[0/5,RFC] Add an interface to discover relationships between namespaces
diff mbox

Message ID 20160721210650.GA10989@outlook.office365.com
State New
Headers show

Commit Message

Andrey Vagin July 21, 2016, 9:06 p.m. UTC
On Thu, Jul 21, 2016 at 04:41:12PM +0200, Michael Kerrisk (man-pages) wrote:
> Hi Andrey,
> 
> On 07/14/2016 08:20 PM, Andrey Vagin wrote:

<snip>

> 
> Could you add here an of the API in detail: what do these FDs refer to,
> and how do you use them to solve the use case? And could you you add
> that info to the commit messages please.

Hi Michael,

A patch for man-pages is attached. It adds the following text to
namespaces(7).

Since  Linux 4.X, the following ioctl(2) calls are supported for names‐
pace file descriptors.  The correct syntax is:

      fd = ioctl(ns_fd, ioctl_type);

where ioctl_type is one of the following:

NS_GET_USERNS
      Returns a file descriptor that refers to an owning  user  names‐
      pace.

NS_GET_PARENT
      Returns  a  file  descriptor  that refers to a parent namespace.
      This ioctl(2) can be used for pid and user namespaces. For  user
      namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
      ing.

In addition to generic ioctl(2) errors, the following specific ones can
occur:

EINVAL NS_GET_PARENT was called for a nonhierarchical namespace.

EPERM  The  requested  namespace  is  outside  of the current namespace
      scope.

ENOENT ns_fd refers to the init namespace.

Thanks,
Andrew

> 
> Thanks,
> 
> Michael
> 
> 
> > [1] https://lkml.org/lkml/2016/7/6/158
> > [2] https://lkml.org/lkml/2016/7/9/101
> > 
> > Cc: "Eric W. Biederman" <ebiederm@xmission.com>
> > Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
> > Cc: "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com>
> > Cc: "W. Trevor King" <wking@tremily.us>
> > Cc: Alexander Viro <viro@zeniv.linux.org.uk>
> > Cc: Serge Hallyn <serge.hallyn@canonical.com>
> > 
> > --
> > 2.5.5
> > 
> > 
> 
> 
> -- 
> Michael Kerrisk
> Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
> Linux/UNIX System Programming Training: http://man7.org/training/
From 4b9194026f901c2247150bb3038c41658700f6dd Mon Sep 17 00:00:00 2001
From: Andrey Vagin <avagin@openvz.org>
Date: Thu, 21 Jul 2016 13:58:06 -0700
Subject: [PATCH] namespace.7: descirbe NS_GET_USERNS and NS_GET-PARENT ioctl-s

Signed-off-by: Andrey Vagin <avagin@openvz.org>
---
 man7/namespaces.7 | 43 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 43 insertions(+)

Comments

Andrey Vagin July 22, 2016, 6:25 p.m. UTC | #1
On Thu, Jul 21, 2016 at 11:48 PM, Michael Kerrisk (man-pages)
<mtk.manpages@gmail.com> wrote:
> Hi Andrey,
>
>
> On 07/21/2016 11:06 PM, Andrew Vagin wrote:
>>
>> On Thu, Jul 21, 2016 at 04:41:12PM +0200, Michael Kerrisk (man-pages)
>> wrote:
>>>
>>> Hi Andrey,
>>>
>>> On 07/14/2016 08:20 PM, Andrey Vagin wrote:
>>
>>
>> <snip>
>>
>>>
>>> Could you add here an of the API in detail: what do these FDs refer to,
>>> and how do you use them to solve the use case? And could you you add
>>> that info to the commit messages please.
>>
>>
>> Hi Michael,
>>
>> A patch for man-pages is attached. It adds the following text to
>> namespaces(7).
>>
>> Since  Linux 4.X, the following ioctl(2) calls are supported for names‐
>> pace file descriptors.  The correct syntax is:
>>
>>       fd = ioctl(ns_fd, ioctl_type);
>>
>> where ioctl_type is one of the following:
>>
>> NS_GET_USERNS
>>       Returns a file descriptor that refers to an owning  user  names‐
>>       pace.
>>
>> NS_GET_PARENT
>>       Returns  a  file  descriptor  that refers to a parent namespace.
>>       This ioctl(2) can be used for pid and user namespaces. For  user
>>       namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
>>       ing.
>>
>> In addition to generic ioctl(2) errors, the following specific ones can
>> occur:
>>
>> EINVAL NS_GET_PARENT was called for a nonhierarchical namespace.
>>
>> EPERM  The  requested  namespace  is  outside  of the current namespace
>>       scope.
>>
>> ENOENT ns_fd refers to the init namespace.
>
>
> Thanks for this. But still part of the question remains unanswered.
> How do we (in user-space) use the file descriptors to answer any of
> the questions that this patch series was designed to solve? (This
> info should be in the commit message and the man-pages patch.)

I'm sorry, but I am not sure that I understand what you ask.

Here are the origin questions:
Someone else then asked me a question that led me to wonder about
generally introspecting on the parental relationships between user
namespaces and the association of other namespaces types with user
namespaces. One use would be visualization, in order to understand the
running system. Another would be to answer the question I already
mentioned: what capability does process X have to perform operations
on a resource governed by namespace Y?

Here is an example which shows how we can get the owning namespace
inode number by using these ioctl-s.

$ ls -l /proc/13929/ns/pid
lrwxrwxrwx 1 root root 0 Jul 22 21:03 /proc/13929/ns/pid -> 'pid:[4026532228]'

$ ./nsowner /proc/13929/ns/pid
user:[4026532227]

The owning user namespace for pid:[4026532228] is user:[4026532227].

The nsowner  tool is cimpiled from this code:

int main(int argc, char *argv[])
{
        char buf[128], path[] = "/proc/self/fd/0123456789";
        int ns, uns, ret;

        ns = open(argv[1], O_RDONLY);
        if (ns < 0)
                return 1;

        uns = ioctl(ns, NS_GET_USERNS);
        if (uns < 0)
                return 1;

        snprintf(path, sizeof(path), "/proc/self/fd/%d", uns);
        ret = readlink(path, buf, sizeof(buf) - 1);
        if (ret < 0)
                return 1;
        buf[ret] = 0;

        printf("%s\n", buf);

        return 0;
}

Does this example answer to the origin question? If it isn't, could
you eloborate what you expect to see here.

And I wrote one more example which show all relationships between
namespaces. It enumirates all processes in a system, collects all
namespaces and determins parent and owning namespaces for each of
them, then it constructs a namespace tree and shows it.

Here is a code: https://gist.github.com/avagin/db805f95e15ffb0af7e559dbb8de4418

Here is an example of output for my test system:
[root@fc24 nsfs]# ./nstree
user:[4026531837]
 \__  mnt:[4026532203]
 \__  ipc:[4026531839]
 \__  user:[4026532224]
     \__  user:[4026532226]
         \__  user:[4026532227]
             \__  pid:[4026532228]
     \__  pid:[4026532225]
         \__  pid:[4026532228]
 \__  user:[4026532221]
     \__  pid:[4026532222]
     \__  user:[4026532223]
 \__  mnt:[4026532211]
 \__  uts:[4026531838]
 \__  cgroup:[4026531835]
 \__  pid:[4026531836]
     \__  pid:[4026532225]
         \__  pid:[4026532228]
     \__  pid:[4026532222]
 \__  mnt:[4026531857]
 \__  mnt:[4026531840]
 \__  net:[4026531957]

Thanks,
Andrew

>
> Thanks,
>
> Michael
>
>
>>>> [1] https://lkml.org/lkml/2016/7/6/158
>>>> [2] https://lkml.org/lkml/2016/7/9/101
>>>>
>>>> Cc: "Eric W. Biederman" <ebiederm@xmission.com>
>>>> Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
>>>> Cc: "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com>
>>>> Cc: "W. Trevor King" <wking@tremily.us>
>>>> Cc: Alexander Viro <viro@zeniv.linux.org.uk>
>>>> Cc: Serge Hallyn <serge.hallyn@canonical.com>
>>>>
>>>> --
>>>> 2.5.5
>>>>
>>>>
>>>
>>>
>>> --
>>> Michael Kerrisk
>>> Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
>>> Linux/UNIX System Programming Training: http://man7.org/training/
>
>
>
> --
> Michael Kerrisk
> Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
> Linux/UNIX System Programming Training: http://man7.org/training/
> _______________________________________________
> Containers mailing list
> Containers@lists.linux-foundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/containers
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 25, 2016, 11:47 a.m. UTC | #2
Hi Andrey,

