[asterisk-dev] Digium card TD2400/TE40XP irq problem

[Matthew Fredrickson]

On Apr 20, 2007, at 3:35 PM, Vasil Kolev wrote:

> В пт, 2007-04-20 в 11:31 -0500, Matthew Fredrickson написа:
>> You might check the trunk version of wct4xxp driver.  I have started a
>> fairly massive and significant overhaul on it that your feedback would
>> be useful on.
>>
>
>
> That's great, in fact after some issues I checked SVN and saw the 
> latest
> update to wct4xxp - can you just elaborate a bit on the nature of the
> changes? I can read (or at least try :) ) through most of them, but an
> explanation will really help chose between either waiting for the next
> release or using the trunk for the production stuff I have.

Most of the changes have to deal with improving our performance.  Most 
of it originally was to improve our performance using DTMF detection on 
the octasic.  It involved breaking out all non essential (and 
deferrable) parts of what was being done in the interrupt handler and 
moving that into a tasklet for bottom half processing.  It also 
includes some updates to reduce the amount of time held by spinlocks, 
especially within interrupt context.  We are seeing some spectacular 
results, especially with DTMF detection enabled on the octasic (the 
former worst case scenario).

>
> Also, this might not be the right thread, and the question was 
> discussed
> some months ago here (but not with a resolution), what about merging
> those in the mainline kernel? The process is ugly, annoying, time and
> nerve consuming, but the kernel people seem to find all the possible
> problems, especially regarding locking, so it seems like this will be
> beneficial to the driver.

There were a lot of opinions in both directions on that thread.  You'll 
have to see it to see what everybody thought.

Matthew Fredrickson