Compiling cardbus devices monolithic doesn't work?

Compiling cardbus devices monolithic doesn't work?

[Rob Landley]

I'm easing my way into the 2.6.0-test series, and everything I've done so far 
has been with monolithic kernels to minimize the number of fun new things 
I've been playing with, and I just can't get the monolithic orinoco_cs to 
find my new thinkpad's built-in wireless networking thingy.

I think it's because even though the sucker's built-in, the bus topology puts 
it behind its own cardbus bridge controller thingy (for no readily apparent 
reason).  Cardbus needs hotplug, hotplug wants to load modules.

Currently I'm booting into 2.4 when I need to use the internet.  I suppose 
it's time to actually fire up rusty's modutils, but before I do that I'm 
curious WHY the compiled-in orinoco_cs driver can't find the card.  
(Presumably because hotplug can't provide it with module arguments, or 
something like that?)  If they DO only work as modules, why are they allowed 
to be compiled monolithic?  (Did I miss a warning in the kconfig help text, 
or should I have read the relevant section of Documentation more closely?)

It's entirely possible there's some other reason my wireless card isn't 
working under 2.6, this is just a working hypothesis.  I'm happy to debug it 
if anybody has suggestions...

Rob

(P.S.  And while I'm at it, what's the relationship between orinoco_cs, 
orinoco, and hermes?  The /proc/modules dependency tree thing says they're 
using each other in a chain.  Probably true, just a bit odd, I thought.  
Couldn't figure out which driver I needed, compiled all three, and it loaded 
ALL of them.  Can't complain, the card works under 2.4.  This is just a 
random "huh?")

Re: Compiling cardbus devices monolithic doesn't work?

[Russell King]

echo $subject | sed s/cardbus/pcmcia/i

On Sun, Aug 17, 2003 at 09:58:34PM -0400, Rob Landley wrote:
> I'm easing my way into the 2.6.0-test series, and everything I've done so far 
> has been with monolithic kernels to minimize the number of fun new things 
> I've been playing with, and I just can't get the monolithic orinoco_cs to 
> find my new thinkpad's built-in wireless networking thingy.

You still need to use cardmgr to bind the driver to the device.
It seems to work for me here on SA11x0 platforms, and I'm not aware
of it breaking at any point in the 2.5 series.

While it is true that Cardbus devices plugged into cardbus slots do
not need cardmgr, PCMCIA devices still do.

> (P.S.  And while I'm at it, what's the relationship between orinoco_cs, 
> orinoco, and hermes?  The /proc/modules dependency tree thing says they're 
> using each other in a chain.  Probably true, just a bit odd, I thought.  
> Couldn't figure out which driver I needed, compiled all three, and it loaded 
> ALL of them.  Can't complain, the card works under 2.4.  This is just a 
> random "huh?")

IIRC hermes provides the low level interface to the device, orinoco
provides the interface between it and the network stack, and orinoco_cs
provides a bridge between the PCMCIA subsystem and orinoco.

-- 
Russell King (rmk@arm.linux.org.uk)                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html

Re: Compiling cardbus devices monolithic doesn't work?

[Rob Landley]

On Monday 18 August 2003 03:44, Russell King wrote:
> echo $subject | sed s/cardbus/pcmcia/i

I thought cardbus was just 32 bit pcmcia...

(I suppose I should go look it up.)

Rob

Re: Compiling cardbus devices monolithic doesn't work?

[Rob Landley]

Ahem, second attempt:

On Monday 18 August 2003 03:44, Russell King wrote:

> You still need to use cardmgr to bind the driver to the device.
> It seems to work for me here on SA11x0 platforms, and I'm not aware
> of it breaking at any point in the 2.5 series.
>
> While it is true that Cardbus devices plugged into cardbus slots do
> not need cardmgr, PCMCIA devices still do.

The hotplug scripts from RH9 are there and seem happy, I thought cardmgr was 
called from them.  (The same setup works in 2.4.21, albeit with modules 
enabled.  I should compile a monolithic 2.4 kernel and see what it does...)

> > (P.S.  And while I'm at it, what's the relationship between orinoco_cs,
> > orinoco, and hermes?  The /proc/modules dependency tree thing says
> > they're using each other in a chain.  Probably true, just a bit odd, I
> > thought. Couldn't figure out which driver I needed, compiled all three,
> > and it loaded ALL of them.  Can't complain, the card works under 2.4. 
> > This is just a random "huh?")
>
> IIRC hermes provides the low level interface to the device, orinoco
> provides the interface between it and the network stack, and orinoco_cs
> provides a bridge between the PCMCIA subsystem and orinoco.

Now I'm confused.  I thought the _cs on the end was short for "cardbus"...

Rob

Re: Compiling cardbus devices monolithic doesn't work?

[Randy.Dunlap]

On Mon, 18 Aug 2003 15:34:37 -0400 Rob Landley <rob@landley.net> wrote:

| Ahem, second attempt:
| 
| On Monday 18 August 2003 03:44, Russell King wrote:
| 
| > You still need to use cardmgr to bind the driver to the device.
| > It seems to work for me here on SA11x0 platforms, and I'm not aware
| > of it breaking at any point in the 2.5 series.
| >
| > While it is true that Cardbus devices plugged into cardbus slots do
| > not need cardmgr, PCMCIA devices still do.
| 
| The hotplug scripts from RH9 are there and seem happy, I thought cardmgr was 
| called from them.  (The same setup works in 2.4.21, albeit with modules 
| enabled.  I should compile a monolithic 2.4 kernel and see what it does...)
| 
| > > (P.S.  And while I'm at it, what's the relationship between orinoco_cs,
| > > orinoco, and hermes?  The /proc/modules dependency tree thing says
| > > they're using each other in a chain.  Probably true, just a bit odd, I
| > > thought. Couldn't figure out which driver I needed, compiled all three,
| > > and it loaded ALL of them.  Can't complain, the card works under 2.4. 
| > > This is just a random "huh?")
| >
| > IIRC hermes provides the low level interface to the device, orinoco
| > provides the interface between it and the network stack, and orinoco_cs
| > provides a bridge between the PCMCIA subsystem and orinoco.
| 
| Now I'm confused.  I thought the _cs on the end was short for "cardbus"...

I would have guess that it was for Card Services... (a PCMCIA software
component, at least architecturally).

--
~Randy
"Everything is relative."