On 07/22/2016 08:25 PM, Andrey Vagin wrote:
> On Thu, Jul 21, 2016 at 11:48 PM, Michael Kerrisk (man-pages)
> <mtk.manpages@gmail.com> wrote:
>> Hi Andrey,
>>
>>
>> On 07/21/2016 11:06 PM, Andrew Vagin wrote:
>>>
>>> On Thu, Jul 21, 2016 at 04:41:12PM +0200, Michael Kerrisk (man-pages)
>>> wrote:
>>>>
>>>> Hi Andrey,
>>>>
>>>> On 07/14/2016 08:20 PM, Andrey Vagin wrote:
>>>
>>>
>>> <snip>
>>>
>>>>
>>>> Could you add here an of the API in detail: what do these FDs refer to,
>>>> and how do you use them to solve the use case? And could you you add
>>>> that info to the commit messages please.
>>>
>>>
>>> Hi Michael,
>>>
>>> A patch for man-pages is attached. It adds the following text to
>>> namespaces(7).
>>>
>>> Since  Linux 4.X, the following ioctl(2) calls are supported for names‐
>>> pace file descriptors.  The correct syntax is:
>>>
>>>       fd = ioctl(ns_fd, ioctl_type);
>>>
>>> where ioctl_type is one of the following:
>>>
>>> NS_GET_USERNS
>>>       Returns a file descriptor that refers to an owning  user  names‐
>>>       pace.
>>>
>>> NS_GET_PARENT
>>>       Returns  a  file  descriptor  that refers to a parent namespace.
>>>       This ioctl(2) can be used for pid and user namespaces. For  user
>>>       namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
>>>       ing.

For each of the above, I think it is worth mentioning that the
close-on-exec flag is set for the returned file descriptor.

>>>
>>> In addition to generic ioctl(2) errors, the following specific ones can
>>> occur:
>>>
>>> EINVAL NS_GET_PARENT was called for a nonhierarchical namespace.
>>>
>>> EPERM  The  requested  namespace  is  outside  of the current namespace
>>>       scope.

Perhaps add "and the caller does not have CAP_SYS_ADMIN" in the initial
user namespace"?

>>>
>>> ENOENT ns_fd refers to the init namespace.
>>
>>
>> Thanks for this. But still part of the question remains unanswered.
>> How do we (in user-space) use the file descriptors to answer any of
>> the questions that this patch series was designed to solve? (This
>> info should be in the commit message and the man-pages patch.)
>
> I'm sorry, but I am not sure that I understand what you ask.
>
> Here are the origin questions:
> Someone else then asked me a question that led me to wonder about
> generally introspecting on the parental relationships between user
> namespaces and the association of other namespaces types with user
> namespaces. One use would be visualization, in order to understand the
> running system. Another would be to answer the question I already
> mentioned: what capability does process X have to perform operations
> on a resource governed by namespace Y?
>
> Here is an example which shows how we can get the owning namespace
> inode number by using these ioctl-s.
>
> $ ls -l /proc/13929/ns/pid
> lrwxrwxrwx 1 root root 0 Jul 22 21:03 /proc/13929/ns/pid -> 'pid:[4026532228]'
>
> $ ./nsowner /proc/13929/ns/pid
> user:[4026532227]
>
> The owning user namespace for pid:[4026532228] is user:[4026532227].
>
> The nsowner  tool is cimpiled from this code:
>
> int main(int argc, char *argv[])
> {
>         char buf[128], path[] = "/proc/self/fd/0123456789";
>         int ns, uns, ret;
>
>         ns = open(argv[1], O_RDONLY);
>         if (ns < 0)
>                 return 1;
>
>         uns = ioctl(ns, NS_GET_USERNS);
>         if (uns < 0)
>                 return 1;
>
>         snprintf(path, sizeof(path), "/proc/self/fd/%d", uns);
>         ret = readlink(path, buf, sizeof(buf) - 1);
>         if (ret < 0)
>                 return 1;
>         buf[ret] = 0;
>
>         printf("%s\n", buf);
>
>         return 0;
> }

So, from my point of view, the important piece that was missing from
your commit message was the note to use readlink("/proc/self/fd/%d")
on the returned FDs. I think that detail needs to be part of the
commit message (and also the man page text). I think it even be
helpful to include the above program as part of the commit message:
it helps people more quickly grasp the API.

> Does this example answer to the origin question?

Yes.

>If it isn't, could
> you eloborate what you expect to see here.
>
> And I wrote one more example which show all relationships between
> namespaces. It enumirates all processes in a system, collects all
> namespaces and determins parent and owning namespaces for each of
> them, then it constructs a namespace tree and shows it.
>
> Here is a code: https://gist.github.com/avagin/db805f95e15ffb0af7e559dbb8de4418

That's great! Thanks!
  
> Here is an example of output for my test system:
> [root@fc24 nsfs]# ./nstree
> user:[4026531837]
>  \__  mnt:[4026532203]
>  \__  ipc:[4026531839]
>  \__  user:[4026532224]
>      \__  user:[4026532226]
>          \__  user:[4026532227]
>              \__  pid:[4026532228]
>      \__  pid:[4026532225]
>          \__  pid:[4026532228]
>  \__  user:[4026532221]
>      \__  pid:[4026532222]
>      \__  user:[4026532223]
>  \__  mnt:[4026532211]
>  \__  uts:[4026531838]
>  \__  cgroup:[4026531835]
>  \__  pid:[4026531836]
>      \__  pid:[4026532225]
>          \__  pid:[4026532228]
>      \__  pid:[4026532222]
>  \__  mnt:[4026531857]
>  \__  mnt:[4026531840]
>  \__  net:[4026531957]

Cheers,

Michael

>>>>> [1] https://lkml.org/lkml/2016/7/6/158
>>>>> [2] https://lkml.org/lkml/2016/7/9/101
Eric W. Biederman July 25, 2016, 1:18 p.m. UTC | #3
"Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:

> Hi Andrey,
>
> On 07/22/2016 08:25 PM, Andrey Vagin wrote:
>> On Thu, Jul 21, 2016 at 11:48 PM, Michael Kerrisk (man-pages)
>> <mtk.manpages@gmail.com> wrote:
>>> Hi Andrey,
>>>
>>>
>>> On 07/21/2016 11:06 PM, Andrew Vagin wrote:
>>>>
>>>> On Thu, Jul 21, 2016 at 04:41:12PM +0200, Michael Kerrisk (man-pages)
>>>> wrote:
>>>>>
>>>>> Hi Andrey,
>>>>>
>>>>> On 07/14/2016 08:20 PM, Andrey Vagin wrote:
>>>>
>>>>
>>>> <snip>
>>>>
>>>>>
>>>>> Could you add here an of the API in detail: what do these FDs refer to,
>>>>> and how do you use them to solve the use case? And could you you add
>>>>> that info to the commit messages please.
>>>>
>>>>
>>>> Hi Michael,
>>>>
>>>> A patch for man-pages is attached. It adds the following text to
>>>> namespaces(7).
>>>>
>>>> Since  Linux 4.X, the following ioctl(2) calls are supported for names‐
>>>> pace file descriptors.  The correct syntax is:
>>>>
>>>>       fd = ioctl(ns_fd, ioctl_type);
>>>>
>>>> where ioctl_type is one of the following:
>>>>
>>>> NS_GET_USERNS
>>>>       Returns a file descriptor that refers to an owning  user  names‐
>>>>       pace.
>>>>
>>>> NS_GET_PARENT
>>>>       Returns  a  file  descriptor  that refers to a parent namespace.
>>>>       This ioctl(2) can be used for pid and user namespaces. For  user
>>>>       namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
>>>>       ing.
>
> For each of the above, I think it is worth mentioning that the
> close-on-exec flag is set for the returned file descriptor.

Hmm.  That is an odd default.

