Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1704743 > unrolled thread

[PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched

Back to article view | Back to linux.kernel

Contents

  [PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched Aleksa Sarai <asarai@suse.com> - 2017-08-06 06:50 +0200
    Re: [PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched ebiederm@xmission.com (Eric W. Biederman) - 2017-08-06 17:10 +0200
    Re: [PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched Peter Zijlstra <peterz@infradead.org> - 2017-08-07 10:50 +0200
      Re: [PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched Jessie Frazelle <jessfraz@google.com> - 2017-08-07 17:20 +0200
        Re: [PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched ebiederm@xmission.com (Eric W. Biederman) - 2017-08-08 17:30 +0200
    [tip:sched/core] sched/debug: Use task_pid_nr_ns in  /proc/$pid/sched tip-bot for Aleksa Sarai <tipbot@zytor.com> - 2017-08-10 14:20 +0200

#1704743 — [PATCH v2] sched: debug: use task_pid_nr_ns in /proc/$pid/sched

It appears as though the addition of the PID namespace did not update
the output code for /proc/*/sched, which resulted in it providing PIDs
that were not self-consistent with the /proc mount. This additionally
made it trivial to detect whether a process was inside &init_pid_ns from
userspace (making container detection trivial[1]). This lead to
situations such as:

  % unshare -pmf
  % mount -t proc proc /proc
  % head -n1 /proc/1/sched
  head (10047, #threads: 1)

Fix this by just using task_pid_nr_ns for the output of /proc/*/sched.
All of the other uses of task_pid_nr in kernel/sched/debug.c are from a
sysctl context and thus don't need to be namespaced.

[1]: https://github.com/jessfraz/amicontained

Cc: <stable@vger.kernel.org>
Cc: Jess Frazelle <acidburn@google.com>
Signed-off-by: Aleksa Sarai <asarai@suse.com>
---
 fs/proc/base.c              | 3 ++-
 include/linux/sched/debug.h | 4 +++-
 kernel/sched/debug.c        | 5 +++--
 3 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/fs/proc/base.c b/fs/proc/base.c
index 719c2e943ea1..98fd8f6df851 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1408,12 +1408,13 @@ static const struct file_operations proc_fail_nth_operations = {
 static int sched_show(struct seq_file *m, void *v)
 {
 	struct inode *inode = m->private;
+	struct pid_namespace *ns = inode->i_sb->s_fs_info;
 	struct task_struct *p;
 
 	p = get_proc_task(inode);
 	if (!p)
 		return -ESRCH;
-	proc_sched_show_task(p, m);
+	proc_sched_show_task(p, ns, m);
 
 	put_task_struct(p);
 
diff --git a/include/linux/sched/debug.h b/include/linux/sched/debug.h
index e0eaee54c5a4..5d58d49e9f87 100644
--- a/include/linux/sched/debug.h
+++ b/include/linux/sched/debug.h
@@ -6,6 +6,7 @@
  */
 
 struct task_struct;
+struct pid_namespace;
 
 extern void dump_cpu_task(int cpu);
 
@@ -34,7 +35,8 @@ extern void sched_show_task(struct task_struct *p);
 
 #ifdef CONFIG_SCHED_DEBUG
 struct seq_file;
-extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
+extern void proc_sched_show_task(struct task_struct *p,
+				 struct pid_namespace *ns, struct seq_file *m);
 extern void proc_sched_set_task(struct task_struct *p);
 #endif
 
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 4fa66de52bd6..ac345115877b 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -872,11 +872,12 @@ static void sched_show_numa(struct task_struct *p, struct seq_file *m)
 #endif
 }
 
-void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
+void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
+						  struct seq_file *m)
 {
 	unsigned long nr_switches;
 
-	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr(p),
+	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr_ns(p, ns),
 						get_nr_threads(p));
 	SEQ_printf(m,
 		"---------------------------------------------------------"
-- 
2.13.3

[toc] | [next] | [standalone]

#1704871

Aleksa Sarai <asarai@suse.com> writes:

> It appears as though the addition of the PID namespace did not update
> the output code for /proc/*/sched, which resulted in it providing PIDs
> that were not self-consistent with the /proc mount. This additionally
> made it trivial to detect whether a process was inside &init_pid_ns from
> userspace (making container detection trivial[1]). This lead to
> situations such as:
>
>   % unshare -pmf
>   % mount -t proc proc /proc
>   % head -n1 /proc/1/sched
>   head (10047, #threads: 1)
>
> Fix this by just using task_pid_nr_ns for the output of /proc/*/sched.
> All of the other uses of task_pid_nr in kernel/sched/debug.c are from a
> sysctl context and thus don't need to be namespaced.
>
> [1]: https://github.com/jessfraz/amicontained
>
> Cc: <stable@vger.kernel.org>
> Cc: Jess Frazelle <acidburn@google.com>
> Signed-off-by: Aleksa Sarai <asarai@suse.com>

