* [Weekly meetings] MoM - 9th of June 2022
@ 2022-06-22 16:23 Matthieu Baerts
0 siblings, 0 replies; only message in thread
From: Matthieu Baerts @ 2022-06-22 16:23 UTC (permalink / raw)
To: MPTCP Upstream
Hello everyone,
Last Thursday, we had our 198th meeting with Mat, Ossama (Intel), Paolo,
Davide (Redhat), Geray (Parkyeri) 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.kernel.org/project/mptcp/list/?state=3
netdev (if mptcp ML is in cc) (by: /):
/
our repo (by: Geliang Tang):
12874730 [mptcp-next,v4,2/2] mptcp: trace MP_FAIL subflow in mptcp_sock
12874728 [mptcp-next,v4,1/2] mptcp: refactor MP_FAIL response timeout:
- it looks like the worker might not be launched reliably (maybe
~2%) → reset timer got removed
- reuse MPTCP socket timer to catch up when a worker has not been
launched
- similar to what is done in TCP timer side
- Squash-to patch needed.
- Paolo might be able to look at that next week if no one else had
to chance to do that before (patches in review)
12870186 [mptcp-next,3/3] Squash to "selftests/bpf: Add bpf_rr test"
12870185 [mptcp-next,2/3] Squash to "selftests/bpf: Add bpf_bkup test"
12870184 [mptcp-next,1/3] Squash to "selftests/bpf: Add bpf_first test"
12867508 [mptcp-next,v7,13/13] selftests/bpf: Add bpf_rr test
12867507 [mptcp-next,v7,12/13] selftests/bpf: Add bpf_rr scheduler
12867506 [mptcp-next,v7,11/13] selftests/bpf: Add bpf_bkup test
12867505 [mptcp-next,v7,10/13] selftests/bpf: Add bpf_bkup scheduler
12867504 [mptcp-next,v7,09/13] Squash to "selftests/bpf: add bpf_first
test"
12867503 [mptcp-next,v7,08/13] Squash to "selftests/bpf: add bpf_first
schedul...
12867502 [mptcp-next,v7,07/13] selftests/bpf: Add mptcp sched structs
12867501 [mptcp-next,v7,06/13] bpf: Add bpf_mptcp_sched_kfunc_set
12867500 [mptcp-next,v7,05/13] Squash to "mptcp: add bpf_mptcp_sched_ops"
12867499 [mptcp-next,v7,04/13] Squash to "mptcp: add get_subflow wrappers"
12867498 [mptcp-next,v7,03/13] mptcp: add scheduled in
mptcp_subflow_context
12867497 [mptcp-next,v7,02/13] Squash to "mptcp: add sched in mptcp_sock"
12867496 [mptcp-next,v7,01/13] Squash to "mptcp: add struct
mptcp_sched_ops":
- Paolo would like to have a look at this series:
- it looks like we might traverse all subflows linked to a
connections a few times even if it is not needed
- for the "default" case, we should avoid these traversals
- (BPF code should not impact the "default" behaviour)
- another optimisation has been removed, might be good to remove
it only for BPF
- may be good to rework how the scheduler architecture before
going further
- (on the bright side, it is good the patches are there and we can
change stuff while easily checking everything is still working as expected)
Pending patches:
- The list of pending patches can be seen on PatchWork:
https://patchwork.kernel.org/project/mptcp/list/?state=*
netdev (if mptcp ML is in cc) (by: Mat Martineau):
12874585 [net-next,2/2] mptcp: move MPTCPOPT_HMAC_LEN to net/mptcp.h
12874584 [net-next,1/2] mptcp: fix conflict with <netinet/in.h>
our repo (by: Dmytro SHYTYI, Geliang Tang, Jiapeng Chong, Matthieu
Baerts):
12282219: RFC: [RESEND,RFC,2/4] tcp: move selected mptcp helpers to
tcp.h/mptcp.h
12282221: RFC: [RESEND,RFC,4/4] tcp: parse tcp options contained in
reset packets
12282223: RFC: [RESEND,RFC,mptpcp-next] mptcp: add ooo prune support
12282225: RFC: [RESEND,1/5] tcp: make two mptcp helpers available to tcp
stack
12282227: RFC: [RESEND,5/5] mptcp: send fastclose if userspace closes
socket with unread data
12321111: RFC: mptcp: Remove redundant assignment to remaining:
- RFC
12858306: Changes Requested: [RFC,mptcp-next,v3] mptcp: Fast Open Mechanism:
- Some discussions with the timestamps but also the mapping
- Not really functional from the MPTCP point of view
- Some special code needed on the MSK side to deal with these data
in the SYN
12870447: Changes Requested: [mptcp-next,v5,1/7] Squash to "mptcp: add
get_subflow wrappers"
12870448: Changes Requested: [mptcp-next,v5,2/7] mptcp: redundant
subflows push pending
12870449: Changes Requested: [mptcp-next,v5,3/7] mptcp: placeholder for
subflow_push_pending
12870450: Changes Requested: [mptcp-next,v5,4/7] mptcp: add
sched_get_retrans wrapper
12870451: Changes Requested: [mptcp-next,v5,5/7] mptcp: redundant
subflows retrans support
12870452: Changes Requested: [mptcp-next,v5,6/7] selftests/bpf: Add
bpf_red scheduler
12870453: Changes Requested: [mptcp-next,v5,7/7] selftests/bpf: Add
bpf_red test:
