码迷,mamicode.com
首页 > 其他好文 > 详细

intel万兆网卡驱动简要分析

时间:2015-01-10 21:08:28      阅读:1113      评论:0      收藏:0      [点我收藏+]

标签:

原创文章,转载请注明: 转载自pagefault

本文链接地址: intel万兆网卡驱动简要分析

这里分析的驱动代码是给予linux kernel 3.4.4

对应的文件在drivers/net/ethernet/intel 目录下,这个分析不涉及到很细节的地方,主要目的是理解下数据在协议栈和驱动之间是如何交互的。

首先我们知道网卡都是pci设备,因此这里每个网卡驱动其实就是一个pci驱动。并且intel这里是把好几个万兆网卡(82599/82598/x540)的驱动做在一起的。

首先我们来看对应的pci_driver的结构体,这里每个pci驱动都是一个pci_driver的结构体,而这里是多个万兆网卡共用这个结构体ixgbe_driver.

1
2
3
4
5
6
7
8
9
10
11
12
static struct pci_driver ixgbe_driver = {
    .name     = ixgbe_driver_name,
    .id_table = ixgbe_pci_tbl,
    .probe    = ixgbe_probe,
    .remove   = __devexit_p(ixgbe_remove),
#ifdef CONFIG_PM
    .suspend  = ixgbe_suspend,
    .resume   = ixgbe_resume,
#endif
    .shutdown = ixgbe_shutdown,
    .err_handler = &ixgbe_err_handler
};


然后是模块初始化方法,这里其实很简单,就是调用pci的驱动注册方法,把ixgbe挂载到pci设备链中。 这里不对pci设备的初始化做太多介绍,我以前的blog有这方面的介绍,想了解的可以去看看。这里我们只需要知道最终内核会调用probe回调来初始化ixgbe。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
char ixgbe_driver_name[] = "ixgbe";
static const char ixgbe_driver_string[] =
                  "Intel(R) 10 Gigabit PCI Express Network Driver";
 
static int __init ixgbe_init_module(void)
{
    int ret;
    pr_info("%s - version %s\n", ixgbe_driver_string, ixgbe_driver_version);
    pr_info("%s\n", ixgbe_copyright);
 
#ifdef CONFIG_IXGBE_DCA
    dca_register_notify(&dca_notifier);
#endif
 
    ret = pci_register_driver(&ixgbe_driver);
    return ret;
}

这里不去追究具体如何调用probe的细节,我们直接来看probe函数,这个函数中通过硬件的信息来确定需要初始化那个驱动(82598/82599/x540),然后核心的驱动结构就放在下面的这个数组中。

1
2
3
4
5
static const struct ixgbe_info *ixgbe_info_tbl[] = {
    [board_82598] = &ixgbe_82598_info,
    [board_82599] = &ixgbe_82599_info,
    [board_X540] = &ixgbe_X540_info,
};

ixgbe_probe函数很长,我们这里就不详细分析了,因为这部分就是对网卡进行初始化。不过我们关注下面几个代码片段。

首先是根据硬件的参数来取得对应的驱动值:

1
const struct ixgbe_info *ii = ixgbe_info_tbl[ent->driver_data];

然后就是如何将不同的网卡驱动挂载到对应的回调中,这里做的很简单,就是通过对应的netdev的结构取得adapter,然后所有的核心操作都是保存在adapter中的,最后将ii的所有回调拷贝给adapter就可以了。我们来看代码:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
    struct net_device *netdev;
    struct ixgbe_adapter *adapter = NULL;
    struct ixgbe_hw *hw;
.....................................
 
    adapter = netdev_priv(netdev);
    pci_set_drvdata(pdev, adapter);
 
    adapter->netdev = netdev;
    adapter->pdev = pdev;
    hw = &adapter->hw;
    hw->back = adapter;
.......................................
    memcpy(&hw->mac.ops, ii->mac_ops, sizeof(hw->mac.ops));
    hw->mac.type  = ii->mac;
 
    /* EEPROM */
    memcpy(&hw->eeprom.ops, ii->eeprom_ops, sizeof(hw->eeprom.ops));
.....................................