Acked-by: "Eric W. Biederman" <ebiederm@xmission.com>

> ---
>  fs/proc/base.c              | 3 ++-
>  include/linux/sched/debug.h | 4 +++-
>  kernel/sched/debug.c        | 5 +++--
>  3 files changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index 719c2e943ea1..98fd8f6df851 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -1408,12 +1408,13 @@ static const struct file_operations proc_fail_nth_operations = {
>  static int sched_show(struct seq_file *m, void *v)
>  {
>  	struct inode *inode = m->private;
> +	struct pid_namespace *ns = inode->i_sb->s_fs_info;
>  	struct task_struct *p;
>  
>  	p = get_proc_task(inode);
>  	if (!p)
>  		return -ESRCH;
> -	proc_sched_show_task(p, m);
> +	proc_sched_show_task(p, ns, m);
>  
>  	put_task_struct(p);
>  
> diff --git a/include/linux/sched/debug.h b/include/linux/sched/debug.h
> index e0eaee54c5a4..5d58d49e9f87 100644
> --- a/include/linux/sched/debug.h
> +++ b/include/linux/sched/debug.h
> @@ -6,6 +6,7 @@
>   */
>  
>  struct task_struct;
> +struct pid_namespace;
>  
>  extern void dump_cpu_task(int cpu);
>  
> @@ -34,7 +35,8 @@ extern void sched_show_task(struct task_struct *p);
>  
>  #ifdef CONFIG_SCHED_DEBUG
>  struct seq_file;
> -extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
> +extern void proc_sched_show_task(struct task_struct *p,
> +				 struct pid_namespace *ns, struct seq_file *m);
>  extern void proc_sched_set_task(struct task_struct *p);
>  #endif
>  
> diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
> index 4fa66de52bd6..ac345115877b 100644
> --- a/kernel/sched/debug.c
> +++ b/kernel/sched/debug.c
> @@ -872,11 +872,12 @@ static void sched_show_numa(struct task_struct *p, struct seq_file *m)
>  #endif
>  }
>  
> -void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
> +void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
> +						  struct seq_file *m)
>  {
>  	unsigned long nr_switches;
>  
> -	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr(p),
> +	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr_ns(p, ns),
>  						get_nr_threads(p));
>  	SEQ_printf(m,
>  		"---------------------------------------------------------"

[toc] | [prev] | [next] | [standalone]

#1705261

On Sun, Aug 06, 2017 at 02:41:41PM +1000, Aleksa Sarai wrote:
> It appears as though the addition of the PID namespace did not update
> the output code for /proc/*/sched, which resulted in it providing PIDs
> that were not self-consistent with the /proc mount. This additionally
> made it trivial to detect whether a process was inside &init_pid_ns from
> userspace (making container detection trivial[1]). This lead to
> situations such as:
> 
>   % unshare -pmf
>   % mount -t proc proc /proc
>   % head -n1 /proc/1/sched
>   head (10047, #threads: 1)
> 
> Fix this by just using task_pid_nr_ns for the output of /proc/*/sched.
> All of the other uses of task_pid_nr in kernel/sched/debug.c are from a
> sysctl context and thus don't need to be namespaced.
> 
> [1]: https://github.com/jessfraz/amicontained
> 
> Cc: <stable@vger.kernel.org>
> Cc: Jess Frazelle <acidburn@google.com>
> Signed-off-by: Aleksa Sarai <asarai@suse.com>

Thanks!

[toc] | [prev] | [next] | [standalone]

#1705635

[Multipart message — attachments visible in raw view] — view raw

Thanks for patching this :) Guess I will find a different rabbit in a
hat to detect if running in a pid namespace ;)

On Mon, Aug 7, 2017 at 4:44 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> On Sun, Aug 06, 2017 at 02:41:41PM +1000, Aleksa Sarai wrote:
>> It appears as though the addition of the PID namespace did not update
>> the output code for /proc/*/sched, which resulted in it providing PIDs
>> that were not self-consistent with the /proc mount. This additionally
>> made it trivial to detect whether a process was inside &init_pid_ns from
>> userspace (making container detection trivial[1]). This lead to
>> situations such as:
>>
>>   % unshare -pmf
>>   % mount -t proc proc /proc
>>   % head -n1 /proc/1/sched
>>   head (10047, #threads: 1)
>>
>> Fix this by just using task_pid_nr_ns for the output of /proc/*/sched.
>> All of the other uses of task_pid_nr in kernel/sched/debug.c are from a
>> sysctl context and thus don't need to be namespaced.
>>
>> [1]: https://github.com/jessfraz/amicontained
>>
>> Cc: <stable@vger.kernel.org>
>> Cc: Jess Frazelle <acidburn@google.com>
>> Signed-off-by: Aleksa Sarai <asarai@suse.com>
>
> Thanks!

