pre-2.0.31 problems

Jon Lewis (
Tue, 13 May 1997 00:48:30 -0400 (EDT)

I've got pre-2.0.31 on my test box and it's not doing well with synfloods.

validated probe(1c30e5cd, 1730e5cd, 19460, 25, 783464920)
validated probe(1c30e5cd, 1730e5cd, 19716, 25, 1121709911)
eth0: Too much work at interrupt, status 0x01
eth0: Too much work at interrupt, status 0x01
eth0: Too much work at interrupt, status 0x01
eth0: Too much work at interrupt, status 0x01
invalid operand: 0000
CPU: 0
EIP: 0010:[<0008c00c>]
EFLAGS: 00010217
eax: 00089ed4 ebx: 0376c414 ecx: 03a6cdf4 edx: 03a6c000
esi: 00000124 edi: 00000000 ebp: 00000246 esp: 03d4ff14
ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018
Process sshd (pid: 90, process nr: 25, stackpage=03d4f000)
Stack: 00141483 0376c414 0376c414 03a6ce8c 001417ab 0376c414 03a6ce8c
00000000 0381a858 00141c32 03a6ce8c 00000000 03a6ce8c 0381a818
03a6ce8c 00000000 03a6ce8c 0381a840 0381a818 00000202 03d4ffa0
Call Trace: [<00141483>] [<001417ab>] [<00141c32>] [<001562db>]
[<00156776>] [<00156688>] [<00116959>]
[<0011d5b7>] [<0010ab4b>]
Code: ff ff ff ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
Aiee, killing interrupt handler

Using `/kernels/' to map addresses to symbols.

Trace: 141483 <sock_wfree+23/30>
Trace: 1417ab <kfree_skb+b7/f8>
Trace: 141c32 <dev_kfree_skb+3e/4c>
Trace: 1562db <arp_purge_send_q+53/64>
Trace: 156776 <arp_expire_request+ee/1d8>
Trace: 156776 <arp_expire_request+ee/1d8>
Trace: 116959 <timer_bh+2ed/334>
Trace: 11d5b7 <do_bottom_half+3b/60>
Trace: 10ab4b <handle_bottom_half+b/20>

Code: (bad)
Code: (bad)
Code: (bad)
Code: incl (%eax)
Segmentation fault

This is stock 2.0.30 + pre-patch-2.0.31.gz + some softdog.c changes of
mine + an ext2 debug patch from Gerard + the O'Reilly >256fd kmalloc fds
patch. It was compiled SMP for my SMP Tomcat II. All these additional
patches have been in use on my systems for some time (the kmalloc fds
patch not as long as others) and I don't think they're the problem as
other systems seem happy with them.

After a reboot, I tried to reproduce this, and did have some trouble doing
so, but with persistance and a synflooder program, I was able to badly
hose the system and detect that there still seem to be memory leaks

sloth:~# free
total used free shared buffers cached
Mem: 63116 61900 1216 5548 524 3500
-/+ buffers: 57876 5240
Swap: 62216 0 62216

There's basically nothing but bare bones basics running:

I managed to get another EIP dump, but the system had gotten so slow I
decided to reboot before using ksymoops, and shutdown apparently failed.
The system became unusable (I can ping it, but nothing else) so I won't be
able to get to it for ksymoops decoding until tomorrow.

FYI...this is a Tomcat II (dual P120), NCR810, IBM 540mb DALS SCSI disk,
SMC Ultra Elite16T, ATI Mach64, SB16.

Jon Lewis <> | Unsolicited commercial e-mail will
Network Administrator | be proof-read for $199/hr.
________Finger for PGP public key_______