pid_ns_for_children set by a task is known only to the task itself,
and it's impossible to identify it from outside.
It's a big problem for checkpoint/restore software like CRIU,
because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
in proccess of their work. If they have a custom pid_ns_for_children
before dump, they must have the same ns after restore. Otherwise,
restored task bumped into enviroment it does not expect.
This patchset solves the problem. It exposes pid_ns_for_children
to ns directory in standard way with the name "pid_for_children":
~# ls /proc/5531/ns -l | grep pid
lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
---
Kirill Tkhai (2):
ns: Allow ns_entries to have custom symlink content
pidns: Expose task pid_ns_for_children to userspace
fs/nsfs.c | 4 +++-
fs/proc/namespaces.c | 1 +
include/linux/proc_ns.h | 2 ++
kernel/pid_namespace.c | 25 +++++++++++++++++++++++++
4 files changed, 31 insertions(+), 1 deletion(-)
--
Signed-off-by: Kirill Tkhai <[email protected]>
Make possible to have link content prefix yyy
different from the link name xxx:
$ readlink /proc/[pid]/ns/xxx
yyy:[4026531838]
This will be used in next patch.
Signed-off-by: Kirill Tkhai <[email protected]>
Reviewed-by: Cyrill Gorcunov <[email protected]>
Acked-by: Andrei Vagin <[email protected]>
---
fs/nsfs.c | 4 +++-
include/linux/proc_ns.h | 1 +
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/fs/nsfs.c b/fs/nsfs.c
index 1656843e87d2..495f12b83a7b 100644
--- a/fs/nsfs.c
+++ b/fs/nsfs.c
@@ -195,9 +195,11 @@ int ns_get_name(char *buf, size_t size, struct task_struct *task,
{
struct ns_common *ns;
int res = -ENOENT;
+ const char *name;
ns = ns_ops->get(task);
if (ns) {
- res = snprintf(buf, size, "%s:[%u]", ns_ops->name, ns->inum);
+ name = ns_ops->real_ns_name ? : ns_ops->name;
+ res = snprintf(buf, size, "%s:[%u]", name, ns->inum);
ns_ops->put(ns);
}
return res;
diff --git a/include/linux/proc_ns.h b/include/linux/proc_ns.h
index 12cb8bd81d2d..88dba3b53375 100644
--- a/include/linux/proc_ns.h
+++ b/include/linux/proc_ns.h
@@ -14,6 +14,7 @@ struct inode;
struct proc_ns_operations {
const char *name;
+ const char *real_ns_name;
int type;
struct ns_common *(*get)(struct task_struct *task);
void (*put)(struct ns_common *ns);
pid_ns_for_children set by a task is known only to the task itself,
and it's impossible to identify it from outside.
It's a big problem for checkpoint/restore software like CRIU,
because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
in proccess of their work.
