> okay, I am reporting again.
> I happened to meet this before, maybe it is a rare case.
> Socket destroy delayed (r=0 w=112)
> Socket destroy delayed (r=0 w=256)
> Any idea? Had anyone seen that before?
When a socket has buffers charged to it the socket cant be destroyed so
the kernel hangs on to it. It indicates one of several things - a socket
that has some state timeouts left to do that are longer than 10 secs
after the close, a socket that has buffers queued up with a very constipated
driver, or a leak.