I just noticed a really peculiar network slowdown with -current from
about a week ago.  The machine has two NICs - an old 16-bit SMC card
(ed0) running 10baseT and a PCI SMC card (tx0) running 100baseTX full

When I scp (either ssh-1.2.27 or OpenSSH) to a remote machine via ed0,
I get a throughput of ~8KB/sec.  When I go via tx0, I get ~3MB/sec
(CPU limited).  If I use ftp via ed0, I get wire speed (just over
1MB/sec).  (I haven't tried 

Looking at the packets sent via ed0 with tcpdump, I find:
15:29:58.140543 remote.ssh > local.4825: . ack 3344 win 33580 (DF) [tos 0x8] 
15:29:58.141278 local.4825 > remote.ssh: . 3344:4804(1460) ack 552 win 17520 (DF) [tos 
15:29:58.340556 remote.ssh > local.4825: . ack 4804 win 33580 (DF) [tos 0x8] 
15:29:58.341180 local.4825 > remote.ssh: . 4804:6264(1460) ack 552 win 17520 (DF) [tos 
15:29:58.540553 remote.ssh > local.4825: . ack 6264 win 33580 (DF) [tos 0x8] 
15:29:58.541172 local.4825 > remote.ssh: . 6264:7724(1460) ack 552 win 17520 (DF) [tos 
15:29:58.740559 remote.ssh > local.4825: . ack 7724 win 33580 (DF) [tos 0x8] 
15:29:58.741185 local.4825 > remote.ssh: . 7724:9184(1460) ack 552 win 17520 (DF) [tos 
15:29:58.940554 remote.ssh > local.4825: . ack 9184 win 33580 (DF) [tos 0x8] 
15:29:58.941164 local.4825 > remote.ssh: . 9184:10644(1460) ack 552 win 17520 (DF) 
[tos 0x8] 
15:29:59.140553 remote.ssh > local.4825: . ack 10644 win 33580 (DF) [tos 0x8] 
15:29:59.141178 local.4825 > remote.ssh: . 10644:12104(1460) ack 552 win 17520 (DF) 
[tos 0x8] 
15:29:59.340556 remote.ssh > local.4825: . ack 12104 win 33580 (DF) [tos 0x8] 

Notice that although it's advertising a window size of 17520, it's
not sending the next packet until the previous packet is ack'd.
When I ftp via ed0, or use ftp or scp via tx0, I get the expected
behaviour - the send side is several packets ahead of the ack's.

I didn't notice this before I upgraded last week (though it was
previously running software from just prior to 4.0-RELEASE).

The possibly unusual things in the machine:
- It has VLAN support compiled in (though not currently in use).
- It has IPFW support compiled in (via a couple of rules to
  ignore an HP OpenView machine that wants to manage every host
  in the company).
- There are also a couple of serial user-land ppp links active.

I can't understand why it only affects ssh on one link.  If it
affected everything on ed0, or if it affected ssh on both NICs,
I'd have some idea where to start looking.

Does anyone have any ideas?


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message

Reply via email to