2001-12-24 09:50:43

by Juergen Sauer

[permalink] [raw]
Subject: VIA Chipsets + USB + SMP == UGLY TRASH

Hi!
Merry X-Mas everywhere !

So, my USB tryout is over.
This is the expierience report:
You should not try to use VIA Chipsets + SMP + USB, that's
the worst thinkable idea. It's junk (usb-Part).

That's why:
1. not solved USB Irq errors in APIC mode, causes:
Error -110, device does not accept ID
USB Host is recognized fine, no device is attaced

This is an error somewhere in the Kernel APIC Irq routing, which may
worked around with "append noapic pirq="your irq" but using such a cutdown
USB System is not a good idea, no relly working bulk-transfers (forget
any devices which depend from it: scanners, camera, sound, isdn,
harddisks, zip etc.)

State: unusable.
Timeframe to be fixed: Unkonwn, Error Reports open open since 1 year

So now I shutdown an connect the scanner (HP 6350) with a 2nd scsi card.

Ciao
J?rgen Sauer

--
J?rgen Sauer - AutomatiX GmbH, +49-4209-4699, [email protected] **
** Das Linux Systemhaus - Service - Support - Server - L?sungen **
http://www.automatix.de to Mail me: remove: -not-for-spawm- **


2001-12-24 10:54:02

by Jens Gecius

[permalink] [raw]
Subject: Re: VIA Chipsets + USB + SMP == UGLY TRASH

Juergen Sauer <[email protected]> writes:

> Hi!
> Merry X-Mas everywhere !

Same to you.

> So, my USB tryout is over.
> This is the expierience report:
> You should not try to use VIA Chipsets + SMP + USB, that's
> the worst thinkable idea. It's junk (usb-Part).

Well, it's not junk, I'd say. It works fine here on my gigabyte
dualboard.

> That's why:
> 1. not solved USB Irq errors in APIC mode, causes:
> Error -110, device does not accept ID
> USB Host is recognized fine, no device is attaced

I used to have problems with APIC and my network card, but that's
over. APIC works just fine with my box (USB mouse, USB scanner).

> This is an error somewhere in the Kernel APIC Irq routing, which may
> worked around with "append noapic pirq="your irq" but using such a
> cutdown

As I was told by Alan, it's not that "cutdown" as you would expect.

--
Tschoe, http://gecius.de/gpg-key.txt - Fingerprint:
Jens 1AAB 67A2 1068 77CA 6B0A 41A4 18D4 A89B 28D0 F097

2001-12-24 14:20:50

by Juergen Sauer

[permalink] [raw]
Subject: Re: VIA Chipsets + USB + SMP == UGLY TRASH

Am Montag, 24. Dezember 2001 11:53 schrieb Jens Gecius:
> Juergen Sauer <[email protected]> writes:
> > Hi!
> > Merry X-Mas everywhere !

> Well, it's not junk, I'd say. It works fine here on my gigabyte
> dualboard.
Exact Chipset ?
This makes many troubles:
00:00.0 Host bridge: VIA Technologies, Inc. VT82C691 [Apollo PRO] (rev c4)
Flags: bus master, medium devsel, latency 8
Build in in Epox epd3va.

> I used to have problems with APIC and my network card, but that's
> over. APIC works just fine with my box (USB mouse, USB scanner).

> As I was told by Alan, it's not that "cutdown" as you would expect.
Would be fine. ;->>

The IRQ routing was struggling over a lousy pci-host-bridge and an idiotic
(and ugly badly documented) irq routing on the epox board.

This one runns only with "noapic pirq=5,10,11,15" and fires up to much
irqs onto the same one, leaving other possible resoures unused. Just have
a look on irq 10, there are three sources of irq working. This fails
sometimes. No wonder why. ;-<<
Also was it impossible to get suitable usb-bus-bandwith-performance for
the scanner (HP 6350) to get it scanning withouts stopping and
re-starting (buffer transmission to sane/xsane).

jojo@pc2:jojo $ uname -a
Linux pc2 2.4.16-xfs #7 SMP Sam Dez 1 17:46:18 CET 2001 i686 unknown

jojo@pc2:jojo $ cat /proc/interrupts
CPU0 CPU1
0: 606544 0 XT-PIC timer
1: 12278 0 XT-PIC keyboard
2: 0 0 XT-PIC cascade
5: 216 0 XT-PIC usb-uhci
8: 3 0 XT-PIC rtc
10: 316151 0 XT-PIC ide2, Ensoniq AudioPCI, nvidia
11: 102769 0 XT-PIC sym53c8xx
12: 72420 0 XT-PIC PS/2 Mouse
14: 1070229 0 XT-PIC eth0, saa7146(1)
15: 2 0 XT-PIC ide1
NMI: 0 0
LOC: 606446 606488
ERR: 0
MIS: 0
jojo@pc2:jojo $


mfG
J. Sauer
--
J?rgen Sauer - AutomatiX GmbH, +49-4209-4699, [email protected] **
** Das Linux Systemhaus - Service - Support - Server - L?sungen **
http://www.automatix.de to Mail me: remove: -not-for-spawm- **

