从ip_queue到nfnetlink_queue(下)

从ip_queue到nfnetlink_queue(下)

本文档的Copyleft归yfydz所有，使用GPL发布，可以自由拷贝，转载，转载时请保持文档的完整性，严

禁用于任何商业用途。
msn: [e-mail]
来源：http://yfydz.cublog.cn

3. 内核空间

内核版本2.6.17.11。
内核空间的代码程序包括net/netfilter/nfnetlink_queue.c和xt_NFQUEUE.c，前者是具体实现，后者

是iptables的一个目标，用来指定数据属于哪个队列。

3.1 数据结构

/* include/linux/netfilter/nfnetlink_queue.h */

// nfqueue netlink消息类型
enum nfqnl_msg_types {
NFQNL_MSG_PACKET,/* packet from kernel to userspace */
NFQNL_MSG_VERDICT,/* verdict from userspace to kernel */
NFQNL_MSG_CONFIG,/* connect to a particular queue */

NFQNL_MSG_MAX
};

// nfqueue netlink消息数据包头
struct nfqnl_msg_packet_hdr {
u_int32_tpacket_id;/* unique ID of packet in queue */
u_int16_thw_protocol;/* hw protocol (network order) */
u_int8_thook;/* netfilter hook */
} __attribute__ ((packed));

// nfqueue netlink消息数据包头硬件部分,MAC地址
struct nfqnl_msg_packet_hw {
u_int16_thw_addrlen;
u_int16_t_pad;
u_int8_thw_addr[8];
} __attribute__ ((packed));

// nfqueue netlink消息数据包64位时间戳
struct nfqnl_msg_packet_timestamp {
aligned_u64sec;
aligned_u64usec;
} __attribute__ ((packed));

// nfqueue netlink属性
enum nfqnl_attr_type {类型
NFQA_UNSPEC,
NFQA_PACKET_HDR,
NFQA_VERDICT_HDR,/* nfqnl_msg_verdict_hrd */
NFQA_MARK,/* u_int32_t nfmark */
NFQA_TIMESTAMP,/* nfqnl_msg_packet_timestamp */
NFQA_IFINDEX_INDEV,/* u_int32_t ifindex */
NFQA_IFINDEX_OUTDEV,/* u_int32_t ifindex */
NFQA_IFINDEX_PHYSINDEV,/* u_int32_t ifindex */
NFQA_IFINDEX_PHYSOUTDEV,/* u_int32_t ifindex */
NFQA_HWADDR,/* nfqnl_msg_packet_hw */
NFQA_PAYLOAD,/* opaque data payload */

__NFQA_MAX
};
#define NFQA_MAX (__NFQA_MAX - 1)

// nfqueue netlink消息数据判定头
struct nfqnl_msg_verdict_hdr {
u_int32_t verdict;
u_int32_t id;
} __attribute__ ((packed));

// nfqueue netlink消息配置命令类型
enum nfqnl_msg_config_cmds {
NFQNL_CFG_CMD_NONE,
NFQNL_CFG_CMD_BIND,
NFQNL_CFG_CMD_UNBIND,
NFQNL_CFG_CMD_PF_BIND,
NFQNL_CFG_CMD_PF_UNBIND,
};

// nfqueue netlink消息配置命令结构
struct nfqnl_msg_config_cmd {
u_int8_tcommand;/* nfqnl_msg_config_cmds */
u_int8_t_pad;
u_int16_tpf;/* AF_xxx for PF_[UN]BIND */
} __attribute__ ((packed));

// nfqueue netlink消息配置模式
enum nfqnl_config_mode {
NFQNL_COPY_NONE, // 不拷贝
NFQNL_COPY_META, // 只拷贝基本信息
NFQNL_COPY_PACKET, // 拷贝整个数据包
};

// nfqueue netlink消息配置参数结构
struct nfqnl_msg_config_params {
u_int32_tcopy_range;
u_int8_tcopy_mode;/* enum nfqnl_config_mode */
} __attribute__ ((packed));

// nfqueue netlink消息配置模式
enum nfqnl_attr_config {
NFQA_CFG_UNSPEC,
NFQA_CFG_CMD,/* nfqnl_msg_config_cmd */
NFQA_CFG_PARAMS,/* nfqnl_msg_config_params */
__NFQA_CFG_MAX
};
#define NFQA_CFG_MAX (__NFQA_CFG_MAX-1)

/* include/linux/netfilter.c */

struct nf_info
{
/* The ops struct which sent us to userspace. */
struct nf_hook_ops *elem;

/* If we're sent to userspace, this keeps housekeeping info */
int pf;
unsigned int hook;
struct net_device *indev, *outdev;
int (*okfn)(struct sk_buff *);
};

