|
> user space tools should not include kernel headers (which is where ipvs.h
> will be residing with your patch). I suggest following:
libipvs.h depends on ip_vs.h, which is in the kernel source tree (typically
/usr/src/linux/include/net/ip_vs.h), so including a kernel header in a
user-space application is necessary (unless the application is distributed
with something similar to ip_vs.h, which seems a little redundant?). Perhaps
we should duplicate the necessary structures from ip_vs.h into libipvs.h,
making a 'user-space-only' header ??
Maybe I'm misunderstanding the role of libipvs.h, but it seems strange to
require a kernel header (ip_vs.h) in a user-space application, even
indirectly.
> That however is even a better idea :).
OK, I'll have a play with it.
The first publicly-usable version of feedbackd (0.2.0) should be out in the
next couple of days. Just ironing out some last-minute bugs.
This will be the standalone version of feedbackd, which alters ipvs tables
directly. We should get the event-passing system sorted so I can integrate
with keepalived & friends.
Jeremy
|
