[PATCH] add child reaper to pid_namespace

Add a per pid_namespace child-reaper. This is needed so processes are reaped within the same pid space and do not spill over to the parent pid space. Its also needed so containers preserve existing semantic that pid == 1 would reap orphaned children. This is based on Eric Biederman's patch: http://lkml.org/lkml/2006/2/6/285 Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com> Signed-off-by: Cedric Le Goater <clg@fr.ibm.com> Cc: Kirill Korotaev <dev@openvz.org> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Herbert Poetzl <herbert@13thfloor.at> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] add child reaper to pid_namespace
Add a per pid_namespace child-reaper. This is needed so processes are reaped within the same pid space and do not spill over to the parent pid space. Its also needed so containers preserve existing semantic that pid == 1 would reap orphaned children. This is based on Eric Biederman's patch: http://lkml.org/lkml/2006/2/6/285 Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com> Signed-off-by: Cedric Le Goater <clg@fr.ibm.com> Cc: Kirill Korotaev <dev@openvz.org> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Herbert Poetzl <herbert@13thfloor.at> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Sukadev Bhattiprolu · Linus Torvalds
1 parent 6cc1b22a4a
Showing 8 changed files with 40 additions and 19 deletions Side-by-side Diff
fs/exec.c
include/linux/pid.h
include/linux/pid_namespace.h
include/linux/sched.h
init/main.c
kernel/exit.c
kernel/pid.c
kernel/signal.c
@@ -38,6 +38,7 @@
 #include <linux/binfmts.h>
 #include <linux/swap.h>
 #include <linux/utsname.h>
+#include <linux/pid_namespace.h>
 #include <linux/module.h>
 #include <linux/namei.h>
 #include <linux/proc_fs.h>
@@ -620,8 +621,8 @@
 	 * Reparenting needs write_lock on tasklist_lock,
 	 * so it is safe to do it under read_lock.
 	 */
-	if (unlikely(tsk->group_leader == child_reaper))
-		child_reaper = tsk;
+	if (unlikely(tsk->group_leader == child_reaper(tsk)))
+		tsk->nsproxy->pid_ns->child_reaper = tsk;
  
 	zap_other_threads(tsk);
 	read_unlock(&tasklist_lock);
@@ -35,8 +35,9 @@
  *
  * Holding a reference to struct pid solves both of these problems.
  * It is small so holding a reference does not consume a lot of
- * resources, and since a new struct pid is allocated when the numeric
- * pid value is reused we don't mistakenly refer to new processes.
+ * resources, and since a new struct pid is allocated when the numeric pid
+ * value is reused (when pids wrap around) we don't mistakenly refer to new
+ * processes.
  */
  
 struct pid
@@ -19,6 +19,7 @@
 	struct kref kref;
 	struct pidmap pidmap[PIDMAP_ENTRIES];
 	int last_pid;
+	struct task_struct *child_reaper;
 };
  
 extern struct pid_namespace init_pid_ns;
@@ -34,6 +35,11 @@
 static inline void put_pid_ns(struct pid_namespace *ns)
 {
 	kref_put(&ns->kref, free_pid_ns);
+}
+
+static inline struct task_struct *child_reaper(struct task_struct *tsk)
+{
+	return tsk->nsproxy->pid_ns->child_reaper;
 }
  
 #endif /* _LINUX_PID_NS_H */
@@ -1400,7 +1400,6 @@
 extern void daemonize(const char *, ...);
 extern int allow_signal(int);
 extern int disallow_signal(int);
-extern struct task_struct *child_reaper;
  
 extern int do_execve(char *, char __user * __user *, char __user * __user *, struct pt_regs *);
 extern long do_fork(unsigned long, unsigned long, struct pt_regs *, unsigned long, int __user *, int __user *);
@@ -51,6 +51,7 @@
 #include <linux/debug_locks.h>
 #include <linux/lockdep.h>
 #include <linux/utsrelease.h>
+#include <linux/pid_namespace.h>
 #include <linux/compile.h>
  
