From: Geliang Tang <tanggeliang@kylinos.cn> v2: - squash patches 5,6 of v1 into one. - patch 3, use lockdep_assert_held instead of spin_is_locked. - update patch 5 to make it run normally in btf-debug mode. Another bpf_iter for BPF path manager to traverse all address entries on userspace_pm_local_addr_list of an MPTCP socket is added in this set. Userspace pm tool is needed by the self test, so add pm_nl_ctl link in patch 1. Geliang Tang (5): selftests/bpf: Add mptcp pm_nl_ctl link selftests/bpf: Use pm_nl_ctl if ip mptcp not supported bpf: Add mptcp_address bpf_iter bpf: Export more helpers used by mptcp_address test selftests/bpf: Add mptcp_address bpf_iter subtest MAINTAINERS | 1 + net/mptcp/bpf.c | 66 +++++ tools/testing/selftests/bpf/Makefile | 3 +- .../testing/selftests/bpf/bpf_experimental.h | 8 + tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c | 1 + .../testing/selftests/bpf/prog_tests/mptcp.c | 235 +++++++++++++++++- tools/testing/selftests/bpf/progs/mptcp_bpf.h | 5 + .../selftests/bpf/progs/mptcp_bpf_iters.c | 39 +++ 8 files changed, 349 insertions(+), 9 deletions(-) create mode 120000 tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> This patch adds a symlink to MPTCP's pm_nl_ctl tool into bpf selftests, and updates Makefile to compile it. This is useful to run MPTCP BPF selftests on systems with an old version of IPRoute2. This tool can be used as an alternative to 'ip mptcp'. In addition, this tool is used as userspace path manager. "csf" and "dsf" commands of pm_nl_ctl for creating and destroying subflows, and "ann" and "rem" commands for signaling ADD_ADDR and RM_ADDR. These commands are not currently supported in 'ip mptcp' yet. MAINTAINERS needs to be updated since a new file is added in a non covered place. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> --- MAINTAINERS | 1 + tools/testing/selftests/bpf/Makefile | 3 ++- tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c | 1 + 3 files changed, 4 insertions(+), 1 deletion(-) create mode 120000 tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: include/trace/events/mptcp.h F: include/uapi/linux/mptcp*.h F: net/mptcp/ F: tools/testing/selftests/bpf/*/*mptcp*.[ch] +F: tools/testing/selftests/bpf/*mptcp*.[ch] F: tools/testing/selftests/net/mptcp/ NETWORKING [TCP] diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -XXX,XX +XXX,XX @@ TEST_GEN_PROGS_EXTENDED = \ flow_dissector_load test_flow_dissector test_tcp_check_syncookie_user \ test_lirc_mode2_user xdping test_cpp runqslower bench bpf_testmod.ko \ xskxceiver xdp_redirect_multi xdp_synproxy veristat xdp_hw_metadata \ - xdp_features bpf_test_no_cfi.ko + xdp_features bpf_test_no_cfi.ko mptcp_pm_nl_ctl TEST_GEN_FILES += liburandom_read.so urandom_read sign-file uprobe_multi @@ -XXX,XX +XXX,XX @@ TRUNNER_EXTRA_FILES := $(OUTPUT)/urandom_read $(OUTPUT)/bpf_testmod.ko \ $(OUTPUT)/xdp_synproxy \ $(OUTPUT)/sign-file \ $(OUTPUT)/uprobe_multi \ + $(OUTPUT)/mptcp_pm_nl_ctl \ ima_setup.sh \ verify_sig_setup.sh \ $(wildcard progs/btf_dump_test_case_*.c) \ diff --git a/tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c b/tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c new file mode 120000 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c @@ -0,0 +1 @@ +../net/mptcp/pm_nl_ctl.c \ No newline at end of file -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> If running MPTCP BPF selftests on systems with an old version of IPRoute2, 'ip mptcp' command is not supported. In this case, instead of skipping the test, falling back to using 'pm_nl_ctl' tool is a better option. This patch adds an 'ip_mptcp' argument for endpoint_add() to control whether to use 'ip mptcp' or 'pm_nl_ctl' to add an endpoint. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- .../testing/selftests/bpf/prog_tests/mptcp.c | 21 ++++++++++++------- 1 file changed, 13 insertions(+), 8 deletions(-) diff --git a/tools/testing/selftests/bpf/prog_tests/mptcp.c b/tools/testing/selftests/bpf/prog_tests/mptcp.c index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/prog_tests/mptcp.c +++ b/tools/testing/selftests/bpf/prog_tests/mptcp.c @@ -XXX,XX +XXX,XX @@ #define ADDR6_3 "dead:beef:3::1" #define ADDR6_4 "dead:beef:4::1" #define PORT_1 10001 +#define PM_CTL "./mptcp_pm_nl_ctl" #define WITH_DATA true #define WITHOUT_DATA false @@ -XXX,XX +XXX,XX @@ static int address_init(void) return -1; } -static int endpoint_add(char *addr, char *flags) +static int endpoint_add(char *addr, char *flags, bool ip_mptcp) { - return SYS_NOFAIL("ip -net %s mptcp endpoint add %s %s", NS_TEST, addr, flags); + if (ip_mptcp) + return SYS_NOFAIL("ip -net %s mptcp endpoint add %s %s", + NS_TEST, addr, flags); + return SYS_NOFAIL("ip netns exec %s %s add %s flags %s", + NS_TEST, PM_CTL, addr, flags); } static int endpoint_init(char *flags, u8 endpoints) { + bool ip_mptcp = true; int ret = -1; if (!endpoints || endpoints > 4) @@ -XXX,XX +XXX,XX @@ static int endpoint_init(char *flags, u8 endpoints) if (SYS_NOFAIL("ip -net %s mptcp limits set add_addr_accepted 4 subflows 4", NS_TEST)) { - printf("'ip mptcp' not supported, skip this test.