Support _debuglink recursions.

2006-10-08  Soren Sandmann <sandmann@daimi.au.dk>

	* binfile.c (find_separate_debug_file): Support _debuglink
	recursions.

	* elfparser.c: Check that the sections have valid types before
	using them.

ChangeLog

@@ -1,3 +1,11 @@
+2006-10-08  Soren Sandmann <sandmann@daimi.au.dk>
+
+	* binfile.c (find_separate_debug_file): Support _debuglink
+	recursions.
+
+	* elfparser.c: Check that the sections have valid types before
+	using them. 
+
 2006-10-08  Soren Sandmann  <sandmann@daimi.au.dk>
 
 	* Delete various debug spew

TODO

@@ -28,25 +28,21 @@ Before 1.0.4:
 * Consider renaming sysprof-module.[ch] to sysprof.[ch] to move it closer to 
   kernel naming conventions. 
 
-* Get rid of include of "../config.h" as that won't work in the kernel.
+* Get rid of include of "../config.h" as that won't work in the latest 
+  kernel. done in HEAD, need to check what if anything breaks with older
+  kernels.
 
 Before 1.2:
 
-* Try reproducing crash when profiling xrender demo
-
 * Fix (potential) performance issues in symbol lookup.
 
 * Elf bugs:
-
-	- when an elf file is read, it should be checked that the various
-	  sections are of the right type. For example the debug information
-	  for emacs is just a stub file where all the sections are NOBITS.
-
 	- Also error handling for bin_parser is necessary.
 
-	- Can .gnu_debuglink recurse?
+	- don't loop infinitely if there are cycles in the debuglink graph.
-		yes, it can, and we should probably not crash if there are
+
-		cycles in the graph.
+	* Find out why all apps have an "In file /usr/bin/<app binary>" below
+	  _libc_main. If possible, maybe make up a name for it.
 
 * Strategies for taking reliable stacktraces.
 
@@ -656,6 +652,25 @@ Later:
 
 -=-=-=-=-=-=-=-=-=-=-=-=-=-=- ALREADY DONE -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
 
+
+- when an elf file is read, it should be checked that the various
+  sections are of the right type. For example the debug information
+  for emacs is just a stub file where all the sections are NOBITS.
+
+* Try reproducing crash when profiling xrender demo
+  - it looks like it crashes when it attempts to read /usr/bin/python
+  - apparently what's going on is that one of the symbols in python's
+    dynamic symbol table has a completely crazy 'st_name' offset.
+  DONE: we didn't actually need to read the name at all,
+  but still should find out why that value is so weird.
+  It looks like there is something strange going on with that file.
+  All the dynsyms have weird info/type values, yet nm and readelf
+  have no problems displaying it.
+	
+- Can .gnu_debuglink recurse?
+  yes, it can, and we should probably not crash if there are
+  cycles in the graph.
+
 * Find out why we are getting bogus symbols reported for /usr/bin/Xorg
   Like this:
 

binfile.c

@@ -88,8 +88,9 @@ separate_debug_file_exists (const char *name, guint32 crc)
 static const char *const debug_file_directory = "/usr/lib/debug";
 
 static ElfParser *
-find_separate_debug_file (ElfParser *elf,
+get_debug_file (ElfParser *elf,
-			  const char *filename)
+		const char *filename,
+		char **new_name)
 {
     const char *basename;
     char *dir;
@@ -104,7 +105,7 @@ find_separate_debug_file (ElfParser *elf,
     basename = elf_parser_get_debug_link (elf, &crc32);
 
     if (!basename)
-	return elf;
+	return NULL;
 
     dir = g_path_get_dirname (filename);
     
@@ -116,7 +117,11 @@ find_separate_debug_file (ElfParser *elf,
     {
 	result = separate_debug_file_exists (tries[i], crc32);
 	if (result)
+	{
+	    if (new_name)
+		*new_name = g_strdup (tries[i]);
 	    break;
+	}
     }
 
     g_free (dir);
@@ -124,15 +129,40 @@ find_separate_debug_file (ElfParser *elf,
     for (i = 0; i < 3; ++i)
 	g_free (tries[i]);
 
