* [PATCH] Make ld.so.cache identical between IA-32 and AMD64
@ 2003-08-25 10:49 Jakub Jelinek
2003-08-25 18:38 ` Ulrich Drepper
0 siblings, 1 reply; 2+ messages in thread
From: Jakub Jelinek @ 2003-08-25 10:49 UTC (permalink / raw)
To: Ulrich Drepper; +Cc: Glibc hackers
Hi!
If the number of old cache libraries in ld.so.cache (ie. those with
entry->hwcap == 0) is odd and ld.so.cache contains both old and new
cache, the 32-bit dynamic linker will not be able to find new cache
created by 64-bit ldconfig or 64-bit ld.so will not be able to find
new cache created by 32-bit ldconfig on AMD64.
The problem is that cache_file_new structure is 32-bit aligned on
IA-32 while 64-bit aligned on AMD64 (it contains uint64_t hwcap).
The following patch fixes it by creating a dummy old cache entry,
which means no ld.so or libc.a loader changes are needed and compatibility
is ensured.
Also, it seems save_cache could clobber memory past end of file_entries
malloced buffer if last entry has entry->hwcap != 0.
2003-08-25 Jakub Jelinek <jakub@redhat.com>
* elf/cache.c (save_cache): Don't write beyond end of file_entries
buffer.
Duplicate last old cache entry if the count would be odd.
--- libc/elf/cache.c.jj 2003-03-15 20:03:52.000000000 -0500
+++ libc/elf/cache.c 2003-08-25 06:31:06.000000000 -0400
@@ -305,6 +305,13 @@ save_cache (const char *cache_name)
if (opt_format != 2)
{
+ /* struct cache_file_new is 64-bit aligned on some arches while
+ only 32-bit aligned on other arches. Duplicate last old
+ cache entry so that new cache in ld.so.cache can be used by
+ both. */
+ if (opt_format != 0)
+ cache_entry_old_count = (cache_entry_old_count + 1) & ~1;
+
/* And the list of all entries in the old format. */
file_entries_size = sizeof (struct cache_file)
+ cache_entry_old_count * sizeof (struct file_entry);
@@ -351,7 +358,7 @@ save_cache (const char *cache_name)
entry = entry->next, ++idx_new)
{
/* First the library. */
- if (opt_format != 2)
+ if (opt_format != 2 && entry->hwcap == 0)
{
file_entries->libs[idx_old].flags = entry->flags;
/* XXX: Actually we can optimize here and remove duplicates. */
@@ -374,7 +381,7 @@ save_cache (const char *cache_name)
++str;
str_offset += len + 1;
/* Then the path. */
- if (opt_format != 2)
+ if (opt_format != 2 && entry->hwcap == 0)
file_entries->libs[idx_old].value = str_offset + pad;
if (opt_format != 0)
file_entries_new->libs[idx_new].value = str_offset;
@@ -388,6 +395,11 @@ save_cache (const char *cache_name)
++idx_old;
}
+ /* Duplicate last old cache entry if needed. */
+ if (opt_format != 2
+ && idx_old < cache_entry_old_count)
+ file_entries->libs[idx_old] = file_entries->libs[idx_old - 1];
+
/* Write out the cache. */
/* Write cache first to a temporary file and rename it later. */
Jakub
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: [PATCH] Make ld.so.cache identical between IA-32 and AMD64
2003-08-25 10:49 [PATCH] Make ld.so.cache identical between IA-32 and AMD64 Jakub Jelinek
@ 2003-08-25 18:38 ` Ulrich Drepper
0 siblings, 0 replies; 2+ messages in thread
From: Ulrich Drepper @ 2003-08-25 18:38 UTC (permalink / raw)
To: Jakub Jelinek; +Cc: Glibc hackers
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Jakub Jelinek wrote:
> 2003-08-25 Jakub Jelinek <jakub@redhat.com>
>
> * elf/cache.c (save_cache): Don't write beyond end of file_entries
> buffer.
> Duplicate last old cache entry if the count would be odd.
I applied the patch, thanks.
- --
- --------------. ,-. 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/SleQ2ijCOnn/RHQRAnAPAJ0TIxQcVk2JscMuWvxAo8z5jtPfvgCgpdBk
gWjHdXvhXUkARdpF/gMBASA=
=ZYzQ
-----END PGP SIGNATURE-----
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2003-08-25 18:38 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-08-25 10:49 [PATCH] Make ld.so.cache identical between IA-32 and AMD64 Jakub Jelinek
2003-08-25 18:38 ` Ulrich Drepper
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).