wireguard-linux/net/mptcp, branch jd/orphan-parallel

mptcp: depends on IPV6 but not as a module

2020-10-21T15:05:40Z

Like TCP, MPTCP cannot be compiled as a module. Obviously, MPTCP IPv6' support also depends on CONFIG_IPV6. But not all functions from IPv6 code are exported. To simplify the code and reduce modifications outside MPTCP, it was decided from the beginning to support MPTCP with IPv6 only if CONFIG_IPV6 was built inlined. That's also why CONFIG_MPTCP_IPV6 was created. More modifications are needed to support CONFIG_IPV6=m. Even if it was not explicit, until recently, we were forcing CONFIG_IPV6 to be built-in because we had "select IPV6" in Kconfig. Now that we have "depends on IPV6", we have to explicitly set "IPV6=y" to force CONFIG_IPV6 not to be built as a module. In other words, we can now only have CONFIG_MPTCP_IPV6=y if CONFIG_IPV6=y. Note that the new dependency might hide the fact IPv6 is not supported in MPTCP even if we have CONFIG_IPV6=m. But selecting IPV6 like we did before was forcing it to be built-in while it was maybe not what the user wants. Reported-by: Geert Uytterhoeven Fixes: 010b430d5df5 ("mptcp: MPTCP_IPV6 should depend on IPV6 instead of selecting it") Signed-off-by: Matthieu Baerts Link: https://lore.kernel.org/r/20201021105154.628257-1-matthieu.baerts@tessares.net Signed-off-by: Jakub Kicinski

mptcp: MPTCP_IPV6 should depend on IPV6 instead of selecting it

2020-10-21T03:54:08Z

MPTCP_IPV6 selects IPV6, thus enabling an optional feature the user may not want to enable. Fix this by making MPTCP_IPV6 depend on IPV6, like is done for all other IPv6 features. Fixes: f870fa0b5768842c ("mptcp: Add MPTCP socket stubs") Signed-off-by: Geert Uytterhoeven Reviewed-by: Matthieu Baerts Link: https://lore.kernel.org/r/20201020073839.29226-1-geert@linux-m68k.org Signed-off-by: Jakub Kicinski

mptcp: MPTCP_KUNIT_TESTS should depend on MPTCP instead of selecting it

2020-10-20T23:50:06Z

MPTCP_KUNIT_TESTS selects MPTCP, thus enabling an optional feature the user may not want to enable. Fix this by making the test depend on MPTCP instead. Fixes: a00a582203dbc43e ("mptcp: move crypto test to KUNIT") Signed-off-by: Geert Uytterhoeven Reviewed-by: Matthieu Baerts Link: https://lore.kernel.org/r/20201019113240.11516-1-geert@linux-m68k.org Signed-off-by: Jakub Kicinski

mptcp: move mptcp_options_received's port initialization

2020-10-20T23:32:36Z

Move mptcp_options_received's port initialization from mptcp_parse_option to mptcp_get_options, put it together with the other fields initializations of mptcp_options_received. Signed-off-by: Geliang Tang Reviewed-by: Matthieu Baerts Signed-off-by: Jakub Kicinski

mptcp: initialize mptcp_options_received's ahmac

2020-10-20T23:32:14Z

Initialize mptcp_options_received's ahmac to zero, otherwise it will be a random number when receiving ADD_ADDR suboption with echo-flag=1. Fixes: 3df523ab582c5 ("mptcp: Add ADD_ADDR handling") Signed-off-by: Geliang Tang Reviewed-by: Matthieu Baerts Signed-off-by: Jakub Kicinski

Merge git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

2020-10-15T19:43:21Z

Minor conflicts in net/mptcp/protocol.h and tools/testing/selftests/net/Makefile. In both cases code was added on both sides in the same place so just keep both. Signed-off-by: Jakub Kicinski

mptcp: subflows garbage collection

2020-10-10T18:04:53Z

The msk can close MP_JOIN subflows if the initial handshake fails. Currently such subflows are kept alive in the conn_list until the msk itself is closed. Beyond the wasted memory, we could end-up sending the DATA_FIN and the DATA_FIN ack on such socket, even after a reset. Fixes: 43b54c6ee382 ("mptcp: Use full MPTCP-level disconnect state machine") Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Jakub Kicinski

mptcp: fix fallback for MP_JOIN subflows

2020-10-10T18:04:53Z

Additional/MP_JOIN subflows that do not pass some initial handshake tests currently causes fallback to TCP. That is an RFC violation: we should instead reset the subflow and leave the the msk untouched. Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/91 Fixes: f296234c98a8 ("mptcp: Add handling of incoming MP_JOIN requests") Reviewed-by: Mat Martineau Signed-off-by: Paolo Abeni Signed-off-by: Jakub Kicinski

net: mptcp: make DACK4/DACK8 usage consistent among all subflows

2020-10-09T15:25:48Z

using packetdrill it's possible to observe the same MPTCP DSN being acked by different subflows with DACK4 and DACK8. This is in contrast with what specified in RFC8684 §3.3.2: if an MPTCP endpoint transmits a 64-bit wide DSN, it MUST be acknowledged with a 64-bit wide DACK. Fix 'use_64bit_ack' variable to make it a property of MPTCP sockets, not TCP subflows. Fixes: a0c1d0eafd1e ("mptcp: Use 32-bit DATA_ACK when possible") Acked-by: Paolo Abeni Signed-off-by: Davide Caratti Reviewed-by: Mat Martineau Signed-off-by: Jakub Kicinski

mptcp: fix infinite loop on recvmsg()/worker() race.

2020-10-09T00:24:04Z

If recvmsg() and the workqueue race to dequeue the data pending on some subflow, the current mapping for such subflow covers several skbs and some of them have not reached yet the received, either the worker or recvmsg() can find a subflow with the data_avail flag set - since the current mapping is valid and in sequence - but no skbs in the receive queue - since the other entity just processed them. The above will lead to an unbounded loop in __mptcp_move_skbs() and a subsequent hang of any task trying to acquiring the msk socket lock. This change addresses the issue stopping the __mptcp_move_skbs() loop as soon as we detect the above race (empty receive queue with data_avail set). Reported-and-tested-by: syzbot+fcf8ca5817d6e92c6567@syzkaller.appspotmail.com Fixes: ab174ad8ef76 ("mptcp: move ooo skbs into msk out of order queue.") Signed-off-by: Paolo Abeni Reviewed-by: Mat Martineau Signed-off-by: Jakub Kicinski