[MPTCP] [Weekly meetings] MoM - 11th of March 2021

[Matthieu Baerts <matthieu.baerts at tessares.net>]

[-- Attachment #1: Type: text/plain, Size: 11870 bytes --]

Hello everyone,

Today, we just had our 139th meeting with Mat and Ossama (Intel), Paolo, 
Davide and Florian (RedHat) and myself (Tessares).

Thanks again for this new good meeting!

Here are the minutes of the meeting:



Accepted patches:
     - The list of accepted patches can be seen on PatchWork:
       https://patchwork.ozlabs.org/project/mptcp/list/?state=3


     netdev (if mptcp ML is in cc) (Davide Caratti, Mat Martineau):

1448927  [net] mptcp: fix length of ADD_ADDR with port sub-option

1447508  [net,9/9] mptcp: free resources when the port number is mismatched
1447507  [net,8/9] mptcp: fix missing wakeup
1447510  [net,7/9] mptcp: fix race in release_cb
1447509  [net,6/9] mptcp: factor out __mptcp_retrans helper()
1447511  [net,5/9] mptcp: reset 'first' and ack_hint on subflow close
1447502  [net,4/9] mptcp: dispose initial struct socket when its subflow 
is cl...
1447504  [net,3/9] mptcp: fix memory accounting on allocation error
1447505  [net,2/9] mptcp: put subflow sock on connect error
1447506  [net,1/9] mptcp: reset last_snd on subflow close


     our repo (by: Davide Caratti, Geliang Tang, Mat Martineau, Matthieu 
Baerts, Paolo Abeni ):

1451044  [mptcp-next] Squash to "mptcp: only admit explicitly supported 
sockop...
1449638  [mptcp-next] Squash-to: "mptcp: move sockopt function into a 
new file"
1449534  [v3,mptcp-next] Squash to "mptcp: add rm_list in 
mptcp_out_options" -...

1449092  [mptcp-next] selftests: mptcp: support capture with other options
1448960  [mptcp-next] Squash to "selftests: mptcp: launch mptcp_connect 
with t...

1448958  [net-next,v3,3/3] mptcp: revert "mptcp: provide subflow aware 
release...
1448957  [net-next,v3,2/3] mptcp: only admit explicitly supported sockopt
1448955  [net-next,v3,1/3] mptcp: move sockopt function into a new file

1447428  [net,v2] mptcp: fix length of ADD_ADDR with port suboption

1447193  [v6,mptcp-next,5/5] selftests: mptcp: remove id 0 address 
testcases
1447192  [v6,mptcp-next,4/5] selftests: mptcp: add addr argument for 
del_addr
1447191  [v6,mptcp-next,3/5] mptcp: remove id 0 address
1447190  [v6,mptcp-next,2/5] mptcp: unify RM_ADDR and RM_SUBFLOW receiving
1447188  [v6,mptcp-next,1/5] mptcp: remove all subflows involving id 0 
address

1447136  [mptcp-next,6/6] Squash to "mptcp: remove multi addresses and 
subflow...
1447135  [mptcp-next,5/6] Squash to "mptcp: remove multi subflows in PM"
1447134  [mptcp-next,4/6] Squash to "mptcp: add rm_list_rx in 
mptcp_pm_data"
1447133  [mptcp-next,3/6] Squash to "mptcp: add rm_list in 
mptcp_options_recei...
1447132  [mptcp-next,2/6] Squash to "mptcp: add rm_list_tx in 
mptcp_pm_data"
1447130  [mptcp-next,1/6] Squash to "mptcp: add rm_list in 
mptcp_out_options"

1446930  [mptcp-net] selftests: mptcp: Restore packet capture option in 
join t...



