What's fd 8? Why extra "read(8, 0xFFBEEA17, 1) (sleeping...)"

carcassone_fr_at_yahoo.com
Date: 09/19/03 

Date: 19 Sep 2003 11:11:06 -0700

I have a problem where scp will stall on a sleep at the end of the
transfer. I've done some poking around with truss on scp -v -v -v and
found that on the machine that works, there is only one "read(8,
0xFFBEEA17, 1)(sleeping...)" near the beginning:

debug1: Found key in /u/app/.ssh/known_hosts:1
debug1: bits set: 1580/3191
read(8, 0xFFBEF33F, 1) (sleeping...)
debug1: ssh_rsa_verify: signature correct
debug1: kex_derive_keys

On the machine that stalls, there is an extra "read(8, 0xFFBEEA17,
1)(sleeping...)" after the file is transferred, causing a stall of 5
minutes:

alarm(1) = 0
read(4, "1F9D90 pD2C0 y01A0A0C183".., 8192) = 8192
write(7, "1F9D90 pD2C0 y01A0A0C183".., 8192) = 8192
read(4, " a 7 ,87EDB01EF6C382D810".., 8192) = 8192
write(7, " a 7 ,87EDB01EF6C382D810".., 8192) = 8192
read(4, " z X Z d88D5 E92 X `E489".., 8192) = 8192
write(7, " z X Z d88D5 E92 X `E489".., 8192) = 8192
read(4, "A802 U GFFF6E9 e l9C KFF".., 1143) = 1143
write(7, "A802 U GFFF6E9 e l9C KFF".., 1143) = 1143
ioctl(1, TIOCGSID, 0xFFBEE7CC) = 0
getsid(0) = 8697
ioctl(1, TIOCGPGRP, 0xFFBEE834) = 0
ioctl(1, TIOCGWINSZ, 0xFFBEE890) = 0
testfile 100%
|*****************************************************
rite(1, "\r t e s t f i l e ".., 158) = 158
alarm(0) = 1
write(1, "\n", 1) = 1
close(4) = 0
write(7, "\0", 1) = 1
read(8, 0xFFBEEA17, 1) (sleeping...)
read(8, "\0", 1) = 1
llseek(0, 0, SEEK_CUR) = 38000
_exit(0)

Can someone tell me what's happening with "read(8, 0xFFBEEA17, 1)"?
What's fd 8? Is it a block vs non_block issue? Is it SIGHUP?

I can send complete log if necessary.

Please reply with post as well as e-mail.
Thanks in advance for your help.