You are here
netconsole（通过网络传输dmesg）的实现 有大用
By: dreamice
Netconsole实例源代码分析
By dremice
dremice.jiang@gmail.com
1．前言····················································································································- 2 -
2．实例配置············································································································- 2 -
2.1 监控主机的配置···········································································································- 2 -
2.2 被监控主机的配置······································································································- 2 -
2.3 实例测试·······················································································································- 3 -
3．Netconsole内核源码分析··················································································- 4 -
3.1 几个重要的数据结构··································································································- 4 -
3.1.1 struct console···················································································································- 4 -
3.1.2 struct netpoll····················································································································- 5 -
3.1.3 struct net_device··············································································································- 5 -
3.1.4 struct netpoll_info············································································································- 6 -
3.2 实现分析·······················································································································- 6 -
3.2.1 Netconsole初始化············································································································- 7 -
3.2.2 具体运行实现···············································································································- 11 -
- 1 -
By: dreamice
1．前言
Netconsole是Linux2.6版内核的一个新的特性。它允许将本机的dmesg系统信息，通过网络的方式传送到另一台主机上。这样，就可以实现远程监控某台机子的kernel panic信息了。使用起来非常方便，也给开发人员调试内核提供了更加便捷的途径。
2．实例配置
由于2.6版本的内核本身已经支持Netconsole，并以模块形式编译进了内核。所以，下面的介绍都是基于2.6版内核，且本身在内核编译的时候，已经选择了将Netconsole以模块进行编译的方式。如果内核本身并没有编译Netconsole这个模块，那就需要重新编译内核了（笔者暂时没有找到其他的解决方案）。
2.1 监控主机的配置
监控主机有两种配置方式：使用本身的syslogd和利用netcat这个工具。Syslogd使用了514这个特定的UDP端口，而netcat工具可以指定任意未被使用的UDP端口进行监视。我这里使用的是netcat的方式进行监视的。安装好netcat后，执行shell命令：
# netcat -l -p 30000 –u
这里使用了端口30000来进行接收，这个端口是任意制定的，只要不发生冲突即可。
2.2 被监控主机的配置
在需要被监控的主机上加载运行netconsole模块，由于netconsole模块需要许多其他模块的依赖，以及在加载时必须配置相关的端口，IP地址等信息，所以shell执行命令如下：
#modprobe netconsole netconsole=6666@10.14.0.225/eth7,30000@10.14.0.220/00:0C:29:E2:87:E8
我们来分析一下这段加载命令。
Modprobe 这个模块加载命令是指在加载该模块时，同时加载其依赖的其它模块；
第一个@前的6666是本地端口，这个我们可以在源码中看到，是一个源代码中默认的本地端口；
10.14.0.225/eth7：这里指配本机的IP地址及网卡名称（通常我们可以用ifconfig –a来查看到）；
30000@10.14.0.220/00:0C:29:E2:87:E8：三个位段分别是，远端监控端口，远端IP地址及远端机子的MAC地址。也就是我们在2.1节中指定的监控端口30000，以及其IP地址和MAC地址了。 - 2 -
By: dreamice
2.3 实例测试
以上配置就算完成了，让我来看一个实例吧。在被监控的主机上运行一个简单的hello world模块，看看我们的printk信息是怎样被发送到监控主机上去的。
模块helloworld源码：
hello.c
1 /*hello.c*/
2 #include
3 #include
4 #include
5 MODULE_LICENSE("Dual BSD/GPL");
6
7 static int hello_init(void)
8 {
9 printk(KERN_ALERT "Hello World\n" );
10 return 0;
11 }
12 static void hello_exit(void)
13 {
14 printk(KERN_ALERT "Goodbye World\n" );
15 }
16
17 module_init(hello_init);
18 module_exit(hello_exit);
Makefile:
1 obj-m:=hello.o
2 KDIR:=/lib/modules/$(shell uname -r)/build
3 PWD:=$(shell pwd)
4
5 default:
6 $(MAKE) -C $(KDIR) M=$(PWD) modules
7
8 clean:
9 $(RM) *.o *.mod.c *.ko *.symvers
执行：
# make
# insmod hello.ko
在监控机的终端上，我们看到以下信息：
netconsole: network logging started
Hello World
第一条是我们先前加载netconsole模块时收到的，第二条是我们加载helloworld模块时
- 3 -
By: dreamice
收到的。
再运行rmmod hello
监控机终端上将看到：
netconsole: network logging started
Hello World
Goodbye World
OK，以上测试顺利完成，当然，我们也可以测试kernel panic的情况，例如操作一个内核的非法地址等。下面，我们开始来分析内核源码的netconsole的实现。
3．Netconsole内核源码分析
3.1 几个重要的数据结构
3.1.1 struct console
Include/linux/console.h
/*
* The interface for a console, or any other device that wants to capture
* console messages (printer driver?)
