ping: [patch 6/6] PIE: Fix back re-run

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

Hi,

originally posted as:
	[patch 6/6] PIE: Fix back re-run
	http://sourceware.org/ml/gdb-patches/2010-03/msg01005.html

Rediffed only.

------------------------------------------------------------------------------

originally posted as:
	[patch] PIE: Fix back re-run [Re: [patch] svr4_exec_displacement success indicator]
	http://sourceware.org/ml/gdb-patches/2010-03/msg00868.html

Rediffed only (+some word change in the accompanying mail text below).

FYI on i686 Linux kernel (+x86_64 kernel for i686 programs)
"set disable-randomization on" acts the same as - apparently being applied -
- "set disable-randomization off" (the state before introducing that
"disable-randomization" GDB patch).  This is a Linux kernel bug; tracked at:
	https://bugzilla.redhat.com/show_bug.cgi?id=220892

------------------------------------------------------------------------------

currently:

$ echo 'main(){}'|gcc -o 1 -fPIE -pie -x c -; ./gdb -nx -ex 'set disable-randomization off' -ex 'b main' -ex r -ex c -ex r ./1
Breakpoint 1 at 0x6b0
Starting program: /home/jkratoch/redhat/gdb-clean/gdb/testsuite/1 
Breakpoint 1, 0x00007fbf73e8c6b0 in main ()
Continuing.
Program exited with code 0140.
Starting program: /home/jkratoch/redhat/gdb-clean/gdb/testsuite/1 
Error in re-setting breakpoint 1: Cannot access memory at address 0x7fbf73e8c6ac

It is since:
	[patch] svr4_exec_displacement success indicator [Re: PIE question]
	http://sourceware.org/ml/gdb-patches/2010-03/msg00336.html


On Mon, 08 Mar 2010 22:53:58 +0100, Jan Kratochvil wrote:
> Attached these changes:
> 
> * svr4_exec_displacement calling convention should have success indicator.
> 
> * Preserving now section_offsets if they are already set, inspired by
>   init_objfile_sect_indices.
> 
> I believe either of parts would be sufficient for this problem.

The part "Preserving now section_offsets if they are already set" has caused
the regression for PIE on native x86* GNU/Linux host.

As I believe for Daniel J.'s seen regression of `qOffsets' the checked-in part
"svr4_exec_displacement calling convention should have success indicator" is
sufficient - I would like to remove the "Preserving now section_offsets if they
are already set" part.


No regressions on {x86_64,x86_64-m32,i686}-fedora12-linux-gnu for the whole
patch series together.


Thanks,
Jan


gdb/
2010-03-29  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Fix re-run of PIE executable.
	* solib-svr4.c (svr4_relocate_main_executable) <symfile_objfile>: Remove
	the part of pre-set SYMFILE_OBJFILE->SECTION_OFFSETS.

gdb/testsuite/
2010-03-29  Jan Kratochvil  <jan.kratochvil@redhat.com>

	Fix re-run of PIE executable.
	* gdb.base/break-interp.exp (test_ld): Turn off "disable-randomization".
	Remove $displacement_main to match the solib-svr4.c change.  New "kill"
	and re-"run" of the inferior.

--- a/gdb/solib-svr4.c
+++ b/gdb/solib-svr4.c
@@ -1985,17 +1985,10 @@ svr4_relocate_main_executable (void)
 {
   CORE_ADDR displacement;
 
-  if (symfile_objfile)
-    {
-      int i;
-
-      /* Remote target may have already set specific offsets by `qOffsets'
-	 which should be preferred.  */
-
-      for (i = 0; i < symfile_objfile->num_sections; i++)
-	if (ANOFFSET (symfile_objfile->section_offsets, i) != 0)
-	  return;
-    }
+  /* SYMFILE_OBJFILE->SECTION_OFFSETS may now contain displacement from the
+     previous run of the inferior.  Re-set it according to the current value,
+     if we can find it out.  But otherwise keep it as for remote target it may
+     have been pre-set by the `qOffsets' packet.  */
 
   if (! svr4_exec_displacement (&displacement))
     return;
--- a/gdb/testsuite/gdb.base/break-interp.exp
+++ b/gdb/testsuite/gdb.base/break-interp.exp
@@ -330,6 +330,9 @@ proc test_ld {file ifmain trynosym displacement} {
     # Print the "PIE (Position Independent Executable) displacement" message.
     gdb_test_no_output "set verbose on"
 
+    # A bit better test coverage.
+    gdb_test "set disable-randomization off"
+
     reach "dl_main" "run segv" $displacement
 
     gdb_test "bt" "#0 +\[^\r\n\]*\\mdl_main\\M.*" "dl bt"
@@ -340,7 +343,13 @@ proc test_ld {file ifmain trynosym displacement} {
 	reach "libfunc" continue "NONE"
 
 	gdb_test "bt" "#0 +\[^\r\n\]*\\mlibfunc\\M\[^\r\n\]*\r\n#1 +\[^\r\n\]*\\mmain\\M.*" "main bt"
+    }
 
+    # Try re-run if the new PIE displacement takes effect.
+    gdb_test "kill" "" "kill" {Kill the program being debugged\? \(y or n\) } "y"
+    reach "dl_main" "run  segv" $displacement
+
+    if $ifmain {
 	test_core $file $displacement
 
 	test_attach $file $displacement