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

Groups > linux.kernel > #1638449 > unrolled thread

[PATCH] fs: add an ioctl to get an owning userns for a superblock

Back to article view | Back to linux.kernel

Contents

  [PATCH] fs: add an ioctl to get an owning userns for a superblock Andrei Vagin <avagin@openvz.org> - 2017-05-10 01:30 +0200
    Re: [PATCH] fs: add an ioctl to get an owning userns for a superblock ebiederm@xmission.com (Eric W. Biederman) - 2017-05-10 02:50 +0200
      Re: [PATCH] fs: add an ioctl to get an owning userns for a superblock ebiederm@xmission.com (Eric W. Biederman) - 2017-05-11 03:40 +0200

#1638449 — [PATCH] fs: add an ioctl to get an owning userns for a superblock

The introduced ioctl returns a file descriptor that refers to a owning
user namespace for a superblock which is associated with a target file
descriptor.

EPERM is returned if the current process doesn't have CAP_SYS_ADMIN in
the returned user namespace.

This information is required to dump and restore mount namespaces. We
need to know to which user namespace a superblock is belonged to.

We already have the SIOCGSKNS ioctl for sockets to get a network
namespace, so it looks reasonable to use the same interface for
superblocks too.

This functionality can be useful for users in order to understand
a running system.

Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andrei Vagin <avagin@openvz.org>
---
 fs/ioctl.c              | 23 +++++++++++++++++++++++
 include/uapi/linux/fs.h |  2 ++
 2 files changed, 25 insertions(+)

diff --git a/fs/ioctl.c b/fs/ioctl.c
index 569db68..22bbf37 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -16,6 +16,8 @@
 #include <linux/buffer_head.h>
 #include <linux/falloc.h>
 #include <linux/sched/signal.h>
+#include <linux/proc_fs.h>
+#include <linux/user_namespace.h>
 
 #include "internal.h"
 
@@ -614,6 +616,25 @@ static int ioctl_file_dedupe_range(struct file *file, void __user *arg)
 	return ret;
 }
 
+static struct ns_common *get_sb_userns(struct ns_common *ns_common)
+{
+	struct user_namespace *ns;
+
+	ns = container_of(ns_common, struct user_namespace, ns);
+
+	return &get_user_ns(ns)->ns;
+}
+
+static int ioctl_fs_sb_userns(struct file *filp)
+{
+	struct super_block *sb = file_inode(filp)->i_sb;
+
+	if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
+		return -EPERM;
+
+	return open_related_ns(&sb->s_user_ns->ns, get_sb_userns);
+}
+
 /*
  * When you add any new common ioctls to the switches above and below
  * please update compat_sys_ioctl() too.
@@ -677,6 +698,8 @@ int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
 
 	case FIDEDUPERANGE:
 		return ioctl_file_dedupe_range(filp, argp);
+	case FS_IOC_SB_USERNS:
+		return ioctl_fs_sb_userns(filp);
 
 	default:
 		if (S_ISREG(inode->i_mode))
diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
index 33423aa..26ef2d5 100644
--- a/include/uapi/linux/fs.h
+++ b/include/uapi/linux/fs.h
@@ -246,6 +246,8 @@ struct fsxattr {
 #define FICLONE		_IOW(0x94, 9, int)
 #define FICLONERANGE	_IOW(0x94, 13, struct file_clone_range)
 #define FIDEDUPERANGE	_IOWR(0x94, 54, struct file_dedupe_range)
+/* Get a file descriptor to an owning userns for a superblock */
+#define FS_IOC_SB_USERNS		_IOR('X', 55, int)
 
 #define	FS_IOC_GETFLAGS			_IOR('f', 1, long)
 #define	FS_IOC_SETFLAGS			_IOW('f', 2, long)
-- 
2.9.3

[toc] | [next] | [standalone]

#1638473

Andrei Vagin <avagin@openvz.org> writes:

> The introduced ioctl returns a file descriptor that refers to a owning
> user namespace for a superblock which is associated with a target file
> descriptor.
>
> EPERM is returned if the current process doesn't have CAP_SYS_ADMIN in
> the returned user namespace.
>
> This information is required to dump and restore mount namespaces. We
> need to know to which user namespace a superblock is belonged to.
>
> We already have the SIOCGSKNS ioctl for sockets to get a network
> namespace, so it looks reasonable to use the same interface for
> superblocks too.
>
> This functionality can be useful for users in order to understand
> a running system.

This will probably work.  And the capability check eases any concerns
I might have that this would be a trivial information leak.