12871207: Changes Requested: [mptcp-next] Squash to "mptcp: redundant
subflows retrans support" v5
- v5 of "BPF redundant scheduler"
12874836: New: [mptcp-next] Squash to "selftests/bpf: Add bpf_first test":
- some clean-up
Issues on Github:
https://github.com/multipath-tcp/mptcp_net-next/issues/
Recently opened (latest from last week: 279)
282 selftests: `simult_flows`: failing with debug kernel: transfer
slower than expected [bug] [selftests]:
- 450ms:
- it is linked to mptcp_connect does a few sleep
- to allow additional subflows to be created
- + a few more ms
- it should not be needed to wait more because we wait for the
2nd subflow to be connected
- or not run this test with a debug kconfig? it already support
slow transfers but here we are super slow
- it could be interesting to measure the usual time we get to see
the distribution and increase the time just for debug kconfig
280 support of ndiffport [question]:
- either announcing the same IP with a different port
- or netlink PM to create multiple subflows to the same IP but
different ports but we also need to open listen sockets
- could be nice to have an mptcpd ndiffport plugin :)
Bugs (opened, flagged as "bug" and assigned)
264 selftests: diag: failing on the public CI with the new
debug.config [bug] [selftests] @matttbe:
- TODO: try to reproduce it → not OK locally
181 implement data_fin ack retransmission for subflow in TIME_WAIT
state [bug] @mjmartineau:
- the priority got reduced but if needed, can be discussed
Bugs (opened and flagged as "bug" and not assigned)
282 selftests: `simult_flows`: failing with debug kernel: transfer
slower than expected [bug] [selftests]
279 [syzkaller] Warning in icsk_get_port [bug] [syzkaller]:
- it looks like there is something to do on MPTCP code
- after having applied the v1 patch, the warning has been seen
only once
- on the upstream side, there are still some bugs there around the
locking side, more work needed and quite hard to reproduce
- maybe causing the single warning Mat saw
- probably best to wait
248 packetdrill: more tests failing due to packets arriving later
than expected [bug] [packetdrill]
In Progress (opened, new feature and assigned)
234 Packetdrill: Support MPC+DATA+checksum error [enhancement]
[packetdrill] @spoorva
167 packetdrill: add coverage for RM_ADDR [enhancement] [packetdrill]
@dcaratti
75 BPF: packet scheduler [enhancement] @geliangtang:
- see patches above
74 BPF: path manager [enhancement] @geliangtang
For later (opened and not assigned assigned)
280 support of ndiffport [question]
273 Packetdrill: test outgoing MP_FASTCLOSE [enhancement] [packetdrill]
269 Allow having a mix of v4/v6 subflows for the same socket
[enhancement]
266 Packetdrill: add MP_FAIL coverage [packetdrill]
236 Review supported sockopts list [enhancement]
222 Netlink event API: add SUBFLOW_CREATED event [enhancement]
215 TCP Urgent pointer and MPTCP [enhancement]
213 add MPTCP man page [enhancement]
208 better handing of ssk memory pressure in the TX path [enhancement]
202 Add sendmsg support for ancillary data [enhancement]
197 more mibs needed [enhancement]
180 Get an update when MPTCP fall back to TCP [enhancement]
177 improve retransmit subflow selection [enhancement]
169 packetdrill: add coverage for ADD_ADDR and MP_JOIN on a different
port [enhancement] [packetdrill]
150 remove completely workqueue usage [enhancement]
141 avoid acquiring mptcp_data_lock() twice in the receive path
[enhancement]
133 PM: Closing the MPTCP connection when last subflow is not the
initial one and its IP address is removed [enhancement]
128 When the last subflow is closed without DATA_FIN and msk
Established, close msk (after a timeout) [enhancement]
79 allow 'force to MPTCP' mode: BPF [enhancement]
78 notify the application (userspace) when a subflow is
added/removed [enhancement]
77 [gs]etsockopt: forward to new/existing SF [enhancement]
76 [gs]etsockopt per subflow: BPF [enhancement]
61 move msk clone after ctx creation [enhancement]
59 (MP)TFO support [enhancement]
57 After a few attempts of failed MPTCP, directly fallback to TCP
for new connections [enhancement]
43 [syzkaller] Change syzkaller to exercise MPTCP inet_diag
interface [enhancement] [syzkaller]
41 reduce indirect call usage [enhancement]
24 Revisit layout of struct mptcp_subflow_context [enhancement]
Recently closed (since last week)
281 selftests: join: `Infinite map` (95) is failing with a debug
kernel: userspace error [bug] [selftests] @geliangtang