>>>>
>>>> In addition to generic ioctl(2) errors, the following specific ones can
>>>> occur:
>>>>
>>>> EINVAL NS_GET_PARENT was called for a nonhierarchical namespace.
>>>>
>>>> EPERM  The  requested  namespace  is  outside  of the current namespace
>>>>       scope.
>
> Perhaps add "and the caller does not have CAP_SYS_ADMIN" in the initial
> user namespace"?

Having looked at that bit of code I don't think capabilities really
have a role to play.

>>>> ENOENT ns_fd refers to the init namespace.
>>>
>>>
>>> Thanks for this. But still part of the question remains unanswered.
>>> How do we (in user-space) use the file descriptors to answer any of
>>> the questions that this patch series was designed to solve? (This
>>> info should be in the commit message and the man-pages patch.)
>>
>> I'm sorry, but I am not sure that I understand what you ask.
>>
>> Here are the origin questions:
>> Someone else then asked me a question that led me to wonder about
>> generally introspecting on the parental relationships between user
>> namespaces and the association of other namespaces types with user
>> namespaces. One use would be visualization, in order to understand the
>> running system. Another would be to answer the question I already
>> mentioned: what capability does process X have to perform operations
>> on a resource governed by namespace Y?
>>
>> Here is an example which shows how we can get the owning namespace
>> inode number by using these ioctl-s.
>>
>> $ ls -l /proc/13929/ns/pid
>> lrwxrwxrwx 1 root root 0 Jul 22 21:03 /proc/13929/ns/pid -> 'pid:[4026532228]'
>>
>> $ ./nsowner /proc/13929/ns/pid
>> user:[4026532227]
>>
>> The owning user namespace for pid:[4026532228] is user:[4026532227].
>>
>> The nsowner  tool is cimpiled from this code:
>>
>> int main(int argc, char *argv[])
>> {
>>         char buf[128], path[] = "/proc/self/fd/0123456789";
>>         int ns, uns, ret;
>>
>>         ns = open(argv[1], O_RDONLY);
>>         if (ns < 0)
>>                 return 1;
>>
>>         uns = ioctl(ns, NS_GET_USERNS);
>>         if (uns < 0)
>>                 return 1;
>>
>>         snprintf(path, sizeof(path), "/proc/self/fd/%d", uns);
>>         ret = readlink(path, buf, sizeof(buf) - 1);
>>         if (ret < 0)
>>                 return 1;
>>         buf[ret] = 0;
>>
>>         printf("%s\n", buf);
>>
>>         return 0;
>> }
>
> So, from my point of view, the important piece that was missing from
> your commit message was the note to use readlink("/proc/self/fd/%d")
> on the returned FDs. I think that detail needs to be part of the
> commit message (and also the man page text). I think it even be
> helpful to include the above program as part of the commit message:
> it helps people more quickly grasp the API.

Please, please make the standard way to compare these things fstat.
That is much less magic than a symlink, and a little more future proof.
Possibly even kcmp.

At some point we will care about migrating a migrating sub-container and we
may have to have some minor changes.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 25, 2016, 2:46 p.m. UTC | #4
Hi Eric,

On 07/25/2016 03:18 PM, Eric W. Biederman wrote:
> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>
>> Hi Andrey,
>>
>> On 07/22/2016 08:25 PM, Andrey Vagin wrote:
>>> On Thu, Jul 21, 2016 at 11:48 PM, Michael Kerrisk (man-pages)
>>> <mtk.manpages@gmail.com> wrote:
>>>> Hi Andrey,
>>>>
>>>>
>>>> On 07/21/2016 11:06 PM, Andrew Vagin wrote:
>>>>>
>>>>> On Thu, Jul 21, 2016 at 04:41:12PM +0200, Michael Kerrisk (man-pages)
>>>>> wrote:
>>>>>>
>>>>>> Hi Andrey,
>>>>>>
>>>>>> On 07/14/2016 08:20 PM, Andrey Vagin wrote:
>>>>>
>>>>>
>>>>> <snip>
>>>>>
>>>>>>
>>>>>> Could you add here an of the API in detail: what do these FDs refer to,
>>>>>> and how do you use them to solve the use case? And could you you add
>>>>>> that info to the commit messages please.
>>>>>
>>>>>
>>>>> Hi Michael,
>>>>>
>>>>> A patch for man-pages is attached. It adds the following text to
>>>>> namespaces(7).
>>>>>
>>>>> Since  Linux 4.X, the following ioctl(2) calls are supported for names‐
>>>>> pace file descriptors.  The correct syntax is:
>>>>>
>>>>>       fd = ioctl(ns_fd, ioctl_type);
>>>>>
>>>>> where ioctl_type is one of the following:
>>>>>
>>>>> NS_GET_USERNS
>>>>>       Returns a file descriptor that refers to an owning  user  names‐
>>>>>       pace.
>>>>>
>>>>> NS_GET_PARENT
>>>>>       Returns  a  file  descriptor  that refers to a parent namespace.
>>>>>       This ioctl(2) can be used for pid and user namespaces. For  user
>>>>>       namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
>>>>>       ing.
>>
>> For each of the above, I think it is worth mentioning that the
>> close-on-exec flag is set for the returned file descriptor.
>
> Hmm.  That is an odd default.

Why do you say that? It's pretty common as the default for various
APIs that create new FDs these days. (There's of course a strong argument
that the original UNIX default was a design blunder...)

>>>>>
>>>>> In addition to generic ioctl(2) errors, the following specific ones can
>>>>> occur:
>>>>>
>>>>> EINVAL NS_GET_PARENT was called for a nonhierarchical namespace.
>>>>>
>>>>> EPERM  The  requested  namespace  is  outside  of the current namespace
>>>>>       scope.
>>
>> Perhaps add "and the caller does not have CAP_SYS_ADMIN" in the initial
>> user namespace"?
>
> Having looked at that bit of code I don't think capabilities really
> have a role to play.

Yes, I caught up with that now. I await to see how this plays out
in the next patch version.

>>>>> ENOENT ns_fd refers to the init namespace.
>>>>
>>>>
>>>> Thanks for this. But still part of the question remains unanswered.
>>>> How do we (in user-space) use the file descriptors to answer any of
>>>> the questions that this patch series was designed to solve? (This
>>>> info should be in the commit message and the man-pages patch.)
>>>
>>> I'm sorry, but I am not sure that I understand what you ask.
>>>
>>> Here are the origin questions:
>>> Someone else then asked me a question that led me to wonder about
>>> generally introspecting on the parental relationships between user
>>> namespaces and the association of other namespaces types with user
>>> namespaces. One use would be visualization, in order to understand the
>>> running system. Another would be to answer the question I already
>>> mentioned: what capability does process X have to perform operations
>>> on a resource governed by namespace Y?
>>>
>>> Here is an example which shows how we can get the owning namespace
>>> inode number by using these ioctl-s.
>>>
>>> $ ls -l /proc/13929/ns/pid
>>> lrwxrwxrwx 1 root root 0 Jul 22 21:03 /proc/13929/ns/pid -> 'pid:[4026532228]'
>>>
>>> $ ./nsowner /proc/13929/ns/pid
>>> user:[4026532227]
>>>
>>> The owning user namespace for pid:[4026532228] is user:[4026532227].
>>>
>>> The nsowner  tool is cimpiled from this code:
>>>
>>> int main(int argc, char *argv[])
>>> {
>>>         char buf[128], path[] = "/proc/self/fd/0123456789";
>>>         int ns, uns, ret;
>>>
>>>         ns = open(argv[1], O_RDONLY);
>>>         if (ns < 0)
>>>                 return 1;
>>>
>>>         uns = ioctl(ns, NS_GET_USERNS);
>>>         if (uns < 0)
>>>                 return 1;
>>>
>>>         snprintf(path, sizeof(path), "/proc/self/fd/%d", uns);
>>>         ret = readlink(path, buf, sizeof(buf) - 1);
>>>         if (ret < 0)
>>>                 return 1;
>>>         buf[ret] = 0;
>>>
>>>         printf("%s\n", buf);
>>>
>>>         return 0;
>>> }
>>
>> So, from my point of view, the important piece that was missing from
>> your commit message was the note to use readlink("/proc/self/fd/%d")
>> on the returned FDs. I think that detail needs to be part of the
>> commit message (and also the man page text). I think it even be
>> helpful to include the above program as part of the commit message:
>> it helps people more quickly grasp the API.
>
> Please, please make the standard way to compare these things fstat.
> That is much less magic than a symlink, and a little more future proof.
> Possibly even kcmp.

