bpf: add BPF_CGROUP_INET_SOCK_RELEASE hook
Subsystems (9)
v3 cover bpf: add BPF_CGROUP_INET_SOCK_RELEASE hook
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.
There is already existing BPF_CGROUP_INET_SOCK_CREATE hook that triggers upon socket creation; let's add new hook (BPF_CGROUP_INET_SOCK_RELEASE) that triggers on socket release.
v3:
* s/CHECK_FAIL/CHECK/ (Andrii Nakryiko)
* s/bpf_prog_attach/bpf_program__attach_cgroup/ (Andrii Nakryiko)
* fix &in_use in BPF program (Andrii Nakryiko)
v2:
* fix compile issue with CONFIG_CGROUP_BPF=n (kernel test robot)
Stanislav Fomichev (4):
bpf: add BPF_CGROUP_INET_SOCK_RELEASE hook
libbpf: add support for BPF_CGROUP_INET_SOCK_RELEASE
bpftool: add support for BPF_CGROUP_INET_SOCK_RELEASE selftests/bpf: test BPF_CGROUP_INET_SOCK_RELEASE
include/linux/bpf-cgroup.h | 4 ++ include/uapi/linux/bpf.h | 1 + kernel/bpf/syscall.c | 3 + net/core/filter.c | 1 + net/ipv4/af_inet.c | 3 + tools/bpf/bpftool/common.c | 1 + tools/include/uapi/linux/bpf.h | 1 + tools/lib/bpf/libbpf.c | 4 ++ .../selftests/bpf/prog_tests/udp_limit.c | 72 +++++++++++++++++++ tools/testing/selftests/bpf/progs/udp_limit.c | 42 +++++++++++ 10 files changed, 132 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/udp_limit.c create mode 100644 tools/testing/selftests/bpf/progs/udp_limit.c
Signature · 1 line
2.27.0.212.ge8ba1cc988-goog
PATCH v3 1/4 bpf: add BPF_CGROUP_INET_SOCK_RELEASE hook
Implement BPF_CGROUP_INET_SOCK_RELEASE hook that triggers on inet socket release. It triggers only for userspace sockets, the same semantics as existing BPF_CGROUP_INET_SOCK_CREATE.
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.
--- include/linux/bpf-cgroup.h | 4 ++++ 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, 12 insertions(+)
include/linux/bpf-cgroup.h +4 -0
@@ -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)
@@ -401,6 +404,7 @@ static inline int bpf_percpu_cgroup_storage_update(struct bpf_map *map,#define BPF_CGROUP_RUN_PROG_INET_INGRESS (sk,skb) ({ 0; })#define BPF_CGROUP_RUN_PROG_INET_EGRESS (sk,skb) ({ 0; })#define BPF_CGROUP_RUN_PROG_INET_SOCK (sk) ({ 0; }) +#define BPF_CGROUP_RUN_PROG_INET_SOCK_RELEASE (sk) ({ 0; })#define BPF_CGROUP_RUN_PROG_INET4_BIND (sk, uaddr) ({ 0; })#define BPF_CGROUP_RUN_PROG_INET6_BIND (sk, uaddr) ({ 0; })#define BPF_CGROUP_RUN_PROG_INET4_POST_BIND (sk) ({ 0; })
include/uapi/linux/bpf.h +1 -0
@@ -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 };
kernel/bpf/syscall.c +3 -0
@@ -1981,6 +1981,7 @@ 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 ;
@@ -2779,6 +2780,7 @@ 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 ;
@@ -2929,6 +2931,7 @@ 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 :
net/core/filter.c +1 -0
@@ -6890,6 +6890,7 @@ 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;
net/ipv4/af_inet.c +3 -0
@@ -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);
Signature · 1 line
2.27.0.212.ge8ba1cc988-goog
PATCH v3 2/4 libbpf: add support for BPF_CGROUP_INET_SOCK_RELEASE
In reply to Stanislav FomichevAdd auto-detection for the cgroup/sock_release programs.