*
* If a console driver is marked CON_BOOT then it will be auto-unregistered
* when the first real console is registered. This is for early-printk drivers.
*/
#define CON_PRINTBUFFER (1)
#define CON_CONSDEV (2) /* Last on the command line */
#define CON_ENABLED (4)
#define CON_BOOT (8)
#define CON_ANYTIME (16) /* Safe to call when cpu is offline */
struct console
{
char name[8];
void (*write)(struct console *, const char *, unsigned);
int (*read)(struct console *, char *, unsigned);
struct tty_driver *(*device)(struct console *, int *);
void (*unblank)(void);
int (*setup)(struct console *, char *);
short flags;
short index; - 4 -
By: dreamice
int cflag;
void *data;
struct console *next;
};
从上面的注释，我们看到，这个结构，是提供给那些需要捕捉console终端信息的设备的。我们捕获的dmesg的信息，实际上是用prink输出的，因此，我们必须构造这么一个结构，并实现其信息捕捉函数，即write函数。
3.1.2 struct netpoll
Include/linux/netpoll.h
struct netpoll {
struct net_device *dev; //指向实际的网卡
char dev_name[16], *name; //名字
void (*rx_hook)(struct netpoll *, int, char *, int); //钩子函数，这里并没有用到
void (*drop)(struct sk_buff *skb); //信息发送函数
u32 local_ip, remote_ip; //本地ip地址和远端ip地址
u16 local_port, remote_port;//本地端口和远端端口
unsigned char local_mac[6], remote_mac[6];//本地mac地址和远端mac地址
};
这是netconsole实现的一个关键数据结构。net_device成员指向了实际的网卡，我们正是通过网络进行监视获取dmesg信息，因此就必须通过实际的网卡来发送数据。实际上，这个数据结构包含了一个网络包发送的所有关键字段（除了有效数据）。
3.1.3 struct net_device
Include/linux/netdevice.h
struct net_device
{
/*
* This is the first field of the "visible" part of this structure
* (i.e. as seen by users in the "Space.c" file). It is the name
* the interface.
*/
char name[IFNAMSIZ];
/* device name hash chain */
struct hlist_node name_hlist;
……
#ifdef CONFIG_NETPOLL
struct netpoll_info *npinfo;
#endif
#ifdef CONFIG_NET_POLL_CONTROLLER
- 5 -
By: dreamice
void (*poll_controller)(struct net_device *dev);
#endif
/* bridge stuff */
struct net_bridge_port *br_port;
#ifdef CONFIG_NET_DIVERT
/* this will get initialized at each interface type init routine */
struct divert_blk *divert;
#endif /* CONFIG_NET_DIVERT */
/* class/net/name entry */
struct class_device class_dev;
/* space for optional statistics and wireless sysfs groups */
struct attribute_group *sysfs_groups[3];
};
net_device结构包含了一个网络设备相关的所有信息，这里并没有全部列出其成员，其中用粉红色标记出来的，正是和netpoll相关的结构。从这里我们可以看到，如果要支持netconsole，那么就必须编译netpoll，而编译netpoll，就必须配置CONFIG_NETPOLL和CONFIG_NET_POLL_CONTROLLER这两个选项。
结构中，npinfo包含了与netpoll相关的一些重要信息。
3.1.4 struct netpoll_info
Include/linux/netpoll.h
struct netpoll_info {
spinlock_t poll_lock; //spinlock防止并发访问
int poll_owner; //所有者
int tries;//如果发送失败，指定了发送信息的次数
int rx_flags;
spinlock_t rx_lock;
struct netpoll *rx_np; /* netpoll that registered an rx_hook */
struct sk_buff_head arp_tx; /* list of arp requests to reply to */
};
在后面的实现函数分析中，将看到这个结构的详细用处。
3.2 实现分析
Drivers/net/netconsole.h
static char config[256]; //定义模块加载时指定参数选项的buf
module_param_string(netconsole, config, 256, 0); //定义模块参数
- 6 -
By: dreamice
//模块的描述，说明了模块加载时，指定参数及选项的格式
MODULE_PARM_DESC(netconsole, " netconsole=[src-port]@[src-ip]/[dev],[tgt-port]@
3.2.1 Netconsole初始化
前面提到，如果要对console的信息进行捕获，必须要实现一个struct console。因此，我们首先来看一看如何注册一个console结构。（看源码的时候，很多地方我使用的是英文注释）
