* [RFC][patch] PR_DETACH: detach from parent
@ 2010-12-24 18:15 Stas Sergeev
0 siblings, 0 replies; 2+ messages in thread
From: Stas Sergeev @ 2010-12-24 18:15 UTC (permalink / raw)
To: linux-kernel
[-- Attachment #1: Type: text/plain, Size: 1693 bytes --]
Hi.
I was trying to use daemon() library call to daemonize the process, and
found out that it is not suitable
for the process that have threads: all the threads are simply being killed.
The reason is that the daemon() function does fork(), then exit() in
parent and setsid() in child, which is
a bit ugly, and doesn't work with threads (fork() doesn't fork threads).
I've found no solution to that, and just hacked up one instead.
The attached patch adds the PR_DETACH prctl command, which detaches the
process from its parent.
With that, the daemon() function can be implemented without the
fork/exit/setsid hacks, and here is an
example of it:
---
#include <stdio.h>
#include <unistd.h>
#include <errno.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <sys/stat.h>
#include <sys/prctl.h>
#ifndef PR_DETACH
#define PR_DETACH 35
#endif
int daemon2(int nochdir, int noclose)
{
int err, fd;
err = prctl(PR_DETACH);
if (err == -1) {
if (errno == EINVAL)
errno = ENOTSUP;
return -1;
}
fd = open("/dev/tty", O_RDWR);
if (fd != -1) {
ioctl(fd, TIOCNOTTY, NULL);
close(fd);
}
if (!nochdir && chdir("/") == -1)
return -1;
if (!noclose) {
fd = open("/dev/null", O_RDWR, 0);
if (fd == -1)
return -1;
if (dup2(fd, STDIN_FILENO) == -1)
return -1;
if (dup2(fd, STDOUT_FILENO) == -1)
return -1;
if (dup2(fd, STDERR_FILENO) == -1)
return -1;
if (fd > 2)
close(fd);
}
return 0;
}
---
With this implementation, all threads survive the daemonization.
Thoughts?
[-- Attachment #2: pr_detach.diff --]
[-- Type: text/plain, Size: 1587 bytes --]
diff --git a/include/linux/prctl.h b/include/linux/prctl.h
index a3baeb2..fbd2451 100644
--- a/include/linux/prctl.h
+++ b/include/linux/prctl.h
@@ -102,4 +102,6 @@
#define PR_MCE_KILL_GET 34
+#define PR_DETACH 35
+
#endif /* _LINUX_PRCTL_H */
diff --git a/kernel/signal.c b/kernel/signal.c
index 4e3cff1..2cd495a 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1450,10 +1450,10 @@ int do_notify_parent(struct task_struct *tsk, int sig)
BUG_ON(sig == -1);
- /* do_notify_parent_cldstop should have been called instead. */
- BUG_ON(task_is_stopped_or_traced(tsk));
+ /* do_notify_parent_cldstop should have been called instead. */
+ BUG_ON(task_is_stopped_or_traced(tsk));
- BUG_ON(!task_ptrace(tsk) &&
+ BUG_ON(!task_ptrace(tsk) && (tsk->flags & PF_EXITING) &&
(tsk->group_leader != tsk || !thread_group_empty(tsk)));
info.si_signo = sig;
diff --git a/kernel/sys.c b/kernel/sys.c
index 7f5a0cd..fa10732 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -1727,6 +1727,21 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
else
error = PR_MCE_KILL_DEFAULT;
break;
+ case PR_DETACH:
+ error = -EPERM;
+ if (me->real_parent == init_pid_ns.child_reaper)
+ break;
+ if (me->group_leader != me)
+ break;
+ exit_ptrace(me);
+ do_notify_parent(me, me->exit_signal);
+ me->exit_signal = SIGCHLD;
+ me->parent = me->real_parent =
+ init_pid_ns.child_reaper;
+ list_move_tail(&me->sibling,
+ &me->real_parent->children);
+ error = 0;
+ break;
default:
error = -EINVAL;
break;
^ permalink raw reply related [flat|nested] 2+ messages in thread
* [RFC][patch] PR_DETACH: detach from parent
@ 2010-12-25 17:45 Stas Sergeev
0 siblings, 0 replies; 2+ messages in thread
From: Stas Sergeev @ 2010-12-25 17:45 UTC (permalink / raw)
To: linux-kernel
[-- Attachment #1: Type: text/plain, Size: 1722 bytes --]
Hi.
