From: Paolo Abeni pabeni@redhat.com

That will avoid an unneeded conditional in both the fast-path and in the fallback case and will simplify a bit the next patch.

Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- net/mptcp/options.c | 6 ++++++ net/mptcp/protocol.c | 6 ------ 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/net/mptcp/options.c b/net/mptcp/options.c index 8a8083207be4..4bdcd2b326bd 100644 --- a/net/mptcp/options.c +++ b/net/mptcp/options.c @@ -1119,6 +1119,12 @@ bool mptcp_incoming_options(struct sock *sk, struct sk_buff *skb) mptcp_data_lock(subflow->conn); if (sk_stream_memory_free(sk)) __mptcp_check_push(subflow->conn, sk); + + /* on fallback we just need to ignore the msk-level snd_una, as + * this is really plain TCP + */ + msk->snd_una = READ_ONCE(msk->snd_nxt); + __mptcp_data_acked(subflow->conn); mptcp_data_unlock(subflow->conn); return true; diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 992b89c75631..9c756d675d4d 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -1004,12 +1004,6 @@ static void __mptcp_clean_una(struct sock *sk) struct mptcp_data_frag *dtmp, *dfrag; u64 snd_una;

- /* on fallback we just need to ignore snd_una, as this is really - * plain TCP - */ - if (__mptcp_check_fallback(msk)) - msk->snd_una = READ_ONCE(msk->snd_nxt); - snd_una = msk->snd_una; list_for_each_entry_safe(dfrag, dtmp, &msk->rtx_queue, list) { if (after64(dfrag->data_seq + dfrag->data_len, snd_una))

From: Paolo Abeni pabeni@redhat.com

Update the existing sockopt test-case to do some basic checks on the newly added counters.

Link: https://github.com/multipath-tcp/mptcp_net-next/issues/385 Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- tools/testing/selftests/net/mptcp/mptcp_sockopt.c | 27 ++++++++++++++++++++++- 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/tools/testing/selftests/net/mptcp/mptcp_sockopt.c b/tools/testing/selftests/net/mptcp/mptcp_sockopt.c index b35148edbf02..5ee710b30f10 100644 --- a/tools/testing/selftests/net/mptcp/mptcp_sockopt.c +++ b/tools/testing/selftests/net/mptcp/mptcp_sockopt.c @@ -51,6 +51,11 @@ struct mptcp_info { __u8 mptcpi_local_addr_used; __u8 mptcpi_local_addr_max; __u8 mptcpi_csum_enabled; + __u32 mptcpi_retransmits; + __u64 mptcpi_bytes_retrans; + __u64 mptcpi_bytes_sent; + __u64 mptcpi_bytes_received; + __u64 mptcpi_bytes_acked; };

struct mptcp_subflow_data { @@ -83,8 +88,10 @@ struct mptcp_subflow_addrs {

struct so_state { struct mptcp_info mi; + struct mptcp_info last_sample; uint64_t mptcpi_rcv_delta; uint64_t tcpi_rcv_delta; + bool pkt_stats_avail; };

#ifndef MIN @@ -322,8 +329,9 @@ static void do_getsockopt_mptcp_info(struct so_state *s, int fd, size_t w) if (ret < 0) die_perror("getsockopt MPTCP_INFO");

- assert(olen == sizeof(i)); + s->pkt_stats_avail = olen >= sizeof(i);

+ s->last_sample = i; if (s->mi.mptcpi_write_seq == 0) s->mi = i;

@@ -562,6 +570,23 @@ static void process_one_client(int fd, int pipefd) do_getsockopts(&s, fd, ret, ret2); if (s.mptcpi_rcv_delta != (uint64_t)ret + 1) xerror("mptcpi_rcv_delta %" PRIu64 ", expect %" PRIu64, s.mptcpi_rcv_delta, ret + 1, s.mptcpi_rcv_delta - ret); + + /* be nice when running on top of older kernel */ + if (s.pkt_stats_avail) { + if (s.last_sample.mptcpi_bytes_sent != ret2) + xerror("mptcpi_bytes_sent %" PRIu64 ", expect %" PRIu64, + s.last_sample.mptcpi_bytes_sent, ret2, + s.last_sample.mptcpi_bytes_sent - ret2); + if (s.last_sample.mptcpi_bytes_received != ret) + xerror("mptcpi_bytes_received %" PRIu64 ", expect %" PRIu64, + s.last_sample.mptcpi_bytes_received, ret, + s.last_sample.mptcpi_bytes_received - ret); + if (s.last_sample.mptcpi_bytes_acked != ret) + xerror("mptcpi_bytes_acked %" PRIu64 ", expect %" PRIu64, + s.last_sample.mptcpi_bytes_acked, ret2, + s.last_sample.mptcpi_bytes_acked - ret2); + } + close(fd); }

From: Paolo Abeni pabeni@redhat.com

Some user-space applications want to monitor the subflows utilization.