--- tools/include/uapi/linux/bpf.h | 1 + tools/lib/bpf/libbpf.c | 4 ++++ 2 files changed, 5 insertions(+)
tools/include/uapi/linux/bpf.h +1 -0
@@ -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 };
tools/lib/bpf/libbpf.c +4 -0
@@ -6917,6 +6917,10 @@ static const struct bpf_sec_def section_defs[] = {BPF_APROG_SEC ("cgroup_skb/egress" ,BPF_PROG_TYPE_CGROUP_SKB ,BPF_CGROUP_INET_EGRESS ),BPF_APROG_COMPAT ("cgroup/skb" ,BPF_PROG_TYPE_CGROUP_SKB ), +BPF_EAPROG_SEC ("cgroup/sock_create" ,BPF_PROG_TYPE_CGROUP_SOCK , +BPF_CGROUP_INET_SOCK_CREATE ), +BPF_EAPROG_SEC ("cgroup/sock_release" ,BPF_PROG_TYPE_CGROUP_SOCK , +BPF_CGROUP_INET_SOCK_RELEASE ),BPF_APROG_SEC ("cgroup/sock" ,BPF_PROG_TYPE_CGROUP_SOCK ,BPF_CGROUP_INET_SOCK_CREATE ), BPF_EAPROG_SEC("cgroup/post_bind4", BPF_PROG_TYPE_CGROUP_SOCK,
Signature · 1 line
2.27.0.212.ge8ba1cc988-goog
PATCH v3 3/4 bpftool: add support for BPF_CGROUP_INET_SOCK_RELEASE
In reply to Stanislav FomichevSupport attaching to BPF_CGROUP_INET_SOCK_RELEASE and properly display attach type upon prog dump.
--- tools/bpf/bpftool/common.c | 1 + 1 file changed, 1 insertion(+)
tools/bpf/bpftool/common.c +1 -0
@@ -33,6 +33,7 @@ const char * const attach_type_name[__MAX_BPF_ATTACH_TYPE] = { [BPF_CGROUP_INET_INGRESS] = "ingress", [BPF_CGROUP_INET_EGRESS] = "egress", [BPF_CGROUP_INET_SOCK_CREATE] = "sock_create", + [BPF_CGROUP_INET_SOCK_RELEASE] = "sock_release", [BPF_CGROUP_SOCK_OPS] = "sock_ops", [BPF_CGROUP_DEVICE] = "device", [BPF_CGROUP_INET4_BIND] = "bind4",
Signature · 1 line
2.27.0.212.ge8ba1cc988-goog
PATCH v3 4/4 selftests/bpf: test BPF_CGROUP_INET_SOCK_RELEASE
In reply to Stanislav FomichevSimple test that enforces a single SOCK_DGRAM socker per cgroup.
--- .../selftests/bpf/prog_tests/udp_limit.c | 72 +++++++++++++++++++ tools/testing/selftests/bpf/progs/udp_limit.c | 42 +++++++++++ 2 files changed, 114 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/udp_limit.c create mode 100644 tools/testing/selftests/bpf/progs/udp_limit.c
tools/testing/selftests/bpf/prog_tests/udp_limit.c new file +72 -0
@@ -0,0 +1,72 @@ +// SPDX-License-Identifier: GPL-2.0 +#include <test_progs.h> +#include "udp_limit.skel.h" + +#include <sys/types.h> +#include <sys/socket.h> + +static int duration; + +void test_udp_limit (void ) +{ +struct udp_limit *skel; +int cgroup_fd; +int fd1, fd2; + + cgroup_fd =test__join_cgroup ("/udp_limit" ); +if (CHECK (cgroup_fd <0 ,"cg-join" ,"errno %d" , errno)) +return ; + + skel =udp_limit__open_and_load (); +if (CHECK (!skel,"skel-load" ,"errno %d" , errno)) + goto close_cgroup_fd; + + skel->links.sock =bpf_program__attach_cgroup (skel->progs.sock, cgroup_fd); + skel->links.sock_release =bpf_program__attach_cgroup (skel->progs.sock_release, cgroup_fd); +if (CHECK (IS_ERR (skel->links.sock) ||IS_ERR (skel->links.sock_release), +"cg-attach" ,"sock %ld sock_release %ld" , +PTR_ERR (skel->links.sock), +PTR_ERR (skel->links.sock_release))) + goto close_skeleton; + +/* BPF program enforces a single UDP socket per cgroup, +* verify that. +*/ + fd1 =socket (AF_INET ,SOCK_DGRAM ,0 ); +if (CHECK (fd1 <0 ,"fd1" ,"errno %d" , errno)) + goto close_skeleton; + + fd2 =socket (AF_INET ,SOCK_DGRAM ,0 ); +if (CHECK (fd2 >=0 ,"fd2" ,"errno %d" , errno)) + goto close_fd1; + +/* We can reopen again after close. */ +close (fd1); + + fd1 =socket (AF_INET ,SOCK_DGRAM ,0 ); +if (CHECK (fd1 <0 ,"fd1-again" ,"errno %d" , errno)) + goto close_skeleton; + +/* Make sure the program was invoked the expected +* number of times: +* - open fd1 - BPF_CGROUP_INET_SOCK_CREATE +* - attempt to openfd2 - BPF_CGROUP_INET_SOCK_CREATE +* - close fd1 - BPF_CGROUP_INET_SOCK_RELEASE +* - open fd1 again - BPF_CGROUP_INET_SOCK_CREATE +*/ +if (CHECK (skel->bss->invocations !=4 ,"bss-invocations" , +"invocations=%d" , skel->bss->invocations)) + goto close_fd1; + +/* We should still have a single socket in use */ +if (CHECK (skel->bss->in_use !=1 ,"bss-in_use" , +"in_use=%d" , skel->bss->in_use)) + goto close_fd1; + +close_fd1: +close (fd1); +close_skeleton: +udp_limit__destroy (skel); +close_cgroup_fd: +close (cgroup_fd); +}
