[PATCH net-next v2 00/10] net: fix sysfs permssions when device changes network

[PATCH net-next v2 00/10] net: fix sysfs permssions when device changes network

[Christian Brauner]

Hey everyone,

This is v2 with fixes based on Greg's and Rafael's review.

(I've tagged this with net-next since it's triggered by a bug for
 network device files but it also touches driver core aspects so it's
 not clear-cut. I can of course split this series into separate
 patchsets.) 
We have been struggling with a bug surrounding the ownership of network
device sysfs files when moving network devices between network
namespaces owned by different user namespaces reported by multiple
users.

Currently, when moving network devices between network namespaces the
ownership of the corresponding sysfs entries is not changed. This leads
to problems when tools try to operate on the corresponding sysfs files.

I also causes a bug when creating a network device in a network
namespaces owned by a user namespace and moving that network device back
to the host network namespaces. Because when a network device is created
in a network namespaces it will be owned by the root user of the user
namespace and all its associated sysfs files will also be owned by the
root user of the corresponding user namespace.
If such a network device has to be moved back to the host network
namespace the permissions will still be set to the root user of the
owning user namespaces of the originating network namespace. This means
unprivileged users can e.g. re-trigger uevents for such incorrectly
owned devices on the host or in other network namespaces. They can also
modify the settings of the device itself through sysfs when they
wouldn't be able to do the same through netlink. Both of these things
are unwanted.

For example, quite a few workloads will create network devices in the
host network namespace. Other tools will then proceed to move such
devices between network namespaces owner by other user namespaces. While
the ownership of the device itself is updated in
net/core/net-sysfs.c:dev_change_net_namespace() the corresponding sysfs
entry for the device is not. Below you'll find that moving a network
device (here a veth device) from a network namespace into another
network namespaces owned by a different user namespace with a different
id mapping. As you can see the permissions are wrong even though it is
owned by the userns root user after it has been moved and can be
interacted with through netlink: 

drwxr-xr-x 5 nobody nobody    0 Jan 25 18:08 .
drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_assign_type
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_len
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 address
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 broadcast
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_changes
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_down_count
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_up_count
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_id
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_port
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dormant
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 duplex
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 flags
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 gro_flush_timeout
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifalias
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifindex
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 iflink
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 link_mode
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 mtu
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 name_assign_type
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 netdev_group
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 operstate
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_id
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_name
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_switch_id
drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 power
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 proto_down
drwxr-xr-x 4 nobody nobody    0 Jan 25 18:09 queues
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 speed
drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 statistics
lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:08 subsystem -> ../../../../class/net
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 tx_queue_len
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 type
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:08 uevent

Constrast this with creating a device of the same type in the network
namespace directly. In this case the device's sysfs permissions will be
correctly updated.
(Please also note, that in a lot of workloads this strategy of creating
 the network device directly in the network device to workaround this
 issue can not be used. Either because the network device is dedicated
 after it has been created or because it used by a process that is
 heavily sandboxed and couldn't create network devices itself.):

drwxr-xr-x 5 root   root      0 Jan 25 18:12 .
drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
-r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_assign_type
-r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_len
-r--r--r-- 1 root   root   4096 Jan 25 18:12 address
-r--r--r-- 1 root   root   4096 Jan 25 18:12 broadcast
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 carrier
-r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_changes
-r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_down_count
-r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_up_count
-r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_id
-r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_port
-r--r--r-- 1 root   root   4096 Jan 25 18:12 dormant
-r--r--r-- 1 root   root   4096 Jan 25 18:12 duplex
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 flags
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 gro_flush_timeout
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 ifalias
-r--r--r-- 1 root   root   4096 Jan 25 18:12 ifindex
-r--r--r-- 1 root   root   4096 Jan 25 18:12 iflink
-r--r--r-- 1 root   root   4096 Jan 25 18:12 link_mode
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 mtu
-r--r--r-- 1 root   root   4096 Jan 25 18:12 name_assign_type
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 netdev_group
-r--r--r-- 1 root   root   4096 Jan 25 18:12 operstate
-r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_id
-r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_name
-r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_switch_id
drwxr-xr-x 2 root   root      0 Jan 25 18:12 power
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 proto_down
drwxr-xr-x 4 root   root      0 Jan 25 18:12 queues
-r--r--r-- 1 root   root   4096 Jan 25 18:12 speed
drwxr-xr-x 2 root   root      0 Jan 25 18:12 statistics
lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:12 subsystem -> ../../../../class/net
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 tx_queue_len
-r--r--r-- 1 root   root   4096 Jan 25 18:12 type
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 uevent

Now, when creating a network device in a network namespace owned by a
user namespace and moving it to the host the permissions will be set to
the id that the user namespace root user has been mapped to on the host
leading to all sorts of permission issues mentioned above:

458752
drwxr-xr-x 5 458752 458752      0 Jan 25 18:12 .
drwxr-xr-x 9 root   root        0 Jan 25 18:08 ..
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_assign_type
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_len
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 address
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 broadcast
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_changes
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_down_count
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_up_count
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_id
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_port
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dormant
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 duplex
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 flags
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 gro_flush_timeout
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 ifalias
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 ifindex
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 iflink
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 link_mode
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 mtu
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 name_assign_type
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 netdev_group
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 operstate
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_id
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_name
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_switch_id
drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 power
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 proto_down
drwxr-xr-x 4 458752 458752      0 Jan 25 18:12 queues
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 speed
drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 statistics
lrwxrwxrwx 1 root   root        0 Jan 25 18:12 subsystem -> ../../../../class/net
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 tx_queue_len
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 type
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 uevent

Fix this by changing the basic sysfs files associated with network
devices when moving them between network namespaces. To this end we add
some infrastructure to sysfs.

The patchset takes care to only do this when the owning user namespaces
changes and the kids differ. So there's only a performance overhead,
when the owning user namespace of the network namespace is different
__and__ the kid mappings for the root user are different for the two
user namespaces:
Assume we have a netdev eth0 which we create in netns1 owned by userns1.
userns1 has an id mapping of 0 100000 100000. Now we move eth0 into
netns2 which is owned by userns2 which also defines an id mapping of 0
100000 100000. In this case sysfs doesn't need updating. The patch will
handle this case and not do any needless work. Now assume eth0 is moved
into netns3 which is owned by userns3 which defines an id mapping of 0
123456 65536. In this case the root user in each namespace corresponds
to different kid and sysfs needs updating.

Thanks!
Christian

Christian Brauner (10):
  sysfs: add sysfs_file_change_owner{_by_name}()
  sysfs: add sysfs_link_change_owner()
  sysfs: add sysfs_group_change_owner()
  sysfs: add sysfs_groups_change_owner()
  sysfs: add sysfs_change_owner()
  device: add device_change_owner()
  drivers/base/power: add dpm_sysfs_change_owner()
  net-sysfs: add netdev_change_owner()
  net-sysfs: add queue_change_owner()
  net: fix sysfs permssions when device changes network namespace

 drivers/base/core.c        |  77 +++++++++++++++++
 drivers/base/power/power.h |   2 +
 drivers/base/power/sysfs.c |  40 +++++++++
 fs/sysfs/file.c            | 171 +++++++++++++++++++++++++++++++++++++
 fs/sysfs/group.c           | 129 ++++++++++++++++++++++++++++
 include/linux/device.h     |   1 +
 include/linux/sysfs.h      |  45 ++++++++++
 net/core/dev.c             |   9 +-
 net/core/net-sysfs.c       | 128 +++++++++++++++++++++++++++
 net/core/net-sysfs.h       |   2 +
 10 files changed, 603 insertions(+), 1 deletion(-)


base-commit: bb6d3fb354c5ee8d6bde2d576eb7220ea09862b9
-- 
2.25.0

[PATCH net-next v2 01/10] sysfs: add sysfs_file_change_owner{_by_name}()

[Christian Brauner]

Add helpers to change owner of a sysfs files.
The ownership of a sysfs object is determined based on the ownership of
the corresponding kobject, i.e. only if the ownership of a kobject is
changed will this function change the ownership of the corresponding
sysfs entry.
This function will be used to correctly account for kobject ownership
changes, e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
-  Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
   - Better naming for sysfs_file_change_owner() to reflect the fact that it
     can be used to change the owner of the kobject itself by passing NULL as
     argument.
- Christian Brauner <christian.brauner@ubuntu.com>:
  - Split sysfs_file_change_owner() into two helpers sysfs_change_owner() and
    sysfs_change_owner_by_name(). The former changes the owner of the kobject
    itself, the latter the owner of the kobject looked up via the name
    argument.
---
 fs/sysfs/file.c       | 82 +++++++++++++++++++++++++++++++++++++++++++
 include/linux/sysfs.h | 14 ++++++++
 2 files changed, 96 insertions(+)