This patch solves the problem, and it exposes pid_ns_for_children
to ns directory in standard way with the name "pid_for_children":
~# ls /proc/5531/ns -l | grep pid
lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
Signed-off-by: Kirill Tkhai <[email protected]>
Reviewed-by: Cyrill Gorcunov <[email protected]>
Acked-by: Andrei Vagin <[email protected]>
---
fs/proc/namespaces.c | 1 +
include/linux/proc_ns.h | 1 +
kernel/pid_namespace.c | 25 +++++++++++++++++++++++++
3 files changed, 27 insertions(+)
diff --git a/fs/proc/namespaces.c b/fs/proc/namespaces.c
index 766f0c637ad1..3803b24ca220 100644
--- a/fs/proc/namespaces.c
+++ b/fs/proc/namespaces.c
@@ -23,6 +23,7 @@ static const struct proc_ns_operations *ns_entries[] = {
#endif
#ifdef CONFIG_PID_NS
&pidns_operations,
+ &pidns_for_children_operations,
#endif
#ifdef CONFIG_USER_NS
&userns_operations,
diff --git a/include/linux/proc_ns.h b/include/linux/proc_ns.h
index 88dba3b53375..58ab28d81fc2 100644
--- a/include/linux/proc_ns.h
+++ b/include/linux/proc_ns.h
@@ -27,6 +27,7 @@ extern const struct proc_ns_operations netns_operations;
extern const struct proc_ns_operations utsns_operations;
extern const struct proc_ns_operations ipcns_operations;
extern const struct proc_ns_operations pidns_operations;
+extern const struct proc_ns_operations pidns_for_children_operations;
extern const struct proc_ns_operations userns_operations;
extern const struct proc_ns_operations mntns_operations;
extern const struct proc_ns_operations cgroupns_operations;
diff --git a/kernel/pid_namespace.c b/kernel/pid_namespace.c
index de461aa0bf9a..4dd02ff0b0bd 100644
--- a/kernel/pid_namespace.c
+++ b/kernel/pid_namespace.c
@@ -374,6 +374,20 @@ static struct ns_common *pidns_get(struct task_struct *task)
return ns ? &ns->ns : NULL;
}
+static struct ns_common *pidns_for_children_get(struct task_struct *task)
+{
+ struct pid_namespace *ns = NULL;
+
+ task_lock(task);
+ if (task->nsproxy) {
+ ns = task->nsproxy->pid_ns_for_children;
+ get_pid_ns(ns);
+ }
+ task_unlock(task);
+
+ return ns ? &ns->ns : NULL;
+}
+
static void pidns_put(struct ns_common *ns)
{
put_pid_ns(to_pid_ns(ns));
@@ -443,6 +457,17 @@ const struct proc_ns_operations pidns_operations = {
.get_parent = pidns_get_parent,
};
+const struct proc_ns_operations pidns_for_children_operations = {
+ .name = "pid_for_children",
+ .real_ns_name = "pid",
+ .type = CLONE_NEWPID,
+ .get = pidns_for_children_get,
+ .put = pidns_put,
+ .install = pidns_install,
+ .owner = pidns_owner,
+ .get_parent = pidns_get_parent,
+};
+
static __init int pid_namespaces_init(void)
{
pid_ns_cachep = KMEM_CACHE(pid_namespace, SLAB_PANIC);
On Thu, 30 Mar 2017 13:27:59 +0300 Kirill Tkhai <[email protected]> wrote:
> pid_ns_for_children set by a task is known only to the task itself,
> and it's impossible to identify it from outside.
>
> It's a big problem for checkpoint/restore software like CRIU,
> because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
> in proccess of their work.
>
> This patch solves the problem, and it exposes pid_ns_for_children
> to ns directory in standard way with the name "pid_for_children":
>
> ~# ls /proc/5531/ns -l | grep pid
> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
>
> --- a/fs/proc/namespaces.c
> +++ b/fs/proc/namespaces.c
> @@ -23,6 +23,7 @@ static const struct proc_ns_operations *ns_entries[] = {
> #endif
> #ifdef CONFIG_PID_NS
> &pidns_operations,
> + &pidns_for_children_operations,
> #endif
This interface should be documented somewhere under Documentation/.
But I can't immediately find where the /proc/pid/ns/ pseudo-files are
documented...
On Thu, Mar 30, 2017 at 03:05:20PM -0700, Andrew Morton wrote:
> On Thu, 30 Mar 2017 13:27:59 +0300 Kirill Tkhai <[email protected]> wrote:
>
> > pid_ns_for_children set by a task is known only to the task itself,
> > and it's impossible to identify it from outside.
> >
> > It's a big problem for checkpoint/restore software like CRIU,
> > because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
> > in proccess of their work.
> >
> > This patch solves the problem, and it exposes pid_ns_for_children
> > to ns directory in standard way with the name "pid_for_children":
> >
> > ~# ls /proc/5531/ns -l | grep pid
> > lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
> > lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
> >
> > --- a/fs/proc/namespaces.c
> > +++ b/fs/proc/namespaces.c
> > @@ -23,6 +23,7 @@ static const struct proc_ns_operations *ns_entries[] = {
> > #endif
> > #ifdef CONFIG_PID_NS
> > &pidns_operations,
> > + &pidns_for_children_operations,
> > #endif
>
> This interface should be documented somewhere under Documentation/.
