From ba6b8de423f8d0dee48d6030288ed81c03ddf9f0 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Mon, 23 Apr 2018 15:39:23 -0700 Subject: bpf: sockmap, map_release does not hold refcnt for pinned maps Relying on map_release hook to decrement the reference counts when a map is removed only works if the map is not being pinned. In the pinned case the ref is decremented immediately and the BPF programs released. After this BPF programs may not be in-use which is not what the user would expect. This patch moves the release logic into bpf_map_put_uref() and brings sockmap in-line with how a similar case is handled in prog array maps. Fixes: 3d9e952697de ("bpf: sockmap, fix leaking maps with attached but not detached progs") Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann --- kernel/bpf/arraymap.c | 3 ++- kernel/bpf/sockmap.c | 4 ++-- kernel/bpf/syscall.c | 4 ++-- 3 files changed, 6 insertions(+), 5 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/arraymap.c b/kernel/bpf/arraymap.c index 14750e7c5ee4..027107f4be53 100644 --- a/kernel/bpf/arraymap.c +++ b/kernel/bpf/arraymap.c @@ -476,7 +476,7 @@ static u32 prog_fd_array_sys_lookup_elem(void *ptr) } /* decrement refcnt of all bpf_progs that are stored in this map */ -void bpf_fd_array_map_clear(struct bpf_map *map) +static void bpf_fd_array_map_clear(struct bpf_map *map) { struct bpf_array *array = container_of(map, struct bpf_array, map); int i; @@ -495,6 +495,7 @@ const struct bpf_map_ops prog_array_map_ops = { .map_fd_get_ptr = prog_fd_array_get_ptr, .map_fd_put_ptr = prog_fd_array_put_ptr, .map_fd_sys_lookup_elem = prog_fd_array_sys_lookup_elem, + .map_release_uref = bpf_fd_array_map_clear, }; static struct bpf_event_entry *bpf_event_entry_gen(struct file *perf_file, diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index a3b21385e947..a73d484b6e4c 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -1831,7 +1831,7 @@ static int sock_map_update_elem(struct bpf_map *map, return err; } -static void sock_map_release(struct bpf_map *map, struct file *map_file) +static void sock_map_release(struct bpf_map *map) { struct bpf_stab *stab = container_of(map, struct bpf_stab, map); struct bpf_prog *orig; @@ -1855,7 +1855,7 @@ const struct bpf_map_ops sock_map_ops = { .map_get_next_key = sock_map_get_next_key, .map_update_elem = sock_map_update_elem, .map_delete_elem = sock_map_delete_elem, - .map_release = sock_map_release, + .map_release_uref = sock_map_release, }; BPF_CALL_4(bpf_sock_map_update, struct bpf_sock_ops_kern *, bpf_sock, diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c index 4ca46df19c9a..ebfe9f29dae8 100644 --- a/kernel/bpf/syscall.c +++ b/kernel/bpf/syscall.c @@ -257,8 +257,8 @@ static void bpf_map_free_deferred(struct work_struct *work) static void bpf_map_put_uref(struct bpf_map *map) { if (atomic_dec_and_test(&map->usercnt)) { - if (map->map_type == BPF_MAP_TYPE_PROG_ARRAY) - bpf_fd_array_map_clear(map); + if (map->ops->map_release_uref) + map->ops->map_release_uref(map); } } -- cgit v1.2.3 From e20f7334837ae47341d8ec4e3170d0b4336a3676 Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Mon, 23 Apr 2018 15:39:28 -0700 Subject: bpf: sockmap, sk_wait_event needed to handle blocking cases In the recvmsg handler we need to add a wait event to support the blocking use cases. Without this we return zero and may confuse user applications. In the wait event any data received on the sk either via sk_receive_queue or the psock ingress list will wake up the sock. Fixes: fa246693a111 ("bpf: sockmap, BPF_F_INGRESS flag for BPF_SK_SKB_STREAM_VERDICT") Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 44 