Pending patches:
     - The list of pending patches can be seen on PatchWork:
       https://patchwork.ozlabs.org/project/mptcp/list/?state=*


     netdev (if mptcp ML is in cc) (by: /):

/


     our repo (by: Florian Westphal, Geliang Tang, Paolo Abeni):

1370700: RFC: [RFC,2/4] tcp: move selected mptcp helpers to tcp.h/mptcp.h
1370702: RFC: [RFC,4/4] tcp: parse tcp options contained in reset packets:
     - WIP

1375893: RFC: [RFC,mptpcp-next] mptcp: add ooo prune support

1395128: RFC: [1/5] tcp: make two mptcp helpers available to tcp stack
1395133: RFC: [5/5] mptcp: send fastclose if userspace closes socket 
with unread data:
     - WIP

1426554: Changes Requested: [PATCHi,iproute2] mptcp: add support for 
event monitoring:
     - WIP

1442954: Needs Review / ACK: [mptcp-next,1/3] mptcp: add tracepoint for 
mptcp_subflow_get_send
1442956: Needs Review / ACK: [mptcp-next,2/3] mptcp: add tracepoint for 
get_mapping_status
1442957: Needs Review / ACK: [mptcp-next,3/3] mptcp: add tracepoint for 
subflow_check_data_avail:
     - Needs review

1450494: New: [RFC,1/2] mptcp: DSS checksum support
1450496: New: [RFC,2/2] mptcp: add MP_FAIL support:
     - Looking for comments
     - Mat will have a look



Issues on Github:
     https://github.com/multipath-tcp/mptcp_net-next/issues/


     Recently opened (latest from last week: 170)

   171  iproute2: support removing ID 0 [enhancement] [iproute2]:
       - kernel support added
       - also in selftests with pm tool
       - we need to modify iproute2 to reflect that


     Bugs (opened, flagged as "bug" and assigned)

   170  possible deadlock in ipv6_sock_mc_close [bug] [syzkaller] @pabeni:
       - linked to the series about sockopt
       - TODO: close it → Done

   146  DATA_FIN is not retransmitted on timeout [bug] @mjmartineau:
       - feel free to share early patches (even broken ones) to have 
early feedback


     Bugs (opened and flagged as "bug" and not assigned)

   162  sendmsg()/recvmsg() fail when an unknown CMSG argument is 
provided [bug]
   137  selftests: simult_flows.sh: unbalanced bwidth tests are unstable 
[bug]
   120  [interop] netnext is dropping packets, causing MPTCP-level 
retransmissions on mptcp.org [bug]
   119  [syzkaller] memory leak in __get_filter [bug] [syzkaller]
   107  Review use of WARN_ON() / WARN_ON_ONCE() [bug]
    65  clearing properly the status in listen() [bug]
    56  msk connection state set without msk lock [bug]


     In Progress (opened and assigned)

   167  packetdrill: add coverage for RM_ADDR [enhancement] 
[packetdrill] @dcaratti
   158  iproute2: change backup mode (MP_PRIO) for active connections 
[enhancement] [iproute2] @dcaratti

   134  Checksum support [enhancement] @geliangtang :
       - RFC patches have been shared

   131  replace some/most pr_debug with trace events [enhancement] 
@geliangtang
       - patches have been shared

    96  Python: add support for IPPROTO_MPTCP [enhancement] @matttbe
    53  MP_TCPRST support [enhancement] @fw-strlen

    52  MP_FAIL support [enhancement] @geliangtang
       - RFC patches have been shared


     Recently closed (since last week)

   168  [syzkaller] WARNING in dst_release [bug] [syzkaller]
   166  iproute2: ADD_ADDR: ports support [enhancement] [iproute2] @pabeni
   156  [syzkaller] WARNING: refcount bug in __tcp_transmit_skb [bug] 
[syzkaller]
   153  [syzkaller] WARNING: refcount bug in mptcp_incoming_options 
[bug] [syzkaller]
   140  RM_ADDR: remove a list of addresses [enhancement] @geliangtang



FYI: Current Roadmap:
     - Bugs: https://github.com/multipath-tcp/mptcp_net-next/projects/2
     - Current/Coming merge window (5.13): 
https://github.com/multipath-tcp/mptcp_net-next/projects/8
     - For later: https://github.com/multipath-tcp/mptcp_net-next/projects/4



TopGit tree has been re-created:
     - removing old refs
     - moved commits:
         - one of Geliang series was split in 2 → now back together
         - BPF patches at the end
     - new Git Markup for us and scripts (CI)



Patches to send to netdev:
     - net:
         - one patch to send
         - Mat is going to do that