Dumping the per subflow tcp_info is not enough, as the PM could close and re-create the subflows under-the-hood, fooling the accounting. Even checking the src/dst addresses used by each subflow could not be enough, because new subflows could re-use the same address/port of the just closed one.

This patch introduces a new socket option, allow dumping all the relevant information all-at-once (everything, everywhere...), in a consistent manner.

Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/388 Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- include/uapi/linux/mptcp.h | 24 +++++++++ net/mptcp/sockopt.c | 127 ++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 149 insertions(+), 2 deletions(-)

diff --git a/include/uapi/linux/mptcp.h b/include/uapi/linux/mptcp.h index a124be6ebbba..ee9c49f949a2 100644 --- a/include/uapi/linux/mptcp.h +++ b/include/uapi/linux/mptcp.h @@ -249,9 +249,33 @@ struct mptcp_subflow_addrs { }; };

+struct mptcp_subflow_info { + __u32 id; + struct mptcp_subflow_addrs addrs; +}; + +struct mptcp_full_info { + __u32 size_tcpinfo_kernel; /* must be 0, set by kernel */ + __u32 size_tcpinfo_user; + __u32 size_sfinfo_kernel; /* must be 0, set by kernel */ + __u32 size_sfinfo_user; + __u32 num_subflows; /* must be 0, set by kernel (real subflow count) */ + __u32 size_arrays_user; /* max subflows that userspace is interested in; + * the buffers at subflow_info/tcp_info + * are respectively at least: + * size_arrays * size_sfinfo_user + * size_arrays * size_tcpinfo_user + * bytes wide + */ + __aligned_u64 subflow_info; + __aligned_u64 tcp_info; + struct mptcp_info mptcp_info; +}; + /* MPTCP socket options */ #define MPTCP_INFO 1 #define MPTCP_TCPINFO 2 #define MPTCP_SUBFLOW_ADDRS 3 +#define MPTCP_FULL_INFO 4

#endif /* _UAPI_MPTCP_H */ diff --git a/net/mptcp/sockopt.c b/net/mptcp/sockopt.c index fa5055d5b029..63f7a09335c5 100644 --- a/net/mptcp/sockopt.c +++ b/net/mptcp/sockopt.c @@ -14,7 +14,8 @@ #include <net/mptcp.h> #include "protocol.h"

-#define MIN_INFO_OPTLEN_SIZE 16 +#define MIN_INFO_OPTLEN_SIZE 16 +#define MIN_FULL_INFO_OPTLEN_SIZE 40

static struct sock *__mptcp_tcp_fallback(struct mptcp_sock *msk) { @@ -981,7 +982,8 @@ static int mptcp_put_subflow_data(struct mptcp_subflow_data *sfd, }

static int mptcp_get_subflow_data(struct mptcp_subflow_data *sfd, - char __user *optval, int __user *optlen) + char __user *optval, + int __user *optlen) { int len, copylen;

@@ -1162,6 +1164,125 @@ static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *o return 0; }