> But I can't immediately find where the /proc/pid/ns/ pseudo-files are
> documented...
I know that they are documented in man7/namespaces.7
https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/man7/namespaces.7#n187
>
>
On 31.03.2017 04:04, Andrei Vagin wrote:
> On Thu, Mar 30, 2017 at 03:05:20PM -0700, Andrew Morton wrote:
>> On Thu, 30 Mar 2017 13:27:59 +0300 Kirill Tkhai <[email protected]> wrote:
>>
>>> pid_ns_for_children set by a task is known only to the task itself,
>>> and it's impossible to identify it from outside.
>>>
>>> It's a big problem for checkpoint/restore software like CRIU,
>>> because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
>>> in proccess of their work.
>>>
>>> This patch solves the problem, and it exposes pid_ns_for_children
>>> to ns directory in standard way with the name "pid_for_children":
>>>
>>> ~# ls /proc/5531/ns -l | grep pid
>>> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
>>> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
>>>
>>> --- a/fs/proc/namespaces.c
>>> +++ b/fs/proc/namespaces.c
>>> @@ -23,6 +23,7 @@ static const struct proc_ns_operations *ns_entries[] = {
>>> #endif
>>> #ifdef CONFIG_PID_NS
>>> &pidns_operations,
>>> + &pidns_for_children_operations,
>>> #endif
>>
>> This interface should be documented somewhere under Documentation/.
>> But I can't immediately find where the /proc/pid/ns/ pseudo-files are
>> documented...
>
> I know that they are documented in man7/namespaces.7
>
> https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/man7/namespaces.7#n187
I suggest the below patch, but it's too early for the man description till
the feature is in mainline, because the man page requires commit id of the feature.
[PATCH] namespaces.7: Document the /proc/[pid]/ns/pid_for_children file
Signed-off-by: Kirill Tkhai <[email protected]>
---
man7/namespaces.7 | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/man7/namespaces.7 b/man7/namespaces.7
index 6dfceaa2a..06041774f 100644
--- a/man7/namespaces.7
+++ b/man7/namespaces.7
@@ -125,6 +125,7 @@ lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 ipc \-> ipc:[4026531839]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 mnt \-> mnt:[4026531840]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 net \-> net:[4026531969]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid \-> pid:[4026531836]
+lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid_for_children -> pid:[4026531834]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 user \-> user:[4026531837]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 uts \-> uts:[4026531838]
.fi
@@ -186,7 +187,14 @@ This file is a handle for the network namespace of the process.
.TP
.IR /proc/[pid]/ns/pid " (since Linux 3.8)"
.\" commit 57e8391d327609cbf12d843259c968b9e5c1838f
-This file is a handle for the PID namespace of the process.
+This file is a handle for the PID namespace of the process. It's
+permanent during the whole process life.
+.TP
+.IR /proc/[pid]/ns/pid_for_children " (since Linux 4.12)"
+.\" commit FIXME
+This file is a handle for the PID namespace of a next born child
+of the process. It's changed after unshare(2) and via setns(2),
+so the file may differ from /proc/[pid]/ns/pid.
.TP
.IR /proc/[pid]/ns/user " (since Linux 3.8)"
.\" commit cde1975bc242f3e1072bde623ef378e547b73f91
On Fri, Mar 31, 2017 at 2:45 AM, Kirill Tkhai <[email protected]> wrote:
> On 31.03.2017 04:04, Andrei Vagin wrote:
>> On Thu, Mar 30, 2017 at 03:05:20PM -0700, Andrew Morton wrote:
>>> On Thu, 30 Mar 2017 13:27:59 +0300 Kirill Tkhai <[email protected]> wrote:
>>>
>>>> pid_ns_for_children set by a task is known only to the task itself,
>>>> and it's impossible to identify it from outside.
