* [PATCH] Fix ppc32 ld.so
@ 2003-05-15 15:18 Jakub Jelinek
2003-05-16 4:43 ` Ulrich Drepper
2003-05-16 13:14 ` Franz Sirl
0 siblings, 2 replies; 3+ messages in thread
From: Jakub Jelinek @ 2003-05-15 15:18 UTC (permalink / raw)
To: Ulrich Drepper, Roland McGrath, Jack Howarth, Franz Sirl; +Cc: Glibc hackers
Hi!
This is something I cannot understand how it ever worked.
If there is a reloc overflow on ppc32, _dl_reloc_overflow tries to print the
name of the symbol. But unless sym is NULL (and thus errsym = refsym) resp.
equal to refsym, it will print something from the refsym's strtab at index
sym->st_name, ie. a random string at best, worst case segfault.
To print sym's name instead of refsym's (though I don't understand why that
is needed, both sym and refsym ought to always have the same symbol name,
otherwise symbol lookup is broken) we'd have to pass down sym_map as well,
but sym_map is not always computed (ATM only if --with-tls).
2003-05-15 Jakub Jelinek <jakub@redhat.com>
* sysdeps/powerpc/powerpc32/dl-machine.c (_dl_reloc_overflow):
Remove sym argument, always use refsym.
(__process_machine_rela): Adjust callers.
* sysdeps/powerpc/powerpc32/dl-machine.h (_dl_reloc_overflow):
Adjust prototype.
--- libc/sysdeps/powerpc/powerpc32/dl-machine.c.jj 2003-04-24 13:56:31.000000000 -0400
+++ libc/sysdeps/powerpc/powerpc32/dl-machine.c 2003-05-15 10:57:20.000000000 -0400
@@ -372,22 +372,20 @@ void
_dl_reloc_overflow (struct link_map *map,
const char *name,
Elf32_Addr *const reloc_addr,
- const Elf32_Sym *sym,
const Elf32_Sym *refsym)
{
char buffer[128];
char *t;
- const Elf32_Sym *errsym = sym ?: refsym;
t = stpcpy (buffer, name);
t = stpcpy (t, " relocation at 0x00000000");
_itoa_word ((unsigned) reloc_addr, t, 16, 0);
- if (errsym)
+ if (refsym)
{
const char *strtab;
strtab = (const void *) D_PTR (map, l_info[DT_STRTAB]);
t = stpcpy (t, " for symbol `");
- t = stpcpy (t, strtab + errsym->st_name);
+ t = stpcpy (t, strtab + refsym->st_name);
t = stpcpy (t, "'");
}
t = stpcpy (t, " out of range");
@@ -424,19 +422,19 @@ __process_machine_rela (struct link_map
case R_PPC_ADDR24:
if (__builtin_expect (finaladdr > 0x01fffffc && finaladdr < 0xfe000000, 0))
- _dl_reloc_overflow (map, "R_PPC_ADDR24", reloc_addr, sym, refsym);
+ _dl_reloc_overflow (map, "R_PPC_ADDR24", reloc_addr, refsym);
*reloc_addr = (*reloc_addr & 0xfc000003) | (finaladdr & 0x3fffffc);
break;
case R_PPC_ADDR16:
if (__builtin_expect (finaladdr > 0x7fff && finaladdr < 0xffff8000, 0))
- _dl_reloc_overflow (map, "R_PPC_ADDR16", reloc_addr, sym, refsym);
+ _dl_reloc_overflow (map, "R_PPC_ADDR16", reloc_addr, refsym);
*(Elf32_Half*) reloc_addr = finaladdr;
break;
case R_PPC_UADDR16:
if (__builtin_expect (finaladdr > 0x7fff && finaladdr < 0xffff8000, 0))
- _dl_reloc_overflow (map, "R_PPC_UADDR16", reloc_addr, sym, refsym);
+ _dl_reloc_overflow (map, "R_PPC_UADDR16", reloc_addr, refsym);
((char *) reloc_addr)[0] = finaladdr >> 8;
((char *) reloc_addr)[1] = finaladdr;
break;
@@ -457,7 +455,7 @@ __process_machine_rela (struct link_map
case R_PPC_ADDR14_BRTAKEN:
case R_PPC_ADDR14_BRNTAKEN:
if (__builtin_expect (finaladdr > 0x7fff && finaladdr < 0xffff8000, 0))
- _dl_reloc_overflow (map, "R_PPC_ADDR14", reloc_addr, sym, refsym);
+ _dl_reloc_overflow (map, "R_PPC_ADDR14", reloc_addr, refsym);
*reloc_addr = (*reloc_addr & 0xffff0003) | (finaladdr & 0xfffc);
if (rinfo != R_PPC_ADDR14)
*reloc_addr = ((*reloc_addr & 0xffdfffff)
@@ -469,7 +467,7 @@ __process_machine_rela (struct link_map
{
Elf32_Sword delta = finaladdr - (Elf32_Word) reloc_addr;
if (delta << 6 >> 6 != delta)
- _dl_reloc_overflow (map, "R_PPC_REL24", reloc_addr, sym, refsym);
+ _dl_reloc_overflow (map, "R_PPC_REL24", reloc_addr, refsym);
