The mips bfd will sign extend 32-bit addresses into 64-bit values,
so if the entry happens to be 0x80000000 or higher, it is turned to
0xffffffff80000000 which points to memory that doesn't exist.
This wasn't an issue until commit
26f8bf63bf36f9062a5cc1afacf71462a
as all addresses were automatically truncated there in the translate
function to 32-bits. When we cleaned up that code, the full 64-bits
were checked leading to many test failures for mips-sde-elf targets
and such.
+2016-01-12 Mike Frysinger <vapier@gentoo.org>
+
+ * interp.c: Include elf-bfd.h.
+ (sim_create_inferior): Truncate pc to 32-bits when EI_CLASS is
+ ELFCLASS32.
+
2016-01-10 Mike Frysinger <vapier@gentoo.org>
* config.in, configure: Regenerate.
#include "getopt.h"
#include "libiberty.h"
#include "bfd.h"
+#include "elf-bfd.h"
#include "gdb/callback.h" /* GDB simulator callback interface */
#include "gdb/remote-sim.h" /* GDB simulator interface */
for (cpu_nr = 0; cpu_nr < sim_engine_nr_cpus (sd); cpu_nr++)
{
sim_cpu *cpu = STATE_CPU (sd, cpu_nr);
- CPU_PC_SET (cpu, (unsigned64) bfd_get_start_address (abfd));
+ sim_cia pc = bfd_get_start_address (abfd);
+
+ /* We need to undo brain-dead bfd behavior where it sign-extends
+ addresses that are supposed to be unsigned. See the mips bfd
+ sign_extend_vma setting. We have to check the ELF data itself
+ in order to handle o32 & n32 ABIs. */
+ if (abfd->tdata.elf_obj_data->elf_header->e_ident[EI_CLASS] ==
+ ELFCLASS32)
+ pc = (unsigned32) pc;
+
+ CPU_PC_SET (cpu, pc);
}
}