I have an Ultrasparc5 working as an NFS server. However, I get no reply when I ping it. It has pf_enable="NO" in rc.conf, and tcpdump shows it is receiving the pings, but does not bother replying to them! It replies to pings from its own shells. I had no problem with 5.x on the same hardware. Interface is hme. [Appears to have problems serving with apache to other local machines (but can serve to self) as well, but this might be an Apache config problem]. Fix: No Idea! How-To-Repeat: if 192.168.0.1 is a Sparc64 running 6.2-PRERELEASE, then, from another machine on the network do: > ping 192.168.0.1
I found some request if netstat -sp icmp shows any changes: Yes, it reports "xx multicast echo requsts ignored" I am pretty sure that the echo requests are fine. I can ping other machines on the network whether running Solaris on Sparc or FreeBSD on i386 with no problem. This is definitely a problem with FreeBSD 6.2-STABLE on Sparc64 (Was RC but I upgraded to try to solve the problem :-).
Andrew, I run 6.2-RELEASE on an Ultra60 (2-way SMP), with nfsd running although not actively serving any NFS clients. I am also not running Apache, as you are. I am able to ping this machine (I receive an ICMP_ECHOREPLY [id=0] packet back corresponding to each ICMP_ECHO [id=8] packet I send to the machine.) Are you familiar with ddb and kernel debugging at all? If so, find out exactly what the kernel is doing with those ICMP_ECHO packets. They would be stored in mbufs by the kernel after the NIC DMAs them into main memory. Watch those mbufs during breakpoints on various network functions. If I were you, I would start by setting breakpoints on ip_input() and icmp_input(). Regards, Jon
State Changed From-To: open->feedback Are you still having the same problem on most recent releases (6.3 or 7.0)?
State Changed From-To: feedback->closed Feedback timeout (~1 year).
Responsible Changed From-To: freebsd-sparc64->gavin Track. If this is still a problem, we can reopen the PR.