/* net/netfilter/nfnetlink_queue.c */
// 队列项结构
struct nfqnl_queue_entry {
struct list_head list;
struct nf_info *info;
struct sk_buff *skb;
unsigned int id;
};

// 队列实例结构
struct nfqnl_instance {
// HASH链表节点
struct hlist_node hlist;/* global list of queues */
atomic_t use;

// 应用程序的pid
int peer_pid;
// 队列最大长度
unsigned int queue_maxlen;
// 数据拷贝范围
unsigned int copy_range;
// 当前队列元素数
unsigned int queue_total;
// 队列丢包数
unsigned int queue_dropped;
// 用户程序判定丢包
unsigned int queue_user_dropped;
// ID序
atomic_t id_sequence;/* 'sequence' of pkt ids */
// 队列号
u_int16_t queue_num;/* number of this queue */
// 拷贝模式
u_int8_t copy_mode;

spinlock_t lock;
// queue entry队列
struct list_head queue_list;/* packets in queue */
};

3.2 内核程序流程

3.2.1 系统初始化

/* net/netfilter/nfnetlink_queue.c */

static int __init nfnetlink_queue_init(void)
{
int i, status = -ENOMEM;
#ifdef CONFIG_PROC_FS
struct proc_dir_entry *proc_nfqueue;
#endif

// 16个HASH链表
for (i = 0; i < INSTANCE_BUCKETS; i++)
INIT_HLIST_HEAD(&instance_table[i]);

// 登记netlink通知
netlink_register_notifier(&nfqnl_rtnl_notifier);
// 登记nfnetlink子系统
status = nfnetlink_subsys_register(&nfqnl_subsys);
if (status < 0) {
printk(KERN_ERR "nf_queue: failed to create netlink socket/n");
goto cleanup_netlink_notifier;
}

#ifdef CONFIG_PROC_FS
// 建立/proc/net/netfilter/nfnetlink_queue文件
proc_nfqueue = create_proc_entry("nfnetlink_queue", 0440,
proc_net_netfilter);
if (!proc_nfqueue)
goto cleanup_subsys;
proc_nfqueue->proc_fops = &nfqnl_file_ops;
#endif

// 登记nfqueue netlink设备通知
register_netdevice_notifier(&nfqnl_dev_notifier);
return status;

#ifdef CONFIG_PROC_FS
cleanup_subsys:
nfnetlink_subsys_unregister(&nfqnl_subsys);
#endif
cleanup_netlink_notifier:
netlink_unregister_notifier(&nfqnl_rtnl_notifier);
return status;
}

3.2.2

// netlink通知,只是定义一个通知回调函数, 在接收到netlink套接字信息时调用
static struct notifier_block nfqnl_rtnl_notifier = {
.notifier_call= nfqnl_rcv_nl_event,
};

static int
nfqnl_rcv_nl_event(struct notifier_block *this,
unsigned long event, void *ptr)
{
struct netlink_notify *n = ptr;

// 就只处理释放事件
if (event == NETLINK_URELEASE &&
n->protocol == NETLINK_NETFILTER && n->pid) {
int i;

/* destroy all instances for this pid */
write_lock_bh(&instances_lock);
for (i = 0; i < INSTANCE_BUCKETS; i++) {
struct hlist_node *tmp, *t2;
struct nfqnl_instance *inst;
struct hlist_head *head = &instance_table[i];
// 释放指定pid的所有子队列信息
hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
if (n->pid == inst->peer_pid)
__instance_destroy(inst);
}
}
write_unlock_bh(&instances_lock);
}
return NOTIFY_DONE;
}

以下两个函数实现释放操作,实际是调用同一个函数,一个需要加锁,一个不需要

static inline void
instance_destroy(struct nfqnl_instance *inst)
{
_instance_destroy2(inst, 1);
}

static inline void
__instance_destroy(struct nfqnl_instance *inst)
{
_instance_destroy2(inst, 0);
}

static void
_instance_destroy2(struct nfqnl_instance *inst, int lock)
{
/* first pull it out of the global list */
if (lock)
write_lock_bh(&instances_lock);

QDEBUG("removing instance %p (queuenum=%u) from hash/n",
inst, inst->queue_num);
// 将队列实例先从链表中移出
hlist_del(&inst->hlist);

if (lock)
write_unlock_bh(&instances_lock);

/* then flush all pending skbs from the queue */
// 将当前队列中所有包的判定都设置DROP
nfqnl_flush(inst, NF_DROP);

/* and finally put the refcount */
// 释放队列实例本身
instance_put(inst);
// 释放模块引用
module_put(THIS_MODULE);
}

3.2.3 子系统

// 子系统定义
static struct nfnetlink_subsystem nfqnl_subsys = {
.name= "nf_queue",
.subsys_id= NFNL_SUBSYS_QUEUE, // NFQUEUE的ID号为3
.cb_count= NFQNL_MSG_MAX, // 3个控制块
.cb= nfqnl_cb,
};

