https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77330
Bug ID: 77330
Summary: __float128 segfaults on 32-bit x86 due to 8-byte
malloc alignment
Product: gcc
Version: 6.2.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: mikulas at artax dot karlin.mff.cuni.cz
Target Milestone: ---
Host: x86_64-pc-linux-gnu
Target: x86_64-pc-linux-gnu
Build: x86_64-pc-linux-gnu
On 32-bit x86 system, glibc malloc aligns allocated memory to 8 bytes. The type
__float128 has 16-byte alignment (__alignof__(__float128) return 16), however
this alignment can't be guaranteed because of 8-byte malloc alignment.
GCC uses aligned SSE instructions when accessing the __float128 type, these
instructions cause a crash when accessing heap memory that is aligned only to 8
bytes.
This program will crash if compiled with "gcc -m32 -march=athlon64" (with or
without optimization).
#include <stdio.h>
#include <stdlib.h>
struct s {
__float128 f1;
};
int main(void)
{
struct s *p = malloc(sizeof(struct s));
printf("%p\n", p);
p->f1 = 1.234;
return 0;
}
The misgenerated code is this:
0804842b <main>:
804842b: 55 push %ebp
804842c: 89 e5 mov %esp,%ebp
804842e: 83 e4 f0 and $0xfffffff0,%esp
8048431: 83 ec 20 sub $0x20,%esp
8048434: c7 04 24 10 00 00 00 movl $0x10,(%esp)
804843b: e8 a0 fe ff ff call 80482e0 <malloc@plt>
8048440: 89 44 24 1c mov %eax,0x1c(%esp)
8048444: 8b 44 24 1c mov 0x1c(%esp),%eax
8048448: 89 44 24 04 mov %eax,0x4(%esp)
804844c: c7 04 24 10 85 04 08 movl $0x8048510,(%esp)
8048453: e8 78 fe ff ff call 80482d0 <printf@plt>
8048458: 8b 44 24 1c mov 0x1c(%esp),%eax
804845c: 66 0f 6f 05 20 85 04 movdqa 0x8048520,%xmm0
8048463: 08
8048464: 0f 29 00 movaps %xmm0,(%eax)
8048467: b8 00 00 00 00 mov $0x0,%eax
804846c: c9 leave
804846d: c3 ret
