mptcp: add bpf_setsockopt support
Subsystems (7)
v3 cover mptcp: add bpf_setsockopt support
From: Gang Yan <[email protected]>
Hi, Matt, Geliang, Paolo
Patch 3-4 have been reviewed by Paolo before, and ready for merge [1][2].
Changelog:
v3:
- Patch 2 keeps the mptcp_setsockopt_all_sf as Geliang suggested.
v2:
- Patches 1 and 2 are new in this series; they address TCP_MAXSEG handling in the bpf_setsockopt() path.
- Patch 4 adds an early return to fix msk->sk_rcvlowat being unexpectedly modified, an issue seen in v1.
- Patch 5 makes the hook safe for the non-tcp master socket: it guards bpf_sock_ops_cb_flags_set() with sk_is_tcp() to prevent out-of-bounds heap reads/writes through tcp_sk(sk)->bpf_sock_ops_cb_flags, and does not set is_locked_tcp_sock for the msk (unlike tcp_call_bpf()). That flag authorizes the verifier's direct tcp_sock-offset field accesses; since the msk is not a tcp_sock, leaving it at the default 0 is safe.
v1:
[1] https://patchwork.kernel.org/project/mptcp/patch/[email protected]/
[2] https://patchwork.kernel.org/project/mptcp/patch/[email protected]/
Gang Yan (7):
mptcp: drop unused @max arg of __mptcp_setsockopt_set_val
mptcp: take TCP_MAXSEG handling into __mptcp_setsockopt_set_val
mptcp: use sockopt_lock/release_sock in sockopt
mptcp: reject sockopt requiring ssks' lock in BPF context
mptcp: enable bpf_setsockopt on the master socket
mptcp: add TCP_CONNECT_CB sock_ops hook
selftests: bpf: verify mptcp bpf_setsockopt from TCP_CONNECT_CB
include/net/mptcp.h | 9 ++ net/core/filter.c | 10 ++ net/ipv4/tcp.c | 1 + net/mptcp/protocol.c | 6 + net/mptcp/protocol.h | 29 ++++ net/mptcp/sockopt.c | 127 +++++++++++------- .../testing/selftests/bpf/prog_tests/mptcp.c | 61 +++++++++ .../selftests/bpf/progs/mptcp_setsockopt.c | 32 +++++ 8 files changed, 223 insertions(+), 52 deletions(-) create mode 100644 tools/testing/selftests/bpf/progs/mptcp_setsockopt.c
Signature · 1 line
2.43.0
PATCH v3 1/7 mptcp: drop unused @max arg of __mptcp_setsockopt_set_val
From: Gang Yan <[email protected]>
The @max argument is never read in the function body. Remove it and the MAX_TCP_KEEP* values passed by the TCP_KEEPIDLE/INTVL/KEEPCNT callers.
--- net/mptcp/sockopt.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)
net/mptcp/sockopt.c +4 -7
@@ -632,7 +632,7 @@ static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_treturn ret; } -static int __mptcp_setsockopt_set_val (struct mptcp_sock *msk,int max, +static int __mptcp_setsockopt_set_val (struct mptcp_sock *msk,int (*set_val)(struct sock *,int ),int *msk_val,int val) {
@@ -876,18 +876,15 @@ static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname, ret =__mptcp_setsockopt_sol_tcp_nodelay (msk, val);break ;case TCP_KEEPIDLE : - ret =__mptcp_setsockopt_set_val (msk,MAX_TCP_KEEPIDLE , - &tcp_sock_set_keepidle_locked, + ret =__mptcp_setsockopt_set_val (msk, &tcp_sock_set_keepidle_locked, &msk->keepalive_idle, val);break ;case TCP_KEEPINTVL : - ret =__mptcp_setsockopt_set_val (msk,MAX_TCP_KEEPINTVL , - &tcp_sock_set_keepintvl, + ret =__mptcp_setsockopt_set_val (msk, &tcp_sock_set_keepintvl, &msk->keepalive_intvl, val);break ;case TCP_KEEPCNT : - ret =__mptcp_setsockopt_set_val (msk,MAX_TCP_KEEPCNT , - &tcp_sock_set_keepcnt, + ret =__mptcp_setsockopt_set_val (msk, &tcp_sock_set_keepcnt, &msk->keepalive_cnt, val);break ;
Signature · 1 line
2.43.0
Geliang Tang
reviewerRe: [PATCH mptcp-next v3 1/7] mptcp: drop unused @max arg of __mptcp_setsockopt_set_val
Quoted 6 lines · Gang Yan
From: Gang Yan <[email protected]>
The @max argument is never read in the function body. Remove it and the
MAX_TCP_KEEP* values passed by the TCP_KEEPIDLE/INTVL/KEEPCNT callers.
No, we shouldn't drop this "max" argument. Instead, we need to add some code to check the "val" argument against the "max" argument in __mptcp_setsockopt_set_val(), something like:
if (val < 1 || val > max)
return -EINVAL;Quoted 57 lines
Signed-off-by Gang Yan <[email protected]>---
net/mptcp/sockopt.c | 11 ++++-------1 file changed, 4 insertions(+), 7 deletions(-)net/mptcp/sockopt.c +0
-0@@ -632,7 +632,7 @@ static intmptcp_setsockopt_sol_tcp_congestion(struct
mptcp_sock*msk,sockptr_treturn ret;
}
-static int __mptcp_setsockopt_set_val(structmptcp_sock*msk, int max,
+static int __mptcp_setsockopt_set_val(structmptcp_sock*msk, int (*set_val)(struct sock *, int),
int*msk_val, int val) {net/mptcp/sockopt.c +0
-0@@ -876,18 +876,15 @@ static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname, ret = __mptcp_setsockopt_sol_tcp_nodelay(msk, val); break;case
TCP_KEEPIDLE:net/mptcp/sockopt.c +0
-1- ret = __mptcp_setsockopt_set_val(msk,
MAX_TCP_KEEPIDLE,net/mptcp/sockopt.c +0
-1-
&tcp_sock_set_keepidle_locked,net/mptcp/sockopt.c +1
-0+ ret = __mptcp_setsockopt_set_val(msk,
&tcp_sock_set_keepidle_locked,
&msk-Quoted 1 line
keepalive_idle, val);break;
caseTCP_KEEPINTVL:net/mptcp/sockopt.c +0
-1- ret = __mptcp_setsockopt_set_val(msk,
MAX_TCP_KEEPINTVL,net/mptcp/sockopt.c +0
-1-
&tcp_sock_set_keepintvl,net/mptcp/sockopt.c +1
-0+ ret = __mptcp_setsockopt_set_val(msk,
&tcp_sock_set_keepintvl,
&msk-Quoted 1 line
keepalive_intvl, val);break;
caseTCP_KEEPCNT:net/mptcp/sockopt.c +0
-1- ret = __mptcp_setsockopt_set_val(msk,
MAX_TCP_KEEPCNT,net/mptcp/sockopt.c +0
-1-
&tcp_sock_set_keepcnt,net/mptcp/sockopt.c +1
-0+ ret = __mptcp_setsockopt_set_val(msk,
&tcp_sock_set_keepcnt,
&msk-Quoted 1 line
keepalive_cnt,val); break;
Re: [PATCH mptcp-next v3 1/7] mptcp: drop unused @max arg of __mptcp_setsockopt_set_val
Quoted 11 lines · Geliang Tang
Quoted 8 lines · Gang Yan
From: Gang Yan <[email protected]>
The @max argument is never read in the function body. Remove it and the
MAX_TCP_KEEP* values passed by the TCP_KEEPIDLE/INTVL/KEEPCNT callers.
Quoted 7 lines
No, we shouldn't drop this "max" argument. Instead, we need to add some code to check the "val" argument against the "max" argument in
__mptcp_setsockopt_set_val(), something like:if (val < 1 || val > max)
return-EINVAL;
Hi Geliang
Thank you for the review and the suggestion to add a range check inside __mptcp_setsockopt_set_val().
But, I noticed that the underlying TCP helper functions already perform their own parameter validation internally.
Given that these helpers already enforce the same bounds, I would like to ask for your opinion: Is it still necessary to add an extra check in the MPTCP layer?
Thanks
Gang
Quoted 59 lines
Quoted 58 lines
Signed-off-by Gang Yan <[email protected]>---
net/mptcp/sockopt.c | 11 ++++------- 1 file changed, 4 insertions(+), 7 deletions(-)diff --git
a/net/mptcp/sockopt.cb/net/mptcp/sockopt.cindexfcf6feb2a9eb..9c4faa0d82ef100644
---a/net/mptcp/sockopt.c
+++b/net/mptcp/sockopt.c
@@ -632,7 +632,7 @@ static int mptcp_setsockopt_sol_tcp_congestion(structmptcp_sock*msk,sockptr_treturn ret;
}
-static int __mptcp_setsockopt_set_val(structmptcp_sock*msk, int max,
+static int __mptcp_setsockopt_set_val(structmptcp_sock*msk, int (*set_val)(struct sock *,
int),
int*msk_val, int val)
{
@@ -876,18 +876,15 @@ static int mptcp_setsockopt_sol_tcp(structmptcp_sock*msk, int optname,
ret =__mptcp_setsockopt_sol_tcp_nodelay(msk, val); break;
caseTCP_KEEPIDLE:
- ret = __mptcp_setsockopt_set_val(msk,
MAX_TCP_KEEPIDLE,
-
&tcp_sock_set_keepidle_locked,
+ ret = __mptcp_setsockopt_set_val(msk,&tcp_sock_set_keepidle_locked,
&msk-keepalive_idle, val);
break;
caseTCP_KEEPINTVL:
- ret = __mptcp_setsockopt_set_val(msk,
MAX_TCP_KEEPINTVL,
-
&tcp_sock_set_keepintvl,
+ ret = __mptcp_setsockopt_set_val(msk,&tcp_sock_set_keepintvl,
&msk-keepalive_intvl, val);
break;
caseTCP_KEEPCNT:
- ret = __mptcp_setsockopt_set_val(msk,
MAX_TCP_KEEPCNT,
-
&tcp_sock_set_keepcnt,
+ ret = __mptcp_setsockopt_set_val(msk,&tcp_sock_set_keepcnt,
&msk-keepalive_cnt,
val);
break;
PATCH v3 2/7 mptcp: take TCP_MAXSEG handling into __mptcp_setsockopt_set_val
In reply to Gang YanFrom: Gang Yan <[email protected]>
mptcp_setsockopt_all_sf is only used in 'TCP_MAXSEG', and it can be replaced with __mptcp_setsockopt_set_val.
--- net/ipv4/tcp.c | 1 + net/mptcp/sockopt.c | 5 ++--- 2 files changed, 3 insertions(+), 3 deletions(-)
net/ipv4/tcp.c +1 -0
@@ -3830,6 +3830,7 @@ int tcp_sock_set_maxseg(struct sock *sk, int val)WRITE_ONCE (tcp_sk (sk)->rx_opt.user_mss, val);return 0 ; } +EXPORT_SYMBOL (tcp_sock_set_maxseg);/* * Socket option code for TCP.
