[Asterisk-Users] Re: TDM400 - channel out to lunch?

[Michael Welter]

>>Not to mention the CPU spikes every <n> seconds.  Rich, while you're 
>>testing, would you keep an eye on 'vmstst 1' and the 'system' (not user) 
>>CPU utilization?
> 
> 
> That cpu spiking is another issue separate from the stability issue
> (I think). Not sure where the discussion of the spiking ended up a 
> few weeks ago, do you remember?

I spoke with Matt O. at Digium tech support.  He was in my machine 
running 'vmstat 1'.  I demonstrated how the "CPU spiking" stopped when I 
unloaded the wcfxs module.  I pointed-out that, during a spike, the 
total number of interrupts did not increase, suggesting that one 
interrupt was being held for 20+ms.  Matt then did the same tests on his 
own machine and observed the same results.  Matt said Mark was looking 
at the problem.

I believe this spiking is at the root of my spandsp problems, causing a 
periodic frame slip and the failure of the fax transmission.

That is the latest information I have.

> 
> I've not tried to dig through the code, but it wouldn't surprise me
> if some temp code exists that might be polling the tdm card (or
> something like that) as an aid towards identifying the stability
> issue. Gut feeling suggests that if stability is truly related to
> tdm design problems (or whatever), then resolving that issue probably
> should be a precursor to chasing the cpu spikes.
> 

Agree.  But if you can get the card to go crazy as a result of physical 
pressure then it would be interesting to see if there is a correlation 
to CPU usage.  Something in wcfxs is causing it to not exit its 
interrupt routine in a normal manner.