-    if (result)
+    return result;
+}
+
+static ElfParser *
+find_separate_debug_file (ElfParser *elf,
+			  const char *filename)
+{
+    ElfParser *debug;
+    char *debug_name = NULL;
+    char *fname;
+
+    fname = g_strdup (filename);
+
+    /* FIXME: there is an infinite loop if there are cycles in
+     * the debug_link graph
+     */
+    do
     {
-	elf_parser_free (elf);
+	debug = get_debug_file (elf, fname, &debug_name);
-	return result;
+
-    }
+	if (debug)
-    else
+	{
-    {
+	    elf_parser_free (elf);
-	return elf;
+	    elf = debug;
+
+	    g_free (fname);
+	    fname = debug_name;
+	}
     }
+    while (debug);
+
+    g_free (fname);
+
+    return elf;
 }
 
 static GHashTable *bin_files;
@@ -163,7 +193,7 @@ bin_file_new (const char *filename)
 	if (bf->elf)
 	    bf->text_offset = elf_parser_get_text_offset (bf->elf);
 	
-	bf->elf = find_separate_debug_file (bf->elf, filename);
+	bf->elf = find_separate_debug_file (bf->elf, filename); /* find_separate_debug_file (bf->elf, filename); */
 	bf->inode = read_inode (filename);
 	bf->filename = g_strdup (filename);
 	bf->undefined_name = g_strdup_printf ("In file %s", filename);
@@ -210,6 +240,14 @@ bin_file_lookup_symbol (BinFile    *bin_file,
 	    return (const BinSymbol *)sym;
 	}
     }
+#if 0
+    else
+	g_print ("no elf file for %s\n", bin_file->filename);
+#endif
+
+#if 0
+    g_print ("%lx undefined in %s (textoffset %d)\n", address + bin_file->text_offset, bin_file->filename, bin_file->text_offset);
+#endif
     
     return (const BinSymbol *)bin_file->undefined_name;
 }

elfparser.c

@@ -19,6 +19,7 @@ struct Section
     gsize		size;
     gboolean		allocated;
     gulong		load_address;
+    guint		type;
 };
 
 struct ElfParser
@@ -80,6 +81,8 @@ section_new (BinRecord *record,
     else
 	section->load_address = 0;
 
+    section->type = bin_record_get_uint (record, "sh_type");
+
     return section;
 }
 
@@ -91,7 +94,8 @@ section_free (Section *section)
 
 static const Section *
 find_section (ElfParser *parser,
-	      const char *name)
+	      const char *name,
+	      guint	  type)
 {
     int i;
 
@@ -103,7 +107,7 @@ find_section (ElfParser *parser,
     {
 	Section *section = parser->sections[i];
 	
-	if (strcmp (section->name, name) == 0)
+	if (strcmp (section->name, name) == 0 && section->type == type)
 	{
 #if 0
 	    g_print ("found it as number %d with offset %d\n", i, section->offset);
@@ -191,6 +195,10 @@ elf_parser_new (const char *filename,
     if (!file)
 	return NULL;
 
+#if 0
+    g_print ("elf parser new : %s\n", filename);
+#endif
+    
     data = (guchar *)g_mapped_file_get_contents (file);
     length = g_mapped_file_get_length (file);
 
@@ -200,6 +208,11 @@ elf_parser_new (const char *filename,
 
     parser = parser_new_from_data (data, length);
 
+#if 0
+    g_print ("Elf file: %s  (debug: %s)\n",
+	     filename, elf_parser_get_debug_link (parser, NULL));
+#endif
+    
     parser->file = file;
 
     return parser;
@@ -370,15 +383,29 @@ read_table (ElfParser *parser,
     {
 	guint info;
 	gulong addr;
-	const char *name;
 	gulong offset;
+#if 0
+	const char *name;
+#endif
 
 	bin_record_index (symbol, i);
 	
 	info = bin_record_get_uint (symbol, "st_info");
+
+#if 0
+	g_print ("info: %d  =>   %d %d\n", info, info & 0xf, info >> 4);
+#endif
+#if 0
+	g_print ("func: %d global %d local: %d\n", STT_FUNC, STB_GLOBAL, STB_LOCAL);
+#endif
+#if 0
+	g_print ("%d, name off: %d\n", i, bin_record_get_uint (symbol, "st_name"));
+#endif
 	addr = bin_record_get_uint (symbol, "st_value");
+#if 0
 	name = bin_record_get_string_indirect (symbol, "st_name",
 					       str_table->offset);
+#endif
 	offset = bin_record_get_offset (symbol);
 	
 	if (addr != 0				&&
@@ -389,10 +416,22 @@ read_table (ElfParser *parser,
 	    parser->symbols[n_functions].address = addr;
 	    parser->symbols[n_functions].offset = offset;
 