>>>>
>>>> It's a big problem for checkpoint/restore software like CRIU,
>>>> because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
>>>> in proccess of their work.
>>>>
>>>> This patch solves the problem, and it exposes pid_ns_for_children
>>>> to ns directory in standard way with the name "pid_for_children":
>>>>
>>>> ~# ls /proc/5531/ns -l | grep pid
>>>> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
>>>> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
>>>>
>>>> --- a/fs/proc/namespaces.c
>>>> +++ b/fs/proc/namespaces.c
>>>> @@ -23,6 +23,7 @@ static const struct proc_ns_operations *ns_entries[] = {
>>>> #endif
>>>> #ifdef CONFIG_PID_NS
>>>> &pidns_operations,
>>>> + &pidns_for_children_operations,
>>>> #endif
>>>
>>> This interface should be documented somewhere under Documentation/.
>>> But I can't immediately find where the /proc/pid/ns/ pseudo-files are
>>> documented...
>>
>> I know that they are documented in man7/namespaces.7
>>
>> https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/man7/namespaces.7#n187
>
> I suggest the below patch, but it's too early for the man description till
> the feature is in mainline, because the man page requires commit id of the feature.
>
> [PATCH] namespaces.7: Document the /proc/[pid]/ns/pid_for_children file
>
> Signed-off-by: Kirill Tkhai <[email protected]>
> ---
> man7/namespaces.7 | 10 +++++++++-
> 1 file changed, 9 insertions(+), 1 deletion(-)
> diff --git a/man7/namespaces.7 b/man7/namespaces.7
> index 6dfceaa2a..06041774f 100644
> --- a/man7/namespaces.7
> +++ b/man7/namespaces.7
> @@ -125,6 +125,7 @@ lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 ipc \-> ipc:[4026531839]
> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 mnt \-> mnt:[4026531840]
> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 net \-> net:[4026531969]
> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid \-> pid:[4026531836]
> +lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid_for_children -> pid:[4026531834]
Minor nit: this needs to be "\-" for the "-"
> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 user \-> user:[4026531837]
> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 uts \-> uts:[4026531838]
> .fi
> @@ -186,7 +187,14 @@ This file is a handle for the network namespace of the process.
> .TP
> .IR /proc/[pid]/ns/pid " (since Linux 3.8)"
> .\" commit 57e8391d327609cbf12d843259c968b9e5c1838f
> -This file is a handle for the PID namespace of the process.
> +This file is a handle for the PID namespace of the process. It's
> +permanent during the whole process life.
> +.TP
> +.IR /proc/[pid]/ns/pid_for_children " (since Linux 4.12)"
> +.\" commit FIXME
> +This file is a handle for the PID namespace of a next born child
> +of the process. It's changed after unshare(2) and via setns(2),
> +so the file may differ from /proc/[pid]/ns/pid.
> .TP
> .IR /proc/[pid]/ns/user " (since Linux 3.8)"
> .\" commit cde1975bc242f3e1072bde623ef378e547b73f91
>
-Kees
--
Kees Cook
Pixel Security
On 31.03.2017 18:06, Kees Cook wrote:
> On Fri, Mar 31, 2017 at 2:45 AM, Kirill Tkhai <[email protected]> wrote:
>> On 31.03.2017 04:04, Andrei Vagin wrote:
>>> On Thu, Mar 30, 2017 at 03:05:20PM -0700, Andrew Morton wrote:
>>>> On Thu, 30 Mar 2017 13:27:59 +0300 Kirill Tkhai <[email protected]> wrote:
>>>>
>>>>> pid_ns_for_children set by a task is known only to the task itself,
>>>>> and it's impossible to identify it from outside.
>>>>>
>>>>> It's a big problem for checkpoint/restore software like CRIU,
>>>>> because it can't correctly handle tasks, that do setns(CLONE_NEWPID)
>>>>> in proccess of their work.
