Hi Greg, Sasha,

Here are two MPTCP patches backports (patches 2-3/4), and one
prerequisite (patch 1/4), that recently failed to apply to the 6.1
stable tree. They prevent some locking issues with MPTCP.

After having cherry-picked patch 1/4 -- a simple refactoring to make a
function more generic -- patch 2/4 applied without any issue.

For patch 3/4, I had to resolve two simple function because two
if-statements around the modified code have curly braces in v6.1, not
later, see commit 976d302fb616 ("mptcp: deduplicate error paths on
endpoint creation").

On top of that, patch 4/4 fixes MPTCP userspace PM selftest that has
been recently broken due to a backport done in v6.1.8.

Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
---
Matthieu Baerts (2):
      mptcp: sockopt: make 'tcp_fastopen_connect' generic
      selftests: mptcp: userspace: fix v4-v6 test in v6.1

Paolo Abeni (2):
      mptcp: fix locking for setsockopt corner-case
      mptcp: fix locking for in-kernel listener creation

 net/mptcp/pm_netlink.c                            | 10 ++++++----
 net/mptcp/sockopt.c                               | 20 ++++++++++++++------
 net/mptcp/subflow.c                               |  2 +-
 tools/testing/selftests/net/mptcp/userspace_pm.sh | 11 +++++++++++
 4 files changed, 32 insertions(+), 11 deletions(-)
---
base-commit: 9012d1ebd3236e1d741ab4264f1d14e276c2e29f
change-id: 20230214-upstream-stable-20230214-linux-6-1-12-rc1-mptcp-fixes-df24a5f41151

Best regards,
-- 
Matthieu Baerts <matthieu.baerts@tessares.net>

Hey Matthieu,

I've ended up doing something slightly different:

On Tue, Feb 14, 2023 at 05:05:06PM +0100, Matthieu Baerts wrote:
>Hi Greg, Sasha,
>
>Here are two MPTCP patches backports (patches 2-3/4), and one
>prerequisite (patch 1/4), that recently failed to apply to the 6.1
>stable tree. They prevent some locking issues with MPTCP.
>
>After having cherry-picked patch 1/4 -- a simple refactoring to make a
>function more generic -- patch 2/4 applied without any issue.

I did the same for 1+2.

>For patch 3/4, I had to resolve two simple function because two
>if-statements around the modified code have curly braces in v6.1, not
>later, see commit 976d302fb616 ("mptcp: deduplicate error paths on
>endpoint creation").

Instead of resolving the conflict, I took 976d302fb616 ("mptcp:
deduplicate error paths on endpoint creation".

>On top of that, patch 4/4 fixes MPTCP userspace PM selftest that has
>been recently broken due to a backport done in v6.1.8.

I didn't have this one, I'll go and queue it up. Thanks!

-- 
Thanks,
Sasha