Serge E. Hallyn | b9d8b45 | 2010-05-04 21:45:38 -0500 | [diff] [blame] | 1 | Sysfs tagging |
| 2 | ------------- |
| 3 | |
| 4 | (Taken almost verbatim from Eric Biederman's netns tagging patch |
| 5 | commit msg) |
| 6 | |
| 7 | The problem. Network devices show up in sysfs and with the network |
| 8 | namespace active multiple devices with the same name can show up in |
| 9 | the same directory, ouch! |
| 10 | |
| 11 | To avoid that problem and allow existing applications in network |
| 12 | namespaces to see the same interface that is currently presented in |
| 13 | sysfs, sysfs now has tagging directory support. |
| 14 | |
| 15 | By using the network namespace pointers as tags to separate out the |
| 16 | the sysfs directory entries we ensure that we don't have conflicts |
| 17 | in the directories and applications only see a limited set of |
| 18 | the network devices. |
| 19 | |
| 20 | Each sysfs directory entry may be tagged with zero or one |
| 21 | namespaces. A sysfs_dirent is augmented with a void *s_ns. If a |
| 22 | directory entry is tagged, then sysfs_dirent->s_flags will have a |
| 23 | flag between KOBJ_NS_TYPE_NONE and KOBJ_NS_TYPES, and s_ns will |
| 24 | point to the namespace to which it belongs. |
| 25 | |
| 26 | Each sysfs superblock's sysfs_super_info contains an array void |
| 27 | *ns[KOBJ_NS_TYPES]. When a a task in a tagging namespace |
| 28 | kobj_nstype first mounts sysfs, a new superblock is created. It |
| 29 | will be differentiated from other sysfs mounts by having its |
| 30 | s_fs_info->ns[kobj_nstype] set to the new namespace. Note that |
| 31 | through bind mounting and mounts propagation, a task can easily view |
| 32 | the contents of other namespaces' sysfs mounts. Therefore, when a |
| 33 | namespace exits, it will call kobj_ns_exit() to invalidate any |
| 34 | sysfs_dirent->s_ns pointers pointing to it. |
| 35 | |
| 36 | Users of this interface: |
| 37 | - define a type in the kobj_ns_type enumeration. |
| 38 | - call kobj_ns_type_register() with its kobj_ns_type_operations which has |
| 39 | - current_ns() which returns current's namespace |
| 40 | - netlink_ns() which returns a socket's namespace |
| 41 | - initial_ns() which returns the initial namesapce |
| 42 | - call kobj_ns_exit() when an individual tag is no longer valid |