Based on kernel version 3.9. Page generated on 2013-05-02 23:11 EST.
1 2 Network Devices, the Kernel, and You! 3 4 5 Introduction 6 ============ 7 The following is a random collection of documentation regarding 8 network devices. 9 10 struct net_device allocation rules 11 ================================== 12 Network device structures need to persist even after module is unloaded and 13 must be allocated with kmalloc. If device has registered successfully, 14 it will be freed on last use by free_netdev. This is required to handle the 15 pathologic case cleanly (example: rmmod mydriver </sys/class/net/myeth/mtu ) 16 17 There are routines in net_init.c to handle the common cases of 18 alloc_etherdev, alloc_netdev. These reserve extra space for driver 19 private data which gets freed when the network device is freed. If 20 separately allocated data is attached to the network device 21 (netdev_priv(dev)) then it is up to the module exit handler to free that. 22 23 MTU 24 === 25 Each network device has a Maximum Transfer Unit. The MTU does not 26 include any link layer protocol overhead. Upper layer protocols must 27 not pass a socket buffer (skb) to a device to transmit with more data 28 than the mtu. The MTU does not include link layer header overhead, so 29 for example on Ethernet if the standard MTU is 1500 bytes used, the 30 actual skb will contain up to 1514 bytes because of the Ethernet 31 header. Devices should allow for the 4 byte VLAN header as well. 32 33 Segmentation Offload (GSO, TSO) is an exception to this rule. The 34 upper layer protocol may pass a large socket buffer to the device 35 transmit routine, and the device will break that up into separate 36 packets based on the current MTU. 37 38 MTU is symmetrical and applies both to receive and transmit. A device 39 must be able to receive at least the maximum size packet allowed by 40 the MTU. A network device may use the MTU as mechanism to size receive 41 buffers, but the device should allow packets with VLAN header. With 42 standard Ethernet mtu of 1500 bytes, the device should allow up to 43 1518 byte packets (1500 + 14 header + 4 tag). The device may either: 44 drop, truncate, or pass up oversize packets, but dropping oversize 45 packets is preferred. 46 47 48 struct net_device synchronization rules 49 ======================================= 50 ndo_open: 51 Synchronization: rtnl_lock() semaphore. 52 Context: process 53 54 ndo_stop: 55 Synchronization: rtnl_lock() semaphore. 56 Context: process 57 Note: netif_running() is guaranteed false 58 59 ndo_do_ioctl: 60 Synchronization: rtnl_lock() semaphore. 61 Context: process 62 63 ndo_get_stats: 64 Synchronization: dev_base_lock rwlock. 65 Context: nominally process, but don't sleep inside an rwlock 66 67 ndo_start_xmit: 68 Synchronization: __netif_tx_lock spinlock. 69 70 When the driver sets NETIF_F_LLTX in dev->features this will be 71 called without holding netif_tx_lock. In this case the driver 72 has to lock by itself when needed. It is recommended to use a try lock 73 for this and return NETDEV_TX_LOCKED when the spin lock fails. 74 The locking there should also properly protect against 75 set_rx_mode. Note that the use of NETIF_F_LLTX is deprecated. 76 Don't use it for new drivers. 77 78 Context: Process with BHs disabled or BH (timer), 79 will be called with interrupts disabled by netconsole. 80 81 Return codes: 82 o NETDEV_TX_OK everything ok. 83 o NETDEV_TX_BUSY Cannot transmit packet, try later 84 Usually a bug, means queue start/stop flow control is broken in 85 the driver. Note: the driver must NOT put the skb in its DMA ring. 86 o NETDEV_TX_LOCKED Locking failed, please retry quickly. 87 Only valid when NETIF_F_LLTX is set. 88 89 ndo_tx_timeout: 90 Synchronization: netif_tx_lock spinlock; all TX queues frozen. 91 Context: BHs disabled 92 Notes: netif_queue_stopped() is guaranteed true 93 94 ndo_set_rx_mode: 95 Synchronization: netif_addr_lock spinlock. 96 Context: BHs disabled 97 98 struct napi_struct synchronization rules 99 ======================================== 100 napi->poll: 101 Synchronization: NAPI_STATE_SCHED bit in napi->state. Device 102 driver's ndo_stop method will invoke napi_disable() on 103 all NAPI instances which will do a sleeping poll on the 104 NAPI_STATE_SCHED napi->state bit, waiting for all pending 105 NAPI activity to cease. 106 Context: softirq 107 will be called with interrupts disabled by netconsole.