tools/testing/selftests/bpf/progs/udp_limit.c new file +42 -0
42 lines not on this page · raw message
Signature · 1 line
2.27.0.212.ge8ba1cc988-goog
Re: [PATCH bpf-next v3 4/4] selftests/bpf: test BPF_CGROUP_INET_SOCK_RELEASE
Quoted 11 lines · Stanislav Fomichev
Simple test that enforces a single
SOCK_DGRAMsocker per cgroup.Signed-off-by Stanislav Fomichev <[email protected]>--- .../selftests/bpf/prog_tests/udp_limit.c | 72 +++++++++++++++++++ tools/testing/selftests/bpf/progs/udp_limit.c | 42 +++++++++++ 2 files changed, 114 insertions(+) create mode 100644 tools/testing/selftests/bpf/prog_tests/udp_limit.c create mode 100644 tools/testing/selftests/bpf/progs/udp_limit.c
[...]
tools/testing/selftests/bpf/prog_tests/udp_limit.c line 40
Quoted 6 lines
tools/testing/selftests/bpf/prog_tests/udp_limit.c new file +6
-0+ fd1 =socket (AF_INET ,SOCK_DGRAM ,0 ); +if (CHECK (fd1 <0 ,"fd1" ,"errno %d" , errno)) + goto close_skeleton; + + fd2 =socket (AF_INET ,SOCK_DGRAM ,0 ); +if (CHECK (fd2 >=0 ,"fd2" ,"errno %d" , errno))
close(fd2);
Quoted 2 lines
(quoted diff) +2
-0+ goto close_fd1; +
[...]
tools/testing/selftests/bpf/prog_tests/udp_limit.c line 71
Quoted 6 lines
tools/testing/selftests/bpf/prog_tests/udp_limit.c new file +6
-0+close_fd1: +close (fd1); +close_skeleton: +udp_limit__destroy (skel); +close_cgroup_fd: +close (cgroup_fd);
nit: choosing between close_fd1 and close_skeleton (which also alternates!) is hard to keep track of. When clean up gets one step beyond trivial, I usually just initialize variables properly and do clean up for all exit scenario in one block:
if (fd1 >= 0)
close(fd1);
udp_limit__destroy(skel);close(cgroup_fd);
It also makes later extensions simpler.
Quoted 7 lines
(quoted diff) +1
-0+}tools/testing/selftests/bpf/progs/udp_limit.c new file +0
-0
tools/testing/selftests/bpf/progs/udp_limit.c line 7
Quoted 7 lines
tools/testing/selftests/bpf/progs/udp_limit.c new file +7
-0@@ -0,0 +1,42 @@ +// SPDX-License-Identifier: GPL-2.0-only + +#include <sys/socket.h> +#include <linux/bpf.h> +#include <bpf/bpf_helpers.h> + +int invocations, in_use;
uninitialized globals can leads to libbpf refusing to load them (due to COM section), Daniel just recently had this problem. So better to always zero-initialize them.
[...]
Re: [PATCH bpf-next v3 3/4] bpftool: add support for BPF_CGROUP_INET_SOCK_RELEASE
In reply to Stanislav FomichevQuoted 6 lines · Stanislav Fomichev
Support attaching to
BPF_CGROUP_INET_SOCK_RELEASEand properly display attach type upon prog dump.Signed-off-by Stanislav Fomichev <[email protected]>---
Quoted 18 lines
tools/bpf/bpftool/common.c | 1 + 1 file changed, 1 insertion(+)tools/bpf/bpftool/common.c +1
-1@@ -33,6 +33,7 @@ const char * const attach_type_name[__MAX_BPF_ATTACH_TYPE] = { [BPF_CGROUP_INET_INGRESS] = "ingress", [BPF_CGROUP_INET_EGRESS] = "egress", [BPF_CGROUP_INET_SOCK_CREATE] = "sock_create", + [BPF_CGROUP_INET_SOCK_RELEASE] = "sock_release", [BPF_CGROUP_SOCK_OPS] = "sock_ops", [BPF_CGROUP_DEVICE] = "device", [BPF_CGROUP_INET4_BIND] = "bind4", --2.27.0.212.ge8ba1cc988-goog