// 子系统回调控制
static struct nfnl_callback nfqnl_cb[NFQNL_MSG_MAX] = {
// 接收数据包,实际没进行定义
[NFQNL_MSG_PACKET]= { .call = nfqnl_recv_unsupp,
.attr_count = NFQA_MAX, },
// 接收判定
[NFQNL_MSG_VERDICT]= { .call = nfqnl_recv_verdict,
.attr_count = NFQA_MAX, },
// 接收配置
[NFQNL_MSG_CONFIG]= { .call = nfqnl_recv_config,
.attr_count = NFQA_CFG_MAX, },
};

3.2.3.1

// 实际没定义
static int
nfqnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
struct nlmsghdr *nlh, struct nfattr *nfqa[], int *errp)
{
return -ENOTSUPP;
}

3.2.3.2 接收判定

该函数接收netlink套接字返回的数据包的判定结果,根据结果对包进行相关处理

static int
nfqnl_recv_verdict(struct sock *ctnl, struct sk_buff *skb,
struct nlmsghdr *nlh, struct nfattr *nfqa[], int *errp)
{
struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
u_int16_t queue_num = ntohs(nfmsg->res_id);

struct nfqnl_msg_verdict_hdr *vhdr;
struct nfqnl_instance *queue;
unsigned int verdict;
struct nfqnl_queue_entry *entry;
int err;

// 判定数据包大小是否有问题
if (nfattr_bad_size(nfqa, NFQA_MAX, nfqa_verdict_min)) {
QDEBUG("bad attribute size/n");
return -EINVAL;
}

// 根据队列号找到队列的实例,并增加计数
queue = instance_lookup_get(queue_num);
if (!queue)
return -ENODEV;

// 检查该队列对应的pid是否和netlink数据包中的pid匹配
if (queue->peer_pid != NETLINK_CB(skb).pid) {
err = -EPERM;
goto err_out_put;
}

// 检查是否返回了判定结果
if (!nfqa[NFQA_VERDICT_HDR-1]) {
err = -EINVAL;
goto err_out_put;
}

// 获取判定结果
vhdr = NFA_DATA(nfqa[NFQA_VERDICT_HDR-1]);
verdict = ntohl(vhdr->verdict);

// 低16位为判定结果, 不能超过NF_MAX_VERDICT(5)
if ((verdict & NF_VERDICT_MASK) > NF_MAX_VERDICT) {
err = -EINVAL;
goto err_out_put;
}

// 根据返回包的ID号在队列中找缓存具体的数据包
entry = find_dequeue_entry(queue, id_cmp, ntohl(vhdr->id));
if (entry == NULL) {
err = -ENOENT;
goto err_out_put;
}

if (nfqa[NFQA_PAYLOAD-1]) {
// 返回了负载内容,说明要进行数据包的修改,如果不修改是不用返回载荷内容的
if (nfqnl_mangle(NFA_DATA(nfqa[NFQA_PAYLOAD-1]),
NFA_PAYLOAD(nfqa[NFQA_PAYLOAD-1]), entry) < 0)
// 修改出错,丢弃数据包
verdict = NF_DROP;
}

// 是否修改数据包的mark值
if (nfqa[NFQA_MARK-1])
entry->skb->nfmark = ntohl(*(u_int32_t *)
NFA_DATA(nfqa[NFQA_MARK-1]));

// 和ip_queue一样,调用nf_reinject()重新将数据包发回netfilter进行处理
// 然后将该entry的内存释放掉
issue_verdict(entry, verdict);

// 减少队列引用计数
instance_put(queue);
return 0;

err_out_put:
instance_put(queue);
return err;
}

3.2.3.3 接收配置

static int
nfqnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
struct nlmsghdr *nlh, struct nfattr *nfqa[], int *errp)
{
struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
u_int16_t queue_num = ntohs(nfmsg->res_id);
struct nfqnl_instance *queue;
int ret = 0;

QDEBUG("entering for msg %u/n", NFNL_MSG_TYPE(nlh->nlmsg_type));
// 数据大小检查
if (nfattr_bad_size(nfqa, NFQA_CFG_MAX, nfqa_cfg_min)) {
QDEBUG("bad attribute size/n");
return -EINVAL;
}

