APIC usb MPS 1.4 and the 2.4.2 kernel

APIC usb MPS 1.4 and the 2.4.2 kernel

[David DeGeorge]

I am running 2.4.2 as obtained from redhat, but I have experienced the same 
problems with a kernel compiled from the 2.4.2 sources at kernel.org.
I am experiencing troubles with enabling MPS 1.4 and USB. I have an ABIT VP6 
motherboard with two stock 733MHz PIIIs.
If I set MPS1.1 in the bios then my IOmega Photoshow usb zip drive works, the 
usb interrupt appears on irq 9 and after a day or two I experience  a hard 
(sysreq doesn't work) lock. It seems usb related since doing usb things i.e. 
mounting the drive sometimes cause the lock.
If I set MPS1.4 in the bios  then the usb interrupt appears on irq 19, whose 
count is alway zero, and the zip drive doesn't get registered. If give the 
noapic command line then things appear to work, irq=9,don't know about the 
hard locks, but booting seems much slower. Of course I can provide much more 
information but I wonder is this a common problem and what are the 
consequences of the noapic command?
David

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Pete Toscano]

[-- Attachment #1: Type: text/plain, Size: 1602 bytes --]

Well, I can't speak for the consequences of noapic (I've wondered as
much myself), but I know that there's been a problem with SMP 2.4
kernels (even the 2.4 test kernels) and USB running on VIA chipsets for
a while now.  I'm told by the linux-usb maintainers that it's a problem
with the PCI IRQ routing for the VIA chipsets, but I've been unable to
get anyone who knows about this to do anything (and I've been asking for
a while).  Alas, since this stuff is beyond me, I just accept the fact
that it'll probably always be broke.

pete

On Mon, 12 Mar 2001, David DeGeorge wrote:

> I am running 2.4.2 as obtained from redhat, but I have experienced the same 
> problems with a kernel compiled from the 2.4.2 sources at kernel.org.
> I am experiencing troubles with enabling MPS 1.4 and USB. I have an ABIT VP6 
> motherboard with two stock 733MHz PIIIs.
> If I set MPS1.1 in the bios then my IOmega Photoshow usb zip drive works, the 
> usb interrupt appears on irq 9 and after a day or two I experience  a hard 
> (sysreq doesn't work) lock. It seems usb related since doing usb things i.e. 
> mounting the drive sometimes cause the lock.
> If I set MPS1.4 in the bios  then the usb interrupt appears on irq 19, whose 
> count is alway zero, and the zip drive doesn't get registered. If give the 
> noapic command line then things appear to work, irq=9,don't know about the 
> hard locks, but booting seems much slower. Of course I can provide much more 
> information but I wonder is this a common problem and what are the 
> consequences of the noapic command?
> David

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Helge Hafting]

Pete Toscano wrote:
> 
> Well, I can't speak for the consequences of noapic (I've wondered as
> much myself), but I know that there's been a problem with SMP 2.4
> kernels (even the 2.4 test kernels) and USB running on VIA chipsets for
> a while now.  I'm told by the linux-usb maintainers that it's a problem
> with the PCI IRQ routing for the VIA chipsets, but I've been unable to
> get anyone who knows about this to do anything (and I've been asking for
> a while).  Alas, since this stuff is beyond me, I just accept the fact
> that it'll probably always be broke.

There seems to be something wrong with 2.4.2 smp even without usb.
My abit bp6 freeze after a day or so.  I have gone back to
2.4.2pre3 and this seems better.  I have to run a few more days
to be sure though.

I don't use USB and no devices malfunction - the machine just freeze
in the middle of the night and don't respond to sysrq 
or pings.  Resetting shows
nothing in the logs.

Helge Hafting

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Greg KH]

On Tue, Mar 13, 2001 at 12:25:13AM -0500, Pete Toscano wrote:
> Well, I can't speak for the consequences of noapic (I've wondered as
> much myself), but I know that there's been a problem with SMP 2.4
> kernels (even the 2.4 test kernels) and USB running on VIA chipsets for
> a while now.  I'm told by the linux-usb maintainers that it's a problem
> with the PCI IRQ routing for the VIA chipsets, but I've been unable to
> get anyone who knows about this to do anything (and I've been asking for
> a while).  Alas, since this stuff is beyond me, I just accept the fact
> that it'll probably always be broke.

Pete,

The Tyan Tiger 133 motherboard that you and I have, just will not work
in SMP mode without the noapic command line option.  Randy Dunlap and I
worked on this back and forth for a while by email, and then I drug the
machine to a Linux Advanced Topics meeting that our local LUG puts on,
and he beat on it for a few hours.