\n"); - test__skip(); - goto fail; + SYS(fail, "ip netns exec %s %s limits 4 4", NS_TEST, PM_CTL); + ip_mptcp = false; } if (endpoints > 1) - ret = endpoint_add(ADDR_2, flags); + ret = endpoint_add(ADDR_2, flags, ip_mptcp); if (endpoints > 2) - ret = ret ?: endpoint_add(ADDR_3, flags); + ret = ret ?: endpoint_add(ADDR_3, flags, ip_mptcp); if (endpoints > 3) - ret = ret ?: endpoint_add(ADDR_4, flags); + ret = ret ?: endpoint_add(ADDR_4, flags, ip_mptcp); fail: return ret; -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> Just like the mptcp_subflow bpf_iter used to implement the MPTCP BPF packet scheduler, another bpf_iter is also needed, named mptcp_address, to traverse all address entries on userspace_pm_local_addr_list of an MPTCP socket for implementing the MPTCP BPF path manager. In kernel space, we walk this list like this: list_for_each_entry(entry, &msk->pm.userspace_pm_local_addr_list, list) kfunc(entry); With the mptcp_address bpf_iter, bpf_for_each() can be used to do the same thing in BPF program: bpf_for_each(mptcp_address, entry, msk) kfunc(entry); This bpf_iter should be invoked under holding the msk pm lock, so use lockdep_assert_held() to assert the lock is holding. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- net/mptcp/bpf.c | 45 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 45 insertions(+) diff --git a/net/mptcp/bpf.c b/net/mptcp/bpf.c index XXXXXXX..XXXXXXX 100644 --- a/net/mptcp/bpf.c +++ b/net/mptcp/bpf.c @@ -XXX,XX +XXX,XX @@ struct bpf_iter_mptcp_subflow_kern { struct list_head *pos; } __aligned(8); +struct bpf_iter_mptcp_address { + __u64 __opaque[2]; +} __aligned(8); + +struct bpf_iter_mptcp_address_kern { + struct mptcp_sock *msk; + struct list_head *pos; +} __aligned(8); + __bpf_kfunc_start_defs(); __bpf_kfunc static struct mptcp_sock *bpf_mptcp_sk(struct sock *sk) @@ -XXX,XX +XXX,XX @@ bpf_iter_mptcp_subflow_destroy(struct bpf_iter_mptcp_subflow *it) { } +__bpf_kfunc static int +bpf_iter_mptcp_address_new(struct bpf_iter_mptcp_address *it, + struct mptcp_sock *msk) +{ + struct bpf_iter_mptcp_address_kern *kit = (void *)it; + + kit->msk = msk; + if (!msk) + return -EINVAL; + + lockdep_assert_held(&msk->pm.lock); + + kit->pos = &msk->pm.userspace_pm_local_addr_list; + return 0; +} + +__bpf_kfunc static struct mptcp_pm_addr_entry * +bpf_iter_mptcp_address_next(struct bpf_iter_mptcp_address *it) +{ + struct bpf_iter_mptcp_address_kern *kit = (void *)it; + + if (!kit->msk || list_is_last(kit->pos, &kit->msk->pm.userspace_pm_local_addr_list)) + return NULL; + + kit->pos = kit->pos->next; + return list_entry(kit->pos, struct mptcp_pm_addr_entry, list); +} + +__bpf_kfunc static void +bpf_iter_mptcp_address_destroy(struct bpf_iter_mptcp_address *it) +{ +} + __bpf_kfunc static struct mptcp_sock *bpf_mptcp_sock_acquire(struct mptcp_sock *msk) { @@ -XXX,XX +XXX,XX @@ BTF_ID_FLAGS(func, bpf_mptcp_subflow_tcp_sock) BTF_ID_FLAGS(func, bpf_iter_mptcp_subflow_new, KF_ITER_NEW | KF_TRUSTED_ARGS) BTF_ID_FLAGS(func, bpf_iter_mptcp_subflow_next, KF_ITER_NEXT | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_iter_mptcp_subflow_destroy, KF_ITER_DESTROY) +BTF_ID_FLAGS(func, bpf_iter_mptcp_address_new, KF_ITER_NEW | KF_TRUSTED_ARGS) +BTF_ID_FLAGS(func, bpf_iter_mptcp_address_next, KF_ITER_NEXT | KF_RET_NULL) +BTF_ID_FLAGS(func, bpf_iter_mptcp_address_destroy, KF_ITER_DESTROY) BTF_ID_FLAGS(func, bpf_mptcp_sock_acquire, KF_ACQUIRE | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_mptcp_sock_release, KF_RELEASE) BTF_KFUNCS_END(bpf_mptcp_common_kfunc_ids) -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> The mptcp_address bpf_iter should be invoked under holding the msk pm lock, so spin_lock_bh() and spin_unlock_bh() are needed in BPF program to hold or release the msk pm lock. And bpf_ipv6_addr_v4mapped() helper is also needed by the mptcp_address bpf_iter selftest program. This patch adds the corresponding BPF wrappers for these helpers and adds them to the mptcp common kfunc_set. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- net/mptcp/bpf.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) diff --git a/net/mptcp/bpf.c b/net/mptcp/bpf.c index XXXXXXX..XXXXXXX 100644 --- a/net/mptcp/bpf.c +++ b/net/mptcp/bpf.c @@ -XXX,XX +XXX,XX @@ __bpf_kfunc static void bpf_mptcp_sock_release(struct mptcp_sock *msk) WARN_ON_ONCE(!sk || !refcount_dec_not_one(&sk->sk_refcnt)); } +__bpf_kfunc static void bpf_spin_lock_bh(spinlock_t *lock) +{ + spin_lock_bh(lock); +} + +__bpf_kfunc static void bpf_spin_unlock_bh(spinlock_t *lock) +{ + spin_unlock_bh(lock); +} + +__bpf_kfunc static bool bpf_ipv6_addr_v4mapped(const struct mptcp_addr_info *a) +{ +#if IS_ENABLED(CONFIG_MPTCP_IPV6) + return ipv6_addr_v4mapped(&a->addr6); +#endif + return false; +} + __bpf_kfunc struct mptcp_subflow_context * bpf_mptcp_subflow_ctx_by_pos(const struct mptcp_sched_data *data, unsigned int pos) { @@ -XXX,XX +XXX,XX @@ BTF_ID_FLAGS(func, bpf_iter_mptcp_address_next, KF_ITER_NEXT | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_iter_mptcp_address_destroy, KF_ITER_DESTROY) BTF_ID_FLAGS(func, bpf_mptcp_sock_acquire, KF_ACQUIRE | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_mptcp_sock_release, KF_RELEASE) +BTF_ID_FLAGS(func, bpf_spin_lock_bh) +BTF_ID_FLAGS(func, bpf_spin_unlock_bh) +BTF_ID_FLAGS(func, bpf_ipv6_addr_v4mapped) BTF_KFUNCS_END(bpf_mptcp_common_kfunc_ids) static const struct btf_kfunc_id_set bpf_mptcp_common_kfunc_set = { -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> This patch adds a test program for the newly added mptcp_address bpf_iter in SEC "cgroup/getsockopt". This test iterates over all address entries on the local address list of userspace PM and check whether each one is an IPv4mapped address. Export mptcp_address helpers bpf_iter_mptcp_address_new/_next/_destroy into bpf_experimental.h. Use bpf_mptcp_sock_acquire() to acquire the msk, then lock