xref: /openbmc/linux/Documentation/networking/driver.rst (revision 9df839a711aee437390b16ee39cf0b5c1620be6a)
1.. SPDX-License-Identifier: GPL-2.0
2
3=====================
4Softnet Driver Issues
5=====================
6
7Probing guidelines
8==================
9
10Address validation
11------------------
12
13Any hardware layer address you obtain for your device should
14be verified.  For example, for ethernet check it with
15linux/etherdevice.h:is_valid_ether_addr()
16
17Close/stop guidelines
18=====================
19
20Quiescence
21----------
22
23After the ndo_stop routine has been called, the hardware must
24not receive or transmit any data.  All in flight packets must
25be aborted. If necessary, poll or wait for completion of
26any reset commands.
27
28Auto-close
29----------
30
31The ndo_stop routine will be called by unregister_netdevice
32if device is still UP.
33
34Transmit path guidelines
35========================
36
37Stop queues in advance
38----------------------
39
40The ndo_start_xmit method must not return NETDEV_TX_BUSY under
41any normal circumstances.  It is considered a hard error unless
42there is no way your device can tell ahead of time when its
43transmit function will become busy.
44
45Instead it must maintain the queue properly.  For example,
46for a driver implementing scatter-gather this means:
47
48.. code-block:: c
49
50	static netdev_tx_t drv_hard_start_xmit(struct sk_buff *skb,
51					       struct net_device *dev)
52	{
53		struct drv *dp = netdev_priv(dev);
54
55		lock_tx(dp);
56		//...
57		/* This is a hard error log it. */
58		if (TX_BUFFS_AVAIL(dp) <= (skb_shinfo(skb)->nr_frags + 1)) {
59			netif_stop_queue(dev);
60			unlock_tx(dp);
61			printk(KERN_ERR PFX "%s: BUG! Tx Ring full when queue awake!\n",
62			       dev->name);
63			return NETDEV_TX_BUSY;
64		}
65
66		//... queue packet to card ...
67		//... update tx consumer index ...
68
69		if (TX_BUFFS_AVAIL(dp) <= (MAX_SKB_FRAGS + 1))
70			netif_stop_queue(dev);
71
72		//...
73		unlock_tx(dp);
74		//...
75		return NETDEV_TX_OK;
76	}
77
78And then at the end of your TX reclamation event handling:
79
80.. code-block:: c
81
82	if (netif_queue_stopped(dp->dev) &&
83	    TX_BUFFS_AVAIL(dp) > (MAX_SKB_FRAGS + 1))
84		netif_wake_queue(dp->dev);
85
86For a non-scatter-gather supporting card, the three tests simply become:
87
88.. code-block:: c
89
90		/* This is a hard error log it. */
91		if (TX_BUFFS_AVAIL(dp) <= 0)
92
93and:
94
95.. code-block:: c
96
97		if (TX_BUFFS_AVAIL(dp) == 0)
98
99and:
100
101.. code-block:: c
102
103	if (netif_queue_stopped(dp->dev) &&
104	    TX_BUFFS_AVAIL(dp) > 0)
105		netif_wake_queue(dp->dev);
106
107Lockless queue stop / wake helper macros
108~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
109
110.. kernel-doc:: include/net/netdev_queues.h
111   :doc: Lockless queue stopping / waking helpers.
112
113No exclusive ownership
114----------------------
115
116An ndo_start_xmit method must not modify the shared parts of a
117cloned SKB.
118
119Timely completions
120------------------
121
122Do not forget that once you return NETDEV_TX_OK from your
123ndo_start_xmit method, it is your driver's responsibility to free
124up the SKB and in some finite amount of time.
125
126For example, this means that it is not allowed for your TX
127mitigation scheme to let TX packets "hang out" in the TX
128ring unreclaimed forever if no new TX packets are sent.
129This error can deadlock sockets waiting for send buffer room
130to be freed up.
131
132If you return NETDEV_TX_BUSY from the ndo_start_xmit method, you
133must not keep any reference to that SKB and you must not attempt
134to free it up.
135