[toc] | [prev] | [next] | [standalone]

#1706649

Jessie Frazelle <jessfraz@google.com> writes:

> Thanks for patching this :) Guess I will find a different rabbit in a
> hat to detect if running in a pid namespace ;)

I want to say that stating the /proc/pid/ns/pid files ought to be an
easy way.  But thinking about it they permissions on those might be a
bit tricky.

Eric

[toc] | [prev] | [next] | [standalone]

#1708546 — [tip:sched/core] sched/debug: Use task_pid_nr_ns in /proc/$pid/sched

Commit-ID:  74dc3384fc7983b78cc46ebb1824968a3db85eb1
Gitweb:     http://git.kernel.org/tip/74dc3384fc7983b78cc46ebb1824968a3db85eb1
Author:     Aleksa Sarai <asarai@suse.com>
AuthorDate: Sun, 6 Aug 2017 14:41:41 +1000
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Thu, 10 Aug 2017 12:18:19 +0200

sched/debug: Use task_pid_nr_ns in /proc/$pid/sched

It appears as though the addition of the PID namespace did not update
the output code for /proc/*/sched, which resulted in it providing PIDs
that were not self-consistent with the /proc mount. This additionally
made it trivial to detect whether a process was inside &init_pid_ns from
userspace, making container detection trivial:

   https://github.com/jessfraz/amicontained

This leads to situations such as:

  % unshare -pmf
  % mount -t proc proc /proc
  % head -n1 /proc/1/sched
  head (10047, #threads: 1)

Fix this by just using task_pid_nr_ns for the output of /proc/*/sched.
All of the other uses of task_pid_nr in kernel/sched/debug.c are from a
sysctl context and thus don't need to be namespaced.

Signed-off-by: Aleksa Sarai <asarai@suse.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Cc: Jess Frazelle <acidburn@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: cyphar@cyphar.com
Link: http://lkml.kernel.org/r/20170806044141.5093-1-asarai@suse.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 fs/proc/base.c              | 3 ++-
 include/linux/sched/debug.h | 4 +++-
 kernel/sched/debug.c        | 5 +++--
 3 files changed, 8 insertions(+), 4 deletions(-)

diff --git a/fs/proc/base.c b/fs/proc/base.c
index 719c2e9..98fd8f6 100644
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1408,12 +1408,13 @@ static const struct file_operations proc_fail_nth_operations = {
 static int sched_show(struct seq_file *m, void *v)
 {
 	struct inode *inode = m->private;
+	struct pid_namespace *ns = inode->i_sb->s_fs_info;
 	struct task_struct *p;
 
 	p = get_proc_task(inode);
 	if (!p)
 		return -ESRCH;
-	proc_sched_show_task(p, m);
+	proc_sched_show_task(p, ns, m);
 
 	put_task_struct(p);
 
diff --git a/include/linux/sched/debug.h b/include/linux/sched/debug.h
index e0eaee5..5d58d49 100644
--- a/include/linux/sched/debug.h
+++ b/include/linux/sched/debug.h
@@ -6,6 +6,7 @@
  */
 
 struct task_struct;
+struct pid_namespace;
 
 extern void dump_cpu_task(int cpu);
 
@@ -34,7 +35,8 @@ extern void sched_show_task(struct task_struct *p);
 
 #ifdef CONFIG_SCHED_DEBUG
 struct seq_file;
-extern void proc_sched_show_task(struct task_struct *p, struct seq_file *m);
+extern void proc_sched_show_task(struct task_struct *p,
+				 struct pid_namespace *ns, struct seq_file *m);
 extern void proc_sched_set_task(struct task_struct *p);
 #endif
 
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 4fa66de..ac34511 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -872,11 +872,12 @@ static void sched_show_numa(struct task_struct *p, struct seq_file *m)
 #endif
 }
 
-void proc_sched_show_task(struct task_struct *p, struct seq_file *m)
+void proc_sched_show_task(struct task_struct *p, struct pid_namespace *ns,
+						  struct seq_file *m)
 {
 	unsigned long nr_switches;
 
-	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr(p),
+	SEQ_printf(m, "%s (%d, #threads: %d)\n", p->comm, task_pid_nr_ns(p, ns),
 						get_nr_threads(p));
 	SEQ_printf(m,
 		"---------------------------------------------------------"

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web