libgo: reduce overhead for memory/block/mutex profiling

    
    Revise the gccgo version of memory/block/mutex profiling to reduce
    runtime overhead. The main change is to collect raw stack traces while
    the profile is on line, then post-process the stacks just prior to the
    point where we are ready to use the final product. Memory profiling
    (at a very low sampling rate) is enabled by default, and the overhead
    of the symbolization / DWARF-reading from backtrace_full was slowing
    things down relative to the main Go runtime.
    
    Reviewed-on: https://go-review.googlesource.com/c/gofrontend/+/171497

From-SVN: r271172

3 files changed, 75 insertions, 7 deletions

diff --git a/libgo/runtime/go-caller.c b/libgo/runtime/go-caller.c
index 2143446..5e31f91 100644
--- a/libgo/runtime/go-caller.c
+++ b/libgo/runtime/go-caller.c
@@ -26,11 +26,13 @@ struct caller
   String file;
   intgo line;
   intgo index;
+  intgo frames;
 };
 
 /* Collect file/line information for a PC value.  If this is called
-   more than once, due to inlined functions, we use the last call, as
-   that is usually the most useful one.  */
+   more than once, due to inlined functions, we record the number of
+   inlined frames but return file/func/line for the last call, as
+   that is usually the most useful one.   */
 
 static int
 callback (void *data, uintptr_t pc __attribute__ ((unused)),
@@ -38,6 +40,8 @@ callback (void *data, uintptr_t pc __attribute__ ((unused)),
 {
   struct caller *c = (struct caller *) data;
 
+  c->frames++;
+
   /* The libbacktrace library says that these strings might disappear,
      but with the current implementation they won't.  We can't easily
      allocate memory here, so for now assume that we can save a
@@ -125,18 +129,19 @@ __go_get_backtrace_state ()
   return back_state;
 }
 
-/* Return function/file/line information for PC.  The index parameter
+/* Return function/file/line/nframes information for PC.  The index parameter
    is the entry on the stack of inlined functions; -1 means the last
-   one.  */
+   one, with *nframes set to the count of inlined frames for this PC.  */
 
 static _Bool
-__go_file_line (uintptr pc, int index, String *fn, String *file, intgo *line)
+__go_file_line (uintptr pc, int index, String *fn, String *file, intgo *line, intgo *nframes)
 {
   struct caller c;
   struct backtrace_state *state;
 
   runtime_memclr (&c, sizeof c);
   c.index = index;
+  c.frames = 0;
   runtime_xadd (&__go_runtime_in_callers, 1);
   state = __go_get_backtrace_state ();
   runtime_xadd (&__go_runtime_in_callers, -1);
@@ -144,6 +149,7 @@ __go_file_line (uintptr pc, int index, String *fn, String *file, intgo *line)
   *fn = c.fn;
   *file = c.file;
   *line = c.line;
+  *nframes = c.frames;
 
   // If backtrace_pcinfo didn't get the function name from the debug
   // info, try to get it from the symbol table.
@@ -222,7 +228,7 @@ runtime_funcfileline (uintptr targetpc, int32 index)
   struct funcfileline_return ret;
 
   if (!__go_file_line (targetpc, index, &ret.retfn, &ret.retfile,
-		       &ret.retline))
+		       &ret.retline, &ret.retframes))
     runtime_memclr (&ret, sizeof ret);
   return ret;
 }
diff --git a/libgo/runtime/go-callers.c b/libgo/runtime/go-callers.c
index 31ff474..4a9c1a7 100644
--- a/libgo/runtime/go-callers.c
+++ b/libgo/runtime/go-callers.c
@@ -63,7 +63,9 @@ callback (void *data, uintptr_t pc, const char *filename, int lineno,
 
   /* Skip thunks and recover functions.  There is no equivalent to
      these functions in the gc toolchain, so returning them here means
-     significantly different results for runtime.Caller(N).  */
+     significantly different results for runtime.Caller(N). See also
+     similar code in runtime/mprof.go that strips out such functions
+     for block/mutex/memory profiles.  */
   if (function != NULL && !arg->keep_thunks)
     {
       const char *p;
@@ -262,3 +264,62 @@ Callers (intgo skip, struct __go_open_array pc)
 
   return ret;
 }
+
+struct callersRaw_data
+{
+  uintptr* pcbuf;
+  int skip;
+  int index;
+  int max;
+};
+
+// Callback function for backtrace_simple.  Just collect pc's.
+// Return zero to continue, non-zero to stop.
+
+static int callback_raw (void *data, uintptr_t pc)
+{
+  struct callersRaw_data *arg = (struct callersRaw_data *) data;
+
+  if (arg->skip > 0)
+    {
+      --arg->skip;
+      return 0;
+    }
+
+  /* On the call to backtrace_simple the pc value was most likely
+     decremented if there was a normal call, since the pc referred to
+     the instruction where the call returned and not the call itself.
+     This was done so that the line number referred to the call
+     instruction.  To make sure the actual pc from the call stack is
+     used, it is incremented here.
+
+     In the case of a signal, the pc was not decremented by
+     backtrace_full but still incremented here.  That doesn't really
+     hurt anything since the line number is right and the pc refers to
+     the same instruction.  */
+
+  arg->pcbuf[arg->index] = pc + 1;
+  arg->index++;
+  return arg->index >= arg->max;
+}
+
+/* runtime_callersRaw is similar to runtime_callers() above, but
+   it returns raw PC values as opposed to file/func/line locations. */
+int32
+runtime_callersRaw (int32 skip, uintptr *pcbuf, int32 m)
+{
+  struct callersRaw_data data;
+  struct backtrace_state* state;
+
+  data.pcbuf = pcbuf;
+  data.skip = skip + 1;
+  data.index = 0;
+  data.max = m;
+  runtime_xadd (&__go_runtime_in_callers, 1);
+  state = __go_get_backtrace_state ();
+  backtrace_simple (state, 0, callback_raw, error_callback, &data);
+  runtime_xadd (&__go_runtime_in_callers, -1);
+
+  return data.index;
+}
+
diff --git a/libgo/runtime/runtime.h b/libgo/runtime/runtime.h
index 9506e3d..71c1a3e 100644
--- a/libgo/runtime/runtime.h
+++ b/libgo/runtime/runtime.h
@@ -485,6 +485,7 @@ struct funcfileline_return
   String retfn;
   String retfile;
   intgo retline;
+  intgo retframes;
 };
 
 struct funcfileline_return