net/mptcp/sockopt.c +2 -3
@@ -889,9 +889,8 @@ static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname, val);break ;case TCP_MAXSEG : - msk->maxseg = val; - ret =mptcp_setsockopt_all_sf (msk,SOL_TCP , optname, optval, - optlen); + ret =__mptcp_setsockopt_set_val (msk, &tcp_sock_set_maxseg, + &msk->maxseg, val);break ;default : ret = -ENOPROTOOPT ;
Signature · 1 line
2.43.0
Geliang Tang
reviewerRe: [PATCH mptcp-next v3 2/7] mptcp: take TCP_MAXSEG handling into __mptcp_setsockopt_set_val
Quoted 39 lines · Gang Yan
From: Gang Yan <[email protected]>
mptcp_setsockopt_all_sfis only used in 'TCP_MAXSEG', and it can be replaced with__mptcp_setsockopt_set_val.Signed-off-by Gang Yan <[email protected]>---
net/ipv4/tcp.c | 1 +
net/mptcp/sockopt.c | 5 ++---2 files changed, 3 insertions(+), 3 deletions(-)net/ipv4/tcp.c +1
-0@@ -3830,6 +3830,7 @@ int tcp_sock_set_maxseg(struct sock *sk, int val) WRITE_ONCE(tcp_sk(sk)->rx_opt.user_mss, val); return 0; } +EXPORT_SYMBOL (tcp_sock_set_maxseg);
/*
* Socket option code for TCP.net/mptcp/sockopt.c +0
-0@@ -889,9 +889,8 @@ static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname, val); break;case
TCP_MAXSEG:net/mptcp/sockopt.c +0
-2- msk->maxseg = val; - ret = mptcp_setsockopt_all_sf(msk, SOL_TCP, optname,optval,
net/mptcp/sockopt.c +1
-1- optlen); + ret = __mptcp_setsockopt_set_val(msk,
&tcp_sock_set_maxseg,
net/mptcp/sockopt.c static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname, · line 893
Quoted 1 line
net/mptcp/sockopt.c +1
-0+ &msk->maxseg, val);
So, here you do need to pass in MAX_TCP_MAXSEG, and you need to consider what would be a reasonable value to define it as.
Quoted 3 lines
break;
default:
ret =-ENOPROTOOPT;
PATCH v3 3/7 mptcp: use sockopt_lock/release_sock in sockopt
In reply to Gang YanFrom: Gang Yan <[email protected]>
TCP and the core socket layer use sockopt_lock_sock() / sockopt_release_sock() in their setsockopt and getsockopt handlers.
Switch the MPTCP socket (msk) level lock_sock()/release_sock() calls to use the BPF-aware wrappers, making the MPTCP sockopt codepaths consistent with the rest of the networking stack.
--- net/mptcp/sockopt.c | 84 ++++++++++++++++++++++----------------------- 1 file changed, 42 insertions(+), 42 deletions(-)
net/mptcp/sockopt.c +42 -42
@@ -78,7 +78,7 @@ static void mptcp_sol_socket_sync_intval(struct mptcp_sock *msk, int optname, instruct mptcp_subflow_context *subflow;struct sock *sk = (struct sock *)msk; -lock_sock (sk); +sockopt_lock_sock (sk);sockopt_seq_inc (msk);mptcp_for_each_subflow (msk, subflow) {
@@ -123,7 +123,7 @@ static void mptcp_sol_socket_sync_intval(struct mptcp_sock *msk, int optname, inunlock_sock_fast (ssk, slow); } -release_sock (sk); +sockopt_release_sock (sk); }static int mptcp_sol_socket_intval (struct mptcp_sock *msk,int optname,int val)
@@ -162,7 +162,7 @@ static int mptcp_setsockopt_sol_socket_tstamp(struct mptcp_sock *msk, int optnamif (ret)return ret; -lock_sock (sk); +sockopt_lock_sock (sk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);
@@ -171,7 +171,7 @@ static int mptcp_setsockopt_sol_socket_tstamp(struct mptcp_sock *msk, int optnamrelease_sock (ssk); } -release_sock (sk); +sockopt_release_sock (sk);return 0 ; }
@@ -237,7 +237,7 @@ static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk,if (ret)return ret; -lock_sock (sk); +sockopt_lock_sock (sk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);
@@ -251,7 +251,7 @@ static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk, ret = err; } -release_sock (sk); +sockopt_release_sock (sk);return ret; }
@@ -276,7 +276,7 @@ static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_tif (ret)return ret; -lock_sock (sk); +sockopt_lock_sock (sk);sockopt_seq_inc (msk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);
@@ -293,7 +293,7 @@ static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_tunlock_sock_fast (ssk, slow); } -release_sock (sk); +sockopt_release_sock (sk);return 0 ; }
@@ -309,10 +309,10 @@ static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,case SO_REUSEADDR :case SO_BINDTODEVICE :case SO_BINDTOIFINDEX : -lock_sock (sk); +sockopt_lock_sock (sk); ssk =__mptcp_nmpc_sk (msk);if (IS_ERR (ssk)) { -release_sock (sk); +sockopt_release_sock (sk);return PTR_ERR (ssk); }
@@ -327,7 +327,7 @@ static int mptcp_setsockopt_sol_socket(struct mptcp_sock *msk, int optname,else if (optname ==SO_BINDTOIFINDEX ) sk->sk_bound_dev_if = ssk->sk_bound_dev_if; } -release_sock (sk); +sockopt_release_sock (sk);return ret;case SO_KEEPALIVE :case SO_PRIORITY :
@@ -405,16 +405,16 @@ static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,case IPV6_V6ONLY :case IPV6_TRANSPARENT :case IPV6_FREEBIND : -lock_sock (sk); +sockopt_lock_sock (sk); ssk =__mptcp_nmpc_sk (msk);if (IS_ERR (ssk)) { -release_sock (sk); +sockopt_release_sock (sk);return PTR_ERR (ssk); } ret =tcp_setsockopt (ssk,SOL_IPV6 , optname, optval, optlen);if (ret !=0 ) { -release_sock (sk); +sockopt_release_sock (sk);return ret; }
@@ -434,7 +434,7 @@ static int mptcp_setsockopt_v6(struct mptcp_sock *msk, int optname,break ; } -release_sock (sk); +sockopt_release_sock (sk);break ; }
@@ -611,7 +611,7 @@ static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_t cap_net_admin =ns_capable (sock_net (sk)->user_ns,CAP_NET_ADMIN ); ret =0 ; -lock_sock (sk); +sockopt_lock_sock (sk);sockopt_seq_inc (msk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);
@@ -628,7 +628,7 @@ static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_tif (ret ==0 )strscpy (msk->ca_name, name,sizeof (msk->ca_name)); -release_sock (sk); +sockopt_release_sock (sk);return ret; }
