On Wed, Jan 17, 2001 at 02:19:01PM +0100, Andi Kleen wrote:
> On Wed, Jan 17, 2001 at 12:40:39PM +0100, Gleb Natapov wrote:
> > On Wed, Jan 17, 2001 at 12:06:52PM +0100, Andi Kleen wrote:
> > > On Wed, Jan 17, 2001 at 09:19:45AM +0100, Gleb Natapov wrote:
> > > > On Tue, Jan 16, 2001 at 09:28:34PM +0300, kuznet@xxxxxxxxxxxxx wrote:
> > > > > Hello!
> > > > >
> > > > > > Resync should be an exception and not the rule IMO.
> > > > >
> > > > > If in your system simultaneous UP of 100 interfaces is not an
> > > > > exception. 8)
> > > > Forget about interfaces :). Suppose I have BGP router with 100.000
> > > > prefixes on
> > > > one interface. When interface goes UP my router daemon adds all this
> > > > prefixes
> > > > to the kernel. This will generate burst of netlink messages right?
> > >
> > > Netlink sendmsg does flow control based on the buffer.
> > >
> >
> > But if there is another process listening to netlink and it wants to know
> > about routing
> > table changes. Will kernel stop the process that adds routes to the routing
> > table until
> > reading process will empty the socket? I hope not.
>
> It does, unless you made the socket non blocking, in which case you would
> get an EAGAIN and could wait using poll(2) for new write space.
> [that's no different from how normal non-blocking networking works]
>
You are trying to say that if I'll connect to my router via 9600 serial line
and run 'ip monitor'
there the routing daemon will not be able to feed routes to the kernel quicker
than ip monitor will
be able to read them and send output via slow serial line?! Somehow 9600 serial
line become a
bottleneck! Are you sure about that, or I misunderstood you?
--
Gleb.
|