Merge tag 'trace-v4.15-rc9' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "With the new ORC unwinder, ftrace stack tracing became disfunctional.

  One was that ORC didn't know how to handle the ftrace callbacks in
  general (which Josh fixed).

  The other was that ORC would just bail if it hit a dynamically
  allocated trampoline. Which means all ftrace stack tracing that
  happens from the function tracer would produce no results (that
  includes killing the max stack size tracer). I added a check to the
  ORC unwinder to see if the trampoline belonged to ftrace, and if it
  did, use the orc entry of the static trampoline that was used to
  create the dynamic one (it would be identical).

  Finally, I noticed that the skip values of the stack tracing were out
  of whack. I went through and fixed them up"

* tag 'trace-v4.15-rc9' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Update stack trace skipping for ORC unwinder
  ftrace, orc, x86: Handle ftrace dynamically allocated trampolines
  x86/ftrace: Fix ORC unwinding from ftrace handlers

3 files changed, 66 insertions, 11 deletions

diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
index 81bb565f4497..7e2baf7304ae 100644
--- a/arch/x86/kernel/Makefile
+++ b/arch/x86/kernel/Makefile
@@ -29,10 +29,13 @@ KASAN_SANITIZE_stacktrace.o				:= n
 KASAN_SANITIZE_paravirt.o				:= n
 
 OBJECT_FILES_NON_STANDARD_relocate_kernel_$(BITS).o	:= y
-OBJECT_FILES_NON_STANDARD_ftrace_$(BITS).o		:= y
 OBJECT_FILES_NON_STANDARD_test_nx.o			:= y
 OBJECT_FILES_NON_STANDARD_paravirt_patch_$(BITS).o	:= y
 
+ifdef CONFIG_FRAME_POINTER
+OBJECT_FILES_NON_STANDARD_ftrace_$(BITS).o		:= y
+endif
+
 # If instrumentation of this dir is enabled, boot hangs during first second.
 # Probably could be more selective here, but note that files related to irqs,
 # boot, dumpstack/stacktrace, etc are either non-interesting or can lead to
diff --git a/arch/x86/kernel/ftrace_64.S b/arch/x86/kernel/ftrace_64.S
index 7cb8ba08beb9..ef61f540cf0a 100644
--- a/arch/x86/kernel/ftrace_64.S
+++ b/arch/x86/kernel/ftrace_64.S
@@ -8,6 +8,7 @@
 #include <asm/ftrace.h>
 #include <asm/export.h>
 #include <asm/nospec-branch.h>
+#include <asm/unwind_hints.h>
 
 	.code64
 	.section .entry.text, "ax"
@@ -20,7 +21,6 @@ EXPORT_SYMBOL(__fentry__)
 EXPORT_SYMBOL(mcount)
 #endif
 
-/* All cases save the original rbp (8 bytes) */
 #ifdef CONFIG_FRAME_POINTER
 # ifdef CC_USING_FENTRY
 /* Save parent and function stack frames (rip and rbp) */
@@ -31,7 +31,7 @@ EXPORT_SYMBOL(mcount)
 # endif
 #else
 /* No need to save a stack frame */
-# define MCOUNT_FRAME_SIZE	8
+# define MCOUNT_FRAME_SIZE	0
 #endif /* CONFIG_FRAME_POINTER */
 
 /* Size of stack used to save mcount regs in save_mcount_regs */
@@ -64,10 +64,10 @@ EXPORT_SYMBOL(mcount)
  */
 .macro save_mcount_regs added=0
 
-	/* Always save the original rbp */
+#ifdef CONFIG_FRAME_POINTER
+	/* Save the original rbp */
 	pushq %rbp
 
-#ifdef CONFIG_FRAME_POINTER
 	/*
 	 * Stack traces will stop at the ftrace trampoline if the frame pointer
 	 * is not set up properly. If fentry is used, we need to save a frame
@@ -105,7 +105,11 @@ EXPORT_SYMBOL(mcount)
 	 * Save the original RBP. Even though the mcount ABI does not
 	 * require this, it helps out callers.
 	 */
+#ifdef CONFIG_FRAME_POINTER
 	movq MCOUNT_REG_SIZE-8(%rsp), %rdx
+#else
+	movq %rbp, %rdx
+#endif
 	movq %rdx, RBP(%rsp)
 
