Skip to contentSkip to conversation

/proc/$PID/sched does not take PID namespace into account

DISCUSSION Emmanuel Deloget 3 participants 5 messages Sep 3, 2013 – Sep 12, 2013 Scheduler

lore
Share & tools
lore
#0 lore raw

Hello,

(please CC me when answering this mail ; and sorry for my borken English).

I noticed that when a process is executed in a PID namespace it can still find his real PID by parsing /proc/$PID_IN_NS/sched.

(in this example I created a new PID namespace and ran /bin/bash at PID 1)

  # uname -a
  Linux my-thinkpad 3.9-1-amd64 #1 SMP Debian 3.9.6-1 x86_64 GNU/Linux
  #
  # pidof bash
  1
  # head -n 1 /proc/1/sched
  bash (14957, #threads: 1)
  # cat /proc/1/stat
  1 (bash) S 0 1 0 34823 220 4202752 11359 (...)

In the root PID namespace

  # pidof bash
  (...) 14957 (...)
  # head -n 1 /proc/14957/sched
  bash (14957, #threads: 1)
  # cat /proc/14957/stat
  14957 (bash) S 14956 14957 23465 34823 14957 4202752 11386 (...)

The principle of least surprise tells me that if my PID is n in a particular PID namespace then every stat or debug info should tell me that my PID is n (i.e. I should not be able to get the real PID of my program). This is a consistency issue: if I get different information from different sources I may not be able to tell which one is the right one.

The issue (if this is really an issue) lies in kernel/sched/debug.c, function proc_sched_show_task(). The code says [1]:

    SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
                        get_nr_threads(p));

I understand that you're not supposed to use the content of a /proc file that has been generated by a kernel source file named "debug.c" in a production environment yet several distributions out there seems to enable this by default (at least the Debian distribution I'm using does it).

I see a few options:

* either it's a bug and it should be corrected (I'm not sure how to do it; the printed PID should reflect the current PID namespace and I don't how how to get this information).

* or it has been decided on purpose (i.e. it's not a bug); /proc/$PID/sched then offers a reliable way to get the real PID of any process, including processes that run in a child PID namespace.

* or it's a bug but it cannot be corrected (kernel ABI ; it has been here for a looooong time - possibly since the PID namespace integration).

(There might be other options here).

In the two last cases there is no need for a patch but I'd be happy if someone explains the reasoning.

Best regards,

-- Emmanuel Deloget

[1]: https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/kernel/sched/debug.c?id=refs/tags/v3.11#n495
[2]: https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/kernel/sched/debug.c?id=refs/tags/v3.11#n118

begin:vcard
fn:Emmanuel Deloget
n:Deloget;Emmanuel
org:efixo / SFR;DATA
adr;quoted-printable:;;67 mont=C3=A9e de St Menet;MARSEILLE;;13011;FRANCE email;internet:[email protected]
title:Team Leader
tel;work:04 88 15 50 77
url:www.sfr.fr
version:2.1
end:vcard

maintainer
#1 lore raw
Quoted 7 lines · Emmanuel Deloget

The issue (if this is really an issue) lies in kernel/sched/debug.c, function proc_sched_show_task(). The code says [1]:

    SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
                        get_nr_threads(p));
Quoted 5 lines

I see a few options:

* either it's a bug and it should be corrected (I'm not sure how to do it; the printed PID should reflect the current PID namespace and I don't how how to get this information).

I suppose something like the below ought to work? -- completely untested, please confirm.

---
Subject: sched, debug: Use PID namespaces

Emmanuel reported that /proc/sched_debug didn't report the right PIDs when using namespaces, cure this.

---
 kernel/sched/debug.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/debug.c +3 -3
index e076bdd..e30bf44 100644
@@ -124,7 +124,7 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		SEQ_printf(m, " ");
 
 	SEQ_printf(m, "%15s %5d %9Ld.%06ld %9Ld %5d ",
-		p->comm, p->pid,
+		p->comm, task_pid(p),,
 		SPLIT_NS(p->se.vruntime),
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
@@ -289,7 +289,7 @@ do {									\
 	P(nr_load_updates);
 	P(nr_uninterruptible);
 	PN(next_balance);
-	P(curr->pid);
+	P(task_pid(curr));
 	PN(clock);
 	P(cpu_load[0]);
 	P(cpu_load[1]);
@@ -492,7 +492,7 @@ void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
 {
 	unsigned long nr_switches;
 
-	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
+	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid(p),
 						get_nr_threads(p));
 	SEQ_printf(m,
 		"---------------------------------------------------------"
#2 lore raw

Hello,

Quoted 7 lines · Peter Zijlstra
Quoted 6 lines · Emmanuel Deloget

The issue (if this is really an issue) lies in kernel/sched/debug.c, function proc_sched_show_task(). The code says [1]:

    SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
                        get_nr_threads(p));
Quoted 7 lines
Quoted 5 lines

I see a few options:

* either it's a bug and it should be corrected (I'm not sure how to do it; the printed PID should reflect the current PID namespace and I don't how how to get this information).

I suppose something like the below ought to work? -- completely untested, please confirm.

Thanks !

I'll test that as soon as possible and I'll report to you.

Best regards,

-- Emmanuel Deloget

Quoted 46 lines

---
Subject: sched, debug: Use PID namespaces

Emmanuel reported that /proc/sched_debug didn't report the right PIDs when using namespaces, cure this.

---
 kernel/sched/debug.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/debug.c +3 -3
index e076bdd..e30bf44 100644
@@ -124,7 +124,7 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		SEQ_printf(m, " ");
 
 	SEQ_printf(m, "%15s %5d %9Ld.%06ld %9Ld %5d ",
-		p->comm, p->pid,
+		p->comm, task_pid(p),,
 		SPLIT_NS(p->se.vruntime),
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
@@ -289,7 +289,7 @@ do {									\
 	P(nr_load_updates);
 	P(nr_uninterruptible);
 	PN(next_balance);
-	P(curr->pid);
+	P(task_pid(curr));
 	PN(clock);
 	P(cpu_load[0]);
 	P(cpu_load[1]);
@@ -492,7 +492,7 @@ void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
 {
 	unsigned long nr_switches;
 
-	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
+	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid(p),
 						get_nr_threads(p));
 	SEQ_printf(m,
 		"---------------------------------------------------------"

.

begin:vcard
fn:Emmanuel Deloget
n:Deloget;Emmanuel
org:efixo / SFR;DATA
adr;quoted-printable:;;67 mont=C3=A9e de St Menet;MARSEILLE;;13011;FRANCE email;internet:[email protected]
title:Team Leader
tel;work:04 88 15 50 77
url:www.sfr.fr
version:2.1
end:vcard

maintainer
#3 lore raw
In reply to Peter Zijlstra
Quoted 52 lines · Peter Zijlstra
Quoted 7 lines · Emmanuel Deloget

The issue (if this is really an issue) lies in kernel/sched/debug.c, function proc_sched_show_task(). The code says [1]:

    SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
                        get_nr_threads(p));
Quoted 5 lines

I see a few options:

* either it's a bug and it should be corrected (I'm not sure how to do it; the printed PID should reflect the current PID namespace and I don't how how to get this information).

I suppose something like the below ought to work? -- completely untested, please confirm.

---
Subject: sched, debug: Use PID namespaces

Emmanuel reported that /proc/sched_debug didn't report the right PIDs when using namespaces, cure this.

---
 kernel/sched/debug.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/debug.c +2 -2
index e076bdd..e30bf44 100644
@@ -124,7 +124,7 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		SEQ_printf(m, " ");
 
 	SEQ_printf(m, "%15s %5d %9Ld.%06ld %9Ld %5d ",
-		p->comm, p->pid,
+		p->comm, task_pid(p),,
 		SPLIT_NS(p->se.vruntime),
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
@@ -289,7 +289,7 @@ do {									\
 	P(nr_load_updates);
 	P(nr_uninterruptible);
 	PN(next_balance);
-	P(curr->pid);
+	P(task_pid(curr));
 	PN(clock);
 	P(cpu_load[0]);
 	P(cpu_load[1]);

kernel/sched/debug.c void proc_sched_show_task(struct task_struct *p, struct seq_file *m) · line 498

Quoted 8 lines
kernel/sched/debug.c +1 -1
@@ -492,7 +492,7 @@ void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
 {
 	unsigned long nr_switches;
 
-	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
+	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid(p),
 						get_nr_threads(p));
 	SEQ_printf(m,
 		"---------------------------------------------------------"

Please do s/task_pid/task_pid_nr/g on this before trying.

#4 lore raw

[tip:sched/core] sched/debug: Take PID namespace into account

In reply to Peter Zijlstra
sched/debug: Take PID namespace into account
Commit-ID:  fc840914e9b07ab4685c195e1e54e58de4f84c03
Gitweb:     http://git.kernel.org/tip/fc840914e9b07ab4685c195e1e54e58de4f84c03
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Mon, 9 Sep 2013 13:01:41 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Thu, 12 Sep 2013 19:14:16 +0200

sched/debug: Take PID namespace into account

Emmanuel reported that /proc/sched_debug didn't report the right PIDs when using namespaces, cure this.

---
 kernel/sched/debug.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
kernel/sched/debug.c +3 -3
index e076bdd..1965599 100644
@@ -124,7 +124,7 @@ print_task(struct seq_file *m, struct rq *rq, struct task_struct *p)
 		SEQ_printf(m, " ");
 
 	SEQ_printf(m, "%15s %5d %9Ld.%06ld %9Ld %5d ",
-		p->comm, p->pid,
+		p->comm, task_pid_nr(p),
 		SPLIT_NS(p->se.vruntime),
 		(long long)(p->nvcsw + p->nivcsw),
 		p->prio);
@@ -289,7 +289,7 @@ do {									\
 	P(nr_load_updates);
 	P(nr_uninterruptible);
 	PN(next_balance);
-	P(curr->pid);
+	SEQ_printf(m, "  .%-30s: %ld\n", "curr->pid", (long)(task_pid_nr(rq->curr)));
 	PN(clock);
 	P(cpu_load[0]);
 	P(cpu_load[1]);
@@ -492,7 +492,7 @@ void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
 {
 	unsigned long nr_switches;
 
-	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, p->pid,
+	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr(p),
 						get_nr_threads(p));
 	SEQ_printf(m,
 		"---------------------------------------------------------"