the msk pm lock and use bpf_for_each(mptcp_address) to walk the local address list of this msk. Invoke bpf_ipv6_addr_v4mapped() in the loop to check whether the address is an IPv4mapped one. Then Add the address ID of each entry to local veriable local_ids. Out of the loop, unlock the msk pm lock and use bpf_mptcp_sock_release() to release the msk. Finally, assign local_ids to global variable ids so that the application can obtain this value. Add a subtest named test_iters_address to load and verify the newly added mptcp_address type bpf_iter example in test_mptcp. Since mptcp_address bpf_iter iterates over all address entries on the local address list of userspace PM. A set of userspace PM helpers are added. Set pm_type and start "pm_nl_ctl events" command to save pm events into the log file /tmp/bpf_userspace_pm_events in userspace_pm_init(). Kill "pm_nl_ctl" command and remove the log file in userspace_pm_cleanup(). Parse the log file in userspace_pm_add_subflow() to get the token, sport and dport values, then use "pm_nl_ctl csf" command to create a subflow. Use the helper userspace_pm_add_subflow() to add 3 new subflow endpoints. Send a byte of message to start the mptcp connection, and receive the message. getsockopt() is invoked to trigger the "cgroup/getsockopt" test program. Check if skel->bss->ids equals 60 to verify whether the program loops three times as expected. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- .../testing/selftests/bpf/bpf_experimental.h | 8 + .../testing/selftests/bpf/prog_tests/mptcp.c | 214 ++++++++++++++++++ tools/testing/selftests/bpf/progs/mptcp_bpf.h | 5 + .../selftests/bpf/progs/mptcp_bpf_iters.c | 39 ++++ 4 files changed, 266 insertions(+) diff --git a/tools/testing/selftests/bpf/bpf_experimental.h b/tools/testing/selftests/bpf/bpf_experimental.h index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/bpf_experimental.h +++ b/tools/testing/selftests/bpf/bpf_experimental.h @@ -XXX,XX +XXX,XX @@ bpf_iter_mptcp_subflow_next(struct bpf_iter_mptcp_subflow *it) __weak __ksym; extern void bpf_iter_mptcp_subflow_destroy(struct bpf_iter_mptcp_subflow *it) __weak __ksym; +struct bpf_iter_mptcp_address; +extern int bpf_iter_mptcp_address_new(struct bpf_iter_mptcp_address *it, + struct mptcp_sock *msk) __weak __ksym; +extern struct mptcp_pm_addr_entry * +bpf_iter_mptcp_address_next(struct bpf_iter_mptcp_address *it) __weak __ksym; +extern void +bpf_iter_mptcp_address_destroy(struct bpf_iter_mptcp_address *it) __weak __ksym; + extern int bpf_wq_init(struct bpf_wq *wq, void *p__map, unsigned int flags) __weak __ksym; extern int bpf_wq_start(struct bpf_wq *wq, unsigned int flags) __weak __ksym; extern int bpf_wq_set_callback_impl(struct bpf_wq *wq, diff --git a/tools/testing/selftests/bpf/prog_tests/mptcp.c b/tools/testing/selftests/bpf/prog_tests/mptcp.c index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/prog_tests/mptcp.c +++ b/tools/testing/selftests/bpf/prog_tests/mptcp.c @@ -XXX,XX +XXX,XX @@ #define ADDR6_4 "dead:beef:4::1" #define PORT_1 10001 #define PM_CTL "./mptcp_pm_nl_ctl" +#define PM_EVENTS "/tmp/bpf_userspace_pm_events" #define WITH_DATA true #define WITHOUT_DATA false @@ -XXX,XX +XXX,XX @@ #endif #define MPTCP_SCHED_NAME_MAX 16 +enum mptcp_pm_type { + MPTCP_PM_TYPE_KERNEL = 0, + MPTCP_PM_TYPE_USERSPACE, + + __MPTCP_PM_TYPE_NR, + __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, +}; + static const unsigned int total_bytes = 10 * 1024 * 1024; static int duration; @@ -XXX,XX +XXX,XX @@ static void send_byte(int fd) ASSERT_EQ(write(fd, &b, sizeof(b)), 1, "send single byte"); } +static int recv_byte(int fd) +{ + char buf[1]; + ssize_t n; + + n = recv(fd, buf, sizeof(buf), 0); + if (CHECK(n <= 0, "recv_byte", "recv")) { + log_err("failed/partial recv"); + return -1; + } + return 0; +} + static int verify_mptcpify(int server_fd, int client_fd) { struct __mptcp_info info; @@ -XXX,XX +XXX,XX @@ static void test_iters_subflow(void) close(cgroup_fd); } +static int userspace_pm_init(enum mptcp_pm_type pm_type) +{ + if (address_init()) + goto fail; + + SYS(fail, "ip netns exec %s sysctl -qw net.mptcp.pm_type=%u", + NS_TEST, pm_type); + SYS(fail, "ip netns exec %s %s limits 4 4", + NS_TEST, PM_CTL); + SYS(fail, "ip netns exec %s %s events >> %s 2>&1 &", + NS_TEST, PM_CTL, PM_EVENTS); + + return 0; +fail: + return -1; +} + +static void userspace_pm_cleanup(void) +{ + SYS_NOFAIL("ip netns exec %s killall %s > /dev/null 2>&1", + NS_TEST, PM_CTL); + SYS_NOFAIL("ip netns exec %s rm -rf %s", NS_TEST, PM_EVENTS); +} + +static int userspace_pm_get_events_line(char *type, char *line) +{ + char buf[BUFSIZ], *str; + size_t len; + int fd; + + fd = open(PM_EVENTS, O_RDONLY); + if (fd < 0) { + log_err("failed to open pm events\n"); + return -1; + } + + len = read(fd, buf, sizeof(buf)); + close(fd); + if (len <= 0) { + log_err("failed to read pm events\n"); + return -1; + } + + str = strstr(buf, type); + if (!str) { + log_err("failed to get type %s pm event\n", type); + return -1; + } + + strcpy(line, str); + return 0; +} + +static int userspace_pm_get_token(int fd) +{ + char line[1024], *str; + __u32 token; + int i; + + /* Wait max 2 sec for the connection to be established */ + for (i = 0; i < 10; i++) { + usleep(200000); /* 0.2s */ + send_byte(fd); + + sync(); + if (userspace_pm_get_events_line("type:2", line)) + continue; + str = strstr(line, "token"); + if (!str) + continue; + if (sscanf(str, "token:%u,", &token) != 1) + continue; + return token; + } + + return 0; +} + +static