It seems that the APIC on this motherboard does not have most of the
pins connected, so that even if we could get the USB interrupt to work
properly (which we couldn't) there would be no benefit to run in APIC
mode.  I was going to run some crude benchmarks on the box with and
without APIC mode just to get an sense if we are missing anything
running in noapic mode, but I haven't gotten around to it yet.

But, Linux does seem to run just fine with USB and SMP in the noapic
mode, which is a lot better than Win2000 can say, as it doesn't even
support the VIA USB chipset on this board at all :)

Hope this helps,

greg k-h

-- 
greg@(kroah|wirex).com
http://immunix.org/~greg

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Pete Toscano]

[-- Attachment #1: Type: text/plain, Size: 1807 bytes --]



On Tue, 13 Mar 2001, Greg KH wrote:

> On Tue, Mar 13, 2001 at 12:25:13AM -0500, Pete Toscano wrote:
> > Well, I can't speak for the consequences of noapic (I've wondered as
> > much myself), but I know that there's been a problem with SMP 2.4
> > kernels (even the 2.4 test kernels) and USB running on VIA chipsets for
> > a while now.  I'm told by the linux-usb maintainers that it's a problem
> > with the PCI IRQ routing for the VIA chipsets, but I've been unable to
> > get anyone who knows about this to do anything (and I've been asking for
> > a while).  Alas, since this stuff is beyond me, I just accept the fact
> > that it'll probably always be broke.

> It seems that the APIC on this motherboard does not have most of the
> pins connected, so that even if we could get the USB interrupt to work
> properly (which we couldn't) there would be no benefit to run in APIC
> mode.  I was going to run some crude benchmarks on the box with and
> without APIC mode just to get an sense if we are missing anything
> running in noapic mode, but I haven't gotten around to it yet.

Very interesting.  I had not heard about this.  Are there any SMP boards
with a VIA chipset that does work well with Linux and USB?  I have an
old P2B-DS that I had replace with this board as I needed more PCI
slots.  Heck, for that matter are there any SMP boards that work well
with Linux and USB that have six or more PCI slots?

> But, Linux does seem to run just fine with USB and SMP in the noapic
> mode, which is a lot better than Win2000 can say, as it doesn't even
> support the VIA USB chipset on this board at all :)

How would this express itself?  I recently upgraded from WinME to Win2k
and it all _seems_ to be working well.  Where would I look to verify
this?

Thanks for the info and the update.

pete

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Greg KH]

On Tue, Mar 13, 2001 at 12:49:54PM -0500, Pete Toscano wrote:
> 
> Very interesting.  I had not heard about this.  Are there any SMP boards
> with a VIA chipset that does work well with Linux and USB?  I have an
> old P2B-DS that I had replace with this board as I needed more PCI
> slots.  Heck, for that matter are there any SMP boards that work well
> with Linux and USB that have six or more PCI slots?

If you really want USB to work in APIC mode on the Tiger 133 board,
spend $25 on a USB pci card.  That should work just fine :)

> > But, Linux does seem to run just fine with USB and SMP in the noapic
> > mode, which is a lot better than Win2000 can say, as it doesn't even
> > support the VIA USB chipset on this board at all :)
> 
> How would this express itself?  I recently upgraded from WinME to Win2k
> and it all _seems_ to be working well.  Where would I look to verify
> this?

All I know is this:
	http://support.microsoft.com/support/kb/articles/Q233/1/63.ASP
as I've not run Win2000 on this machine, thankfully....

Also the message board at:
	http://www.usbman.com
is full of comments about the VIA chipsets and the difficulties of
getting them to work properly under Windows.

> Thanks for the info and the update.

No problem, sorry I didn't get back to you sooner with this.

greg k-h

-- 
greg@(kroah|wirex).com
http://immunix.org/~greg

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Stephen Wille Padnos]

Pete Toscano wrote:
> 
> Very interesting.  I had not heard about this.  Are there any SMP boards
> with a VIA chipset that does work well with Linux and USB?  I have an
> old P2B-DS that I had replace with this board as I needed more PCI
> slots.  Heck, for that matter are there any SMP boards that work well
> with Linux and USB that have six or more PCI slots?
> 

Well, I use an old SuperMicro ( http://www.supermicro.com ) P6DNH
board.  It has 8 PCI slots, 3 ISA slots, an I960 I2O processor (four of
the PCI slots are on a secondary bus), and supports dual Pentium Pro
CPUs and 1G RAM (EDO - it's 4 years old).

They have newer boards with 6 PCI (64 bit, 66 MHz) + 1 AGP slot.  Their
boards are very high quality - though you'll pay for the reliability in
$$$.


