From: Jakub Jelinek <jakub@redhat.com>
To: Ulrich Drepper <drepper@redhat.com>,
Roland McGrath <roland@redhat.com>, Wolfram Gloger <wg@malloc.de>
Cc: Glibc hackers <libc-hacker@sources.redhat.com>
Subject: [PATCH] Make sure MALLOC_ALIGNMENT is at least long double's alignment
Date: Tue, 28 Feb 2006 15:21:00 -0000 [thread overview]
Message-ID: <20060228152100.GH30252@sunsite.mff.cuni.cz> (raw)
Hi!
The recent switch to 128-bit long double on ppc32 broke malloc,
as on this 32-bit arch long double is 16 byte aligned, but malloc
was only guaranteeing 8 byte aligment so far.
On sparc32 and s390 this is not a problem, since __alignof__ (long double)
is 8.
But, apparently some of the quick checks for invalid pointers don't
work when MALLOC_ALIGNMENT is bigger than 2 * SIZE_SZ, in that case
the pointer returned to the user must be MALLOC_ALIGNMENT aligned,
not the chunk pointer (which is always user pointer - 2 * SIZE_SZ).
With this, ppc32 glibc passes make check.
I haven't checked though if this pessimizes code on i?86/x86_64 or other
important arches, will do soon. If it does, then we might consider doing:
#define check_align_OK(chunkptr, userptr) \
(MALLOC_ALIGNMENT > 2 * SIZE_SZ \
? __builtin_expect ((uintptr_t) userptr & MALLOC_ALIGN_MASK, 0) \
: __builtin_expect ((uintptr_t) chunkptr & MALLOC_ALIGN_MASK, 0))
and using check_align_OK (oldp, oldmem) etc.
2006-02-28 Jakub Jelinek <jakub@redhat.com>
* malloc/malloc.c (MALLOC_ALIGNMENT): Set to __alignof__ (long double)
if long double is more aligned than 2 * SIZE_SZ.
(public_rEALLOc, _int_free, _int_realloc): Check that *mem is aligned,
rather than corresponding mem2chunk pointer.
--- libc/malloc/malloc.c 2005-12-30 09:04:02.000000000 +0100
+++ libc/malloc/malloc.c 2006-02-28 15:30:20.000000000 +0100
@@ -188,7 +188,8 @@
Changing default word sizes:
INTERNAL_SIZE_T size_t
- MALLOC_ALIGNMENT 2 * sizeof(INTERNAL_SIZE_T)
+ MALLOC_ALIGNMENT MAX (2 * sizeof(INTERNAL_SIZE_T),
+ __alignof__ (long double))
Configuration and functionality options:
@@ -380,7 +381,8 @@ extern "C" {
#ifndef MALLOC_ALIGNMENT
-#define MALLOC_ALIGNMENT (2 * SIZE_SZ)
+#define MALLOC_ALIGNMENT (2 * SIZE_SZ < __alignof__ (long double) \
+ ? __alignof__ (long double) : 2 * SIZE_SZ)
#endif
/* The corresponding bit mask value */
@@ -3468,7 +3470,7 @@ public_rEALLOc(Void_t* oldmem, size_t by
Therefore we can exclude some size values which might appear
here by accident or by "design" from some intruder. */
if (__builtin_expect ((uintptr_t) oldp > (uintptr_t) -oldsize, 0)
- || __builtin_expect ((uintptr_t) oldp & MALLOC_ALIGN_MASK, 0))
+ || __builtin_expect ((uintptr_t) oldmem & MALLOC_ALIGN_MASK, 0))
{
malloc_printerr (check_action, "realloc(): invalid pointer", oldmem);
return NULL;
@@ -4282,7 +4284,7 @@ _int_free(mstate av, Void_t* mem)
Therefore we can exclude some size values which might appear
here by accident or by "design" from some intruder. */
if (__builtin_expect ((uintptr_t) p > (uintptr_t) -size, 0)
- || __builtin_expect ((uintptr_t) p & MALLOC_ALIGN_MASK, 0))
+ || __builtin_expect ((uintptr_t) mem & MALLOC_ALIGN_MASK, 0))
{
errstr = "free(): invalid pointer";
errout:
@@ -4628,7 +4630,7 @@ _int_realloc(mstate av, Void_t* oldmem,
oldsize = chunksize(oldp);
/* Simple tests for old block integrity. */
- if (__builtin_expect ((uintptr_t) oldp & MALLOC_ALIGN_MASK, 0))
+ if (__builtin_expect ((uintptr_t) oldmem & MALLOC_ALIGN_MASK, 0))
{
errstr = "realloc(): invalid pointer";
errout:
Jakub
next reply other threads:[~2006-02-28 15:21 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-02-28 15:21 Jakub Jelinek [this message]
2006-03-02 15:04 ` Jakub Jelinek
2006-03-02 15:53 ` Ulrich Drepper
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20060228152100.GH30252@sunsite.mff.cuni.cz \
--to=jakub@redhat.com \
--cc=drepper@redhat.com \
--cc=libc-hacker@sources.redhat.com \
--cc=roland@redhat.com \
--cc=wg@malloc.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).