double lock on socket with diskless pre2.0.1

David Monro (
Mon, 13 May 1996 00:45:34 +1000 (EST)

I get a bunch of 'double lock on socket at xxxxx' messages about 1 in 2 times
when I boot my diskless pre2.0.1 box (using nfsroot and an eprom containing
etherboot 1.1). They seem to appear just as the system is spawning the gettys
on the various consoles (so after all the system daemons are started etc). I
haven't seen this on previous kernels (meaning all of 1.3.9x at least).

I get about 16 in a row, coming from two places: most are from
skb_recv_datagram+36 (which is the lock_sock() the top of skb_recv_datagram),
the rest are from udp_sendto+200 (which I think is the lock_sock() on line 393
of linux/net/ipv4/udp.c). The system seems fine afterwards, but it looks a
little alarming...

Oh, ethernet is an NE2000, nfs server is another pre2.0.1 box.

Just thought people might like to know ;-)