-- 
Stephen Wille Padnos
Programmer, Engineer, Problem Solver
swpadnos@adelphia.net

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[John R Lenton]

On Tue, Mar 13, 2001 at 12:49:54PM -0500, Pete Toscano wrote:
> Very interesting.  I had not heard about this.  Are there any SMP boards
> with a VIA chipset that does work well with Linux and USB?  I have an
> old P2B-DS that I had replace with this board as I needed more PCI
> slots.  Heck, for that matter are there any SMP boards that work well
> with Linux and USB that have six or more PCI slots?

My 694D Pro (MS-6321) has been working fine once I got the heat
problem off my hands. USB works, as long as the MPS is set at
1.1. It's a SMP board with VIA's "Apollo Pro133A" chipset, and
the vt82c686a.

-- 
John Lenton (john@grulic.org.ar) -- Random fortune:
I matematici lo fanno in teoria, oppure lo portano al limite.

RE: APIC usb MPS 1.4 and the 2.4.2 kernel

[Juha Saarinen]

Greg,

:: It seems that the APIC on this motherboard does not have most of the
:: pins connected, so that even if we could get the USB interrupt to work
:: properly (which we couldn't) there would be no benefit to run in APIC
:: mode.  I was going to run some crude benchmarks on the box with and
:: without APIC mode just to get an sense if we are missing anything
:: running in noapic mode, but I haven't gotten around to it yet.

So for Tyan Tigers, is it better to compile the kernel without APIC? About
to install Linux on a dual 500 P3 here.

:: But, Linux does seem to run just fine with USB and SMP in the noapic
:: mode, which is a lot better than Win2000 can say, as it doesn't even
:: support the VIA USB chipset on this board at all :)

There's a Win2K patch for VIA chip sets now... I've got USB working under
Win2K here.

-- Juha

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[Pete Toscano]

[-- Attachment #1: Type: text/plain, Size: 1316 bytes --]



On Wed, 14 Mar 2001, Juha Saarinen wrote:

> Greg,
> 
> :: It seems that the APIC on this motherboard does not have most of the
> :: pins connected, so that even if we could get the USB interrupt to work
> :: properly (which we couldn't) there would be no benefit to run in APIC
> :: mode.  I was going to run some crude benchmarks on the box with and
> :: without APIC mode just to get an sense if we are missing anything
> :: running in noapic mode, but I haven't gotten around to it yet.
> 
> So for Tyan Tigers, is it better to compile the kernel without APIC? About
> to install Linux on a dual 500 P3 here.

AFAIK, the option to compile w/o APIC is only for UP systems.  If you
want to use both of your processors, you have to compile in APIC
support, but just disable it when loading the kernel (ie. for lilo,
'append="noapic"')

> :: But, Linux does seem to run just fine with USB and SMP in the noapic
> :: mode, which is a lot better than Win2000 can say, as it doesn't even
> :: support the VIA USB chipset on this board at all :)
> 
> There's a Win2K patch for VIA chip sets now... I've got USB working under
> Win2K here.

That would explain why it works for me.  Now, if only I didn't have
devices that need to have their BIOSes upgraded via a Windows .exe...

pete


[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

RE: APIC usb MPS 1.4 and the 2.4.2 kernel

[Juha Saarinen]

:: AFAIK, the option to compile w/o APIC is only for UP systems.  If you
:: want to use both of your processors, you have to compile in APIC
:: support, but just disable it when loading the kernel (ie. for lilo,
:: 'append="noapic"')

I haven't seen the beginning of the APIC/VIA/Tyan thread, but isn't there a
way to check if the APIC's OK, instead of resorting to workarounds like the
above?

:: That would explain why it works for me.  Now, if only I didn't have
:: devices that need to have their BIOSes upgraded via a Windows .exe...

;-).

That's a good point, actually. I can't recall a single BIOS or device BIOS
flasher that works under anything but DOS (or in some cases, Windows). Is
there any work being done on this for Linux?

-- Juha

Re: APIC usb MPS 1.4 and the 2.4.2 kernel

[idalton]

On Tue, Mar 13, 2001 at 12:25:13AM -0500, Pete Toscano wrote:
> Well, I can't speak for the consequences of noapic (I've wondered as
> much myself), but I know that there's been a problem with SMP 2.4
> kernels (even the 2.4 test kernels) and USB running on VIA chipsets for
> a while now.  I'm told by the linux-usb maintainers that it's a problem
> with the PCI IRQ routing for the VIA chipsets, but I've been unable to
> get anyone who knows about this to do anything (and I've been asking for
> a while).  Alas, since this stuff is beyond me, I just accept the fact
> that it'll probably always be broke.

I have the same problem with my Triton II dual-pentium 200 system, for
what it's worth.

-- Ferret