*reloc_addr = (*reloc_addr & 0xfc000003) | (delta & 0x3fffffc);
}
break;
@@ -568,7 +566,7 @@ __process_machine_rela (struct link_map
inline void do_reloc16 (const char *r_name, Elf32_Addr value)
{
if (__builtin_expect (value > 0x7fff && value < 0xffff8000, 0))
- _dl_reloc_overflow (map, r_name, reloc_addr, sym, refsym);
+ _dl_reloc_overflow (map, r_name, reloc_addr, refsym);
*(Elf32_Half *) reloc_addr = value;
}
inline void do_reloc16_LO (const char *r_name, Elf32_Addr value)
--- libc/sysdeps/powerpc/powerpc32/dl-machine.h.jj 2003-04-24 13:56:31.000000000 -0400
+++ libc/sysdeps/powerpc/powerpc32/dl-machine.h 2003-05-15 10:57:58.000000000 -0400
@@ -347,7 +347,6 @@ extern void __process_machine_rela (stru
extern void _dl_reloc_overflow (struct link_map *map,
const char *name,
Elf32_Addr *const reloc_addr,
- const Elf32_Sym *sym,
const Elf32_Sym *refsym) attribute_hidden;
/* Perform the relocation specified by RELOC and SYM (which is fully resolved).
Jakub
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix ppc32 ld.so
2003-05-15 15:18 [PATCH] Fix ppc32 ld.so Jakub Jelinek
@ 2003-05-16 4:43 ` Ulrich Drepper
2003-05-16 13:14 ` Franz Sirl
1 sibling, 0 replies; 3+ messages in thread
From: Ulrich Drepper @ 2003-05-16 4:43 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Roland McGrath, Jack Howarth, Franz Sirl, Glibc hackers
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Jakub Jelinek wrote:
> 2003-05-15 Jakub Jelinek <jakub@redhat.com>
>
> * sysdeps/powerpc/powerpc32/dl-machine.c (_dl_reloc_overflow):
> Remove sym argument, always use refsym.
> (__process_machine_rela): Adjust callers.
> * sysdeps/powerpc/powerpc32/dl-machine.h (_dl_reloc_overflow):
> Adjust prototype.
I've checked in the patch and also changed the ppc64 code.
- --
- --------------. ,-. 444 Castro Street
Ulrich Drepper \ ,-----------------' \ Mountain View, CA 94041 USA
Red Hat `--' drepper at redhat.com `---------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
iD8DBQE+xGds2ijCOnn/RHQRAt/zAKCX2u/f+ZE9nF8lwvaM7h0XZ7Yj2ACgj1cb
RF+pNumB0tQeWFu0vFSKX/s=
=D6va
-----END PGP SIGNATURE-----
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] Fix ppc32 ld.so
2003-05-15 15:18 [PATCH] Fix ppc32 ld.so Jakub Jelinek
2003-05-16 4:43 ` Ulrich Drepper
@ 2003-05-16 13:14 ` Franz Sirl
1 sibling, 0 replies; 3+ messages in thread
From: Franz Sirl @ 2003-05-16 13:14 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Ulrich Drepper, Roland McGrath, Jack Howarth, Glibc hackers
At 17:16 15.05.2003, Jakub Jelinek wrote:
>Hi!
>
>This is something I cannot understand how it ever worked.
>If there is a reloc overflow on ppc32, _dl_reloc_overflow tries to print the
>name of the symbol. But unless sym is NULL (and thus errsym = refsym) resp.
>equal to refsym, it will print something from the refsym's strtab at index
>sym->st_name, ie. a random string at best, worst case segfault.
>To print sym's name instead of refsym's (though I don't understand why that
>is needed, both sym and refsym ought to always have the same symbol name,
>otherwise symbol lookup is broken) we'd have to pass down sym_map as well,
>but sym_map is not always computed (ATM only if --with-tls).
Hmm, IIRC that was a mere result of try and error. I think I used the first
version that produced sensible results with the (small) binutils testcases.
Certainly you have more insight in that relocation code and I believe you
if you say that it just worked by luck :-).
Franz.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2003-05-16 13:14 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-05-15 15:18 [PATCH] Fix ppc32 ld.so Jakub Jelinek
2003-05-16 4:43 ` Ulrich Drepper
2003-05-16 13:14 ` Franz Sirl
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).