diff --git a/fs/sysfs/file.c b/fs/sysfs/file.c
index 130fc6fbcc03..8f2607de2456 100644
--- a/fs/sysfs/file.c
+++ b/fs/sysfs/file.c
@@ -558,3 +558,85 @@ void sysfs_remove_bin_file(struct kobject *kobj,
 	kernfs_remove_by_name(kobj->sd, attr->attr.name);
 }
 EXPORT_SYMBOL_GPL(sysfs_remove_bin_file);
+
+static int internal_change_owner(struct kernfs_node *kn, struct kobject *kobj)
+{
+	kuid_t uid;
+	kgid_t gid;
+	struct iattr newattrs = {
+		.ia_valid = ATTR_UID | ATTR_GID,
+	};
+
+	kobject_get_ownership(kobj, &uid, &gid);
+	newattrs.ia_uid = uid;
+	newattrs.ia_gid = gid;
+
+	return kernfs_setattr(kn, &newattrs);
+}
+
+/**
+ *	sysfs_file_change_owner_by_name - change owner of a file.
+ *	@kobj:	object.
+ *	@name:	name of the file to change.
+ *
+ * To change the ownership of a sysfs object, the caller must first change the
+ * uid/gid of the kobject and then call this function. Usually this will be
+ * taken care of by the relevant subsystem, e.g. moving a network device
+ * between network namespaces owned by different user namespaces will change
+ * the uid/gid of the kobject to the uid/gid of the root user in the user
+ * namespace. Calling this function afterwards will cause the sysfs object to
+ * reflect the new uid/gid.
+ */
+int sysfs_file_change_owner_by_name(struct kobject *kobj, const char *name)
+{
+	struct kernfs_node *kn;
+	int error;
+
+	if (!name)
+		return -EINVAL;
+
+	if (!kobj->state_in_sysfs)
+		return -EINVAL;
+
+	kn = kernfs_find_and_get(kobj->sd, name);
+	if (!kn)
+		return -ENOENT;
+
+	error = internal_change_owner(kn, kobj);
+
+	kernfs_put(kn);
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(sysfs_file_change_owner_by_name);
+
+/**
+ *	sysfs_file_change_owner - change owner of a file.
+ *	@kobj:	object.
+ *
+ * To change the ownership of a sysfs object, the caller must first change the
+ * uid/gid of the kobject and then call this function. Usually this will be
+ * taken care of by the relevant subsystem, e.g. moving a network device
+ * between network namespaces owned by different user namespaces will change
+ * the uid/gid of the kobject to the uid/gid of the root user in the user
+ * namespace. Calling this function afterwards will cause the sysfs object to
+ * reflect the new uid/gid.
+ */
+int sysfs_file_change_owner(struct kobject *kobj)
+{
+	struct kernfs_node *kn;
+	int error;
+
+	if (!kobj->state_in_sysfs)
+		return -EINVAL;
+
+	kernfs_get(kobj->sd);
+
+	kn = kobj->sd;
+	error = internal_change_owner(kn, kobj);
+
+	kernfs_put(kn);
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(sysfs_file_change_owner);
diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
index fa7ee503fb76..4b3c3b76ff80 100644
--- a/include/linux/sysfs.h
+++ b/include/linux/sysfs.h
@@ -310,6 +310,9 @@ static inline void sysfs_enable_ns(struct kernfs_node *kn)
 	return kernfs_enable_ns(kn);
 }
 
+int sysfs_file_change_owner(struct kobject *kobj);
+int sysfs_file_change_owner_by_name(struct kobject *kobj, const char *name);
+
 #else /* CONFIG_SYSFS */
 
 static inline int sysfs_create_dir_ns(struct kobject *kobj, const void *ns)
@@ -522,6 +525,17 @@ static inline void sysfs_enable_ns(struct kernfs_node *kn)
 {
 }
 
+static inline int int sysfs_file_change_owner(struct kobject *kobj)
+{
+	return 0;
+}
+
+static inline int sysfs_file_change_owner_by_name(struct kobject *kobj,
+						  const char *name)
+{
+	return 0;
+}
+
 #endif /* CONFIG_SYSFS */
 
 static inline int __must_check sysfs_create_file(struct kobject *kobj,
-- 
2.25.0

[PATCH net-next v2 02/10] sysfs: add sysfs_link_change_owner()

[Christian Brauner]

Add a helper to change the owner of a sysfs link.
The ownership of a sysfs object is determined based on the ownership of
the corresponding kobject, i.e. only if the ownership of a kobject is
changed will this function change the ownership of the corresponding
sysfs entry.
This function will be used to correctly account for kobject ownership
changes, e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
-  Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
   - Add comment how ownership of sysfs object is changed.
---
 fs/sysfs/file.c       | 46 +++++++++++++++++++++++++++++++++++++++++++
 include/linux/sysfs.h |  9 +++++++++
 2 files changed, 55 insertions(+)

diff --git a/fs/sysfs/file.c b/fs/sysfs/file.c
index 8f2607de2456..8b20245f359c 100644
--- a/fs/sysfs/file.c
+++ b/fs/sysfs/file.c
@@ -574,6 +574,52 @@ static int internal_change_owner(struct kernfs_node *kn, struct kobject *kobj)
 	return kernfs_setattr(kn, &newattrs);
 }
 
+/**
+ *	sysfs_link_change_owner - change owner of a link.
+ *	@kobj:	object of the kernfs_node the symlink is located in.
+ *	@targ:	object of the kernfs_node the symlink points to.
+ *	@name:	name of the link.
+ *
+ * To change the ownership of a sysfs object, the caller must first change the
+ * uid/gid of the kobject and then call this function. Usually this will be
+ * taken care of by the relevant subsystem, e.g. moving a network device
+ * between network namespaces owned by different user namespaces will change
+ * the uid/gid of the kobject to the uid/gid of the root user in the user
+ * namespace. Calling this function afterwards will cause the sysfs object to
+ * reflect the new uid/gid.
+ */
+int sysfs_link_change_owner(struct kobject *kobj, struct kobject *targ,
+			    const char *name)
+{
+	struct kernfs_node *parent, *kn = NULL;
+	int error;
+
+	if (!kobj)
+		parent = sysfs_root_kn;
+	else
+		parent = kobj->sd;
+
+	if (!targ->state_in_sysfs)
+		return -EINVAL;
+
+	error = -ENOENT;
+	kn = kernfs_find_and_get_ns(parent, name, targ->sd->ns);
+	if (!kn)
+		goto out;
+
+	error = -EINVAL;
+	if (kernfs_type(kn) != KERNFS_LINK)
+		goto out;
+	if (kn->symlink.target_kn->priv != targ)
+		goto out;
+
+	error = internal_change_owner(kn, targ);
+
+out:
+	kernfs_put(kn);
+	return error;
+}
+
 /**
  *	sysfs_file_change_owner_by_name - change owner of a file.
  *	@kobj:	object.
diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
index 4b3c3b76ff80..238f3d7b1fa0 100644
--- a/include/linux/sysfs.h
+++ b/include/linux/sysfs.h
@@ -312,6 +312,8 @@ static inline void sysfs_enable_ns(struct kernfs_node *kn)
 
 int sysfs_file_change_owner(struct kobject *kobj);
 int sysfs_file_change_owner_by_name(struct kobject *kobj, const char *name);
+int sysfs_link_change_owner(struct kobject *kobj, struct kobject *targ,
+			    const char *name);
 
 #else /* CONFIG_SYSFS */
 
@@ -536,6 +538,13 @@ static inline int sysfs_file_change_owner_by_name(struct kobject *kobj,
 	return 0;
 }
 
+static inline int sysfs_link_change_owner(struct kobject *kobj,
+					  struct kobject *targ,
+					  const char *name)
+{
+	return 0;
+}
+
 #endif /* CONFIG_SYSFS */
 
 static inline int __must_check sysfs_create_file(struct kobject *kobj,
-- 
2.25.0

[PATCH net-next v2 03/10] sysfs: add sysfs_group_change_owner()

[Christian Brauner]

Add a helper to change the owner of a sysfs group.
The ownership of a sysfs object is determined based on the ownership of
the corresponding kobject, i.e. only if the ownership of a kobject is
changed will this function change the ownership of the corresponding
sysfs entry.
This function will be used to correctly account for kobject ownership
changes, e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
-  Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
   - Add comment how ownership of sysfs object is changed.
---
 fs/sysfs/group.c      | 93 +++++++++++++++++++++++++++++++++++++++++++
 include/linux/sysfs.h |  8 ++++
 2 files changed, 101 insertions(+)

diff --git a/fs/sysfs/group.c b/fs/sysfs/group.c
index c4ab045926b7..7d158d59d097 100644
--- a/fs/sysfs/group.c
+++ b/fs/sysfs/group.c
@@ -13,6 +13,7 @@
 #include <linux/dcache.h>
 #include <linux/namei.h>
 #include <linux/err.h>
+#include <linux/fs.h>
 #include "sysfs.h"
 
