? soacceptqueue errantly set to 128 ?

[Aranjedeath]

This is an exploratory bug, because I don't know enough about go to answer this myself and file a better bug. I've been asked to file it on behalf of IPFS.

They, apparently, use go-sockaddr.

As an IPFS user, I noticed that one of the TCP sockets bound with a 128 depth soacceptqueue. My operating system is freebsd, and as such has a kernel tunable kern.ipc.soacceptqueue. It's set to 4096, not 128. An IPFS person suggests this socket is bound or managed by go-sockaddr.

Any idea why it would not respect my OS settings? Or perhaps I should bark up golang's tree?

[jbenet]

I borrowed directly from the golang source-- i dont think i changed that, but havent checked. look at the sources in golang-- we shouldn't be doing anything different here.

cc @whyrusleeping ??

[Stebalien]

This looks like golang/go#11412 and I believe this has been fixed. Can you try again?

[Aranjedeath]

Can you spot the fix? When I wrote this bug it did not matter if you set somaxconn or soacceptqueue (both are in fact set), ipfs ignored them and set to 128. I can try it again and check, though.

[Stebalien]

@Aranjedeath nevermind; the fix isn't merged yet. The fix is here, FYI.

When I wrote this bug it did not matter if you set somaxconn or soacceptqueue

Then I'm not entirely sure this will fix it. It should have been using somaxconn.

[Stebalien]

As far as I can tell, this is a go-wide bug (all go code has this problem, from what I can tell) so I'm going to close this issue. If that's not the case, tell me and I'll look deeper.