[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[MirageOS-devel] network stack issues

Hi all,

this is a summary of the issues we met to make synjitsu works again. We managed 
to get a working snapshot to run the experiments, but there are few issues 
which needs fixing in the longer term.

- There is still a deadlock/race in mirage-net-xen which cause packet loss when 
someone write on the network very early (the whole point of synjitsu). I was 
not really able to track down the issue precisely, but it appears in 1.3.0 (so 
we reverted back to 1.2 to run the experiments). This is tracked by 
https://github.com/mirage/mirage-net-xen/issues/20 The bug causes the initial 
SYNACK to not be sent at all, so the unikernel is receiving the next SYN after 
the timeout (after 1s).

- TCP option unmarshalling sometimes goes wrong. This bug is not very bad, it 
just causes some packets to get dropped, and I can reproduce it quite easily on 
my home network but not in the lab. This is tracked by 
https://github.com/mirage/mirage-tcpip/issues/106 This might be caused by 
malformed packets living on my network, though ...

- The stack should destroy connection after receiving a RST. This is tracked by 
https://github.com/mirage/mirage-tcpip/issues/107. This is not critical, but 
will cause a lot of unnecessary traffic on the network.

MirageOS-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.