276 <linux/mptcp.h> conflict with <netinet/in.h> in C library [bug]
FYI: Current Roadmap:
- Bugs: https://github.com/multipath-tcp/mptcp_net-next/projects/2
- Current/Coming merge window (5.20/6.0):
https://github.com/multipath-tcp/mptcp_net-next/projects/15
- For later: https://github.com/multipath-tcp/mptcp_net-next/projects/4
Patches to send to netdev:
- Fixes for other trees: /
- Fixes for -net:
- [1c60f24cb2a5] mptcp: invoke MP_FAIL response when needed
(Geliang Tang):
- Recently applied
- There will be a Squash-to patch for fix an issue, see above
- Fixes for net-next:
- [a95484faf5da] mptcp: fix conflict with <netinet/in.h> (Ossama
Othman):
- sent, in review upstream
- Features for net-next:
- [4af562ddbbb1] mptcp: move MPTCPOPT_HMAC_LEN to net/mptcp.h
(Geliang Tang):
- sent, in review upstream
- Features for net-next (next): /
- Features for other trees:
- [4daf3d747cc0] mptcp: add struct mptcp_sched_ops (Geliang Tang)
- [eaa6a27fa969] mptcp: add a new sysctl scheduler (Geliang Tang)
- [728c1107887f] mptcp: add sched in mptcp_sock (Geliang Tang)
- [d6e834993498] mptcp: add scheduled in mptcp_subflow_context
(Geliang Tang)
- [8907ce414e3c] mptcp: add get_subflow wrappers (Geliang Tang)
- [94260f8ab147] bpf: Add bpf_mptcp_sched_ops (Geliang Tang)
- [71e1ac1b8e6e] bpf: Add bpf_mptcp_sched_kfunc_set (Geliang Tang)
- [abbcd58e8175] selftests/bpf: Add mptcp sched structs (Geliang
Tang)
- [49153fb815cb] selftests/bpf: Add bpf_first scheduler (Geliang
Tang)
- [9880ce6ad1dd] selftests/bpf: Add bpf_first test (Geliang Tang)
- [cbcc76ae580a] selftests/bpf: Add bpf_bkup scheduler (Geliang
Tang)
- [97a6d924f9cb] selftests/bpf: Add bpf_bkup test (Geliang Tang)
- [eb49a09a9182] selftests/bpf: Add bpf_rr scheduler (Geliang Tang)
- [eab5d98e92ee] selftests/bpf: Add bpf_rr test (Geliang Tang):
- can wait, still some open discussions upstream
Patches from upstream:
- One series from Eric:
-
https://patchwork.kernel.org/project/netdevbpf/list/?series=648713&state=*&archive=both
- net: reduce tcp_memory_allocated inflation
- It will probably be applied soon
- there are still some stuff to do on the MPTCP side:
- see patch 6/7:
- Removal of SK_RECLAIM_CHUNK and SK_RECLAIM_THRESHOLD
is left to MPTCP maintainers as a follow up.
-
https://patchwork.kernel.org/project/netdevbpf/patch/20220609063412.2205738-7-eric.dumazet@gmail.com/
- best to wait for it to be applied before proposing other stuff
- Mat will create an issue for that
Extra tests:
- news about Syzkaller? (Christoph / Mat):
- Focusing on the WARN, see above, ticket 279
- news about interop with mptcp.org/other stacks? (Christoph):
- /
- news about Intel's kbuild? (Mat):
- Find same intermittent failure as what has been reported by
the public CI
- packetdrill (Davide):
- /
- Patchew (Davide):
- Still alive
- WebUI: Can be a bit slow but we have to deal with that :)
- CI (Matth):
- /
TopGit Tree:
- currently net is sync with upstream only when net-next has been
sync with net upstream
- we did that to reduce conflicts in code we don't maintain
- but maybe not a source of issue
- TODO: Matth: we can try to turn this one and abort if we see it
takes us time → Done
Cross platforms:
- use CONFIG_TUN_VNET_CROSS_LE?
-> This option allows TUN/TAP and MACVTAP device drivers in a
little-endian kernel to parse vnet headers that come from a big-endian
legacy virtio device.
-> Userspace programs can control the feature using the
TUNSETVNETBE and TUNGETVNETBE ioctls.
- not a lot of doc about it
- could it help to setup VM to detect LE/BE issues?
Netdev:
-
https://netdevconf.info/0x16/news.html?netdev-0x16-call-for-submissions
- Important dates:
Closing of CFS: Wed, Sept. 7, 2022
Notification by: Thu, Sept. 15, 2022
Conference dates: Oct 24th - 28th, 2022
- For plumbers:
eBPF & Networking Proposals (Open)
Submissions close: August 10, 2022 - Speakers notified: from
August 12, 2022
Next meeting:
- Next one on Thursday, the 23rd of June.
- 15:00 UTC (8am PST, 4pm CET, 11am CST)
- Still open to everyone!
- https://annuel2.framapad.org/p/mptcp_upstreaming_20220623
Feel free to comment on these topics and suggest new ones for the next
meeting!
Talk to you on Thursday,
Matt
--
Tessares | Belgium | Hybrid Access Solutions
www.tessares.net
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2022-06-22 16:24 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-06-22 16:23 [Weekly meetings] MoM - 9th of June 2022 Matthieu Baerts
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).