int userspace_pm_add_subflow(__u32 token, char *addr, __u8 id) +{ + bool ipv6 = strstr(addr, ":"); + char line[1024], *str; + __u32 sport, dport; + + if (userspace_pm_get_events_line("type:2", line)) + return -1; + + str = strstr(line, "sport"); + if (!str || sscanf(str, "sport:%u,dport:%u,", &sport, &dport) != 2) { + log_err("add_subflow error, str=%s\n", str); + return -1; + } + + str = ipv6 ? (strstr(addr, ".") ? "::ffff:"ADDR_1 : ADDR6_1) : ADDR_1; + SYS_NOFAIL("ip netns exec %s %s csf lip %s lid %u rip %s rport %u token %u", + NS_TEST, PM_CTL, addr, id, str, dport, token); + + return 0; +} + +static void run_iters_address(void) +{ + int server_fd, client_fd, accept_fd; + int is_mptcp, err; + socklen_t len; + __u32 token; + + server_fd = start_mptcp_server(AF_INET6, "::ffff:"ADDR_1, PORT_1, 0); + if (!ASSERT_OK_FD(server_fd, "start_mptcp_server")) + return; + + client_fd = connect_to_fd(server_fd, 0); + if (!ASSERT_OK_FD(client_fd, "connect_to_fd")) + goto close_server; + + accept_fd = accept(server_fd, NULL, NULL); + if (!ASSERT_OK_FD(accept_fd, "accept")) + goto close_client; + + token = userspace_pm_get_token(client_fd); + if (!token) + goto close_client; + recv_byte(accept_fd); + usleep(200000); /* 0.2s */ + + err = userspace_pm_add_subflow(token, "::ffff:"ADDR_2, 10); + err = err ?: userspace_pm_add_subflow(token, "::ffff:"ADDR_3, 20); + err = err ?: userspace_pm_add_subflow(token, "::ffff:"ADDR_4, 30); + if (!ASSERT_OK(err, "userspace_pm_add_subflow")) + goto close_accept; + + send_byte(accept_fd); + recv_byte(client_fd); + + len = sizeof(is_mptcp); + /* mainly to trigger the BPF program */ + err = getsockopt(client_fd, SOL_TCP, TCP_IS_MPTCP, &is_mptcp, &len); + if (ASSERT_OK(err, "getsockopt(client_fd, TCP_IS_MPTCP)")) + ASSERT_EQ(is_mptcp, 1, "is_mptcp"); + +close_accept: + close(accept_fd); +close_client: + close(client_fd); +close_server: + close(server_fd); +} + +static void test_iters_address(void) +{ + struct mptcp_bpf_iters *skel; + struct nstoken *nstoken; + int cgroup_fd; + int err; + + cgroup_fd = test__join_cgroup("/iters_address"); + if (!ASSERT_OK_FD(cgroup_fd, "join_cgroup: iters_address")) + return; + + skel = mptcp_bpf_iters__open_and_load(); + if (!ASSERT_OK_PTR(skel, "skel_open_load: iters_address")) + goto close_cgroup; + + skel->links.iters_address = bpf_program__attach_cgroup(skel->progs.iters_address, + cgroup_fd); + if (!ASSERT_OK_PTR(skel->links.iters_address, "attach getsockopt")) + goto skel_destroy; + + nstoken = create_netns(); + if (!ASSERT_OK_PTR(nstoken, "create_netns")) + goto skel_destroy; + + err = userspace_pm_init(MPTCP_PM_TYPE_USERSPACE); + if (!ASSERT_OK(err, "userspace_pm_init: iters_address")) + goto close_netns; + + run_iters_address(); + + /* 10 + 20 + 30 = 60 */ + ASSERT_EQ(skel->bss->ids, 60, "address ids"); + + userspace_pm_cleanup(); +close_netns: + cleanup_netns(nstoken); +skel_destroy: + mptcp_bpf_iters__destroy(skel); +close_cgroup: + close(cgroup_fd); +} + static struct nstoken *sched_init(char *flags, char *sched) { struct nstoken *nstoken; @@ -XXX,XX +XXX,XX @@ void test_mptcp(void) test_subflow(); if (test__start_subtest("iters_subflow")) test_iters_subflow(); + if (test__start_subtest("iters_address")) + test_iters_address(); if (test__start_subtest("default")) test_default(); if (test__start_subtest("first")) diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf.h b/tools/testing/selftests/bpf/progs/mptcp_bpf.h index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/progs/mptcp_bpf.h +++ b/tools/testing/selftests/bpf/progs/mptcp_bpf.h @@ -XXX,XX +XXX,XX @@ bpf_mptcp_subflow_ctx(const struct sock *sk) __ksym; extern struct sock * bpf_mptcp_subflow_tcp_sock(const struct mptcp_subflow_context *subflow) __ksym; +extern void bpf_spin_lock_bh(spinlock_t *lock) __ksym; +extern void bpf_spin_unlock_bh(spinlock_t *lock) __ksym; + +extern bool bpf_ipv6_addr_v4mapped(const struct mptcp_addr_info *a) __ksym; + extern void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, bool scheduled) __ksym; diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c +++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c @@ -XXX,XX +XXX,XX @@ int iters_subflow(struct bpf_sockopt *ctx) bpf_mptcp_sock_release(msk); return 1; } + +SEC("cgroup/getsockopt") +int iters_address(struct bpf_sockopt *ctx) +{ + struct mptcp_pm_addr_entry *entry; + struct bpf_sock *sk = ctx->sk; + struct mptcp_sock *msk; + int local_ids = 0; + + if (!sk || sk->protocol != IPPROTO_MPTCP || + ctx->level != SOL_TCP || ctx->optname != TCP_IS_MPTCP) + return 1; + + msk = bpf_mptcp_sk((struct sock *)sk); + if (msk->pm.server_side) + return 1; + + msk = bpf_mptcp_sock_acquire(msk); + if (!msk) + return 1; + bpf_spin_lock_bh(&msk->pm.lock); + bpf_for_each(mptcp_address, entry, msk) { + /* Here MPTCP-specific path manager kfunc can be called: + * this test is not doing anything really useful, only to + * verify the iteration works. + */ + + if (!bpf_ipv6_addr_v4mapped(&entry->addr)) + break; + + local_ids += entry->addr.id; + } + bpf_spin_unlock_bh(&msk->pm.lock); + bpf_mptcp_sock_release(msk); + + ids = local_ids; + + return 1; +} -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> v3: - rename mptcp_address to mptcp_userspace_pm_addr as I mentioned in [1]. - rebased. [1] https://patchwork.kernel.org/project/mptcp/patch/b6b53fd730a0edd0024963468d078488ad5e67aa.1729588019.git.tanggeliang@kylinos.cn/ v2: - squash patches 5,6 of v1 into one. - patch 3, use lockdep_assert_held instead of