//
// 根据队列号找到队列的实例,并增加计数
queue = instance_lookup_get(queue_num);
if (nfqa[NFQA_CFG_CMD-1]) {
// 配置命令,由于可能是进行新建queue操作,所以此时的queue值可能为空
//
struct nfqnl_msg_config_cmd *cmd;
cmd = NFA_DATA(nfqa[NFQA_CFG_CMD-1]);
QDEBUG("found CFG_CMD/n");

switch (cmd->command) {
case NFQNL_CFG_CMD_BIND:
if (queue)
return -EBUSY;
// 绑定命令,就是新建一个queue和对应的pid绑定
queue = instance_create(queue_num, NETLINK_CB(skb).pid);
if (!queue)
return -EINVAL;
break;
case NFQNL_CFG_CMD_UNBIND:
// 取消绑定
if (!queue)
return -ENODEV;
// 检查pid是否匹配
if (queue->peer_pid != NETLINK_CB(skb).pid) {
ret = -EPERM;
goto out_put;
}
// 是否队列实例
instance_destroy(queue);
break;
case NFQNL_CFG_CMD_PF_BIND:
// 绑定协议族, 将nfqueue handler绑定到指定的协议
QDEBUG("registering queue handler for pf=%u/n",
ntohs(cmd->pf));
ret = nf_register_queue_handler(ntohs(cmd->pf), &nfqh);
break;
case NFQNL_CFG_CMD_PF_UNBIND:
// 取消协议族的绑定
QDEBUG("unregistering queue handler for pf=%u/n",
ntohs(cmd->pf));
/* This is a bug and a feature. We can unregister
* other handlers(!) */
ret = nf_unregister_queue_handler(ntohs(cmd->pf));
break;
default:
ret = -EINVAL;
break;
}
} else {
// 如果不是配置命令,检查queue是否存在,pid是否匹配
if (!queue) {
QDEBUG("no config command, and no instance ENOENT/n");
ret = -ENOENT;
goto out_put;
}

if (queue->peer_pid != NETLINK_CB(skb).pid) {
QDEBUG("no config command, and wrong pid/n");
ret = -EPERM;
goto out_put;
}
}

if (nfqa[NFQA_CFG_PARAMS-1]) {
// 配置参数
struct nfqnl_msg_config_params *params;

if (!queue) {
ret = -ENOENT;
goto out_put;
}
params = NFA_DATA(nfqa[NFQA_CFG_PARAMS-1]);
// 设置数据拷贝模式
nfqnl_set_mode(queue, params->copy_mode,
ntohl(params->copy_range));
}

out_put:
// 减少引用计数
// 除了初始化函数和释放函数外,所有其他处理函数的计数增加和减少操作都是成对出现的
instance_put(queue);
return ret;
}

其中队列实例建立函数如下:

static struct nfqnl_instance *
instance_create(u_int16_t queue_num, int pid)
{
struct nfqnl_instance *inst;

QDEBUG("entering for queue_num=%u, pid=%d/n", queue_num, pid);

write_lock_bh(&instances_lock);
//
// 根据队列号找到队列的实例,这里是不增加计数的
if (__instance_lookup(queue_num)) {
// 理论上是不可能进入这里的
inst = NULL;
QDEBUG("aborting, instance already exists/n");
goto out_unlock;
}

// 分配queue实例空间, 初始化参数
inst = kzalloc(sizeof(*inst), GFP_ATOMIC);
if (!inst)
goto out_unlock;

inst->queue_num = queue_num;
inst->peer_pid = pid;
inst->queue_maxlen = NFQNL_QMAX_DEFAULT;
inst->copy_range = 0xfffff;
inst->copy_mode = NFQNL_COPY_NONE;
atomic_set(&inst->id_sequence, 0);
/* needs to be two, since we _put() after creation */
// 初始引用计数为2,因为nfqnl_recv_config()会释放掉一次
atomic_set(&inst->use, 2);
spin_lock_init(&inst->lock);
INIT_LIST_HEAD(&inst->queue_list);

if (!try_module_get(THIS_MODULE))
goto out_free;

// 将该队列实例添加到总的队列HASH链表中
hlist_add_head(&inst->hlist,
&instance_table[instance_hashfn(queue_num)]);

write_unlock_bh(&instances_lock);

QDEBUG("successfully created new instance/n");

return inst;

out_free:
kfree(inst);
out_unlock:
write_unlock_bh(&instances_lock);
return NULL;
}

其中nf_queue_handler定义如下, 主要是定义数据进入协议队列函数,这个就是数据包进入nf_queue的

进入点:

static struct nf_queue_handler nfqh = {
.name = "nf_queue",
.outfn= &nfqnl_enqueue_packet,
};

static int
nfqnl_enqueue_packet(struct sk_buff *skb, struct nf_info *info,
unsigned int queuenum, void *data)
{
int status = -EINVAL;
struct sk_buff *nskb;
struct nfqnl_instance *queue;
struct nfqnl_queue_entry *entry;

QDEBUG("entered/n");
//
// 根据队列号找到队列的实例,并增加计数
queue = instance_lookup_get(queuenum);
if (!queue) {
QDEBUG("no queue instance matching/n");
return -EINVAL;
}

