Matthieu Baerts
Matthieu Baerts
Linked to #379, local addresses are being tracked on the kernel side to be able to send `REMOVE_ADDR` for subflows that have been created (`MP_JOIN` (and `MP_CAPABLE`)) but not announced...
KCIDB is collecting results from different CIs. It would be nice to publish our tests results there to have all tests in a central point, but also to benefit from...
The recent switch to GH Actions reduces the usage of Cirrus CI -- still needed for the tests with a debug kconfig -- but it duplicates CI related code. Also,...
It can be helpful to regularly monitor which part of the code is covered or not by our test suite. Adding gcov supports should not be difficult to do: https://www.kernel.org/doc/html/latest/dev-tools/gcov.html...
When GSO is used, we can now have packets with a size > 2¹⁶ (see commit 7c4e983c4f3c) which is not compatible with MPTCP's DLL (Data-Level Length) field (see issue #450...
As reported by Willem on [netdev](https://lore.kernel.org/netdev/[email protected]/T/), some inputs are randomised (`mptcp_connect.c` and `mptcp_inq.c`). It was certainly used to increase the coverage but we don't print anything about that and there...
When discussing about the packet scheduler API at the last meeting, it sounds very likely the current packet scheduler will not react by queuing more packets if some subflows only...
This test is sometimes failing, not often. An example: ``` # Created /tmp/tmp.LFVXW6Eoza (size 1024 KB) containing data sent by client # Created /tmp/tmp.ZIPnR2U18Y (size 1024 KB) containing data sent...
In case of network issue (blackhole, wrong path, etc.), it might be interesting to re-established subflows that got closed after a TCP-level error (timeout, RST, etc.) as it was maybe...
The number of subflows, when using the in-kernel PM, is currently limited to (like the entrypoints/received add_addr). Limiting endpoints to 8 makes sense (arrays are used) but not subflows: it...