设备接口层NetPoll机制
虞展
NetPoll是设备接口层基于NAPI模式提供的一种纯粹的轮询接收数据包机制，它只依赖网络设备驱动，不依赖中断机制和协议栈，通过NetPoll，可以实现UDP报文的收发。设计它的目的是为了在中断异常或者协议栈异常的情况下，提供一种可以对外发送报文的机制，这样可以让系统告诉外界自身的状况。
drivers/net/netconsole.c就是基于NetPoll实现的一种将本机内核日志发送给远程主机的驱动，关于它的介绍参考内核文档Documentation/networking/netconsole.txt。
数据结构
NetPoll对象: netpoll
struct netpoll {
// NetPool对象必须和一个网络设备对象绑定，调用者指定网络设备名字，
// 由NetPool来关联dev指针
struct net_device *dev;
char dev_name[IFNAMSIZ];
const char *name; // 使用NetPoll的模块名称
// 如果需要通过NetPool接收，则指定接收回调
void (*rx_hook)(struct netpoll *, int, char *, int);
// 本地和远端IP、端口、远端mac地址
u32 local_ip, remote_ip;
u16 local_port, remote_port;
u8 remote_mac[ETH_ALEN];
};
NetPoll信息: netpoll_info
struct netpoll_info {
atomic_t refcnt; // 引用计数
int rx_flags; // 使能NETPOLL_RX_ENABLED表示要接收数据
spinlock_t rx_lock;
struct netpoll *rx_np; /* netpoll that registered an rx_hook 只可以有一个接收对象 */
struct sk_buff_head arp_tx; /* list of arp requests to reply to */
// 待发送skb
struct sk_buff_head txq;
// 延时发送任务，处理函数为queue_process()
struct delayed_work tx_work;
};
NetPoll提供了接口netpoll_parse_options()可以让调用者指定一个字符串配置来初始化netpoll对象中的ip端口等地址信息。
模块初始化: netpoll_init()
系统初始化时只初始化一个skb队列。
static struct sk_buff_head skb_pool;
static int __init netpoll_init(void)
{
skb_queue_head_init(&skb_pool);
return 0;
}
core_initcall(netpoll_init);
注册NetPoll对象: netpoll_setup()
int netpoll_setup(struct netpoll *np)
{
struct net_device *ndev = NULL;
struct in_device *in_dev;
struct netpoll_info *npinfo;
unsigned long flags;
int err;
// 根据网络设备名称查找网络设备对象，必须指定
if (np->dev_name)
ndev = dev_get_by_name(&init_net, np->dev_name);
if (!ndev) {
printk(KERN_ERR "%s: %s doesn't exist, aborting.\n",
np->name, np->dev_name);
return -ENODEV;
}
np->dev = ndev;
// 如果该网络设备对象还没有关联netpool_info对象，则为其新建一个；已有则增加引用计数
if (!ndev->npinfo) {
npinfo = kmalloc(sizeof(*npinfo), GFP_KERNEL);
if (!npinfo) {
err = -ENOMEM;
goto release;
}
// 初始化netpoll_info对象
npinfo->rx_flags = 0;
npinfo->rx_np = NULL;
spin_lock_init(&npinfo->rx_lock);
skb_queue_head_init(&npinfo->arp_tx);
skb_queue_head_init(&npinfo->txq);
INIT_DELAYED_WORK(&npinfo->tx_work, queue_process);
// 引用计数为1
atomic_set(&npinfo->refcnt, 1);
} else {
npinfo = ndev->npinfo;
atomic_inc(&npinfo->refcnt);
}
// 网络设备对象要支持NetPool就必须要提供ndo_poll_controller()回调
if (!ndev->netdev_ops->ndo_poll_controller) {
printk(KERN_ERR "%s: %s doesn't support polling, aborting.\n",
np->name, np->dev_name);
err = -ENOTSUPP;
goto release;
}
// 如果网络设备没有打开，则尝试打开它
if (!netif_running(ndev)) {
unsigned long atmost, atleast;
printk(KERN_INFO "%s: device %s not up yet, forcing it\n",
np->name, np->dev_name);
rtnl_lock();
err = dev_open(ndev); // 打开网络设备
rtnl_unlock();
if (err) {
printk(KERN_ERR "%s: failed to open %s\n", np->name, ndev->name);
goto release;
}
// 等待网络设备ok
atleast = jiffies + HZ/10;
atmost = jiffies + 4*HZ;
while (!netif_carrier_ok(ndev)) {
if (time_after(jiffies, atmost)) {
printk(KERN_NOTICE "%s: timeout waiting for carrier\n", np->name);
break;
}
cond_resched();
}
/* If carrier appears to come up instantly, we don't
* trust it and pause so that we don't pump all our
* queued console messages into the bitbucket.
*/
// carrier就绪后，可能并不是立刻就可以发送数据，稍微等一会儿
if (time_before(jiffies, atleast)) {
printk(KERN_NOTICE "%s: carrier detect appears"
" untrustworthy, waiting 4 seconds\n",
np->name);
msleep(4000);
}
}
// 如果没有指定本机IP，则尝试从网络设备中读取一个
if (!np->local_ip) {
rcu_read_lock();
in_dev = __in_dev_get_rcu(ndev);
if (!in_dev || !in_dev->ifa_list) {
rcu_read_unlock();
printk(KERN_ERR "%s: no IP address for %s, aborting\n",
np->name, np->dev_name);
err = -EDESTADDRREQ;
goto release;
}
np->local_ip = ntohl(in_dev->ifa_list->ifa_local);
rcu_read_unlock();
printk(KERN_INFO "%s: local IP %d.%d.%d.%d\n",
np->name, HIPQUAD(np->local_ip));
}
// 如果要接收数据，则使能NETPOLL_RX_ENABLED标记，并且将netpool对象记录到netpoll_info对象中
if (np->rx_hook) {
spin_lock_irqsave(&npinfo->rx_lock, flags);
npinfo->rx_flags |= NETPOLL_RX_ENABLED;
npinfo->rx_np = np;
spin_unlock_irqrestore(&npinfo->rx_lock, flags);
}
// 提前分配一定数量的skb，存入全局skb_poll队列中
refill_skbs();
// 将netpoll_info对象存入网络设备对象中
ndev->npinfo = npinfo;
/* avoid racing with NAPI reading npinfo */
synchronize_rcu();
return 0;
release:
if (!ndev->npinfo)
kfree(npinfo);
np->dev = NULL;
dev_put(ndev);
return err;
}
发送数据: netpoll_send_udp()
NetPoll只能发送和接收UDP报文。
void netpoll_send_udp(struct netpoll *np, const char *msg, int len)
{
int total_len, eth_len, ip_len, udp_len;
struct sk_buff *skb;
struct udphdr *udph;
struct iphdr *iph;
struct ethhdr *eth;
// 计算udp、ip和mac各层报文总长度
udp_len = len + sizeof(*udph);
ip_len = eth_len = udp_len + sizeof(*iph);
total_len = eth_len + ETH_HLEN + NET_IP_ALIGN;
// 找一个skb
skb = find_skb(np, total_len, total_len - len);
if (!skb)
return;
// 将待发送数据拷贝到skb中
skb_copy_to_linear_data(skb, msg, len);
skb->len += len;
// 构造UDP报文首部
skb_push(skb, sizeof(*udph));
skb_reset_transport_header(skb);
udph = udp_hdr(skb);
udph->source = htons(np->local_port);
udph->dest = htons(np->remote_port);
udph->len = htons(udp_len);
udph->check = 0;
udph->check = csum_tcpudp_magic(htonl(np->local_ip),
htonl(np->remote_ip),
udp_len, IPPROTO_UDP,
csum_partial(udph, udp_len, 0));
if (udph->check == 0)
udph->check = CSUM_MANGLED_0;
// 构造IP报文首部
skb_push(skb, sizeof(*iph));
skb_reset_network_header(skb);
iph = ip_hdr(skb);
/* iph->version = 4; iph->ihl = 5; */
put_unaligned(0x45, (unsigned char *)iph);
iph->tos = 0;
put_unaligned(htons(ip_len), &(iph->tot_len));
iph->id = 0;
iph->frag_off = 0;
iph->ttl = 64;
iph->protocol = IPPROTO_UDP;
iph->check = 0;
put_unaligned(htonl(np->local_ip), &(iph->saddr));
put_unaligned(htonl(np->remote_ip), &(iph->daddr));
iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
// 构造以太网帧首部
eth = (struct ethhdr *) skb_push(skb, ETH_HLEN);
skb_reset_mac_header(skb);
skb->protocol = eth->h_proto = htons(ETH_P_IP);
memcpy(eth->h_source, np->dev->dev_addr, ETH_ALEN);
memcpy(eth->h_dest, np->remote_mac, ETH_ALEN);
// 指定出口网络设备
skb->dev = np->dev;
// 发送数据包
netpoll_send_skb(np, skb);
}
netpoll_send_skb()
static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
{
int status = NETDEV_TX_BUSY;
unsigned long tries;
struct net_device *dev = np->dev;
const struct net_device_ops *ops = dev->netdev_ops;
struct netpoll_info *npinfo = np->dev->npinfo;
// 网络设备当前无法工作，丢弃发送报文
if (!npinfo || !netif_running(dev) || !netif_device_present(dev)) {
