SMP Woes

SMP Woes

[Bobby Hitt]

Hello,

I recently posted this message to the linux-kernel list:
------
I have an ASUS CUV4XD dual CPU mobo that I've been using with Window 2000
Professional for the last two years. It worked fine, W2K used both CPUs with
no problems whatsoever. I recently upgraded my W2K system, and decided to
use the ASUS board in a system I was building to use as a Linux gateway. The
problem I'm having is the system refuses to boot when I created a kernel
with SMP support. If I turn off SMP and rebuild the kernel, boots fine. I
have two other SMP systems, very old mother boards, 5+ years old. They both
run Linux fine with SMP. I even took the source tree from a working system,
built a new kernel, locks up on bootup. The last message displayed :

Total of 2 processors activated ( 3186.68 BogoMIPS)
Enabling IO-APIC IRQs
Setting 2 in the phys_id_present_map
... changing IO-APIC physical APIC ID to 2 ... ok.

I contacted Alan Cox about this, he said he was working on his MBa and
hadn't been involved with kernel development for a while. He suggest adding:

append="noapic pci=usepirqmask"

to the lilo.conf file, which I did. No change.

Any body have similar problems or suggestions?
------

This was for kernel 2.4.23.

I've since figured out that problem. I found that the BIOS had an option
"MPS 1/4 Support" which was checked. Turned that off, and now the system
boots fine, but now have a new problem with my RAID configuration. When
"raidstart /dev/md0" runs, it gives this error:

cannot determine md version: 6

The only thing I've done is to enable smp support in the kernel, which
brings up another issue. I had just installed the source tree for 2.4.26.
smp support was already checked, so all I did was to enable raid, reiserfs
support, etc. I didn't turn on anything un-necessary. With "vga=791" set in
the lilo.conf file, on reboot now all I get is a black screen and system
lockup. When I went back to turn off smp, now the build fails with lots of
"num_cpus" (or close to that) appear everywhere. The only way I can now
build a new kernel is to save my .config file, wipe the directory,
re-install the source tree, copy my .config file back and rebuild. With
regard to the display issue, I'll disable the display for now. The raid
issue is the main thing.

TIA,

Bobby

Re: SMP Woes

[Vladimir G. Ivanovic]

I have an ASUS CUV4X-DLS dual-processor motherboard that I've been
running for three years (?).

I use grub; here is my grub.conf file:

   timeout=10
   splashimage=(hd0,0)/grub/splash.xpm.gz
   default=0

   title Fedora Core (2.4.22-1.2179.nptlsmp)
           root (hd0,0)
           kernel /vmlinuz-2.4.22-1.2179.nptlsmp root=/dev/sda4 hdb=ide-scsi ide2=0x9400,0x9002 ide3=0x8800,0x8402
           initrd /initrd-2.4.22-1.2179.nptlsmp.img

   title Fedora Core (2.4.22-1.2174.nptlsmp)
           root (hd0,0)
           kernel /vmlinuz-2.4.22-1.2174.nptlsmp root=/dev/sda4 hdb=ide-scsi ide2=0x9400,0x9002 ide3=0x8800,0x8402
           initrd /initrd-2.4.22-1.2174.nptlsmp.img

   title Memtest86 v3.1
           kernel /memtest86-3.1

But to get Linux SMP to run, I had to turn off MPS 1.4 support in the
BIOS. I also turn off automatic IDE detection, but then enable the
Primary IDE channel. (I have an 133 MHz IDE controller hooked up to a
200 GB drive.)

Let me know if you want more details about my setup.

--- Vladimir

------------------------------------------------------------------------
Vladimir G. Ivanovic                        http://leonora.org/~vladimir
2770 Cowper St.                                         vladimir@acm.org
Palo Alto, CA 94306-2447                                 +1 650 678 8014
------------------------------------------------------------------------

