network connections stalls

please CC me I'm not on the list.

Hi,

I need some help with this bug.

I've got two athlon boxes which produce this same problem with all kernels from 2.4.18-pre4 back to 2.4.0 which is a far as I went.

It seems to be only triggered by a particular string of binary data which I've narrowed down (it came from a pdf document). which I'll attach now. its 735 bytes, if I change it or remove a byte it doesn't cause the problem.

The machines are both athlon 1.1GHZ on ASUS A7V133 mobo's, thats the only simularitys between them ones scsi the others ide ones a via-rhine nic and the other is a tulip.

output from tcpump:
14:28:25.184415 hathor.tsa.46893 > alhazred.tsa.1234: S 2668390347:2668390347(0) win 5840 <mss 1460,sackOK,timestamp 18451301 0,nop,wscale 0> (DF) 14:28:25.184461 alhazred.tsa.1234 > hathor.tsa.46893: S 2674935738:2674935738(0) ack 2668390348 win 5792 <mss 1460,sackOK,timestamp 102014 18451301,nop,wscale 0> (DF) 14:28:25.184616 hathor.tsa.46893 > alhazred.tsa.1234: . ack 1 win 5840 <nop,nop,timestamp 18451301 102014> (DF)
14:28:25.184954 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18451301 102014> (DF)
14:28:25.393041 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18451322 102014> (DF)
14:28:25.813051 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18451364 102014> (DF)
14:28:26.653052 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18451448 102014> (DF)
14:28:28.333082 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18451616 102014> (DF)
14:28:31.182923 hathor.tsa.46893 > alhazred.tsa.1234: F 735:735(0) ack 1 win 5840 <nop,nop,timestamp 18451901 102014> (DF)
14:28:31.182947 alhazred.tsa.1234 > hathor.tsa.46893: . ack 1 win 5792 <nop,nop,timestamp 102614 18451616,nop,nop,sack sack 1 {735:736} > (DF) 14:28:31.693138 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18451952 102614> (DF)
14:28:38.413253 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18452624 102614> (DF)
14:28:51.853463 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18453968 102614> (DF)
14:29:18.733896 hathor.tsa.46893 > alhazred.tsa.1234: P 1:735(734) ack 1 win 5840 <nop,nop,timestamp 18456656 102614> (DF)

I've tried this with both netcat and rsh, which produce the same results. I've also tested sending the same string from a solaris box which results in the same error.

I've also reduce the kernel to a bare minimum of drivers. I can send the .config if anyone wants it.

The problem does not show its self on the loopback device.

I am willing to put some effort into debugging this if someone can help me figure out where to start. If there is any other info required let me know.

Thanks.

Does the "InErrs" TCP counter in /proc/net/snmp increment on the receiver when this occurs?

It smells of bad checksumming...

Yep it increments by 7

@@ -1,8 +1,8 @@
 Ip: Forwarding DefaultTTL InReceives InHdrErrors InAddrErrors ForwDatagrams InUnknownProtos InDiscards InDelivers OutRequests OutDiscards OutNoRoutes ReasmTimeout ReasmReqds ReasmOKs ReasmFails FragOKs FragFails FragCreates
-Ip: 2 64 192 0 0 0 0 0 148 248 0 0 0 0 0 0 0 0 0
+Ip: 2 64 322 0 0 0 0 0 256 359 0 0 0 0 0 0 0 0 0
 Icmp: InMsgs InErrors InDestUnreachs InTimeExcds InParmProbs InSrcQuenchs InRedirects InEchos InEchoReps InTimestamps InTimestampReps InAddrMasks InAddrMaskReps OutMsgs OutErrors OutDestUnreachs OutTimeExcds OutParmProbs OutSrcQuenchs OutRedirects OutEchos OutEchoReps OutTimestamps OutTimestampReps OutAddrMasks OutAddrMaskReps
-Icmp: 44 0 19 0 0 0 0 25 0 0 0 0 0 25 0 0 0 0 0 0 0 25 0 0 0 0 0
+Icmp: 66 0 41 0 0 0 0 25 0 0 0 0 0 25 0 0 0 0 0 0 0 25 0 0 0 0 0
 Tcp: RtoAlgorithm RtoMin RtoMax MaxConn ActiveOpens PassiveOpens AttemptFails EstabResets CurrEstab InSegs OutSegs RetransSegs InErrs OutRsts
-Tcp: 0 0 0 0 0 0 0 0 3 144 101 2 0 0
+Tcp: 0 0 0 0 0 0 0 0 3 250 166 2 7 0
 Udp: InDatagrams NoPorts InErrors OutDatagrams
-Udp: 3 0 0 122
+Udp: 3 0 0 168