[Home]

Summary:ASTERISK-09764: SIP compile warnings, cashes crash on invite received
Reporter:Tjardick van der Kraan (tjardick)Labels:
Date Opened:2007-06-27 10:54:54Date Closed:2008-01-28 13:54:04.000-0600
Priority:CriticalRegression?No
Status:Closed/CompleteComponents:Channels/chan_sip/General
Versions:Frequency of
Occurrence
Related
Issues:
Environment:Attachments:( 0) bt-0010069.txt
Description:compiling of sip channel on CentOS 4 on i64 (dell 3250 with itanium 1.4) gives compile warnings on SIP channel and does a segmentation fault on invite. see additional info for compile warnings.

****** ADDITIONAL INFORMATION ******

Warnings on compile:

  [CC] chan_sip.c -> chan_sip.o
{standard input}: Assembler messages:
{standard input}:94982: Warning: Use of 'addl' may violate WAW dependency 'GR%, % in 1 - 127' (impliedf), specific resource number is 54
{standard input}:94981: Warning: This is the location of the conflicting usage
{standard input}:94993: Warning: Use of 'ld8.mov' may violate RAW dependency 'GR%, % in 1 - 127' (impliedf), specific resource number is 54
{standard input}:94992: Warning: This is the location of the conflicting usage
{standard input}:94993: Warning: Use of 'ld8.mov' may violate WAW dependency 'GR%, % in 1 - 127' (impliedf), specific resource number is 54
{standard input}:94992: Warning: This is the location of the conflicting usage
Comments:By: Tjardick van der Kraan (tjardick) 2007-06-27 13:11:06

Forgot to add uname -a

Linux localhost.localdomain 2.6.9-55.EL #1 SMP Wed May 2 20:52:25 EEST 2007 ia64 ia64 ia64 GNU/Linux
By: Russell Bryant (russell) 2007-06-27 13:37:37

Do you have a backtrace from the crash?
By: Tjardick van der Kraan (tjardick) 2007-07-03 07:12:57

We have moved to 1.4.6 since. I haven't been able to repeat it on 1.4.6 but the machine is not having much use for the moment. I propose to close down this bug and i will re-open a new one in case i run into the same problems and then attach a backtrace directly.
By: Tjardick van der Kraan (tjardick) 2007-07-03 07:13:58

I did notice that the same compile warnings are there in app_sms as well when i compiled it. Not sure if the warnings above can hurt...
By: Joshua C. Colp (jcolp) 2007-07-05 08:25:16

Per your request I'm suspending this bug for now, if it happens again feel free to reopen... but remember that none of the developers have access to that architecture...
By: Tjardick van der Kraan (tjardick) 2007-07-09 08:15:49

While following the backtrace documentation on how to create a proper backtrace, i recompiled asterisk without the optimization. Without this optimization, asterisk does not give the mentioned compile error and it doesn't crash either.

I assume it is an incompatibility with the code optimization on the ia64 platform?

Further more my company is happy to provide and itanium server for the developers for a certain periode to be able to perform code testing/optimization etc, but would like to coordinate that with someone from digium to avoid giving open access to the machine etc. If there is interest, please have someone from digium contact me directly.
By: Tjardick van der Kraan (tjardick) 2007-07-09 08:16:19

last test where done on 1.4.6
By: Russell Bryant (russell) 2007-07-09 15:53:27

Can you try to get a backtrace of the crash with optimizations enabled?  There is a chance it could still be helpful, but most likely not.
By: Tjardick van der Kraan (tjardick) 2007-07-10 04:10:02

All output from the core: gdb, bt, bt full, thread apply all bt.
By: Joshua C. Colp (jcolp) 2007-08-06 12:24:47

Can you please contact me at jcolp@digium.com to arrange access? Thank you.
By: Michiel van Baak (mvanbaak) 2007-09-08 07:18:46

tjardick, file: Any progress on this one?
By: Tjardick van der Kraan (tjardick) 2007-09-10 03:21:24

A seperate machine for debugging by digium is now up and running mail send to address above requesting ssh key and IP end last week. So i hope we will have some more details on this soon.


By: Tjardick van der Kraan (tjardick) 2007-09-19 01:51:22

No reply from jcolp@digium.com re-sending a mail to check where this is at.
By: Tjardick van der Kraan (tjardick) 2007-10-24 03:39:44

Is there anything you guys need? i see now login on the machine so far? feel free to give me a shout if anything more is needed by contacting me on my mail address.
By: Olle Johansson (oej) 2007-11-15 06:00:58.000-0600

FILE::: WAKE UP! /O :-)
By: Michiel van Baak (mvanbaak) 2008-01-10 07:13:46.000-0600

Can you confirm this one on the latest 1.4 svn ?
By: Jason Parker (jparker) 2008-01-28 13:54:03.000-0600

After looking over various postings on other mailing lists, I've discovered that there are still some issues in gcc/as on ia64 that ends up with bad instructions.  I would very highly recommend upgrading binutils as well as gcc (bonus points if you can get a binutils that existed at the time the gcc version was released).  Note that this is exactly in line with your report of this issue going away when disabling optimizations.  There really isn't anything we can do on our end, in this case.

For what it's worth, WAW and RAW stand for "write after write" and "read after write".


Closing.