>>>>> "bh" == Bobby Hitt <Bob.Hitt@bscnet.com> writes:

    bh> Hello,
    bh> 
    bh> I recently posted this message to the linux-kernel list:
    bh> ------
    bh> I have an ASUS CUV4XD dual CPU mobo that I've been using with Window 2000
    bh> Professional for the last two years. It worked fine, W2K used both CPUs with
    bh> no problems whatsoever. I recently upgraded my W2K system, and decided to
    bh> use the ASUS board in a system I was building to use as a Linux gateway. The
    bh> problem I'm having is the system refuses to boot when I created a kernel
    bh> with SMP support. If I turn off SMP and rebuild the kernel, boots fine. I
    bh> have two other SMP systems, very old mother boards, 5+ years old. They both
    bh> run Linux fine with SMP. I even took the source tree from a working system,
    bh> built a new kernel, locks up on bootup. The last message displayed :
    bh> 
    bh> Total of 2 processors activated ( 3186.68 BogoMIPS)
    bh> Enabling IO-APIC IRQs
    bh> Setting 2 in the phys_id_present_map
    bh> .. changing IO-APIC physical APIC ID to 2 ... ok.
    bh> 
    bh> I contacted Alan Cox about this, he said he was working on his MBa and
    bh> hadn't been involved with kernel development for a while. He suggest adding:
    bh> 
    bh> append="noapic pci=usepirqmask"
    bh> 
    bh> to the lilo.conf file, which I did. No change.
    bh> 
    bh> Any body have similar problems or suggestions?
    bh> ------
    bh> 
    bh> This was for kernel 2.4.23.
    bh> 
    bh> I've since figured out that problem. I found that the BIOS had an option
    bh> "MPS 1/4 Support" which was checked. Turned that off, and now the system
    bh> boots fine, but now have a new problem with my RAID configuration. When
    bh> "raidstart /dev/md0" runs, it gives this error:
    bh> 
    bh> cannot determine md version: 6
    bh> 
    bh> The only thing I've done is to enable smp support in the kernel, which
    bh> brings up another issue. I had just installed the source tree for 2.4.26.
    bh> smp support was already checked, so all I did was to enable raid, reiserfs
    bh> support, etc. I didn't turn on anything un-necessary. With "vga=791" set in
    bh> the lilo.conf file, on reboot now all I get is a black screen and system
    bh> lockup. When I went back to turn off smp, now the build fails with lots of
    bh> "num_cpus" (or close to that) appear everywhere. The only way I can now
    bh> build a new kernel is to save my .config file, wipe the directory,
    bh> re-install the source tree, copy my .config file back and rebuild. With
    bh> regard to the display issue, I'll disable the display for now. The raid
    bh> issue is the main thing.
    bh> 
    bh> TIA,
    bh> 
    bh> Bobby
    bh> 
    bh> -
    bh> To unsubscribe from this list: send the line "unsubscribe linux-smp" in
    bh> the body of a message to majordomo@vger.kernel.org
    bh> More majordomo info at  http://vger.kernel.org/majordomo-info.html
    bh> 

Re: SMP Woes

[Bobby Hitt]

Howdy,

I figured out the MPS 1.4 issue, it was enabled in my BIOS. I just
downloaded 2.6.5 and installed it. Everything works except the video mode. I
have "vga=791" in lilo.conf, which has worked on every kernel I've ever
built. 2.6.5 just ignores it. Ireadup a little on the vesafb, it has a lot
more stuff about configuring then I've seen before. In a bit I'm going to
reboot with "vga=ask" to see if I can use something different.

Bobby

----- Original Message ----- 
From: "Vladimir G. Ivanovic" <vladimir@leonora.org>
To: "Bobby Hitt" <Bob.Hitt@bscnet.com>
Cc: <linux-smp@vger.kernel.org>
Sent: Thursday, April 22, 2004 1:21 PM
Subject: Re: SMP Woes


