On Sun, 08 Mar 2015 06:16:04 +0000
Antti Kantee <[email protected]> wrote:

> > I hope this the right list.
> 
> Yes, it is.  If there is something else than politeness that lead you to 
> suspect this isn't, please let me know if I can adjust some wordings 
> somewhere.
> 

Since the version I'm using is bundled with NetBSD 6, I had some
concern that tech-kern or something like that might be more appropriate.


> > Am I missing an -l option, or is this simply not implemented right now?
> 
> the latter.  After spending quite a bit of effort on select/poll in 
> hijack.c, and since kqueue was seemingly even more difficult, I took the 
> "wait until someone complains" approach.  IIRC you're the first to 
> complain in 4 years, so that approach paid off.
> 

Makes perfect sense - don't write code you don't need :)


> Is there any reason you want to hijack nginx itself (slow)?  If your 
> goal is using nginx at decent speeds against a userlevel tcp/ip stack, 
> you are probably happier with the "rumprun" approach, i.e. building 
> everything into one bundle.  ISTR that Mato already built nginx as a 
> rumprun stack on Xen, so technically it should be possible, but I'm not 
> 100% up to speed on the location for the tools required.  But we can 
> figure it out, if that's what you're after (mmmm, free guinea pigs).
> 

I was working with rumphijack as I'm trying to stick to completely
unmodified NetBSD binaries. It's a moonshot-type project that might not
be feasible, or maybe just far outside my abilities. Have to start
somewhere, though.

The basic goal is container style virtualization on NetBSD: something
like FreeBSD jails, Linux LXC, etc. I realized that chroot + a rump
network stack is about halfway there and was giving it a shot. Good for
production, not really - but a decent proof of concept. I was using
nginx as it's a real-world program that's trivial to test, and easy to
get going in a chroot filesystem.

-- 
Aaron B. <[email protected]>

Reply via email to