mbox series

[net-next,0/4] mptcp: various small improvements

Message ID 20241021-net-next-mptcp-misc-6-13-v1-0-1ef02746504a@kernel.org (mailing list archive)
Headers show
Series mptcp: various small improvements | expand

Message

Matthieu Baerts (NGI0) Oct. 21, 2024, 3:14 p.m. UTC
The following patches are not related to each other.

- Patch 1: Avoid sending advertisements on stale subflows, reducing
  risks on loosing them.

- Patch 2: Annotate data-races around subflow->fully_established, using
  READ/WRITE_ONCE().

- Patch 3: A small clean-up on the PM side, avoiding a bit of duplicated
  code.

- Patch 4: Use "Middlebox interference" MP_TCPRST code in reaction to a
  packet received without MPTCP options in the middle of a connection.

Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
---
Davide Caratti (1):
      mptcp: use "middlebox interference" RST when no DSS

Gang Yan (1):
      mptcp: annotate data-races around subflow->fully_established

Geliang Tang (1):
      mptcp: implement mptcp_pm_connection_closed

Matthieu Baerts (NGI0) (1):
      mptcp: pm: send ACK on non-stale subflows

 net/mptcp/diag.c       |  2 +-
 net/mptcp/options.c    |  4 ++--
 net/mptcp/pm.c         |  3 +++
 net/mptcp/pm_netlink.c | 14 +++++++++++---
 net/mptcp/protocol.c   |  8 +++-----
 net/mptcp/protocol.h   |  6 +++---
 net/mptcp/subflow.c    | 16 ++++++++++------
 7 files changed, 33 insertions(+), 20 deletions(-)
---
base-commit: 7cb08476e19fb3d0dce618df7c11713434553e27
change-id: 20241018-net-next-mptcp-misc-6-13-c34335423ea0

Best regards,

Comments

patchwork-bot+netdevbpf@kernel.org Oct. 28, 2024, 11 p.m. UTC | #1
Hello:

This series was applied to netdev/net-next.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Mon, 21 Oct 2024 17:14:02 +0200 you wrote:
> The following patches are not related to each other.
> 
> - Patch 1: Avoid sending advertisements on stale subflows, reducing
>   risks on loosing them.
> 
> - Patch 2: Annotate data-races around subflow->fully_established, using
>   READ/WRITE_ONCE().
> 
> [...]

Here is the summary with links:
  - [net-next,1/4] mptcp: pm: send ACK on non-stale subflows
    https://git.kernel.org/netdev/net-next/c/a42f3076648e
  - [net-next,2/4] mptcp: annotate data-races around subflow->fully_established
    https://git.kernel.org/netdev/net-next/c/581c8cbfa934
  - [net-next,3/4] mptcp: implement mptcp_pm_connection_closed
    https://git.kernel.org/netdev/net-next/c/5add80bfdc46
  - [net-next,4/4] mptcp: use "middlebox interference" RST when no DSS
    https://git.kernel.org/netdev/net-next/c/46a3282b87b1

You are awesome, thank you!
Jakub Kicinski Oct. 29, 2024, 11:50 p.m. UTC | #2
On Mon, 21 Oct 2024 17:14:02 +0200 Matthieu Baerts (NGI0) wrote:
> The following patches are not related to each other.
> 
> - Patch 1: Avoid sending advertisements on stale subflows, reducing
>   risks on loosing them.
> 
> - Patch 2: Annotate data-races around subflow->fully_established, using
>   READ/WRITE_ONCE().
> 
> - Patch 3: A small clean-up on the PM side, avoiding a bit of duplicated
>   code.
> 
> - Patch 4: Use "Middlebox interference" MP_TCPRST code in reaction to a
>   packet received without MPTCP options in the middle of a connection.

Could be a coincidence but looks like we got two flakes in mptcp-join
since yesterday (different sub-cases). What's more sad is that it looks
like our subcase parsing is broken, when I look at the subcase view
all sub-cases are marked as passing :( Could you take a look?
Matthieu Baerts (NGI0) Oct. 30, 2024, 5:07 p.m. UTC | #3
Hi Jakub,

On 30/10/2024 00:50, Jakub Kicinski wrote:
> On Mon, 21 Oct 2024 17:14:02 +0200 Matthieu Baerts (NGI0) wrote:
>> The following patches are not related to each other.
>>
>> - Patch 1: Avoid sending advertisements on stale subflows, reducing
>>   risks on loosing them.
>>
>> - Patch 2: Annotate data-races around subflow->fully_established, using
>>   READ/WRITE_ONCE().
>>
>> - Patch 3: A small clean-up on the PM side, avoiding a bit of duplicated
>>   code.
>>
>> - Patch 4: Use "Middlebox interference" MP_TCPRST code in reaction to a
>>   packet received without MPTCP options in the middle of a connection.
> 
> Could be a coincidence but looks like we got two flakes in mptcp-join
> since yesterday (different sub-cases).

Thank you for having reported this. It looks like a coincidence, but I
will monitor that.

> What's more sad is that it looks
> like our subcase parsing is broken, when I look at the subcase view
> all sub-cases are marked as passing :( Could you take a look?

I think the issue is with the 'retry' feature, when the nested tests are
parsed. I hope I fixed this in:

  https://github.com/linux-netdev/nipa/pull/43

Cheers,
Matt