http://sourceware.org/ml/gdb-patches/2013-05/msg00042.html
Subject: [patch] Fix gcore for vDSO (on ppc64)
Hi,
on ppc64 GNU/Linux one gets in live process or kernel-generated core file:
#0 0x00003fff9e946b3c in __pause_nocancel () from /lib64/libc.so.6
#1 0x000000001000080c in handle_alrm (signo=14) at small.c:16
#2 <signal handler called>
#3 0x00003fff9e946b3c in __pause_nocancel () from /lib64/libc.so.6
#4 0x0000000010000914 in main (argc=2, argv=0x3fffcd00dec8) at small.c:32
while with gcore generated core file one gets:
#0 0x00003fff9e946b3c in __pause_nocancel () from /lib64/libc.so.6
#1 0x000000001000080c in handle_alrm (signo=14) at small.c:16
#2 0x00003fff9ea70448 in ?? ()
#3 0x0000000010000914 in main (argc=2, argv=0x3fffcd00dec8) at small.c:32
This is because on ppc64 GNU/Linux the signal handler frame is in vDSO (and
not in libc as on x86_64 GNU/Linux). And if one has kernel-debuginfo
installed GDB gcore thinks it can omit the vDSO page from core file.
As vDSO cannot be reliably found from link_map it should be rather fully
present in the core file. Which also gcore_create_callback tries to do but it
gets mistaken by the separate debug info.
It seems pretty obvious patch to me.
Thanks,
Jan
gdb/
2013-05-03 Jan Kratochvil <jan.kratochvil@redhat.com>
* gcore.c (gcore_create_callback): Ignore sections with
separate_debug_objfile_backlink != NULL.
--- gdb-7.6-orig/gdb/gcore.c 2013-01-01 07:32:42.000000000 +0100
+++ gdb-7.6/gdb/gcore.c 2013-05-03 03:31:34.795312996 +0200
@@ -428,8 +428,9 @@ gcore_create_callback (CORE_ADDR vaddr,
This BFD was synthesized from reading target memory,
we don't want to omit that. */
- if (((vaddr >= start && vaddr + size <= end)
- || (start >= vaddr && end <= vaddr + size))
+ if (objfile->separate_debug_objfile_backlink == NULL
+ && ((vaddr >= start && vaddr + size <= end)
+ || (start >= vaddr && end <= vaddr + size))
&& !(bfd_get_file_flags (abfd) & BFD_IN_MEMORY))
{
flags &= ~(SEC_LOAD | SEC_HAS_CONTENTS);