iscream solo giving out bad pkts?

glimmerman

Hi all

Lately I seem to be getting a LOAD of rx errors occurring on the aforementioned overpriced service from eircom. I'm running a linux box acting as a nat and router, for my small network at home. I never had these probs before tho... so just wondering if anyone else is seeing the same? ifconfig ppp0 gets the goods...
Has eircom done something to the dslams?

Hmmm, something else just occurs.... if the errors are even more than the correclty received packets, I guess my cap is decreasing twice as fast....

On on OT note, I just got my first bill.... aaaaarggghhh!!!

longword

Originally posted by glimmerman
Lately I seem to be getting a LOAD of rx errors occurring on the aforementioned overpriced service from eircom.

I'm on the more overpriced Enhanced service from the Crown Alley exchange. Zero PPP packet errors out of 400MB down, 60MB up since my last reboot (been messing with machines and kernels) - 288 AAL5 frame errors out of about 700,000. I had some problems initially, but they were because of a crap motherboard that couldn't handle the IRQ load of my cheap PCI ADSL card - it's now working fine in a motherboard that should otherwise be half as fast.

On on OT note, I just got my first bill.... aaaaarggghhh!!!

I know that feeling. With my first bill I could have bought a totally butt-kicking PC.

bricks

ppp0 Link encap:Point-to-Point Protocol
inet addr:159.134.x.y P-t-P:159.134.x.z Mask:255.255.255.255
UP POINTOPOINT RUNNING NOARP MULTICAST MTU:1492 Metric:1
RX packets:942083 errors:0 dropped:0 overruns:0 frame:0
TX packets:737581 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:10

Also on the Physical interface that the Modem is plugged into:

eth1 Link encap:Ethernet HWaddr 00:10:4B:AE:97:FB
inet addr:1.1.1.1 Bcast:1.1.1.255 Mask:255.255.255.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:953449 errors:0 dropped:0 overruns:0 frame:0
TX packets:738745 errors:0 dropped:0 overruns:0 carrier:0
collisions:43 txqueuelen:100
Interrupt:10 Base address:0xfc00

One thing I do get a lot of is this error:
Sep 8 20:49:19 gateway pppoe[732]: Bogus PPPoE length field (50)

I get one 24x7 consistently 1 every 2 minutes in the /var/log/messages.

longword

I get nothing in /var/log/messages.

My setup works like this: Conexant ADSL driver presents itself as an ATM device to the standard Linux kernel ATM layer. The br2684 kernel module interfaces with that to produce a bridged ethernet device called nas0. On top of that pppd with an rp-pppoe plugin talks to the kernel pppoe module to do the PPPoE link.

Here's my bridged ethernet device. I swapped the card into a new machine yesterday so the numbers are a bit low...

nas0 Link encap:Ethernet HWaddr 00:80:AD:B1:A7:FA
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:18853 errors:0 dropped:0 overruns:0 frame:0
TX packets:16231 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:100
RX bytes:10506256 (10.0 Mb) TX bytes:1562997 (1.4 Mb)

The card driver reports this at the AAL5 layer (turning ATM into ethernet packets)...

AAL 5 Channel Stats: Receive Transmit
Good Bytes 11407008 1062240
Good Frames 18957 16332
Discarded Bytes 14592
Discarded Frames 12
CRC Errors 11
Invalid Length Errors 1
Timeout Errors 0

glimmerman

sheesh, so it prolly IS something in my setup.... anyway, just for general info I guess, I'm running Rh7.3 with the default 2.4.18-3 kernel, using the alcatel 1 port ethernet modem... hackity hack hack... hmmm,

ppp0 Link encap:Point-to-Point Protocol
inet addr:159.134.xxx.xxx
P-t-P:159.134.xxx.xMask:255.255.255.255
UP POINTOPOINT RUNNING NOARP MULTICAST MTU:1492 Metric:1
RX packets:57239 errors:64110 dropped:0 overruns:0 frame:0
TX packets:75075 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:3
RX bytes:30858680 (29.4 Mb) TX bytes:8639687 (8.2 Mb)

jd

Originally posted by bricks


One thing I do get a lot of is this error:
Sep 8 20:49:19 gateway pppoe[732]: Bogus PPPoE length field (50)

I get one 24x7 consistently 1 every 2 minutes in the /var/log/messages.

Bet you its every 125 seconds..
router is checking for membership of multicast groups and 125 seconds is the default query interval

jd

Have a look here,,
http://multicast.eircom.net/

glimmerman

So it seems the problem has something to do with MTU sizes...
I switched from using tkpppoe to just using the scripts under it, namely the standard RoaringPenguin stuff (adsl-start et al).


There is an option in the /etc/ppp/pppoe.conf file, namely:
CLAMPMSS=1412

which had to be turned on in my config (tkpppoe does not allow this as far as I could tell)... this seems to refrag the packets on my local lan down from ethernet sizes to pppoe sizes.

the only thing that still puzzles is why I didn't have the pkt errors problem before.... but that may have something to do with removing the eircom sw from the win98 machine on the internal NATted net which we tested dsl with: perhaps this had set the mtu to the corrrect size.

thanks for the help

longword

Originally posted by glimmerman
CLAMPMSS=1412

1412? That's a bit harsh. Off the top of my head I think 1452 is what it should be. I use an iptables rule instead (you may need an updated kernel and/or iptables binary to do this)...

iptables -A FORWARD -p tcp -m tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu

glimmerman

I'll check it out tonight, but I'm sure you're right... I think the difference in sizes has to do with ppp headers being inserted?

Anyway, cheers for that

longword

Originally posted by glimmerman
I'll check it out tonight, but I'm sure you're right... I think the difference in sizes has to do with ppp headers being inserted?

PPPoE thieves 8 bytes out of your normal ethernet frame of 1500 bytes. The MSS is the maximum TCP payload, so that's 1500 less the 8 PPPoE bytes, less the IP header (20 bytes) and less the TCP header (another 20), so you arrive at 1452.