(1) 当内核接收到本地的IP包, 在传递给上层协议处理之前,
先进行碎片重组.IP包片段之间的标识号(id)是相同的.
当IP包片偏量(frag_off)第14位(IP_MF)为1时, 表示该IP包有后继片段.
片偏量的低13位则为该片段在完整数据包中的偏移量, 以8字节为单位. 当IP_MF位为0时,
表示IP包是最后一块碎片.
(2) 碎片重组由重组队列完成, 每一重组队列对应于(daddr,saddr,protocol,id)构成的键值,
它们存在于ipq结构构成的散列链之中. 重组队列将IP包按照将片段偏量的顺序进行排列,
当所有的片段都到齐后, 就可以将队列中的包碎片按顺序拼合成一个完整的IP包.
(3) 如果30秒后重组队列内包未到齐, 则重组过程失败, 重组队列被释放,
同时向发送方以ICMP协议通知失败信息.
重组队列的内存消耗不得大于256k(sysctl_ipfrag_high_thresh),
否则将会调用(ip_evictor)释放每支散列尾端的重组队列.
; net/ipv4/ip_input.c:
/*
* Deliver IP Packets to the higher protocol layers.
*/
int ip_local_deliver(struct sk_buff *skb)
{
struct iphdr *iph = skb->nh.iph;
/*
* Reassemble IP fragments.
*/
if (iph->frag_off & htons(IP_MF|IP_OFFSET)) {
skb = ip_defrag(skb);
if (!skb)
return 0;
}
return NF_HOOK(PF_INET, NF_IP_LOCAL_IN, skb, skb->dev, NULL,
ip_local_deliver_finish);
}
; net/ipv4/ip_fragment.c:
/* NOTE. Logic of IP defragmentation is parallel to corresponding IPv6
* code now. If you change something here, _PLEASE_ update ipv6/reassembly.c
* as well. Or notify me, at least. --ANK
*/
/* Fragment cache limits. We will commit 256K at one time. Should we
* cross that limit we will prune down to 192K. This should cope with
* even the most extreme cases without allowing an attacker to measurably
* harm machine performance.
*/
int sysctl_ipfrag_high_thresh = 256*1024; (256k字节)
int sysctl_ipfrag_low_thresh = 192*1024; (192k字节)
/* Important NOTE! Fragment queue must be destroyed before MSL expires.
* RFC791 is wrong proposing to prolongate timer each fragment arrival by TTL.
*/
int sysctl_ipfrag_time = IP_FRAG_TIME; (30秒)
struct ipfrag_skb_cb
{
struct inet_skb_parm h;
int offset;
};
#define FRAG_CB(skb) ((struct ipfrag_skb_cb*)((skb)->cb))
/* Describe an entry in the "incomplete datagrams" queue. */
struct ipq {
struct ipq *next; /* linked list pointers */
u32 saddr;
u32 daddr;
u16 id;
u8 protocol;
u8 last_in;
#define COMPLETE 4
#define FIRST_IN 2
#define LAST_IN 1
struct sk_buff *fragments; /* linked list of received fragments */
int len; /* total length of original datagram */
int meat;
spinlock_t lock;
atomic_t refcnt;
struct timer_list timer; /* when will this queue expire? */
struct ipq **pprev;
int iif; /* Device index - for icmp replies */
};
/* Hash table. */
#define IPQ_HASHSZ 64
/* Per-bucket lock is easy to add now. */
static struct ipq *ipq_hash[IPQ_HASHSZ];
static rwlock_t ipfrag_lock = RW_LOCK_UNLOCKED;
int ip_frag_nqueues = 0;
/* Process an incoming IP datagram fragment. */
struct sk_buff *ip_defrag(struct sk_buff *skb)
{
struct iphdr *iph = skb->nh.iph;
struct ipq *qp;
struct net_device *dev;
IP_INC_STATS_BH(IpReasmReqds);
/* Start by cleaning up the memory. */
if (atomic_read(&ip_frag_mem) > sysctl_ipfrag_high_thresh)
ip_evictor();
dev = skb->dev;
/* Lookup (or create) queue header */
if ((qp = ip_find(iph)) != NULL) {
struct sk_buff *ret = NULL;
spin_lock(&qp->lock);
ip_frag_queue(qp, skb);
if (qp->last_in == (FIRST_IN|LAST_IN) && 首包与尾包已收到
qp->meat == qp->len) 队列中字节数恰好等于队列尾部
ret = ip_frag_reasm(qp, dev); 重组碎片
spin_unlock(&qp->lock);
ipq_put(qp);
return ret;
}
IP_INC_STATS_BH(IpReasmFails);
kfree_skb(skb);
return NULL;
}
/*
* Was: ((((id) >> 1) ^ (saddr) ^ (daddr) ^ (prot)) & (IPQ_HASHSZ - 1))
*
* I see, I see evil hand of bigendian mafia. On Intel all the packets hit
* one hash bucket with this hash function. 8)
*/
static __inline__ unsigned int ipqhashfn(u16 id, u32 saddr, u32 daddr, u8 prot)
{
unsigned int h = saddr ^ daddr;
h ^= (h>>16)^id;
h ^= (h>>8)^prot;
return h & (IPQ_HASHSZ - 1);
}
/* Find the correct entry in the "incomplete datagrams" queue for
* this IP datagram, and create new one, if nothing is found.
*/
static inline struct ipq *ip_find(struct iphdr *iph)
{
__u16 id = iph->id;
__u32 saddr = iph->saddr;
__u32 daddr = iph->daddr;
__u8 protocol = iph->protocol;
unsigned int hash = ipqhashfn(id, saddr, daddr, protocol);
struct ipq *qp;
read_lock(&ipfrag_lock);
for(qp = ipq_hash[hash]; qp; qp = qp->next) {
if(qp->id == id &&
qp->saddr == saddr &&
qp->daddr == daddr &&
qp->protocol == protocol) {
atomic_inc(&qp->refcnt);
read_unlock(&ipfrag_lock);
return qp;
}
}
read_unlock(&ipfrag_lock);
return ip_frag_create(hash, iph); 创建一条IP片段队列
}
/* Add an entry to the 'ipq' queue for a newly received IP datagram. */
static struct ipq *ip_frag_create(unsigned hash, struct iphdr *iph)
{
struct ipq *qp;
if ((qp = frag_alloc_queue()) == NULL)
goto out_nomem;
qp->protocol = iph->protocol;
qp->last_in = 0;
qp->id = iph->id;
qp->saddr = iph->saddr;
qp->daddr = iph->daddr;
qp->len = 0;
qp->meat = 0;
qp->fragments = NULL;
qp->iif = 0;
/* Initialize a timer for this entry. */
init_timer(&qp->timer);
qp->timer.data = (unsigned long) qp; /* pointer to queue */
qp->timer.function = ip_expire; /* expire function */
qp->lock = SPIN_LOCK_UNLOCKED;
atomic_set(&qp->refcnt, 1);
return ip_frag_intern(hash, qp);
out_nomem:
NETDEBUG(printk(KERN_ERR "ip_frag_create: no memory left !
"));
return NULL;
}
extern __inline__ struct ipq *frag_alloc_queue(void)
{
struct ipq *qp = kmalloc(sizeof(struct ipq), GFP_ATOMIC);
if(!qp)
return NULL;
atomic_add(sizeof(struct ipq), &ip_frag_mem);
return qp;
}
/* Creation primitives. */
static struct ipq *ip_frag_intern(unsigned int hash, struct ipq *qp_in)
{
struct ipq *qp;
write_lock(&ipfrag_lock);
#ifdef CONFIG_SMP
/* With SMP race we have to recheck hash table, because
* such entry could be created on other cpu, while we
* promoted read lock to write lock.
*/
for(qp = ipq_hash[hash]; qp; qp = qp->next) {
if(qp->id == qp_in->id &&
qp->saddr == qp_in->saddr &&
qp->daddr == qp_in->daddr &&
qp->protocol == qp_in->protocol) {
atomic_inc(&qp->refcnt);
write_unlock(&ipfrag_lock);
qp_in->last_in |= COMPLETE;
ipq_put(qp_in);
return qp;
}
}
#endif
qp = qp_in;
if (!mod_timer(&qp->timer, jiffies + sysctl_ipfrag_time))
atomic_inc(&qp->refcnt); 安装重组队列超时监视器
atomic_inc(&qp->refcnt);
if((qp->next = ipq_hash[hash]) != NULL)
qp->next->pprev = &qp->next;
ipq_hash[hash] = qp;
qp->pprev = &ipq_hash[hash];
ip_frag_nqueues++;
write_unlock(&ipfrag_lock);
return qp;
}
/*
* Oops, a fragment queue timed out. Kill it and send an ICMP reply.
