb->sk = NULL; return rc; @@ -114,17 +122,10 @@ static void smap_do_verdict(struct smap_psock *psock, struct sk_buff *skb) struct sock *sk; int rc; - /* Because we use per cpu values to feed input from sock redirect - * in BPF program to do_sk_redirect_map() call we need to ensure we - * are not preempted. RCU read lock is not sufficient in this case - * with CONFIG_PREEMPT_RCU enabled so we must be explicit here. - */ - preempt_disable(); rc = smap_verdict_func(psock, skb); switch (rc) { case SK_REDIRECT: - sk = do_sk_redirect_map(); - preempt_enable(); + sk = do_sk_redirect_map(skb); if (likely(sk)) { struct smap_psock *peer = smap_psock_sk(sk); @@ -141,8 +142,6 @@ static void smap_do_verdict(struct smap_psock *psock, struct sk_buff *skb) /* Fall through and free skb otherwise */ case SK_DROP: default: - if (rc != SK_REDIRECT) - preempt_enable(); kfree_skb(skb); } } |