@@ -707,11 +707,11 @@ static int mptcp_setsockopt_sol_ip_set(struct mptcp_sock *msk, int optname,if (err !=0 )return err; -lock_sock (sk); +sockopt_lock_sock (sk); ssk =__mptcp_nmpc_sk (msk);if (IS_ERR (ssk)) { -release_sock (sk); +sockopt_release_sock (sk);return PTR_ERR (ssk); }
@@ -732,13 +732,13 @@ static int mptcp_setsockopt_sol_ip_set(struct mptcp_sock *msk, int optname, READ_ONCE(inet_sk(sk)->local_port_range));break ;default : -release_sock (sk); +sockopt_release_sock (sk);WARN_ON_ONCE (1 );return -EOPNOTSUPP ; }sockopt_seq_inc (msk); -release_sock (sk); +sockopt_release_sock (sk);return 0 ; }
@@ -754,7 +754,7 @@ static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,if (err !=0 )return err; -lock_sock (sk); +sockopt_lock_sock (sk);sockopt_seq_inc (msk); val =READ_ONCE (inet_sk (sk)->tos);mptcp_for_each_subflow (msk, subflow) {
@@ -765,7 +765,7 @@ static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,__ip_sock_set_tos (ssk, val);unlock_sock_fast (ssk, slow); } -release_sock (sk); +sockopt_release_sock (sk);return 0 ; }
@@ -794,7 +794,7 @@ static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, intint ret;/* Limit to first subflow, before the connection establishment */ -lock_sock (sk); +sockopt_lock_sock (sk); ssk =__mptcp_nmpc_sk (msk);if (IS_ERR (ssk)) { ret =PTR_ERR (ssk);
@@ -804,7 +804,7 @@ static int mptcp_setsockopt_first_sf_only(struct mptcp_sock *msk, int level, int ret =tcp_setsockopt (ssk, level, optname, optval, optlen); unlock: -release_sock (sk); +sockopt_release_sock (sk);return ret; }
@@ -857,7 +857,7 @@ static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname,if (ret)return ret; -lock_sock (sk); +sockopt_lock_sock (sk);switch (optname) {case TCP_INQ :if (val <0 || val >1 )
@@ -896,7 +896,7 @@ static int mptcp_setsockopt_sol_tcp(struct mptcp_sock *msk, int optname, ret = -ENOPROTOOPT ; } -release_sock (sk); +sockopt_release_sock (sk);return ret; }
@@ -920,9 +920,9 @@ int mptcp_setsockopt(struct sock *sk, int level, int optname,* is in TCP fallback, when TCP socket options are passed through * to the one remaining subflow. */ -lock_sock (sk); +sockopt_lock_sock (sk); ssk =__mptcp_tcp_fallback (msk); -release_sock (sk); +sockopt_release_sock (sk);if (ssk)return tcp_setsockopt (ssk, level, optname, optval, optlen);
@@ -945,7 +945,7 @@ static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, intstruct sock *ssk;int ret; -lock_sock (sk); +sockopt_lock_sock (sk); ssk = msk->first;if (ssk) goto get;
@@ -960,7 +960,7 @@ static int mptcp_getsockopt_first_sf_only(struct mptcp_sock *msk, int level, int ret =tcp_getsockopt (ssk, level, optname, optval, optlen); out: -release_sock (sk); +sockopt_release_sock (sk);return ret; }
@@ -1131,7 +1131,7 @@ static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval, infoptr = optval + sfd.size_subflow_data; -lock_sock (sk); +sockopt_lock_sock (sk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);
@@ -1144,7 +1144,7 @@ static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval,tcp_get_info (ssk, &info);if (copy_to_user (infoptr, &info, sfd.size_user)) { -release_sock (sk); +sockopt_release_sock (sk);return -EFAULT ; }
@@ -1154,7 +1154,7 @@ static int mptcp_getsockopt_tcpinfo(struct mptcp_sock *msk, char __user *optval, } } -release_sock (sk); +sockopt_release_sock (sk); sfd.num_subflows = sfcount;
@@ -1223,7 +1223,7 @@ static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *o addrptr = optval + sfd.size_subflow_data; -lock_sock (sk); +sockopt_lock_sock (sk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);
@@ -1236,7 +1236,7 @@ static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *omptcp_get_sub_addrs (ssk, &a);if (copy_to_user (addrptr, &a, sfd.size_user)) { -release_sock (sk); +sockopt_release_sock (sk);return -EFAULT ; }
@@ -1246,7 +1246,7 @@ static int mptcp_getsockopt_subflow_addrs(struct mptcp_sock *msk, char __user *o } } -release_sock (sk); +sockopt_release_sock (sk); sfd.num_subflows = sfcount;
@@ -1332,7 +1332,7 @@ static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optva sizeof(struct mptcp_subflow_info)); tcpinfoptr =u64_to_user_ptr (mfi.tcp_info); -lock_sock (sk); +sockopt_lock_sock (sk);mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);struct mptcp_subflow_info sfinfo;
@@ -1362,7 +1362,7 @@ static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optva tcpinfoptr += mfi.size_tcpinfo_user; sfinfoptr += mfi.size_sfinfo_user; } -release_sock (sk); +sockopt_release_sock (sk); mfi.num_subflows = sfcount;if (mptcp_put_full_info (&mfi, optval, copylen, optlen))
@@ -1371,7 +1371,7 @@ static int mptcp_getsockopt_full_info(struct mptcp_sock *msk, char __user *optvareturn 0 ; fail_release: -release_sock (sk); +sockopt_release_sock (sk);return -EFAULT ; }
@@ -1526,9 +1526,9 @@ int mptcp_getsockopt(struct sock *sk, int level, int optname,* is in TCP fallback, when socket options are passed through * to the one remaining subflow. */ -lock_sock (sk); +sockopt_lock_sock (sk); ssk =__mptcp_tcp_fallback (msk); -release_sock (sk); +sockopt_release_sock (sk);if (ssk)return tcp_getsockopt (ssk, level, optname, optval, option);
Signature · 1 line
2.43.0
PATCH v3 4/7 mptcp: reject sockopt requiring ssks' lock in BPF context
In reply to Gang YanFrom: Gang Yan <[email protected]>
Several MPTCP setsockopt handlers need to acquire the subflow lock via lock_sock(ssk) to propagate settings to each subflow. This lock can sleep and is therefore not usable in BPF context where sleeping is forbidden.
