Number of interfaces limitation

David W. Hankins David_Hankins at isc.org
Wed Jul 19 17:31:30 UTC 2006


On Wed, Jul 19, 2006 at 02:46:08PM +0200, Shane Kerr wrote:
> It is a linked list now; the server allocates memory for each interface
> present on startup.

The internal representation was never the bottleneck (but you're
right about select()...wouldn't it be nice if we just had one
UDP socket for all interfaces?), the representation from the system's
kernel was.

See RT #5058.  KOZUKA Masahiro submitted a patch that dynamically
resizes the buffer we pass the kernel to feed us interface
descriptors.  If that's too small, the call will fail.  He
encountered problems at a relatively low number of interfaces:
merely 38.

I haven't looked closely at the patch, except on a very cursory
level it appears to do the right thing.


Ultimately, I'd prefer to reuse the bind9 interface discovery
code rather than to solve all these problems ourselves in parallel.

Mark still bears the battle scars that ultimately lead to that
implementation.

But there's an expediency issue here.

-- 
David W. Hankins	"If you don't do it right the first time,
Software Engineer		you'll just have to do it again."
Internet Systems Consortium, Inc.	-- Jack T. Hankins


More information about the dhcp-users mailing list