 
@@ -457,3 +458,95 @@ int __compat_only_sysfs_link_entry_to_kobj(struct kobject *kobj,
 	return PTR_ERR_OR_ZERO(link);
 }
 EXPORT_SYMBOL_GPL(__compat_only_sysfs_link_entry_to_kobj);
+
+static int sysfs_group_attrs_change_owner(struct kernfs_node *grp_kn,
+					  const struct attribute_group *grp,
+					  struct iattr *newattrs)
+{
+	struct kernfs_node *kn;
+	int error;
+
+	if (grp->attrs) {
+		struct attribute *const *attr;
+
+		for (attr = grp->attrs; *attr; attr++) {
+			kn = kernfs_find_and_get(grp_kn, (*attr)->name);
+			if (!kn)
+				return -ENOENT;
+
+			error = kernfs_setattr(kn, newattrs);
+			kernfs_put(kn);
+			if (error)
+				return error;
+		}
+	}
+
+	if (grp->bin_attrs) {
+		struct bin_attribute *const *bin_attr;
+
+		for (bin_attr = grp->bin_attrs; *bin_attr; bin_attr++) {
+			kn = kernfs_find_and_get(grp_kn, (*bin_attr)->attr.name);
+			if (!kn)
+				return -ENOENT;
+
+			error = kernfs_setattr(kn, newattrs);
+			kernfs_put(kn);
+			if (error)
+				return error;
+		}
+	}
+
+	return 0;
+}
+
+/**
+ * sysfs_group_change_owner - change owner of an attribute group.
+ * @kobj:	The kobject containing the group.
+ * @grp:	The attribute group.
+ *
+ * To change the ownership of a sysfs object, the caller must first change the
+ * uid/gid of the kobject and then call this function. Usually this will be
+ * taken care of by the relevant subsystem, e.g. moving a network device
+ * between network namespaces owned by different user namespaces will change
+ * the uid/gid of the kobject to the uid/gid of the root user in the user
+ * namespace. Calling this function afterwards will cause the sysfs object to
+ * reflect the new uid/gid.
+ *
+ * Returns 0 on success or error code on failure.
+ */
+int sysfs_group_change_owner(struct kobject *kobj,
+			     const struct attribute_group *grp)
+{
+	struct kernfs_node *grp_kn;
+	kuid_t uid;
+	kgid_t gid;
+	int error;
+	struct iattr newattrs = {
+		.ia_valid = ATTR_UID | ATTR_GID,
+	};
+
+	if (!kobj->state_in_sysfs)
+		return -EINVAL;
+
+	if (grp->name) {
+		grp_kn = kernfs_find_and_get(kobj->sd, grp->name);
+	} else {
+		kernfs_get(kobj->sd);
+		grp_kn = kobj->sd;
+	}
+	if (!grp_kn)
+		return -ENOENT;
+
+	kobject_get_ownership(kobj, &uid, &gid);
+	newattrs.ia_uid = uid;
+	newattrs.ia_gid = gid;
+
+	error = kernfs_setattr(grp_kn, &newattrs);
+	if (!error)
+		error = sysfs_group_attrs_change_owner(grp_kn, grp, &newattrs);
+
+	kernfs_put(grp_kn);
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(sysfs_group_change_owner);
diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
index 238f3d7b1fa0..dd211f0a654b 100644
--- a/include/linux/sysfs.h
+++ b/include/linux/sysfs.h
@@ -314,6 +314,8 @@ int sysfs_file_change_owner(struct kobject *kobj);
 int sysfs_file_change_owner_by_name(struct kobject *kobj, const char *name);
 int sysfs_link_change_owner(struct kobject *kobj, struct kobject *targ,
 			    const char *name);
+int sysfs_group_change_owner(struct kobject *kobj,
+			     const struct attribute_group *groups);
 
 #else /* CONFIG_SYSFS */
 
@@ -545,6 +547,12 @@ static inline int sysfs_link_change_owner(struct kobject *kobj,
 	return 0;
 }
 
+static inline int sysfs_group_change_owner(struct kobject *kobj,
+			 const struct attribute_group **groups)
+{
+	return 0;
+}
+
 #endif /* CONFIG_SYSFS */
 
 static inline int __must_check sysfs_create_file(struct kobject *kobj,
-- 
2.25.0

[PATCH net-next v2 04/10] sysfs: add sysfs_groups_change_owner()

[Christian Brauner]

Add a helper to change the owner of a set of sysfs groups file.
The ownership of a sysfs object is determined based on the ownership of
the corresponding kobject, i.e. only if the ownership of a kobject is
changed will this function change the ownership of the corresponding
sysfs entry.
This function will be used to correctly account for kobject ownership
changes, e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
-  Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
   - Add comment how ownership of sysfs object is changed.
---
 fs/sysfs/group.c      | 36 ++++++++++++++++++++++++++++++++++++
 include/linux/sysfs.h |  8 ++++++++
 2 files changed, 44 insertions(+)

diff --git a/fs/sysfs/group.c b/fs/sysfs/group.c
index 7d158d59d097..0fc7b119a5dd 100644
--- a/fs/sysfs/group.c
+++ b/fs/sysfs/group.c
@@ -550,3 +550,39 @@ int sysfs_group_change_owner(struct kobject *kobj,
 	return error;
 }
 EXPORT_SYMBOL_GPL(sysfs_group_change_owner);
+
+/**
+ * sysfs_groups_change_owner - change owner of a set of attribute groups.
+ * @kobj:	The kobject containing the groups.
+ * @groups:	The attribute groups.
+ *
+ * To change the ownership of a sysfs object, the caller must first change the
+ * uid/gid of the kobject and then call this function. Usually this will be
+ * taken care of by the relevant subsystem, e.g. moving a network device
+ * between network namespaces owned by different user namespaces will change
+ * the uid/gid of the kobject to the uid/gid of the root user in the user
+ * namespace. Calling this function afterwards will cause the sysfs object to
+ * reflect the new uid/gid.
+ *
+ * Returns 0 on success or error code on failure.
+ */
+int sysfs_groups_change_owner(struct kobject *kobj,
+			      const struct attribute_group **groups)
+{
+	int error = 0, i;
+
+	if (!kobj->state_in_sysfs)
+		return -EINVAL;
+
+	if (!groups)
+		return 0;
+
+	for (i = 0; groups[i]; i++) {
+		error = sysfs_group_change_owner(kobj, groups[i]);
+		if (error)
+			break;
+	}
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(sysfs_groups_change_owner);
diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
index dd211f0a654b..d13c6cc8e487 100644
--- a/include/linux/sysfs.h
+++ b/include/linux/sysfs.h
@@ -314,6 +314,8 @@ int sysfs_file_change_owner(struct kobject *kobj);
 int sysfs_file_change_owner_by_name(struct kobject *kobj, const char *name);
 int sysfs_link_change_owner(struct kobject *kobj, struct kobject *targ,
 			    const char *name);
+int sysfs_groups_change_owner(struct kobject *kobj,
+			      const struct attribute_group **groups);
 int sysfs_group_change_owner(struct kobject *kobj,
 			     const struct attribute_group *groups);
 
@@ -547,6 +549,12 @@ static inline int sysfs_link_change_owner(struct kobject *kobj,
 	return 0;
 }
 
+static inline int sysfs_groups_change_owner(struct kobject *kobj,
+			  const struct attribute_group **groups)
+{
+	return 0;
+}
+
 static inline int sysfs_group_change_owner(struct kobject *kobj,
 			 const struct attribute_group **groups)
 {
-- 
2.25.0

[PATCH net-next v2 05/10] sysfs: add sysfs_change_owner()

[Christian Brauner]

Add a helper to change the owner of sysfs objects.
The ownership of a sysfs object is determined based on the ownership of
the corresponding kobject, i.e. only if the ownership of a kobject is
changed will this function change the ownership of the corresponding
sysfs entry.
This function will be used to correctly account for kobject ownership
changes, e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
-  Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
   - Add comment how ownership of sysfs object is changed.
---
 fs/sysfs/file.c       | 43 +++++++++++++++++++++++++++++++++++++++++++
 include/linux/sysfs.h |  6 ++++++
 2 files changed, 49 insertions(+)

diff --git a/fs/sysfs/file.c b/fs/sysfs/file.c
index 8b20245f359c..f88b8bf2fcc3 100644
--- a/fs/sysfs/file.c
+++ b/fs/sysfs/file.c
@@ -686,3 +686,46 @@ int sysfs_file_change_owner(struct kobject *kobj)
 	return error;
 }
 EXPORT_SYMBOL_GPL(sysfs_file_change_owner);
+
+/**
+ *	sysfs_change_owner - change owner of the given object.
+ *	@kobj:	object.
+ *
+ * To change the ownership of a sysfs object, the caller must first change the
+ * uid/gid of the kobject and then call this function. Usually this will be
+ * taken care of by the relevant subsystem, e.g. moving a network device
+ * between network namespaces owned by different user namespaces will change
+ * the uid/gid of the kobject to the uid/gid of the root user in the user
+ * namespace. Calling this function afterwards will cause the sysfs object to
+ * reflect the new uid/gid.
+ */
+int sysfs_change_owner(struct kobject *kobj)
+{
+	int error;
+	const struct kobj_type *ktype;
+
+	if (!kobj->state_in_sysfs)
+		return -EINVAL;
+
+	error = sysfs_file_change_owner(kobj);
+	if (error)
+		return error;
+
+	ktype = get_ktype(kobj);
+	if (ktype) {
+		struct attribute **kattr;
+
+		for (kattr = ktype->default_attrs; kattr && *kattr; kattr++) {
+			error = sysfs_file_change_owner_by_name(kobj, (*kattr)->name);
+			if (error)
+				return error;
+		}
+
+		error = sysfs_groups_change_owner(kobj, ktype->default_groups);
+		if (error)
+			return error;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(sysfs_change_owner);
diff --git a/include/linux/sysfs.h b/include/linux/sysfs.h
index d13c6cc8e487..ce5165a111bc 100644
--- a/include/linux/sysfs.h
+++ b/include/linux/sysfs.h
@@ -312,6 +312,7 @@ static inline void sysfs_enable_ns(struct kernfs_node *kn)
 
 int sysfs_file_change_owner(struct kobject *kobj);
 int sysfs_file_change_owner_by_name(struct kobject *kobj, const char *name);
+int sysfs_change_owner(struct kobject *kobj);
 int sysfs_link_change_owner(struct kobject *kobj, struct kobject *targ,
 			    const char *name);
 int sysfs_groups_change_owner(struct kobject *kobj,
@@ -549,6 +550,11 @@ static inline int sysfs_link_change_owner(struct kobject *kobj,
 	return 0;
 }
 
+static inline int sysfs_change_owner(struct kobject *kobj)
+{
+	return 0;
+}
+
 static inline int sysfs_groups_change_owner(struct kobject *kobj,
 			  const struct attribute_group **groups)
 {
-- 
2.25.0

[PATCH net-next v2 06/10] device: add device_change_owner()

[Christian Brauner]

Add a helper to change the owner of a device's sysfs entries. This
needs to happen when the ownership of a device is changed, e.g. when
moving network devices between network namespaces.
The ownership of a device is determined based on the ownership of
the corresponding kobject, i.e. only if the ownership of a kobject is
changed will this function change the ownership of the corresponding
device's sysfs entries.
This function will be used to correctly account for ownership changes,
e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
unchanged
---
 drivers/base/core.c    | 73 ++++++++++++++++++++++++++++++++++++++++++
 include/linux/device.h |  1 +
 2 files changed, 74 insertions(+)

diff --git a/drivers/base/core.c b/drivers/base/core.c
index 42a672456432..915766c20a47 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -3458,6 +3458,79 @@ int device_move(struct device *dev, struct device *new_parent,
 }
 EXPORT_SYMBOL_GPL(device_move);
 
+static int device_attrs_change_owner(struct device *dev)
+{
+	struct kobject *kobj = &dev->kobj;
+	struct class *class = dev->class;
+	const struct device_type *type = dev->type;
+	int error;
+
+	if (class) {
+		error = sysfs_groups_change_owner(kobj, class->dev_groups);
+		if (error)
+			return error;
+	}
+
+	if (type) {
+		error = sysfs_groups_change_owner(kobj, type->groups);
+		if (error)
+			return error;
+	}
+
+	error = sysfs_groups_change_owner(kobj, dev->groups);
+	if (error)
+		return error;
+
+	if (device_supports_offline(dev) && !dev->offline_disabled) {
+		error = sysfs_file_change_owner_by_name(kobj, dev_attr_online.attr.name);
+		if (error)
+			return error;
+	}
+
+	return 0;
+}
+
+/**
+ * device_change_owner - change the owner of an existing device.
+ * @dev: device.
+ */
+int device_change_owner(struct device *dev)
+{
+	int error;
+	struct kobject *kobj = &dev->kobj;
+
+	dev = get_device(dev);
+	if (!dev)
+		return -EINVAL;
+
+	error = sysfs_change_owner(kobj);
+	if (error)
+		goto out;
+
+	error = sysfs_file_change_owner_by_name(kobj, dev_attr_uevent.attr.name);
+	if (error)
+		goto out;
+
+	error = device_attrs_change_owner(dev);
+	if (error)
+		goto out;
+
+#ifdef CONFIG_BLOCK
+	if (sysfs_deprecated && dev->class == &block_class)
+		goto out;
+#endif
+
+	error = sysfs_link_change_owner(&dev->class->p->subsys.kobj, &dev->kobj,
+					dev_name(dev));
+	if (error)
+		goto out;
+
+out:
+	put_device(dev);
+	return error;
+}
+EXPORT_SYMBOL_GPL(device_change_owner);
+
 /**
  * device_shutdown - call ->shutdown() on each device to shutdown.
  */
diff --git a/include/linux/device.h b/include/linux/device.h
index 0cd7c647c16c..63ebd9e3f6b0 100644
--- a/include/linux/device.h
+++ b/include/linux/device.h
@@ -817,6 +817,7 @@ extern struct device *device_find_child_by_name(struct device *parent,
 extern int device_rename(struct device *dev, const char *new_name);
 extern int device_move(struct device *dev, struct device *new_parent,
 		       enum dpm_order dpm_order);
+extern int device_change_owner(struct device *dev);
 extern const char *device_get_devnode(struct device *dev,
 				      umode_t *mode, kuid_t *uid, kgid_t *gid,
 				      const char **tmp);
-- 
2.25.0

[PATCH net-next v2 07/10] drivers/base/power: add dpm_sysfs_change_owner()

[Christian Brauner]

Add a helper to change the owner of a device's power entries. This
needs to happen when the ownership of a device is changed, e.g. when
moving network devices between network namespaces.
The ownership of a device's power entries is determined based on the
ownership of the corresponding kobject, i.e. only if the ownership of a
kobject is changed will this function change the ownership of the
corresponding sysfs entries.
This function will be used to correctly account for ownership changes,
e.g. when moving network devices between network namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
- "Rafael J. Wysocki" <rafael@kernel.org>:
  -  Fold if (dev->power.wakeup && dev->power.wakeup->dev) check into
     if (device_can_wakeup(dev)) check since the former can never be true if
     the latter is false.

- Christian Brauner <christian.brauner@ubuntu.com>:
  - Place (dev->power.wakeup && dev->power.wakeup->dev) check under
    CONFIG_PM_SLEEP ifdefine since it will wakeup_source will only be available
    when this config option is set.
---
 drivers/base/core.c        |  4 ++++
 drivers/base/power/power.h |  2 ++
 drivers/base/power/sysfs.c | 40 ++++++++++++++++++++++++++++++++++++++
 3 files changed, 46 insertions(+)

diff --git a/drivers/base/core.c b/drivers/base/core.c
index 915766c20a47..74ca07f9eb2b 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -3515,6 +3515,10 @@ int device_change_owner(struct device *dev)
 	if (error)
 		goto out;
 
+	error = dpm_sysfs_change_owner(dev);
+	if (error)
+		goto out;
+
 #ifdef CONFIG_BLOCK
 	if (sysfs_deprecated && dev->class == &block_class)
 		goto out;
diff --git a/drivers/base/power/power.h b/drivers/base/power/power.h
index 444f5c169a0b..f68490d0811b 100644
--- a/drivers/base/power/power.h
+++ b/drivers/base/power/power.h
@@ -74,6 +74,7 @@ extern int pm_qos_sysfs_add_flags(struct device *dev);
 extern void pm_qos_sysfs_remove_flags(struct device *dev);
 extern int pm_qos_sysfs_add_latency_tolerance(struct device *dev);
 extern void pm_qos_sysfs_remove_latency_tolerance(struct device *dev);
+extern int dpm_sysfs_change_owner(struct device *dev);
 
 #else /* CONFIG_PM */
 
@@ -88,6 +89,7 @@ static inline void pm_runtime_remove(struct device *dev) {}
 
 static inline int dpm_sysfs_add(struct device *dev) { return 0; }
 static inline void dpm_sysfs_remove(struct device *dev) {}
+static inline int dpm_sysfs_change_owner(struct device *dev) { return 0; }
 
 #endif
 
diff --git a/drivers/base/power/sysfs.c b/drivers/base/power/sysfs.c
index d7d82db2e4bc..a18ceafd921e 100644
--- a/drivers/base/power/sysfs.c
+++ b/drivers/base/power/sysfs.c
@@ -684,6 +684,46 @@ int dpm_sysfs_add(struct device *dev)
 	return rc;
 }
 
+int dpm_sysfs_change_owner(struct device *dev)
+{
+	int rc;
+
+	if (device_pm_not_required(dev))
+		return 0;
+
+	rc = sysfs_group_change_owner(&dev->kobj, &pm_attr_group);
+	if (rc)
+		return rc;
+
+	if (pm_runtime_callbacks_present(dev)) {
+		rc = sysfs_group_change_owner(&dev->kobj,
+					      &pm_runtime_attr_group);
+		if (rc)
+			return rc;
+	}
+	if (device_can_wakeup(dev)) {
+		rc = sysfs_group_change_owner(&dev->kobj,
+					      &pm_wakeup_attr_group);
+		if (rc)
+			return rc;
+
+#ifdef CONFIG_PM_SLEEP
+		if (dev->power.wakeup && dev->power.wakeup->dev) {
+			rc = device_change_owner(dev->power.wakeup->dev);
+			if (rc)
+				return rc;
+		}
+#endif
+	}
+	if (dev->power.set_latency_tolerance) {
+		rc = sysfs_group_change_owner(&dev->kobj,
+				&pm_qos_latency_tolerance_attr_group);
+		if (rc)
+			return rc;
+	}
+	return 0;
+}
+
 int wakeup_sysfs_add(struct device *dev)
 {
 	return sysfs_merge_group(&dev->kobj, &pm_wakeup_attr_group);
-- 
2.25.0

[PATCH net-next v2 08/10] net-sysfs: add netdev_change_owner()

[Christian Brauner]

Add a function to change the owner of a network device when it is moved
between network namespaces.

Currently, when moving network devices between network namespaces the
ownership of the corresponding sysfs entries is not changed. This leads
to problems when tools try to operate on the corresponding sysfs files.
This leads to a bug whereby a network device that is created in a
network namespaces owned by a user namespace will have its corresponding
sysfs entry owned by the root user of the corresponding user namespace.
If such a network device has to be moved back to the host network
namespace the permissions will still be set to the user namespaces. This
means unprivileged users can e.g. trigger uevents for such incorrectly
owned devices. They can also modify the settings of the device itself.
Both of these things are unwanted.

