IP packet drop

we made the observation that in case an application does not pick up packets from the ip stack ‘in time’, packets are dropped.

We assume that the reason behind is that queues run full. (packets are not discarded with a timeout, are they?)

Now, we would like to know how we can find out the queue(buffer) sizes and whether those are configurable on per socket basis, on per interface basis And/or on per ip stack basis.

Dheerajs,

They are definitely configurable on a per socket basis (see setsockopt with options SO_SNDBUF/SO_RCVBUF). The default size is 16K per socket and you can set it as high as around 196K.

You might also be able to configure other buffers but these are likely the ones you want to change.

Tim

Thanks Tim. I got it.