| Message ID | cover.1627547666.git.geliangtang@xiaomi.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | MP_FAIL support | expand |
On Thu, 29 Jul 2021, Geliang Tang wrote: > From: Geliang Tang <geliangtang@xiaomi.com> > > v7: > - The single subflow case is handled by sending MP_FAIL + RST instead > of sending MP_FAIL echo. > Hi Geliang - I ran some tests using your checksum test patch, and verified the single subflow behavior. So I think this looks good for the export branch: Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com> Thanks, Mat > v6: only rebased patch 1 > - move the struct member 'fail_seq' behind 'ext_copy'. > - define OPTION_MPTCP_FAIL to BIT(12), BIT(11) is used by DSS > - move the MP_FAIL writing code at the beginning of mptcp_write_options, > and add the 'unlikely' tag. > - tag: export/20210728T080904 > > v5: > - patch 1, change "ret = true;" to "return true;" > - patch 3, in the single-subflow case, send MP_FAIL and receive the > echo, then temporarily handled by reset. > > v4: > - just deal with the multiple subflows case, put the single subflow > case into the new 'infinite mapping' part. > > v3: > - respond with MP_FAIL > - add single subflow check > - add infinite mapping sending and receiving > - export/20210626T054902 > > v2: > - MP_FAIL logic: > * Peer B send a DSS to peer A, and the data has been modify by the > middleboxes, then peer A detects the bad checksum. > * In the multiple subflows case, peer A sends MP_FAIL+RST back to peer B, > and peer A discards the data following the bad data sequence number. Peer > B receives this MP_FAIL+RST, and close this subflow. > * In the single subflow case, using the simple implementation, peer A > sends MP_FAIL back to peer B, and peer A fallback to a regular TCP. Peer > B receives this MP_FAIL, and fallback to a regular TCP. > > Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/52 > > Geliang Tang (5): > mptcp: MP_FAIL suboption sending > mptcp: MP_FAIL suboption receiving > mptcp: send out MP_FAIL when data checksum fails > mptcp: add the mibs for MP_FAIL > selftests: mptcp: add MP_FAIL mibs check > > include/net/mptcp.h | 5 +- > net/mptcp/mib.c | 2 + > net/mptcp/mib.h | 2 + > net/mptcp/options.c | 76 ++++++++++++++++++- > net/mptcp/pm.c | 5 ++ > net/mptcp/protocol.h | 19 +++++ > net/mptcp/subflow.c | 16 ++++ > .../testing/selftests/net/mptcp/mptcp_join.sh | 38 ++++++++++ > 8 files changed, 158 insertions(+), 5 deletions(-) > > -- > 2.31.1 > > > -- Mat Martineau Intel
Hi Geliang, Mat, Paolo, On 29/07/2021 10:40, Geliang Tang wrote: > From: Geliang Tang <geliangtang@xiaomi.com> > > v7: > - The single subflow case is handled by sending MP_FAIL + RST instead > of sending MP_FAIL echo. (...) > Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/52 > > Geliang Tang (5): > mptcp: MP_FAIL suboption sending > mptcp: MP_FAIL suboption receiving > mptcp: send out MP_FAIL when data checksum fails > mptcp: add the mibs for MP_FAIL > selftests: mptcp: add MP_FAIL mibs check Thank you for the patches and reviews! Now in our tree with Mat's RvB tags: - 6bf0897526b8: mptcp: MP_FAIL suboption sending - 913f5ee5f5e8: mptcp: MP_FAIL suboption receiving - cce924a902eb: mptcp: send out MP_FAIL when data checksum fails - e14356532033: mptcp: add the mibs for MP_FAIL - 77405167d756: selftests: mptcp: add MP_FAIL mibs check - Results: 7706daf6fd74..d2d5ed438884 Builds and tests are now in progress: https://cirrus-ci.com/github/multipath-tcp/mptcp_net-next/export/20210804T085052 https://github.com/multipath-tcp/mptcp_net-next/actions/workflows/build-validation.yml?query=branch:export/20210804T085052 Cheers, Matt
From: Geliang Tang <geliangtang@xiaomi.com> v7: - The single subflow case is handled by sending MP_FAIL + RST instead of sending MP_FAIL echo. v6: only rebased patch 1 - move the struct member 'fail_seq' behind 'ext_copy'. - define OPTION_MPTCP_FAIL to BIT(12), BIT(11) is used by DSS - move the MP_FAIL writing code at the beginning of mptcp_write_options, and add the 'unlikely' tag. - tag: export/20210728T080904 v5: - patch 1, change "ret = true;" to "return true;" - patch 3, in the single-subflow case, send MP_FAIL and receive the echo, then temporarily handled by reset. v4: - just deal with the multiple subflows case, put the single subflow case into the new 'infinite mapping' part. v3: - respond with MP_FAIL - add single subflow check - add infinite mapping sending and receiving - export/20210626T054902 v2: - MP_FAIL logic: * Peer B send a DSS to peer A, and the data has been modify by the middleboxes, then peer A detects the bad checksum. * In the multiple subflows case, peer A sends MP_FAIL+RST back to peer B, and peer A discards the data following the bad data sequence number. Peer B receives this MP_FAIL+RST, and close this subflow. * In the single subflow case, using the simple implementation, peer A sends MP_FAIL back to peer B, and peer A fallback to a regular TCP. Peer B receives this MP_FAIL, and fallback to a regular TCP. Closes: https://github.com/multipath-tcp/mptcp_net-next/issues/52 Geliang Tang (5): mptcp: MP_FAIL suboption sending mptcp: MP_FAIL suboption receiving mptcp: send out MP_FAIL when data checksum fails mptcp: add the mibs for MP_FAIL selftests: mptcp: add MP_FAIL mibs check include/net/mptcp.h | 5 +- net/mptcp/mib.c | 2 + net/mptcp/mib.h | 2 + net/mptcp/options.c | 76 ++++++++++++++++++- net/mptcp/pm.c | 5 ++ net/mptcp/protocol.h | 19 +++++ net/mptcp/subflow.c | 16 ++++ .../testing/selftests/net/mptcp/mptcp_join.sh | 38 ++++++++++ 8 files changed, 158 insertions(+), 5 deletions(-)