The short-term solution is to make any sockopt operation that requires subflow-level lock fail with -EOPNOTSUPP when called from BPF context.
--- net/mptcp/sockopt.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+)
net/mptcp/sockopt.c +27 -0
@@ -185,6 +185,9 @@ static int mptcp_setsockopt_sol_socket_int(struct mptcp_sock *msk, int optname,if (ret)return ret; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +switch (optname) {case SO_KEEPALIVE :case SO_DEBUG :
@@ -218,6 +221,9 @@ static int mptcp_setsockopt_sol_socket_timestamping(struct mptcp_sock *msk,struct so_timestamping timestamping;int ret; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +if (optlen ==sizeof (timestamping)) {if (copy_from_sockptr (×tamping, optval,sizeof (timestamping)))
@@ -265,6 +271,9 @@ static int mptcp_setsockopt_sol_socket_linger(struct mptcp_sock *msk, sockptr_tsockptr_t kopt;int ret; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +if (optlen <sizeof (ling))return -EINVAL ;
@@ -598,6 +607,9 @@ static int mptcp_setsockopt_sol_tcp_congestion(struct mptcp_sock *msk, sockptr_tbool cap_net_admin;int ret; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +if (optlen <1 )return -EINVAL ;
@@ -639,6 +651,9 @@ static int __mptcp_setsockopt_set_val(struct mptcp_sock *msk,struct mptcp_subflow_context *subflow;int err =0 ; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +mptcp_for_each_subflow (msk, subflow) {struct sock *ssk =mptcp_subflow_tcp_sock (subflow);int ret;
@@ -662,6 +677,9 @@ static int __mptcp_setsockopt_sol_tcp_cork(struct mptcp_sock *msk, int val)struct mptcp_subflow_context *subflow;struct sock *sk = (struct sock *)msk; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +sockopt_seq_inc (msk); msk->cork = !!val;mptcp_for_each_subflow (msk, subflow) {
@@ -682,6 +700,9 @@ static int __mptcp_setsockopt_sol_tcp_nodelay(struct mptcp_sock *msk, int val)struct mptcp_subflow_context *subflow;struct sock *sk = (struct sock *)msk; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +sockopt_seq_inc (msk); msk->nodelay = !!val;mptcp_for_each_subflow (msk, subflow) {
@@ -749,6 +770,9 @@ static int mptcp_setsockopt_v4_set_tos(struct mptcp_sock *msk, int optname,struct sock *sk = (struct sock *)msk;int err, val; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; + err =ip_setsockopt (sk,SOL_IP , optname, optval, optlen);if (err !=0 )
@@ -1632,6 +1656,9 @@ int mptcp_set_rcvlowat(struct sock *sk, int val)if (sk->sk_protocol ==IPPROTO_TCP )return -EINVAL ; +if (has_current_bpf_ctx ()) +return -EOPNOTSUPP ; +if (sk->sk_userlocks &SOCK_RCVBUF_LOCK ) cap = sk->sk_rcvbuf >>1 ;else
Signature · 1 line
2.43.0
PATCH v3 5/7 mptcp: enable bpf_setsockopt on the master socket
In reply to Gang YanFrom: Gang Yan <[email protected]>
bpf_setsockopt() currently cannot be used on mptcp master sockets:__bpf_setsockopt() dispatches by level to the protocol-agnostic sol_*_sockopt() helpers, which either reject the msk (sk_protocol == IPPROTO_MPTCP) and the ssk (sk_is_tcp() is false) or bypass mptcp's own dispatch (e.g. SOL_IP going straight to do_ip_setsockopt()).
This patch routes any level to mptcp_setsockopt(), which already handles all levels.
