1 21. Introduction 3 4Linux distinguishes between administrative and operational state of an 5interface. Administrative state is the result of "ip link set dev 6<dev> up or down" and reflects whether the administrator wants to use 7the device for traffic. 8 9However, an interface is not usable just because the admin enabled it 10- ethernet requires to be plugged into the switch and, depending on 11a site's networking policy and configuration, an 802.1X authentication 12to be performed before user data can be transferred. Operational state 13shows the ability of an interface to transmit this user data. 14 15Thanks to 802.1X, userspace must be granted the possibility to 16influence operational state. To accommodate this, operational state is 17split into two parts: Two flags that can be set by the driver only, and 18a RFC2863 compatible state that is derived from these flags, a policy, 19and changeable from userspace under certain rules. 20 21 222. Querying from userspace 23 24Both admin and operational state can be queried via the netlink 25operation RTM_GETLINK. It is also possible to subscribe to RTMGRP_LINK 26to be notified of updates. This is important for setting from userspace. 27 28These values contain interface state: 29 30ifinfomsg::if_flags & IFF_UP: 31 Interface is admin up 32ifinfomsg::if_flags & IFF_RUNNING: 33 Interface is in RFC2863 operational state UP or UNKNOWN. This is for 34 backward compatibility, routing daemons, dhcp clients can use this 35 flag to determine whether they should use the interface. 36ifinfomsg::if_flags & IFF_LOWER_UP: 37 Driver has signaled netif_carrier_on() 38ifinfomsg::if_flags & IFF_DORMANT: 39 Driver has signaled netif_dormant_on() 40 41TLV IFLA_OPERSTATE 42 43contains RFC2863 state of the interface in numeric representation: 44 45IF_OPER_UNKNOWN (0): 46 Interface is in unknown state, neither driver nor userspace has set 47 operational state. Interface must be considered for user data as 48 setting operational state has not been implemented in every driver. 49IF_OPER_NOTPRESENT (1): 50 Unused in current kernel (notpresent interfaces normally disappear), 51 just a numerical placeholder. 52IF_OPER_DOWN (2): 53 Interface is unable to transfer data on L1, f.e. ethernet is not 54 plugged or interface is ADMIN down. 55IF_OPER_LOWERLAYERDOWN (3): 56 Interfaces stacked on an interface that is IF_OPER_DOWN show this 57 state (f.e. VLAN). 58IF_OPER_TESTING (4): 59 Unused in current kernel. 60IF_OPER_DORMANT (5): 61 Interface is L1 up, but waiting for an external event, f.e. for a 62 protocol to establish. (802.1X) 63IF_OPER_UP (6): 64 Interface is operational up and can be used. 65 66This TLV can also be queried via sysfs. 67 68TLV IFLA_LINKMODE 69 70contains link policy. This is needed for userspace interaction 71described below. 72 73This TLV can also be queried via sysfs. 74 75 763. Kernel driver API 77 78Kernel drivers have access to two flags that map to IFF_LOWER_UP and 79IFF_DORMANT. These flags can be set from everywhere, even from 80interrupts. It is guaranteed that only the driver has write access, 81however, if different layers of the driver manipulate the same flag, 82the driver has to provide the synchronisation needed. 83 84__LINK_STATE_NOCARRIER, maps to !IFF_LOWER_UP: 85 86The driver uses netif_carrier_on() to clear and netif_carrier_off() to 87set this flag. On netif_carrier_off(), the scheduler stops sending 88packets. The name 'carrier' and the inversion are historical, think of 89it as lower layer. 90 91netif_carrier_ok() can be used to query that bit. 92 93__LINK_STATE_DORMANT, maps to IFF_DORMANT: 94 95Set by the driver to express that the device cannot yet be used 96because some driver controlled protocol establishment has to 97complete. Corresponding functions are netif_dormant_on() to set the 98flag, netif_dormant_off() to clear it and netif_dormant() to query. 99 100On device allocation, networking core sets the flags equivalent to 101netif_carrier_ok() and !netif_dormant(). 102 103 104Whenever the driver CHANGES one of these flags, a workqueue event is 105scheduled to translate the flag combination to IFLA_OPERSTATE as 106follows: 107 108!netif_carrier_ok(): 109 IF_OPER_LOWERLAYERDOWN if the interface is stacked, IF_OPER_DOWN 110 otherwise. Kernel can recognise stacked interfaces because their 111 ifindex != iflink. 112 113netif_carrier_ok() && netif_dormant(): 114 IF_OPER_DORMANT 115 116netif_carrier_ok() && !netif_dormant(): 117 IF_OPER_UP if userspace interaction is disabled. Otherwise 118 IF_OPER_DORMANT with the possibility for userspace to initiate the 119 IF_OPER_UP transition afterwards. 120 121 1224. Setting from userspace 123 124Applications have to use the netlink interface to influence the 125RFC2863 operational state of an interface. Setting IFLA_LINKMODE to 1 126via RTM_SETLINK instructs the kernel that an interface should go to 127IF_OPER_DORMANT instead of IF_OPER_UP when the combination 128netif_carrier_ok() && !netif_dormant() is set by the 129driver. Afterwards, the userspace application can set IFLA_OPERSTATE 130to IF_OPER_DORMANT or IF_OPER_UP as long as the driver does not set 131netif_carrier_off() or netif_dormant_on(). Changes made by userspace 132are multicasted on the netlink group RTMGRP_LINK. 133 134So basically a 802.1X supplicant interacts with the kernel like this: 135 136-subscribe to RTMGRP_LINK 137-set IFLA_LINKMODE to 1 via RTM_SETLINK 138-query RTM_GETLINK once to get initial state 139-if initial flags are not (IFF_LOWER_UP && !IFF_DORMANT), wait until 140 netlink multicast signals this state 141-do 802.1X, eventually abort if flags go down again 142-send RTM_SETLINK to set operstate to IF_OPER_UP if authentication 143 succeeds, IF_OPER_DORMANT otherwise 144-see how operstate and IFF_RUNNING is echoed via netlink multicast 145-set interface back to IF_OPER_DORMANT if 802.1X reauthentication 146 fails 147-restart if kernel changes IFF_LOWER_UP or IFF_DORMANT flag 148 149if supplicant goes down, bring back IFLA_LINKMODE to 0 and 150IFLA_OPERSTATE to a sane value. 151 152A routing daemon or dhcp client just needs to care for IFF_RUNNING or 153waiting for operstate to go IF_OPER_UP/IF_OPER_UNKNOWN before 154considering the interface / querying a DHCP address. 155 156 157For technical questions and/or comments please e-mail to Stefan Rompf 158(stefan at loplof.de). 159