Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 11 Apr 2000 11:05:18 -0500
From:      "Duncan, Eric A." <eric@cdc.net>
To:        <freebsd-smp@freebsd.org>
Subject:   Serious SMP problems on Dell PE 2450 dual P-IIIs 733mhz
Message-ID:  <NDBBIMCDELJLMJDJPOFICEONCDAA.eric@cdc.net>

next in thread | raw e-mail | index | archive | help

Basically, I enable SMP support and the system hangs on the PCI probing on
boot up.  I originally wanted to install 3.4, but had to install 4.0 which I
seem to like more now (read below).  Here's the configuration:

Dell Power Edge 2450
Dual 733mhz 133mhz bus P-IIIs
512megs
Adaptec 7899 U2W SCSI (2940 ahc driver)
  ^- Important info below.

Here's the story:

FreeBSD 3.4
----------------
Upon the setup from the 3.4 floppies (ftp install), the kernel never found
the SCSI controller (ahc0).  It never even probed it during the PCI scan.  I
even tried a 3.3-STABLE kernel without success.  I built a custom kernel on
another box including the ahc0 driver.  I even tried wiring down devices,
calling scsibus, setting them manually, setting with '?'s.  I tried it all
and built a new kernel every time.  Even making sure the boot loader wasn't
di'ing it.  All kernels were gzipped as kernel.gz and replaced the kernel.gz
off of the boot floppies at first.  This was how I was able to boot the new
kernel off of the floppies.  I even tried downloading and created the ISOs
for the 3.4 and 4.0 kernels without success.

The ahc controller was never probed upon boot.  It never even showed 'ahc0:
not found' when I had it compiled in the kernel.  Nothing.  Very odd!

FreeBSD 4.0
----------------
Immediately FBSD4 found the ahc0 controller and loaded the correct driver.
It found my tape backup devices, ata cdrom, and 2 da 18gig SCSI drives
without any problems.

I continued to install the kernel, ports, packages, etc.  The system is
running great with all devices configured!

I tried to build a custom kernel.  To disable drivers and clean up the
kernel, firewall, smp, etc.  Everything was going great and running smooth.
Until the SMP part.  As I mentioned above, the kernel halts, with no errors,
it simply locks the machine up while booting.  I tried even using the
GENERIC kernel, which enables a lot and works just fine.  Until I simply
enable SMP support.

And example SMP setup I tried is (which seems to be what the system asks
for):
options SMP
options APIC_IO
options NCPU=2
options NBUS=4
options NAPIC=2
options NINTR=28

Yes, I tried disabling all but the first two lines.  The kernel would halt
immediately telling me I has more then 1 apic and I had to rebuild my kernel
(it numbered NAPIC=2 on the error).  It also reported back errors that I had
4 buses and to increase my NBUS=4 and that I had 28 intr and to increase
NINTR to 28.  I even tried to build a very minimal kernel, disabling
everything but the SMP.  Same results.  It locks up every time after the
ISA configuration of devices, before the 'waiting 15seconds for scsi devices
to settle' statement.

Now, I haven't tried OVER SIZING any of the above.  I had to ship the
servers off today.  But I will be back working on them in a few days and
would love any advice.

Another odd story.  Just for argument sake, I compiled a 3.3-STABLE kernel
with SMP support with the params above, and the floppies loaded it correctly
enabling both CPU0 and CPU1 on boot!  Just no SCSI card/Harddrives were
found because of the 3.x problem of not detecting my SCSI board.

So, I now have a 3.x kernel with SMP but no HD.  And a 4.0 kernel with HDs,
but no SMP support.

Yes, I tried changing all sorts of settings in the BIOS and Utility config.
Changing mem IOs, IRQs, disabling the second bus, etc etc etc.  Everything
combination I can think of.  Disabling the IDE bus (used for the CDROM),
disabling the floppy, etc.  I had another PowerEdge 2450 with dual P-IIIs
600mhz and same controller, but with embedded RAID from Dell.  The
3.3-KERNEL never found the HDs, and the 4.0 kernel never enabled the SMP.
Same results so it isn't a hardware fault I believe.

HELP!  A $1,200 second CPU laying to waste in this system isn't too good at
all.


Thanks in advance,
Eric Duncan

eduncan@idealmusic.com



To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-smp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?NDBBIMCDELJLMJDJPOFICEONCDAA.eric>