For those following this thread, the userland .core files do not include kernel address space, and it is apparent from the stack traces that while each .core file may be different, the root problem is the same, and occurs within the kernel address space. Some research has led i3luefire and I to kernel option SYSCALL_DEBUG. This is not enabled by default, and the code it adds to the kernel was out of date and required some patching.
We now have a program designed to capture one second of syscalls and arguments in a dmesg buffer, then capture a kernel core dump, and are hoping we will obtain useful information once it successfully runs on the system under test.
Last edited by jggimi; 1st January 2016 at 06:22 AM.
Reason: clarity, typo
|