This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Segfault running C6670 cycle approx. simulator under DSS (CCSv5.1M7 linux)

Hi All,

   I have a very strange problem running the C6670 cycle approximate simulation model under the CCsV5 linux DebugServer batch environment. I get the following segfault:

Now the strange thing about this is that when I run this under my userid, it works fine. When it is run under anyone else's userid, the result is the segfault above. It looks like there is some initialization of the DDR memory interface and the code is trying to write into a file. The environment is the same in all cases (same PATH, same LD_LIBRARY_PATH and same CLASSPATH, same dss.sh script). The version we are currently using is: CCSV5.1M7 linux. I saw that there is a later linux version available: 5.1M8.5 that I will download. However, looking at the release notes did not reveal anything on this sort of problem. Have you seen anything like this before? Is there something special in the environment I may have missed?

Thanks,

Mark Brown

  • Hi Mark,

    I'm wondering if it is a permissions thing. Do you have admin rights while other accounts do not?

    ki

  • Ki,

       I thought about the same thing. I checked and all userids are members of the same groups. As far as I know I do not have any special authority. In any case the application should not segfault. Can your developers identify from the stack trace what file is being accessed. If we know this, it may give us a clue as to where in the file system to look.

    Thanks,

    Mark Brown

  • Ki,

       Any progress on this issue and my last request? This is holding up many developers.

    Thanks,

    Mark Brown

  • Sorry, not yet. I am requesting support from the simulation team. I don't know if you want to try RC1 (M9). I can't guarantee that there is a specific fix for the above issue but it may be worth a shot

    Thansk

    ki

  • The above issue is occuring because of not having write access to the file <CCS_installation>/\ccsv4\simulation_csp_ny\env\ccs\drivers\flat_mem_handler_asserts.log

    Can you please check if you are able to bring up the simulation after giving write permission for this file. Meanwhile we are working on fixing this issue.

    Thanks and regards

    Abhilash

     

     

  • Abhilash,

       Yes we also arrived at the same conclusion. What is happening is that if the log file doesn't exist, the first user's session creates it using that user's permission mask. Currently all our users have a mask that provides only the owner with write access. Therefore when another user attempts to update the same file, the permission error occurs. However a segfault should not occur here but should at least notify the user with a message. What we have done to temporarily fix this is to pre-create the file and grant group write permission so all usres can potentially write to this file. We have also started using the functional-level simulator which does not access this file.

       This brings up a philosophical issue here. Why are all users accessing the same file? Additionally why does the file reside in the release area? We are using batch simulation with potentially many concurrent users. This is typically how batch simulation is used. The log file protocol you have setup does not support this mode of use. Is there a way to setup the environment to support this multi-user mode?

    Thanks,

    Mark

  • Mark,
           Currently there are few files which get generated by default and there is no option to turn this "OFF". in the configuration file for the 6670 cycle approximate simulator. For functional simulator these files are not generated. We would add support to disable generation of these file and optional provide support through configuration file to generate if needed.

    We would plan to add this support in the future simulator release. As of now , I hope you are not blocked by this issue.

    Thanks and regards
    Abhilash