> I have an ASUS CUV4X-DLS dual-processor motherboard that I've been
> running for three years (?).
>
> I use grub; here is my grub.conf file:
>
>    timeout=10
>    splashimage=(hd0,0)/grub/splash.xpm.gz
>    default=0
>
>    title Fedora Core (2.4.22-1.2179.nptlsmp)
>            root (hd0,0)
>            kernel /vmlinuz-2.4.22-1.2179.nptlsmp root=/dev/sda4
hdb=ide-scsi ide2=0x9400,0x9002 ide3=0x8800,0x8402
>            initrd /initrd-2.4.22-1.2179.nptlsmp.img
>
>    title Fedora Core (2.4.22-1.2174.nptlsmp)
>            root (hd0,0)
>            kernel /vmlinuz-2.4.22-1.2174.nptlsmp root=/dev/sda4
hdb=ide-scsi ide2=0x9400,0x9002 ide3=0x8800,0x8402
>            initrd /initrd-2.4.22-1.2174.nptlsmp.img
>
>    title Memtest86 v3.1
>            kernel /memtest86-3.1
>
> But to get Linux SMP to run, I had to turn off MPS 1.4 support in the
> BIOS. I also turn off automatic IDE detection, but then enable the
> Primary IDE channel. (I have an 133 MHz IDE controller hooked up to a
> 200 GB drive.)
>
> Let me know if you want more details about my setup.
>
> --- Vladimir
>
> ------------------------------------------------------------------------
> Vladimir G. Ivanovic                        http://leonora.org/~vladimir
> 2770 Cowper St.                                         vladimir@acm.org
> Palo Alto, CA 94306-2447                                 +1 650 678 8014
> ------------------------------------------------------------------------
>
> >>>>> "bh" == Bobby Hitt <Bob.Hitt@bscnet.com> writes:
>
>     bh> Hello,
>     bh>
>     bh> I recently posted this message to the linux-kernel list:
>     bh> ------
>     bh> I have an ASUS CUV4XD dual CPU mobo that I've been using with
Window 2000
>     bh> Professional for the last two years. It worked fine, W2K used both
CPUs with
>     bh> no problems whatsoever. I recently upgraded my W2K system, and
decided to
>     bh> use the ASUS board in a system I was building to use as a Linux
gateway. The
>     bh> problem I'm having is the system refuses to boot when I created a
kernel
>     bh> with SMP support. If I turn off SMP and rebuild the kernel, boots
fine. I
>     bh> have two other SMP systems, very old mother boards, 5+ years old.
They both
>     bh> run Linux fine with SMP. I even took the source tree from a
working system,
>     bh> built a new kernel, locks up on bootup. The last message displayed
:
>     bh>
>     bh> Total of 2 processors activated ( 3186.68 BogoMIPS)
>     bh> Enabling IO-APIC IRQs
>     bh> Setting 2 in the phys_id_present_map
>     bh> .. changing IO-APIC physical APIC ID to 2 ... ok.
>     bh>
>     bh> I contacted Alan Cox about this, he said he was working on his MBa
and
>     bh> hadn't been involved with kernel development for a while. He
suggest adding:
>     bh>
>     bh> append="noapic pci=usepirqmask"
>     bh>
>     bh> to the lilo.conf file, which I did. No change.
>     bh>
>     bh> Any body have similar problems or suggestions?
>     bh> ------
>     bh>
>     bh> This was for kernel 2.4.23.
>     bh>
>     bh> I've since figured out that problem. I found that the BIOS had an
option
>     bh> "MPS 1/4 Support" which was checked. Turned that off, and now the
system
>     bh> boots fine, but now have a new problem with my RAID configuration.
When
>     bh> "raidstart /dev/md0" runs, it gives this error:
>     bh>
>     bh> cannot determine md version: 6
>     bh>
>     bh> The only thing I've done is to enable smp support in the kernel,
which
>     bh> brings up another issue. I had just installed the source tree for
2.4.26.
>     bh> smp support was already checked, so all I did was to enable raid,
reiserfs
>     bh> support, etc. I didn't turn on anything un-necessary. With
"vga=791" set in
>     bh> the lilo.conf file, on reboot now all I get is a black screen and
system
>     bh> lockup. When I went back to turn off smp, now the build fails with
lots of
>     bh> "num_cpus" (or close to that) appear everywhere. The only way I
can now
>     bh> build a new kernel is to save my .config file, wipe the directory,
>     bh> re-install the source tree, copy my .config file back and rebuild.
With
>     bh> regard to the display issue, I'll disable the display for now. The
raid
>     bh> issue is the main thing.
>     bh>
>     bh> TIA,
>     bh>
>     bh> Bobby
>     bh>
>     bh> -
>     bh> To unsubscribe from this list: send the line "unsubscribe
linux-smp" in
>     bh> the body of a message to majordomo@vger.kernel.org
>     bh> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>     bh>
>