[Asterisk-Users] Intel 537EP chipset, revisited

[Lee Howard]

On 2004.06.11 11:41 Steven Critchfield wrote:

> You next step really would to be to go looking through the module load
> function and maybe through some debugging code in there to have it
> tell
> you more about what is going on. Luckily enough, if you didn't crash
> the
> machine with your module, you can probably load and unload safely
> until
> you break something in the module. This is one of those nice things
> about linux drivers from a developer stand point.

It turns out that the Linux serial driver had claimed the modem device, 
and thus the wcfxo module was unable to use it.  Once I disabled the 
Linux serial driver (for testing, not a permanent solution, of course) 
then the wcfxo module was able to "pick it up".  The module load still 
failed, however:

Zapata Telephony Interface Registered on major 196
Setting hook state to 0 (08)
Registered Span 1 ('WCFXO/0') with 1 channels
Span ('WCFXO/0') is new master
wcfxo: Out of space to write register 06 with e0
wcfxo: Out of space to write register 0f with 00
Failed to initailize DAA, giving up...
Unregistering Span 'WCFXO/0' with 1 channels
Zapata Telephony Interface Unloaded

Digging through more archives I see that someone else has gotten this 
far, also:

http://lists.digium.com/pipermail/asterisk-users/2004-February/035818.html

Following the comments on:

http://www.digium.com/index.php?menu=faq#Errors_1

... this person assumed that the failure was due to the IRQ being 
shared between devices.  However, in my case the device had its own 
IRQ.  Reading through more archives, I see that the general reason for 
those "Out of space to write register" errors is due to interrupt 
problems, and the general advise has been to change PCI slots.  So I 
tried this.  Depending on the presense of other PCI devices in the 
system as well as their configuration in the PCI slots, I would either 
get the outcome described above, or worse, I would get:

Zapata Telephony Interface Registered on major 196
Setting hook state to 0 (08)
Registered Span 1 ('WCFXO/0') with 1 channels
Span ('WCFXO/0') is new master
Unexpected IRQ trap at vector c8
Unexpected IRQ trap at vector ad

... and then the system locks so that it requires a power-cycle/reset.  
IRQs were never shared in any configuration.

So, at this point I have satisfied my curiosity, and I'm going to 
conclude that the wcfxo driver incompatibility with the Intel 537EP 
chipset is not merely a problem with PCI device IDs, but also that 
there is some deeper incompatibility.  If anyone is interested in 
taking this farther, or if my conclusions are incorrect, just let me 
know, I'll be happy to entertain the experience.

Thanks.

Lee.