--- include/net/mptcp.h | 9 +++++++++ net/core/filter.c | 7 +++++++ 2 files changed, 16 insertions(+)
include/net/mptcp.h +9 -0
@@ -237,6 +237,9 @@ static inline __be32 mptcp_reset_option(const struct sk_buff *skb) }void mptcp_active_detect_blackhole (struct sock *sk,bool expired); + +int mptcp_setsockopt (struct sock *sk,int level,int optname, +sockptr_t optval,unsigned int optlen);#else static inline void mptcp_init (void )
@@ -314,6 +317,12 @@ static inline struct request_sock *mptcp_subflow_reqsk_alloc(const struct requesstatic inline __be32 mptcp_reset_option (const struct sk_buff *skb) {return htonl (0u ); }static inline void mptcp_active_detect_blackhole (struct sock *sk,bool expired) { } + +static inline int mptcp_setsockopt (struct sock *sk,int level,int optname, +sockptr_t optval,unsigned int optlen) +{ +return -EINVAL ; +}#endif /* CONFIG_MPTCP */ #if IS_ENABLED (CONFIG_MPTCP_IPV6 )
net/core/filter.c +7 -0
@@ -5683,6 +5683,13 @@ static int __bpf_setsockopt(struct sock *sk, int level, int optname,if (!sk_fullsock (sk))return -EINVAL ; +/* Route any bpf_setsockopt on the mptcp socket to mptcp_setsockopt, +* which handles all levels. +*/ +if (IS_ENABLED (CONFIG_MPTCP ) && sk->sk_protocol ==IPPROTO_MPTCP ) +return mptcp_setsockopt (sk, level, optname, +KERNEL_SOCKPTR (optval), optlen); +if (level ==SOL_SOCKET )return sol_socket_sockopt (sk, optname, optval, &optlen, false);else if (IS_ENABLED (CONFIG_INET ) && level ==SOL_IP )
Signature · 1 line
2.43.0
PATCH v3 6/7 mptcp: add TCP_CONNECT_CB sock_ops hook
In reply to Gang YanFrom: Gang Yan <[email protected]>
This patch adds a helper named 'mptcp_call_bpf' like tcp_call_bpf. Invoke the new helper from mptcp_connect() with BPF_SOCK_OPS_TCP_CONNECT_CB, placed after the subflow lock is acquired and before tcp_connect(). At this point the msk lock is held by __inet_stream_connect(), mirroring the placement of TCP_CONNECT_CB in tcp_v4_connect()/tcp_v6_connect().
'bpf_sock_ops_cb_flags_set' can be called via msk, so using sk_is_tcp() to avoid this issue.
--- net/core/filter.c | 3 +++ net/mptcp/protocol.c | 6 ++++++ net/mptcp/protocol.h | 29 +++++++++++++++++++++++++++++ 3 files changed, 38 insertions(+)
net/core/filter.c +3 -0
@@ -6097,6 +6097,9 @@ BPF_CALL_2(bpf_sock_ops_cb_flags_set, struct bpf_sock_ops_kern *, bpf_sock,if (!IS_ENABLED (CONFIG_INET ) || !sk_fullsock (sk))return -EINVAL ; +if (!sk_is_tcp (sk)) +return -EOPNOTSUPP ; +tcp_sk (sk)->bpf_sock_ops_cb_flags = val;return argval & (~BPF_SOCK_OPS_ALL_CB_FLAGS );
net/mptcp/protocol.c +6 -0
@@ -4149,6 +4149,12 @@ static int mptcp_connect(struct sock *sk, struct sockaddr_unsized *uaddr,if (!msk->fastopening)lock_sock (ssk); +/* Notify cgroup BPF on the msk before initiating the subflow connect. +* Mirrors BPF_SOCK_OPS_TCP_CONNECT_CB; msk lock is held by the +* caller (__inet_stream_connect) and ssk is held before. +*/ +mptcp_call_bpf (sk,BPF_SOCK_OPS_TCP_CONNECT_CB ,0 ,NULL ); +/* the following mirrors closely a very small chunk of code from * __inet_stream_connect() */
net/mptcp/protocol.h +29 -0
@@ -1374,4 +1374,33 @@ mptcp_token_join_cookie_init_state(struct mptcp_subflow_request_sock *subflow_restatic inline void mptcp_join_cookie_init (void ) {}#endif +#ifdef CONFIG_BPF_JIT +static inline int mptcp_call_bpf (struct sock *sk,int op,u32 nargs,u32 *args) +{ +struct bpf_sock_ops_kern sock_ops; +int ret; + +memset (&sock_ops,0 , offsetof(struct bpf_sock_ops_kern , temp)); + +if (sk_fullsock (sk)) { + sock_ops.is_fullsock =1 ; +sock_owned_by_me (sk); + } + + sock_ops.sk = sk; + sock_ops.op = op; + +if (nargs >0 ) +memcpy (sock_ops.args, args, nargs *sizeof (*args)); + + ret =BPF_CGROUP_RUN_PROG_SOCK_OPS (&sock_ops); +return ret ==0 ? sock_ops.reply :-1 ; +} +#else +static inline int mptcp_call_bpf (struct sock *sk,int op,u32 nargs,u32 *args) +{ +return -1 ; +} +#endif +#endif /* __MPTCP_PROTOCOL_H */
Signature · 1 line
2.43.0
PATCH v3 7/7 selftests: bpf: verify mptcp bpf_setsockopt from TCP_CONNECT_CB
In reply to Gang YanFrom: Gang Yan <[email protected]>
Add a BPF sockops program attached to BPF_CGROUP_SOCK_OPS that fires on BPF_SOCK_OPS_TCP_CONNECT_CB and exercises bpf_setsockopt() on the mptcp master socket (msk).
Two scenarios are covered by the new "setsockopt" subtest:
- TCP_INQ must succeed and the value (1) must be observable from userspace via getsockopt() on the mptcp socket.
- TCP_CONGESTION needs the subflow lock and is therefore rejected in bpf context with -EOPNOTSUPP.
