Message ID | 20200911143022.414783-1-nicolas.rybowski@tessares.net |
---|---|
State | Superseded |
Headers | show |
Series | [bpf-next,v2,1/5] bpf: expose is_mptcp flag to bpf_tcp_sock | expand |
Previously it was not possible to make a distinction between plain TCP sockets and MPTCP subflow sockets on the BPF_PROG_TYPE_SOCK_OPS hook. This patch series now enables a fine control of subflow sockets. In its current state, it allows to put different sockopt on each subflow from a same MPTCP connection (socket mark, TCP congestion algorithm, ...) using BPF programs. It should also be the basis of exposing MPTCP-specific fields through BPF. v1 -> v2: - add basic mandatory selftests for the new helper and is_mptcp field (Alexei) - rebase on latest bpf-next Nicolas Rybowski (5): bpf: expose is_mptcp flag to bpf_tcp_sock mptcp: attach subflow socket to parent cgroup bpf: add 'bpf_mptcp_sock' structure and helper bpf: selftests: add MPTCP test base bpf: selftests: add bpf_mptcp_sock() verifier tests include/linux/bpf.h | 33 +++++ include/uapi/linux/bpf.h | 15 +++ kernel/bpf/verifier.c | 30 +++++ net/core/filter.c | 13 +- net/mptcp/Makefile | 2 + net/mptcp/bpf.c | 72 +++++++++++ net/mptcp/subflow.c | 27 ++++ scripts/bpf_helpers_doc.py | 2 + tools/include/uapi/linux/bpf.h | 15 +++ tools/testing/selftests/bpf/config | 1 + tools/testing/selftests/bpf/network_helpers.c | 37 +++++- tools/testing/selftests/bpf/network_helpers.h | 3 + .../testing/selftests/bpf/prog_tests/mptcp.c | 119 ++++++++++++++++++ tools/testing/selftests/bpf/progs/mptcp.c | 48 +++++++ tools/testing/selftests/bpf/verifier/sock.c | 63 ++++++++++ 15 files changed, 474 insertions(+), 6 deletions(-) create mode 100644 net/mptcp/bpf.c create mode 100644 tools/testing/selftests/bpf/prog_tests/mptcp.c create mode 100644 tools/testing/selftests/bpf/progs/mptcp.c
On Fri, Sep 11, 2020 at 9:46 AM Nicolas Rybowski <nicolas.rybowski@tessares.net> wrote: > > This patch adds verifier side tests for the new bpf_mptcp_sock() helper. > > Here are the new tests : > - NULL bpf_sock is correctly handled > - We cannot access a field from bpf_mptcp_sock if the latter is NULL > - We can access a field from bpf_mptcp_sock if the latter is not NULL > - We cannot modify a field from bpf_mptcp_sock. > > Note that "token" is currently the only field in bpf_mptcp_sock. > > Currently, there is no easy way to test the token field since we cannot > get back the mptcp_sock in userspace, this could be a future amelioration. > > Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net> > Signed-off-by: Nicolas Rybowski <nicolas.rybowski@tessares.net> Acked-by: Song Liu <songliubraving@fb.com>
On Fri, Sep 11, 2020 at 8:07 AM Nicolas Rybowski <nicolas.rybowski@tessares.net> wrote: > > is_mptcp is a field from struct tcp_sock used to indicate that the > current tcp_sock is part of the MPTCP protocol. > > In this protocol, a first socket (mptcp_sock) is created with > sk_protocol set to IPPROTO_MPTCP (=262) for control purpose but it > isn't directly on the wire. This is the role of the subflow (kernel) > sockets which are classical tcp_sock with sk_protocol set to > IPPROTO_TCP. The only way to differentiate such sockets from plain TCP > sockets is the is_mptcp field from tcp_sock. > > Such an exposure in BPF is thus required to be able to differentiate > plain TCP sockets from MPTCP subflow sockets in BPF_PROG_TYPE_SOCK_OPS > programs. > > The choice has been made to silently pass the case when CONFIG_MPTCP is > unset by defaulting is_mptcp to 0 in order to make BPF independent of > the MPTCP configuration. Another solution is to make the verifier fail > in 'bpf_tcp_sock_is_valid_ctx_access' but this will add an additional > '#ifdef CONFIG_MPTCP' in the BPF code and a same injected BPF program > will not run if MPTCP is not set. > > An example use-case is provided in > https://github.com/multipath-tcp/mptcp_net-next/tree/scripts/bpf/examples > > Suggested-by: Matthieu Baerts <matthieu.baerts@tessares.net> > Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net> > Acked-by: Mat Martineau <mathew.j.martineau@linux.intel.com> > Signed-off-by: Nicolas Rybowski <nicolas.rybowski@tessares.net> > --- > include/uapi/linux/bpf.h | 1 + > net/core/filter.c | 9 ++++++++- > tools/include/uapi/linux/bpf.h | 1 + > 3 files changed, 10 insertions(+), 1 deletion(-) > > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h > index 7dd314176df7..7d179eada1c3 100644 > --- a/include/uapi/linux/bpf.h > +++ b/include/uapi/linux/bpf.h > @@ -4060,6 +4060,7 @@ struct bpf_tcp_sock { > __u32 delivered; /* Total data packets delivered incl. rexmits */ > __u32 delivered_ce; /* Like the above but only ECE marked packets */ > __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ > + __u32 is_mptcp; /* Is MPTCP subflow? */ Shall we have an __u32 flags, and make is_mptcp a bit of it? Thanks, Song [...]