 	/* Copy the parent address into %rsi (second parameter) */
@@ -148,7 +152,7 @@ EXPORT_SYMBOL(mcount)
 
 ENTRY(function_hook)
 	retq
-END(function_hook)
+ENDPROC(function_hook)
 
 ENTRY(ftrace_caller)
 	/* save_mcount_regs fills in first two parameters */
@@ -184,7 +188,7 @@ GLOBAL(ftrace_graph_call)
 /* This is weak to keep gas from relaxing the jumps */
 WEAK(ftrace_stub)
 	retq
-END(ftrace_caller)
+ENDPROC(ftrace_caller)
 
 ENTRY(ftrace_regs_caller)
 	/* Save the current flags before any operations that can change them */
@@ -255,7 +259,7 @@ GLOBAL(ftrace_regs_caller_end)
 
 	jmp ftrace_epilogue
 
-END(ftrace_regs_caller)
+ENDPROC(ftrace_regs_caller)
 
 
 #else /* ! CONFIG_DYNAMIC_FTRACE */
@@ -313,9 +317,10 @@ ENTRY(ftrace_graph_caller)
 	restore_mcount_regs
 
 	retq
-END(ftrace_graph_caller)
+ENDPROC(ftrace_graph_caller)
 
-GLOBAL(return_to_handler)
+ENTRY(return_to_handler)
+	UNWIND_HINT_EMPTY
 	subq  $24, %rsp
 
 	/* Save the return values */
@@ -330,4 +335,5 @@ GLOBAL(return_to_handler)
 	movq (%rsp), %rax
 	addq $24, %rsp
 	JMP_NOSPEC %rdi
+END(return_to_handler)
 #endif
diff --git a/arch/x86/kernel/unwind_orc.c b/arch/x86/kernel/unwind_orc.c
index be86a865087a..1f9188f5357c 100644
--- a/arch/x86/kernel/unwind_orc.c
+++ b/arch/x86/kernel/unwind_orc.c
@@ -74,8 +74,50 @@ static struct orc_entry *orc_module_find(unsigned long ip)
 }
 #endif
 
+#ifdef CONFIG_DYNAMIC_FTRACE
+static struct orc_entry *orc_find(unsigned long ip);
+
+/*
+ * Ftrace dynamic trampolines do not have orc entries of their own.
+ * But they are copies of the ftrace entries that are static and
+ * defined in ftrace_*.S, which do have orc entries.
+ *
+ * If the undwinder comes across a ftrace trampoline, then find the
+ * ftrace function that was used to create it, and use that ftrace
+ * function's orc entrie, as the placement of the return code in
+ * the stack will be identical.
+ */
+static struct orc_entry *orc_ftrace_find(unsigned long ip)
+{
+	struct ftrace_ops *ops;
+	unsigned long caller;
+
+	ops = ftrace_ops_trampoline(ip);
+	if (!ops)
+		return NULL;
+
+	if (ops->flags & FTRACE_OPS_FL_SAVE_REGS)
+		caller = (unsigned long)ftrace_regs_call;
+	else
+		caller = (unsigned long)ftrace_call;
+
+	/* Prevent unlikely recursion */
+	if (ip == caller)
+		return NULL;
+
+	return orc_find(caller);
+}
+#else
+static struct orc_entry *orc_ftrace_find(unsigned long ip)
+{
+	return NULL;
+}
+#endif
+
 static struct orc_entry *orc_find(unsigned long ip)
 {
+	static struct orc_entry *orc;
+
 	if (!orc_init)
 		return NULL;
 
@@ -111,7 +153,11 @@ static struct orc_entry *orc_find(unsigned long ip)
 				  __stop_orc_unwind_ip - __start_orc_unwind_ip, ip);
 
 	/* Module lookup: */
-	return orc_module_find(ip);
+	orc = orc_module_find(ip);
+	if (orc)
+		return orc;
+
+	return orc_ftrace_find(ip);
 }
 
 static void orc_sort_swap(void *_a, void *_b, int size)