wross@engin.umich.edu (Wendy Ross) (12/04/90)
After installing the 3002 patch and rebooting my machine, I
noticed a rather bizarre error:
Starting tcpip daemons:
/etc/syslogd: ERROR: '57' getsockname, SRC not found, continuing without SRC support
syslogd
sendmail
portmap
inetd
As a result, none of the above daemons started automatically.
After narrowing everything down, I discovered that the 3001 and 3002
ps -e commands return slightly different things. 3001 version has
five columns and 3002 has four columns. Thus, when /etc/rc.tcpip
checks for the srcmstr running, the line should be changed from
src_running=`ps -e | awk '$5 == "srcmstr" { print $1 }'` <version 3001>
to
src_running=`ps -e | awk '$4 == "srcmstr" { print $1 }'` <version 3002>
I was rather perplexed, because although there was mention in the update
documentation that /etc/rc.tcpip *was* changed in the 3002 release, a
restore -T on the update bosnet.tcpip.obj file did not confirm that
/etc/rc.tcpip was one of the changed files. Additionally, the machines
that were updated did not show a change in this file.
I did a dd from the stacked tape to get these files onto my server so I
could apply the patch across the network.
Am I missing something here, or have other persons had this problem
when moving to 3002?
-Wendy
wross@caen.engin.umich.edu
<insert amusing quip here>RS0THP@ROHVM1.BITNET (Dr. Thomas Pierce) (12/05/90)
I just installed release 3002 from 1/4" tape, and my rc.tcpip
was slightly different than yours.
This release also told me that /etc/rc.tcpip would be changed.
My version is:
src_running=`ps -e | awk '$NF == "srcmstr" {print $1; exit} '`
Perhaps your tape is a 'early version' of release 3002?RS0THP@ROHVM1.BITNET (Dr. Thomas Pierce) (12/05/90)
Another interesting trick with patch tape 3002.
After I installed it and uncommented the routed line in the new rc.tcpip,
I discovered that I could telnet out and rlogin out of the RS/6000, and
I could rlogin from other machines to the RS/6000. However, I could not
telnet into the RS/6000 from any other machines on the network. Any help
would be appreciated.
You may send messages to me at:
rs0thp@rohvm1 bitnet
or
rs0thp@rohmhaas.com internet
Thanks again for any help.clarke@arnor.UUCP (Ed Clarke/204800000;07-033,2630) (12/08/90)
From article <1990Dec4.031214.11374@engin.umich.edu>, by wross@engin.umich.edu (Wendy Ross):
- ps -e commands return slightly different things. 3001 version has
- five columns and 3002 has four columns. Thus, when /etc/rc.tcpip
- checks for the srcmstr running, the line should be changed from
- src_running=`ps -e | awk '$5 == "srcmstr" { print $1 }'` <version 3001>
- to
- src_running=`ps -e | awk '$4 == "srcmstr" { print $1 }'` <version 3002>
Hmmmm - I just applied 3002 yesterday. As indicated, ps returns four
columns instead of five. In my case, rc.tcpip WAS updated to:
src_running=`ps -e | awk '$NF == "srcmstr" { print $1; exit }'`
and the 'routed' uncommenting problem mentioned in another posting did
not occur. I updated from tape, not from diskettes. On the other hand,
ioctl and fcntl seem to be severely broken now. Anyone see some stupid
mistake in the following code?
int tty;
tty = open( "/dev/tty1" , 0);
printf("tty is %d\n",tty);
if (ioctl(tty, TIOCEXCL, NULL) < 0) {
perror("ioctl TIOCEXCL");
exit(1);
}
Fairly simple and a really bad thing to break. The fcntl error is buried
in some xntp code ( ntp_io.c if you're interested ).