spin_is_locked. - update patch 5 to make it run normally in btf-debug mode. Another bpf_iter for BPF path manager to traverse all address entries on userspace_pm_local_addr_list of an MPTCP socket is added in this set. Userspace pm tool is needed by the self test, so add pm_nl_ctl link in patch 1. Geliang Tang (5): selftests/bpf: Add mptcp pm_nl_ctl link selftests/bpf: Use pm_nl_ctl if ip mptcp not supported bpf: Add mptcp_userspace_pm_addr bpf_iter bpf: Export more helpers for mptcp_userspace_pm_addr test selftests/bpf: Add mptcp_userspace_pm_addr bpf_iter subtest MAINTAINERS | 1 + net/mptcp/bpf.c | 67 +++++ tools/testing/selftests/bpf/Makefile | 4 +- .../testing/selftests/bpf/bpf_experimental.h | 8 + tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c | 1 + .../testing/selftests/bpf/prog_tests/mptcp.c | 235 +++++++++++++++++- tools/testing/selftests/bpf/progs/mptcp_bpf.h | 5 + .../selftests/bpf/progs/mptcp_bpf_iters.c | 39 +++ 8 files changed, 351 insertions(+), 9 deletions(-) create mode 120000 tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> This patch adds a symlink to MPTCP's pm_nl_ctl tool into bpf selftests, and updates Makefile to compile it. This is useful to run MPTCP BPF selftests on systems with an old version of IPRoute2. This tool can be used as an alternative to 'ip mptcp'. In addition, this tool is used as userspace path manager. "csf" and "dsf" commands of pm_nl_ctl for creating and destroying subflows, and "ann" and "rem" commands for signaling ADD_ADDR and RM_ADDR. These commands are not currently supported in 'ip mptcp' yet. MAINTAINERS needs to be updated since a new file is added in a non covered place. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org> --- MAINTAINERS | 1 + tools/testing/selftests/bpf/Makefile | 4 +++- tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c | 1 + 3 files changed, 5 insertions(+), 1 deletion(-) create mode 120000 tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c diff --git a/MAINTAINERS b/MAINTAINERS index XXXXXXX..XXXXXXX 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -XXX,XX +XXX,XX @@ F: include/trace/events/mptcp.h F: include/uapi/linux/mptcp*.h F: net/mptcp/ F: tools/testing/selftests/bpf/*/*mptcp*.[ch] +F: tools/testing/selftests/bpf/*mptcp*.[ch] F: tools/testing/selftests/net/mptcp/ NETWORKING [TCP] diff --git a/tools/testing/selftests/bpf/Makefile b/tools/testing/selftests/bpf/Makefile index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/Makefile +++ b/tools/testing/selftests/bpf/Makefile @@ -XXX,XX +XXX,XX @@ TEST_GEN_PROGS_EXTENDED = \ xdp_redirect_multi \ xdp_synproxy \ xdping \ - xskxceiver + xskxceiver \ + mptcp_pm_nl_ctl TEST_GEN_FILES += liburandom_read.so urandom_read sign-file uprobe_multi @@ -XXX,XX +XXX,XX @@ TRUNNER_EXTRA_FILES := $(OUTPUT)/urandom_read $(OUTPUT)/bpf_testmod.ko \ $(OUTPUT)/xdp_synproxy \ $(OUTPUT)/sign-file \ $(OUTPUT)/uprobe_multi \ + $(OUTPUT)/mptcp_pm_nl_ctl \ ima_setup.sh \ verify_sig_setup.sh \ $(wildcard progs/btf_dump_test_case_*.c) \ diff --git a/tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c b/tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c new file mode 120000 index XXXXXXX..XXXXXXX --- /dev/null +++ b/tools/testing/selftests/bpf/mptcp_pm_nl_ctl.c @@ -0,0 +1 @@ +../net/mptcp/pm_nl_ctl.c \ No newline at end of file -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> If running MPTCP BPF selftests on systems with an old version of IPRoute2, 'ip mptcp' command is not supported. In this case, instead of skipping the test, falling back to using 'pm_nl_ctl' tool is a better option. This patch adds an 'ip_mptcp' argument for endpoint_add() to control whether to use 'ip mptcp' or 'pm_nl_ctl' to add an endpoint. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- .../testing/selftests/bpf/prog_tests/mptcp.c | 21 ++++++++++++------- 1 file changed, 13 insertions(+), 8 deletions(-) diff --git a/tools/testing/selftests/bpf/prog_tests/mptcp.c b/tools/testing/selftests/bpf/prog_tests/mptcp.c index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/prog_tests/mptcp.c +++ b/tools/testing/selftests/bpf/prog_tests/mptcp.c @@ -XXX,XX +XXX,XX @@ #define ADDR6_3 "dead:beef:3::1" #define ADDR6_4 "dead:beef:4::1" #define PORT_1 10001 +#define PM_CTL "./mptcp_pm_nl_ctl" #define WITH_DATA true #define WITHOUT_DATA false @@ -XXX,XX +XXX,XX @@ static int address_init(void) return -1; } -static int endpoint_add(char *addr, char *flags) +static int endpoint_add(char *addr, char *flags, bool ip_mptcp) { - return SYS_NOFAIL("ip -net %s mptcp endpoint add %s %s", NS_TEST, addr, flags); + if (ip_mptcp) + return SYS_NOFAIL("ip -net %s mptcp endpoint add %s %s", + NS_TEST, addr, flags); + return SYS_NOFAIL("ip netns exec %s %s add %s flags %s", + NS_TEST, PM_CTL, addr, flags); } static int endpoint_init(char *flags, u8 endpoints) { + bool ip_mptcp = true; int ret = -1; if (!endpoints || endpoints > 4) @@ -XXX,XX +XXX,XX @@ static int endpoint_init(char *flags, u8 endpoints) if (SYS_NOFAIL("ip -net %s mptcp limits set add_addr_accepted 4 subflows 4", NS_TEST)) { - printf("'ip mptcp' not supported, skip this test.