This is a follow up to the topic discussed in recent pubblic mtg.
Introduces unique id for accurate subflow stats tracking and
aggregate mptcp counters, plus some minimal self-tests.
The tests themself do not take in account support for running on
older kernel.
This is on top of "mptcp: a bunch of data race fixes".
There should be non trivial conflicts with:
"mptcp: use get_retrans wrapper".
v4 -> v5:
- changed again binary layout for MPTCP_FULL_INFO structs (Matttbe)
- fixed 32bit build issue
- reordered the patches, less controversial first
v3 -> v4:
- change binary layout for MPTCP_FULL_INFO structs (Florian)
v2 -> v3:
- address Matttbe comments on patch 1, 2 and 5, see the indivdual
patches changelog for the details
v1 -> v2:
- introduce MPTCP_FULL_INFO instead of overloading a tcp_info field
- add related self-tests
- fix a couple of subflow_id initialization bugs
Paolo Abeni (6):
mptcp: move snd_una update earlier for fallback socket.
mptcp: track some aggregate data counters.
selftests: mptcp: explicitly tests aggregate counters
mptcp: add subflow unique id
mptcp: introduce MPTCP_FULL_INFO getsockopt
selftests: mptcp: add MPTCP_FULL_INFO testcase
include/uapi/linux/mptcp.h | 29 ++++
net/mptcp/options.c | 14 +-
net/mptcp/protocol.c | 24 ++-
net/mptcp/protocol.h | 9 +-
net/mptcp/sockopt.c | 155 +++++++++++++++++-
net/mptcp/subflow.c | 2 +
.../selftests/net/mptcp/mptcp_sockopt.c | 118 ++++++++++++-
7 files changed, 332 insertions(+), 19 deletions(-)
--
2.40.1