Re: [PATCH 2/2] tools/nolibc: x86-64: Fix startup code bug

From: Louvian Lyndal <louvianlyndal@gmail.com>
To: Ammar Faizi <ammar.faizi@students.amikom.ac.id>
Cc: Willy Tarreau <w@1wt.eu>,
	Paul Walmsley <paul.walmsley@sifive.com>,
	Palmer Dabbelt <palmer@dabbelt.com>,
	Albert Ou <aou@eecs.berkeley.edu>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Andy Lutomirski <luto@kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
	David Laight <David.Laight@aculab.com>,
	Peter Cordes <peter@cordes.ca>,
	Bedirhan KURT <windowz414@gnuweeb.org>
Subject: Re: [PATCH 2/2] tools/nolibc: x86-64: Fix startup code bug
Date: Fri, 15 Oct 2021 16:41:20 +0700	[thread overview]
Message-ID: <CAP2ubg+p9sxJRKVhRAqYE0RKVU0Xz81YYy+=fysRiScMtAV2LQ@mail.gmail.com> (raw)
In-Reply-To: <6sZ9qpcJvtqCksJQVaiZyA-ammarfaizi2@gnuweeb.org>

On Fri, Oct 15, 2021 at 3:57 PM Ammar Faizi wrote:
>
> Hi,
>
> This is a code to test.
>
> Compile with:
>   gcc -O3 -ggdb3 -nostdlib -o test test.c
>
> Technical explanation:
> The System V ABI mandates the %rsp must be 16-byte aligned before
> performing a function call, but the current nolibc.h violates it.
>
> This %rsp alignment violation makes the callee can't align its stack
> properly. Note that the callee may have a situation where it requires
> vector aligned move. For example, `movaps` with memory operand w.r.t.
> xmm registers, it requires the src/dst address be 16-byte aligned.
>
> Since the callee can't align its stack properly, it will segfault when
> executing `movaps`. The following C code is the reproducer and test
> to ensure the bug really exists and this patch fixes it.

Hello,
With the current nolibc.h, the program segfault on movaps:
Program received signal SIGSEGV, Segmentation fault.
0x0000555555555032 in dump_argv (argv=0x7fffffffe288, argc=1) at test.c:15
15        const char str[] = "\nDumping argv...\n";
(gdb) x/20i main
   0x555555555000 <main>:    endbr64
   0x555555555004 <main+4>:    push   %r14
   0x555555555006 <main+6>:    push   %r13
   0x555555555008 <main+8>:    mov    %edi,%r13d
   0x55555555500b <main+11>:    push   %r12
   0x55555555500d <main+13>:    push   %rbp
   0x55555555500e <main+14>:    mov    %rdx,%rbp
   0x555555555011 <main+17>:    mov    $0xa,%edx
   0x555555555016 <main+22>:    push   %rbx
   0x555555555017 <main+23>:    mov    %rsi,%rbx
   0x55555555501a <main+26>:    sub    $0x8,%rsp
   0x55555555501e <main+30>:    movdqa 0xffa(%rip),%xmm0        # 0x555555556020
   0x555555555026 <main+38>:    mov    %dx,-0x68(%rsp)
   0x55555555502b <main+43>:    lea    -0x78(%rsp),%r12
   0x555555555030 <main+48>:    xor    %edx,%edx
=> 0x555555555032 <main+50>:    movaps %xmm0,-0x78(%rsp)
   0x555555555037 <main+55>:    nopw   0x0(%rax,%rax,1)
   0x555555555040 <main+64>:    add    $0x1,%rdx
   0x555555555044 <main+68>:    cmpb   $0x0,(%r12,%rdx,1)
   0x555555555049 <main+73>:    jne    0x555555555040 <main+64>
(gdb) p $rsp-0x78
$1 = (void *) 0x7fffffffe1c8
(gdb)

Apparently it's because $rsp-0x78 is not multiple of 16. After this
patchset, it works fine. gcc version 11.1.0

Tested-by: Louvian Lyndal <louvianlyndal@gmail.com>