--- .../testing/selftests/bpf/prog_tests/mptcp.c | 61 +++++++++++++++++++ .../selftests/bpf/progs/mptcp_setsockopt.c | 32 ++++++++++ 2 files changed, 93 insertions(+) create mode 100644 tools/testing/selftests/bpf/progs/mptcp_setsockopt.c
tools/testing/selftests/bpf/prog_tests/mptcp.c +61 -0
@@ -18,6 +18,7 @@#include "mptcp_bpf_rr.skel.h" #include "mptcp_bpf_red.skel.h" #include "mptcp_bpf_burst.skel.h" +#include "mptcp_setsockopt.skel.h" #define NS_TEST "mptcp_ns"#define ADDR_1 "10.0.1.1"
@@ -813,6 +814,64 @@ static void test_burst(void)mptcp_bpf_burst__destroy (skel); } +static void test_setsockopt (void ) +{ +struct mptcp_setsockopt *skel; +struct netns_obj *netns; +int cgroup_fd, server_fd, client_fd; +int err; +int inq; +socklen_t len; + + cgroup_fd =test__join_cgroup ("/mptcp_setsockopt" ); +if (!ASSERT_OK_FD (cgroup_fd,"join_cgroup" )) +return ; + + skel =mptcp_setsockopt__open_and_load (); +if (!ASSERT_OK_PTR (skel,"skel_open_load" )) + goto close_cgroup; + + skel->links.mptcp_connect_cb = +bpf_program__attach_cgroup (skel->progs.mptcp_connect_cb, + cgroup_fd); +if (!ASSERT_OK_PTR (skel->links.mptcp_connect_cb,"attach connect_cb" )) + goto skel_destroy; + + netns =netns_new (NS_TEST , true); +if (!ASSERT_OK_PTR (netns,"netns_new" )) + goto skel_destroy; + + server_fd =start_mptcp_server (AF_INET ,NULL ,0 ,0 ); +if (!ASSERT_OK_FD (server_fd,"start_mptcp_server" )) + goto close_netns; + + client_fd =connect_to_fd (server_fd,0 ); +if (!ASSERT_OK_FD (client_fd,"connect_to_fd" )) + goto close_server; + +/* TCP_INQ should be set successfullly */ +ASSERT_EQ (skel->bss->connect_cb_inq_ret,0 ,"connect_cb TCP_INQ ret" ); + + len =sizeof (inq); + err =getsockopt (client_fd,SOL_TCP ,TCP_INQ , &inq, &len); +if (ASSERT_OK (err,"getsockopt TCP_INQ" )) +ASSERT_EQ (inq,1 ,"TCP_INQ value" ); + +/* TCP_CONGESTION should be -EOPNOTSUPP */ +ASSERT_EQ (skel->bss->connect_cb_cc_ret, -EOPNOTSUPP , +"connect_cb TCP_CONGESTION ret" ); + +close (client_fd); +close_server: +close (server_fd); +close_netns: +netns_free (netns); +skel_destroy: +mptcp_setsockopt__destroy (skel); +close_cgroup: +close (cgroup_fd); +} +void test_mptcp (void ) {if (test__start_subtest ("base" ))
@@ -835,4 +894,6 @@ void test_mptcp(void)test_red ();if (test__start_subtest ("burst" ))test_burst (); +if (test__start_subtest ("setsockopt" )) +test_setsockopt (); }
tools/testing/selftests/bpf/progs/mptcp_setsockopt.c new file +32 -0
@@ -0,0 +1,32 @@ +#include "bpf_tracing_net.h" +#include "mptcp_bpf.h" + +#ifndef TCP_INQ +#define TCP_INQ 36 +#endif + +int connect_cb_inq_ret; +int connect_cb_cc_ret; + +char cc_reno[TCP_CA_NAME_MAX ] ="reno" ; + +SEC("sockops") +int mptcp_connect_cb (struct bpf_sock_ops *skops) +{ +struct bpf_sock *sk = skops->sk; +int one =1 ; + +if (skops->op !=BPF_SOCK_OPS_TCP_CONNECT_CB ) +return 1 ; + +if (!sk || sk->protocol !=IPPROTO_MPTCP ) +return 1 ; + + connect_cb_inq_ret = +bpf_setsockopt (skops,SOL_TCP ,TCP_INQ , &one,sizeof (one)); + connect_cb_cc_ret = +bpf_setsockopt (skops,SOL_TCP ,TCP_CONGESTION , + cc_reno,sizeof (cc_reno)); + +return 1 ; +}
Signature · 1 line
2.43.0
Hi Gang,
Thank you for your modifications, that's great!
But sadly, our CI spotted some issues with it when trying to build it.
You can find more details there:
https://github.com/multipath-tcp/mptcp_net-next/actions/runs/30233020759
Status: failure
Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/acb09ec8b8b1
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1134827
Feel free to reply to this email if you cannot access logs, if you need some support to fix the error, if this doesn't seem to be caused by your modifications or if the error is a false positive one.
Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)
Hi Gang,
Thank you for your modifications, that's great!
Our CI did some validations and here is its report:
- KVM Validation: normal (except selftest_mptcp_join): Script error! ❓
- KVM Validation: normal (only selftest_mptcp_join): Script error! ❓
- KVM Validation: debug (except selftest_mptcp_join): Script error! ❓
- KVM Validation: debug (only selftest_mptcp_join): Script error! ❓
- KVM Validation: btf-normal (only bpftest_all): Script error! ❓
- KVM Validation: btf-debug (only bpftest_all): Script error! ❓
- Task: https://github.com/multipath-tcp/mptcp_net-next/actions/runs/30233020793
Initiator: Patchew Applier
Commits: https://github.com/multipath-tcp/mptcp_net-next/commits/acb09ec8b8b1
Patchwork: https://patchwork.kernel.org/project/mptcp/list/?series=1134827
If there are some issues, you can reproduce them using the same environment as the one used by the CI thanks to a docker image, e.g.:
$ cd [kernel source code]
$ docker run -v "${PWD}:${PWD}:rw" -w "${PWD}" --privileged --rm -it \
--pull always mptcp/mptcp-upstream-virtme-docker:latest \
auto-normalFor more details:
https://github.com/multipath-tcp/mptcp-upstream-virtme-docker
Please note that despite all the efforts that have been already done to have a stable tests suite when executed on a public CI like here, it is possible some reported issues are not due to your modifications. Still, do not hesitate to help us improve that ;-)
Cheers,
MPTCP GH Action bot
Bot operated by Matthieu Baerts (NGI0 Core)
Geliang Tang
reviewerHi Gang,
Quoted 6 lines · Gang Yan
From: Gang Yan <[email protected]>
Hi, Matt, Geliang, Paolo
Patch 3-4 have been reviewed by Paolo before, and ready for merge [1][2].
