* [MPTCP] [Weekly meetings] MoM - 13th of February 2020
@ 2020-02-14 15:27 Matthieu Baerts
0 siblings, 0 replies; only message in thread
From: Matthieu Baerts @ 2020-02-14 15:27 UTC (permalink / raw)
To: mptcp
[-- Attachment #1: Type: text/plain, Size: 10764 bytes --]
Hello,
Yesterday, we had our 86th meeting with Mat, Peter and Ossama (Intel
OTC), Christoph (Apple), Paolo and Davide (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
- By /
Pending patches:
- The list of pending patches can be seen on PatchWork:
https://patchwork.ozlabs.org/project/mptcp/list/?state=*
- By Florian Westphal, Mat Martineau, Matthieu Baerts, Paolo Abeni,
Peter Krystad
1196109: RFC: [10/10,RFC] selftests:mptcp: decrease timeout to 100 sec:
- same as last week
1233392: Needs Review / ACK: subflow: place outgoing connections on the
join list:
- by Florian
- *Waiting for review*
1233648: Needs Review / ACK: [RFC,1/3] mptcp: Check connection state
before attempting send
1233650: Needs Review / ACK: [RFC,2/3] mptcp: Use per-subflow storage
for DATA_FIN sequence number
1233649: Needs Review / ACK: [RFC,3/3] mptcp: Only send DATA_FIN with
final mapping:
- by Mat
- needed for "net" ideally
- *waiting for review* : should not be too long/hard
1235143: New: mptcp: Protect subflow socket options before connection
completes:
- got some reviews
- v2 is needed (at least a change in the commit message + switch to
other function for tcp_[gs]etsockopt)
1235200: New: [1/2] mptcp: Re-factor mptcp_crypto_hmac_sha()
1235198: New: [2/2] mptcp: Make v1 changes for ADD_ADDR and RM_ADDR options:
- by Peter
- got a few reviews
1236411: New: [RFC,1/6] Squash-to: "mptcp: Add ADD_ADDR handling"
1236416: New: [RFC,2/6] Squash-to: "mptcp: Add path manager interface"
1236413: New: [RFC,3/6] Squash-to: "mptcp: Add handling of incoming
MP_JOIN requests"
1236414: New: [RFC,4/6] Squash-to: "mptcp: Add handling of outgoing
MP_JOIN requests"
1236415: New: [RFC,5/6] Squash-to: "mptcp: Implement path manager
interface commands":
- by Paolo
- WIP
- got reviews from Florian
1236417: New: [RFC,6/6] mptcp: add netlink based PM:
- can be nice to review this one because it introduces a netlink API
- once it's upstream, hard to change
- *waiting for review*
1236945: New: mptcp: select CRYPTO
1236960: New: [RFC,net] mptcp: select CRYPTO:
- kconfig complains there is a missing dependency
- should we send it?:
- yes. If not accepted, we squash it (see above patch)
- we will need it for MP_JOIN
1237408: New: [RFC,1/5] mptcp: perform mptcp ack update during read
operations
1237409: New: [RFC,2/5] tcp: permit skb stealing in recv actor
1237410: New: [RFC,3/5] mptcp: add and use mptcp_data_ready helper
1237411: New: [RFC,4/5] mptcp: update mptcp ack sequence from work queue
1237412: New: [RFC,5/5] mptcp: add rmem queue accounting:
- by Florian
- got some reviews by Paolo
FYI: Current Roadmap:
- Part 3 (next merge window):
- PM:
- locking
- basic → netlink support
- MP_JOIN
- ADD_ADDR for MPTCP v1
- clean the current patches
- need to make sure we continue to support the protocol →
workaround for shared received windows
- other items:
- Full DATA_FIN support [WIP]:
- could be nice to have it: if ready
- Shared recv window (drop data received on other subflows) [TODO]:
- could be nice to have
- could be nice to share a prototype/RFC with TCP
maintainers (Eric)
- Active backup support [WIP]
- Opti in TCP option structures (unions) [to be rebased)
- ADD_ADDR for MPTCPv1 [TODO]
- PM basic [WIP]
- fix sender ID in MP_JOIN
- Part 4 (to fully support MPTCP):
- Shared recv window (full support)
- IPv6 - IPv4 mapped support
- not dropping MPTCP options (ADD_ADDR, etc.)
- FAST_CLOSE
- full MPTCP v1 support (reliable add_addr, etc.)
