[PATCH] support lookup table for regular kernel markers

[Masami Hiramatsu <mhiramat at redhat dot com>]

Hi,

Here is a patch for supporting regular kernel markers
which I posted on LKML. This is for explaining how
new c-style format can be used on systemtap/other tracers.

runtime/regular-marker.format is a lookup table for
regular kernel markers. This file has pairs of
c-style format and printf-style format as below;

<marker name><c-style format>:<printf-style format>

Module.markers provides marker names and c-style formats,
so systemtap can look up corresponding printf-style formats
from this file. So, we can use printf-format parser for
regular kernel markers.

Even if the kernel updates its format, what we have to do is
just add new pairs.

Thank you,

-- 
Masami Hiramatsu

Software Engineer
Hitachi Computer Products (America) Inc.
Software Solutions Division

e-mail: mhiramat@redhat.com

---
 Makefile.am                   |    2 -
 Makefile.in                   |    2 -
 runtime/regular-marker.format |    2 +
 tapsets.cxx                   |   60 +++++++++++++++++++++++++++++++++++++-----
 4 files changed, 57 insertions(+), 9 deletions(-)

Index: systemtap/runtime/regular-marker.format
===================================================================
--- /dev/null	1970-01-01 00:00:00.000000000 +0000
+++ systemtap/runtime/regular-marker.format	2008-06-13 16:49:51.000000000 -0400
@@ -0,0 +1,2 @@
+irq_entry(int irq_id, int kernel_mode):irq_id %d kernel_mode %d
+irq_exit(void): 
Index: systemtap/tapsets.cxx
===================================================================
--- systemtap.orig/tapsets.cxx	2008-06-13 16:49:39.000000000 -0400
+++ systemtap/tapsets.cxx	2008-06-13 16:52:48.000000000 -0400
@@ -6765,10 +6765,11 @@
 {
   mark_derived_probe (systemtap_session &s,
                       const string& probe_name, const string& probe_format,
+                      const string& probe_signature,
                       probe* base_probe, probe_point* location);
 
   systemtap_session& sess;
-  string probe_name, probe_format;
+  string probe_name, probe_format, probe_signature;
   vector <struct mark_arg *> mark_args;
   bool target_symbol_seen;
 
@@ -6980,9 +6981,10 @@
 mark_derived_probe::mark_derived_probe (systemtap_session &s,
                                         const string& p_n,
                                         const string& p_f,
+                                        const string& p_s,
                                         probe* base, probe_point* loc):
   derived_probe (base, new probe_point(*loc) /* .components soon rewritten */),
-  sess (s), probe_name (p_n), probe_format (p_f),
+  sess (s), probe_name (p_n), probe_format (p_f), probe_signature (p_s),
   target_symbol_seen (false)
 {
   // create synthetic probe point name; preserve condition
@@ -7001,8 +7003,14 @@
   target_symbol_seen = v.target_symbol_seen;
 
   if (sess.verbose > 2)
-    clog << "marker-based " << name << " mark=" << probe_name
-	 << " fmt='" << probe_format << "'" << endl;
+    {
+      clog << "marker-based " << name << " mark=" << probe_name
+	   << " fmt='" << probe_format << "'";
+      if (probe_signature != probe_format)
+	clog << " sig='" << probe_signature << "'" << endl;
+      else
+	clog << endl;
+    }
 }
 
 
@@ -7253,7 +7261,7 @@
       s.op->newline () << "{";
       s.op->line() << " .name=" << lex_cast_qstring(probes[i]->probe_name)
 		   << ",";
-      s.op->line() << " .format=" << lex_cast_qstring(probes[i]->probe_format)
+      s.op->line() << " .format=" << lex_cast_qstring(probes[i]->probe_signature)
 		   << ",";
       s.op->line() << " .pp=" << lex_cast_qstring (*probes[i]->sole_location())
 		   << ",";
@@ -7329,6 +7337,7 @@
   typedef pair<mark_cache_const_iterator_t, mark_cache_const_iterator_t>
     mark_cache_const_iterator_pair_t;
   mark_cache_t mark_cache;
+  map<string, string> regular_mark;
 
 public:
   mark_builder(): cache_initialized(false) {}
@@ -7340,6 +7349,7 @@
         if (s.verbose > 3)
           clog << "mark_builder releasing cache" << endl;
 	mark_cache.clear();
+        regular_mark.clear();
       }
   }
 