最后需要关注的就是设置网卡属性,这些属性一般来说都是通过ethtool 可以设置的属性(比如tso/checksum等),这里我们就截取一部分:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
    netdev->features = NETIF_F_SG |
               NETIF_F_IP_CSUM |
               NETIF_F_IPV6_CSUM |
               NETIF_F_HW_VLAN_TX |
               NETIF_F_HW_VLAN_RX |
               NETIF_F_HW_VLAN_FILTER |
               NETIF_F_TSO |
               NETIF_F_TSO6 |
               NETIF_F_RXHASH |
               NETIF_F_RXCSUM;
 
    netdev->hw_features = netdev->features;
 
    switch (adapter->hw.mac.type) {
    case ixgbe_mac_82599EB:
    case ixgbe_mac_X540:
        netdev->features |= NETIF_F_SCTP_CSUM;
        netdev->hw_features |= NETIF_F_SCTP_CSUM |
                       NETIF_F_NTUPLE;
        break;
    default:
        break;
    }
 
    netdev->hw_features |= NETIF_F_RXALL;
..................................................
 
    netdev->priv_flags |= IFF_UNICAST_FLT;
    netdev->priv_flags |= IFF_SUPP_NOFCS;
 
    if (adapter->flags & IXGBE_FLAG_SRIOV_ENABLED)
        adapter->flags &= ~(IXGBE_FLAG_RSS_ENABLED |
                    IXGBE_FLAG_DCB_ENABLED);
...................................................................
    if (pci_using_dac) {
        netdev->features |= NETIF_F_HIGHDMA;
        netdev->vlan_features |= NETIF_F_HIGHDMA;
    }
 
    if (adapter->flags2 & IXGBE_FLAG2_RSC_CAPABLE)
        netdev->hw_features |= NETIF_F_LRO;
    if (adapter->flags2 & IXGBE_FLAG2_RSC_ENABLED)
        netdev->features |= NETIF_F_LRO;

然后我们来看下中断的注册,因为万兆网卡大部分都是多对列网卡(配合msix),因此对于上层软件来说,就好像有多个网卡一样,它们之间的数据是相互独立的,这里读的话主要是napi驱动的poll方法,后面我们会分析这个.

到了这里或许要问那么网卡是如何挂载回调给上层,从而上层来发送数据呢,这里是这样子的,每个网络设备都有一个回调函数表(比如ndo_start_xmit)来供上层调用,而在ixgbe中的话,就是ixgbe_netdev_ops,下面就是这个结构,不过只是截取了我们很感兴趣的几个地方.

不过这里注意,读回调并不在里面,这是因为写是软件主动的,而读则是硬件主动的。现在ixgbe是NAPI的,因此它的poll回调是ixgbe_poll,是中断注册时候通过netif_napi_add添加进去的。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
static const struct net_device_ops ixgbe_netdev_ops = {
    .ndo_open       = ixgbe_open,
    .ndo_stop       = ixgbe_close,
    .ndo_start_xmit     = ixgbe_xmit_frame,
    .ndo_select_queue   = ixgbe_select_queue,
    .ndo_set_rx_mode    = ixgbe_set_rx_mode,
    .ndo_validate_addr  = eth_validate_addr,
    .ndo_set_mac_address    = ixgbe_set_mac,
    .ndo_change_mtu     = ixgbe_change_mtu,
    .ndo_tx_timeout     = ixgbe_tx_timeout,
.................................................
    .ndo_set_features = ixgbe_set_features,
    .ndo_fix_features = ixgbe_fix_features,
};

这里我们最关注的其实就是ndo_start_xmit回调,这个回调就是驱动提供给协议栈的发送回调接口。我们来看这个函数.

