[patch] Fix crash on `info files' with gdbserver

[patch] Fix crash on `info files' with gdbserver

[Jan Kratochvil]

Hi,

while playing with gdbserver GDB just crashed for me.

The problem is exec_ops get pushed to the target stack for the loaded shared
libraries despite exec_bfd remains NULL which is not expected by exec.c.

Breakpoint 1, add_target_sections (sections=0x21372e0, sections_end=0x2137580) at exec.c:471
471               push_target (&exec_ops);
(gdb) bt
#0  add_target_sections (sections=0x21372e0, sections_end=0x2137580) at exec.c:471
#1  in solib_map_sections (so=0x21cd6f0) at solib.c:507
#2  in update_solib_list (from_tty=1, target=0x1d8fee0) at solib.c:804
#3  in solib_add (pattern=0x0, from_tty=1, target=0x1d8fee0, readsyms=1) at solib.c:889
#4  in enable_break (info=0x210f000, from_tty=1) at solib-svr4.c:1624
#5  in svr4_solib_create_inferior_hook (from_tty=1) at solib-svr4.c:2234
#6  in solib_create_inferior_hook (from_tty=1) at solib.c:1172
#7  in reload_shared_libraries (ignored=0x0, from_tty=1, e=0x1fc5cd0) at solib.c:1308

So it is questionable whether to keep exec_ops unpushed or whether to handle
NULL exec_bfd there.  I chose the latter.

The set_section_command part has no code change effect, just there was a bug.

No regressions on {x86_64,x86_64-m32,i686}-fedorarawhide-linux-gnu and with
non-extended gdbserver.

I will check it in.


Thanks,
Jan


gdb/
2012-02-08  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Fix crash on loaded shlibs without loaded exec_bfd.
	* exec.c (exec_files_info): Do not crash on NULL EXEC_BFD.
	(set_section_command): Replace exec_bfd by p->bfd.

gdb/testsuite/
2012-02-08  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* gdb.server/server-exec-info.exp: New file.

--- a/gdb/exec.c
+++ b/gdb/exec.c
@@ -761,7 +761,10 @@ print_section_info (struct target_section_table *t, bfd *abfd)
 static void
 exec_files_info (struct target_ops *t)
 {
-  print_section_info (current_target_sections, exec_bfd);
+  if (exec_bfd)
+    print_section_info (current_target_sections, exec_bfd);
+  else
+    puts_filtered (_("\t<no file loaded>\n"));
 
   if (vmap)
     {
@@ -813,9 +816,9 @@ set_section_command (char *args, int from_tty)
   table = current_target_sections;
   for (p = table->sections; p < table->sections_end; p++)
     {
-      if (!strncmp (secname, bfd_section_name (exec_bfd,
+      if (!strncmp (secname, bfd_section_name (p->bfd,
 					       p->the_bfd_section), seclen)
-	  && bfd_section_name (exec_bfd, p->the_bfd_section)[seclen] == '\0')
+	  && bfd_section_name (p->bfd, p->the_bfd_section)[seclen] == '\0')
 	{
 	  offset = secaddr - p->addr;
 	  p->addr += offset;
--- /dev/null
+++ b/gdb/testsuite/gdb.server/server-exec-info.exp
@@ -0,0 +1,29 @@
+# Copyright (C) 2012 Free Software Foundation, Inc.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation; either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+load_lib gdbserver-support.exp
+
+set testfile "no-execfile-info"
+set srcfile "server.c"
+#if { [build_executable ${testfile}.exp ${testfile} ${srcfile}] == -1} {
+if [prepare_for_testing ${testfile}.exp ${testfile} ${srcfile}] {
+    return -1
+}
+
+gdb_test "file" ".*" "file" \
+	 {Discard symbol table from `.*'\? \(y or n\) } "y"
+gdbserver_run ""
+gdb_test "set sysroot remote:"
+gdb_test "info files" "\r\nLocal exec file:\r\n\t<no file loaded>"

Re: [patch] Fix crash on `info files' with gdbserver

[Pedro Alves]

On 02/08/2012 08:49 PM, Jan Kratochvil wrote:

> while playing with gdbserver GDB just crashed for me.
> 
> The problem is exec_ops get pushed to the target stack for the loaded shared
> libraries despite exec_bfd remains NULL which is not expected by exec.c.
> 
> Breakpoint 1, add_target_sections (sections=0x21372e0, sections_end=0x2137580) at exec.c:471
> 471               push_target (&exec_ops);
> (gdb) bt
> #0  add_target_sections (sections=0x21372e0, sections_end=0x2137580) at exec.c:471
> #1  in solib_map_sections (so=0x21cd6f0) at solib.c:507
> #2  in update_solib_list (from_tty=1, target=0x1d8fee0) at solib.c:804
> #3  in solib_add (pattern=0x0, from_tty=1, target=0x1d8fee0, readsyms=1) at solib.c:889
> #4  in enable_break (info=0x210f000, from_tty=1) at solib-svr4.c:1624
> #5  in svr4_solib_create_inferior_hook (from_tty=1) at solib-svr4.c:2234
> #6  in solib_create_inferior_hook (from_tty=1) at solib.c:1172
> #7  in reload_shared_libraries (ignored=0x0, from_tty=1, e=0x1fc5cd0) at solib.c:1308
> 
> So it is questionable whether to keep exec_ops unpushed or whether to handle
> NULL exec_bfd there.  I chose the latter.

That's the right choice.  The exec target nowadays handles reading memory
from all target sections, no matter whether they came from the executable
or from shared libraries.  So e.g., "set trust-readonly-sections" works
with the DSOs target sections too.

-- 
Pedro Alves

[commit] [patch] Fix crash on `info files' with gdbserver

[Jan Kratochvil]

On Wed, 08 Feb 2012 22:21:12 +0100, Pedro Alves wrote:
> That's the right choice.  The exec target nowadays handles reading memory
> from all target sections, no matter whether they came from the executable
> or from shared libraries.  So e.g., "set trust-readonly-sections" works
> with the DSOs target sections too.

Checked in:
	http://sourceware.org/ml/gdb-cvs/2012-02/msg00065.html

Thanks for the review.


Jan