Testing: what happens when client crashes and restarts.

[Rishabh04-02]

Support for recovery from client crash.

[Rishabh04-02]

Output of the test case https://pastebin.com/raw/RRWYJf84
Test case added libreswan/libreswan@c0008e5
Commit in the PR libreswan/libreswan#263

[letoams]

On Wed, 7 Aug 2019, Rishabh wrote: Output https://pastebin.com/raw/RRWYJf84

#check if tunnel is up 64 road # 65 ipsec whack --trafficstatus 66 whack: is Pluto running? connect() for "/run/pluto/pluto.ctl" failed (111 Connection refused) I would remove this. It is confusing. Do you expect a tunnel to be up? Since we just hard killed pluto and not yet restarted it, in fact the kernel has the tunnels up but obviously traffic status which is the pluto view is gone since pluto is no longer running. After this, you do an ipsec restart which is okay. But since that clears the kernel state there is no use of running trafficstatus then either without generating some pings first. So remove the trafficstatus there as well. These might look like small issues, but 2 years from now when this test case starts failing, people dont remember any of this clearly, and will get confused. I know because I wrote most test cases and are still too often confused by the writing of my younger self :) Maybe rename 8-east-road-final.sh to 8-east-final.sh. There is no point in getting the road status _again_, and then you can change the message there from "A tunnel" to "two tunnels", since east will have both of the road tunnels established. That will also be less confusing to our future selves. Paul

[Rishabh04-02]

Thanks Paul for the review. I've fixed all the issues in this test case.

and then you can change the message there from "A tunnel" to "two tunnels", since east will have both of the road tunnels established.

But why does it have Two tunnels. Is it because we killed pluto before it could send the termination request?