For example, workloads will create network devices in the host network
namespace. Other tools will then proceed to move such devices between
network namespaces owner by other user namespaces. While the ownership
of the device itself is updated in
net/core/net-sysfs.c:dev_change_net_namespace() the corresponding sysfs
entry for the device is not:

drwxr-xr-x 5 nobody nobody    0 Jan 25 18:08 .
drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_assign_type
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_len
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 address
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 broadcast
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_changes
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_down_count
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_up_count
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_id
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_port
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dormant
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 duplex
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 flags
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 gro_flush_timeout
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifalias
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifindex
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 iflink
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 link_mode
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 mtu
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 name_assign_type
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 netdev_group
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 operstate
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_id
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_name
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_switch_id
drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 power
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 proto_down
drwxr-xr-x 4 nobody nobody    0 Jan 25 18:09 queues
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 speed
drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 statistics
lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:08 subsystem -> ../../../../class/net
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 tx_queue_len
-r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 type
-rw-r--r-- 1 nobody nobody 4096 Jan 25 18:08 uevent

However, if a device is created directly in the network namespace then
the device's sysfs permissions will be correctly updated:

drwxr-xr-x 5 root   root      0 Jan 25 18:12 .
drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
-r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_assign_type
-r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_len
-r--r--r-- 1 root   root   4096 Jan 25 18:12 address
-r--r--r-- 1 root   root   4096 Jan 25 18:12 broadcast
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 carrier
-r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_changes
-r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_down_count
-r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_up_count
-r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_id
-r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_port
-r--r--r-- 1 root   root   4096 Jan 25 18:12 dormant
-r--r--r-- 1 root   root   4096 Jan 25 18:12 duplex
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 flags
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 gro_flush_timeout
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 ifalias
-r--r--r-- 1 root   root   4096 Jan 25 18:12 ifindex
-r--r--r-- 1 root   root   4096 Jan 25 18:12 iflink
-r--r--r-- 1 root   root   4096 Jan 25 18:12 link_mode
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 mtu
-r--r--r-- 1 root   root   4096 Jan 25 18:12 name_assign_type
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 netdev_group
-r--r--r-- 1 root   root   4096 Jan 25 18:12 operstate
-r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_id
-r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_name
-r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_switch_id
drwxr-xr-x 2 root   root      0 Jan 25 18:12 power
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 proto_down
drwxr-xr-x 4 root   root      0 Jan 25 18:12 queues
-r--r--r-- 1 root   root   4096 Jan 25 18:12 speed
drwxr-xr-x 2 root   root      0 Jan 25 18:12 statistics
lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:12 subsystem -> ../../../../class/net
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 tx_queue_len
-r--r--r-- 1 root   root   4096 Jan 25 18:12 type
-rw-r--r-- 1 root   root   4096 Jan 25 18:12 uevent

Now, when creating a network device in a network namespace owned by a
user namespace and moving it to the host the permissions will be set to
the id that the user namespace root user has been mapped to on the host
leading to all sorts of permission issues:

458752
drwxr-xr-x 5 458752 458752      0 Jan 25 18:12 .
drwxr-xr-x 9 root   root        0 Jan 25 18:08 ..
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_assign_type
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_len
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 address
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 broadcast
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_changes
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_down_count
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_up_count
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_id
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_port
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dormant
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 duplex
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 flags
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 gro_flush_timeout
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 ifalias
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 ifindex
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 iflink
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 link_mode
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 mtu
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 name_assign_type
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 netdev_group
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 operstate
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_id
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_name
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_switch_id
drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 power
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 proto_down
drwxr-xr-x 4 458752 458752      0 Jan 25 18:12 queues
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 speed
drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 statistics
lrwxrwxrwx 1 root   root        0 Jan 25 18:12 subsystem -> ../../../../class/net
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 tx_queue_len
-r--r--r-- 1 458752 458752   4096 Jan 25 18:12 type
-rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 uevent

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
unchanged
---
 net/core/net-sysfs.c | 27 +++++++++++++++++++++++++++
 net/core/net-sysfs.h |  2 ++
 2 files changed, 29 insertions(+)

diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index 4c826b8bf9b1..4fda021edf6d 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -1767,6 +1767,33 @@ int netdev_register_kobject(struct net_device *ndev)
 	return error;
 }
 
+/* Change owner for sysfs entries when moving network devices across network
+ * namespaces owned by different user namespaces.
+ */
+int netdev_change_owner(struct net_device *ndev, const struct net *net_old,
+			const struct net *net_new)
+{
+	struct device *dev = &ndev->dev;
+	kuid_t old_uid, new_uid;
+	kgid_t old_gid, new_gid;
+	int error;
+
+	net_ns_get_ownership(net_old, &old_uid, &old_gid);
+	net_ns_get_ownership(net_new, &new_uid, &new_gid);
+
+	/* The network namespace was changed but the owning user namespace is
+	 * identical so there's no need to change the owner of sysfs entries.
+	 */
+	if (uid_eq(old_uid, new_uid) && gid_eq(old_gid, new_gid))
+		return 0;
+
+	error = device_change_owner(dev);
+	if (error)
+		return error;
+
+	return 0;
+}
+
 int netdev_class_create_file_ns(const struct class_attribute *class_attr,
 				const void *ns)
 {
diff --git a/net/core/net-sysfs.h b/net/core/net-sysfs.h
index 006876c7b78d..8a5b04c2699a 100644
--- a/net/core/net-sysfs.h
+++ b/net/core/net-sysfs.h
@@ -8,5 +8,7 @@ void netdev_unregister_kobject(struct net_device *);
 int net_rx_queue_update_kobjects(struct net_device *, int old_num, int new_num);
 int netdev_queue_update_kobjects(struct net_device *net,
 				 int old_num, int new_num);
+int netdev_change_owner(struct net_device *, const struct net *net_old,
+			const struct net *net_new);
 
 #endif
-- 
2.25.0

[PATCH net-next v2 09/10] net-sysfs: add queue_change_owner()

[Christian Brauner]

Add a function to change the owner of the queue entries for a network device
when it is moved between network namespaces.

Currently, when moving network devices between network namespaces the
ownership of the corresponding queue sysfs entries are not changed. This leads
to problems when tools try to operate on the corresponding sysfs files. Fix
this.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
- kbuild test robot <lkp@intel.com> via sparse:
  - Make net_rx_queue_change_owner() static since it's not exported.
---
 net/core/net-sysfs.c | 101 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 101 insertions(+)

diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index 4fda021edf6d..3f93f996c7a3 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -944,6 +944,22 @@ static int rx_queue_add_kobject(struct net_device *dev, int index)
 	kobject_put(kobj);
 	return error;
 }
+
+static int rx_queue_change_owner(struct net_device *dev, int index)
+{
+	struct netdev_rx_queue *queue = dev->_rx + index;
+	struct kobject *kobj = &queue->kobj;
+	int error;
+
+	error = sysfs_change_owner(kobj);
+	if (error)
+		return error;
+
+	if (dev->sysfs_rx_queue_group)
+		error = sysfs_group_change_owner(kobj, dev->sysfs_rx_queue_group);
+
+	return error;
+}
 #endif /* CONFIG_SYSFS */
 
 int
@@ -981,6 +997,28 @@ net_rx_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
 #endif
 }
 
+static int net_rx_queue_change_owner(struct net_device *dev, int num)
+{
+#ifdef CONFIG_SYSFS
+	int error = 0;
+	int i;
+
+#ifndef CONFIG_RPS
+	if (!dev->sysfs_rx_queue_group)
+		return 0;
+#endif
+	for (i = 0; i < num; i++) {
+		error = rx_queue_change_owner(dev, i);
+		if (error)
+			break;
+	}
+
+	return error;
+#else
+	return 0;
+#endif
+}
+
 #ifdef CONFIG_SYSFS
 /*
  * netdev_queue sysfs structures and functions.
@@ -1486,6 +1524,22 @@ static int netdev_queue_add_kobject(struct net_device *dev, int index)
 	kobject_put(kobj);
 	return error;
 }
+
+static int tx_queue_change_owner(struct net_device *ndev, int index)
+{
+	struct netdev_queue *queue = ndev->_tx + index;
+	struct kobject *kobj = &queue->kobj;
+	int error;
+
+	error = sysfs_change_owner(kobj);
+	if (error)
+		return error;
+
+#ifdef CONFIG_BQL
+	error = sysfs_group_change_owner(kobj, &dql_group);
+#endif
+	return error;
+}
 #endif /* CONFIG_SYSFS */
 
 int
@@ -1520,6 +1574,24 @@ netdev_queue_update_kobjects(struct net_device *dev, int old_num, int new_num)
 #endif /* CONFIG_SYSFS */
 }
 
+static int net_tx_queue_change_owner(struct net_device *dev, int num)
+{
+#ifdef CONFIG_SYSFS
+	int error = 0;
+	int i;
+
+	for (i = 0; i < num; i++) {
+		error = tx_queue_change_owner(dev, i);
+		if (error)
+			break;
+	}
+
+	return error;
+#else
+	return 0;
+#endif /* CONFIG_SYSFS */
+}
+
 static int register_queue_kobjects(struct net_device *dev)
 {
 	int error = 0, txq = 0, rxq = 0, real_rx = 0, real_tx = 0;
@@ -1554,6 +1626,31 @@ static int register_queue_kobjects(struct net_device *dev)
 	return error;
 }
 
+static int queue_change_owner(struct net_device *ndev)
+{
+	int error = 0, real_rx = 0, real_tx = 0;
+
+#ifdef CONFIG_SYSFS
+	if (ndev->queues_kset) {
+		error = sysfs_change_owner(&ndev->queues_kset->kobj);
+		if (error)
+			return error;
+	}
+	real_rx = ndev->real_num_rx_queues;
+#endif
+	real_tx = ndev->real_num_tx_queues;
+
+	error = net_rx_queue_change_owner(ndev, real_rx);
+	if (error)
+		return error;
+
+	error = net_tx_queue_change_owner(ndev, real_tx);
+	if (error)
+		return error;
+
+	return 0;
+}
+
 static void remove_queue_kobjects(struct net_device *dev)
 {
 	int real_rx = 0, real_tx = 0;
@@ -1791,6 +1888,10 @@ int netdev_change_owner(struct net_device *ndev, const struct net *net_old,
 	if (error)
 		return error;
 
