PR10524: reduce massive-uprobe script modules' .data consumption

The general approach is to rip out task_finder_tgt's from all over the
place (including the unwindsym vmcbs, and the stap_uprobe_specs), and
instead have a small handful of them: one for all unwindsyms, and one
per abstract probed process (PID or NAME).  These are in turn shared
by all new stap_unwind_specs by index.

Before:  probe process("./stap").function("*")

-rw-r--r--. 1 fche users 11775283 2009-09-08 20:26 /var/tmp/fche/systemtap/cache/96/stap_96c0479d674db55ec98d8a8750a790e7_7989596.ko
   text    data     bss     dec     hex filename
 445158 8351944 4306472 13103574 c7f1d6 /var/tmp/fche/systemtap/cache/96/stap_96c0479d674db55ec98d8a8750a790e7_7989596.ko

After: (Note how data shrank, though text gained a bit in const-init-data.)

-rw-r--r--. 1 fche users  4021569 2009-09-08 20:27 /var/tmp/fche/systemtap/cache/e4/stap_e46e88634efd850b1586e81c231c239a_8058419.ko
   text    data     bss     dec     hex filename
1896511    2192 4324808 6223511  5ef697 /var/tmp/fche/systemtap/cache/e4/stap_e46e88634efd850b1586e81c231c239a_8058419.ko

* tapsets.cxx (uprobe_derived_probe_group): Rewrite emit_module_decls,
  and adjust emit_module_init.
* runtime/sym.c (_stp_sym_init): Initialize unwindsyms-shared vmcb.
* runtime/sym.h (_stp_module): Remove *vmcb field.
* translate.cxx (emit_module_init, dump_unwindsyms): Adapt.
* translate.h (assert_0_indent): Flush output before possibly assert-failing.

1 files changed, 4 insertions, 32 deletions

diff --git a/translate.cxx b/translate.cxx
index 88968d45..e32f932a 100644
--- a/translate.cxx
+++ b/translate.cxx
@@ -1161,22 +1161,12 @@ c_unparser::emit_module_init ()
   o->newline(-1) << "}";
   o->newline() << "#endif";
 
-  // PR10228: set up symbol table-related task finders
-  o->newline() << "#if defined(STP_NEED_VMA_TRACKER) && defined(CONFIG_UTRACE)";
+  // PR10228: set up symbol table-related task finders.
+  o->newline() << "#if defined(STP_NEED_VMA_TRACKER)";
   o->newline() << "_stp_sym_init();";
-  o->newline() << "for (i=0; i<_stp_num_modules; i++) {";
-  o->newline(1) << "if (_stp_modules[i]->vmcb) {";
-  o->newline(1) << "rc = stap_register_task_finder_target (_stp_modules[i]->vmcb);";
-  o->newline() << "if (rc) {"; 
-  o->newline(1) << "_stp_error ("
-                << "\"couldn't initialize task-finder for %s\\n\","
-                << "_stp_modules[i]->vmcb->procname);";
-  o->newline() << "goto out;";
-  o->newline(-1) << "}";
-  o->newline(-1) << "}";
-  o->newline(-1) << "}";
   o->newline() << "#endif";
-
+  // NB: we don't need per-_stp_module task_finders, since a single common one
+  // set up in runtime/sym.c's _stp_sym_init() will scan through all _stp_modules.
   o->newline() << "(void) probe_point;";
   o->newline() << "(void) i;";
   o->newline() << "(void) j;";
@@ -4857,27 +4847,9 @@ dump_unwindsyms (Dwfl_Module *m,
 
   mainfile = canonicalize_file_name(mainfile);
 
-  // PR10228: populate the task_finder_vmcb.
-  if (modname[0] == '/') // user-space module
-    {
-      // NB: runtime/sym.c
-      c->output << "static struct stap_task_finder_target _stp_vmcb_" << stpmod_idx << "= {\n";
-      c->output << "#ifdef CONFIG_UTRACE\n";
-      c->output << ".procname = " << lex_cast_qstring (mainfile) << ",\n";
-      c->output << ".mmap_callback = &_stp_tf_mmap_cb,\n";
-      c->output << ".munmap_callback = &_stp_tf_munmap_cb,\n";
-      c->output << "#endif\n";
-      c->output << "};\n";
-    }
-
   c->output << "static struct _stp_module _stp_module_" << stpmod_idx << " = {\n";
   c->output << ".name = " << lex_cast_qstring (modname) << ", \n";
   c->output << ".path = " << lex_cast_qstring (mainfile) << ",\n";
-
-  // PR10228: populate the task_finder_vmcb.
-  if (modname[0] == '/') // user-space module
-      c->output << ".vmcb = & _stp_vmcb_" << stpmod_idx << ",\n";
-
   c->output << ".dwarf_module_base = 0x" << hex << base << dec << ", \n";
   c->output << ".eh_frame_addr = 0x" << hex << eh_addr << dec << ", \n";