@@ -7370,6 +7380,7 @@
       cache_initialized = true;
       string module_markers_path = "/lib/modules/" + sess.kernel_release
 	  + "/build/Module.markers";
+      string regular_format_path = sess.runtime_path + "/regular-marker.format";
 
       ifstream module_markers;
       module_markers.open(module_markers_path.c_str(), ifstream::in);
@@ -7381,6 +7392,33 @@
 	  return;
 	}
 
+      // read regular format database
+      ifstream regular_format;
+      regular_format.open(regular_format_path.c_str(), ifstream::in);
+      if (regular_format)
+        {
+        string format;
+	  do
+	    {
+	      getline(regular_format, format);
+	      string::size_type idx = format.find(":");
+	      if (idx != string::npos)
+		{
+		  regular_mark[format.substr(0, idx)] =
+		      format.substr(idx + 1);
+		}
+	    }
+	  while (! regular_format.eof());
+	  regular_format.close();
+	  if (sess.verbose>3)
+	    {
+	      map<string, string>::iterator p;
+	      for (p = regular_mark.begin(); p != regular_mark.end(); p++)
+		clog << "format:'" << p->first << "'->'"<< p->second
+		     << "'" << endl;
+	    }
+        }
+
       string name, module, format;
       do
         {
@@ -7439,17 +7477,25 @@
       if (! rc)
         {
 	  bool add_result = true;
+	  map<string, string>::iterator p;
+          string format;
+          // search regular format
+	  p = regular_mark.find(it->first + it->second);
+	  if (p != regular_mark.end())
+	    format = p->second;
+	  else
+	    format = it->second;
 
 	  // Match format strings (if the user specified one)
 	  if (has_mark_format && fnmatch(mark_format_val.c_str(),
-					 it->second.c_str(), 0))
+					 format.c_str(), 0))
 	    add_result = false;
 
 	  if (add_result)
 	    {
 	      derived_probe *dp
 		= new mark_derived_probe (sess,
-					  it->first, it->second,
+					  it->first, format, it->second,
 					  base, loc);
 	      finished_results.push_back (dp);
 	    }
Index: systemtap/Makefile.am
===================================================================
--- systemtap.orig/Makefile.am	2008-06-13 16:49:39.000000000 -0400
+++ systemtap/Makefile.am	2008-06-13 16:49:51.000000000 -0400
@@ -186,7 +186,7 @@
 	find $(distdir) -name 'stap' -o -name '*.log' -o -name '*.sum' -o -name 'site.exp' | xargs rm -rf
 
 install-data-local:
-	(cd $(srcdir)/runtime; for f in *.[ch]; do $(INSTALL_DATA) -D $$f $(DESTDIR)$(pkgdatadir)/runtime/$$f; done)
+	(cd $(srcdir)/runtime; for f in *.[ch] regular-marker.format; do $(INSTALL_DATA) -D $$f $(DESTDIR)$(pkgdatadir)/runtime/$$f; done)
 	(cd $(srcdir)/runtime/unwind; find . \( -name '*.[ch]' \) -print \
 		| while read f; do $(INSTALL_DATA) -D $$f $(DESTDIR)$(pkgdatadir)/runtime/unwind/$$f; done)
 	(cd $(srcdir)/runtime/transport; for f in *.[ch]; \
Index: systemtap/Makefile.in
===================================================================
--- systemtap.orig/Makefile.in	2008-06-13 16:49:39.000000000 -0400
+++ systemtap/Makefile.in	2008-06-13 16:49:51.000000000 -0400
@@ -1555,7 +1555,7 @@
 	find $(distdir) -name 'stap' -o -name '*.log' -o -name '*.sum' -o -name 'site.exp' | xargs rm -rf
 
 install-data-local:
-	(cd $(srcdir)/runtime; for f in *.[ch]; do $(INSTALL_DATA) -D $$f $(DESTDIR)$(pkgdatadir)/runtime/$$f; done)
+	(cd $(srcdir)/runtime; for f in *.[ch] regular-marker.format; do $(INSTALL_DATA) -D $$f $(DESTDIR)$(pkgdatadir)/runtime/$$f; done)
 	(cd $(srcdir)/runtime/unwind; find . \( -name '*.[ch]' \) -print \
 		| while read f; do $(INSTALL_DATA) -D $$f $(DESTDIR)$(pkgdatadir)/runtime/unwind/$$f; done)
 	(cd $(srcdir)/runtime/transport; for f in *.[ch]; \