// 如果该子队列拷贝模式是NFQNL_COPY_NONE,出错返回
if (queue->copy_mode == NFQNL_COPY_NONE) {
QDEBUG("mode COPY_NONE, aborting/n");
status = -EAGAIN;
goto err_out_put;
}

// 分配一个队列项entry
entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
if (entry == NULL) {
if (net_ratelimit())
printk(KERN_ERR
"nf_queue: OOM in nfqnl_enqueue_packet()/n");
status = -ENOMEM;
goto err_out_put;
}

entry->info = info;
entry->skb = skb;
// 数据包的ID是顺序增加的
entry->id = atomic_inc_return(&queue->id_sequence);

// 构建一个netlink协议的skb包
nskb = nfqnl_build_packet_message(queue, entry, &status);
if (nskb == NULL)
goto err_out_free;

spin_lock_bh(&queue->lock);

// pid是否存在,pid为0的进程不存在
if (!queue->peer_pid)
goto err_out_free_nskb;

// 队列长度是否过长
if (queue->queue_total >= queue->queue_maxlen) {
queue->queue_dropped++;
status = -ENOSPC;
if (net_ratelimit())
printk(KERN_WARNING "ip_queue: full at %d entries, "
"dropping packets(s). Dropped: %d/n",
queue->queue_total, queue->queue_dropped);
goto err_out_free_nskb;
}

/* nfnetlink_unicast will either free the nskb or add it to a socket */
// 将新构造的netlink数据包发送给上层的netlink套接字
status = nfnetlink_unicast(nskb, queue->peer_pid, MSG_DONTWAIT);
if (status < 0) {
queue->queue_user_dropped++;
goto err_out_unlock;
}
// 将队列项entry放入队列
__enqueue_entry(queue, entry);

spin_unlock_bh(&queue->lock);
// 减少队列计数
instance_put(queue);
return status;

err_out_free_nskb:
kfree_skb(nskb);

err_out_unlock:
spin_unlock_bh(&queue->lock);

err_out_free:
kfree(entry);
err_out_put:
instance_put(queue);
return status;
}

// 构造netlink数据包
static struct sk_buff *
nfqnl_build_packet_message(struct nfqnl_instance *queue,
struct nfqnl_queue_entry *entry, int *errp)
{
unsigned char *old_tail;
size_t size;
size_t data_len = 0;
struct sk_buff *skb;
struct nfqnl_msg_packet_hdr pmsg;
struct nlmsghdr *nlh;
struct nfgenmsg *nfmsg;
// entry info, 可得到inif,outif,hook等
struct nf_info *entinf = entry->info;
// entry skb, 原始skb
struct sk_buff *entskb = entry->skb;
struct net_device *indev;
struct net_device *outdev;
unsigned int tmp_uint;

QDEBUG("entered/n");

/* all macros expand to constant values at compile time */
// 头部固定长度
size = NLMSG_SPACE(sizeof(struct nfgenmsg)) +
+ NFA_SPACE(sizeof(struct nfqnl_msg_packet_hdr))
+ NFA_SPACE(sizeof(u_int32_t))/* ifindex */
+ NFA_SPACE(sizeof(u_int32_t))/* ifindex */
#ifdef CONFIG_BRIDGE_NETFILTER
+ NFA_SPACE(sizeof(u_int32_t))/* ifindex */
+ NFA_SPACE(sizeof(u_int32_t))/* ifindex */
#endif
+ NFA_SPACE(sizeof(u_int32_t))/* mark */
+ NFA_SPACE(sizeof(struct nfqnl_msg_packet_hw))
+ NFA_SPACE(sizeof(struct nfqnl_msg_packet_timestamp));
// 数据包出网卡
outdev = entinf->outdev;

spin_lock_bh(&queue->lock);

switch (queue->copy_mode) {
case NFQNL_COPY_META:
case NFQNL_COPY_NONE:
// 这两种拷贝类型数据长度为0
data_len = 0;
break;

case NFQNL_COPY_PACKET:
// 拷贝整个包
if (entskb->ip_summed == CHECKSUM_HW &&
(*errp = skb_checksum_help(entskb,
outdev == NULL))) {
// 校验和检查失败
spin_unlock_bh(&queue->lock);
return NULL;
}
if (queue->copy_range == 0 // 为0表示不限制拷贝范围长度
|| queue->copy_range > entskb->len) // 拷贝限制大于数据包长
// 数据长度为实际数据包长度
data_len = entskb->len;
else
// 数据长度为限制的拷贝长度限制
data_len = queue->copy_range;
// 将data_len对齐后添加包头长度
size += NFA_SPACE(data_len);
break;

default:
*errp = -EINVAL;
spin_unlock_bh(&queue->lock);
return NULL;
}

spin_unlock_bh(&queue->lock);
// 分配skb
skb = alloc_skb(size, GFP_ATOMIC);
if (!skb)
goto nlmsg_failure;