 #include <asm/io.h>
@@ -626,8 +627,6 @@
 }
 __setup("initcall_debug", initcall_debug_setup);
  
-struct task_struct *child_reaper = &init_task;
-
 extern initcall_t __initcall_start[], __initcall_end[];
  
 static void __init do_initcalls(void)
@@ -727,7 +726,7 @@
 	 * assumptions about where in the task array this
 	 * can be found.
 	 */
-	child_reaper = current;
+	init_pid_ns.child_reaper = current;
  
 	cad_pid = task_pid(current);
  
@@ -22,6 +22,7 @@
 #include <linux/file.h>
 #include <linux/binfmts.h>
 #include <linux/nsproxy.h>
+#include <linux/pid_namespace.h>
 #include <linux/ptrace.h>
 #include <linux/profile.h>
 #include <linux/mount.h>
@@ -48,7 +49,6 @@
 #include <asm/mmu_context.h>
  
 extern void sem_exit (void);
-extern struct task_struct *child_reaper;
  
 static void exit_mm(struct task_struct * tsk);
  
@@ -260,7 +260,8 @@
 }
  
 /**
- * reparent_to_init - Reparent the calling kernel thread to the init task.
+ * reparent_to_init - Reparent the calling kernel thread to the init task
+ * of the pid space that the thread belongs to.
  *
  * If a kernel thread is launched as a result of a system call, or if
  * it ever exits, it should generally reparent itself to init so that
@@ -278,8 +279,8 @@
 	ptrace_unlink(current);
 	/* Reparent to init */
 	remove_parent(current);
-	current->parent = child_reaper;
-	current->real_parent = child_reaper;
+	current->parent = child_reaper(current);
+	current->real_parent = child_reaper(current);
 	add_parent(current);
  
 	/* Set the exit signal to SIGCHLD so we signal init on exit */
@@ -662,7 +663,8 @@
  * When we die, we re-parent all our children.
  * Try to give them to another thread in our thread
  * group, and if no such member exists, give it to
- * the global child reaper process (ie "init")
+ * the child reaper process (ie "init") in our pid
+ * space.
  */
 static void
 forget_original_parent(struct task_struct *father, struct list_head *to_release)
@@ -673,7 +675,7 @@
 	do {
 		reaper = next_thread(reaper);
 		if (reaper == father) {
-			reaper = child_reaper;
+			reaper = child_reaper(father);
 			break;
 		}
 	} while (reaper->exit_state);
@@ -859,8 +861,13 @@
 		panic("Aiee, killing interrupt handler!");
 	if (unlikely(!tsk->pid))
 		panic("Attempted to kill the idle task!");
-	if (unlikely(tsk == child_reaper))
-		panic("Attempted to kill init!");
+	if (unlikely(tsk == child_reaper(tsk))) {
+		if (tsk->nsproxy->pid_ns != &init_pid_ns)
+			tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper;
+		else
+			panic("Attempted to kill init!");
+	}
+
  
 	if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
 		current->ptrace_message = code;
@@ -65,7 +65,8 @@
 	.pidmap = {
 		[ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
 	},
-	.last_pid = 0
+	.last_pid = 0,
+	.child_reaper = &init_task
 };
  
 /*
@@ -24,6 +24,9 @@
 #include <linux/signal.h>
 #include <linux/capability.h>
 #include <linux/freezer.h>
+#include <linux/pid_namespace.h>
+#include <linux/nsproxy.h>
+
 #include <asm/param.h>
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
@@ -1877,8 +1880,12 @@
 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
 			continue;
  