/* netconsole module initialization */
static int init_netconsole(void)
{
/* The config is the module parameter, which contains the capture options */
//这个config就是上面的模块参数，在加载模块的时候它已经保存了必要的配置信息
if(strlen(config))
option_setup(config); //解析模块参数
if(!configured) { //用户加载时，参数不对，配置失败
printk("netconsole: not configured, aborting\n");
return 0;
}
/* Setup the netpoll capture operations */
if(netpoll_setup(&np)) //初始化建立netpoll的配置，np是一个struct netpoll结构全局变量，//指定了netpoll的相关信息
return -EINVAL;
/* register netconsole that can get the printk info */
register_console(&netconsole); //注册netconsole
printk(KERN_INFO "netconsole: network logging started\n");
return 0;
}
我们首先看一看结构netconsole的定义，去解析对终端信息捕捉的具体操作：
/* Initialize the console capture... */
static struct console netconsole = {
.name = "netcon",
.flags = CON_ENABLED | CON_PRINTBUFFER,
.write = write_msg //这就是对终端信息的捕捉函数
};
来看一下option_setup:
static int configured = 0;
static int option_setup(char *opt)
{
- 7 -
By: dreamice
/* parse the console capture options */
configured = !netpoll_parse_options(&np, opt);
return 1;
}
netpoll_parse_options()实现对模块参数的解析，并提取信息，保存到np结构中。这里不再详细解析该函数的实现，有兴趣可以阅读源代码。
下面分析netpoll_setup()函数的实现，在分析之前，先看一下np结构这个全局变量的定义：
static struct netpoll np = {
.name = "netconsole", //名字定义
.dev_name = "eth0", //默认的网络接口卡名称
.local_port = 6665, //默认的本地端口定义
.remote_port = 6666, //默认的远程监控端口定义
.remote_mac = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff}, //默认的远程监控机MAC地址，实际上是一个广播地址
.drop = netpoll_queue, //定义了netpoll在处理console信息的发送函数
};
int netpoll_setup(struct netpoll *np)
{
struct net_device *ndev = NULL;
struct in_device *in_dev;
struct netpoll_info *npinfo;
unsigned long flags;
//检查是否指定了网络接口卡，如果指定了，并查找系统中是否存在这样一块网卡
if (np->dev_name)
/* find the network interface card */
ndev = dev_get_by_name(np->dev_name);
if (!ndev) {
printk(KERN_ERR "%s: %s doesn't exist, aborting.\n",
np->name, np->dev_name);
return -1;
}
np->dev = ndev; //为np结构指定网络接口卡
if (!ndev->npinfo) { //这里就是准备初始化前面说到的net_device结构这个重要的npinfo成员了
npinfo = kmalloc(sizeof(*npinfo), GFP_KERNEL);
if (!npinfo)
goto release;
npinfo->rx_flags = 0;
npinfo->rx_np = NULL; //不指定hook函数 - 8 -
By: dreamice
spin_lock_init(&npinfo->poll_lock); //初始化自旋锁
npinfo->poll_owner = -1; //暂时无引用
npinfo->tries = MAX_RETRIES; //设定如果发送失败的话，重复的最大次数
spin_lock_init(&npinfo->rx_lock);
skb_queue_head_init(&npinfo->arp_tx); //和arp相关的
} else
npinfo = ndev->npinfo;
if (!ndev->poll_controller) { //检查netdev是否设定了poll_controller，否则将不能支持netconsole。它保证了在不使能中断的情况下，就可以发送skb，而且它并不是在执行终端例程的情况下被执行的。
printk(KERN_ERR "%s: %s doesn't support polling, aborting.\n",
np->name, np->dev_name);
goto release;
}
if (!netif_running(ndev)) { //检查网卡是否启动
unsigned long atmost, atleast;
printk(KERN_INFO "%s: device %s not up yet, forcing it\n",
np->name, np->dev_name);
rtnl_lock();
if (dev_change_flags(ndev, ndev->flags | IFF_UP) < 0) {
printk(KERN_ERR "%s: failed to open %s\n",
np->name, np->dev_name);
rtnl_unlock();
goto release;
}
rtnl_unlock();
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
- 9 -
By: dreamice
* trust it and pause so that we don't pump all our
* queued console messages into the bitbucket.