它的实现很简单,就是选取对应的队列,然后调用ixgbe_xmit_frame_ring来发送数据。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
static netdev_tx_t ixgbe_xmit_frame(struct sk_buff *skb,
                    struct net_device *netdev)
{
    struct ixgbe_adapter *adapter = netdev_priv(netdev);
    struct ixgbe_ring *tx_ring;
 
    if (skb->len <= 0) {
        dev_kfree_skb_any(skb);
        return NETDEV_TX_OK;
    }
 
    /*
     * The minimum packet size for olinfo paylen is 17 so pad the skb
     * in order to meet this minimum size requirement.
     */
    if (skb->len < 17) {
        if (skb_padto(skb, 17))
            return NETDEV_TX_OK;
        skb->len = 17;
    }
//取得对应的队列
    tx_ring = adapter->tx_ring[skb->queue_mapping];
//发送数据
    return ixgbe_xmit_frame_ring(skb, adapter, tx_ring);
}

而在ixgbe_xmit_frame_ring中,我们就关注两个地方,一个是tso(什么是TSO,请自行google),一个是如何发送.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
    tso = ixgbe_tso(tx_ring, first, &hdr_len);
    if (tso < 0)
        goto out_drop;
    else if (!tso)
        ixgbe_tx_csum(tx_ring, first);
 
    /* add the ATR filter if ATR is on */
    if (test_bit(__IXGBE_TX_FDIR_INIT_DONE, &tx_ring->state))
        ixgbe_atr(tx_ring, first);
 
#ifdef IXGBE_FCOE
xmit_fcoe:
#endif /* IXGBE_FCOE */
    ixgbe_tx_map(tx_ring, first, hdr_len);

调用ixgbe_tso处理完tso之后,就会调用ixgbe_tx_map来发送数据。而ixgbe_tx_map所做的最主要是两步,第一步请求DMA,第二步写寄存器,通知网卡发送数据.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
    dma = dma_map_single(tx_ring->dev, skb->data, size, DMA_TO_DEVICE);
    if (dma_mapping_error(tx_ring->dev, dma))
        goto dma_error;
 
    /* record length, and DMA address */
    dma_unmap_len_set(first, len, size);
    dma_unmap_addr_set(first, dma, dma);
 
    tx_desc->read.buffer_addr = cpu_to_le64(dma);
 
    for (;;) {
        while (unlikely(size > IXGBE_MAX_DATA_PER_TXD)) {
            tx_desc->read.cmd_type_len =
                cmd_type | cpu_to_le32(IXGBE_MAX_DATA_PER_TXD);
 
            i++;
            tx_desc++;
            if (i == tx_ring->count) {
                tx_desc = IXGBE_TX_DESC(tx_ring, 0);
                i = 0;
            }
 
            dma += IXGBE_MAX_DATA_PER_TXD;
            size -= IXGBE_MAX_DATA_PER_TXD;
 
            tx_desc->read.buffer_addr = cpu_to_le64(dma);
            tx_desc->read.olinfo_status = 0;
        }
 
...................................................
        data_len -= size;
 
        dma = skb_frag_dma_map(tx_ring->dev, frag, 0, size,
                       DMA_TO_DEVICE);
..........................................................
 
        frag++;
    }
.................................
    tx_ring->next_to_use = i;
 
    /* notify HW of packet */
    writel(i, tx_ring->tail);
.................

上面的操作是异步的,也就是说此时内核还不能释放SKB,而是网卡硬件发送完数据之后,会再次产生中断通知内核,然后内核才能释放内存.接下来我们来看这部分代码。

首先来看的是中断注册的代码,这里我们假设启用了MSIX,那么网卡的中断注册回调就是ixgbe_request_msix_irqs函数,这里我们可以看到调用request_irq函数来注册回调,并且每个队列都有自己的中断号。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
static int ixgbe_request_msix_irqs(struct ixgbe_adapter *adapter)
{
    struct net_device *netdev = adapter->netdev;
    int q_vectors = adapter->num_msix_vectors - NON_Q_VECTORS;
    int vector, err;
    int ri = 0, ti = 0;
 