As in fstat() to get the st_ino field, right?

Cheers,

Michael

> At some point we will care about migrating a migrating sub-container and we
> may have to have some minor changes.
>
> Eric
>
Serge E. Hallyn July 25, 2016, 2:54 p.m. UTC | #5
Quoting Michael Kerrisk (man-pages) (mtk.manpages@gmail.com):
> Hi Eric,
> 
> On 07/25/2016 03:18 PM, Eric W. Biederman wrote:
> >"Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
> >
> >>Hi Andrey,
> >>
> >>On 07/22/2016 08:25 PM, Andrey Vagin wrote:
> >>>On Thu, Jul 21, 2016 at 11:48 PM, Michael Kerrisk (man-pages)
> >>><mtk.manpages@gmail.com> wrote:
> >>>>Hi Andrey,
> >>>>
> >>>>
> >>>>On 07/21/2016 11:06 PM, Andrew Vagin wrote:
> >>>>>
> >>>>>On Thu, Jul 21, 2016 at 04:41:12PM +0200, Michael Kerrisk (man-pages)
> >>>>>wrote:
> >>>>>>
> >>>>>>Hi Andrey,
> >>>>>>
> >>>>>>On 07/14/2016 08:20 PM, Andrey Vagin wrote:
> >>>>>
> >>>>>
> >>>>><snip>
> >>>>>
> >>>>>>
> >>>>>>Could you add here an of the API in detail: what do these FDs refer to,
> >>>>>>and how do you use them to solve the use case? And could you you add
> >>>>>>that info to the commit messages please.
> >>>>>
> >>>>>
> >>>>>Hi Michael,
> >>>>>
> >>>>>A patch for man-pages is attached. It adds the following text to
> >>>>>namespaces(7).
> >>>>>
> >>>>>Since  Linux 4.X, the following ioctl(2) calls are supported for names‐
> >>>>>pace file descriptors.  The correct syntax is:
> >>>>>
> >>>>>      fd = ioctl(ns_fd, ioctl_type);
> >>>>>
> >>>>>where ioctl_type is one of the following:
> >>>>>
> >>>>>NS_GET_USERNS
> >>>>>      Returns a file descriptor that refers to an owning  user  names‐
> >>>>>      pace.
> >>>>>
> >>>>>NS_GET_PARENT
> >>>>>      Returns  a  file  descriptor  that refers to a parent namespace.
> >>>>>      This ioctl(2) can be used for pid and user namespaces. For  user
> >>>>>      namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
> >>>>>      ing.
> >>
> >>For each of the above, I think it is worth mentioning that the
> >>close-on-exec flag is set for the returned file descriptor.
> >
> >Hmm.  That is an odd default.
> 
> Why do you say that? It's pretty common as the default for various
> APIs that create new FDs these days. (There's of course a strong argument
> that the original UNIX default was a design blunder...)
> 
> >>>>>
> >>>>>In addition to generic ioctl(2) errors, the following specific ones can
> >>>>>occur:
> >>>>>
> >>>>>EINVAL NS_GET_PARENT was called for a nonhierarchical namespace.
> >>>>>
> >>>>>EPERM  The  requested  namespace  is  outside  of the current namespace
> >>>>>      scope.
> >>
> >>Perhaps add "and the caller does not have CAP_SYS_ADMIN" in the initial
> >>user namespace"?
> >
> >Having looked at that bit of code I don't think capabilities really
> >have a role to play.
> 
> Yes, I caught up with that now. I await to see how this plays out
> in the next patch version.

Thanks - that had caught my eye but I hadn't had time to look into the
justification for this.  Hiding this kind of thing indeed seems wrong to
me, unless there is a really good justification for it, i.e. a way
to use that info in an exploit.

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Eric W. Biederman July 25, 2016, 2:59 p.m. UTC | #6
"Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:

> Hi Eric,
>
> On 07/25/2016 03:18 PM, Eric W. Biederman wrote:
>> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>>
>>> Hi Andrey,
>>>
>>> On 07/22/2016 08:25 PM, Andrey Vagin wrote:
>>>> On Thu, Jul 21, 2016 at 11:48 PM, Michael Kerrisk (man-pages)
>>>> <mtk.manpages@gmail.com> wrote:
>>>>> Hi Andrey,
>>>>>
>>>>>
>>>>> On 07/21/2016 11:06 PM, Andrew Vagin wrote:
>>>>>>
[snip]
>>>>>> where ioctl_type is one of the following:
>>>>>>
>>>>>> NS_GET_USERNS
>>>>>>       Returns a file descriptor that refers to an owning  user  names‐
>>>>>>       pace.
>>>>>>
>>>>>> NS_GET_PARENT
>>>>>>       Returns  a  file  descriptor  that refers to a parent namespace.
>>>>>>       This ioctl(2) can be used for pid and user namespaces. For  user
>>>>>>       namespaces,  NS_GET_PARENT and NS_GET_USERNS have the same mean‐
>>>>>>       ing.
>>>
>>> For each of the above, I think it is worth mentioning that the
>>> close-on-exec flag is set for the returned file descriptor.
>>
>> Hmm.  That is an odd default.
>
> Why do you say that? It's pretty common as the default for various
> APIs that create new FDs these days. (There's of course a strong argument
> that the original UNIX default was a design blunder...)

Interesting.  I haven't kept up on that, but it seems reasonable.

[snip]
>>> So, from my point of view, the important piece that was missing from
>>> your commit message was the note to use readlink("/proc/self/fd/%d")
>>> on the returned FDs. I think that detail needs to be part of the
>>> commit message (and also the man page text). I think it even be
>>> helpful to include the above program as part of the commit message:
>>> it helps people more quickly grasp the API.
>>
>> Please, please make the standard way to compare these things fstat.
>> That is much less magic than a symlink, and a little more future proof.
>> Possibly even kcmp.
>
> As in fstat() to get the st_ino field, right?

Both the st_ino and st_dev fields.

The most likely change to support checkpoint/restart in the future is to
preserve st_ino across migrations and instantiate a different instance
of nsfs to hold the inode numbers from the previous machine.

We would need to handle the preservation carefully or else there is
a chance that two namespace file descriptors (collected from different
sources) with different st_dev and st_ino fields may actuall refer to
the same object.

Which is a long way of saying we have the st_dev field please use it,
it may matter at some point.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Eric W. Biederman July 25, 2016, 3:17 p.m. UTC | #7
"Serge E. Hallyn" <serge@hallyn.com> writes:

> Quoting Michael Kerrisk (man-pages) (mtk.manpages@gmail.com):
>> Hi Eric,
>> 
>> On 07/25/2016 03:18 PM, Eric W. Biederman wrote:
>> >"Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>> >
>> >>Hi Andrey,
>> >>
>> >>On 07/22/2016 08:25 PM, Andrey Vagin wrote:
>> >>Perhaps add "and the caller does not have CAP_SYS_ADMIN" in the initial
>> >>user namespace"?
>> >
>> >Having looked at that bit of code I don't think capabilities really
>> >have a role to play.
>> 
>> Yes, I caught up with that now. I await to see how this plays out
>> in the next patch version.
>
> Thanks - that had caught my eye but I hadn't had time to look into the
> justification for this.  Hiding this kind of thing indeed seems wrong to
> me, unless there is a really good justification for it, i.e. a way
> to use that info in an exploit.

To avoid breaking checkpoint/restart we need to limit information to the
namespaces the caller is a member of for the user and pid namespaces.

This roughly duplicates the parentage checks in ns_capable.

Conceptually this is the same as limiting .. in a chroot environment.

Eric

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Andrey Vagin July 26, 2016, 2:54 a.m. UTC | #8
On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:

[snip]

> [snip]
> >>> So, from my point of view, the important piece that was missing from
> >>> your commit message was the note to use readlink("/proc/self/fd/%d")
> >>> on the returned FDs. I think that detail needs to be part of the
> >>> commit message (and also the man page text). I think it even be
> >>> helpful to include the above program as part of the commit message:
> >>> it helps people more quickly grasp the API.
> >>
> >> Please, please make the standard way to compare these things fstat.
> >> That is much less magic than a symlink, and a little more future proof.
> >> Possibly even kcmp.

I like the idea to use kcmp to compare namespaces. I am going to add this
functionality to kcmp and describe all these in the man page.

> >
> > As in fstat() to get the st_ino field, right?
> 
> Both the st_ino and st_dev fields.
> 
> The most likely change to support checkpoint/restart in the future is to
> preserve st_ino across migrations and instantiate a different instance
> of nsfs to hold the inode numbers from the previous machine.

It sounds tricky. BTW: Actually this is not only one places where we have
this sort of problem. For example, now mount id-s are not preserved when
a container is migrated. The same problem is applied to tmpfs, where
inode numbers are not preserved for files. 

> 
> We would need to handle the preservation carefully or else there is
> a chance that two namespace file descriptors (collected from different
> sources) with different st_dev and st_ino fields may actuall refer to
> the same object.
> 
> Which is a long way of saying we have the st_dev field please use it,
> it may matter at some point.
> 
> Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 26, 2016, 8:03 a.m. UTC | #9
On 07/26/2016 04:54 AM, Andrew Vagin wrote:
> On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
>> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>
> [snip]
>
>> [snip]
>>>>> So, from my point of view, the important piece that was missing from
>>>>> your commit message was the note to use readlink("/proc/self/fd/%d")
>>>>> on the returned FDs. I think that detail needs to be part of the
>>>>> commit message (and also the man page text). I think it even be
>>>>> helpful to include the above program as part of the commit message:
>>>>> it helps people more quickly grasp the API.
>>>>
>>>> Please, please make the standard way to compare these things fstat.
>>>> That is much less magic than a symlink, and a little more future proof.
>>>> Possibly even kcmp.
>
> I like the idea to use kcmp to compare namespaces. I am going to add this
> functionality to kcmp and describe all these in the man page.

Hi Andrey,

Can you briefly sketch out the proposed API and how it would be used?
I'd find it useful to see that even before the implementation.

Cheers,

Michael
Andrey Vagin July 26, 2016, 6:25 p.m. UTC | #10
On Tue, Jul 26, 2016 at 10:03:25AM +0200, Michael Kerrisk (man-pages) wrote:
> On 07/26/2016 04:54 AM, Andrew Vagin wrote:
> > On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
> > > "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
> > 
> > [snip]
> > 
> > > [snip]
> > > > > > So, from my point of view, the important piece that was missing from
> > > > > > your commit message was the note to use readlink("/proc/self/fd/%d")
> > > > > > on the returned FDs. I think that detail needs to be part of the
> > > > > > commit message (and also the man page text). I think it even be
> > > > > > helpful to include the above program as part of the commit message:
> > > > > > it helps people more quickly grasp the API.
> > > > > 
> > > > > Please, please make the standard way to compare these things fstat.
> > > > > That is much less magic than a symlink, and a little more future proof.
> > > > > Possibly even kcmp.
> > 
> > I like the idea to use kcmp to compare namespaces. I am going to add this
> > functionality to kcmp and describe all these in the man page.
> 
> Hi Andrey,
> 
> Can you briefly sketch out the proposed API and how it would be used?
> I'd find it useful to see that even before the implementation.

Sure. If a process wants to compare two namespaces, it needs to get file
descriptors for them (open /proc/PID/ns/XXX, use new ioctl-s, find a
process which has them),
and then it calls kcmp(pid1, pid2, KCMP_NSFD, ns_fd1, ns_fd2)

For example, if we want to compare pid namespaces for 1 and 2 processes:

pid = getpid();
ns_fd1 = open("/proc/1/ns/pid")
ns_fd2 = open("/proc/2/ns/pid")

if (!kcmp(pid, pid, KCMP_NSFD, ns_fd1, ns_fd2))
	printf("Both processes live in the same pid namespace\n");

Thanks,
Andrew
> 
> Cheers,
> 
> Michael
> 
> 
> -- 
> Michael Kerrisk
> Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/
> Linux/UNIX System Programming Training: http://man7.org/training/
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
W. Trevor King July 26, 2016, 6:32 p.m. UTC | #11
On Tue, Jul 26, 2016 at 11:25:24AM -0700, Andrew Vagin wrote:
> Sure. If a process wants to compare two namespaces, it needs to get file
> descriptors for them (open /proc/PID/ns/XXX, use new ioctl-s, find a
> process which has them),
> and then it calls kcmp(pid1, pid2, KCMP_NSFD, ns_fd1, ns_fd2)

If you use the new ioctl-s to get ns_fd2, do you walk your local /proc
to find pid2?

Cheers,
Trevor
Andrey Vagin July 26, 2016, 7:11 p.m. UTC | #12
On Tue, Jul 26, 2016 at 11:32:25AM -0700, W. Trevor King wrote:
> On Tue, Jul 26, 2016 at 11:25:24AM -0700, Andrew Vagin wrote:
> > Sure. If a process wants to compare two namespaces, it needs to get file
> > descriptors for them (open /proc/PID/ns/XXX, use new ioctl-s, find a
> > process which has them),
> > and then it calls kcmp(pid1, pid2, KCMP_NSFD, ns_fd1, ns_fd2)
> 
> If you use the new ioctl-s to get ns_fd2, do you walk your local /proc
> to find pid2?

If you use the new ioctl-s to get nf_fd2, you will have it in the
current process, so pid2 will be getpid().

pidX identifies a process where to find fdX.

man 2 kcmp:
 The kcmp() system call can be used to check whether the  two processes
 identified  by  pid1  and  pid2 share a kernel resource such as virtual
 memory, file descriptors, and so on.

> 
> Cheers,
> Trevor
> 
> -- 
> This email may be signed or encrypted with GnuPG (http://www.gnupg.org).
> For more information, see http://en.wikipedia.org/wiki/Pretty_Good_Privacy


--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 26, 2016, 7:17 p.m. UTC | #13
Hello Andrew,

On 26 July 2016 at 20:25, Andrew Vagin <avagin@virtuozzo.com> wrote:
> On Tue, Jul 26, 2016 at 10:03:25AM +0200, Michael Kerrisk (man-pages) wrote:
>> On 07/26/2016 04:54 AM, Andrew Vagin wrote:
>> > On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
>> > > "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>> >
>> > [snip]
>> >
>> > > [snip]
>> > > > > > So, from my point of view, the important piece that was missing from
>> > > > > > your commit message was the note to use readlink("/proc/self/fd/%d")
>> > > > > > on the returned FDs. I think that detail needs to be part of the
>> > > > > > commit message (and also the man page text). I think it even be
>> > > > > > helpful to include the above program as part of the commit message:
>> > > > > > it helps people more quickly grasp the API.
>> > > > >
>> > > > > Please, please make the standard way to compare these things fstat.
>> > > > > That is much less magic than a symlink, and a little more future proof.
>> > > > > Possibly even kcmp.
>> >
>> > I like the idea to use kcmp to compare namespaces. I am going to add this
>> > functionality to kcmp and describe all these in the man page.
>>
>> Hi Andrey,
>>
>> Can you briefly sketch out the proposed API and how it would be used?
>> I'd find it useful to see that even before the implementation.
>
> Sure. If a process wants to compare two namespaces, it needs to get file
> descriptors for them (open /proc/PID/ns/XXX, use new ioctl-s, find a
> process which has them),
> and then it calls kcmp(pid1, pid2, KCMP_NSFD, ns_fd1, ns_fd2)
>
> For example, if we want to compare pid namespaces for 1 and 2 processes:
>

What's the purpose of the following line, and the use of 'pid' in the
kcmp() call?:

> pid = getpid();
> ns_fd1 = open("/proc/1/ns/pid")
> ns_fd2 = open("/proc/2/ns/pid")
>
> if (!kcmp(pid, pid, KCMP_NSFD, ns_fd1, ns_fd2))
>         printf("Both processes live in the same pid namespace\n");


Thanks,

Michael
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Eric W. Biederman July 26, 2016, 7:38 p.m. UTC | #14
Andrew Vagin <avagin@virtuozzo.com> writes:

> On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
>> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>
> [snip]
>
>> [snip]
>> >>> So, from my point of view, the important piece that was missing from
>> >>> your commit message was the note to use readlink("/proc/self/fd/%d")
>> >>> on the returned FDs. I think that detail needs to be part of the
>> >>> commit message (and also the man page text). I think it even be
>> >>> helpful to include the above program as part of the commit message:
>> >>> it helps people more quickly grasp the API.
>> >>
>> >> Please, please make the standard way to compare these things fstat.
>> >> That is much less magic than a symlink, and a little more future proof.
>> >> Possibly even kcmp.
>
> I like the idea to use kcmp to compare namespaces. I am going to add this
> functionality to kcmp and describe all these in the man page.
>
>> >
>> > As in fstat() to get the st_ino field, right?
>> 
>> Both the st_ino and st_dev fields.
>> 
>> The most likely change to support checkpoint/restart in the future is to
>> preserve st_ino across migrations and instantiate a different instance
>> of nsfs to hold the inode numbers from the previous machine.
>
> It sounds tricky. BTW: Actually this is not only one places where we have
> this sort of problem. For example, now mount id-s are not preserved when
> a container is migrated. The same problem is applied to tmpfs, where
> inode numbers are not preserved for files.

Agreed.

Interesting. Interesting. Interesting.

I am not completely convinced that improving kcmp solves it for
everything but improving kcmp sounds good enough to be very interesting
and enough to solve a practical case (migration in migration).  Plus
improving kcmp is cheap and easy.

I would propose:

KCMP_OBJECT
    Check whether a file descriptor idx1 in the process pid1 refers to
    the same underlying object as file descriptor idx2 in the process
    pid2.

The default case would be checking to see if to file descriptors refer
to the same inode.  But for weird cases (like proc pid directories, or
sysfs files) the comparison could look deeper.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Andrey Vagin July 26, 2016, 8:39 p.m. UTC | #15
On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:
> Hello Andrew,
> 
> On 26 July 2016 at 20:25, Andrew Vagin <avagin@virtuozzo.com> wrote:
> > On Tue, Jul 26, 2016 at 10:03:25AM +0200, Michael Kerrisk (man-pages) wrote:
> >> On 07/26/2016 04:54 AM, Andrew Vagin wrote:
> >> > On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
> >> > > "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
> >> >
> >> > [snip]
> >> >
> >> > > [snip]
> >> > > > > > So, from my point of view, the important piece that was missing from
> >> > > > > > your commit message was the note to use readlink("/proc/self/fd/%d")
> >> > > > > > on the returned FDs. I think that detail needs to be part of the
> >> > > > > > commit message (and also the man page text). I think it even be
> >> > > > > > helpful to include the above program as part of the commit message:
> >> > > > > > it helps people more quickly grasp the API.
> >> > > > >
> >> > > > > Please, please make the standard way to compare these things fstat.
> >> > > > > That is much less magic than a symlink, and a little more future proof.
> >> > > > > Possibly even kcmp.
> >> >
> >> > I like the idea to use kcmp to compare namespaces. I am going to add this
> >> > functionality to kcmp and describe all these in the man page.
> >>
> >> Hi Andrey,
> >>
> >> Can you briefly sketch out the proposed API and how it would be used?
> >> I'd find it useful to see that even before the implementation.
> >
> > Sure. If a process wants to compare two namespaces, it needs to get file
> > descriptors for them (open /proc/PID/ns/XXX, use new ioctl-s, find a
> > process which has them),
> > and then it calls kcmp(pid1, pid2, KCMP_NSFD, ns_fd1, ns_fd2)
> >
> > For example, if we want to compare pid namespaces for 1 and 2 processes:
> >
> 
> What's the purpose of the following line, and the use of 'pid' in the
> kcmp() call?:

It's the existing interface of kcmp. It's used to check whether the
two processes identified  by pid1  and  pid2 share a kernel resource
such as virtual memory, file descriptors, and so on.

If we want to compare two file descriptors of the current process,
it is one of cases for which kcmp can be used. We can call kcmp to
compare two namespaces which are opened in other processes.

Thanks,
Andrew

> 
> > pid = getpid();
> > ns_fd1 = open("/proc/1/ns/pid")
> > ns_fd2 = open("/proc/2/ns/pid")
> >
> > if (!kcmp(pid, pid, KCMP_NSFD, ns_fd1, ns_fd2))
> >         printf("Both processes live in the same pid namespace\n");
> 
> Thanks,
> 
> Michael
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 28, 2016, 10:45 a.m. UTC | #16
On 07/26/2016 10:39 PM, Andrew Vagin wrote:
> On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:
>> Hello Andrew,
>>
>> On 26 July 2016 at 20:25, Andrew Vagin <avagin@virtuozzo.com> wrote:
>>> On Tue, Jul 26, 2016 at 10:03:25AM +0200, Michael Kerrisk (man-pages) wrote:
>>>> On 07/26/2016 04:54 AM, Andrew Vagin wrote:
>>>>> On Mon, Jul 25, 2016 at 09:59:43AM -0500, Eric W. Biederman wrote:
>>>>>> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>>>>>
>>>>> [snip]
>>>>>
>>>>>> [snip]
>>>>>>>>> So, from my point of view, the important piece that was missing from
>>>>>>>>> your commit message was the note to use readlink("/proc/self/fd/%d")
>>>>>>>>> on the returned FDs. I think that detail needs to be part of the
>>>>>>>>> commit message (and also the man page text). I think it even be
>>>>>>>>> helpful to include the above program as part of the commit message:
>>>>>>>>> it helps people more quickly grasp the API.
>>>>>>>>
>>>>>>>> Please, please make the standard way to compare these things fstat.
>>>>>>>> That is much less magic than a symlink, and a little more future proof.
>>>>>>>> Possibly even kcmp.
>>>>>
>>>>> I like the idea to use kcmp to compare namespaces. I am going to add this
>>>>> functionality to kcmp and describe all these in the man page.
>>>>
>>>> Hi Andrey,
>>>>
>>>> Can you briefly sketch out the proposed API and how it would be used?
>>>> I'd find it useful to see that even before the implementation.
>>>
>>> Sure. If a process wants to compare two namespaces, it needs to get file
>>> descriptors for them (open /proc/PID/ns/XXX, use new ioctl-s, find a
>>> process which has them),
>>> and then it calls kcmp(pid1, pid2, KCMP_NSFD, ns_fd1, ns_fd2)
>>>
>>> For example, if we want to compare pid namespaces for 1 and 2 processes:
>>>
>>
>> What's the purpose of the following line, and the use of 'pid' in the
>> kcmp() call?:
>
> It's the existing interface of kcmp.  It's used to check whether the
> two processes identified  by pid1  and  pid2 share a kernel resource
> such as virtual memory, file descriptors, and so on.


Yes, understood, but it seems a slightly weird use of the interface,
since in general pid1 will be the same as pid2 in this use case,
whereas in the other use cases, pid1 and pid2 are generally not
equal.

> If we want to compare two file descriptors of the current process,
> it is one of cases for which kcmp can be used. We can call kcmp to
> compare two namespaces which are opened in other processes.

Is there really a use case there? I assume we're talking about the
scenario where a process in one namespace opens a /proc/PID/ns/*
file descriptor and passes that FD to another process via a UNIX
domain socket. Is that correct?

So, supposing that we want to build a map of the relationships
between namespaces using the proposed kcmp() API, and there are
say N namespaces? Does this mena we make (N * (N-1) / 2) calls
to kcmp()?

Cheers,

Michael

>>> pid = getpid();
>>> ns_fd1 = open("/proc/1/ns/pid")
>>> ns_fd2 = open("/proc/2/ns/pid")
>>>
>>> if (!kcmp(pid, pid, KCMP_NSFD, ns_fd1, ns_fd2))
>>>         printf("Both processes live in the same pid namespace\n");
>>
>> Thanks,
>>
>> Michael
>
Eric W. Biederman July 28, 2016, 12:56 p.m. UTC | #17
"Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:

> On 07/26/2016 10:39 PM, Andrew Vagin wrote:
>> On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:

>> If we want to compare two file descriptors of the current process,
>> it is one of cases for which kcmp can be used. We can call kcmp to
>> compare two namespaces which are opened in other processes.
>
> Is there really a use case there? I assume we're talking about the
> scenario where a process in one namespace opens a /proc/PID/ns/*
> file descriptor and passes that FD to another process via a UNIX
> domain socket. Is that correct?
>
> So, supposing that we want to build a map of the relationships
> between namespaces using the proposed kcmp() API, and there are
> say N namespaces? Does this mena we make (N * (N-1) / 2) calls
> to kcmp()?

Potentially.  The numbers are small enough O(N^2) isn't fatal.

Where kcmp shines is that it allows migration to happen.  Inode numbers
to change (which they very much will today), and still have things work.

We can keep it O(Nlog(N)) by taking advantage of not just the equality
but the ordering relationship.  Although Ugh.  One disadvantage of
kcmp currently is that the way the ordering relationship is defined
the order is not preserved over migration :(

Eric

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 28, 2016, 7 p.m. UTC | #18
Hi Eric,

On 07/28/2016 02:56 PM, Eric W. Biederman wrote:
> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>
>> On 07/26/2016 10:39 PM, Andrew Vagin wrote:
>>> On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:
>
>>> If we want to compare two file descriptors of the current process,
>>> it is one of cases for which kcmp can be used. We can call kcmp to
>>> compare two namespaces which are opened in other processes.
>>
>> Is there really a use case there? I assume we're talking about the
>> scenario where a process in one namespace opens a /proc/PID/ns/*
>> file descriptor and passes that FD to another process via a UNIX
>> domain socket. Is that correct?
>>
>> So, supposing that we want to build a map of the relationships
>> between namespaces using the proposed kcmp() API, and there are
>> say N namespaces? Does this mena we make (N * (N-1) / 2) calls
>> to kcmp()?
>
> Potentially.  The numbers are small enough O(N^2) isn't fatal.

Define "small", please.

O(N^2) makes me nervous about what other use cases lurk out
there that may get bitten by this.

> Where kcmp shines is that it allows migration to happen.  Inode numbers
> to change (which they very much will today), and still have things work.


> We can keep it O(Nlog(N)) by taking advantage of not just the equality
> but the ordering relationship.  Although Ugh.

Yes, that sounds pretty ugly...

>One disadvantage of
> kcmp currently is that the way the ordering relationship is defined
> the order is not preserved over migration :(

So, does kcmp() fully solve the proble(s) at hand? It sounds like
not, if I understand your last point correctly.
Eric W. Biederman July 29, 2016, 6:05 p.m. UTC | #19
"Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:

> Hi Eric,
>
> On 07/28/2016 02:56 PM, Eric W. Biederman wrote:
>> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>>
>>> On 07/26/2016 10:39 PM, Andrew Vagin wrote:
>>>> On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:
>>
>>>> If we want to compare two file descriptors of the current process,
>>>> it is one of cases for which kcmp can be used. We can call kcmp to
>>>> compare two namespaces which are opened in other processes.
>>>
>>> Is there really a use case there? I assume we're talking about the
>>> scenario where a process in one namespace opens a /proc/PID/ns/*
>>> file descriptor and passes that FD to another process via a UNIX
>>> domain socket. Is that correct?
>>>
>>> So, supposing that we want to build a map of the relationships
>>> between namespaces using the proposed kcmp() API, and there are
>>> say N namespaces? Does this mena we make (N * (N-1) / 2) calls
>>> to kcmp()?
>>
>> Potentially.  The numbers are small enough O(N^2) isn't fatal.
>
> Define "small", please.
>
> O(N^2) makes me nervous about what other use cases lurk out
> there that may get bitten by this.

Worst case for N (One namespace per thread) is about 60k.
A typical heavy use case may be 1000 namespaces of any type.
So we are talking about O(N^2) that rarely happens and should be done in
a couple of seconds.

>> Where kcmp shines is that it allows migration to happen.  Inode numbers
>> to change (which they very much will today), and still have things work.
>
>
>> We can keep it O(Nlog(N)) by taking advantage of not just the equality
>> but the ordering relationship.  Although Ugh.
>
> Yes, that sounds pretty ugly...

Actually having thought about this a little more if kcmp returns an
ordering by inode and migration preserves the relative order of
the inodes (which should just be a creation order) it should be quite
solvable.

Switch from an order by inode number to an order by object creation
time, and guarantee that all creations are have an order (which with
task_list_lock we practically already have) and it should be even easier
to create.  (A 64bit nanosecond resolution timestamp is good for 544
years of uptime).  A 64bit number that increments each time an object is
created should have an even better lifespan.

I don't know if we can find a way to give that guarantee for other kcmp
comparisons but it is worth a thought.

>>One disadvantage of
>> kcmp currently is that the way the ordering relationship is defined
>> the order is not preserved over migration :(
>
> So, does kcmp() fully solve the proble(s) at hand? It sounds like
> not, if I understand your last point correctly.

There are 3 possibilities I see for migration in migration, ordered
in order of implementation difficulty.
1) Have a clear signal that migration happened and a nested migration
   needs to restart.
2) Use kcmp so that only the relative order needs to be preserved.
3) Preserve the device number and inode numbers.

At a practical level I think (2) may actually in net be the simplest.
It requires a little more care to implement and you have to opt in,
but it should not require any rolling back of activity (merely careful
ordering of object creation).

I definititely like kcmp knowing how to compare things by inode
(aka st_dev, st_inode) because then even if you have to restart
the comparisons after a migration the exact details you are comparing
are hidden and so it is easier to support and harder to get wrong.

I can imagine how to preserve inode numbers by creating a new instance
of nsfs instance and using the old inode numbers upon restore.  I don't
currently see how we could possibly preserve st_dev over migration short of
a device number namespace.

So if we are going to continue with making device numbers be a legacy
attribute applications should not care about we need a way to compare
things by not looking at st_dev.  Which brings us back to kcmp.

Hmm.  Hotplugging as disk and plugging it back likely will change the
device number and give the same kind of challenge with st_dev (although
you can't keep a file descriptor open across that kind of event).  So
certainly a hotplug event on a device should be enough to say don't care
about the device number.

Eric

--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Michael Kerrisk (man-pages) July 31, 2016, 9:31 p.m. UTC | #20
Hi Eric,

On 07/29/2016 08:05 PM, Eric W. Biederman wrote:
> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>
>> Hi Eric,
>>
>> On 07/28/2016 02:56 PM, Eric W. Biederman wrote:
>>> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
>>>
>>>> On 07/26/2016 10:39 PM, Andrew Vagin wrote:
>>>>> On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:
>>>
>>>>> If we want to compare two file descriptors of the current process,
>>>>> it is one of cases for which kcmp can be used. We can call kcmp to
>>>>> compare two namespaces which are opened in other processes.
>>>>
>>>> Is there really a use case there? I assume we're talking about the
>>>> scenario where a process in one namespace opens a /proc/PID/ns/*
>>>> file descriptor and passes that FD to another process via a UNIX
>>>> domain socket. Is that correct?
>>>>
>>>> So, supposing that we want to build a map of the relationships
>>>> between namespaces using the proposed kcmp() API, and there are
>>>> say N namespaces? Does this mena we make (N * (N-1) / 2) calls
>>>> to kcmp()?
>>>
>>> Potentially.  The numbers are small enough O(N^2) isn't fatal.
>>
>> Define "small", please.
>>
>> O(N^2) makes me nervous about what other use cases lurk out
>> there that may get bitten by this.
>
> Worst case for N (One namespace per thread) is about 60k.

I'm getting an education here: where does the 60k number come from?

> A typical heavy use case may be 1000 namespaces of any type.
> So we are talking about O(N^2) that rarely happens and should be done in
> a couple of seconds.

I don't know whether that's acceptable for the migration use case,
but seems quite bad for the visualization use case.

>>> Where kcmp shines is that it allows migration to happen.  Inode numbers
>>> to change (which they very much will today), and still have things work.
>>
>>
>>> We can keep it O(Nlog(N)) by taking advantage of not just the equality
>>> but the ordering relationship.  Although Ugh.
>>
>> Yes, that sounds pretty ugly...
>
> Actually having thought about this a little more if kcmp returns an
> ordering by inode and migration preserves the relative order of
> the inodes (which should just be a creation order) it should be quite
> solvable.
>
> Switch from an order by inode number to an order by object creation
> time, and guarantee that all creations are have an order (which with
> task_list_lock we practically already have) and it should be even easier
> to create.  (A 64bit nanosecond resolution timestamp is good for 544
> years of uptime).  A 64bit number that increments each time an object is
> created should have an even better lifespan.
>
> I don't know if we can find a way to give that guarantee for other kcmp
> comparisons but it is worth a thought.