That said can we hold off just a little bit.  If open_fs work actually
turns into a real interface that would seem to be the perfect place
to stick this functionality.

Eric

>
> Cc: Alexander Viro <viro@zeniv.linux.org.uk>
> Cc: Eric W. Biederman <ebiederm@xmission.com>
> Signed-off-by: Andrei Vagin <avagin@openvz.org>
> ---
>  fs/ioctl.c              | 23 +++++++++++++++++++++++
>  include/uapi/linux/fs.h |  2 ++
>  2 files changed, 25 insertions(+)
>
> diff --git a/fs/ioctl.c b/fs/ioctl.c
> index 569db68..22bbf37 100644
> --- a/fs/ioctl.c
> +++ b/fs/ioctl.c
> @@ -16,6 +16,8 @@
>  #include <linux/buffer_head.h>
>  #include <linux/falloc.h>
>  #include <linux/sched/signal.h>
> +#include <linux/proc_fs.h>
> +#include <linux/user_namespace.h>
>  
>  #include "internal.h"
>  
> @@ -614,6 +616,25 @@ static int ioctl_file_dedupe_range(struct file *file, void __user *arg)
>  	return ret;
>  }
>  
> +static struct ns_common *get_sb_userns(struct ns_common *ns_common)
> +{
> +	struct user_namespace *ns;
> +
> +	ns = container_of(ns_common, struct user_namespace, ns);
> +
> +	return &get_user_ns(ns)->ns;
> +}
> +
> +static int ioctl_fs_sb_userns(struct file *filp)
> +{
> +	struct super_block *sb = file_inode(filp)->i_sb;
> +
> +	if (!ns_capable(sb->s_user_ns, CAP_SYS_ADMIN))
> +		return -EPERM;
> +
> +	return open_related_ns(&sb->s_user_ns->ns, get_sb_userns);
> +}
> +
>  /*
>   * When you add any new common ioctls to the switches above and below
>   * please update compat_sys_ioctl() too.
> @@ -677,6 +698,8 @@ int do_vfs_ioctl(struct file *filp, unsigned int fd, unsigned int cmd,
>  
>  	case FIDEDUPERANGE:
>  		return ioctl_file_dedupe_range(filp, argp);
> +	case FS_IOC_SB_USERNS:
> +		return ioctl_fs_sb_userns(filp);
>  
>  	default:
>  		if (S_ISREG(inode->i_mode))
> diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
> index 33423aa..26ef2d5 100644
> --- a/include/uapi/linux/fs.h
> +++ b/include/uapi/linux/fs.h
> @@ -246,6 +246,8 @@ struct fsxattr {
>  #define FICLONE		_IOW(0x94, 9, int)
>  #define FICLONERANGE	_IOW(0x94, 13, struct file_clone_range)
>  #define FIDEDUPERANGE	_IOWR(0x94, 54, struct file_dedupe_range)
> +/* Get a file descriptor to an owning userns for a superblock */
> +#define FS_IOC_SB_USERNS		_IOR('X', 55, int)
>  
>  #define	FS_IOC_GETFLAGS			_IOR('f', 1, long)
>  #define	FS_IOC_SETFLAGS			_IOW('f', 2, long)

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

#1639144

Andrei Vagin <avagin@virtuozzo.com> writes:

> On Tue, May 09, 2017 at 07:34:00PM -0500, Eric W. Biederman wrote:
>> Andrei Vagin <avagin@openvz.org> writes:
>> 
>> > The introduced ioctl returns a file descriptor that refers to a owning
>> > user namespace for a superblock which is associated with a target file
>> > descriptor.
>> >
>> > EPERM is returned if the current process doesn't have CAP_SYS_ADMIN in
>> > the returned user namespace.
>> >
>> > This information is required to dump and restore mount namespaces. We
>> > need to know to which user namespace a superblock is belonged to.
>> >
>> > We already have the SIOCGSKNS ioctl for sockets to get a network
>> > namespace, so it looks reasonable to use the same interface for
>> > superblocks too.
>> >
>> > This functionality can be useful for users in order to understand
>> > a running system.
>> 
>> This will probably work.  And the capability check eases any concerns
>> I might have that this would be a trivial information leak.
>> 
>> That said can we hold off just a little bit.  If open_fs work actually
>> turns into a real interface that would seem to be the perfect place
>> to stick this functionality.
>
> Sure, we can. Do you know any place where to read more information about
> open_fs? I think I have heared a few times about this idea, but it would be
> good to get more details.


Look for David Howells <dhowells@redhat.com> recent patches on lkml he
has implemented an initial rfc for it.

Eric

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web