-		/* Init gets no signals it doesn't want.  */
-		if (current == child_reaper)
+		/*
+		 * Init of a pid space gets no signals it doesn't want from
+		 * within that pid space. It can of course get signals from
+		 * its parent pid space.
+		 */
+		if (current == child_reaper(current))
 			continue;
  
 		if (sig_kernel_stop(signr)) {
...	...	@@ -38,6 +38,7 @@
38	38	#include <linux/binfmts.h>
39	39	#include <linux/swap.h>
40	40	#include <linux/utsname.h>
	41	+#include <linux/pid_namespace.h>
41	42	#include <linux/module.h>
42	43	#include <linux/namei.h>
43	44	#include <linux/proc_fs.h>
...	...	@@ -620,8 +621,8 @@
620	621	* Reparenting needs write_lock on tasklist_lock,
621	622	* so it is safe to do it under read_lock.
622	623	*/
623		- if (unlikely(tsk->group_leader == child_reaper))
624		- child_reaper = tsk;
	624	+ if (unlikely(tsk->group_leader == child_reaper(tsk)))
	625	+ tsk->nsproxy->pid_ns->child_reaper = tsk;
625	626
626	627	zap_other_threads(tsk);
627	628	read_unlock(&tasklist_lock);
...	...	@@ -35,8 +35,9 @@
35	35	*
36	36	* Holding a reference to struct pid solves both of these problems.
37	37	* It is small so holding a reference does not consume a lot of
38		- * resources, and since a new struct pid is allocated when the numeric
39		- * pid value is reused we don't mistakenly refer to new processes.
	38	+ * resources, and since a new struct pid is allocated when the numeric pid
	39	+ * value is reused (when pids wrap around) we don't mistakenly refer to new
	40	+ * processes.
40	41	*/
41	42
42	43	struct pid
...	...	@@ -19,6 +19,7 @@
19	19	struct kref kref;
20	20	struct pidmap pidmap[PIDMAP_ENTRIES];
21	21	int last_pid;
	22	+ struct task_struct *child_reaper;
22	23	};
23	24
24	25	extern struct pid_namespace init_pid_ns;
...	...	@@ -34,6 +35,11 @@
34	35	static inline void put_pid_ns(struct pid_namespace *ns)
35	36	{
36	37	kref_put(&ns->kref, free_pid_ns);
	38	+}
	39	+
	40	+static inline struct task_struct child_reaper(struct task_struct tsk)
	41	+{
	42	+ return tsk->nsproxy->pid_ns->child_reaper;
37	43	}
38	44
39	45	#endif /* _LINUX_PID_NS_H */
...	...	@@ -1400,7 +1400,6 @@
1400	1400	extern void daemonize(const char *, ...);
1401	1401	extern int allow_signal(int);
1402	1402	extern int disallow_signal(int);
1403		-extern struct task_struct *child_reaper;
1404	1403
1405	1404	extern int do_execve(char , char __user __user , char __user __user , struct pt_regs );
1406	1405	extern long do_fork(unsigned long, unsigned long, struct pt_regs , unsigned long, int __user , int __user *);
...	...	@@ -51,6 +51,7 @@
51	51	#include <linux/debug_locks.h>
52	52	#include <linux/lockdep.h>
53	53	#include <linux/utsrelease.h>
	54	+#include <linux/pid_namespace.h>
54	55	#include <linux/compile.h>
55	56
56	57	#include <asm/io.h>
...	...	@@ -626,8 +627,6 @@
626	627	}
627	628	__setup("initcall_debug", initcall_debug_setup);
628	629
629		-struct task_struct *child_reaper = &init_task;
630		-
631	630	extern initcall_t __initcall_start[], __initcall_end[];
632	631
633	632	static void __init do_initcalls(void)
...	...	@@ -727,7 +726,7 @@
727	726	* assumptions about where in the task array this
728	727	* can be found.
729	728	*/
730		- child_reaper = current;
	729	+ init_pid_ns.child_reaper = current;
731	730
732	731	cad_pid = task_pid(current);
733	732
...	...	@@ -22,6 +22,7 @@
22	22	#include <linux/file.h>
23	23	#include <linux/binfmts.h>
24	24	#include <linux/nsproxy.h>
	25	+#include <linux/pid_namespace.h>
