Re: serial ports - pxa vs. 8250

I am writing in regards to Mickeyl's pxa-serial/8250-serial
IRQ/namespace conflict patch (pxa-serial-hack.patch) and a had a
couple questions. I am attempting to fix this conflict on an embedded
system I have (not the hx4700, but suffers from the same issue). My
hardware configuration includes 3 serial ports driven by the PXA
driver as well as as a PCMCIA modem with an 8250 UART. Due to the
namespace conflict I get the following error when
trying to insmod Russell King's 8250.ko module:

===code===============================
[root_at_ARMCore pcmcia]$ insmod 8250.ko
insmod: error inserting '8250.ko': -1 Device or resource busy
======================================

I took a look at the source for the patch and it looks like it would
resolve the issue if I use it. However, I asked Russell King and he
suggested that the patch would cause more problems than it fixes? Do
his claims sound true?

The following is a copy of my message to him (>) and Russell's reply (>>):

===FWD============================================
>> Something like the following [pxa-serial-hack.patch] to be added to
8250.c >> would prevent a musical chairs issue with /dev/ttyS0 -
/dev/ttyS2?:
>
>Wrong. It's complete crap. It may look like it solves the problem
>but it just papers over the problem.
>
>For example, using console=ttyS0 vs console=ttyS3 does not do what's
>expected - there's no guarantee that one or other driver will end up
>being the console.
>
>> From what you say, it sounds like this modification shouldn't have
>> been necessary if the PXA team had followed the standards, however it
>> looks like my issue could be resolved if I made these mods [used the
>> patch]? You're thoughts, suggestions?
>
>Due to the above, re-using the namespace is just not on for mainline
>kernels. Sorry.
===================================================

Any thought's or suggestions are greatly appreciated.

Thanks.

-Nathan
Received on Sat Aug 06 2005 - 23:55:09 EDT