2005-09-29 15:58:13

by John Reiser

[permalink] [raw]
Subject: ptrace unexpected SIGTRAP (trace bit) on x86, x86_64 kernel 2.6.13.2

Ptrace is giving unexpected SIGTRAP (trace bit) in kernel 2.6.13.2
on both x86 and x86_64.

The 8-instruction program below just execve()s itself over and over.
When run under gdb, the first user-visible SIGTRAP is expected due to
the 'int3'. But the second user-visible SIGTRAP is unexpected, as
there is no reason to trap.

Changing the line "nop; int3" to "nop; nop" gives a program that
just spins merrily when run under /bin/bash. But gdb sees a SIGTRAP,
with the $pc pointing after the second 'nop'. When run under strace
(strace gdb ./execve; (gdb) run), the process spins merrily with
no unexpected SIGTRAP.


-----execve.S
#include <asm/unistd.h>

/*
gcc -o execve -nostartfiles -nostdlib execve.S
gdb ./execve
run
p/x $ps
# 0x202
c
p/x $ps
# 0x302 TF (0x100) set, but should not be
*/

_start: .globl _start
nop; int3
popl %ebp # argc
movl (%esp),%ebx # same filename from argv[0]
movl %esp,%ecx # same argv
lea 4(%esp,%ebp,4),%edx # same envp
movl $__NR_execve,%eax # here we go 'round the mulberry bush, ...
int $0x80
-----end of execve.S

Previous history, and translation for x86_64 are at:
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=144805#c23

--
John Reiser, [email protected]


2005-09-30 00:47:57

by Daniel Jacobowitz

[permalink] [raw]
Subject: Re: ptrace unexpected SIGTRAP (trace bit) on x86, x86_64 kernel 2.6.13.2

On Thu, Sep 29, 2005 at 08:58:25AM -0700, John Reiser wrote:
> Ptrace is giving unexpected SIGTRAP (trace bit) in kernel 2.6.13.2
> on both x86 and x86_64.
>
> The 8-instruction program below just execve()s itself over and over.
> When run under gdb, the first user-visible SIGTRAP is expected due to
> the 'int3'. But the second user-visible SIGTRAP is unexpected, as
> there is no reason to trap.
>
> Changing the line "nop; int3" to "nop; nop" gives a program that
> just spins merrily when run under /bin/bash. But gdb sees a SIGTRAP,
> with the $pc pointing after the second 'nop'. When run under strace
> (strace gdb ./execve; (gdb) run), the process spins merrily with
> no unexpected SIGTRAP.

Most likely GDB is receiving the SIGTRAP generated by execve(). I
don't know why it would appear to be after the int3 rather than before.
GDB has strictly limited support for programs which exec, mostly due to
thorny user interface issues.

--
Daniel Jacobowitz
CodeSourcery, LLC