25	26	#include <linux/ptrace.h>
26	27	#include <linux/profile.h>
27	28	#include <linux/mount.h>
...	...	@@ -48,7 +49,6 @@
48	49	#include <asm/mmu_context.h>
49	50
50	51	extern void sem_exit (void);
51		-extern struct task_struct *child_reaper;
52	52
53	53	static void exit_mm(struct task_struct * tsk);
54	54
...	...	@@ -260,7 +260,8 @@
260	260	}
261	261
262	262	/**
263		- * reparent_to_init - Reparent the calling kernel thread to the init task.
	263	+ * reparent_to_init - Reparent the calling kernel thread to the init task
	264	+ * of the pid space that the thread belongs to.
264	265	*
265	266	* If a kernel thread is launched as a result of a system call, or if
266	267	* it ever exits, it should generally reparent itself to init so that
...	...	@@ -278,8 +279,8 @@
278	279	ptrace_unlink(current);
279	280	/* Reparent to init */
280	281	remove_parent(current);
281		- current->parent = child_reaper;
282		- current->real_parent = child_reaper;
	282	+ current->parent = child_reaper(current);
	283	+ current->real_parent = child_reaper(current);
283	284	add_parent(current);
284	285
285	286	/* Set the exit signal to SIGCHLD so we signal init on exit */
...	...	@@ -662,7 +663,8 @@
662	663	* When we die, we re-parent all our children.
663	664	* Try to give them to another thread in our thread
664	665	* group, and if no such member exists, give it to
665		- * the global child reaper process (ie "init")
	666	+ * the child reaper process (ie "init") in our pid
	667	+ * space.
666	668	*/
667	669	static void
668	670	forget_original_parent(struct task_struct father, struct list_head to_release)
...	...	@@ -673,7 +675,7 @@
673	675	do {
674	676	reaper = next_thread(reaper);
675	677	if (reaper == father) {
676		- reaper = child_reaper;
	678	+ reaper = child_reaper(father);
677	679	break;
678	680	}
679	681	} while (reaper->exit_state);
...	...	@@ -859,8 +861,13 @@
859	861	panic("Aiee, killing interrupt handler!");
860	862	if (unlikely(!tsk->pid))
861	863	panic("Attempted to kill the idle task!");
862		- if (unlikely(tsk == child_reaper))
863		- panic("Attempted to kill init!");
	864	+ if (unlikely(tsk == child_reaper(tsk))) {
	865	+ if (tsk->nsproxy->pid_ns != &init_pid_ns)
	866	+ tsk->nsproxy->pid_ns->child_reaper = init_pid_ns.child_reaper;
	867	+ else
	868	+ panic("Attempted to kill init!");
	869	+ }
	870	+
864	871
865	872	if (unlikely(current->ptrace & PT_TRACE_EXIT)) {
866	873	current->ptrace_message = code;
...	...	@@ -65,7 +65,8 @@
65	65	.pidmap = {
66	66	[ 0 ... PIDMAP_ENTRIES-1] = { ATOMIC_INIT(BITS_PER_PAGE), NULL }
67	67	},
68		- .last_pid = 0
	68	+ .last_pid = 0,
	69	+ .child_reaper = &init_task
69	70	};
70	71
71	72	/*
...	...	@@ -24,6 +24,9 @@
24	24	#include <linux/signal.h>
25	25	#include <linux/capability.h>
26	26	#include <linux/freezer.h>
	27	+#include <linux/pid_namespace.h>
	28	+#include <linux/nsproxy.h>
	29	+
27	30	#include <asm/param.h>
28	31	#include <asm/uaccess.h>
29	32	#include <asm/unistd.h>
...	...	@@ -1877,8 +1880,12 @@
1877	1880	if (sig_kernel_ignore(signr)) /* Default is nothing. */
1878	1881	continue;
1879	1882
1880		- /* Init gets no signals it doesn't want. */
1881		- if (current == child_reaper)
	1883	+ /*
	1884	+ * Init of a pid space gets no signals it doesn't want from
	1885	+ * within that pid space. It can of course get signals from
	1886	+ * its parent pid space.
	1887	+ */
	1888	+ if (current == child_reaper(current))
1882	1889	continue;
1883	1890
1884	1891	if (sig_kernel_stop(signr)) {