+static int mptcp_get_full_info(struct mptcp_full_info *mfi, + char __user *optval, + int __user *optlen) +{ + int len; + + BUILD_BUG_ON(offsetof(struct mptcp_full_info, mptcp_info) != + MIN_FULL_INFO_OPTLEN_SIZE); + + if (get_user(len, optlen)) + return -EFAULT; + + if (len < MIN_FULL_INFO_OPTLEN_SIZE) + return -EINVAL; + + memset(mfi, 0, sizeof(*mfi)); + if (copy_from_user(mfi, optval, MIN_FULL_INFO_OPTLEN_SIZE)) + return -EFAULT; + + if (mfi->size_tcpinfo_kernel || + mfi->size_sfinfo_kernel || + mfi->num_subflows) + return -EINVAL; + + if (mfi->size_sfinfo_user > INT_MAX || + mfi->size_tcpinfo_user > INT_MAX) + return -EINVAL; + + return len - MIN_FULL_INFO_OPTLEN_SIZE; +} + +static int mptcp_put_full_info(struct mptcp_full_info *mfi, + char __user *optval, + u32 copylen, + int __user *optlen) +{ + copylen += MIN_FULL_INFO_OPTLEN_SIZE; + if (put_user(copylen, optlen)) + return -EFAULT; + + if (copy_to_user(optval, mfi, copylen)) + return -EFAULT; + return 0; +} + +static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optval, + int __user *optlen) +{ + unsigned int sfcount = 0, copylen = 0; + struct mptcp_subflow_context *subflow; + struct sock *sk = (struct sock *)msk; + void __user *tcpinfoptr, *sfinfoptr; + struct mptcp_full_info mfi; + int len; + + len = mptcp_get_full_info(&mfi, optval, optlen); + if (len < 0) + return len; + + /* don't bother filling the mptcp info if there is not enough + * user-space-provided storage + */ + if (len > 0) { + mptcp_diag_fill_info(msk, &mfi.mptcp_info); + copylen += min_t(unsigned int, len, sizeof(struct mptcp_info)); + } + + mfi.size_tcpinfo_kernel = sizeof(struct tcp_info); + mfi.size_tcpinfo_user = min_t(unsigned int, mfi.size_tcpinfo_user, + sizeof(struct tcp_info)); + sfinfoptr = u64_to_user_ptr(mfi.subflow_info); + mfi.size_sfinfo_kernel = sizeof(struct mptcp_subflow_info); + mfi.size_sfinfo_user = min_t(unsigned int, mfi.size_sfinfo_user, + sizeof(struct mptcp_subflow_info)); + tcpinfoptr = u64_to_user_ptr(mfi.tcp_info); + + lock_sock(sk); + mptcp_for_each_subflow(msk, subflow) { + struct sock *ssk = mptcp_subflow_tcp_sock(subflow); + struct mptcp_subflow_info sfinfo; + struct tcp_info tcp_info; + + if (sfcount++ >= mfi.size_arrays_user) + continue; + + /* fetch addr/tcp_info only if the user space buffers + * are wide enough + */ + memset(&sfinfo, 0, sizeof(sfinfo)); + sfinfo.id = subflow->subflow_id; + if (mfi.size_sfinfo_user > + offsetof(struct mptcp_subflow_info, addrs)) + mptcp_get_sub_addrs(ssk, &sfinfo.addrs); + if (copy_to_user(sfinfoptr, &sfinfo, mfi.size_sfinfo_user)) + goto fail_release; + + if (mfi.size_tcpinfo_user) { + tcp_get_info(ssk, &tcp_info); + if (copy_to_user(tcpinfoptr, &tcp_info, + mfi.size_tcpinfo_user)) + goto fail_release; + } + + tcpinfoptr += mfi.size_tcpinfo_user; + sfinfoptr += mfi.size_sfinfo_user; + } + release_sock(sk); + + mfi.num_subflows = sfcount; + if (mptcp_put_full_info(&mfi, optval, copylen, optlen)) + return -EFAULT; + + return 0; + +fail_release: + release_sock(sk); + return -EFAULT; +} + static int mptcp_put_int_option(struct mptcp_sock *msk, char __user *optval, int __user *optlen, int val) { @@ -1235,6 +1356,8 @@ static int mptcp_getsockopt_sol_mptcp(struct mptcp_sock *msk, int optname, switch (optname) { case MPTCP_INFO: return mptcp_getsockopt_info(msk, optval, optlen); + case MPTCP_FULL_INFO: + return mptcp_getsockopt_full_info(msk, optval, optlen); case MPTCP_TCPINFO: return mptcp_getsockopt_tcpinfo(msk, optval, optlen); case MPTCP_SUBFLOW_ADDRS:

From: Paolo Abeni pabeni@redhat.com

Add a testcase explicitly triggering the newly introduce MPTCP_FULL_INFO getsockopt.