old_tail= skb->tail;
// netlink信息头放在skb的tailroom中
nlh = NLMSG_PUT(skb, 0, 0,
NFNL_SUBSYS_QUEUE << 8 | NFQNL_MSG_PACKET,
sizeof(struct nfgenmsg));
nfmsg = NLMSG_DATA(nlh);
// 协议族
nfmsg->nfgen_family = entinf->pf;
// 版本
nfmsg->version = NFNETLINK_V0;
// 队列号
nfmsg->res_id = htons(queue->queue_num);
// 包ID号
pmsg.packet_id = htonl(entry->id);
// 硬件协议
pmsg.hw_protocol= htons(entskb->protocol);
// nf的hook点
pmsg.hook= entinf->hook;

NFA_PUT(skb, NFQA_PACKET_HDR, sizeof(pmsg), &pmsg);
// 数据进入网卡
indev = entinf->indev;
if (indev) {
tmp_uint = htonl(indev->ifindex);
#ifndef CONFIG_BRIDGE_NETFILTER
NFA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint), &tmp_uint);
#else
if (entinf->pf == PF_BRIDGE) {
// 如果是桥协议族,填入物理网卡和进入网卡参数
/* Case 1: indev is physical input device, we need to
* look for bridge group (when called from
* netfilter_bridge) */
NFA_PUT(skb, NFQA_IFINDEX_PHYSINDEV, sizeof(tmp_uint),
&tmp_uint);
/* this is the bridge group "brX" */
tmp_uint = htonl(indev->br_port->br->dev->ifindex);
NFA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
&tmp_uint);
} else {
/* Case 2: indev is bridge group, we need to look for
* physical device (when called from ipv4) */
// 填入输入网卡信息
NFA_PUT(skb, NFQA_IFINDEX_INDEV, sizeof(tmp_uint),
&tmp_uint);
if (entskb->nf_bridge
&& entskb->nf_bridge->physindev) {
// 如果存在桥信息和物理进入网卡信息,填入
tmp_uint = htonl(entskb->nf_bridge->physindev->ifindex

);
NFA_PUT(skb, NFQA_IFINDEX_PHYSINDEV,
sizeof(tmp_uint), &tmp_uint);
}
}
#endif
}

// 数据包发出网卡
if (outdev) {
tmp_uint = htonl(outdev->ifindex);
#ifndef CONFIG_BRIDGE_NETFILTER
// 没定义桥模块时直接填入发出网卡信息
NFA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint), &tmp_uint);
#else
if (entinf->pf == PF_BRIDGE) {
// 桥协议组, 分别填入物理发出网卡和发出网卡信息
/* Case 1: outdev is physical output device, we need to
* look for bridge group (when called from
* netfilter_bridge) */
NFA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV, sizeof(tmp_uint),
&tmp_uint);
/* this is the bridge group "brX" */
tmp_uint = htonl(outdev->br_port->br->dev->ifindex);
NFA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
&tmp_uint);
} else {
/* Case 2: outdev is bridge group, we need to look for
* physical output device (when called from ipv4) */
// 填入发出网卡信息
NFA_PUT(skb, NFQA_IFINDEX_OUTDEV, sizeof(tmp_uint),
&tmp_uint);
if (entskb->nf_bridge
&& entskb->nf_bridge->physoutdev) {
// 如果存在桥信息和物理发出网卡信息,填入
tmp_uint = htonl(entskb->nf_bridge->physoutdev-

>ifindex);
NFA_PUT(skb, NFQA_IFINDEX_PHYSOUTDEV,
sizeof(tmp_uint), &tmp_uint);
}
}
#endif
}

if (entskb->nfmark) {
// 如果数据包MARK值不为0, 填入
tmp_uint = htonl(entskb->nfmark);
NFA_PUT(skb, NFQA_MARK, sizeof(u_int32_t), &tmp_uint);
}

if (indev && entskb->dev
&& entskb->dev->hard_header_parse) {
// 填入输入网卡的硬件信息
struct nfqnl_msg_packet_hw phw;

phw.hw_addrlen =
entskb->dev->hard_header_parse(entskb,
phw.hw_addr);
phw.hw_addrlen = htons(phw.hw_addrlen);
NFA_PUT(skb, NFQA_HWADDR, sizeof(phw), &phw);
}

if (entskb->tstamp.off_sec) {
// 时间戳
struct nfqnl_msg_packet_timestamp ts;

ts.sec = cpu_to_be64(entskb->tstamp.off_sec);
ts.usec = cpu_to_be64(entskb->tstamp.off_usec);

