USB host drivers test results (2.5.19) and problem.

USB host drivers test results (2.5.19) and problem.

[Stelian Pop]

Hi,

Now that a 2.5 kernel boots again on my laptop (thanks to the ACPI
IRQ routing problems being resolved :-) ), I tested the different
USB host drivers available (I used to use usb-uhci before).

lspci:
	00:07.2 USB Controller: Intel Corp. 82371AB/EB/MB PIIX4 USB (rev 01)

Results (only simple tests with a USB mouse performed):
	ehci-hcd.o
		does load, no output messages, does NOT work
	ohci-hcd.o 
		does load, no output messages, does NOT work
	uhci-hcd.o 
		does load, does work correctly
	usb-uhci-hcd.o
		does load, does work correctly.


1. Shouldn't the ehci/ohci drivers give some error on loading,
since I obviously don't have the hardware ?

2 When doing a rmmod on one of the two last drivers, 
the kernel oopses with a (hand copied trace):

	bad EIP: c784fbda <[usbcore]rh_report_status+63/143>
	
	c0118f34 <__run_task_queue+61/6a>
	c784f677 <[usbcore]usb_reset_device+1d4/2df>
	c011be58 <timer_bh+22b/27c>
	c0118e48 <bh_action+26/74>
	c0118d79 <tasklet_hi_action+38/59>
	c0118bb7 <do_softirq+57/b1>
	c01099a0 <do_IRQ+da/e5>
	c0108656 <common_interrupt+22/28>
	c01abd54 <acpi_processor_idle+e9/1eb>
	c0106b54 <gunzip+45c/4bc>
	c0106bd0 <default_idle+0/27>
	c01abc6b <acpi_processor_idle+0/1eb>
	c0106bd0 <default_idle+0/27>
	c0106c5a <cpu_idle+2e/3d>
	c0105000 <_stext+0/0>

Another one:
	bad EIP: <c78ed365> <[yenta_socket]yenta_allocate_resources+d/41>

	Adhoc c011be58 <timer_bh+22b/27c>
	Adhoc c0118e48 <bh_action+26/74>
	Adhoc c0118d79 <tasklet_hi_action+38/59>
	Adhoc c0118bb7 <do_softirq+57/b1>
	Adhoc c01099a0 <do_IRQ+da/e5>
	Adhoc c0108656 <common_interrupt+22/28>
	Adhoc c01abd54 <acpi_processor_idle+e9/1eb>

I'm not sure what this trace has to do with USB at all, but I haven't
been able to reproduce the oops otherwise...

Stelian.
-- 
Stelian Pop <stelian.pop@fr.alcove.com>
Alcove - http://www.alcove.com

Re: USB host drivers test results (2.5.19) and problem.

[David Woodhouse]

stelian.pop@fr.alcove.com said:
>  1. Shouldn't the ehci/ohci drivers give some error on loading, since
> I obviously don't have the hardware ? 

How do they know that? You could have it in your hand and be just about to
insert it.

--
dwmw2

Re: USB host drivers test results (2.5.19) and problem.

[Stelian Pop]

On Fri, May 31, 2002 at 03:54:02PM +0100, David Woodhouse wrote:

> >  1. Shouldn't the ehci/ohci drivers give some error on loading, since
> > I obviously don't have the hardware ? 
> 
> How do they know that? You could have it in your hand and be just about to
> insert it.

PCI hotplug ? I fergot about that one...

However, maybe even a hotplug aware driver should print out a message
upon initialization telling 'no hardware found'... It could be useful
when you want to know the difference between: 
	* bad driver, use another one
	* good driver, but doesn't recognize/init the hardware because it's buggy.

Stelian.
-- 
Stelian Pop <stelian.pop@fr.alcove.com>
Alcove - http://www.alcove.com

Re: USB host drivers test results (2.5.19) and problem.

[Greg KH]

On Fri, May 31, 2002 at 03:34:29PM +0200, Stelian Pop wrote:
> 1. Shouldn't the ehci/ohci drivers give some error on loading,
> since I obviously don't have the hardware ?