+	error = queue_change_owner(ndev);
+	if (error)
+		return error;
+
 	return 0;
 }
 
-- 
2.25.0

[PATCH net-next v2 10/10] net: fix sysfs permssions when device changes network namespace

[Christian Brauner]

Now that we moved all the helpers in place and make use netdev_change_owner()
to fixup the permissions when moving network devices between network
namespaces.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
---
/* v2 */
unchanged
---
 net/core/dev.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index a69e8bd7ed74..0f9c4684fcbd 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -10016,6 +10016,7 @@ EXPORT_SYMBOL(unregister_netdev);
 
 int dev_change_net_namespace(struct net_device *dev, struct net *net, const char *pat)
 {
+	struct net *net_old = dev_net(dev);
 	int err, new_nsid, new_ifindex;
 
 	ASSERT_RTNL();
@@ -10031,7 +10032,7 @@ int dev_change_net_namespace(struct net_device *dev, struct net *net, const char
 
 	/* Get out if there is nothing todo */
 	err = 0;
-	if (net_eq(dev_net(dev), net))
+	if (net_eq(net_old, net))
 		goto out;
 
 	/* Pick the destination device name, and ensure
@@ -10107,6 +10108,12 @@ int dev_change_net_namespace(struct net_device *dev, struct net *net, const char
 	err = device_rename(&dev->dev, dev->name);
 	WARN_ON(err);
 
+	/* Adapt owner in case owning user namespace of target network
+	 * namespace is different from the original one.
+	 */
+	err = netdev_change_owner(dev, net_old, net);
+	WARN_ON(err);
+
 	/* Add the device back in the hashes */
 	list_netdevice(dev);
 
-- 
2.25.0

Re: [PATCH net-next v2 08/10] net-sysfs: add netdev_change_owner()

[Greg Kroah-Hartman]

On Mon, Feb 17, 2020 at 05:14:34PM +0100, Christian Brauner wrote:
> Add a function to change the owner of a network device when it is moved
> between network namespaces.
> 
> Currently, when moving network devices between network namespaces the
> ownership of the corresponding sysfs entries is not changed. This leads
> to problems when tools try to operate on the corresponding sysfs files.
> This leads to a bug whereby a network device that is created in a
> network namespaces owned by a user namespace will have its corresponding
> sysfs entry owned by the root user of the corresponding user namespace.
> If such a network device has to be moved back to the host network
> namespace the permissions will still be set to the user namespaces. This
> means unprivileged users can e.g. trigger uevents for such incorrectly
> owned devices. They can also modify the settings of the device itself.
> Both of these things are unwanted.
> 
> For example, workloads will create network devices in the host network
> namespace. Other tools will then proceed to move such devices between
> network namespaces owner by other user namespaces. While the ownership
> of the device itself is updated in
> net/core/net-sysfs.c:dev_change_net_namespace() the corresponding sysfs
> entry for the device is not:
> 
> drwxr-xr-x 5 nobody nobody    0 Jan 25 18:08 .
> drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_assign_type
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_len
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 address
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 broadcast
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_changes
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_down_count
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_up_count
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_id
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_port
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dormant
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 duplex
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 flags
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 gro_flush_timeout
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifalias
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifindex
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 iflink
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 link_mode
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 mtu
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 name_assign_type
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 netdev_group
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 operstate
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_id
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_name
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_switch_id
> drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 power
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 proto_down
> drwxr-xr-x 4 nobody nobody    0 Jan 25 18:09 queues
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 speed
> drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 statistics
> lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:08 subsystem -> ../../../../class/net
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 tx_queue_len
> -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 type
> -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:08 uevent
> 
> However, if a device is created directly in the network namespace then
> the device's sysfs permissions will be correctly updated:
> 
> drwxr-xr-x 5 root   root      0 Jan 25 18:12 .
> drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_assign_type
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_len
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 address
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 broadcast
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 carrier
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_changes
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_down_count
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_up_count
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_id
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_port
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 dormant
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 duplex
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 flags
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 gro_flush_timeout
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 ifalias
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 ifindex
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 iflink
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 link_mode
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 mtu
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 name_assign_type
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 netdev_group
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 operstate
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_id
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_name
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_switch_id
> drwxr-xr-x 2 root   root      0 Jan 25 18:12 power
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 proto_down
> drwxr-xr-x 4 root   root      0 Jan 25 18:12 queues
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 speed
> drwxr-xr-x 2 root   root      0 Jan 25 18:12 statistics
> lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:12 subsystem -> ../../../../class/net
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 tx_queue_len
> -r--r--r-- 1 root   root   4096 Jan 25 18:12 type
> -rw-r--r-- 1 root   root   4096 Jan 25 18:12 uevent
> 
> Now, when creating a network device in a network namespace owned by a
> user namespace and moving it to the host the permissions will be set to
> the id that the user namespace root user has been mapped to on the host
> leading to all sorts of permission issues:
> 
> 458752
> drwxr-xr-x 5 458752 458752      0 Jan 25 18:12 .
> drwxr-xr-x 9 root   root        0 Jan 25 18:08 ..
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_assign_type
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_len
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 address
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 broadcast
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_changes
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_down_count
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_up_count
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_id
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_port
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dormant
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 duplex
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 flags
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 gro_flush_timeout
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 ifalias
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 ifindex
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 iflink
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 link_mode
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 mtu
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 name_assign_type
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 netdev_group
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 operstate
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_id
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_name
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_switch_id
> drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 power
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 proto_down
> drwxr-xr-x 4 458752 458752      0 Jan 25 18:12 queues
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 speed
> drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 statistics
> lrwxrwxrwx 1 root   root        0 Jan 25 18:12 subsystem -> ../../../../class/net
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 tx_queue_len
> -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 type
> -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 uevent
> 
> Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
> ---
> /* v2 */
> unchanged
> ---
>  net/core/net-sysfs.c | 27 +++++++++++++++++++++++++++
>  net/core/net-sysfs.h |  2 ++
>  2 files changed, 29 insertions(+)
> 
> diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
> index 4c826b8bf9b1..4fda021edf6d 100644
> --- a/net/core/net-sysfs.c
> +++ b/net/core/net-sysfs.c
> @@ -1767,6 +1767,33 @@ int netdev_register_kobject(struct net_device *ndev)
>  	return error;
>  }
>  
> +/* Change owner for sysfs entries when moving network devices across network
> + * namespaces owned by different user namespaces.
> + */
> +int netdev_change_owner(struct net_device *ndev, const struct net *net_old,
> +			const struct net *net_new)
> +{
> +	struct device *dev = &ndev->dev;
> +	kuid_t old_uid, new_uid;
> +	kgid_t old_gid, new_gid;
> +	int error;
> +
> +	net_ns_get_ownership(net_old, &old_uid, &old_gid);
> +	net_ns_get_ownership(net_new, &new_uid, &new_gid);
> +
> +	/* The network namespace was changed but the owning user namespace is
> +	 * identical so there's no need to change the owner of sysfs entries.
> +	 */
> +	if (uid_eq(old_uid, new_uid) && gid_eq(old_gid, new_gid))
> +		return 0;
> +
> +	error = device_change_owner(dev);

Ok, maybe I'm slow here, but what actually changed the gid/uid here?
How did it change?  All you did was look up the old uid/gid and the new
uid/gid.  But what set the device to the new one?

All of these functions are just really odd to me, one would think that a
"change owner" function would have the new owner in the paramter to know
what to change it to?  Your documentation says "owner must be changed
before calling this function", but how did that get changed and who
changed it?

Why not just pass it as part of the function itself?

Otherwise it looks really odd, like the above call.  As I can't see how
anything changes here at all by reading this code.  And that's a huge
sign of a bad API, when the maintainer of the subsystem can not even
understand how someone is using it with a single function call :)

thanks,

greg k-h

Re: [PATCH net-next v2 01/10] sysfs: add sysfs_file_change_owner{_by_name}()

[Greg Kroah-Hartman]

