diff mbox series

[bpf-next,1/4] bpf: add BPF_CGROUP_INET_SOCK_RELEASE hook

Message ID 20200626000929.217930-1-sdf@google.com
State New
Headers show
Series [bpf-next,1/4] bpf: add BPF_CGROUP_INET_SOCK_RELEASE hook | expand

Commit Message

Stanislav Fomichev June 26, 2020, 12:09 a.m. UTC 
Sometimes it's handy to know when the socket gets freed.
In particular, we'd like to try to use a smarter allocation
of ports for bpf_bind and explore the possibility of
limiting the number of SOCK_DGRAM sockets the process can have.
Adding a release pair to existing BPF_CGROUP_INET_SOCK_CREATE
can unlock both of the mentioned features.

The only questionable part here is the sock->sk check
in the inet_release. Looking at the places where we
do 'sock->sk = NULL', I don't understand how it can race
with inet_release and why the check is there (it's been
there since the initial git import). Otherwise, the
change itself is pretty simple, we add a BPF hook
to the inet_release and avoid calling it for kernel
sockets.

Signed-off-by: Stanislav Fomichev <sdf@google.com>
---
 include/linux/bpf-cgroup.h | 3 +++
 include/uapi/linux/bpf.h   | 1 +
 kernel/bpf/syscall.c       | 3 +++
 net/core/filter.c          | 1 +
 net/ipv4/af_inet.c         | 3 +++
 5 files changed, 11 insertions(+)
diff mbox series

Patch

diff --git a/include/linux/bpf-cgroup.h b/include/linux/bpf-cgroup.h
index c66c545e161a..b4fd09fe67bd 100644
--- a/include/linux/bpf-cgroup.h
+++ b/include/linux/bpf-cgroup.h
@@ -210,6 +210,9 @@  int bpf_percpu_cgroup_storage_update(struct bpf_map *map, void *key,
 #define BPF_CGROUP_RUN_PROG_INET_SOCK(sk)				       \
 	BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET_SOCK_CREATE)
 
+#define BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk)			       \
+	BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET_SOCK_RELEASE)
+
 #define BPF_CGROUP_RUN_PROG_INET4_POST_BIND(sk)				       \
 	BPF_CGROUP_RUN_SK_PROG(sk, BPF_CGROUP_INET4_POST_BIND)
 
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index c65b374a5090..d7aea1d0167a 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -226,6 +226,7 @@  enum bpf_attach_type {
 	BPF_CGROUP_INET4_GETSOCKNAME,
 	BPF_CGROUP_INET6_GETSOCKNAME,
 	BPF_XDP_DEVMAP,
+	BPF_CGROUP_INET_SOCK_RELEASE,
 	__MAX_BPF_ATTACH_TYPE
 };
 
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 4d530b1d5683..2a3d4b8f95c7 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -1994,6 +1994,7 @@  bpf_prog_load_check_attach(enum bpf_prog_type prog_type,
 	case BPF_PROG_TYPE_CGROUP_SOCK:
 		switch (expected_attach_type) {
 		case BPF_CGROUP_INET_SOCK_CREATE:
+		case BPF_CGROUP_INET_SOCK_RELEASE:
 		case BPF_CGROUP_INET4_POST_BIND:
 		case BPF_CGROUP_INET6_POST_BIND:
 			return 0;
@@ -2792,6 +2793,7 @@  attach_type_to_prog_type(enum bpf_attach_type attach_type)
 		return BPF_PROG_TYPE_CGROUP_SKB;
 		break;
 	case BPF_CGROUP_INET_SOCK_CREATE:
+	case BPF_CGROUP_INET_SOCK_RELEASE:
 	case BPF_CGROUP_INET4_POST_BIND:
 	case BPF_CGROUP_INET6_POST_BIND:
 		return BPF_PROG_TYPE_CGROUP_SOCK;
@@ -2942,6 +2944,7 @@  static int bpf_prog_query(const union bpf_attr *attr,
 	case BPF_CGROUP_INET_INGRESS:
 	case BPF_CGROUP_INET_EGRESS:
 	case BPF_CGROUP_INET_SOCK_CREATE:
+	case BPF_CGROUP_INET_SOCK_RELEASE:
 	case BPF_CGROUP_INET4_BIND:
 	case BPF_CGROUP_INET6_BIND:
 	case BPF_CGROUP_INET4_POST_BIND:
diff --git a/net/core/filter.c b/net/core/filter.c
index 209482a4eaa2..7bcac182383c 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -6855,6 +6855,7 @@  static bool __sock_filter_check_attach_type(int off,
 	case offsetof(struct bpf_sock, priority):
 		switch (attach_type) {
 		case BPF_CGROUP_INET_SOCK_CREATE:
+		case BPF_CGROUP_INET_SOCK_RELEASE:
 			goto full_access;
 		default:
 			return false;
diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c
index 02aa5cb3a4fd..965a96ea1168 100644
--- a/net/ipv4/af_inet.c
+++ b/net/ipv4/af_inet.c
@@ -411,6 +411,9 @@  int inet_release(struct socket *sock)
 	if (sk) {
 		long timeout;
 
+		if (!sk->sk_kern_sock)
+			BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE(sk);
+
 		/* Applications forget to leave groups before exiting */
 		ip_mc_drop_socket(sk);