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 44 insertions(+) (limited to 'kernel') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index a73d484b6e4c..aaf50ec77c94 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -43,6 +43,7 @@ #include #include #include +#include #define SOCK_CREATE_FLAG_MASK \ (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY) @@ -732,6 +733,26 @@ out_err: return err; } +static int bpf_wait_data(struct sock *sk, + struct smap_psock *psk, int flags, + long timeo, int *err) +{ + int rc; + + DEFINE_WAIT_FUNC(wait, woken_wake_function); + + add_wait_queue(sk_sleep(sk), &wait); + sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); + rc = sk_wait_event(sk, &timeo, + !list_empty(&psk->ingress) || + !skb_queue_empty(&sk->sk_receive_queue), + &wait); + sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); + remove_wait_queue(sk_sleep(sk), &wait); + + return rc; +} + static int bpf_tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock, int flags, int *addr_len) { @@ -755,6 +776,7 @@ static int bpf_tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, return tcp_recvmsg(sk, msg, len, nonblock, flags, addr_len); lock_sock(sk); +bytes_ready: while (copied != len) { struct scatterlist *sg; struct sk_msg_buff *md; @@ -809,6 +831,28 @@ static int bpf_tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, } } + if (!copied) { + long timeo; + int data; + int err = 0; + + timeo = sock_rcvtimeo(sk, nonblock); + data = bpf_wait_data(sk, psock, flags, timeo, &err); + + if (data) { + if (!skb_queue_empty(&sk->sk_receive_queue)) { + release_sock(sk); + smap_release_sock(psock, sk); + copied = tcp_recvmsg(sk, msg, len, nonblock, flags, addr_len); + return copied; + } + goto bytes_ready; + } + + if (err) + copied = err; + } + release_sock(sk); smap_release_sock(psock, sk); return copied; -- cgit v1.2.3 From 4fcfdfb83391c74e62683469289db42a143440ac Mon Sep 17 00:00:00 2001 From: John Fastabend Date: Mon, 23 Apr 2018 15:39:33 -0700 Subject: bpf: sockmap, fix double page_put on ENOMEM error in redirect path In the case where the socket memory boundary is hit the redirect path returns an ENOMEM error. However, before checking for this condition the redirect scatterlist buffer is setup with a valid page and length. This is never unwound so when the buffers are released latter in the error path we do a put_page() and clear the scatterlist fields. But, because the initial error happens before completing the scatterlist buffer we end up with both the original buffer and the redirect buffer pointing to the same page resulting in duplicate put_page() calls. To fix this simply move the initial configuration of the redirect scatterlist buffer below the sock memory check. Found this while running TCP_STREAM test with netperf using Cilium. Fixes: fa246693a111 ("bpf: sockmap, BPF_F_INGRESS flag for BPF_SK_SKB_STREAM_VERDICT") Signed-off-by: John Fastabend Signed-off-by: Daniel Borkmann --- kernel/bpf/sockmap.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'kernel') diff --git a/kernel/bpf/sockmap.c b/kernel/bpf/sockmap.c index aaf50ec77c94..634415c7fbcd 100644 --- a/kernel/bpf/sockmap.c +++ b/kernel/bpf/sockmap.c @@ -524,8 +524,6 @@ static int bpf_tcp_ingress(struct sock *sk, int apply_bytes, i = md->sg_start; do { - r->sg_data[i] = md->sg_data[i]; - size = (apply && apply_bytes < md->sg_data[i].length) ? apply_bytes : md->sg_data[i].length; @@ -536,6 +534,7 @@ static int bpf_tcp_ingress(struct sock *sk, int apply_bytes, } sk_mem_charge(sk, size); + r->sg_data[i] = md->sg_data[i]; r->sg_data[i].length = size; md->sg_data[i].length -= size; md->sg_data[i].offset += size; -- cgit v1.2.3