installation problem: network not working
Hi,
I tried to install debian woody on a quite standard i386 machine (AMD
K6-2 500MHz, 128 MB RAM, 30 GB Maxtor IDE hard drive, Realtek RTL 8029
NIC).
I made 6 floppies, rescue.bin, root.bin, driver-{1,2,3,4}.bin from the
images in debian/dists/woody/main/disks-i386/3.0.17-2001-11-18/images-1.44.
I booted, configured the keyboard, partitioned the HD, made and
activated swap (64MB), made and mounted / and /home (2GB and 25GB),
loaded kernel and drivers from the rescue and driver floppies, loaded
the ne2k-pci module, configured my network (IP addr, netmask, gateway,
DNS server, host & domain name). Until here, everything works as
expected.
The machine I want to install on, is behind a Linux router doing
packet filtering and NAT. This is configured correctly and works.
Then I tried to install the base system via network. I chose to install
woody and then chose the default server http://http.us.debian.org to
install from. No proxy.
The network connection is made but then the installation process
hangs. The new machine I want to install on, doesn't send TCP ACK's
for the TCP segments it receives. Following is tcpdump output on the
Linux router. frickel.bogus is 10.0.0.3 and is the machine to be
installed, isnogud.escape.de is the router:
07:28:40.945538 frickel.bogus.1024 > isnogud.escape.de.domain: 131+ A? http.us.debian.org. (36)
07:28:40.946345 isnogud.escape.de.domain > frickel.bogus.1024: 131 6/9/9 A gluck.external.hp.com, (478) (DF)
07:28:43.945118 frickel.bogus.iad2 > gluck.external.hp.com.http: S 956071101:956071101(0) win 16060 <mss 1460,sackOK,timestamp 349407 0,nop,wscale 0> (DF)
07:28:44.264705 gluck.external.hp.com.http > frickel.bogus.iad2: S 1480498670:1480498670(0) ack 956071102 win 5792 <mss 1412,sackOK,timestamp 2281719381 349407,nop,wscale 0> (DF)
07:28:44.265074 frickel.bogus.iad2 > gluck.external.hp.com.http: . ack 1 win 16060 <nop,nop,timestamp 349439 2281719381> (DF)
07:28:44.265499 frickel.bogus.iad2 > gluck.external.hp.com.http: P 1:123(122) ack 1 win 16060 <nop,nop,timestamp 349439 2281719381> (DF)
07:28:44.610153 gluck.external.hp.com.http > frickel.bogus.iad2: . ack 123 win 5792 <nop,nop,timestamp 2281719412 349439> (DF)
07:28:44.644727 gluck.external.hp.com.http > frickel.bogus.iad2: . 1:1401(1400) ack 123 win 5792 <nop,nop,timestamp 2281719414 349439> (DF)
07:28:44.660423 gluck.external.hp.com.http > frickel.bogus.iad2: . 1401:2801(1400) ack 123 win 5792 <nop,nop,timestamp 2281719414 349439> (DF)
07:28:44.665049 frickel.bogus.iad2 > gluck.external.hp.com.http: . ack 2801 win 15400 <nop,nop,timestamp 349479 2281719414> (DF)
07:28:45.060419 gluck.external.hp.com.http > frickel.bogus.iad2: . 2801:4201(1400) ack 123 win 5792 <nop,nop,timestamp 2281719462 349479> (DF)
07:28:45.076631 gluck.external.hp.com.http > frickel.bogus.iad2: . 4201:5601(1400) ack 123 win 5792 <nop,nop,timestamp 2281719462 349479> (DF)
07:28:45.092288 gluck.external.hp.com.http > frickel.bogus.iad2: . 5601:7001(1400) ack 123 win 5792 <nop,nop,timestamp 2281719462 349479> (DF)
07:28:45.205005 frickel.bogus.iad2 > gluck.external.hp.com.http: . ack 7001 win 15400 <nop,nop,timestamp 349533 2281719462> (DF)
07:28:45.562473 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281719512 349533> (DF)
07:28:45.578155 gluck.external.hp.com.http > frickel.bogus.iad2: . 8401:9801(1400) ack 123 win 5792 <nop,nop,timestamp 2281719512 349533> (DF)
07:28:45.593848 gluck.external.hp.com.http > frickel.bogus.iad2: . 9801:11201(1400) ack 123 win 5792 <nop,nop,timestamp 2281719512 349533> (DF)