__kfree_skb(skb);
return;
}
// 首次发送
if (skb_queue_len(&npinfo->txq) == 0 && !netpoll_owner_active(dev)) {
struct netdev_queue *txq;
unsigned long flags;
txq = netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
local_irq_save(flags);
// 如果发送失败，这里尝试发送多次
for (tries = jiffies_to_usecs(1)/USEC_PER_POLL;tries > 0; --tries) {
if (__netif_tx_trylock(txq)) {
// 发送给网络设备驱动
if (!netif_tx_queue_stopped(txq))
status = ops->ndo_start_xmit(skb, dev);
__netif_tx_unlock(txq);
// 发送成功
if (status == NETDEV_TX_OK)
break;
}
// 每个寻魂都尝试去轮询一次设备，检查是否有数据可接收
netpoll_poll(np);
udelay(USEC_PER_POLL);
}
local_irq_restore(flags);
}
// 本轮发送没有成功，将skb放入发送队列
if (status != NETDEV_TX_OK) {
skb_queue_tail(&npinfo->txq, skb);
schedule_delayed_work(&npinfo->tx_work,0);
}
}
延时发送任务: queue_process()
static void queue_process(struct work_struct *work)
{
struct netpoll_info *npinfo = container_of(work, struct netpoll_info, tx_work.work);
struct sk_buff *skb;
unsigned long flags;
// 遍历发送待发送队列
while ((skb = skb_dequeue(&npinfo->txq))) {
struct net_device *dev = skb->dev;
const struct net_device_ops *ops = dev->netdev_ops;
struct netdev_queue *txq;
// 网络设备对象不再工作状态
if (!netif_device_present(dev) || !netif_running(dev)) {
__kfree_skb(skb);
continue;
}
txq = netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
local_irq_save(flags);
__netif_tx_lock(txq, smp_processor_id());
// 发送数据包
if (netif_tx_queue_stopped(txq) ||
netif_tx_queue_frozen(txq) ||
ops->ndo_start_xmit(skb, dev) != NETDEV_TX_OK) {
// 发送失败把skb再放回发送队列首部
skb_queue_head(&npinfo->txq, skb);
__netif_tx_unlock(txq);
local_irq_restore(flags);
// 再次调度
schedule_delayed_work(&npinfo->tx_work, HZ/10);
return;
}
__netif_tx_unlock(txq);
local_irq_restore(flags);
}
}
接收数据
在NetPoll发送数据的过程中，会在某些时机通过netpoll_poll()尝试轮询设备，看网络设备当前是否有数据要接收。
netpoll_poll()
void netpoll_poll(struct netpoll *np)
{
struct net_device *dev = np->dev;
const struct net_device_ops *ops = dev->netdev_ops;
if (!dev || !netif_running(dev) || !ops->ndo_poll_controller)
return;
// 调用驱动程序ndo_poll_controller()回调
ops->ndo_poll_controller(dev);
// 调用驱动程序的poll()回调轮询设备，驱动程序可以在poll()中完成NAPI接收
poll_napi(dev);
// 处理ARP相应
service_arp_queue(dev->npinfo);
// 清理
zap_completion_queue();
}
poll_napi()
/*
* Check whether delayed processing was scheduled for our NIC. If so,
* we attempt to grab the poll lock and use ->poll() to pump the card.
* If this fails, either we've recursed in ->poll() or it's already
* running on another CPU.
*
* Note: we don't mask interrupts with this lock because we're using
* trylock here and interrupts are already disabled in the softirq
* case. Further, we test the poll_owner to avoid recursion on UP
* systems where the lock doesn't exist.
*
* In cases where there is bi-directional communications, reading only
* one message at a time can lead to packets being dropped by the
* network adapter, forcing superfluous retries and possibly timeouts.
* Thus, we set our budget to greater than 1.
*/
static int poll_one_napi(struct netpoll_info *npinfo,
struct napi_struct *napi, int budget)
{
int work;
/* net_rx_action's ->poll() invocations and our's are
* synchronized by this test which is only made while
* holding the napi->poll_lock.