     - net-next: (now open)
         - maybe better to wait for the ones linked to the sockopts
         - many others to send in the tree, e.g. ↓
         - Max 15 for the bot, first:

a4101da4b3a5c selftests: mptcp: add testcases for removing addrs
685fe8b66e42a selftests: mptcp: set addr id for removing testcases
d5d25a8d8f398 selftests: mptcp: add invert argument for chk_rm_nr
07876d7463004 mptcp: remove a list of addrs when flushing
90ef847bd22a3 mptcp: remove multi addresses and subflows in PM
36053c7035e75 mptcp: remove multi subflows in PM
6c9d8edc9334e mptcp: remove multi addresses in PM
1968574ef8a22 mptcp: add rm_list_rx in mptcp_pm_data
c3bd6ec8a59a8 mptcp: add rm_list in mptcp_options_received
a58326a355c54 mptcp: add rm_list_tx in mptcp_pm_data
1d33189486aa5 mptcp: add rm_list in mptcp_out_options

         - After:

20a09f718f161 selftests: mptcp: timeout testcases for multi addresses
f90eb60003759 selftests: mptcp: add cfg_do_w for cfg_remove
ffddb115a77d3 mptcp: move to next addr when timeout
c1a91e3969cc6 mptcp: drop unused subflow in mptcp_pm_subflow_established
4cb5e98ec7ec1 mptcp: skip connecting the connected address
a4d44afce8205 mptcp: drop argument port from mptcp_pm_announce_addr

         - Mat is going to do that



Sockopts:
     - script to monitor missing sockopts?:
         - grep in .h file + grep in mptcp/sockopts.c
         - ran by the CI
     - but if we want that, we need to fix the ones that are already 
missing :)
     - BPF and NFT socktopts have been omitted on purpose:
         - Maybe best to continue
     - → we wait to have feedback something is missing/not supported, 
more urgent topics to cover



Extra tests:
     - news about Syzkaller? (Christoph):
         - /

     - news about interop with mptcp.org? (Christoph):
         - /

     - news about Intel's kbuild? (Mat):
         - some troubles that are not caused by our stuff
         - no need to worry, they are fixing that
         - report should be back soon

     - packetdrill (Davide):
         - working on RM_ADDR and ADD_ADDR with port support

     - CI (Matth):
         - Moving stuff in TopGit tree



Offload: Checksums:
     - After a discussion with Intel Ethernet Drivers team
     - We don't depend on GRO/GSO feature but it could be done also only 
in this case
     - typically would be an additional feature than TCP checksum
     - the NIC could help calculate checksum but one complex point is 
the alignment of the data: if it is between to 64 bits → receiving side 
seems a lot more work than the transmitting side

     - For us, would be good to be helped by the NIC to do some GRO, 
re-assembling packets. Some HW can do that for TCP.

     - Idea for them would be to start doing optimisations per frame 
(e.g. checksum if possible)

     - could be done partly by the HW? e.g. checksum complete, add data 
for that in MPTCP SKB ext:
         - to be done by HW, alignment is important!
         - maybe a solution is to have different checksum flavours 
depending on the alignment but doesn't seem nice
         - (sorry, I lost the end of the discussion, maybe clearer by 
emails with schema if someone needs more details)



Offload: MPTCP frame detection:
     - Would it help? To help with TC and others to detect flows
     - Maybe to gather all MPTCP connections from different subflows and 
NICs, we could move all the MPTCP connection to the same queue → 
Reducing contention
     - Useful for not so many flows and a lot of traffic
     - would be good to load-balance subflows of the same connection to 
the same queue but not really possible to implement



SW Checksum:
     - When can we drop a packet if the checksum is wrong?
     - Before giving it to the userspace! But not too late for TCP?
     - We can no longer give to userspace data that are part of DSS that 
have been received but not covering the full DSS mapping yet → we need 
to validate the checksum of the whole DSS mapping



Checksum on the Internet:
     - It is important to have it, too many nasty middleboxes.
     - Fine if the network is fully controlled but most likely not often 
the case.



Next meeting:
     - On Thursday, the 18th of March.
     - /!\ *NOT* the usual UTC time: 15:00 UTC (8am PDT, 4pm CET, 11pm CST)
     - Still open to everyone!
     - https://annuel2.framapad.org/p/mptcp_upstreaming_20210318



Feel free to comment on these points and propose new ones for the next
meeting!

Talk to you on Thursday,
Matt
-- 
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net