linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] pid: restore the old behaviour of the ns_last_pid sysctl
@ 2017-11-03 20:31 Andrei Vagin
  2017-11-06 14:47 ` Oleg Nesterov
  0 siblings, 1 reply; 3+ messages in thread
From: Andrei Vagin @ 2017-11-03 20:31 UTC (permalink / raw)
  To: Andrew Morton; +Cc: linux-kernel, Andrei Vagin, Gargi Sharma, Oleg Nesterov

CRIU uses ns_last_pid to fork a process with a specified pid. For
example, if we need to create a process with the pid of 10000,
we write 9999 into /proc/sys/kernel/ns_last_pid

$ echo 9999 > /proc/sys/kernel/ns_last_pid; sh -c 'echo $$'
10000

This behaviour has been broken and now if we write 9999 to ns_last_pid,
a process will get the pid 9999. This patch restores the old behaviour.

fixes: ("pid: replace pid bitmap implementation with IDR API")
Cc: Gargi Sharma <gs051095@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrei Vagin <avagin@openvz.org>
---
 kernel/pid_namespace.c | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index fea2c24fa460..504dadb1d920 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -287,6 +287,7 @@ static int pid_ns_ctl_handler(struct ctl_table *table, int write,
 {
 	struct pid_namespace *pid_ns = task_active_pid_ns(current);
 	struct ctl_table tmp = *table;
+	int ret, next;
 
 	if (write && !ns_capable(pid_ns->user_ns, CAP_SYS_ADMIN))
 		return -EPERM;
@@ -297,8 +298,18 @@ static int pid_ns_ctl_handler(struct ctl_table *table, int write,
 	 * it should synchronize its usage with external means.
 	 */
 
-	tmp.data = &pid_ns->idr.idr_next;
-	return proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
+	next = idr_get_cursor(&pid_ns->idr) - 1;
+
+	tmp.data = &next;
+	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
+	if (ret < 0)
+		return ret;
+
+	if (!write)
+		return 0;
+
+	idr_set_cursor(&pid_ns->idr, next + 1);
+	return 0;
 }
 
 extern int pid_max;
-- 
2.13.6

^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] pid: restore the old behaviour of the ns_last_pid sysctl
  2017-11-03 20:31 [PATCH] pid: restore the old behaviour of the ns_last_pid sysctl Andrei Vagin
@ 2017-11-06 14:47 ` Oleg Nesterov
  2017-11-06 18:31   ` [PATCH v2] " Andrei Vagin
  0 siblings, 1 reply; 3+ messages in thread
From: Oleg Nesterov @ 2017-11-06 14:47 UTC (permalink / raw)
  To: Andrei Vagin; +Cc: Andrew Morton, linux-kernel, Gargi Sharma

On 11/03, Andrei Vagin wrote:
>
> @@ -297,8 +298,18 @@ static int pid_ns_ctl_handler(struct ctl_table *table, int write,
>  	 * it should synchronize its usage with external means.
>  	 */
>  
> -	tmp.data = &pid_ns->idr.idr_next;
> -	return proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
> +	next = idr_get_cursor(&pid_ns->idr) - 1;
> +
> +	tmp.data = &next;
> +	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
> +	if (ret < 0)
> +		return ret;
> +
> +	if (!write)
> +		return 0;
> +
> +	idr_set_cursor(&pid_ns->idr, next + 1);
> +	return 0;

Ah yes, we should also take "write" into account, I forgot it is readable.
Can't resist, to me

	err = proc_dointvec_minmax(...);
	if (!err && write)
		idr_set_cursor(...);

	return err;

looks a bit more readable, but this is matter of taste of course.


Acked-by: Oleg Nesterov <oleg@redhat.com>

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [PATCH v2] pid: restore the old behaviour of the ns_last_pid sysctl
  2017-11-06 14:47 ` Oleg Nesterov
@ 2017-11-06 18:31   ` Andrei Vagin
  0 siblings, 0 replies; 3+ messages in thread
From: Andrei Vagin @ 2017-11-06 18:31 UTC (permalink / raw)
  To: Andrew Morton; +Cc: linux-kernel, Andrei Vagin, Gargi Sharma, Oleg Nesterov

CRIU uses ns_last_pid to fork a process with a specified pid. For
example, if we need to create a process with the pid of 10000,
we write 9999 into /proc/sys/kernel/ns_last_pid

$ echo 9999 > /proc/sys/kernel/ns_last_pid; sh -c 'echo $$'
10000

This behaviour has been broken and now if we write 9999 to ns_last_pid,
a process will get the pid 9999. This patch restores the old behaviour.

v2: make code a bit more readable // Oleg

fixes: ("pid: replace pid bitmap implementation with IDR API")
Cc: Gargi Sharma <gs051095@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrei Vagin <avagin@openvz.org>
---
 kernel/pid_namespace.c | 11 +++++++++--
 1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index fea2c24fa460..0b53eef7d34b 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -287,6 +287,7 @@ static int pid_ns_ctl_handler(struct ctl_table *table, int write,
 {
 	struct pid_namespace *pid_ns = task_active_pid_ns(current);
 	struct ctl_table tmp = *table;
+	int ret, next;
 
 	if (write && !ns_capable(pid_ns->user_ns, CAP_SYS_ADMIN))
 		return -EPERM;
@@ -297,8 +298,14 @@ static int pid_ns_ctl_handler(struct ctl_table *table, int write,
 	 * it should synchronize its usage with external means.
 	 */
 
-	tmp.data = &pid_ns->idr.idr_next;
-	return proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
+	next = idr_get_cursor(&pid_ns->idr) - 1;
+
+	tmp.data = &next;
+	ret = proc_dointvec_minmax(&tmp, write, buffer, lenp, ppos);
+	if (!ret && write)
+		idr_set_cursor(&pid_ns->idr, next + 1);
+
+	return ret;
 }
 
 extern int pid_max;
-- 
2.13.6

^ permalink raw reply related	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2017-11-06 18:31 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-11-03 20:31 [PATCH] pid: restore the old behaviour of the ns_last_pid sysctl Andrei Vagin
2017-11-06 14:47 ` Oleg Nesterov
2017-11-06 18:31   ` [PATCH v2] " Andrei Vagin

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).