*/
if (time_before(jiffies, atleast)) {
printk(KERN_NOTICE "%s: carrier detect appears"
" untrustworthy, waiting 4 seconds\n",
np->name);
msleep(4000);
}
}
/* initialize the local mac, that why we need not set the local mac for the options */
//初始化本地mac地址，在加载模块的时候，我们只是指定了监控机的mac地址，所以这里要进行本地mac地址的初始化。
if (is_zero_ether_addr(np->local_mac) && ndev->dev_addr)
memcpy(np->local_mac, ndev->dev_addr, 6);
if (!np->local_ip) {//检查是否指定了本地ip，如果没有的话，主动去获取
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);
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));
}
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);
}
/* fill up the skb queue */
refill_skbs(); - 10 -
By: dreamice
/* last thing to do is link it to the net device structure */
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 -1;
}
以上，初始化完成。
3.2.2 具体运行实现
在上一节最开始就提到了netconsole这个结构的定义，它里面的一个重要的write函数的初始化，正是这个函数实现了对终端信息的捕捉。下面，我们就来分析一下write_msg()的具体实现：
#define MAX_PRINT_CHUNK 1000
static void write_msg(struct console *con, const char *msg, unsigned int len)
{
int frag, left;
unsigned long flags;
if (!np.dev)
return;
local_irq_save(flags);
//处理msg的发送，由于发送的包大小不能操作长度MAX_PRINT_CHUNK，所以可能分多次发送这些信息
for(left = len; left; ) {
/* the transport msg should NOT larger than MAX_PRINT_CHUNK*/
frag = min(left, MAX_PRINT_CHUNK);
netpoll_send_udp(&np, msg, frag); /* send the msg */
msg += frag;
left -= frag;
}
- 11 -
By: dreamice
local_irq_restore(flags);
}
我们看到，真正发送信息，实际上调用的是netpoll_send_udp()函数：
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;
udp_len = len + sizeof(*udph);
ip_len = eth_len = udp_len + sizeof(*iph);
total_len = eth_len + ETH_HLEN + NET_IP_ALIGN;
skb = find_skb(np, total_len, total_len - len);
if (!skb)
return;
memcpy(skb->data, msg, len);
skb->len += len;
udph = (struct udphdr *) skb_push(skb, sizeof(*udph));
udph->source = htons(np->local_port);
udph->dest = htons(np->remote_port);
udph->len = htons(udp_len);
udph->check = 0;
iph = (struct iphdr *)skb_push(skb, sizeof(*iph));
/* 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);
- 12 -
By: dreamice
eth = (struct ethhdr *) skb_push(skb, ETH_HLEN);
eth->h_proto = htons(ETH_P_IP);
memcpy(eth->h_source, np->local_mac, 6);
memcpy(eth->h_dest, np->remote_mac, 6);
skb->dev = np->dev;
netpoll_send_skb(np, skb);
}
这个函数看起来并不复杂，主要是对一些控制信息的设置，最后调用了netpoll_send_skp()来发送数据。接下来看看netpoll_send_skp()是怎样实现的：
static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
{
int status;
struct netpoll_info *npinfo;
//相关检查
if (!np || !np->dev || !netif_running(np->dev)) {
__kfree_skb(skb);
return;
}
npinfo = np->dev->npinfo;
/* avoid recursion */
//检查是不是本地cpu在处理发送信息
if (npinfo->poll_owner == smp_processor_id() ||
np->dev->xmit_lock_owner == smp_processor_id()) {
/* for our own cpu */
//本地cpu的情况，检查netpoll结构是否指定了drop函数，回到3.2.1节的初始化，我们为该结构指定的drop函数是netpoll_queue（在后面分析该函数的实现）
if (np->drop) /* invoke the netpoll_queue of np to send the msg, the work_queue */
np->drop(skb);
else
__kfree_skb(skb);
return;
}
//非本地cpu以及drop函数并没有实现的情况
do {
/* the times of try to send the msg if it fails */
npinfo->tries--;
- 13 -
By: dreamice
netif_tx_lock(np->dev);
/*
* network drivers do not expect to be called if the queue is
* stopped.