Link: https://github.com/multipath-tcp/mptcp_net-next/issues/388 Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Co-developed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- tools/testing/selftests/net/mptcp/mptcp_sockopt.c | 93 ++++++++++++++++++++++- 1 file changed, 91 insertions(+), 2 deletions(-)

diff --git a/tools/testing/selftests/net/mptcp/mptcp_sockopt.c b/tools/testing/selftests/net/mptcp/mptcp_sockopt.c index 5ee710b30f10..926b0be87c99 100644 --- a/tools/testing/selftests/net/mptcp/mptcp_sockopt.c +++ b/tools/testing/selftests/net/mptcp/mptcp_sockopt.c @@ -86,9 +86,38 @@ struct mptcp_subflow_addrs { #define MPTCP_SUBFLOW_ADDRS 3 #endif

+#ifndef MPTCP_FULL_INFO +struct mptcp_subflow_info { + __u32 id; + struct mptcp_subflow_addrs addrs; +}; + +struct mptcp_full_info { + __u32 size_tcpinfo_kernel; /* must be 0, set by kernel */ + __u32 size_tcpinfo_user; + __u32 size_sfinfo_kernel; /* must be 0, set by kernel */ + __u32 size_sfinfo_user; + __u32 num_subflows; /* must be 0, set by kernel (real subflow count) */ + __u32 size_arrays_user; /* max subflows that userspace is interested in; + * the buffers at subflow_info/tcp_info + * are respectively at least: + * size_arrays * size_sfinfo_user + * size_arrays * size_tcpinfo_user + * bytes wide + */ + __aligned_u64 subflow_info; + __aligned_u64 tcp_info; + struct mptcp_info mptcp_info; +}; + +#define MPTCP_FULL_INFO 4 +#endif + struct so_state { struct mptcp_info mi; struct mptcp_info last_sample; + struct tcp_info tcp_info; + struct mptcp_subflow_addrs addrs; uint64_t mptcpi_rcv_delta; uint64_t tcpi_rcv_delta; bool pkt_stats_avail; @@ -370,6 +399,8 @@ static void do_getsockopt_tcp_info(struct so_state *s, int fd, size_t r, size_t olen -= sizeof(struct mptcp_subflow_data); assert(olen == ti.d.size_user);

+ s->tcp_info = ti.ti[0]; + if (ti.ti[0].tcpi_bytes_sent == w && ti.ti[0].tcpi_bytes_received == r) goto done; @@ -391,7 +422,7 @@ static void do_getsockopt_tcp_info(struct so_state *s, int fd, size_t r, size_t do_getsockopt_bogus_sf_data(fd, MPTCP_TCPINFO); }

-static void do_getsockopt_subflow_addrs(int fd) +static void do_getsockopt_subflow_addrs(struct so_state *s, int fd) { struct sockaddr_storage remote, local; socklen_t olen, rlen, llen; @@ -439,6 +470,7 @@ static void do_getsockopt_subflow_addrs(int fd)

assert(memcmp(&local, &addrs.addr[0].ss_local, sizeof(local)) == 0); assert(memcmp(&remote, &addrs.addr[0].ss_remote, sizeof(remote)) == 0); + s->addrs = addrs.addr[0];

memset(&addrs, 0, sizeof(addrs));

@@ -459,13 +491,70 @@ static void do_getsockopt_subflow_addrs(int fd) do_getsockopt_bogus_sf_data(fd, MPTCP_SUBFLOW_ADDRS); }