07:28:45.609548 gluck.external.hp.com.http > frickel.bogus.iad2: . 11201:12601(1400) ack 123 win 5792 <nop,nop,timestamp 2281719512 349533> (DF)
07:28:45.941293 arp who-has frickel.bogus tell isnogud.escape.de
07:28:45.941596 arp reply frickel.bogus is-at 0:40:5:32:74:76
07:28:47.146109 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281719665 349533> (DF)
07:28:50.187231 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281719971 349533> (DF)
07:28:56.273892 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281720583 349533> (DF)
07:29:08.493381 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281721807 349533> (DF)
07:29:32.984272 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281724255 349533> (DF)
07:29:37.981577 arp who-has frickel.bogus tell isnogud.escape.de
07:29:37.981620 arp who-has isnogud.escape.de tell frickel.bogus
07:29:37.981703 arp reply isnogud.escape.de is-at 0:40:5:64:c4:80
07:30:21.993070 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281729151 349533> (DF)
07:30:26.991778 arp who-has frickel.bogus tell isnogud.escape.de
07:30:26.992068 arp reply frickel.bogus is-at 0:40:5:32:74:76
07:31:59.910416 gluck.external.hp.com.http > frickel.bogus.iad2: P 7001:8401(1400) ack 123 win 5792 <nop,nop,timestamp 2281738943 349533> (DF)
07:32:04.901886 arp who-has isnogud.escape.de tell frickel.bogus
07:32:04.901942 arp reply isnogud.escape.de is-at 0:40:5:64:c4:80
08:13:16.749301 frickel.bogus.iad2 > gluck.external.hp.com.http: F 123:123(0) ack 12601 win 15400 <nop,nop,timestamp 616704 2281719512> (DF)
08:13:17.060877 arp who-has frickel.bogus tell isnogud.escape.de
08:13:17.061212 arp reply frickel.bogus is-at 0:40:5:32:74:76
08:13:17.061263 gluck.external.hp.com.http > frickel.bogus.iad2: R 1480511271:1480511271(0) win 0 (DF)
The TCP connection establishment works ok. Then a few data are
downloaded from gluck.external.hp.com. The debian installation system
on frickel acknowledges up to byte 7001 at 07:28:45.205005.
gluck.external.hp.com the continues to send bytes 7001 to 12600 in
four TCP segments but frickel does not acknowledge them. This seems
to be a bug in the debian installation system. Looks like a kernel
bug. After a timeout, gluck.external.hp.com retransmits the first
unacknowledged segment with bytes 7001-8400, but still gets no ACK.
But frickel.bogus has not crashed. At 8:13:16, I pressed CTRL-C to
interrupt the installation and got back into some menu, where I could
again select to install the base system. I did this but again with no
success. When interrupted, the TCP socket on frickel still seems to
be open, since it sends a TCP FIN to gluck.external.hp.com when the
socket is closed. gluck.external.hp.com, however, seems to already
have closed its socket and sends a RST.
I repeated the installation several times (5 or 6 times or so) from
the beginning, i.e. rebooting from the rescue disk, but without
success: Sometimes the installation just hangs as described above, and
sometimes the network connection also hangs but the installation
screen after a while shows an error message about a corrupted package
(apt_0.5.4 AFAIR).
Is this a known problem and is there a work-around? OTOH, I installed
debian woody successfully on my notebook using the same floppy disk.
That, however, was at my job without NAT. But the tcpdump output does
not look like a NAT problem.
If you answer, please CC: to my email address, since I'm not on the
mailing list.
urs
Reply to: