With the in-kernel path-manager, it is possible to change the 'fullmesh' flag. The code in mptcp_pm_nl_fullmesh() expects to change it only on 'subflow' endpoints, to recreate more or less subflows using the linked address.
Unfortunately, the set_flags() hook was a bit more permissive, and allowed 'implicit' endpoints to get the 'fullmesh' flag while it is not allowed before.
That's what syzbot found, triggering the following warning:
WARNING: CPU: 0 PID: 6499 at net/mptcp/pm_netlink.c:1496 __mark_subflow_endp_available net/mptcp/pm_netlink.c:1496 [inline] WARNING: CPU: 0 PID: 6499 at net/mptcp/pm_netlink.c:1496 mptcp_pm_nl_fullmesh net/mptcp/pm_netlink.c:1980 [inline] WARNING: CPU: 0 PID: 6499 at net/mptcp/pm_netlink.c:1496 mptcp_nl_set_flags net/mptcp/pm_netlink.c:2003 [inline] WARNING: CPU: 0 PID: 6499 at net/mptcp/pm_netlink.c:1496 mptcp_pm_nl_set_flags+0x974/0xdc0 net/mptcp/pm_netlink.c:2064 Modules linked in: CPU: 0 UID: 0 PID: 6499 Comm: syz.1.413 Not tainted 6.13.0-rc5-syzkaller-00172-gd1bf27c4e176 #0 Hardware name: Google Compute Engine/Google Compute Engine, BIOS Google 09/13/2024 RIP: 0010:__mark_subflow_endp_available net/mptcp/pm_netlink.c:1496 [inline] RIP: 0010:mptcp_pm_nl_fullmesh net/mptcp/pm_netlink.c:1980 [inline] RIP: 0010:mptcp_nl_set_flags net/mptcp/pm_netlink.c:2003 [inline] RIP: 0010:mptcp_pm_nl_set_flags+0x974/0xdc0 net/mptcp/pm_netlink.c:2064 Code: 01 00 00 49 89 c5 e8 fb 45 e8 f5 e9 b8 fc ff ff e8 f1 45 e8 f5 4c 89 f7 be 03 00 00 00 e8 44 1d 0b f9 eb a0 e8 dd 45 e8 f5 90 <0f> 0b 90 e9 17 ff ff ff 89 d9 80 e1 07 38 c1 0f 8c c9 fc ff ff 48 RSP: 0018:ffffc9000d307240 EFLAGS: 00010293 RAX: ffffffff8bb72e03 RBX: 0000000000000000 RCX: ffff88807da88000 RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000 RBP: ffffc9000d307430 R08: ffffffff8bb72cf0 R09: 1ffff1100b842a5e R10: dffffc0000000000 R11: ffffed100b842a5f R12: ffff88801e2e5ac0 R13: ffff88805c214800 R14: ffff88805c2152e8 R15: 1ffff1100b842a5d FS: 00005555619f6500(0000) GS:ffff8880b8600000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000020002840 CR3: 00000000247e6000 CR4: 00000000003526f0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: <TASK> genl_family_rcv_msg_doit net/netlink/genetlink.c:1115 [inline] genl_family_rcv_msg net/netlink/genetlink.c:1195 [inline] genl_rcv_msg+0xb14/0xec0 net/netlink/genetlink.c:1210 netlink_rcv_skb+0x1e3/0x430 net/netlink/af_netlink.c:2542 genl_rcv+0x28/0x40 net/netlink/genetlink.c:1219 netlink_unicast_kernel net/netlink/af_netlink.c:1321 [inline] netlink_unicast+0x7f6/0x990 net/netlink/af_netlink.c:1347 netlink_sendmsg+0x8e4/0xcb0 net/netlink/af_netlink.c:1891 sock_sendmsg_nosec net/socket.c:711 [inline] __sock_sendmsg+0x221/0x270 net/socket.c:726 ____sys_sendmsg+0x52a/0x7e0 net/socket.c:2583 ___sys_sendmsg net/socket.c:2637 [inline] __sys_sendmsg+0x269/0x350 net/socket.c:2669 do_syscall_x64 arch/x86/entry/common.c:52 [inline] do_syscall_64+0xf3/0x230 arch/x86/entry/common.c:83 entry_SYSCALL_64_after_hwframe+0x77/0x7f RIP: 0033:0x7f5fe8785d29 Code: ff ff c3 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 40 00 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 a8 ff ff ff f7 d8 64 89 01 48 RSP: 002b:00007fff571f5558 EFLAGS: 00000246 ORIG_RAX: 000000000000002e RAX: ffffffffffffffda RBX: 00007f5fe8975fa0 RCX: 00007f5fe8785d29 RDX: 0000000000000000 RSI: 0000000020000480 RDI: 0000000000000007 RBP: 00007f5fe8801b08 R08: 0000000000000000 R09: 0000000000000000 R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000 R13: 00007f5fe8975fa0 R14: 00007f5fe8975fa0 R15: 00000000000011f4 </TASK>
Here, syzbot managed to set the 'fullmesh' flag on an 'implicit' and used -- according to 'id_avail_bitmap' -- endpoint, causing the PM to try decrement the local_addr_used counter which is only incremented for the 'subflow' endpoint.
Note that 'no type' endpoints -- not 'subflow', 'signal', 'implicit' -- are fine, because their ID will not be marked as used in the 'id_avail' bitmap, and setting 'fullmesh' can help forcing the creation of subflow when receiving an ADD_ADDR.
Fixes: 73c762c1f07d ("mptcp: set fullmesh flag in pm_netlink") Cc: stable@vger.kernel.org Reported-by: syzbot+cd16e79c1e45f3fe0377@syzkaller.appspotmail.com Closes: https://lore.kernel.org/6786ac51.050a0220.216c54.00a6.GAE@google.com Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/540 Reviewed-by: Mat Martineau martineau@kernel.org Signed-off-by: Matthieu Baerts (NGI0) matttbe@kernel.org --- net/mptcp/pm_netlink.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/net/mptcp/pm_netlink.c b/net/mptcp/pm_netlink.c index 98ac73938bd8196e196d5ee8c264784ba8d37645..572d160edca33c0a941203d8ae0b0bde0f2ef3e2 100644 --- a/net/mptcp/pm_netlink.c +++ b/net/mptcp/pm_netlink.c @@ -2020,7 +2020,8 @@ int mptcp_pm_nl_set_flags(struct sk_buff *skb, struct genl_info *info) return -EINVAL; } if ((addr.flags & MPTCP_PM_ADDR_FLAG_FULLMESH) && - (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)) { + (entry->flags & (MPTCP_PM_ADDR_FLAG_SIGNAL | + MPTCP_PM_ADDR_FLAG_IMPLICIT))) { spin_unlock_bh(&pernet->lock); GENL_SET_ERR_MSG(info, "invalid addr flags"); return -EINVAL;