SAMBA-1-0-4

[Doug Bissett]

I installed samba-1-0-4.wpi on my ThinkPad, when I installed eCS 2.0
GA. It seems to work fine. Then, I tried it on my quad core system,
with a new install of eCS 2.0 GA. When I attempt to start the SAMBA
server, I get this (from PROCDUMP):
==========================
IBM OS/2 Dump Formatter for a retail or an hstrict SMP kernel.
Formatter is --> Internal revision 14.103a_SMP
Dump file is --> Internal revision 14.104a_SMP (process dump)

Symbol (c:\os2\pdpsi\pmdf\warp45_s\os2krnlr.sym) linked
Symbol (c:\os2\pdpsi\pmdf\warp45_s\pmviop.sym) linked
Symbol (c:\os2\pdpsi\pmdf\warp45_s\pmspl.sym) linked
Symbol (c:\os2\pdpsi\pmdf\warp45_s\pmctls.sym) linked
Symbol (c:\os2\pdpsi\pmdf\warp45_s\pmmerge.sym) linked
Symbol (c:\os2\pdpsi\pmdf\warp45_s\pmwp.sym) linked
Symbol (c:\os2\pdpsi\pmdf\warp45_s\doscall1.sym) linked

Current slot number: 0096

Slot Pid Ppid Csid Ord Sta Pri pTSD pPTDA pTCB Disp SG
Name
*0096# 0064 0062 0064 0001 run 0300 f9095000 fe6345b8 f972ae70 0bf0 1c
SMBD

eax=00000000 ebx=00000000 ecx=00000000 edx=00000001 esi=0042f370
edi=0042f960
eip=1d1c56cb esp=0042f330 ebp=0042f3b8 iopl=0 -- -- -- nv up ei pl nz
na po nc
cs=005b ss=0053 ds=0053 es=0053 fs=150b gs=0000 cr2=00000000
cr3=00000000 p=02
005b:1d1c56cb 89c3 mov ebx,eax
#
Loading version data ... please wait
Loading Process Information ... please wait


Loading structure info for WSeB/ACP SMP ... please wait.
....Structure definition file (warp45_s\kernel.sdf) loaded.
....Structure definition file (warp45_s\shell.sdf) loaded.
Done.
===============================
(there may be some line wraps). Nothing in POPUPLOG.OS2.

I back leveled the SAMBA server core files section to what was in
samba-1-0-3.wpi and it is now working great.

[Herwig Bauernfeind]

Post by Doug Bissett
I installed samba-1-0-4.wpi on my ThinkPad, when I installed eCS 2.0
GA. It seems to work fine. Then, I tried it on my quad core system,
with a new install of eCS 2.0 GA. When I attempt to start the SAMBA

Well, this is not unexpected.

The quadcore bug is located in the OS/2-specific random number
generator, which which was "borrowed" from the Apache webserver (which
at least in theory should suffer from the same defect on quadcore
machines). Virtually the bug affects most binaries both in the Samba
command line utility package and in the Samba server core file package
(all the GUI stuff is not affected!).

Please, note that the bug will not show up on single core and dualcore
machines (it was hunted down, because I now have 2 identical machines at
hand, one has a dualcore, one has a quadcore Xeon).

Thanks,
Herwig B.

[Doug Bissett]

On Tue, 8 Jun 2010 06:05:15 UTC, Herwig Bauernfeind

Post by Herwig Bauernfeind

Post by Doug Bissett
I installed samba-1-0-4.wpi on my ThinkPad, when I installed eCS 2.0
GA. It seems to work fine. Then, I tried it on my quad core system,
with a new install of eCS 2.0 GA. When I attempt to start the SAMBA

Well, this is not unexpected.

It was because eCS 2.0 GA is supposed to be stable for as many
machines as possible. However, that isn't your problem. In my
situation, I had just got back from a trip, so a working LAN
attachment was absolutely required to get my systems synchronized
again. I had to go back to the PEER networking, temporarily, since I
didn't have the time to play with SAMBA.

Post by Herwig Bauernfeind
The quadcore bug is located in the OS/2-specific random number
generator, which which was "borrowed" from the Apache webserver (which
at least in theory should suffer from the same defect on quadcore
machines). Virtually the bug affects most binaries both in the Samba
command line utility package and in the Samba server core file package
(all the GUI stuff is not affected!).

Okay. It would seem that a truely "random" number is really not
required in a case like this. Something a little simpler might be in
order.

Post by Herwig Bauernfeind
Please, note that the bug will not show up on single core and dualcore
machines (it was hunted down, because I now have 2 identical machines at
hand, one has a dualcore, one has a quadcore Xeon).

Great. I will leave it to you to figure it out, but I will be willing
to test the results. Meanwhile, I am happy with the performance of
both the 1.0.3 server core and the updated client.

FWIW, I wouldn't bet that it won't show up in a dual core machine (and
possibly even on a single core). It probably depends on a lot of
things, like machine speed and machine load, to make it happen. Fewer
cores probably make it less likely, but it is probably not impossible.

Post by Herwig Bauernfeind
Thanks,
Herwig B.

Good luck...