Date: Fri, 01 Aug 1997 14:17:04 +0200 From: Uros Juvan <uros.juvan@arnes.si> To: freebsd-bugs@FreeBSD.ORG Subject: possible bug in ptrace(2) - PT_DETACH Message-ID: <97Aug1.141718mest.4308-143%2B229@tosc.arnes.si>
next in thread | raw e-mail | index | archive | help
Look Forwarded Mesage below:
------- Forwarded Message
Received: from kanin.arnes.si ([193.2.1.66]) by tosc.arnes.si with SMTP id
<4308-143>; Fri, 1 Aug 1997 11:34:40 +0100
Received: from revenue.regent.e-technik.tu-muenchen.de by kanin.arnes.si
with SMTP using DNS (PP) id <19393-0@kanin.arnes.si>;
Fri, 1 Aug 1997 11:33:56 +0200
Received: from red.regent.e-technik.tu-muenchen.de
(pes@red.regent.e-technik.tu-muenchen.de [129.187.231.159])
by revenue.regent.e-technik.tu-muenchen.de (8.8.5/8.6.9) with ESMTP
id LAA08988 ; Fri, 1 Aug 1997 11:33:47 +0200
From: "Peter.Schauer" <Peter.Schauer@regent.e-technik.tu-muenchen.de>
Received: (pes@localhost) by red.regent.e-technik.tu-muenchen.de (8.8.5/8.6.9)
id LAA16690 ; Fri, 1 Aug 1997 11:33:46 +0200
Message-Id: <199708010933.LAA16690@red.regent.e-technik.tu-muenchen.de>
Subject: Re: gdb and signals
To: uros.juvan@arnes.si (Uros Juvan)
Date: Fri, 1 Aug 97 11:33:45 MET DST
Cc: shebs@cygnus.com
In-Reply-To: <97Aug1.095519mest.4308-143+225@tosc.arnes.si>; from "Uros Juvan"
at Aug 1, 97 9:55 am
X-Mailer: ELM [version 2.3 PL6]
This looks like a bug in the freebsd kernel, GDB behaves as expected
when I tried the example under SunOs 4.1.
The following GDB code is executed when the detach command is issued:
int siggnal = 0;
if (from_tty)
{
char *exec_file = get_exec_file (0);
if (exec_file == 0)
exec_file = "";
printf_unfiltered ("Detaching from program: %s %s\n", exec_file,
target_pid_to_str (inferior_pid));
gdb_flush (gdb_stdout);
}
if (args)
siggnal = atoi (args);
detach (siggnal);
and detach is:
errno = 0;
ptrace (PT_DETACH, inferior_pid, (PTRACE_ARG3_TYPE) 1, signal);
if (errno)
perror_with_name ("ptrace");
Traditionally passing signal 0 to PT_DETACH means to cancel the current
signal, but it seems that freebsd doesn't honor the request.
> Hello.
>
> Is this kind of behavio a bug or a feature:
>
> test.c program:
> ----------------
> #include <stdlib.h>
> #include <stdio.h>
>
> main()
> {
> while (1) {
> sleep(1);
> }
> }
> -----------------
> > Then I run this program in background:
> # ./test &
> [1] 8407
> > Then I run gdb
> # gdb
> GDB is free software and you are welcome to distribute copies of it
> under certain conditions; type "show copying" to see the conditions.
> There is absolutely no warranty for GDB; type "show warranty" for details.
> GDB 4.16 (i386-unknown-freebsd), Copyright 1996 Free Software Foundation, Inc.
> (gdb) attach 8407
> Attaching to process 8407
> 0x804d853 in ?? ()
> (gdb) info handle
> Signal Stop Print Pass to program Description
>
> SIGHUP Yes Yes Yes Hangup
> SIGINT Yes Yes No Interrupt
> .
> .
> .
> (gdb) c
> Continuing.
> ^C
> Program received signal SIGINT, Interrupt.
> 0x804d853 in ?? ()
> (gdb) detach
> Detaching from program: process 8407
> (gdb) quit
> [1]+ Interrupt ./test
> # jobs
> #
>
> -------------------------------------------------------------------------------
> As far as I read in info pages this should not happen. gdb should not deliver
> SIGINT to process after it has been detached.
>
> With regards,
> Uros Juvan
> ARNES
- --
Peter Schauer pes@regent.e-technik.tu-muenchen.de
------- End of Forwarded Message
I run FreeBSD V2.2.1 and my sys_process.c (PT_DETACH) part looks like this:
----------
case PT_DETACH:
if ((unsigned)uap->data >= NSIG)
return EINVAL;
PHOLD(p);
if (uap->req == PT_STEP) {
if ((error = ptrace_single_step (p))) {
PRELE(p);
return error;
}
}
if (uap->addr != (caddr_t)1) {
fill_eproc (p, &p->p_addr->u_kproc.kp_eproc);
if ((error = ptrace_set_pc (p, (u_int)uap->addr))) {
PRELE(p);
return error;
}
}
PRELE(p);
if (uap->req == PT_DETACH) {
/* reset process parent */
if (p->p_oppid != p->p_pptr->p_pid) {
struct proc *pp;
pp = pfind(p->p_oppid);
proc_reparent(p, pp ? pp : initproc);
}
p->p_flag &= ~(P_TRACED | P_WAITED);
p->p_oppid = 0;
/* should we send SIGCHLD? */ <- that's what I'm worried about...
}
----------
With regards,
Uros Juvan
ARNES
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?97Aug1.141718mest.4308-143%2B229>