2001-12-24 18:52:05

by Greg KH

[permalink] [raw]
Subject: Re: [Linux-usb-users] VIA Chipsets + USB + SMP == UGLY TRASH

On Mon, Dec 24, 2001 at 10:32:49AM +0100, Juergen Sauer wrote:
> Hi!
> Merry X-Mas everywhere !
>
> So, my USB tryout is over.
> This is the expierience report:
> You should not try to use VIA Chipsets + SMP + USB, that's
> the worst thinkable idea. It's junk (usb-Part).

Depends on the motherboard. What one do you have?

> That's why:
> 1. not solved USB Irq errors in APIC mode, causes:
> Error -110, device does not accept ID
> USB Host is recognized fine, no device is attaced
>
> This is an error somewhere in the Kernel APIC Irq routing, which may
> worked around with "append noapic pirq="your irq" but using such a cutdown
> USB System is not a good idea, no relly working bulk-transfers (forget
> any devices which depend from it: scanners, camera, sound, isdn,
> harddisks, zip etc.)

I have a SMP motherboard that requires this (noapic setting). With that
set, everything works fine including bulk transfers and all other USB
devices that I have.

What error messages do you have showing that bulk transfers do not work?

thanks,

greg k-h

2001-12-24 19:27:40

by Graham TerMarsch

[permalink] [raw]
Subject: Re: [Linux-usb-users] VIA Chipsets + USB + SMP == UGLY TRASH

On Monday 24 December 2001 10:47, Greg KH wrote:
> On Mon, Dec 24, 2001 at 10:32:49AM +0100, Juergen Sauer wrote:
> > Hi!
> > Merry X-Mas everywhere !
> >
> > So, my USB tryout is over.
> > This is the expierience report:
> > You should not try to use VIA Chipsets + SMP + USB, that's
> > the worst thinkable idea. It's junk (usb-Part).
>
> Depends on the motherboard. What one do you have?

I've had similar problems here on an AOpen AK73Pro(A) motherboard. Its a
single-processor board based on KT133A, and since 2.4.3 haven't been able
to get USB working on it at all.

Dug through the old USB-users mailing lists and tried a wide variety of
things; "noapic", changing IRQs, PnP On/Off in the BIOS, but had no luck.
Best I was able to discern was that the board is using MPS1.4 but has no
option in the BIOS to switch it back to MPS1.1.

Would love to get it working so that I could finally get my Epson1200U
scanner working again, but am out of things to try (any suggestions would
be MORE than welcome).

> > That's why:
> > 1. not solved USB Irq errors in APIC mode, causes:
> > Error -110, device does not accept ID

Exactly the same error I'm getting here. Doesn't matter whether its the
USB scanner or printer that I plug in to the USB ports, I always end up
getting that error.

--
Graham TerMarsch
Howling Frog Internet Development, Inc. http://www.howlingfrog.com

2001-12-25 03:25:54

by Jack F. Vogel

[permalink] [raw]
Subject: Re: [Linux-usb-users] VIA Chipsets + USB + SMP == UGLY TRASH

On Monday 24 December 2001 10:47 am, Greg KH wrote:
> On Mon, Dec 24, 2001 at 10:32:49AM +0100, Juergen Sauer wrote:
> > Hi!
> > Merry X-Mas everywhere !
> >

And a Happy New Year to us all :)

> > So, my USB tryout is over.
> > This is the expierience report:
> > You should not try to use VIA Chipsets + SMP + USB, that's
> > the worst thinkable idea. It's junk (usb-Part).
>
> Depends on the motherboard. What one do you have?
>
..... [ snipped]
>
> I have a SMP motherboard that requires this (noapic setting). With that
> set, everything works fine including bulk transfers and all other USB
> devices that I have.

I also have an SMP system using VIA. Its an Asus dual coppermine, I
do not recall the exact board number.

I don't have a large variety of USB devices on it. But what I have used
works fine.

This motherboard does have an IOAPIC, so does not require the noapic
option.

One thing of note, there has been a bit of activity in the apic code recently,
you never did say what kernel version this was using, maybe a recent
vintage would help??

Merry New Year :)

--
Jack F. Vogel
IBM xSeries Linux Solutions
[email protected] (at work)
[email protected] (at home)