Memory access violations occur
when a kernel is reading or writing out of bounds of an object or reading
uninitialized memory. This can manifest itself in multiple ways like a simulator
crash or hang. It can also cause simulator results to be non-repeatable. The x86simulator --valgrind option will find memory access
violations in kernel source code.
This option allows detection of memory access violations in kernel source code during x86 simulation with Valgrind. The following kinds of access violations can be detected.
- Out-of-bounds write
- Out-of-bounds read
- Read of uninitialized memory
There are two ways to use this option:
-
x86simulator --valgrind - This runs the simulation with access violation detection turned on. At the end of the simulation, Valgrind prints a report on access violations. If there are none, the report ends with ERROR SUMMARY: 0 errors from 0 contexts. Otherwise, the report lists each access violation found. This includes a stack trace, which highlights the line number in the kernel source code where the access violations occurred.
-
x86simulator --valgrind-gdb - This runs the simulation with access violation detection turned on and debug with GDB. The simulation comes up in GDB and is halted at main(). At this point you can set additional breakpoints. After continuing, the simulation will stop if an access violation is detected. At this point you can inspect local variables and the stack to diagnose the problem.
In either case, some arguments can be added to the Valgrind
command using flag --valgrind-args='list of arguments for
valgrind'. For example:
--valgrind-args='-v --leak-check=no --track-origins=yes' will not
track memory leakage and would display the overall stack when discovering an access
violation.