Do not put unrelated code into a single series. This series is for adding bpf_setsockopt support; other patches do not belong here.
Quoted 4 lines
Changelog:
v3:
- Patch 2 keeps themptcp_setsockopt_all_sfas Geliang suggested.
I recommend that you hold off on sending this patch for now: "Let's wait until IPV6_TCLASS is merged before posting this patch." Rather than keeping an unused mptcp_setsockopt_all_sf.
Do not send code that has not passed compilation to the mailing list.
Quoted 48 lines
v2:
- Patches 1 and 2 are new in this series; they addressTCP_MAXSEGhandling in thebpf_setsockopt()path.
- Patch 4 adds an early return to fixmsk->sk_rcvlowatbeing unexpectedly modified, an issue seen in v1.
- Patch 5 makes the hook safe for the non-tcp master socket: it guards
bpf_sock_ops_cb_flags_set()withsk_is_tcp()to prevent out-of- bounds
heap reads/writes through tcp_sk(sk)->bpf_sock_ops_cb_flags, and does
not setis_locked_tcp_sockfor the msk (unliketcp_call_bpf()). That
flag authorizes the verifier's direct tcp_sock-offset field accesses;
since the msk is not atcp_sock, leaving it at the default 0 is safe.v1:
Link:
https://patchwork.kernel.org/project/mptcp/cover/[email protected]/[1]
https://patchwork.kernel.org/project/mptcp/patch/[email protected]/
[2]
https://patchwork.kernel.org/project/mptcp/patch/[email protected]/Gang Yan (7):
mptcp: drop unused @max arg of__mptcp_setsockopt_set_val
mptcp: takeTCP_MAXSEGhandling into__mptcp_setsockopt_set_val
mptcp: use sockopt_lock/release_sock in sockopt
mptcp: reject sockopt requiring ssks' lock in BPF context
mptcp: enablebpf_setsockopton the master socket
mptcp: addTCP_CONNECT_CBsock_opshook
selftests: bpf: verify mptcpbpf_setsockoptfromTCP_CONNECT_CBinclude/net/mptcp.h | 9 ++ net/core/filter.c | 10 ++ net/ipv4/tcp.c | 1 + net/mptcp/protocol.c | 6 + net/mptcp/protocol.h | 29 ++++ net/mptcp/sockopt.c | 127 +++++++++++----- --
.../testing/selftests/bpf/prog_tests/mptcp.c | 61 +++++++++ .../selftests/bpf/progs/mptcp_setsockopt.c | 32 +++++8 files changed, 223 insertions(+), 52 deletions(-)create mode 100644
tools/testing/selftests/bpf/progs/mptcp_setsockopt.c
Quoted 21 lines · Geliang Tang
Hi Gang,
Quoted 8 lines · Gang Yan
From: Gang Yan <[email protected]>
Hi, Matt, Geliang, Paolo
Patch 3-4 have been reviewed by Paolo before, and ready for merge [1][2].
Do not put unrelated code into a single series. This series is for adding
bpf_setsockoptsupport; other patches do not belong here.Quoted 5 lines
Changelog:
v3:
- Patch 2 keeps themptcp_setsockopt_all_sfas Geliang suggested.
Quoted 6 lines
I recommend that you hold off on sending this patch for now: "Let's wait until
IPV6_TCLASSis merged before posting this patch." Rather than keeping an unusedmptcp_setsockopt_all_sf.Do not send code that has not passed compilation to the mailing list.
Hi Geliang
I sincerely apologize for the noise. I'll drop this patch and wait for the IPV6_TCLASS merge.
Thanks
Gang
Quoted 51 lines
Quoted 50 lines
v2:
- Patches 1 and 2 are new in this series; they addressTCP_MAXSEGhandling in thebpf_setsockopt()path.
- Patch 4 adds an early return to fixmsk->sk_rcvlowatbeing unexpectedly modified, an issue seen in v1.
- Patch 5 makes the hook safe for the non-tcp master socket: it guards
bpf_sock_ops_cb_flags_set()withsk_is_tcp()to prevent out-of- bounds
heap reads/writes through tcp_sk(sk)->bpf_sock_ops_cb_flags, and does
not setis_locked_tcp_sockfor the msk (unliketcp_call_bpf()). That
flag authorizes the verifier's direct tcp_sock-offset field accesses;
since the msk is not atcp_sock, leaving it at the default 0 is safe.v1:
Link:
https://patchwork.kernel.org/project/mptcp/cover/[email protected]/[1]
https://patchwork.kernel.org/project/mptcp/patch/[email protected]/
[2]
https://patchwork.kernel.org/project/mptcp/patch/[email protected]/Gang Yan (7):
mptcp: drop unused @max arg of__mptcp_setsockopt_set_val
mptcp: takeTCP_MAXSEGhandling into__mptcp_setsockopt_set_val
mptcp: use sockopt_lock/release_sock in sockopt
mptcp: reject sockopt requiring ssks' lock in BPF context
mptcp: enablebpf_setsockopton the master socket
mptcp: addTCP_CONNECT_CBsock_opshook
selftests: bpf: verify mptcpbpf_setsockoptfromTCP_CONNECT_CBinclude/net/mptcp.h | 9 ++ net/core/filter.c | 10 ++ net/ipv4/tcp.c | 1 + net/mptcp/protocol.c | 6 + net/mptcp/protocol.h | 29 ++++ net/mptcp/sockopt.c | 127 +++++++++++----- -- .../testing/selftests/bpf/prog_tests/mptcp.c | 61 +++++++++ .../selftests/bpf/progs/mptcp_setsockopt.c | 32 +++++ 8 files changed, 223 insertions(+), 52 deletions(-)create mode 100644
tools/testing/selftests/bpf/progs/mptcp_setsockopt.c