+static void do_getsockopt_mptcp_full_info(struct so_state *s, int fd) +{ + size_t data_size = sizeof(struct mptcp_full_info); + struct mptcp_subflow_info sfinfo[2]; + struct tcp_info tcp_info[2]; + struct mptcp_full_info mfi; + socklen_t olen; + int ret; + + memset(&mfi, 0, data_size); + memset(tcp_info, 0, sizeof(tcp_info)); + memset(sfinfo, 0, sizeof(sfinfo)); + + mfi.size_tcpinfo_user = sizeof(struct tcp_info); + mfi.size_sfinfo_user = sizeof(struct mptcp_subflow_info); + mfi.size_arrays_user = 2; + mfi.subflow_info = (unsigned long)&sfinfo[0]; + mfi.tcp_info = (unsigned long)&tcp_info[0]; + olen = data_size; + + ret = getsockopt(fd, SOL_MPTCP, MPTCP_FULL_INFO, &mfi, &olen); + if (ret < 0) { + if (errno == EOPNOTSUPP) { + perror("MPTCP_FULL_INFO test skipped"); + return; + } + xerror("getsockopt MPTCP_FULL_INFO"); + } + + assert(olen <= data_size); + assert(mfi.size_tcpinfo_kernel > 0); + assert(mfi.size_tcpinfo_user == + MIN(mfi.size_tcpinfo_kernel, sizeof(struct tcp_info))); + assert(mfi.size_sfinfo_kernel > 0); + assert(mfi.size_sfinfo_user == + MIN(mfi.size_sfinfo_kernel, sizeof(struct mptcp_subflow_info))); + assert(mfi.num_subflows == 1); + + /* Tolerate future extension to mptcp_info struct and running newer + * test on top of older kernel. + * Anyway any kernel supporting MPTCP_FULL_INFO must at least include + * the following in mptcp_info. + */ + assert(olen > (socklen_t)__builtin_offsetof(struct mptcp_full_info, tcp_info)); + assert(mfi.mptcp_info.mptcpi_subflows == 0); + assert(mfi.mptcp_info.mptcpi_bytes_sent == s->last_sample.mptcpi_bytes_sent); + assert(mfi.mptcp_info.mptcpi_bytes_received == s->last_sample.mptcpi_bytes_received); + + assert(sfinfo[0].id == 1); + assert(tcp_info[0].tcpi_bytes_sent == s->tcp_info.tcpi_bytes_sent); + assert(tcp_info[0].tcpi_bytes_received == s->tcp_info.tcpi_bytes_received); + assert(!memcmp(&sfinfo->addrs, &s->addrs, sizeof(struct mptcp_subflow_addrs))); +} + static void do_getsockopts(struct so_state *s, int fd, size_t r, size_t w) { do_getsockopt_mptcp_info(s, fd, w);

do_getsockopt_tcp_info(s, fd, r, w);

- do_getsockopt_subflow_addrs(fd); + do_getsockopt_subflow_addrs(s, fd); + + if (r) + do_getsockopt_mptcp_full_info(s, fd); }

static void connect_one_server(int fd, int pipefd)

From: Paolo Abeni pabeni@redhat.com

The MPTCP code always set the msk state to TCP_CLOSE before calling performing the fast-close. Move such state transition in mptcp_do_fastclose() to avoid some code duplication.

Signed-off-by: Paolo Abeni pabeni@redhat.com Reviewed-by: Matthieu Baerts matthieu.baerts@tessares.net Signed-off-by: Matthieu Baerts matthieu.baerts@tessares.net --- net/mptcp/protocol.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index 4ebd6e9aa949..f65eec3e0d22 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -2655,6 +2655,7 @@ static void mptcp_do_fastclose(struct sock *sk) struct mptcp_subflow_context *subflow, *tmp; struct mptcp_sock *msk = mptcp_sk(sk);

+ inet_sk_state_store(sk, TCP_CLOSE); mptcp_for_each_subflow_safe(msk, subflow, tmp) __mptcp_close_ssk(sk, mptcp_subflow_tcp_sock(subflow), subflow, MPTCP_CF_FASTCLOSE); @@ -2692,10 +2693,9 @@ static void mptcp_worker(struct work_struct *work) * even if it is orphaned and in FIN_WAIT2 state */ if (sock_flag(sk, SOCK_DEAD)) { - if (mptcp_should_close(sk)) { - inet_sk_state_store(sk, TCP_CLOSE); + if (mptcp_should_close(sk)) mptcp_do_fastclose(sk); - } + if (sk->sk_state == TCP_CLOSE) { __mptcp_destroy_sock(sk); goto unlock; @@ -2938,7 +2938,6 @@ static void __mptcp_destroy_sock(struct sock *sk) void __mptcp_unaccepted_force_close(struct sock *sk) { sock_set_flag(sk, SOCK_DEAD); - inet_sk_state_store(sk, TCP_CLOSE); mptcp_do_fastclose(sk); __mptcp_destroy_sock(sk); } @@ -2980,7 +2979,6 @@ bool __mptcp_close(struct sock *sk, long timeout) /* If the msk has read data, or the caller explicitly ask it, * do the MPTCP equivalent of TCP reset, aka MPTCP fastclose */ - inet_sk_state_store(sk, TCP_CLOSE); mptcp_do_fastclose(sk); timeout = 0; } else if (mptcp_close_state(sk)) {