tracing, x86: add low level support for ftrace return tracing

Impact: add infrastructure for function-return tracing

Add low level support for ftrace return tracing.

This plug-in stores return addresses on the thread_info structure of
the current task.

The index of the current return address is initialized when the task
is the first one (init) and when a process forks (the child). It is
not needed when a task does a sys_execve because after this syscall,
it still needs to return on the kernel functions it called.

Note that the code of return_to_handler has been suggested by Steven
Rostedt as almost all of the ideas of improvements in this V3.

For purpose of security, arch/x86/kernel/process_32.c is not traced
because __switch_to() changes the current task during its execution.
That could cause inconsistency in the stored return address of this
function even if I didn't have any crash after testing with tracing on
this function enabled.

Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

arch/x86/kernel/ftrace.c

@@ -18,10 +18,173 @@
 #include <linux/list.h>
 
 #include <asm/ftrace.h>
+#include <linux/ftrace.h>
 #include <asm/nops.h>
+#include <asm/nmi.h>
 
 
-static unsigned char ftrace_nop[MCOUNT_INSN_SIZE];
+
+#ifdef CONFIG_FUNCTION_RET_TRACER
+
+/*
+ * These functions are picked from those used on
+ * this page for dynamic ftrace. They have been
+ * simplified to ignore all traces in NMI context.
+ */
+static atomic_t in_nmi;
+
+void ftrace_nmi_enter(void)
+{
+	atomic_inc(&in_nmi);
+}
+
+void ftrace_nmi_exit(void)
+{
+	atomic_dec(&in_nmi);
+}
+
+/*
+ * Synchronize accesses to return adresses stack with
+ * interrupts.
+ */
+static raw_spinlock_t ret_stack_lock;
+
+/* Add a function return address to the trace stack on thread info.*/
+static int push_return_trace(unsigned long ret, unsigned long long time,
+				unsigned long func)
+{
+	int index;
+	struct thread_info *ti;
+	unsigned long flags;
+	int err = 0;
+
+	raw_local_irq_save(flags);
+	__raw_spin_lock(&ret_stack_lock);
+
+	ti = current_thread_info();
+	/* The return trace stack is full */
+	if (ti->curr_ret_stack == FTRACE_RET_STACK_SIZE - 1) {
+		err = -EBUSY;
+		goto out;
+	}
+
+	index = ++ti->curr_ret_stack;
+	ti->ret_stack[index].ret = ret;
+	ti->ret_stack[index].func = func;
+	ti->ret_stack[index].calltime = time;
+
+out:
+	__raw_spin_unlock(&ret_stack_lock);
+	raw_local_irq_restore(flags);
+	return err;
+}
+
+/* Retrieve a function return address to the trace stack on thread info.*/
+static void pop_return_trace(unsigned long *ret, unsigned long long *time,
+				unsigned long *func)
+{
+	struct thread_info *ti;
+	int index;
+	unsigned long flags;
+
+	raw_local_irq_save(flags);
+	__raw_spin_lock(&ret_stack_lock);
+
+	ti = current_thread_info();
+	index = ti->curr_ret_stack;
+	*ret = ti->ret_stack[index].ret;
+	*func = ti->ret_stack[index].func;
+	*time = ti->ret_stack[index].calltime;
+	ti->curr_ret_stack--;
+
+	__raw_spin_unlock(&ret_stack_lock);
+	raw_local_irq_restore(flags);
+}
+
+/*
+ * Send the trace to the ring-buffer.
+ * @return the original return address.
+ */
+unsigned long ftrace_return_to_handler(void)
+{
+	struct ftrace_retfunc trace;
+	pop_return_trace(&trace.ret, &trace.calltime, &trace.func);
+	trace.rettime = cpu_clock(raw_smp_processor_id());
+	ftrace_function_return(&trace);
+
+	return trace.ret;
+}
+
+/*
+ * Hook the return address and push it in the stack of return addrs
+ * in current thread info.
+ */
+asmlinkage
+void prepare_ftrace_return(unsigned long *parent, unsigned long self_addr)
+{
+	unsigned long old;
+	unsigned long long calltime;
+	int faulted;
+	unsigned long return_hooker = (unsigned long)
+				&return_to_handler;
+
+	/* Nmi's are currently unsupported */
+	if (atomic_read(&in_nmi))
+		return;
+
+	/*
+	 * Protect against fault, even if it shouldn't
+	 * happen. This tool is too much intrusive to
+	 * ignore such a protection.
+	 */
+	asm volatile(
+		"1: movl (%[parent_old]), %[old]\n"
+		"2: movl %[return_hooker], (%[parent_replaced])\n"
+		"   movl $0, %[faulted]\n"
+
+		".section .fixup, \"ax\"\n"
+		"3: movl $1, %[faulted]\n"
+		".previous\n"
+
+		".section __ex_table, \"a\"\n"
+		"   .long 1b, 3b\n"
+		"   .long 2b, 3b\n"
+		".previous\n"
+
+		: [parent_replaced] "=rm" (parent), [old] "=r" (old),
+		  [faulted] "=r" (faulted)
+		: [parent_old] "0" (parent), [return_hooker] "r" (return_hooker)
+		: "memory"
+	);
+
+	if (WARN_ON(faulted)) {
+		unregister_ftrace_return();
+		return;
+	}
+
+	if (WARN_ON(!__kernel_text_address(old))) {
+		unregister_ftrace_return();
+		*parent = old;
+		return;
+	}
+
+	calltime = cpu_clock(raw_smp_processor_id());
+
+	if (push_return_trace(old, calltime, self_addr) == -EBUSY)
+		*parent = old;
+}
+
+static int __init init_ftrace_function_return(void)
+{
+	ret_stack_lock = (raw_spinlock_t)__RAW_SPIN_LOCK_UNLOCKED;
+	return 0;
+}
+device_initcall(init_ftrace_function_return);
+
+
+#endif
+
+#ifdef CONFIG_DYNAMIC_FTRACE
 
 union ftrace_code_union {
 	char code[MCOUNT_INSN_SIZE];
@@ -31,17 +194,11 @@ union ftrace_code_union {
 	} __attribute__((packed));
 };
 
-
 static int ftrace_calc_offset(long ip, long addr)
 {
 	return (int)(addr - ip);
 }
 
-unsigned char *ftrace_nop_replace(void)
-{
-	return ftrace_nop;
-}
-
 unsigned char *ftrace_call_replace(unsigned long ip, unsigned long addr)
 {
 	static union ftrace_code_union calc;
@@ -183,6 +340,15 @@ do_ftrace_mod_code(unsigned long ip, void *new_code)
 }
 
 
+
+
+static unsigned char ftrace_nop[MCOUNT_INSN_SIZE];
+
+unsigned char *ftrace_nop_replace(void)
+{
+	return ftrace_nop;
+}
+
 int
 ftrace_modify_code(unsigned long ip, unsigned char *old_code,
 		   unsigned char *new_code)
@@ -292,3 +458,4 @@ int __init ftrace_dyn_arch_init(void *data)
 
 	return 0;
 }
+#endif