>>>>>
>>>>> This patch solves the problem, and it exposes pid_ns_for_children
>>>>> to ns directory in standard way with the name "pid_for_children":
>>>>>
>>>>> ~# ls /proc/5531/ns -l | grep pid
>>>>> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid -> pid:[4026531836]
>>>>> lrwxrwxrwx 1 root root 0 Jan 14 16:38 pid_for_children -> pid:[4026532286]
>>>>>
>>>>> --- a/fs/proc/namespaces.c
>>>>> +++ b/fs/proc/namespaces.c
>>>>> @@ -23,6 +23,7 @@ static const struct proc_ns_operations *ns_entries[] = {
>>>>> #endif
>>>>> #ifdef CONFIG_PID_NS
>>>>> &pidns_operations,
>>>>> + &pidns_for_children_operations,
>>>>> #endif
>>>>
>>>> This interface should be documented somewhere under Documentation/.
>>>> But I can't immediately find where the /proc/pid/ns/ pseudo-files are
>>>> documented...
>>>
>>> I know that they are documented in man7/namespaces.7
>>>
>>> https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/man7/namespaces.7#n187
>>
>> I suggest the below patch, but it's too early for the man description till
>> the feature is in mainline, because the man page requires commit id of the feature.
>>
>> [PATCH] namespaces.7: Document the /proc/[pid]/ns/pid_for_children file
>>
>> Signed-off-by: Kirill Tkhai <[email protected]>
>> ---
>> man7/namespaces.7 | 10 +++++++++-
>> 1 file changed, 9 insertions(+), 1 deletion(-)
>> diff --git a/man7/namespaces.7 b/man7/namespaces.7
>> index 6dfceaa2a..06041774f 100644
>> --- a/man7/namespaces.7
>> +++ b/man7/namespaces.7
>> @@ -125,6 +125,7 @@ lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 ipc \-> ipc:[4026531839]
>> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 mnt \-> mnt:[4026531840]
>> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 net \-> net:[4026531969]
>> lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid \-> pid:[4026531836]
>> +lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid_for_children -> pid:[4026531834]
>
> Minor nit: this needs to be "\-" for the "-"
Thanks, Kees. The updated version is below:
[PATCH] namespaces.7: Document the /proc/[pid]/ns/pid_for_children file
Signed-off-by: Kirill Tkhai <[email protected]>
---
man7/namespaces.7 | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
diff --git a/man7/namespaces.7 b/man7/namespaces.7
index 6dfceaa2a..b4e9b13f0 100644
--- a/man7/namespaces.7
+++ b/man7/namespaces.7
@@ -125,6 +125,7 @@ lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 ipc \-> ipc:[4026531839]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 mnt \-> mnt:[4026531840]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 net \-> net:[4026531969]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid \-> pid:[4026531836]
+lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 pid_for_children \-> pid:[4026531834]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 user \-> user:[4026531837]
lrwxrwxrwx. 1 mtk mtk 0 Apr 28 12:46 uts \-> uts:[4026531838]
.fi
@@ -186,7 +187,14 @@ This file is a handle for the network namespace of the process.
.TP
.IR /proc/[pid]/ns/pid " (since Linux 3.8)"
.\" commit 57e8391d327609cbf12d843259c968b9e5c1838f
-This file is a handle for the PID namespace of the process.
+This file is a handle for the PID namespace of the process. It's
+permanent during the whole process life.
+.TP
+.IR /proc/[pid]/ns/pid_for_children " (since Linux 4.12)"
+.\" commit FIXME
+This file is a handle for the PID namespace of a next born child
+of the process. It's changed after unshare(2) and via setns(2),
+so the file may differ from /proc/[pid]/ns/pid.
.TP
.IR /proc/[pid]/ns/user " (since Linux 3.8)"
.\" commit cde1975bc242f3e1072bde623ef378e547b73f91