    for (vector = 0; vector < q_vectors; vector++) {
        struct ixgbe_q_vector *q_vector = adapter->q_vector[vector];
        struct msix_entry *entry = &adapter->msix_entries[vector];
.......................................................................
        err = request_irq(entry->vector, &ixgbe_msix_clean_rings, 0,
                  q_vector->name, q_vector);
        if (err) {
            e_err(probe, "request_irq failed for MSIX interrupt "
                  "Error: %d\n", err);
            goto free_queue_irqs;
        }
        /* If Flow Director is enabled, set interrupt affinity */
        if (adapter->flags & IXGBE_FLAG_FDIR_HASH_CAPABLE) {
            /* assign the mask for this irq */
            irq_set_affinity_hint(entry->vector,
                          &q_vector->affinity_mask);
        }
    }
 
..............................................
 
    return 0;
 
free_queue_irqs:
    ...............................
    return err;
}

而对应的中断回调是ixgbe_msix_clean_rings,而这个函数呢,做的事情很简单(需要熟悉NAPI的原理,我以前的blog有介绍),就是调用napi_schedule来重新加入软中断处理.

1
2
3
4
5
6
7
8
9
10
11
static irqreturn_t ixgbe_msix_clean_rings(int irq, void *data)
{
    struct ixgbe_q_vector *q_vector = data;
 
    /* EIAM disabled interrupts (on this vector) for us */
 
    if (q_vector->rx.ring || q_vector->tx.ring)
        napi_schedule(&q_vector->napi);
 
    return IRQ_HANDLED;
}

而NAPI驱动我们知道,最终是会调用网卡驱动挂载的poll回调,在ixgbe中,对应的回调就是ixgbe_poll,那么也就是说这个函数要做两个工作,一个是处理读,一个是处理写完之后的清理.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
int ixgbe_poll(struct napi_struct *napi, int budget)
{
    struct ixgbe_q_vector *q_vector =
                container_of(napi, struct ixgbe_q_vector, napi);
    struct ixgbe_adapter *adapter = q_vector->adapter;
    struct ixgbe_ring *ring;
    int per_ring_budget;
    bool clean_complete = true;
 
#ifdef CONFIG_IXGBE_DCA
    if (adapter->flags & IXGBE_FLAG_DCA_ENABLED)
        ixgbe_update_dca(q_vector);
#endif
//清理写
    ixgbe_for_each_ring(ring, q_vector->tx)
        clean_complete &= !!ixgbe_clean_tx_irq(q_vector, ring);
 
    /* attempt to distribute budget to each queue fairly, but don‘t allow
     * the budget to go below 1 because we‘ll exit polling */
    if (q_vector->rx.count > 1)
        per_ring_budget = max(budget/q_vector->rx.count, 1);
    else
        per_ring_budget = budget;
//读数据,并清理已完成的
    ixgbe_for_each_ring(ring, q_vector->rx)
        clean_complete &= ixgbe_clean_rx_irq(q_vector, ring,
                             per_ring_budget);
 
    /* If all work not completed, return budget and keep polling */
    if (!clean_complete)
        return budget;
 
    /* all work done, exit the polling mode */
    napi_complete(napi);
    if (adapter->rx_itr_setting & 1)
        ixgbe_set_itr(q_vector);
    if (!test_bit(__IXGBE_DOWN, &adapter->state))
        ixgbe_irq_enable_queues(adapter, ((u64)1 << q_vector->v_idx));
 
    return 0;
}

intel万兆网卡驱动简要分析

标签:

原文地址:http://blog.csdn.net/g__hk/article/details/42586463

(0)
(1)
   
举报
评论 一句话评论(0
登录后才能评论!
© 2014 mamicode.com 版权所有  联系我们:gaon5@hotmail.com
迷上了代码!