Re: 8250 serial console fixes -- issue

[Kumar Gala <galak@kernel.crashing.org>]

On Feb 2, 2006, at 2:05 AM, Kumar Gala wrote:

>
> On Feb 1, 2006, at 11:54 PM, Kumar Gala wrote:
>
>>
>> On Feb 1, 2006, at 7:47 PM, Alan Cox wrote:
>>
>>> On Mer, 2006-02-01 at 19:21 -0600, Kumar Gala wrote:
>>>> This patch introduces an issue for me an embedded PowerPC SoC  
>>>> using the
>>>> 8250 driver.
>>>>
>>>> The simple description of my issue is this:  I'm using the  
>>>> serial port for
>>>> both a terminal and console.  I run fdisk on a /dev/hda.  Before  
>>>> this
>>>> patch I would get the prompt for fdisk immediately.  After this  
>>>> patch I
>>>> have to hit return before the prompt is displayed.
>>>>
>>>> I know that's not a lot of info, but just let me know what else  
>>>> you need
>>>> to help debug this.
>>>>
>>>> I'm guessing something about the UARTs on the PowerPC maybe bit  
>>>> a little
>>>> non-standard.
>>>
>>> I wonder if I've swapped one race for another. Can you revert  
>>> just the
>>> line which forces THRI on and test with the rest of the change  
>>> please.
>>
>> This doesn't seem to help.
>
> I realized a bit later that the kernel I build with your suggested  
> change wasn't the one I was testing.  After loading the proper  
> kernel this does make the issue I was seeing go away.

After some further testing with this change, the majority of issues I  
was seeing go away.  However, able to lock the console up. If I  
revert the wait_for_xmitr() to always use BOTH_EMPTY the lock up goes  
away.  (Obviously, this is effectively reverting the whole patch).

- kumar