Yes, they should not load.  I'll try to look into this.

> 2 When doing a rmmod on one of the two last drivers, 
> the kernel oopses with a (hand copied trace):

If you remove _any_ pci driver module from 2.5.19 you get an oops, this
isn't limited to the USB drivers right now :)

thanks,

greg k-h

Re: USB host drivers test results (2.5.19) and problem.

[Greg KH]

On Fri, May 31, 2002 at 03:54:02PM +0100, David Woodhouse wrote:
> 
> stelian.pop@fr.alcove.com said:
> >  1. Shouldn't the ehci/ohci drivers give some error on loading, since
> > I obviously don't have the hardware ? 
> 
> How do they know that? You could have it in your hand and be just about to
> insert it.

They should not load, like any other pci driver that should not load if
you don't have the hardware present for it.

thanks,

greg k-h

Re: USB host drivers test results (2.5.19) and problem.

[Stelian Pop]

On Fri, May 31, 2002 at 09:38:36AM -0700, Greg KH wrote:

> > 2 When doing a rmmod on one of the two last drivers, 
> > the kernel oopses with a (hand copied trace):
> 
> If you remove _any_ pci driver module from 2.5.19 you get an oops, this
> isn't limited to the USB drivers right now :)

Yep, I saw that afterwards when shutting down my laptop.

Sorry for having incriminated the USB subsystem for that :-)

Stelian.
-- 
Stelian Pop <stelian.pop@fr.alcove.com>
Alcove - http://www.alcove.com

Re: USB host drivers test results (2.5.19) and problem.

[Stelian Pop]

On Fri, May 31, 2002 at 09:39:14AM -0700, Greg KH wrote:

> > >  1. Shouldn't the ehci/ohci drivers give some error on loading, since
> > > I obviously don't have the hardware ? 
> > 
> > How do they know that? You could have it in your hand and be just about to
> > insert it.
> 
> They should not load, like any other pci driver that should not load if
> you don't have the hardware present for it.

What about the PCI hotplug case, as David suggested ?

Stelian.
-- 
Stelian Pop <stelian.pop@fr.alcove.com>
Alcove - http://www.alcove.com

Re: USB host drivers test results (2.5.19) and problem.

[Greg KH]

On Fri, May 31, 2002 at 08:44:42PM +0200, Stelian Pop wrote:
> On Fri, May 31, 2002 at 09:39:14AM -0700, Greg KH wrote:
> 
> > > >  1. Shouldn't the ehci/ohci drivers give some error on loading, since
> > > > I obviously don't have the hardware ? 
> > > 
> > > How do they know that? You could have it in your hand and be just about to
> > > insert it.
> > 
> > They should not load, like any other pci driver that should not load if
> > you don't have the hardware present for it.
> 
> What about the PCI hotplug case, as David suggested ?

When the pci hotplug core sees a new device it calls out to
/sbin/hotplug to load any available driver (if necessary).  So if you
later plug in the device, everything will work just fine.

Well, that's the way it's _supposed_ to work, but the pci core changes
in 2.5.19 seem to have broken this too :(

/me looks for Pat Mochel...

thanks,

greg k-h

Re: USB host drivers test results (2.5.19) and problem.

[Brad Hards]

On Sat, 1 Jun 2002 01:07, Stelian Pop wrote:
> However, maybe even a hotplug aware driver should print out a message
> upon initialization telling 'no hardware found'... It could be useful
> when you want to know the difference between:
> 	* bad driver, use another one
> 	* good driver, but doesn't recognize/init the hardware because it's buggy.
Even if you have the driver compiled in, and there is no hotplug, you still 
don't want every driver complaining about not having suitable hardware. Think 
distribution kernels (or "kernel that handles my heterogenous lab") and 
network device drivers...

Brad


-- 
http://conf.linux.org.au. 22-25Jan2003. Perth, Australia. Birds in Black.