+#if 0
+	    g_print ("name: %s\n",
+		     bin_record_get_string_indirect (symbol, "st_name",
+						     str_table->offset));
+	    
+	    g_print ("%lx\n", parser->symbols[n_functions].address);
+#endif
+	    
 	    n_functions++;
 	}
     }
 
+#if 0
+    g_print ("%d functions found \n", n_functions);
+#endif
+    
     bin_record_free (symbol);
 
 #if 0
@@ -413,10 +452,10 @@ read_table (ElfParser *parser,
 static void
 read_symbols (ElfParser *parser)
 {
-    const Section *symtab = find_section (parser, ".symtab");
+    const Section *symtab = find_section (parser, ".symtab", SHT_SYMTAB);
-    const Section *strtab = find_section (parser, ".strtab");
+    const Section *strtab = find_section (parser, ".strtab", SHT_STRTAB);
-    const Section *dynsym = find_section (parser, ".dynsym");
+    const Section *dynsym = find_section (parser, ".dynsym", SHT_SYMTAB);
-    const Section *dynstr = find_section (parser, ".dynstr");
+    const Section *dynstr = find_section (parser, ".dynstr", SHT_STRTAB);
 
     if (symtab && strtab)
     {
@@ -513,9 +552,13 @@ elf_parser_lookup_symbol (ElfParser *parser,
     if (parser->n_symbols == 0)
 	return NULL;
 
-    text = find_section (parser, ".text");
+    text = find_section (parser, ".text", SHT_PROGBITS);
     if (!text)
-	return NULL;
+    {
+	text = find_section (parser, ".text", SHT_NOBITS);
+	if (!text)
+	    return NULL;
+    }
     
     address += text->load_address;
 
@@ -558,10 +601,14 @@ elf_parser_get_text_offset (ElfParser *parser)
 
     g_return_val_if_fail (parser != NULL, (gulong)-1);
     
-    text = find_section (parser, ".text");
+    text = find_section (parser, ".text", SHT_PROGBITS);
 
     if (!text)
-	return (gulong)-1;
+    {
+	text = find_section (parser, ".text", SHT_NOBITS);
+	if (!text)
+	    return (gulong)-1;
+    }
 
     return text->offset;
 }
@@ -569,7 +616,8 @@ elf_parser_get_text_offset (ElfParser *parser)
 const char *
 elf_parser_get_debug_link (ElfParser *parser, guint32 *crc32)
 {
-    const Section *debug_link = find_section (parser, ".gnu_debuglink");
+    const Section *debug_link = find_section (parser, ".gnu_debuglink",
+					      SHT_PROGBITS);
     const gchar *result;
 
     if (!debug_link)
@@ -581,7 +629,8 @@ elf_parser_get_debug_link (ElfParser *parser, guint32 *crc32)
 
     bin_parser_align (parser->parser, 4);
 
-    *crc32 = bin_parser_get_uint32 (parser->parser);
+    if (crc32)
+	*crc32 = bin_parser_get_uint32 (parser->parser);
     
     return result;
 }
@@ -589,7 +638,7 @@ elf_parser_get_debug_link (ElfParser *parser, guint32 *crc32)
 const guchar *
 elf_parser_get_eh_frame   (ElfParser   *parser)
 {
-    const Section *eh_frame = find_section (parser, ".eh_frame");
+    const Section *eh_frame = find_section (parser, ".eh_frame", SHT_PROGBITS);
 
     if (eh_frame)
 	return bin_parser_get_data (parser->parser) + eh_frame->offset;

process.c

@@ -586,6 +586,11 @@ process_lookup_symbol (Process *process, gulong address)
     g_print ("address before: \n");
 #endif
 
+#if 0
+    g_print ("%s is mapped at %lx + %lx\n", map->filename, map->start, map->offset);
+    g_print ("incoming address: %lx\n", address);
+#endif
+    
     address -= map->start;
     address += map->offset;