Method to capture vital system statistics during the core dump of a application program.
Original Publication Date: 2002-Jun-24
Included in the Prior Art Database: 2003-Jun-20
Method to capture vital system statistics during the core dump of a application program. Disclosed is a method for capturing vital system statistics in the core dump of an application program running on a UNIX* operating system. Often vital information is needed to debug an application core dump, but it is not collected as it is outside of the process space that is part of a code dump. A file system may be full and the application core dumps in response. Another process frees this processes shared memory and this processes core dump. The bug is really in the process that forked this process and the developer needs to see the relationship between the two. The proposed method would provide system calls, commands, or extended attribute on an executable that would direct the kernel to collect extra information in the core dump. The system calls would direct the kernel to collect extra information that this application knows it will need in the event that the application core dumps. The new commands and extended attributes would allow for an administrator to instruct the kernel to collect extra information in the core dump for applications not using the system calls. This extra information could be the parent processes stack trace, all the processes in the system process table, the amount of used and free memory and paging space, the amount of used and free space in the file systems where the process has open file descriptors. After the kernel is finished writing its part of the code dump it could also trigger commands to run which gather additional information that can be appended to the core dump or create an additional file that is associated with the core dump.