I was trying to use daemon() library call to daemonize the process, and
found out that it is not suitable
for the process that have threads: all the threads are simply being killed.
The reason is that the daemon() function does fork(), then exit() in
parent and setsid() in child, which is
a bit ugly, and doesn't work with threads (fork() doesn't fork threads).
I've found no solution to that, and just hacked up one instead.
The attached patch adds the PR_DETACH prctl command, which detaches the
process from its parent.
With that, the daemon() function can be implemented without the
fork/exit/setsid hacks, and here is an
example of it:
---
#include <stdio.h>
#include <unistd.h>
#include <errno.h>
#include <fcntl.h>
#include <sys/ioctl.h>
#include <sys/stat.h>
#include <sys/prctl.h>
#ifndef PR_DETACH
#define PR_DETACH 35
#endif
int daemon2(int nochdir, int noclose)
{
int err, fd;
err = prctl(PR_DETACH);
if (err == -1) {
if (errno == EINVAL)
errno = ENOTSUP;
return -1;
}
fd = open("/dev/tty", O_RDWR);
if (fd != -1) {
ioctl(fd, TIOCNOTTY, NULL);
close(fd);
}
if (!nochdir && chdir("/") == -1)
return -1;
if (!noclose) {
fd = open("/dev/null", O_RDWR, 0);
if (fd == -1)
return -1;
if (dup2(fd, STDIN_FILENO) == -1)
return -1;
if (dup2(fd, STDOUT_FILENO) == -1)
return -1;
if (dup2(fd, STDERR_FILENO) == -1)
return -1;
if (fd > 2)
close(fd);
}
return 0;
}
---
With this implementation, all threads survive the daemonization.
Thoughts?
[-- Attachment #2: pr_detach.diff --]
[-- Type: text/plain, Size: 1587 bytes --]
diff --git a/include/linux/prctl.h b/include/linux/prctl.h
index a3baeb2..fbd2451 100644
--- a/include/linux/prctl.h
+++ b/include/linux/prctl.h
@@ -102,4 +102,6 @@
#define PR_MCE_KILL_GET 34
+#define PR_DETACH 35
+
#endif /* _LINUX_PRCTL_H */
diff --git a/kernel/signal.c b/kernel/signal.c
index 4e3cff1..2cd495a 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -1450,10 +1450,10 @@ int do_notify_parent(struct task_struct *tsk, int sig)
BUG_ON(sig == -1);
- /* do_notify_parent_cldstop should have been called instead. */
- BUG_ON(task_is_stopped_or_traced(tsk));
+ /* do_notify_parent_cldstop should have been called instead. */
+ BUG_ON(task_is_stopped_or_traced(tsk));
- BUG_ON(!task_ptrace(tsk) &&
+ BUG_ON(!task_ptrace(tsk) && (tsk->flags & PF_EXITING) &&
(tsk->group_leader != tsk || !thread_group_empty(tsk)));
info.si_signo = sig;
diff --git a/kernel/sys.c b/kernel/sys.c
index 7f5a0cd..fa10732 100644
--- a/kernel/sys.c
+++ b/kernel/sys.c
@@ -1727,6 +1727,21 @@ SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
else
error = PR_MCE_KILL_DEFAULT;
break;
+ case PR_DETACH:
+ error = -EPERM;
+ if (me->real_parent == init_pid_ns.child_reaper)
+ break;
+ if (me->group_leader != me)
+ break;
+ exit_ptrace(me);
+ do_notify_parent(me, me->exit_signal);
+ me->exit_signal = SIGCHLD;
+ me->parent = me->real_parent =
+ init_pid_ns.child_reaper;
+ list_move_tail(&me->sibling,
+ &me->real_parent->children);
+ error = 0;
+ break;
default:
error = -EINVAL;
break;
^ permalink raw reply related [flat|nested] 2+ messages in thread
end of thread, other threads:[~2011-01-22 8:34 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-12-25 17:45 [RFC][patch] PR_DETACH: detach from parent Stas Sergeev
-- strict thread matches above, loose matches on Subject: below --
2010-12-24 18:15 Stas Sergeev
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox