The MPTCP data path is quite complex and hard to understend even
without some foggy comments referring to modified code and/or
completely misleading from the beginning.
Update a few of them to more accurately describing the current
status.
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Daire reported a user-space application hang-up when the
peer is forcibly closed before the data transfer completion.
The relevant application expects the peer to either
do an application-level clean shutdown or a transport-level
connection reset.
We can accommodate a such user by extending the fastclose
usage: at fd close time, if the msk socket has some unread
data, and at FIN_WAIT timeout.
Note that at MPTCP close time we must ensure that the TCP
subflows will reset: set the linger socket option to a suitable
value.
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
When an mptcp socket is closed due to an incoming FASTCLOSE
option, so specific sk_err is set and later syscall will
fail usually with EPIPE.
Align the current fastclose error handling with TCP reset,
properly setting the socket error according to the current
msk state and propagating such error.
Additionally sendmsg() is currently not handling properly
the sk_err, always returning EPIPE.
Reviewed-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The mptcp socket and its subflow sockets in accept queue can't be
released after the process exit.
While the release of a mptcp socket in listening state, the
corresponding tcp socket will be released too. Meanwhile, the tcp
socket in the unaccept queue will be released too. However, only init
subflow is in the unaccept queue, and the joined subflow is not in the
unaccept queue, which makes the joined subflow won't be released, and
therefore the corresponding unaccepted mptcp socket will not be released
to.
This can be reproduced easily with following steps:
1. create 2 namespace and veth:
$ ip netns add mptcp-client
$ ip netns add mptcp-server
$ sysctl -w net.ipv4.conf.all.rp_filter=0
$ ip netns exec mptcp-client sysctl -w net.mptcp.enabled=1
$ ip netns exec mptcp-server sysctl -w net.mptcp.enabled=1
$ ip link add red-client netns mptcp-client type veth peer red-server \
netns mptcp-server
$ ip -n mptcp-server address add 10.0.0.1/24 dev red-server
$ ip -n mptcp-server address add 192.168.0.1/24 dev red-server
$ ip -n mptcp-client address add 10.0.0.2/24 dev red-client
$ ip -n mptcp-client address add 192.168.0.2/24 dev red-client
$ ip -n mptcp-server link set red-server up
$ ip -n mptcp-client link set red-client up
2. configure the endpoint and limit for client and server:
$ ip -n mptcp-server mptcp endpoint flush
$ ip -n mptcp-server mptcp limits set subflow 2 add_addr_accepted 2
$ ip -n mptcp-client mptcp endpoint flush
$ ip -n mptcp-client mptcp limits set subflow 2 add_addr_accepted 2
$ ip -n mptcp-client mptcp endpoint add 192.168.0.2 dev red-client id \
1 subflow
3. listen and accept on a port, such as 9999. The nc command we used
here is modified, which makes it use mptcp protocol by default.
$ ip netns exec mptcp-server nc -l -k -p 9999
4. open another *two* terminal and use each of them to connect to the
server with the following command:
$ ip netns exec mptcp-client nc 10.0.0.1 9999
Input something after connect to trigger the connection of the second
subflow. So that there are two established mptcp connections, with the
second one still unaccepted.
5. exit all the nc command, and check the tcp socket in server namespace.
And you will find that there is one tcp socket in CLOSE_WAIT state
and can't release forever.
Fix this by closing all of the unaccepted mptcp socket in
mptcp_subflow_queue_clean() with __mptcp_close().
Now, we can ensure that all unaccepted mptcp sockets will be cleaned by
__mptcp_close() before they are released, so mptcp_sock_destruct(), which
is used to clean the unaccepted mptcp socket, is not needed anymore.
The selftests for mptcp is ran for this commit, and no new failures.
Fixes: f296234c98 ("mptcp: Add handling of incoming MP_JOIN requests")
Fixes: 6aeed90450 ("mptcp: fix race on unaccepted mptcp sockets")
Cc: stable@vger.kernel.org
Reviewed-by: Jiang Biao <benbjiang@tencent.com>
Reviewed-by: Mengen Sun <mengensun@tencent.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Menglong Dong <imagedong@tencent.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
We will soon introduce an optional per-netns ehash.
This means we cannot use tcp_hashinfo directly in most places.
Instead, access it via net->ipv4.tcp_death_row.hashinfo.
The access will be valid only while initialising tcp_hashinfo
itself and creating/destroying each netns.
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
GENL_ADMIN_PERM checks that the user has CAP_NET_ADMIN in the initial
namespace by calling netlink_capable(). Instead, use GENL_UNS_ADMIN_PERM
which uses netlink_ns_capable(). This checks that the caller has
CAP_NET_ADMIN in the current user namespace.
See also
commit 4a92602aa1 ("openvswitch: allow management from inside user namespaces")
which introduced this mechanism. See also
commit 5617c6cd6f ("nl80211: Allow privileged operations from user namespaces")
which introduced this for nl80211.
Signed-off-by: Thomas Haller <thaller@redhat.com>
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
We had historically not checked that genlmsghdr.reserved
is 0 on input which prevents us from using those precious
bytes in the future.
One use case would be to extend the cmd field, which is
currently just 8 bits wide and 256 is not a lot of commands
for some core families.
To make sure that new families do the right thing by default
put the onus of opting out of validation on existing families.
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Acked-by: Paul Moore <paul@paul-moore.com> (NetLabel)
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading sysctl_max_skb_frags, it can be changed concurrently.
Thus, we need to add READ_ONCE() to its readers.
Fixes: 5f74f82ea3 ("net:Add sysctl_max_skb_frags")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
If the mptcp socket creation fails due to a CGROUP_INET_SOCK_CREATE
eBPF program, the MPTCP protocol ends-up leaking all the subflows:
the related cleanup happens in __mptcp_destroy_sock() that is not
invoked in such code path.
Address the issue moving the subflow sockets cleanup in the
mptcp_destroy_common() helper, which is invoked in every msk cleanup
path.
Additionally get rid of the intermediate list_splice_init step, which
is an unneeded relic from the past.
The issue is present since before the reported root cause commit, but
any attempt to backport the fix before that hash will require a complete
rewrite.
Fixes: e16163b6e2 ("mptcp: refactor shutdown and close")
Reported-by: Nguyen Dinh Phi <phind.uet@gmail.com>
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Co-developed-by: Nguyen Dinh Phi <phind.uet@gmail.com>
Signed-off-by: Nguyen Dinh Phi <phind.uet@gmail.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
New subflows are created within the kernel using O_NONBLOCK, so
EINPROGRESS is the expected return value from kernel_connect().
__mptcp_subflow_connect() has the correct logic to consider EINPROGRESS
to be a successful case, but it has also used that error code as its
return value.
Before v5.19 this was benign: all the callers ignored the return
value. Starting in v5.19 there is a MPTCP_PM_CMD_SUBFLOW_CREATE generic
netlink command that does use the return value, so the EINPROGRESS gets
propagated to userspace.
Make __mptcp_subflow_connect() always return 0 on success instead.
Fixes: ec3edaa7ca ("mptcp: Add handling of outgoing MP_JOIN requests")
Fixes: 702c2f646d ("mptcp: netlink: allow userspace-driven subflow establishment")
Acked-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Link: https://lore.kernel.org/r/20220725205231.87529-1-mathew.j.martineau@linux.intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
While reading these sysctl variables, they can be changed concurrently.
Thus, we need to add READ_ONCE() to their readers.
- .sysctl_rmem
- .sysctl_rwmem
- .sysctl_rmem_offset
- .sysctl_wmem_offset
- sysctl_tcp_rmem[1, 2]
- sysctl_tcp_wmem[1, 2]
- sysctl_decnet_rmem[1]
- sysctl_decnet_wmem[1]
- sysctl_tipc_rmem[1]
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading sysctl_tcp_workaround_signed_windows, it can be changed
concurrently. Thus, we need to add READ_ONCE() to its readers.
Fixes: 15d99e02ba ("[TCP]: sysctl to allow TCP window > 32767 sans wscale")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While reading sysctl_tcp_moderate_rcvbuf, it can be changed
concurrently. Thus, we need to add READ_ONCE() to its readers.
Fixes: 1da177e4c3 ("Linux-2.6.12-rc2")
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Paolo Abeni
Jakub Kicinski
Menglong Dong
Benjamin Hesmans
Dmytro Shytyi
Kuniyuki Iwashima
Thomas Haller
Geliang Tang
Matthieu Baerts
Mat Martineau