\n"); - test__skip(); - goto fail; + SYS(fail, "ip netns exec %s %s limits 4 4", NS_TEST, PM_CTL); + ip_mptcp = false; } if (endpoints > 1) - ret = endpoint_add(ADDR_2, flags); + ret = endpoint_add(ADDR_2, flags, ip_mptcp); if (endpoints > 2) - ret = ret ?: endpoint_add(ADDR_3, flags); + ret = ret ?: endpoint_add(ADDR_3, flags, ip_mptcp); if (endpoints > 3) - ret = ret ?: endpoint_add(ADDR_4, flags); + ret = ret ?: endpoint_add(ADDR_4, flags, ip_mptcp); fail: return ret; -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> Just like the mptcp_subflow bpf_iter used to implement the MPTCP BPF packet scheduler, another bpf_iter is also needed, named mptcp_userspace_pm_addr, to traverse all address entries on userspace_pm_local_addr_list of an MPTCP socket for implementing the MPTCP BPF path manager. In kernel space, we walk this list like this: mptcp_for_each_userspace_pm_addr(msk, entry) kfunc(entry); With the mptcp_userspace_pm_addr bpf_iter, bpf_for_each() can be used to do the same thing in BPF program: bpf_for_each(mptcp_userspace_pm_addr, entry, msk) kfunc(entry); This bpf_iter should be invoked under holding the msk pm lock, so use lockdep_assert_held() to assert the lock is holding. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- net/mptcp/bpf.c | 46 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) diff --git a/net/mptcp/bpf.c b/net/mptcp/bpf.c index XXXXXXX..XXXXXXX 100644 --- a/net/mptcp/bpf.c +++ b/net/mptcp/bpf.c @@ -XXX,XX +XXX,XX @@ struct bpf_iter_mptcp_subflow_kern { struct list_head *pos; } __aligned(8); +struct bpf_iter_mptcp_userspace_pm_addr { + __u64 __opaque[2]; +} __aligned(8); + +struct bpf_iter_mptcp_userspace_pm_addr_kern { + struct mptcp_sock *msk; + struct list_head *pos; +} __aligned(8); + __bpf_kfunc_start_defs(); __bpf_kfunc static struct mptcp_sock *bpf_mptcp_sk(struct sock *sk) @@ -XXX,XX +XXX,XX @@ bpf_iter_mptcp_subflow_destroy(struct bpf_iter_mptcp_subflow *it) { } +__bpf_kfunc static int +bpf_iter_mptcp_userspace_pm_addr_new(struct bpf_iter_mptcp_userspace_pm_addr *it, + struct mptcp_sock *msk) +{ + struct bpf_iter_mptcp_userspace_pm_addr_kern *kit = (void *)it; + + kit->msk = msk; + if (!msk) + return -EINVAL; + + lockdep_assert_held(&msk->pm.lock); + + kit->pos = &msk->pm.userspace_pm_local_addr_list; + return 0; +} + +__bpf_kfunc static struct mptcp_pm_addr_entry * +bpf_iter_mptcp_userspace_pm_addr_next(struct bpf_iter_mptcp_userspace_pm_addr *it) +{ + struct bpf_iter_mptcp_userspace_pm_addr_kern *kit = (void *)it; + + if (!kit->msk || list_is_last(kit->pos, + &kit->msk->pm.userspace_pm_local_addr_list)) + return NULL; + + kit->pos = kit->pos->next; + return list_entry(kit->pos, struct mptcp_pm_addr_entry, list); +} + +__bpf_kfunc static void +bpf_iter_mptcp_userspace_pm_addr_destroy(struct bpf_iter_mptcp_userspace_pm_addr *it) +{ +} + __bpf_kfunc static struct mptcp_sock *bpf_mptcp_sock_acquire(struct mptcp_sock *msk) { @@ -XXX,XX +XXX,XX @@ BTF_ID_FLAGS(func, bpf_mptcp_subflow_tcp_sock) BTF_ID_FLAGS(func, bpf_iter_mptcp_subflow_new, KF_ITER_NEW | KF_TRUSTED_ARGS) BTF_ID_FLAGS(func, bpf_iter_mptcp_subflow_next, KF_ITER_NEXT | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_iter_mptcp_subflow_destroy, KF_ITER_DESTROY) +BTF_ID_FLAGS(func, bpf_iter_mptcp_userspace_pm_addr_new, KF_ITER_NEW | KF_TRUSTED_ARGS) +BTF_ID_FLAGS(func, bpf_iter_mptcp_userspace_pm_addr_next, KF_ITER_NEXT | KF_RET_NULL) +BTF_ID_FLAGS(func, bpf_iter_mptcp_userspace_pm_addr_destroy, KF_ITER_DESTROY) BTF_ID_FLAGS(func, bpf_mptcp_sock_acquire, KF_ACQUIRE | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_mptcp_sock_release, KF_RELEASE) BTF_KFUNCS_END(bpf_mptcp_common_kfunc_ids) -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> The mptcp_userspace_pm_addr bpf_iter should be invoked under holding the msk pm lock, so spin_lock_bh() and spin_unlock_bh() are needed in BPF program to hold or release the msk pm lock. And bpf_ipv6_addr_v4mapped() helper is also needed by the mptcp_userspace_pm_addr bpf_iter selftest program. This patch adds the corresponding BPF wrappers for these helpers and adds them to the mptcp common kfunc_set. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- net/mptcp/bpf.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) diff --git a/net/mptcp/bpf.c b/net/mptcp/bpf.c index XXXXXXX..XXXXXXX 100644 --- a/net/mptcp/bpf.c +++ b/net/mptcp/bpf.c @@ -XXX,XX +XXX,XX @@ __bpf_kfunc static void bpf_mptcp_sock_release(struct mptcp_sock *msk) WARN_ON_ONCE(!sk || !refcount_dec_not_one(&sk->sk_refcnt)); } +__bpf_kfunc static void bpf_spin_lock_bh(spinlock_t *lock) +{ + spin_lock_bh(lock); +} + +__bpf_kfunc static void bpf_spin_unlock_bh(spinlock_t *lock) +{ + spin_unlock_bh(lock); +} + +__bpf_kfunc static bool bpf_ipv6_addr_v4mapped(const struct mptcp_addr_info *a) +{ +#if IS_ENABLED(CONFIG_MPTCP_IPV6) + return ipv6_addr_v4mapped(&a->addr6); +#endif + return false; +} + __bpf_kfunc struct mptcp_subflow_context * bpf_mptcp_subflow_ctx_by_pos(const struct mptcp_sched_data *data, unsigned int pos) { @@ -XXX,XX +XXX,XX @@ BTF_ID_FLAGS(func, bpf_iter_mptcp_userspace_pm_addr_next, KF_ITER_NEXT | KF_RET_ BTF_ID_FLAGS(func, bpf_iter_mptcp_userspace_pm_addr_destroy, KF_ITER_DESTROY) BTF_ID_FLAGS(func, bpf_mptcp_sock_acquire, KF_ACQUIRE | KF_RET_NULL) BTF_ID_FLAGS(func, bpf_mptcp_sock_release, KF_RELEASE) +BTF_ID_FLAGS(func, bpf_spin_lock_bh) +BTF_ID_FLAGS(func, bpf_spin_unlock_bh) +BTF_ID_FLAGS(func, bpf_ipv6_addr_v4mapped) BTF_KFUNCS_END(bpf_mptcp_common_kfunc_ids) static const struct btf_kfunc_id_set bpf_mptcp_common_kfunc_set = { -- 2.45.2 From: Geliang Tang <tanggeliang@kylinos.cn> This patch adds a test program for the newly added mptcp_userspace_pm_addr bpf_iter in SEC "cgroup/getsockopt". This test iterates over all address entries on the local address list of userspace PM and check whether each one is an IPv4mapped address. Export