1 http://sourceware.org/ml/gdb-patches/2011-02/msg00630.html
2 Subject: [patch] [i386] Fix {,un}prelinked libraries for attach/core-load
6 please see comments in the patch. The adjusted testcase FAILs on i386.
8 "Prelink", March 4, 2004 - by Jakub Jelinek:
9 http://people.redhat.com/jakub/prelink.pdf
10 primarily section 7 - REL to RELA conversion
12 An example of unprelinked -> prelinked library change:
14 Type Offset VirtAddr PhysAddr FileSiz MemSiz Flg Align
15 - LOAD 0x000000 0x00000000 0x00000000 0x00538 0x00538 R E 0x1000
16 - LOAD 0x000538 0x00001538 0x00001538 0x00100 0x00110 RW 0x1000
17 - DYNAMIC 0x000550 0x00001550 0x00001550 0x000c8 0x000c8 RW 0x4
18 - NOTE 0x0000f4 0x000000f4 0x000000f4 0x00024 0x00024 R 0x4
19 - GNU_EH_FRAME 0x0004e8 0x000004e8 0x000004e8 0x00014 0x00014 R 0x4
20 + LOAD 0x000000 0x411b3000 0x411b3000 0x00558 0x00558 R E 0x1000
21 + LOAD 0x000558 0x411b4558 0x411b4558 0x00100 0x00110 RW 0x1000
22 + DYNAMIC 0x000570 0x411b4570 0x411b4570 0x000c8 0x000c8 RW 0x4
23 + NOTE 0x0000f4 0x411b30f4 0x411b30f4 0x00024 0x00024 R 0x4
24 + GNU_EH_FRAME 0x000508 0x411b3508 0x411b3508 0x00014 0x00014 R 0x4
25 GNU_STACK 0x000000 0x00000000 0x00000000 0x00000 0x00000 RW 0x4
27 So far GDB expected all such displacements will be always PAGE_SIZE aligned.
28 This applies for example for re-prelinking of an already prelinked file.
29 But it does not apply for prelinking of an unprelinked file or unprelinking of
30 a prelinked file, there can be arbitrary displacement.
32 It affects i386 (=i686, prelink doc reports also ARM and MIPS) which uses REL.
33 x86_64 always uses RELA, therefore I have not noticed it so far. i386 still
36 This affects both attachment to a PID and core file loads.
38 This applies in real world if you transfer a core file between hosts and try to
39 backtrace them, libraries of both hosts may differ whether they are / are not
42 I could implement some (displacement-forgiving and prelink-modifications
43 forgiving) comparison of both DYNAMIC segments found. But I do not think it is
44 useful, if the DYNAMIC address from linkmap vs. bfd do not match it is still a
45 better chance to try a displacement to make them match. Keeping the file
46 relocation cannot work anyway when the DYNAMIC address is verified as wrong.
48 No regressions on {x86_64,x86_64-m32,i686}-fedora15-linux-gnu.
50 Mostly do you agree the DYNAMIC content does not have to be verifed?
51 Do you have any comments on the in-code long comments?
59 2011-02-22 Jan Kratochvil <jan.kratochvil@redhat.com>
61 Fix libraries displacement if they change whether they were prelinked.
62 * solib-svr4.c (LM_ADDR_CHECK): Set L_ADDR even if the DYNAMIC pointer
63 does not match. Comment why.
66 2011-02-22 Jan Kratochvil <jan.kratochvil@redhat.com>
68 * gdb.base/break-interp-lib.c (v, vptr): New variables.
69 * gdb.base/break-interp.exp (test_attach): New comment.
71 --- a/gdb/solib-svr4.c
72 +++ b/gdb/solib-svr4.c
73 @@ -237,11 +237,11 @@ LM_ADDR_CHECK (struct so_list *so, bfd *abfd)
75 Even on PPC it must be zero-aligned at least for MINPAGESIZE. */
77 + l_addr = l_dynaddr - dynaddr;
79 if ((l_addr & (minpagesize - 1)) == 0
80 && (l_addr & align) == ((l_dynaddr - dynaddr) & align))
82 - l_addr = l_dynaddr - dynaddr;
85 printf_unfiltered (_("Using PIC (Position Independent Code) "
86 "prelink displacement %s for \"%s\".\n"),
87 @@ -249,9 +249,20 @@ LM_ADDR_CHECK (struct so_list *so, bfd *abfd)
91 - warning (_(".dynamic section for \"%s\" "
92 - "is not at the expected address "
93 - "(wrong library or version mismatch?)"), so->so_name);
95 + /* There is no way to verify the library file matches. prelink
96 + can during prelinking of an unprelinked file (or unprelinking
97 + of a prelinked file) shift the DYNAMIC segment by arbitrary
98 + offset without any page size alignment. There is no way to
99 + find out the ELF header and/or Program Headers for a limited
100 + verification if it they match. One could do a verification
101 + of the DYNAMIC segment. Still the found address is the best
102 + one GDB could find. */
104 + warning (_(".dynamic section for \"%s\" "
105 + "is not at the expected address "
106 + "(wrong library or version mismatch?)"), so->so_name);
111 --- a/gdb/testsuite/gdb.base/break-interp-lib.c
112 +++ b/gdb/testsuite/gdb.base/break-interp-lib.c
117 +/* Force REL->RELA conversion on i386, see "Prelink", March 4, 2004. */
119 +volatile int *vptr = &v[1];
122 libfunc (const char *action)
124 --- a/gdb/testsuite/gdb.base/break-interp.exp
125 +++ b/gdb/testsuite/gdb.base/break-interp.exp
126 @@ -352,6 +352,14 @@ proc test_attach {file displacement {relink_args ""}} {
127 # test simplicity, we merged this test and the test above by not
128 # restoring $INTERP after $EXEC prelink. $INTERP gets restored
131 + # `(wrong library or version mismatch?)' messages are printed for
132 + # $binfile_lib on platforms converting REL->RELA relocations by
133 + # prelink (such as on i386). There is no reliable way to verify
134 + # the library file matches the running library in such case but
135 + # GDB at least attempts to set the right displacement. We test
136 + # `libfunc' is present in the backtrace and therefore the
137 + # displacement has been guessed right.
139 if [prelink$relink $relink_args [file tail $exec]] {
140 # /proc/PID/exe cannot be loaded as it is "EXECNAME (deleted)".