ksymoops.cc fails compile with pentium-gcc; oops with samba

Teunis Peters (teunis@usa.net)
Tue, 29 Jul 1997 11:31:12 -0600 (MDT)

gcc [or g++] ksymoops.cc -o ksymoops
fails with 'Internal compiler error'

I now have a ksymoops thanks to trevor@jpj.net :)

Thank you all who provided g++ libs :) I guess I have to keep a closer
eye on glibc-2 updates....

oops with samba ksym-oopsed to :
Using `/kernel/System.2.1.43' to map addresses to symbols.

>>EIP: c0175f4e <dev_ifsioc+162/4f8>
Trace: c0176636 <net_alias_devsetup+62/15c>
Trace: c019154a <inet_autobind+76/9c>
Trace: c0171095 <sys_socketpair+35/34c>
Trace: c012c3bf <sys_getdents+103/258>
Trace: c01095da <ret_with_reschedule+19/2f>

Code: c0175f4e <dev_ifsioc+162/4f8>

[and the 'oops']:
sigil kernel: Unable to handle kernel paging request at virtual address 20202020
- this line is very unhealthy! (it's also sometimes other bad numbers like 0)
sigil kernel: current->tss.cr3 = 00be2000, (r3 = 00be2000
sigil kernel: *pde = 00000000
sigil kernel: Oops: 0000
sigil kernel: CPU: 0
sigil kernel: EIP: 0010:[<c0175f4e>]
sigil kernel: EFLAGS: 00010212
sigil kernel: eax: 00000000 ebx: c1816764 ecx: 00000000 edx: 20202020
sigil kernel: esi: 20202020 edi: c0d43f38 ebp: 00000008 esp: c0d43f00
sigil kernel: ds: 0018 es: 0018 ss: 0018
sigil kernel: Process smbclient (pid: 985, process nr: 55, stackpage=c0d43000)
sigil kernel: Stack: bfffe77c c0744860 bfffe77c 00008912 c086d968 c0db7040 c0d42000 00000020
sigil kernel: 00000000 c0d43f38 000003e0 bfffe7a4 00000400 bfffe784 00000000 00000000
sigil kernel: 00000000 00000000 00000000 00000000 00000000 00000000 c0176636 bfffe77c
sigil kernel: Call Trace: [<c0176636>] [<c019154a>] [<c0171095>] [<c012c3bf>] [<c01095da>]
sigil kernel: Code: ac aa 84 c0 75 fa 66 8b 43 46 66 89 44 24 48 8b 4b 64 89 4c

Any idea how to trace this?
(I can regenerate with 2.1.44pre3 or the like too... [or 2.1.42 or 2.1.41
or ... but I erased those])

Incidentally, while I _DO_ use pentium-gcc; I use _PRECOMPILED_
glibc/glibc++ (thanks all :) - just in case.

CFLAGS/HOSTSCFLAGS -O2 -fomit-frame-pointer \
-mno-schedule-prologue -fno-peephole

for the most part I don't see any problems. Samba seems to be a wildcard.
I run a LOT of other things compiled this way and samba is the ONLY thing
to have failed (recently) - 2.1.47 got rid of the X problems <grin>.
[X compiled with pentium-gcc because precompiled binaries failed]

FWIW it doesn't seem to matter if smbfs is compiled or not.
(though judging from the trace it's the network ioctl that's failing)

Suspicion is that samba's generating an invalid pointer (but how?)

- Teunis

Oh, I do have ONE program that farely consistently generates bad code but
I suspect it's my program's fault <grin> - not that I can remember how to
generate a debug trace from pentium-gcc <sigh>. [I keep meaning to send