steven.grunza@dgms.com said:
> 			syslog(LOG_ERR, "tftpd: write: %m\n");
> 			syslog(LOG_ERR, "tftpd: send_data failed, size = %d"
> 			", ret_val = %d, errno = %d, peer = %d, dp = %p\n",
> 			size, ret_val, errno, peer, dp);
> 			syslog(LOG_ERR, "tftpd: %s\n", strerror(errno) ); 

This is suboptimal - syslog() does overwrite "errno", so you get
these inconsistent reports:

> Apr 30 01:35:32 scrappy tftpd[2228]: tftpd: write: Host is down
> Apr 30 01:35:32 scrappy tftpd[2228]: tftpd: send_data failed, size = 512,
> ret_val = -1, errno = 9, peer = 0, dp = 0x4dec
> Apr 30 01:35:32 scrappy tftpd[2228]: tftpd: Bad file descriptor 

For the original problem - I believe the EHOSTDOWN does occur if no
ARP entry for the IP packet destination can be obtained.
Is it possible that the SUN console doesn't answer ARP requests?
("tcpdump" could help to find out)
If yes - you could try to install a static ARP entry on the server.

best regards
Matthias