disaster with 2.4.14+preempt

[J Sloan <jjs@pobox.com>]

HI all,

I upgraded my test box (Compaq 6500,
4 CPUs, 1.2 GB RAM) to Red Hat 7.2,
which went very nicely. It seems stable
and tux is very very fast.

I have been pleased with 2.4.14 on several
desktop systems, so my next step was to
build 2.4.14 on the Compaq - I decided to
apply the 2.4.14-1-preempt-patch as well.

I built/booted the new kernel, and started
my usual dbench run to check it out -

Ironically, I then witnessed a disastrous
file corruption, the likes of which I'd never
seen in years of using ext2 - in short, the
old lockup that was previously fixed is back,
nastier than ever - and this time it caused
the box to scribble on /usr/lib before dying.

A few seconds after starting "dbench 16",
the server locked hard. (When it came back
up, many programs were not functional, with
claims of invalid or corrupt libraries)

I was however able to hand copy the oops
from the console and run through ksymoops
after bringing the box backup again:

Invalid operand: 0000
CPU:    2
EIP:    0010:[<c01300ee>]  Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010082
eax: 00010086   ebx: ea416dap   ecx: c0269f20   edx: 00000004
esi: c1a9e1c0   edi: f7bb3000   ebp: 00000000   esp: c2213eb8
ds: 0018   es: 0018  ss: 0018
Process swapper (pid:0, stackpage=c2213000)
Stack: c0238c3e 0000004a 00000082 00000001 00000003 00000046 f628fbc0
ea416da0
       00000082 f7bb3000 00000000 c0135ab6 c019ed8a c0423f00 00000000
00000001
       00000000 f7b916dc c019edef ea416da0 00000000 00000012 00000000
00000040
Call Trace:  [<c0135ab6>] [<c019ed8a>] [<c019edef>] [<c0108a3e>]
[<c0108ccb>]
[<c0105240>] [<c0105240>] [<c0105240>] [<c010526c>] [<c01052f2>]
[<c011611f>]
Code: 0f 0b 58 5a 8b 7e 08 85 ff 74 10 6a 4c 68 3e 8c 23 c0 e8 bb

>>EIP; c01300ee <__free_pages_ok+1e/2e0>   <=====
Trace; c0135ab6 <bounce_end_io_write+46/f0>
Trace; c019ed8a <do_ida_intr+1ba/2c0>
Trace; c019edef <do_ida_intr+21f/2c0>
Trace; c0108a3e <handle_IRQ_event+5e/90>
Trace; c0108ccb <do_IRQ+bb/140>
Trace; c0105240 <default_idle+0/40>
Trace; c0105240 <default_idle+0/40>
Trace; c0105240 <default_idle+0/40>
Trace; c010526c <default_idle+2c/40>
Trace; c01052f2 <cpu_idle+52/70>
Trace; c011611f <printk+15f/1a0>
Code;  c01300ee <__free_pages_ok+1e/2e0>
00000000 <_EIP>:
Code;  c01300ee <__free_pages_ok+1e/2e0>   <=====
   0:   0f 0b                     ud2a      <=====
Code;  c01300f0 <__free_pages_ok+20/2e0>
   2:   58                        pop    %eax
Code;  c01300f1 <__free_pages_ok+21/2e0>
   3:   5a                        pop    %edx
Code;  c01300f2 <__free_pages_ok+22/2e0>
   4:   8b 7e 08                  mov    0x8(%esi),%edi
Code;  c01300f5 <__free_pages_ok+25/2e0>
   7:   85 ff                     test   %edi,%edi
Code;  c01300f7 <__free_pages_ok+27/2e0>
   9:   74 10                     je     1b <_EIP+0x1b> c0130109
<__free_pages_ok+39/2e0>
Code;  c01300f9 <__free_pages_ok+29/2e0>
   b:   6a 4c                     push   $0x4c
Code;  c01300fb <__free_pages_ok+2b/2e0>
   d:   68 3e 8c 23 c0            push   $0xc0238c3e
Code;  c0130100 <__free_pages_ok+30/2e0>
  12:   e8 bb 00 00 00            call   d2 <_EIP+0xd2> c01301c0
<__free_pages_ok+f0/2e0>

<0>Kernel panic: Aiee, killing interrupt handler!

In interrupt handler - not syncing