NFA_PUT(skb, NFQA_TIMESTAMP, sizeof(ts), &ts);
}

if (data_len) {
// 填入数据包长, 以struct nfattr结构方式
struct nfattr *nfa;
int size = NFA_LENGTH(data_len);

if (skb_tailroom(skb) < (int)NFA_SPACE(data_len)) {
printk(KERN_WARNING "nf_queue: no tailroom!/n");
goto nlmsg_failure;
}

nfa = (struct nfattr *)skb_put(skb, NFA_ALIGN(size));
nfa->nfa_type = NFQA_PAYLOAD;
nfa->nfa_len = size;

if (skb_copy_bits(entskb, 0, NFA_DATA(nfa), data_len))
BUG();
}
// netlink信息长度,新tail减老的tail值
nlh->nlmsg_len = skb->tail - old_tail;
return skb;

nlmsg_failure:
nfattr_failure:
if (skb)
kfree_skb(skb);
*errp = -EINVAL;
if (net_ratelimit())
printk(KERN_ERR "nf_queue: error creating packet message/n");
return NULL;
}

3.2.4 登记nfqueue netlink设备通知

static struct notifier_block nfqnl_dev_notifier = {
.notifier_call= nfqnl_rcv_dev_event,
};

static int
nfqnl_rcv_dev_event(struct notifier_block *this,
unsigned long event, void *ptr)
{
struct net_device *dev = ptr;
// 只处理设备释放事件,如果网卡DOWN了,就会进行相关处理
/* Drop any packets associated with the downed device */
if (event == NETDEV_DOWN)
nfqnl_dev_drop(dev->ifindex);
return NOTIFY_DONE;
}

/* drop all packets with either indev or outdev == ifindex from all queue
* instances */
static void
nfqnl_dev_drop(int ifindex)
{
int i;

QDEBUG("entering for ifindex %u/n", ifindex);

/* this only looks like we have to hold the readlock for a way too long
* time, issue_verdict(), nf_reinject(), ... - but we always only
* issue NF_DROP, which is processed directly in nf_reinject() */
read_lock_bh(&instances_lock);

// 查找所有队列
for (i = 0; i < INSTANCE_BUCKETS; i++) {
struct hlist_node *tmp;
struct nfqnl_instance *inst;
struct hlist_head *head = &instance_table[i];

hlist_for_each_entry(inst, tmp, head, hlist) {
struct nfqnl_queue_entry *entry;
while ((entry = find_dequeue_entry(inst, dev_cmp,
ifindex)) != NULL)
// 一旦数据包的进入或发出网卡是DOWN掉的网卡,就丢弃该数据包
issue_verdict(entry, NF_DROP);
}
}

read_unlock_bh(&instances_lock);
}

// 比较设备,不论是in还是out的设备,只要和ifindex符合的就匹配成功
static int
dev_cmp(struct nfqnl_queue_entry *entry, unsigned long ifindex)
{
struct nf_info *entinf = entry->info;

if (entinf->indev)
if (entinf->indev->ifindex == ifindex)
return 1;

if (entinf->outdev)
if (entinf->outdev->ifindex == ifindex)
return 1;

return 0;
}

3.2.5 /proc

就是以前介绍的2.6.*中用于实现/proc只读文件的seq操作

static struct file_operations nfqnl_file_ops = {
.owner = THIS_MODULE,
.open = nfqnl_open,
.read = seq_read,
.llseek = seq_lseek,
.release = seq_release_private,
};

static int nfqnl_open(struct inode *inode, struct file *file)
{
struct seq_file *seq;
struct iter_state *is;
int ret;

is = kzalloc(sizeof(*is), GFP_KERNEL);
if (!is)
return -ENOMEM;
// 打开nfqueue netlink的顺序操作
// 文件内容就是16个HASH表中的各项的参数,最多65536项
ret = seq_open(file, &nfqnl_seq_ops);
if (ret < 0)
goto out_free;
seq = file->private_data;
seq->private = is;
return ret;
out_free:
kfree(is);
return ret;
}

static int seq_show(struct seq_file *s, void *v)
{
const struct nfqnl_instance *inst = v;

// 该/proc文件中最大可能会有65536行, 每行表示一个子queue的信息
return seq_printf(s, "%5d %6d %5d %1d %5d %5d %5d %8d %2d/n",
inst->queue_num,
inst->peer_pid, inst->queue_total,
inst->copy_mode, inst->copy_range,
inst->queue_dropped, inst->queue_user_dropped,
atomic_read(&inst->id_sequence),
atomic_read(&inst->use));
}

3.3 NFQUEUE目标

该目标很简单,返回一个无符号32位值,该值的生成就是提供一个16位的队列号,然后左移16位作为结果的

高16位,低16位置为NF_QUEUE(3).

