It turns out tile suffered from the same problem as S390. However,
disabling CFI information for the __startcontext on tile was not
sufficient to fix the problem; I think the backtracer will just
blindly try to follow the link register (lr) in that case.
Instead, the change adds a cfi_undefined directive for "lr"
and then arranges to call __startcontext directly when the new
context starts, rather than just synthesizing a return to it.
In addition to being a bit easier now to understand the control
flow, this also allows the cfi_undefined directive to be placed in
a way that causes it to be in force at the address that the "lr"
from the called function points to.
+2015-07-21 Chris Metcalf <cmetcalf@ezchip.com>
+
+ * sysdeps/unix/sysv/linux/tile/makecontext.c (__makecontext): Call
+ __startcontext to initialize the new context.
+ * sysdeps/unix/sysv/linux/tile/setcontext.S (__startcontext): Set
+ up CFI directive to forbid further backtracing.
+
2015-07-21 Marko Myllynen <myllynen@redhat.com>
* charmaps/ANSI_X3.110-1983: Remove obsolete repertoire map
}
va_end (ap);
- /* Pass (*func) to __startcontext in pc. */
- ucp->uc_mcontext.pc = (long) func;
+ /* Start in the trampoline. */
+ ucp->uc_mcontext.pc = (long) __startcontext;
/* Set stack pointer. */
ucp->uc_mcontext.sp = (long) sp;
- /* Set the return address to trampoline. */
- ucp->uc_mcontext.lr = (long) __startcontext;
+ /* Pass FUNC to __startcontext in r31. */
+ ucp->uc_mcontext.gregs[31] = (long) func;
/* Pass ucp->uc_link to __startcontext in r30. */
ucp->uc_mcontext.gregs[30] = (long) ucp->uc_link;
weak_alias (__setcontext, setcontext)
ENTRY (__startcontext)
+ cfi_undefined (lr)
FEEDBACK_ENTER(__startcontext)
+ jalr r31
BEQZ r30, 1f
{
move r0, r30