*/
// 接收软中断函数正在调度该napi对象
if (!test_bit(NAPI_STATE_SCHED, &napi->state))
return budget;
// 调度一次poll()回调
npinfo->rx_flags |= NETPOLL_RX_DROP;
atomic_inc(&trapped);
set_bit(NAPI_STATE_NPSVC, &napi->state);
work = napi->poll(napi, budget);
clear_bit(NAPI_STATE_NPSVC, &napi->state);
atomic_dec(&trapped);
npinfo->rx_flags &= ~NETPOLL_RX_DROP;
return budget - work;
}
static void poll_napi(struct net_device *dev)
{
struct napi_struct *napi;
int budget = 16;
list_for_each_entry(napi, &dev->napi_list, dev_list) {
if (napi->poll_owner != smp_processor_id() &&
spin_trylock(&napi->poll_lock)) {
budget = poll_one_napi(dev->npinfo, napi, budget);
spin_unlock(&napi->poll_lock);
// 配额已用尽
if (!budget)
break;
}
}
}
netpoll_rx()
在数据包的接收路径上，会优先通过调用netpoll_receive_skb()或者netpoll_rx()让NetPoll有机会处理数据包，如果NetPoll返回数据包已被处理的结果，那么整个数据包接收路径也会终止。
static inline int netpoll_rx(struct sk_buff *skb)
{
struct netpoll_info *npinfo = skb->dev->npinfo;
unsigned long flags;
int ret = 0;
// 如果接收模式没有开启，返回0让数据包继续正常的接收路径
if (!npinfo || (!npinfo->rx_np && !npinfo->rx_flags))
return 0;
spin_lock_irqsave(&npinfo->rx_lock, flags);
/* check rx_flags again with the lock held */
if (npinfo->rx_flags && __netpoll_rx(skb))
ret = 1;
spin_unlock_irqrestore(&npinfo->rx_lock, flags);
return ret;
}
__netpoll_rx()
// 接收ARP报文开关，以及打开它后可以将所有的输入报文丢弃
static atomic_t trapped;
int __netpoll_rx(struct sk_buff *skb)
{
int proto, len, ulen;
struct iphdr *iph;
struct udphdr *uh;
struct netpoll_info *npi = skb->dev->npinfo;
struct netpoll *np = npi->rx_np;
if (!np)
goto out;
// 必须是以太类型帧
if (skb->dev->type != ARPHRD_ETHER)
goto out;
// 如果需要响应ARP报文，则将skb放入apr_rx队列中，返回1表示skb已经被处理
if (skb->protocol == htons(ETH_P_ARP) && atomic_read(&trapped)) {
skb_queue_tail(&npi->arp_tx, skb);
return 1;
}
// 必须是IP报文，并且必须是发给本机的
proto = ntohs(eth_hdr(skb)->h_proto);
if (proto != ETH_P_IP)
goto out;
if (skb->pkt_type == PACKET_OTHERHOST)
goto out;
if (skb_shared(skb))
goto out;
// IP报文基本判断&&校验和检查
iph = (struct iphdr *)skb->data;
if (!pskb_may_pull(skb, sizeof(struct iphdr)))
goto out;
if (iph->ihl < 5 || iph->version != 4)
goto out;
if (!pskb_may_pull(skb, iph->ihl*4))
goto out;
if (ip_fast_csum((u8 *)iph, iph->ihl) != 0)
goto out;
len = ntohs(iph->tot_len);
if (skb->len < len || len < iph->ihl*4)
goto out;
/*
* Our transport medium may have padded the buffer out.
* Now We trim to the true length of the frame.
*/
if (pskb_trim_rcsum(skb, len))
goto out;
// 必须是UDP报文
if (iph->protocol != IPPROTO_UDP)
goto out;
len -= iph->ihl*4;
uh = (struct udphdr *)(((char *)iph) + iph->ihl*4);
ulen = ntohs(uh->len);
// 检验udp报文信息
if (ulen != len)
goto out;
if (checksum_udp(skb, uh, ulen, iph->saddr, iph->daddr))
goto out;
if (np->local_ip && np->local_ip != ntohl(iph->daddr))
goto out;
if (np->remote_ip && np->remote_ip != ntohl(iph->saddr))
goto out;
if (np->local_port && np->local_port != ntohs(uh->dest))
goto out;
// 将报文传递给接收者，已经剥掉了UDP首部
np->rx_hook(np, ntohs(uh->source), (char *)(uh+1), ulen - sizeof(struct udphdr));
// 释放skb，返回1表示报文已经被接收
kfree_skb(skb);
return 1;
out:
if (atomic_read(&trapped)) {
kfree_skb(skb);
return 1;
}
return 0;
}
类似资料：