#define NF_VERDICT_MASK 0x0000ffff
#define NF_VERDICT_BITS 16

#define NF_VERDICT_QMASK 0xffff0000
#define NF_VERDICT_QBITS 16

#define NF_QUEUE_NR(x) (((x << NF_VERDICT_QBITS) & NF_VERDICT_QMASK) | NF_QUEUE)

static unsigned int
target(struct sk_buff **pskb,
const struct net_device *in,
const struct net_device *out,
unsigned int hooknum,
const struct xt_target *target,
const void *targinfo,
void *userinfo)
{
const struct xt_NFQ_info *tinfo = targinfo;

return NF_QUEUE_NR(tinfo->queuenum);
}

在iptables命令行就可以将指定的数据包设置为进入指定的子队列,例:

iptables -A INPUT -s 1.1.1.1 -d 2.2.2.2 -j NFQUEUE --queue-num 100

将从1.1.1.1到2.2.2.2的包发送到子队列100.

3.4 NFQUEUE包处理

和正常netfilter数据包处理一样, 要进行NFQUEUE的数据包也进入nf_hook_slow()函数处理:

/* net/netfilter/core.c */
int nf_hook_slow(int pf, unsigned int hook, struct sk_buff **pskb,
struct net_device *indev,
struct net_device *outdev,
int (*okfn)(struct sk_buff *),
int hook_thresh)
{
......
// 对于NFQUEUE的包,看verdict的低16位是否为NF_QUEUE
} else if ((verdict & NF_VERDICT_MASK) == NF_QUEUE) {
NFDEBUG("nf_hook: Verdict = QUEUE./n");
// 进入nf_queue进行处理
if (!nf_queue(pskb, elem, pf, hook, indev, outdev, okfn,
verdict >> NF_VERDICT_BITS))
goto next_hook;
......

/* net/netfilter/nf_queue.c */

// nf_queue()函数和以前2.4基本是相同的,从这里是看不出ip_queue和nf_queue的区别,
// 每个协议族还是只有一个QUEUE的handler,但这时挂接的nf_queue的handler
// 的处理函数nfqnl_enqueue_packet()

/*
* Any packet that leaves via this function must come back
* through nf_reinject().
*/
int nf_queue(struct sk_buff **skb,
struct list_head *elem,
int pf, unsigned int hook,
struct net_device *indev,
struct net_device *outdev,
int (*okfn)(struct sk_buff *),
unsigned int queuenum)
{
int status;
struct nf_info *info;
#ifdef CONFIG_BRIDGE_NETFILTER
struct net_device *physindev = NULL;
struct net_device *physoutdev = NULL;
#endif
struct nf_afinfo *afinfo;

/* QUEUE == DROP if noone is waiting, to be safe. */
read_lock(&queue_handler_lock);
if (!queue_handler[pf]) {
read_unlock(&queue_handler_lock);
kfree_skb(*skb);
return 1;
}

afinfo = nf_get_afinfo(pf);
if (!afinfo) {
read_unlock(&queue_handler_lock);
kfree_skb(*skb);
return 1;
}

info = kmalloc(sizeof(*info) + afinfo->route_key_size, GFP_ATOMIC);
if (!info) {
if (net_ratelimit())
printk(KERN_ERR "OOM queueing packet %p/n",
*skb);
read_unlock(&queue_handler_lock);
kfree_skb(*skb);
return 1;
}

*info = (struct nf_info) {
(struct nf_hook_ops *)elem, pf, hook, indev, outdev, okfn };

/* If it's going away, ignore hook. */
if (!try_module_get(info->elem->owner)) {
read_unlock(&queue_handler_lock);
kfree(info);
return 0;
}

/* Bump dev refs so they don't vanish while packet is out */
if (indev) dev_hold(indev);
if (outdev) dev_hold(outdev);

#ifdef CONFIG_BRIDGE_NETFILTER
if ((*skb)->nf_bridge) {
physindev = (*skb)->nf_bridge->physindev;
if (physindev) dev_hold(physindev);
physoutdev = (*skb)->nf_bridge->physoutdev;
if (physoutdev) dev_hold(physoutdev);
}
#endif
afinfo->saveroute(*skb, info);
status = queue_handler[pf]->outfn(*skb, info, queuenum,
queue_handler[pf]->data);

read_unlock(&queue_handler_lock);

if (status < 0) {
/* James M doesn't say fuck enough. */
if (indev) dev_put(indev);
if (outdev) dev_put(outdev);
#ifdef CONFIG_BRIDGE_NETFILTER
if (physindev) dev_put(physindev);
if (physoutdev) dev_put(physoutdev);
#endif
module_put(info->elem->owner);
kfree(info);
kfree_skb(*skb);

return 1;
}

return 1;
}

4. 结论

nf_queue扩展了ip_queue的功能，使用类似802.1qVLAN的技术，将数据包打上不同的“标签”使之归到

不同的队列，而不再象ip_queue那样只支持一个队列，这样就可以使最多65536个应用程序接收内核数据

包，从而分别进行更仔细
的分类处理。