status of serialio bug

[John Heim]

I've been looking into the speakup bug that causes  you to get an error 
message when loading some speakup modules for serial hardware synths. What 
I've found is that the speakup code calls a function named request_resource. 
That function is apparently part of the standard linux kernel code. The 
source code module says its copyright Linux Torbalds. In case you're not 
aware of it, Linus Torvalds is the guy who created linux.

Anyway, the request_resource function does nothing other than to add an item 
to a linked list unless the item would overlap an item already in the list. 
I take this to mean that the linked list is nothing more than a list of 
reserved resources and you're supposed to check the list before you use a 
resource. But it doesn't actually prevent you from using the resource.

The problem is that the serial port space is definately already reserved 
when speakup asks for it even if you have nothing using the serial port. I 
added code to the speakup module to dump the linked list and it shows that 
the entire range from 0x000 through 0xcf7 is taken by something named "PCI 
Bus 0000:00". The serial port's address is 0x3f8 so that definately falls in 
that range. Therefore, it will always fail when speakup calls 
request_resource to reserve the serial port.

The only other possibility is that speakup is checking the wrong linked 
list. It's checking a list named ioport_resource  which certainly sounds 
like the right list.  And why would something called the PCI Bus have the 
serial port address space reserved? That makes no sense to me either.

So my plan is to subscribe to the kernel coders email list and ask for 
information on that function call.