- after a few attempts of failed MPTCP, we fallback to TCP
(like TFO is doing)
- PM server
- Part 5 (extra needed for prod):
- opti/perfs
- TFO
- PM netlink
- PM bpf
- Scheduler bpf
- syncookies
- [gs]etsockopt per subflow
Part 3 remaining stuff:
- Review commit messages:
- needed for the next merge window: not urgent
- PM:
- Paolo is looking at that
- waiting for reviews
- patches 1 → 5 should not change a lot
- patches 6 is going to change a bit
- reduce locks when mptcp_stream_accept (currently 3 locks) but we
might need to add an helper in inet API:
- TODO
- MP_JOIN self-tests:
- TODO
- there is a dependence on the PM (with netlink)
- and even a dependence in the other way because it is used to
validate the PM
- → so part of the new PM
- packetdrill:
- David is working on that
- https://github.com/multipath-tcp/packetdrill/pull/1
- TODO: *Matth* : try it (with the CI) and if positive, we merge
- we might need to update DSS related tests when Florian's
patches are going to be merged
- Davide is testing two scripts related to TCP fallback but
they are failing at the end:
- https://paste.centos.org/view/94fdbe8d :
====
`../common/defaults.sh`
+0 socket(..., SOCK_STREAM, IPPROTO_MPTCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0
+0 < S 0:0(0) win 32792 <mss 1460,sackOK,nop,nop,nop,wscale
7,mpcapable v1 flags[flag_h] nokey>
+0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale
8,mpcapable v1 flags[flag_h] key[skey] >
+0.01 < . 1:1(0) ack 1 win 257 <mpcapable v1 flags[flag_h]
key[ckey=2,skey]>
+0 accept(3, ..., ...) = 4
// server does 2 writes, 100 bytes each
+0.1 < P. 1:201(200) ack 1 win 225
//+0 > . 1:1(0) ack 201 <dss dack8=1 ssn=1 dll=0 nocs>
+0 > . 1:1(0) ack 201
+0 read(4, ..., 200) = 200
+0 write(4, ..., 100) = 100
+0 > P. 1:101(100) ack 201
+0 < . 201:201(0) ack 101 win 225
====
- failing at line 16
- there is a RST with DSS options
- https://paste.centos.org/view/875e1ea0 :
====
`../common/defaults.sh`
+0 socket(..., SOCK_STREAM, IPPROTO_MPTCP) = 3
+0 setsockopt(3, SOL_SOCKET, SO_REUSEADDR, [1], 4) = 0
+0 bind(3, ..., ...) = 0
+0 listen(3, 1) = 0
+0 < S 0:0(0) win 32792 <mss 1460,sackOK,nop,nop,nop,wscale
7,mpcapable v1 flags[flag_h] nokey>
+0 > S. 0:0(0) ack 1 <mss 1460,nop,nop,sackOK,nop,wscale
8,mpcapable v1 flags[flag_h] key[skey] >
+0.01 < . 1:1(0) ack 1 win 257 <mpcapable v1 flags[flag_h]
key[ckey=2,skey]>
+0 accept(3, ..., ...) = 4
// server does 2 writes, 100 bytes each
+0 write(4, ..., 100) = 100
+0 > P. 1:101(100) ack 1 <dss dack8=1 dsn8=1 ssn=1 dll=100 nocs,
nop, nop>
+0.1 < . 1:1(0) ack 101 win 225
+0 write(4, ..., 100) = 100
+0 > P. 101:201(100) ack 1
+0.1 < . 1:1(0) ack 201 win 225
====
- failing at line 18
- there are DSS options while we should not
- Question is: when should we fallback to TCP when we don't
receive MPTCP option with the TCP ACK?:
- related to §3.7 of
https://www.rfc-editor.org/authors/rfc8684.html#name-fallback
- it is possible that there are some bugs: not
respecting the RFC
- better to always include the DSS ACK
- receiver: to validate the path, it needs to receive a
full window of data with DSS ACK
- sender: to validate the path: get the 4th ACK
- What if the 3rd ACK is lost/OOO?
- TODO *Christoph/Paolo* : To be continued on the ML
- there are also a couple of patches at the end of the series we
will have to rebase/squash:
- TODO
- ADD_ADDR for MPTCPv1:
- Peter is looking at that
- show the options processing
- the echo will required the PM to be validated
- DATA_FIN:
- Mat is looking at that
- once the pending patches are accepted, they will be sent to
netdev (at least the 2 first ones):
- https://patchwork.ozlabs.org/patch/1233648/
- https://patchwork.ozlabs.org/patch/1233650/
- for the last one, it should not be an issue with one subflow
but improve the logic:
- https://patchwork.ozlabs.org/patch/1233649/
- still a better protection and small
- *waiting for reviews*
- Shared received windows:
- Florian is looking at that (indirectly)
Extra tests:
- news about Syzkaller? (see prev meeting):
- running on the "export" branch
- looking good so far
- new about Intel's kbuild? (Mat):
- one test where MPTCP is not ON by default but maybe tries
random config's:
- found an issue with CRYPTO (patch sent)
- one test where MPTCP is ON by default:
- everything OK
- MPTCP should be enabled on other builds, waiting for an answer
[gs]etsockopt:
- idea:
- new getsockopt to get "ids" of subflows
- new [gs]etsockopt to pass that to a specific subflow:
https://github.com/bhesmans/mptcp/commit/0b2c48a366d98d94ef195d7dbb834e321441a6dc
- idea for later
- get notified when there are changes in subflows?:
- http://man7.org/linux/man-pages/man3/cmsg.3.html ?
- could be interesting to be notified from the userspace when
there is a fallback to TCP
- apps will create MPTCP socket: it there is a fallback, it
could be nice to be notified (push event)
Next meeting:
- We propose to have the next meeting on Thursday, the 20th of
February.
- Usual time: 17:00 UTC (9am PST, 6pm CET)
- Still open to everyone!
- https://annuel2.framapad.org/p/mptcp_upstreaming_20200220
Feel free to comment on these points and propose new ones for the next
meeting!
Talk to you next week,
Matt
--
Matthieu Baerts | R&D Engineer
matthieu.baerts(a)tessares.net
Tessares SA | Hybrid Access Solutions
www.tessares.net
1 Avenue Jean Monnet, 1348 Louvain-la-Neuve, Belgium
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2020-02-14 15:27 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-02-14 15:27 [MPTCP] [Weekly meetings] MoM - 13th of February 2020 Matthieu Baerts
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.