mptcp_userspace_pm_addr helpers _new/_next/_destroy into bpf_experimental.h. Use bpf_mptcp_sock_acquire() to acquire the msk, then lock the msk pm lock and use bpf_for_each(mptcp_userspace_pm_addr) to walk the local address list of this msk. Invoke bpf_ipv6_addr_v4mapped() in the loop to check whether the address is an IPv4mapped one. Then Add the address ID of each entry to local veriable local_ids. Out of the loop, unlock the msk pm lock and use bpf_mptcp_sock_release() to release the msk. Finally, assign local_ids to global variable ids so that the application can obtain this value. Add a subtest named test_iters_address to load and verify the newly added mptcp_userspace_pm_addr type bpf_iter example in test_mptcp. Since mptcp_userspace_pm_addr bpf_iter iterates over all address entries on the local address list of userspace PM. A set of userspace PM helpers are added. Set pm_type and start "pm_nl_ctl events" command to save pm events into the log file /tmp/bpf_userspace_pm_events in userspace_pm_init(). Kill "pm_nl_ctl" command and remove the log file in userspace_pm_cleanup(). Parse the log file in userspace_pm_add_subflow() to get the token, sport and dport values, then use "pm_nl_ctl csf" command to create a subflow. Use the helper userspace_pm_add_subflow() to add 3 new subflow endpoints. Send a byte of message to start the mptcp connection, and receive the message. getsockopt() is invoked to trigger the "cgroup/getsockopt" test program. Check if skel->bss->ids equals 60 to verify whether the program loops three times as expected. Signed-off-by: Geliang Tang <tanggeliang@kylinos.cn> --- .../testing/selftests/bpf/bpf_experimental.h | 8 + .../testing/selftests/bpf/prog_tests/mptcp.c | 214 ++++++++++++++++++ tools/testing/selftests/bpf/progs/mptcp_bpf.h | 5 + .../selftests/bpf/progs/mptcp_bpf_iters.c | 39 ++++ 4 files changed, 266 insertions(+) diff --git a/tools/testing/selftests/bpf/bpf_experimental.h b/tools/testing/selftests/bpf/bpf_experimental.h index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/bpf_experimental.h +++ b/tools/testing/selftests/bpf/bpf_experimental.h @@ -XXX,XX +XXX,XX @@ bpf_iter_mptcp_subflow_next(struct bpf_iter_mptcp_subflow *it) __weak __ksym; extern void bpf_iter_mptcp_subflow_destroy(struct bpf_iter_mptcp_subflow *it) __weak __ksym; +struct bpf_iter_mptcp_userspace_pm_addr; +extern int bpf_iter_mptcp_userspace_pm_addr_new(struct bpf_iter_mptcp_userspace_pm_addr *it, + struct mptcp_sock *msk) __weak __ksym; +extern struct mptcp_pm_addr_entry * +bpf_iter_mptcp_userspace_pm_addr_next(struct bpf_iter_mptcp_userspace_pm_addr *it) __weak __ksym; +extern void +bpf_iter_mptcp_userspace_pm_addr_destroy(struct bpf_iter_mptcp_userspace_pm_addr *it) __weak __ksym; + extern int bpf_wq_init(struct bpf_wq *wq, void *p__map, unsigned int flags) __weak __ksym; extern int bpf_wq_start(struct bpf_wq *wq, unsigned int flags) __weak __ksym; extern int bpf_wq_set_callback_impl(struct bpf_wq *wq, diff --git a/tools/testing/selftests/bpf/prog_tests/mptcp.c b/tools/testing/selftests/bpf/prog_tests/mptcp.c index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/prog_tests/mptcp.c +++ b/tools/testing/selftests/bpf/prog_tests/mptcp.c @@ -XXX,XX +XXX,XX @@ #define ADDR6_4 "dead:beef:4::1" #define PORT_1 10001 #define PM_CTL "./mptcp_pm_nl_ctl" +#define PM_EVENTS "/tmp/bpf_userspace_pm_events" #define WITH_DATA true #define WITHOUT_DATA false @@ -XXX,XX +XXX,XX @@ #endif #define MPTCP_SCHED_NAME_MAX 16 +enum mptcp_pm_type { + MPTCP_PM_TYPE_KERNEL = 0, + MPTCP_PM_TYPE_USERSPACE, + + __MPTCP_PM_TYPE_NR, + __MPTCP_PM_TYPE_MAX = __MPTCP_PM_TYPE_NR - 1, +}; + static const unsigned int total_bytes = 10 * 1024 * 1024; static int duration; @@ -XXX,XX +XXX,XX @@ static void send_byte(int fd) ASSERT_EQ(write(fd, &b, sizeof(b)), 1, "send single byte"); } +static int recv_byte(int fd) +{ + char buf[1]; + ssize_t n; + + n = recv(fd, buf, sizeof(buf), 0); + if (CHECK(n <= 0, "recv_byte", "recv")) { + log_err("failed/partial recv"); + return -1; + } + return 0; +} + static int verify_mptcpify(int server_fd, int client_fd) { struct __mptcp_info info; @@ -XXX,XX +XXX,XX @@ static void test_iters_subflow(void) close(cgroup_fd); } +static int userspace_pm_init(enum mptcp_pm_type pm_type) +{ + if (address_init()) + goto fail; + + SYS(fail, "ip netns exec %s sysctl -qw net.mptcp.pm_type=%u", + NS_TEST, pm_type); + SYS(fail, "ip netns exec %s %s limits 4 4", + NS_TEST, PM_CTL); + SYS(fail, "ip netns exec %s %s events >> %s 2>&1 &", + NS_TEST, PM_CTL, PM_EVENTS); + + return 0; +fail: + return -1; +} + +static void userspace_pm_cleanup(void) +{ + SYS_NOFAIL("ip netns exec %s killall %s > /dev/null 2>&1", + NS_TEST, PM_CTL); + SYS_NOFAIL("ip netns exec %s rm -rf %s", NS_TEST, PM_EVENTS); +} + +static int userspace_pm_get_events_line(char *type, char *line) +{ + char buf[BUFSIZ], *str; + size_t len; + int fd; + + fd = open(PM_EVENTS, O_RDONLY); + if (fd < 0) { + log_err("failed to open pm events\n"); + return -1; + } + + len = read(fd, buf, sizeof(buf)); + close(fd); + if (len <= 0) { + log_err("failed to read pm events\n"); + return -1; + } + + str = strstr(buf, type); + if (!str) { + log_err("failed to get type %s pm event\n", type); + return -1; + } + + strcpy(line, str); + return 0; +} + +static int userspace_pm_get_token(int fd) +{ + char line[1024], *str; + __u32 token; + int i; + + /* Wait max 2 sec for the connection to be established */ + for (i = 0; i < 10; i++) { + usleep(200000); /* 0.2s */ + send_byte(fd); + + sync(); + if (userspace_pm_get_events_line("type:2", line)) + continue; + str = strstr(line, "token"); + if (!str) + continue; + if (sscanf(str, "token:%u,", &token) != 1) + continue; + return token; + } + + return 0; +} + +static int userspace_pm_add_subflow(__u32 token, char *addr, __u8 id) +{ + bool ipv6 = strstr(addr, ":"); + char line[1024], *str; + __u32 sport, dport; + + if (userspace_pm_get_events_line("type:2", line)) + return -1; + + str = strstr(line, "sport"); + if (!str || sscanf(str, "sport:%u,dport:%u,", &sport, &dport) != 2) { + log_err("add_subflow error, str=%s\n", str); + return -1; + } + + str = ipv6 ? (strstr(addr, ".") ? "::ffff:"ADDR_1 : ADDR6_1) : ADDR_1; + SYS_NOFAIL("ip netns exec %s %s csf lip %s lid %u rip %s rport %u token %u", + NS_TEST, PM_CTL, addr, id, str, dport, token); + + return 0; +} + +static void run_iters_address(void) +{ + int server_fd, client_fd, accept_fd; + int is_mptcp, err; + socklen_t len; + __u32 token; + + server_fd = start_mptcp_server(AF_INET6, "::ffff:"ADDR_1, PORT_1, 0); + if (!ASSERT_OK_FD(server_fd, "start_mptcp_server")) + return; + + client_fd = connect_to_fd(server_fd, 0); + if (!ASSERT_OK_FD(client_fd, "connect_to_fd")) + goto close_server; + + accept_fd = accept(server_fd, NULL, NULL); + if (!ASSERT_OK_FD(accept_fd, "accept")) + goto close_client; + + token = userspace_pm_get_token(client_fd); + if (!token) + goto close_client; + recv_byte(accept_fd); + usleep(200000); /* 0.2s */ + + err = userspace_pm_add_subflow(token, "::ffff:"ADDR_2, 10); + err = err ?: userspace_pm_add_subflow(token, "::ffff:"ADDR_3, 20); + err = err ?: userspace_pm_add_subflow(token, "::ffff:"ADDR_4, 30); + if (!ASSERT_OK(err, "userspace_pm_add_subflow")) + goto close_accept; + + send_byte(accept_fd); + recv_byte(client_fd); + + len = sizeof(is_mptcp); + /* mainly to trigger the BPF program */ + err = getsockopt(client_fd, SOL_TCP, TCP_IS_MPTCP, &is_mptcp, &len); + if (ASSERT_OK(err, "getsockopt(client_fd, TCP_IS_MPTCP)")) + ASSERT_EQ(is_mptcp, 1, "is_mptcp"); + +close_accept: + close(accept_fd); +close_client: + close(client_fd); +close_server: + close(server_fd); +} + +static void test_iters_address(void) +{ + struct mptcp_bpf_iters *skel; + struct netns_obj *netns; + int cgroup_fd; + int err; + + cgroup_fd = test__join_cgroup("/iters_address"); + if (!ASSERT_OK_FD(cgroup_fd, "join_cgroup: iters_address")) + return; + + skel = mptcp_bpf_iters__open_and_load(); + if (!ASSERT_OK_PTR(skel, "skel_open_load: iters_address")) + goto close_cgroup; + + skel->links.iters_address = bpf_program__attach_cgroup(skel->progs.iters_address, + cgroup_fd); + if (!ASSERT_OK_PTR(skel->links.iters_address, "attach getsockopt")) + goto skel_destroy; + + netns = netns_new(NS_TEST, true); + if (!ASSERT_OK_PTR(netns, "netns_new")) + goto skel_destroy; + + err = userspace_pm_init(MPTCP_PM_TYPE_USERSPACE); + if (!ASSERT_OK(err, "userspace_pm_init: iters_address")) + goto close_netns; + + run_iters_address(); + + /* 10 + 20 + 30 = 60 */ + ASSERT_EQ(skel->bss->ids, 60, "address ids"); + + userspace_pm_cleanup(); +close_netns: + netns_free(netns); +skel_destroy: + mptcp_bpf_iters__destroy(skel); +close_cgroup: + close(cgroup_fd); +} + static struct netns_obj *sched_init(char *flags, char *sched) { struct netns_obj *netns; @@ -XXX,XX +XXX,XX @@ void test_mptcp(void) test_subflow(); if (test__start_subtest("iters_subflow")) test_iters_subflow(); + if (test__start_subtest("iters_address")) + test_iters_address(); if (test__start_subtest("default")) test_default(); if (test__start_subtest("first")) diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf.h b/tools/testing/selftests/bpf/progs/mptcp_bpf.h index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/progs/mptcp_bpf.h +++ b/tools/testing/selftests/bpf/progs/mptcp_bpf.h @@ -XXX,XX +XXX,XX @@ extern void mptcp_subflow_set_scheduled(struct mptcp_subflow_context *subflow, extern struct mptcp_subflow_context * bpf_mptcp_subflow_ctx_by_pos(const struct mptcp_sched_data *data, unsigned int pos) __ksym; +extern void bpf_spin_lock_bh(spinlock_t *lock) __ksym; +extern void bpf_spin_unlock_bh(spinlock_t *lock) __ksym; + +extern bool bpf_ipv6_addr_v4mapped(const struct mptcp_addr_info *a) __ksym; + #endif diff --git a/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c b/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c index XXXXXXX..XXXXXXX 100644 --- a/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c +++ b/tools/testing/selftests/bpf/progs/mptcp_bpf_iters.c @@ -XXX,XX +XXX,XX @@ int iters_subflow(struct bpf_sockopt *ctx) bpf_mptcp_sock_release(msk); return 1; } + +SEC("cgroup/getsockopt") +int iters_address(struct bpf_sockopt *ctx) +{ + struct mptcp_pm_addr_entry *entry; + struct bpf_sock *sk = ctx->sk; + struct mptcp_sock *msk; + int local_ids = 0; + + if (!sk || sk->protocol != IPPROTO_MPTCP || + ctx->level != SOL_TCP || ctx->optname != TCP_IS_MPTCP) + return 1; + + msk = bpf_mptcp_sk((struct sock *)sk); + if (msk->pm.server_side) + return 1; + + msk = bpf_mptcp_sock_acquire(msk); + if (!msk) + return 1; + bpf_spin_lock_bh(&msk->pm.lock); + bpf_for_each(mptcp_userspace_pm_addr, entry, msk) { + /* Here MPTCP-specific path manager kfunc can be called: + * this test is not doing anything really useful, only to + * verify the iteration works. + */ + + if (!bpf_ipv6_addr_v4mapped(&entry->addr)) + break; + + local_ids += entry->addr.id; + } + bpf_spin_unlock_bh(&msk->pm.lock); + bpf_mptcp_sock_release(msk); + + ids = local_ids; + + return 1; +} -- 2.45.2