*/
status = NETDEV_TX_BUSY;
if (!netif_queue_stopped(np->dev))
/* send the package directly */
//直接调用了底层的skb数据包发送函数
status = np->dev->hard_start_xmit(skb, np->dev);
netif_tx_unlock(np->dev);
/* success */
if(!status) {
npinfo->tries = MAX_RETRIES; /* reset */
return;
}
/* transmit busy */
//发送繁忙的情况，进行poll操作，并尝试重新发包
netpoll_poll(np);
udelay(50);
} while (npinfo->tries > 0);
}
最后，剩下最重要的netpoll_queue()函数的实现分析了。它的实现具有一定的神奇之处，慢慢来剖析吧。
void netpoll_queue(struct sk_buff *skb)
{
unsigned long flags;
if (queue_depth == MAX_QUEUE_DEPTH) {
__kfree_skb(skb);
return;
}
spin_lock_irqsave(&queue_lock, flags);
if (!queue_head)
queue_head = skb;
else
queue_tail->next = skb;
queue_tail = skb;
- 14 -
By: dreamice
queue_depth++;
spin_unlock_irqrestore(&queue_lock, flags);
schedule_work(&send_queue);
}
从代码已开始，就看到了工作队列，先不说这个函数的实现，补习一下工作队列的相关知识。工作队列是和tasklet一类的东西，但本质上存在极大的差异。表现在：
1、 工作队列可以睡眠；
2、 工作队列可以运行在多cpu上（默认是同一处理器上）；
3、 工作队类不必以原子化执行，它还可以延迟执行；
相比于tasklet，工作队列在实时性上就显得不足了。Tasklet可以在很短的时间内很快执行，并且以原子模式执行。关于工作队列更详细的描述，请参考LDD3（p204—206）。
回到这个程序，我们在看一下在这个函数之前的一些声明和定义：
static void queue_process(void *p)
{
unsigned long flags;
struct sk_buff *skb;
while (queue_head) {
spin_lock_irqsave(&queue_lock, flags);
skb = queue_head;
queue_head = skb->next;
if (skb == queue_tail)
queue_head = NULL;
queue_depth--;
spin_unlock_irqrestore(&queue_lock, flags);
dev_queue_xmit(skb);
}
}
static DECLARE_WORK(send_queue, queue_process, NULL);
我们看到，在这里定义了work queue（send_queue）,并指定了调用函数queue_process，没有传递参数。
可以看到，queue_process()函数，最终调用函数dev_queue_xmit实现了数据包的发送。
重新回到函数netpoll_queue()，它实际上只是做了一些wrok queue深度的检查，并最终调用函数schedule_work(&send_queue);实现对工作队列的调度，也就是最终能够调度到queue_process()来完成数据包的发送任务。
4．总结
Netconsole提供了一种通过网络监控调试信息的便捷的方法，配置也十分简单。其源代码实现主要也精炼，由于使用了工作队列的机制，它可以安全的工作在中断上下文中。 - 15 -
By: dreamice
但是，netconsole的使用仍有一些限制，正如kernel中netconsole.txt文档说的：“only IP networks, UDP packets and ethernet devices are supported”，它只能工作在IP网络中，并且使用不可靠的UDP连接，目前只支持以太网络设备。
期待在新的内核版本中，netconsole会有新的发展。 - 16 -