On Sat, 2012-01-21 at 20:49 -0800, Eric W. Biederman wrote:
> Interesting.
>
> The fact that this happens in native_set_pte would suggest that we are
> trying to write to a page table that does not exist. So this might
> be a layer below kexec_load that has the problem.
>
> Do you have the kernel you were testing? A disassembly of the
> native_set_pte, machine_kexec_prepare and sys_kexec_load
> would be interesting, for attempting to trace this back to what went
> wrong.
I did some work into investigating it today, looks like it's simple to trigger it using the following code:
#include <fcntl.h>
int main(void)
{
char dummy[4096] = {0};
syscall(246, 0xffffffff81008000, 1, dummy, 2);
return 0;
}
I'll continue trying to figure out whats wrong, but hopefully this lead will help in reproducing it easily.
--
Sasha.
