Summary:	ASTERISK-00822: Linux locks hard when deselecting loopback (on SMP system)
Reporter:	steveu (steveu)	Labels:
Date Opened:	2004-01-13 09:43:57.000-0600	Date Closed:	2004-09-25 02:54:42
Priority:	Critical	Regression?	No
Status:	Closed/Complete	Components:	Core/General
Versions:		Frequency of Occurrence
Related Issues:
Environment:		Attachments:
Description:	If I use zttool to enable loopback on a Tormenta 2 (E400P) card all seems well. When I try to remove the loop the whole computer locks solid. I am using the drivers from today's CVS ****** ADDITIONAL INFORMATION ****** This couldn't be duplicated in a non-SMP enviroment, so it appears to just affect SMP systems.
Comments:	By: jrollyson (jrollyson) 2004-01-14 03:08:35.000-0600 I can freeze zttool itself this way with a T400P by looping back an unconfigured span, but I can't freeze the whole system. zttool still dies on a SIGTERM or a SIGKILL though. When I go back in, I can't unloop at all until I configure the span. I can't duplicate at all on a properly configured span. Maybe something is slightly different in the E400P. Did you have the span configured in /etc/zaptel.conf and did you remember to ztcfg? By: steveu (steveu) 2004-01-17 20:49:13.000-0600 Of course I ran ztcfg :-) The span was configured and functioning properly. Maybe the issue is that I use an dual Xeon. Did you test with an SMP box? Someone on IRC said they get the same thing with a dual xeon box. By: izo (izo) 2004-01-17 21:39:58.000-0600 I have same config, E400P + Dual Xeon + SMP kernel on debian, locks after deselecting loopback By: Brian West (bkw918) 2004-01-17 22:12:22.000-0600 Happen to have zaptel compiled with SMP support? By: izo (izo) 2004-01-18 09:34:02.000-0600 works in same way with and without smp support By: Brian West (bkw918) 2004-01-18 19:50:34.000-0600 Do you have hyper threading turned on or off? By: steveu (steveu) 2004-01-19 07:54:54.000-0600 Super-hyper-duper-threading is go! :-) so I effectively am running on a 4 CPU machine. By: jrollyson (jrollyson) 2004-01-19 08:06:37.000-0600 Ok. Anybody able to determine if this is related to hyperthreading? By: jrollyson (jrollyson) 2004-01-19 08:46:27.000-0600 [09:29] <coppice> Tyan 2665, 2 x Xeon 2.4GHz, HT on. TE410P, E400P and TDM400P installed. Only the E400P drivers loaded and configured when it crashes By: Brian West (bkw918) 2004-01-29 23:34:38.000-0600 Turn hyper threading OFF! By: Paul Cadach (pcadach) 2004-02-08 00:31:47.000-0600 How hyper-threading relies on troubes with meetme and other zaptel-related functionalies? edited on: 02-08-04 00:33 By: zoa (zoa) 2004-02-25 07:45:19.000-0600 could you tell me if this got fixed with the recent zaptel/meetme fix ? By: Mark Spencer (markster) 2004-02-29 01:45:56.000-0600 Does the problem occur with TE410P as well or is it only when running an E400P? By: James Golovich (jamesgolovich) 2004-02-29 03:08:33.000-0600 I experience the same thing as jrollyson.  Using a T400P trying to loop up an unconfigured span causes zttool to never return from the "loop up span" stage.  but it doesn't hose up the whole system.  I didn't think there was too much different on a E400P vs T400P but who knows. I tried looping up an unconfigured span on TE410P (configured as T1) and it returns from the loop up and zttool continues to work.  So I'd say that TE410P performs better than T400P here. By: mwood (mwood) 2004-03-05 18:56:40.000-0600 This bug should be fixed in cvs. Open it back up if your problem is not fixed.