new helper: signal_pt_regs()

Always equal to task_pt_regs(current); defined only when we are in
signal delivery.  It may be different from current_pt_regs() - e.g.
architectures like m68k may have pt_regs location on exception
different from that on a syscall and signals (just as ptrace handling)
may happen on exceptions as well as on syscalls.

When they are equal, it's often better to have signal_pt_regs
defined (in asm/ptrace.h) as current_pt_regs - that tends to be
optimized better than default would be.  However, optimisation is
the only reason why we might want an arch-specific definition;
if current_pt_regs() and task_pt_regs(current) have different
values, the latter one is right.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

1 files changed, 9 insertions, 0 deletions

diff --git a/include/linux/ptrace.h b/include/linux/ptrace.h
index 7aefbae2452e..b8e6dcec78ae 100644
--- a/include/linux/ptrace.h
+++ b/include/linux/ptrace.h
@@ -333,6 +333,15 @@ static inline void user_single_step_siginfo(struct task_struct *tsk,
 #define ptrace_signal_deliver(regs, cookie) do { } while (0)
 #endif
 
+/*
+ * unlike current_pt_regs(), this one is equal to task_pt_regs(current)
+ * on *all* architectures; the only reason to have a per-arch definition
+ * is optimisation.
+ */
+#ifndef signal_pt_regs
+#define signal_pt_regs() task_pt_regs(current)
+#endif
+
 extern int task_current_syscall(struct task_struct *target, long *callno,
                                 unsigned long args[6], unsigned int maxargs,
                                 unsigned long *sp, unsigned long *pc);