Summary:	ASTERISK-00292: TE410P: Double/missed interrupt detected
Reporter:	oliver (oliver)	Labels:
Date Opened:	2003-09-21 09:06:53	Date Closed:	2011-06-07 14:00:42
Priority:	Critical	Regression?	No
Status:	Closed/Complete	Components:	Core/General
Versions:		Frequency of Occurrence
Related Issues:
Environment:		Attachments:
Description:	I just got a TE410P installed into one of my systems and it has now happened twice in a week already that the machine stopped responding, scrolling these on the console: TE410P: Double/missed interrupt detected The machine previously had an E100P installed and ran for long periods without any driver-related problems. ****** ADDITIONAL INFORMATION ****** RedHat 9 Kernel 2.4.21 Single-CPU Xeon @2.4 GHz, Supermicro. Latest CVS of Zaptel, libpri and Asterisk.
Comments:	By: Mark Spencer (markster) 2003-09-21 10:24:36 Does running single CPU make a difference? By: oliver (oliver) 2003-09-21 11:01:46 Hey Mark, I am already running single CPU. However, I did find a pretty reliable way to reproduce it.  I have an extension set up on the machine with the TE410P, which puts the call through IAX2, to a TDM400P-connected analog line on a second Asterisk-box.  When I originate a voice-call to that extension, through an E1, connected to the TE410P, everything seems fine.  As soon as I originate a data-call through it, using analog 33K6-modems, the Asterisk-box with the TE410P freezes with the afforementioned problem. Could it be possible that the code which detects a data-call and disables the echo-canceller in that case is somehow to blame? BTW: If I originate the data-call from the TDM400P-side to the E1, everything seems fine.  I tried both about 3 times in a row.  All 3 calls from the E1-side froze the box with the TE410P, all calls originated from the TDM400P-side didn't affect it. The last thing I see on the Asterisk-console of the box with the TE410P before it hangs is this: DEBUG[475157]: File chan_zap.c, Line 1967 (zt_setoption): Set option TONE VERIFY, mode: MUTECONF/MAX(2) on Zap/31-1 DEBUG[114696]: File chan_iax2.c, Line 3941 (socket_read): Ooh, voice format changed to 4 edited on: 09-21-03 10:45 By: oliver (oliver) 2003-09-21 12:30:30 Another note: To prove my theory, I just disabled echo cancellation in "zapata.conf" (echocancel=no, echocancelwhenbridged=no) on the machine with the TE410P, and originated a few more data-calls through the E1.  It doesn't appear to freeze anymore with the EC disabled, so it looks like the freezes are indeed caused by the EC beeing automatically disabled in-call, when a data-call is detected.  I tried some 6 calls with the EC disabled, without freezing the machine.  The first modem-call after I re-enabled EC froze it again.  This time, I saw a kernel-message I hadn't seen before: Tried to load 00000000 into 00000004, but got 00000002 instead It just appeared once, in the midst of the continuesly scrolling "TE410P: Double/missed interrupt detected". PS: I'm using a completely default build of Zaptel, with no changes to the Makefile or source. edited on: 09-21-03 12:09 By: Mark Spencer (markster) 2003-09-21 15:07:21 This is caused by your use of serial console combined with the printk.  Simply removing the printk will cause this error to go away.  Note that serial console printk's block interrupts for extended periods of time.