On Fri, Sep 11, 2020 at 8:15 AM Nicolas Rybowski <nicolas.rybowski@tessares.net> wrote: > > In order to precisely identify the parent MPTCP connection of a subflow, > it is required to access the mptcp_sock's token which uniquely identify a > MPTCP connection. > > This patch adds a new structure 'bpf_mptcp_sock' exposing the 'token' field > of the 'mptcp_sock' extracted from a subflow's 'tcp_sock'. It also adds the > declaration of a new BPF helper of the same name to expose the newly > defined structure in the userspace BPF API. > > This is the foundation to expose more MPTCP-specific fields through BPF. > > Currently, it is limited to the field 'token' of the msk but it is > easily extensible. > > Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net> > Acked-by: Mat Martineau <mathew.j.martineau@linux.intel.com> > Signed-off-by: Nicolas Rybowski <nicolas.rybowski@tessares.net> Acked-by: Song Liu <songliubraving@fb.com>
On Mon, Sep 14, 2020 at 11:21 AM Song Liu <song@kernel.org> wrote: > > On Fri, Sep 11, 2020 at 8:07 AM Nicolas Rybowski > <nicolas.rybowski@tessares.net> wrote: > > > > is_mptcp is a field from struct tcp_sock used to indicate that the > > current tcp_sock is part of the MPTCP protocol. > > > > In this protocol, a first socket (mptcp_sock) is created with > > sk_protocol set to IPPROTO_MPTCP (=262) for control purpose but it > > isn't directly on the wire. This is the role of the subflow (kernel) > > sockets which are classical tcp_sock with sk_protocol set to > > IPPROTO_TCP. The only way to differentiate such sockets from plain TCP > > sockets is the is_mptcp field from tcp_sock. > > > > Such an exposure in BPF is thus required to be able to differentiate > > plain TCP sockets from MPTCP subflow sockets in BPF_PROG_TYPE_SOCK_OPS > > programs. > > > > The choice has been made to silently pass the case when CONFIG_MPTCP is > > unset by defaulting is_mptcp to 0 in order to make BPF independent of > > the MPTCP configuration. Another solution is to make the verifier fail > > in 'bpf_tcp_sock_is_valid_ctx_access' but this will add an additional > > '#ifdef CONFIG_MPTCP' in the BPF code and a same injected BPF program > > will not run if MPTCP is not set. > > > > An example use-case is provided in > > https://github.com/multipath-tcp/mptcp_net-next/tree/scripts/bpf/examples > > > > Suggested-by: Matthieu Baerts <matthieu.baerts@tessares.net> > > Acked-by: Matthieu Baerts <matthieu.baerts@tessares.net> > > Acked-by: Mat Martineau <mathew.j.martineau@linux.intel.com> > > Signed-off-by: Nicolas Rybowski <nicolas.rybowski@tessares.net> > > --- > > include/uapi/linux/bpf.h | 1 + > > net/core/filter.c | 9 ++++++++- > > tools/include/uapi/linux/bpf.h | 1 + > > 3 files changed, 10 insertions(+), 1 deletion(-) > > > > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h > > index 7dd314176df7..7d179eada1c3 100644 > > --- a/include/uapi/linux/bpf.h > > +++ b/include/uapi/linux/bpf.h > > @@ -4060,6 +4060,7 @@ struct bpf_tcp_sock { > > __u32 delivered; /* Total data packets delivered incl. rexmits */ > > __u32 delivered_ce; /* Like the above but only ECE marked packets */ > > __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ > > + __u32 is_mptcp; /* Is MPTCP subflow? */ > > Shall we have an __u32 flags, and make is_mptcp a bit of it? > Bitfields are slow and more annoying to rewrite in verifier, so having an __u32 field is actually good. > Thanks, > Song > [...]
diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h index 7dd314176df7..7d179eada1c3 100644 --- a/include/uapi/linux/bpf.h +++ b/include/uapi/linux/bpf.h @@ -4060,6 +4060,7 @@ struct bpf_tcp_sock { __u32 delivered; /* Total data packets delivered incl. rexmits */ __u32 delivered_ce; /* Like the above but only ECE marked packets */ __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ + __u32 is_mptcp; /* Is MPTCP subflow? */ }; struct bpf_sock_tuple { diff --git a/net/core/filter.c b/net/core/filter.c index d266c6941967..dab48528dceb 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -5837,7 +5837,7 @@ bool bpf_tcp_sock_is_valid_access(int off, int size, enum bpf_access_type type, struct bpf_insn_access_aux *info) { if (off < 0 || off >= offsetofend(struct bpf_tcp_sock, - icsk_retransmits)) + is_mptcp)) return false; if (off % size != 0) @@ -5971,6 +5971,13 @@ u32 bpf_tcp_sock_convert_ctx_access(enum bpf_access_type type, case offsetof(struct bpf_tcp_sock, icsk_retransmits): BPF_INET_SOCK_GET_COMMON(icsk_retransmits); break; + case offsetof(struct bpf_tcp_sock, is_mptcp): +#ifdef CONFIG_MPTCP + BPF_TCP_SOCK_GET_COMMON(is_mptcp); +#else + *insn++ = BPF_MOV32_IMM(si->dst_reg, 0); +#endif + break; } return insn - insn_buf; diff --git a/tools/include/uapi/linux/bpf.h b/tools/include/uapi/linux/bpf.h index 7dd314176df7..7d179eada1c3 100644 --- a/tools/include/uapi/linux/bpf.h +++ b/tools/include/uapi/linux/bpf.h @@ -4060,6 +4060,7 @@ struct bpf_tcp_sock { __u32 delivered; /* Total data packets delivered incl. rexmits */ __u32 delivered_ce; /* Like the above but only ECE marked packets */ __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ + __u32 is_mptcp; /* Is MPTCP subflow? */ }; struct bpf_sock_tuple {