On Mon, Feb 17, 2020 at 05:14:27PM +0100, Christian Brauner wrote:
> Add helpers to change owner of a sysfs files.
> The ownership of a sysfs object is determined based on the ownership of
> the corresponding kobject, i.e. only if the ownership of a kobject is
> changed will this function change the ownership of the corresponding
> sysfs entry.
> This function will be used to correctly account for kobject ownership
> changes, e.g. when moving network devices between network namespaces.
> 
> Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
> ---
> /* v2 */
> -  Greg Kroah-Hartman <gregkh@linuxfoundation.org>:
>    - Better naming for sysfs_file_change_owner() to reflect the fact that it
>      can be used to change the owner of the kobject itself by passing NULL as
>      argument.
> - Christian Brauner <christian.brauner@ubuntu.com>:
>   - Split sysfs_file_change_owner() into two helpers sysfs_change_owner() and
>     sysfs_change_owner_by_name(). The former changes the owner of the kobject
>     itself, the latter the owner of the kobject looked up via the name
>     argument.
> ---
>  fs/sysfs/file.c       | 82 +++++++++++++++++++++++++++++++++++++++++++
>  include/linux/sysfs.h | 14 ++++++++
>  2 files changed, 96 insertions(+)
> 
> diff --git a/fs/sysfs/file.c b/fs/sysfs/file.c
> index 130fc6fbcc03..8f2607de2456 100644
> --- a/fs/sysfs/file.c
> +++ b/fs/sysfs/file.c
> @@ -558,3 +558,85 @@ void sysfs_remove_bin_file(struct kobject *kobj,
>  	kernfs_remove_by_name(kobj->sd, attr->attr.name);
>  }
>  EXPORT_SYMBOL_GPL(sysfs_remove_bin_file);
> +
> +static int internal_change_owner(struct kernfs_node *kn, struct kobject *kobj)
> +{
> +	kuid_t uid;
> +	kgid_t gid;
> +	struct iattr newattrs = {
> +		.ia_valid = ATTR_UID | ATTR_GID,
> +	};
> +
> +	kobject_get_ownership(kobj, &uid, &gid);
> +	newattrs.ia_uid = uid;
> +	newattrs.ia_gid = gid;
> +
> +	return kernfs_setattr(kn, &newattrs);
> +}
> +
> +/**
> + *	sysfs_file_change_owner_by_name - change owner of a file.
> + *	@kobj:	object.
> + *	@name:	name of the file to change.
> + *
> + * To change the ownership of a sysfs object, the caller must first change the
> + * uid/gid of the kobject and then call this function.

Why have the caller do this?  Why not pass the uid/gid as a parameter
here?  That would make it totally obvious as to what is happening here,
right?

Otherwise this function is depending on someone doing something before
calling it, and that's going to be a very very hard thing to always
ensure/audit.

thanks,

greg k-h

Re: [PATCH net-next v2 08/10] net-sysfs: add netdev_change_owner()

[Christian Brauner]

On Mon, Feb 17, 2020 at 05:28:03PM +0100, Greg Kroah-Hartman wrote:
> On Mon, Feb 17, 2020 at 05:14:34PM +0100, Christian Brauner wrote:
> > Add a function to change the owner of a network device when it is moved
> > between network namespaces.
> > 
> > Currently, when moving network devices between network namespaces the
> > ownership of the corresponding sysfs entries is not changed. This leads
> > to problems when tools try to operate on the corresponding sysfs files.
> > This leads to a bug whereby a network device that is created in a
> > network namespaces owned by a user namespace will have its corresponding
> > sysfs entry owned by the root user of the corresponding user namespace.
> > If such a network device has to be moved back to the host network
> > namespace the permissions will still be set to the user namespaces. This
> > means unprivileged users can e.g. trigger uevents for such incorrectly
> > owned devices. They can also modify the settings of the device itself.
> > Both of these things are unwanted.
> > 
> > For example, workloads will create network devices in the host network
> > namespace. Other tools will then proceed to move such devices between
> > network namespaces owner by other user namespaces. While the ownership
> > of the device itself is updated in
> > net/core/net-sysfs.c:dev_change_net_namespace() the corresponding sysfs
> > entry for the device is not:
> > 
> > drwxr-xr-x 5 nobody nobody    0 Jan 25 18:08 .
> > drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_assign_type
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_len
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 address
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 broadcast
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_changes
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_down_count
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_up_count
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_id
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_port
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dormant
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 duplex
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 flags
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 gro_flush_timeout
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifalias
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifindex
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 iflink
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 link_mode
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 mtu
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 name_assign_type
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 netdev_group
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 operstate
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_id
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_name
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_switch_id
> > drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 power
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 proto_down
> > drwxr-xr-x 4 nobody nobody    0 Jan 25 18:09 queues
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 speed
> > drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 statistics
> > lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:08 subsystem -> ../../../../class/net
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 tx_queue_len
> > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 type
> > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:08 uevent
> > 
> > However, if a device is created directly in the network namespace then
> > the device's sysfs permissions will be correctly updated:
> > 
> > drwxr-xr-x 5 root   root      0 Jan 25 18:12 .
> > drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_assign_type
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_len
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 address
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 broadcast
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 carrier
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_changes
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_down_count
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_up_count
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_id
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_port
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 dormant
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 duplex
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 flags
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 gro_flush_timeout
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 ifalias
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 ifindex
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 iflink
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 link_mode
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 mtu
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 name_assign_type
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 netdev_group
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 operstate
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_id
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_name
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_switch_id
> > drwxr-xr-x 2 root   root      0 Jan 25 18:12 power
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 proto_down
> > drwxr-xr-x 4 root   root      0 Jan 25 18:12 queues
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 speed
> > drwxr-xr-x 2 root   root      0 Jan 25 18:12 statistics
> > lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:12 subsystem -> ../../../../class/net
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 tx_queue_len
> > -r--r--r-- 1 root   root   4096 Jan 25 18:12 type
> > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 uevent
> > 
> > Now, when creating a network device in a network namespace owned by a
> > user namespace and moving it to the host the permissions will be set to
> > the id that the user namespace root user has been mapped to on the host
> > leading to all sorts of permission issues:
> > 
> > 458752
> > drwxr-xr-x 5 458752 458752      0 Jan 25 18:12 .
> > drwxr-xr-x 9 root   root        0 Jan 25 18:08 ..
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_assign_type
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_len
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 address
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 broadcast
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_changes
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_down_count
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_up_count
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_id
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_port
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dormant
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 duplex
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 flags
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 gro_flush_timeout
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 ifalias
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 ifindex
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 iflink
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 link_mode
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 mtu
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 name_assign_type
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 netdev_group
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 operstate
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_id
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_name
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_switch_id
> > drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 power
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 proto_down
> > drwxr-xr-x 4 458752 458752      0 Jan 25 18:12 queues
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 speed
> > drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 statistics
> > lrwxrwxrwx 1 root   root        0 Jan 25 18:12 subsystem -> ../../../../class/net
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 tx_queue_len
> > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 type
> > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 uevent
> > 
> > Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
> > ---
> > /* v2 */
> > unchanged
> > ---
> >  net/core/net-sysfs.c | 27 +++++++++++++++++++++++++++
> >  net/core/net-sysfs.h |  2 ++
> >  2 files changed, 29 insertions(+)
> > 
> > diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
> > index 4c826b8bf9b1..4fda021edf6d 100644
> > --- a/net/core/net-sysfs.c
> > +++ b/net/core/net-sysfs.c
> > @@ -1767,6 +1767,33 @@ int netdev_register_kobject(struct net_device *ndev)
> >  	return error;
> >  }
> >  
> > +/* Change owner for sysfs entries when moving network devices across network
> > + * namespaces owned by different user namespaces.
> > + */
> > +int netdev_change_owner(struct net_device *ndev, const struct net *net_old,
> > +			const struct net *net_new)
> > +{
> > +	struct device *dev = &ndev->dev;
> > +	kuid_t old_uid, new_uid;
> > +	kgid_t old_gid, new_gid;
> > +	int error;
> > +
> > +	net_ns_get_ownership(net_old, &old_uid, &old_gid);
> > +	net_ns_get_ownership(net_new, &new_uid, &new_gid);
> > +
> > +	/* The network namespace was changed but the owning user namespace is
> > +	 * identical so there's no need to change the owner of sysfs entries.
> > +	 */
> > +	if (uid_eq(old_uid, new_uid) && gid_eq(old_gid, new_gid))
> > +		return 0;
> > +
> > +	error = device_change_owner(dev);
> 
> Ok, maybe I'm slow here, but what actually changed the gid/uid here?
> How did it change?  All you did was look up the old uid/gid and the new
> uid/gid.  But what set the device to the new one?
> 
> All of these functions are just really odd to me, one would think that a
> "change owner" function would have the new owner in the paramter to know
> what to change it to?  Your documentation says "owner must be changed
> before calling this function", but how did that get changed and who
> changed it?
> 
> Why not just pass it as part of the function itself?
> 
> Otherwise it looks really odd, like the above call.  As I can't see how
> anything changes here at all by reading this code.  And that's a huge
> sign of a bad API, when the maintainer of the subsystem can not even
> understand how someone is using it with a single function call :)

So we've been briefly discussing this in the first iteration of this
series where I also explained why I in the first iteration I didn't add
explicit uid_t and gid_t parameters. I sugggested adding uid_t/gid_t
parameters in that thread but from your response in
https://lore.kernel.org/lkml/20200212160402.GA1799124@kroah.com/:

> > So ownership only changes if the kobject's uid/gid have been changed.
> > So when to stick with the networking example, when a network device is
> > moved into a new network namespace, the uid/gid of the kobject will be
> > changed to the root user of the owning user namespace of that network
> > namespace. So when the move of the network device has completed and
> > kobject_get_ownership() is called it will now return a different
> > uid/gid.
> 
> Ok, then this needs to say "change the uid/gid of the kobject to..." in
> order to explain what it is now being set to.  Otherwise this is really
> confusing if you only read the kerneldoc, right?

From this answer I took it that you did prefer not adding uid/gid
parameters. If that has changed or I misunderstood you then I can change
all these functions.

For clarity, for network namespaces: when the move of a network device
has completed dev_change_net_namespace() is called which changes the
ownership of the kobject associated with the net device. Now
kobject_get_ownership() will return the updated uids/gids. In any case,
asking explicitly: do you want to have those sysfs and device functions
take explicit uid and gid parameters? That sounds nicer to me as well. I
just didn't want to expose it because we alreay had this dynamic way of
setting ownership via kobject_get_ownership(). 

