operstates.rst (6659B)
1.. SPDX-License-Identifier: GPL-2.0 2 3================== 4Operational States 5================== 6 7 81. Introduction 9=============== 10 11Linux distinguishes between administrative and operational state of an 12interface. Administrative state is the result of "ip link set dev 13<dev> up or down" and reflects whether the administrator wants to use 14the device for traffic. 15 16However, an interface is not usable just because the admin enabled it 17- ethernet requires to be plugged into the switch and, depending on 18a site's networking policy and configuration, an 802.1X authentication 19to be performed before user data can be transferred. Operational state 20shows the ability of an interface to transmit this user data. 21 22Thanks to 802.1X, userspace must be granted the possibility to 23influence operational state. To accommodate this, operational state is 24split into two parts: Two flags that can be set by the driver only, and 25a RFC2863 compatible state that is derived from these flags, a policy, 26and changeable from userspace under certain rules. 27 28 292. Querying from userspace 30========================== 31 32Both admin and operational state can be queried via the netlink 33operation RTM_GETLINK. It is also possible to subscribe to RTNLGRP_LINK 34to be notified of updates while the interface is admin up. This is 35important for setting from userspace. 36 37These values contain interface state: 38 39ifinfomsg::if_flags & IFF_UP: 40 Interface is admin up 41 42ifinfomsg::if_flags & IFF_RUNNING: 43 Interface is in RFC2863 operational state UP or UNKNOWN. This is for 44 backward compatibility, routing daemons, dhcp clients can use this 45 flag to determine whether they should use the interface. 46 47ifinfomsg::if_flags & IFF_LOWER_UP: 48 Driver has signaled netif_carrier_on() 49 50ifinfomsg::if_flags & IFF_DORMANT: 51 Driver has signaled netif_dormant_on() 52 53TLV IFLA_OPERSTATE 54------------------ 55 56contains RFC2863 state of the interface in numeric representation: 57 58IF_OPER_UNKNOWN (0): 59 Interface is in unknown state, neither driver nor userspace has set 60 operational state. Interface must be considered for user data as 61 setting operational state has not been implemented in every driver. 62 63IF_OPER_NOTPRESENT (1): 64 Unused in current kernel (notpresent interfaces normally disappear), 65 just a numerical placeholder. 66 67IF_OPER_DOWN (2): 68 Interface is unable to transfer data on L1, f.e. ethernet is not 69 plugged or interface is ADMIN down. 70 71IF_OPER_LOWERLAYERDOWN (3): 72 Interfaces stacked on an interface that is IF_OPER_DOWN show this 73 state (f.e. VLAN). 74 75IF_OPER_TESTING (4): 76 Interface is in testing mode, for example executing driver self-tests 77 or media (cable) test. It can't be used for normal traffic until tests 78 complete. 79 80IF_OPER_DORMANT (5): 81 Interface is L1 up, but waiting for an external event, f.e. for a 82 protocol to establish. (802.1X) 83 84IF_OPER_UP (6): 85 Interface is operational up and can be used. 86 87This TLV can also be queried via sysfs. 88 89TLV IFLA_LINKMODE 90----------------- 91 92contains link policy. This is needed for userspace interaction 93described below. 94 95This TLV can also be queried via sysfs. 96 97 983. Kernel driver API 99==================== 100 101Kernel drivers have access to two flags that map to IFF_LOWER_UP and 102IFF_DORMANT. These flags can be set from everywhere, even from 103interrupts. It is guaranteed that only the driver has write access, 104however, if different layers of the driver manipulate the same flag, 105the driver has to provide the synchronisation needed. 106 107__LINK_STATE_NOCARRIER, maps to !IFF_LOWER_UP: 108 109The driver uses netif_carrier_on() to clear and netif_carrier_off() to 110set this flag. On netif_carrier_off(), the scheduler stops sending 111packets. The name 'carrier' and the inversion are historical, think of 112it as lower layer. 113 114Note that for certain kind of soft-devices, which are not managing any 115real hardware, it is possible to set this bit from userspace. One 116should use TLV IFLA_CARRIER to do so. 117 118netif_carrier_ok() can be used to query that bit. 119 120__LINK_STATE_DORMANT, maps to IFF_DORMANT: 121 122Set by the driver to express that the device cannot yet be used 123because some driver controlled protocol establishment has to 124complete. Corresponding functions are netif_dormant_on() to set the 125flag, netif_dormant_off() to clear it and netif_dormant() to query. 126 127On device allocation, both flags __LINK_STATE_NOCARRIER and 128__LINK_STATE_DORMANT are cleared, so the effective state is equivalent 129to netif_carrier_ok() and !netif_dormant(). 130 131 132Whenever the driver CHANGES one of these flags, a workqueue event is 133scheduled to translate the flag combination to IFLA_OPERSTATE as 134follows: 135 136!netif_carrier_ok(): 137 IF_OPER_LOWERLAYERDOWN if the interface is stacked, IF_OPER_DOWN 138 otherwise. Kernel can recognise stacked interfaces because their 139 ifindex != iflink. 140 141netif_carrier_ok() && netif_dormant(): 142 IF_OPER_DORMANT 143 144netif_carrier_ok() && !netif_dormant(): 145 IF_OPER_UP if userspace interaction is disabled. Otherwise 146 IF_OPER_DORMANT with the possibility for userspace to initiate the 147 IF_OPER_UP transition afterwards. 148 149 1504. Setting from userspace 151========================= 152 153Applications have to use the netlink interface to influence the 154RFC2863 operational state of an interface. Setting IFLA_LINKMODE to 1 155via RTM_SETLINK instructs the kernel that an interface should go to 156IF_OPER_DORMANT instead of IF_OPER_UP when the combination 157netif_carrier_ok() && !netif_dormant() is set by the 158driver. Afterwards, the userspace application can set IFLA_OPERSTATE 159to IF_OPER_DORMANT or IF_OPER_UP as long as the driver does not set 160netif_carrier_off() or netif_dormant_on(). Changes made by userspace 161are multicasted on the netlink group RTNLGRP_LINK. 162 163So basically a 802.1X supplicant interacts with the kernel like this: 164 165- subscribe to RTNLGRP_LINK 166- set IFLA_LINKMODE to 1 via RTM_SETLINK 167- query RTM_GETLINK once to get initial state 168- if initial flags are not (IFF_LOWER_UP && !IFF_DORMANT), wait until 169 netlink multicast signals this state 170- do 802.1X, eventually abort if flags go down again 171- send RTM_SETLINK to set operstate to IF_OPER_UP if authentication 172 succeeds, IF_OPER_DORMANT otherwise 173- see how operstate and IFF_RUNNING is echoed via netlink multicast 174- set interface back to IF_OPER_DORMANT if 802.1X reauthentication 175 fails 176- restart if kernel changes IFF_LOWER_UP or IFF_DORMANT flag 177 178if supplicant goes down, bring back IFLA_LINKMODE to 0 and 179IFLA_OPERSTATE to a sane value. 180 181A routing daemon or dhcp client just needs to care for IFF_RUNNING or 182waiting for operstate to go IF_OPER_UP/IF_OPER_UNKNOWN before 183considering the interface / querying a DHCP address. 184 185 186For technical questions and/or comments please e-mail to Stefan Rompf 187(stefan at loplof.de).