Okay. So, this is a pathway to O(Nlog(N)) at least then?

>>> One disadvantage of
>>> kcmp currently is that the way the ordering relationship is defined
>>> the order is not preserved over migration :(
>>
>> So, does kcmp() fully solve the proble(s) at hand? It sounds like
>> not, if I understand your last point correctly.
>
> There are 3 possibilities I see for migration in migration, ordered
> in order of implementation difficulty.
> 1) Have a clear signal that migration happened and a nested migration
>    needs to restart.
> 2) Use kcmp so that only the relative order needs to be preserved.
> 3) Preserve the device number and inode numbers.
>
> At a practical level I think (2) may actually in net be the simplest.
> It requires a little more care to implement and you have to opt in,
> but it should not require any rolling back of activity (merely careful
> ordering of object creation).
>
> I definititely like kcmp knowing how to compare things by inode
> (aka st_dev, st_inode) because then even if you have to restart
> the comparisons after a migration the exact details you are comparing
> are hidden and so it is easier to support and harder to get wrong.
>
> I can imagine how to preserve inode numbers by creating a new instance
> of nsfs instance and using the old inode numbers upon restore.  I don't
> currently see how we could possibly preserve st_dev over migration short of
> a device number namespace.
>
> So if we are going to continue with making device numbers be a legacy
> attribute applications should not care about we need a way to compare
> things by not looking at st_dev.  Which brings us back to kcmp.
>
> Hmm.  Hotplugging as disk and plugging it back likely will change the
> device number and give the same kind of challenge with st_dev (although
> you can't keep a file descriptor open across that kind of event).  So
> certainly a hotplug event on a device should be enough to say don't care
> about the device number.

Okay.

Thanks,

Michael
Andrey Vagin Aug. 1, 2016, 11:01 p.m. UTC | #21
On Fri, Jul 29, 2016 at 01:05:48PM -0500, Eric W. Biederman wrote:
> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
> 
> > Hi Eric,
> >
> > On 07/28/2016 02:56 PM, Eric W. Biederman wrote:
> >> "Michael Kerrisk (man-pages)" <mtk.manpages@gmail.com> writes:
> >>
> >>> On 07/26/2016 10:39 PM, Andrew Vagin wrote:
> >>>> On Tue, Jul 26, 2016 at 09:17:31PM +0200, Michael Kerrisk (man-pages) wrote:
> >>
> >>>> If we want to compare two file descriptors of the current process,
> >>>> it is one of cases for which kcmp can be used. We can call kcmp to
> >>>> compare two namespaces which are opened in other processes.
> >>>
> >>> Is there really a use case there? I assume we're talking about the
> >>> scenario where a process in one namespace opens a /proc/PID/ns/*
> >>> file descriptor and passes that FD to another process via a UNIX
> >>> domain socket. Is that correct?
> >>>
> >>> So, supposing that we want to build a map of the relationships
> >>> between namespaces using the proposed kcmp() API, and there are
> >>> say N namespaces? Does this mena we make (N * (N-1) / 2) calls
> >>> to kcmp()?
> >>
> >> Potentially.  The numbers are small enough O(N^2) isn't fatal.
> >
> > Define "small", please.
> >
> > O(N^2) makes me nervous about what other use cases lurk out
> > there that may get bitten by this.
> 
> Worst case for N (One namespace per thread) is about 60k.
> A typical heavy use case may be 1000 namespaces of any type.
> So we are talking about O(N^2) that rarely happens and should be done in
> a couple of seconds.
> 
> >> Where kcmp shines is that it allows migration to happen.  Inode numbers
> >> to change (which they very much will today), and still have things work.
> >
> >
> >> We can keep it O(Nlog(N)) by taking advantage of not just the equality
> >> but the ordering relationship.  Although Ugh.
> >
> > Yes, that sounds pretty ugly...
> 
> Actually having thought about this a little more if kcmp returns an
> ordering by inode and migration preserves the relative order of
> the inodes (which should just be a creation order) it should be quite
> solvable.
> 
> Switch from an order by inode number to an order by object creation
> time, and guarantee that all creations are have an order (which with
> task_list_lock we practically already have) and it should be even easier
> to create.  (A 64bit nanosecond resolution timestamp is good for 544
> years of uptime).  A 64bit number that increments each time an object is
> created should have an even better lifespan.
> 
> I don't know if we can find a way to give that guarantee for other kcmp
> comparisons but it is worth a thought.
> 
> >>One disadvantage of
> >> kcmp currently is that the way the ordering relationship is defined
> >> the order is not preserved over migration :(
> >
> > So, does kcmp() fully solve the proble(s) at hand? It sounds like
> > not, if I understand your last point correctly.
> 
> There are 3 possibilities I see for migration in migration, ordered
> in order of implementation difficulty.
> 1) Have a clear signal that migration happened and a nested migration
>    needs to restart.
> 2) Use kcmp so that only the relative order needs to be preserved.
> 3) Preserve the device number and inode numbers.
> 
> At a practical level I think (2) may actually in net be the simplest.
> It requires a little more care to implement and you have to opt in,
> but it should not require any rolling back of activity (merely careful
> ordering of object creation).
> 
> I definititely like kcmp knowing how to compare things by inode
> (aka st_dev, st_inode) because then even if you have to restart
> the comparisons after a migration the exact details you are comparing
> are hidden and so it is easier to support and harder to get wrong.
> 
> I can imagine how to preserve inode numbers by creating a new instance
> of nsfs instance and using the old inode numbers upon restore.  I don't
> currently see how we could possibly preserve st_dev over migration short of
> a device number namespace.

I think we can avoid comparing st_dev if we will compare inode numbers
for parent user namespaces.

Namespaces looks like a tree where user-namespaces are directories and
other namespaces are files.

A namespace can be described by a path in this imaginary file system,
which looks like /userns1/userns2/XXXns.

In this case we need to guarantee uniq names inside each directories and
that they will be not changed over migration.

> 
> So if we are going to continue with making device numbers be a legacy
> attribute applications should not care about we need a way to compare
> things by not looking at st_dev.  Which brings us back to kcmp.
> 
> Hmm.  Hotplugging as disk and plugging it back likely will change the
> device number and give the same kind of challenge with st_dev (although
> you can't keep a file descriptor open across that kind of event).  So
> certainly a hotplug event on a device should be enough to say don't care
> about the device number.
> 
> Eric
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch
diff mbox

diff --git a/man7/namespaces.7 b/man7/namespaces.7
index 98ed3e5..207e4a5 100644
--- a/man7/namespaces.7
+++ b/man7/namespaces.7
@@ -149,6 +149,49 @@  even if all processes in the namespace terminate.
 The file descriptor can be passed to
 .BR setns (2).
 
+Since Linux 4.X, the following
+.BR ioctl (2)
+calls are supported for namespace file descriptors.
+The correct syntax is:
+.PP
+.RS
+.nf
+.IB fd " = ioctl(" ns_fd ", " ioctl_type ");"
+.fi
+.RE
+.PP
+where
+.I ioctl_type
+is one of the following:
+.TP
+.B NS_GET_USERNS
+Returns a file descriptor that refers to an owning user namespace.
+.TP
+.B NS_GET_PARENT
+Returns a file descriptor that refers to a parent namespace. This
+.BR ioctl (2)
+can be used for pid and user namespaces. For user namespaces,
+.B NS_GET_PARENT
+and
+.B NS_GET_USERNS
+have the same meaning.
+.PP
+In addition to generic
+.BR ioctl (2)
+errors, the following specific ones can occur:
+.PP
+.TP
+.B EINVAL
+.B NS_GET_PARENT
+was called for a nonhierarchical namespace.
+.TP
+.B EPERM
+The requested namespace is outside of the current namespace scope.
+.TP
+.B ENOENT
+.IB ns_fd
+refers to the init namespace.
+.PP
 In Linux 3.7 and earlier, these files were visible as hard links.
 Since Linux 3.8, they appear as symbolic links.
 If two processes are in the same namespace, then the inode numbers of their