Thanks!
Christian

Re: [PATCH net-next v2 08/10] net-sysfs: add netdev_change_owner()

[Greg Kroah-Hartman]

On Mon, Feb 17, 2020 at 05:58:54PM +0100, Christian Brauner wrote:
> On Mon, Feb 17, 2020 at 05:28:03PM +0100, Greg Kroah-Hartman wrote:
> > On Mon, Feb 17, 2020 at 05:14:34PM +0100, Christian Brauner wrote:
> > > Add a function to change the owner of a network device when it is moved
> > > between network namespaces.
> > > 
> > > Currently, when moving network devices between network namespaces the
> > > ownership of the corresponding sysfs entries is not changed. This leads
> > > to problems when tools try to operate on the corresponding sysfs files.
> > > This leads to a bug whereby a network device that is created in a
> > > network namespaces owned by a user namespace will have its corresponding
> > > sysfs entry owned by the root user of the corresponding user namespace.
> > > If such a network device has to be moved back to the host network
> > > namespace the permissions will still be set to the user namespaces. This
> > > means unprivileged users can e.g. trigger uevents for such incorrectly
> > > owned devices. They can also modify the settings of the device itself.
> > > Both of these things are unwanted.
> > > 
> > > For example, workloads will create network devices in the host network
> > > namespace. Other tools will then proceed to move such devices between
> > > network namespaces owner by other user namespaces. While the ownership
> > > of the device itself is updated in
> > > net/core/net-sysfs.c:dev_change_net_namespace() the corresponding sysfs
> > > entry for the device is not:
> > > 
> > > drwxr-xr-x 5 nobody nobody    0 Jan 25 18:08 .
> > > drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_assign_type
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 addr_len
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 address
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 broadcast
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_changes
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_down_count
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 carrier_up_count
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_id
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dev_port
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 dormant
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 duplex
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 flags
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 gro_flush_timeout
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifalias
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 ifindex
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 iflink
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 link_mode
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 mtu
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 name_assign_type
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 netdev_group
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 operstate
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_id
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_port_name
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 phys_switch_id
> > > drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 power
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 proto_down
> > > drwxr-xr-x 4 nobody nobody    0 Jan 25 18:09 queues
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 speed
> > > drwxr-xr-x 2 nobody nobody    0 Jan 25 18:09 statistics
> > > lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:08 subsystem -> ../../../../class/net
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:09 tx_queue_len
> > > -r--r--r-- 1 nobody nobody 4096 Jan 25 18:09 type
> > > -rw-r--r-- 1 nobody nobody 4096 Jan 25 18:08 uevent
> > > 
> > > However, if a device is created directly in the network namespace then
> > > the device's sysfs permissions will be correctly updated:
> > > 
> > > drwxr-xr-x 5 root   root      0 Jan 25 18:12 .
> > > drwxr-xr-x 9 nobody nobody    0 Jan 25 18:08 ..
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_assign_type
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 addr_len
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 address
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 broadcast
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 carrier
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_changes
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_down_count
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 carrier_up_count
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_id
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 dev_port
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 dormant
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 duplex
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 flags
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 gro_flush_timeout
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 ifalias
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 ifindex
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 iflink
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 link_mode
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 mtu
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 name_assign_type
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 netdev_group
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 operstate
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_id
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_port_name
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 phys_switch_id
> > > drwxr-xr-x 2 root   root      0 Jan 25 18:12 power
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 proto_down
> > > drwxr-xr-x 4 root   root      0 Jan 25 18:12 queues
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 speed
> > > drwxr-xr-x 2 root   root      0 Jan 25 18:12 statistics
> > > lrwxrwxrwx 1 nobody nobody    0 Jan 25 18:12 subsystem -> ../../../../class/net
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 tx_queue_len
> > > -r--r--r-- 1 root   root   4096 Jan 25 18:12 type
> > > -rw-r--r-- 1 root   root   4096 Jan 25 18:12 uevent
> > > 
> > > Now, when creating a network device in a network namespace owned by a
> > > user namespace and moving it to the host the permissions will be set to
> > > the id that the user namespace root user has been mapped to on the host
> > > leading to all sorts of permission issues:
> > > 
> > > 458752
> > > drwxr-xr-x 5 458752 458752      0 Jan 25 18:12 .
> > > drwxr-xr-x 9 root   root        0 Jan 25 18:08 ..
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_assign_type
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 addr_len
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 address
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 broadcast
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_changes
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_down_count
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 carrier_up_count
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_id
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dev_port
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 dormant
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 duplex
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 flags
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 gro_flush_timeout
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 ifalias
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 ifindex
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 iflink
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 link_mode
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 mtu
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 name_assign_type
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 netdev_group
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 operstate
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_id
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_port_name
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 phys_switch_id
> > > drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 power
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 proto_down
> > > drwxr-xr-x 4 458752 458752      0 Jan 25 18:12 queues
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 speed
> > > drwxr-xr-x 2 458752 458752      0 Jan 25 18:12 statistics
> > > lrwxrwxrwx 1 root   root        0 Jan 25 18:12 subsystem -> ../../../../class/net
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 tx_queue_len
> > > -r--r--r-- 1 458752 458752   4096 Jan 25 18:12 type
> > > -rw-r--r-- 1 458752 458752   4096 Jan 25 18:12 uevent
> > > 
> > > Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
> > > ---
> > > /* v2 */
> > > unchanged
> > > ---
> > >  net/core/net-sysfs.c | 27 +++++++++++++++++++++++++++
> > >  net/core/net-sysfs.h |  2 ++
> > >  2 files changed, 29 insertions(+)
> > > 
> > > diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
> > > index 4c826b8bf9b1..4fda021edf6d 100644
> > > --- a/net/core/net-sysfs.c
> > > +++ b/net/core/net-sysfs.c
> > > @@ -1767,6 +1767,33 @@ int netdev_register_kobject(struct net_device *ndev)
> > >  	return error;
> > >  }
> > >  
> > > +/* Change owner for sysfs entries when moving network devices across network
> > > + * namespaces owned by different user namespaces.
> > > + */
> > > +int netdev_change_owner(struct net_device *ndev, const struct net *net_old,
> > > +			const struct net *net_new)
> > > +{
> > > +	struct device *dev = &ndev->dev;
> > > +	kuid_t old_uid, new_uid;
> > > +	kgid_t old_gid, new_gid;
> > > +	int error;
> > > +
> > > +	net_ns_get_ownership(net_old, &old_uid, &old_gid);
> > > +	net_ns_get_ownership(net_new, &new_uid, &new_gid);
> > > +
> > > +	/* The network namespace was changed but the owning user namespace is
> > > +	 * identical so there's no need to change the owner of sysfs entries.
> > > +	 */
> > > +	if (uid_eq(old_uid, new_uid) && gid_eq(old_gid, new_gid))
> > > +		return 0;
> > > +
> > > +	error = device_change_owner(dev);
> > 
> > Ok, maybe I'm slow here, but what actually changed the gid/uid here?
> > How did it change?  All you did was look up the old uid/gid and the new
> > uid/gid.  But what set the device to the new one?
> > 
> > All of these functions are just really odd to me, one would think that a
> > "change owner" function would have the new owner in the paramter to know
> > what to change it to?  Your documentation says "owner must be changed
> > before calling this function", but how did that get changed and who
> > changed it?
> > 
> > Why not just pass it as part of the function itself?
> > 
> > Otherwise it looks really odd, like the above call.  As I can't see how
> > anything changes here at all by reading this code.  And that's a huge
> > sign of a bad API, when the maintainer of the subsystem can not even
> > understand how someone is using it with a single function call :)
> 
> So we've been briefly discussing this in the first iteration of this
> series where I also explained why I in the first iteration I didn't add
> explicit uid_t and gid_t parameters. I sugggested adding uid_t/gid_t
> parameters in that thread but from your response in
> https://lore.kernel.org/lkml/20200212160402.GA1799124@kroah.com/:
> 
> > > So ownership only changes if the kobject's uid/gid have been changed.
> > > So when to stick with the networking example, when a network device is
> > > moved into a new network namespace, the uid/gid of the kobject will be
> > > changed to the root user of the owning user namespace of that network
> > > namespace. So when the move of the network device has completed and
> > > kobject_get_ownership() is called it will now return a different
> > > uid/gid.
> > 
> > Ok, then this needs to say "change the uid/gid of the kobject to..." in
> > order to explain what it is now being set to.  Otherwise this is really
> > confusing if you only read the kerneldoc, right?
> 
> From this answer I took it that you did prefer not adding uid/gid
> parameters. If that has changed or I misunderstood you then I can change
> all these functions.

Sorry for the confusion.  Yes, the documentation is better now, but the
function parameters are still just as bad.  I think you need to add
uid/gid to the parameters to make it obvious what is happening for when
you just read the code, without seeing the function comments as no one
is going to dig them up somewhere else in another file.

> For clarity, for network namespaces: when the move of a network device
> has completed dev_change_net_namespace() is called which changes the
> ownership of the kobject associated with the net device. Now
> kobject_get_ownership() will return the updated uids/gids. In any case,
> asking explicitly: do you want to have those sysfs and device functions
> take explicit uid and gid parameters?

Yes please.

> That sounds nicer to me as well. I just didn't want to expose it
> because we alreay had this dynamic way of setting ownership via
> kobject_get_ownership(). 

I think that's totally not intutive :)

thanks,

greg k-h