*/
static void ip_expire(unsigned long arg)
{
struct ipq *qp = (struct ipq *) arg;
spin_lock(&qp->lock);
if (qp->last_in & COMPLETE)
goto out;
ipq_kill(qp); 将重组队列从散列中删除
IP_INC_STATS_BH(IpReasmTimeout);
IP_INC_STATS_BH(IpReasmFails);
if ((qp->last_in&FIRST_IN) && qp->fragments != NULL) {
struct sk_buff *head = qp->fragments;
/* Send an ICMP "Fragment Reassembly Timeout" message. */
if ((head->dev = dev_get_by_index(qp->iif)) != NULL) {
icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);
dev_put(head->dev);
}
}
out:
spin_unlock(&qp->lock);
ipq_put(qp); 释放重组队列成员
}
/* Kill ipq entry. It is not destroyed immediately,
* because caller (and someone more) holds reference count.
*/
static __inline__ void ipq_kill(struct ipq *ipq)
{
if (del_timer(&ipq->timer))
atomic_dec(&ipq->refcnt);
if (!(ipq->last_in & COMPLETE)) {
ipq_unlink(ipq);
atomic_dec(&ipq->refcnt);
ipq->last_in |= COMPLETE;
}
}
static __inline__ void __ipq_unlink(struct ipq *qp)
{
if(qp->next)
qp->next->pprev = qp->pprev;
*qp->pprev = qp->next;
ip_frag_nqueues--;
}
static __inline__ void ipq_put(struct ipq *ipq)
{
if (atomic_dec_and_test(&ipq->refcnt))
ip_frag_destroy(ipq);
}
/* Complete destruction of ipq. */
static void ip_frag_destroy(struct ipq *qp)
{
struct sk_buff *fp;
BUG_TRAP(qp->last_in&COMPLETE);
BUG_TRAP(del_timer(&qp->timer) == 0);
/* Release all fragment data. */
fp = qp->fragments;
while (fp) {
struct sk_buff *xp = fp->next;
frag_kfree_skb(fp);
fp = xp;
}
/* Finally, release the queue descriptor itself. */
frag_free_queue(qp);
}
/* Memory Tracking Functions. */
extern __inline__ void frag_kfree_skb(struct sk_buff *skb)
{
atomic_sub(skb->truesize, &ip_frag_mem);
kfree_skb(skb);
}
extern __inline__ void frag_free_queue(struct ipq *qp)
{
atomic_sub(sizeof(struct ipq), &ip_frag_mem);
kfree(qp);
}
/* Memory limiting on fragments. Evictor trashes the oldest
* fragment queue until we are back under the low threshold.
*/
static void ip_evictor(void) 重组队列驱逐器
{
int i, progress;
do {
if (atomic_read(&ip_frag_mem) <= sysctl_ipfrag_low_thresh)
return;
progress = 0;
/* FIXME: Make LRU queue of frag heads. -DaveM */
for (i = 0; i < IPQ_HASHSZ; i++) {
struct ipq *qp;
if (ipq_hash[ i ] == NULL)
continue;
write_lock(&ipfrag_lock);
if ((qp = ipq_hash[ i ]) != NULL) {
/* find the oldest queue for this hash bucket */
while (qp->next)
qp = qp->next; 取每个散列的最后一项
__ipq_unlink(qp);
write_unlock(&ipfrag_lock);
spin_lock(&qp->lock);
if (del_timer(&qp->timer))
atomic_dec(&qp->refcnt);
qp->last_in |= COMPLETE;
spin_unlock(&qp->lock);
ipq_put(qp);
IP_INC_STATS_BH(IpReasmFails);
progress = 1;
continue;
}
write_unlock(&ipfrag_lock);
}
} while (progress);
}
/* Add new segment to existing queue. */
static void ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
{
struct iphdr *iph = skb->nh.iph;
struct sk_buff *prev, *next;
int flags, offset;
int ihl, end;
if (qp->last_in & COMPLETE)
goto err;
offset = ntohs(iph->frag_off); 取片偏移量描述字
flags = offset & ~IP_OFFSET; 取片标志
offset &= IP_OFFSET; 求片偏移量
offset <<= 3; /* offset is in 8-byte chunks */
ihl = iph->ihl * 4;
/* Determine the position of this fragment. */
end = offset + (ntohs(iph->tot_len) - ihl); 求该片段尾部的数据偏移量
/* Is this the final fragment? */
if ((flags & IP_MF) == 0) { 最后一片段
/* If we already have some bits beyond end
* or have different end, the segment is corrrupted.
*/
if (end < qp->len || 最后一个片段的未尾小于队列内最大的未尾
((qp->last_in & LAST_IN) && end != qp->len))
goto err;
qp->last_in |= LAST_IN;
qp->len = end; 设取队列最大未尾
} else { 是中间某个片段
if (end&7) { 如果片段尾部不在8字节上对齐
end &= ~7;
if (skb->ip_summed != CHECKSUM_UNNECESSARY)
skb->ip_summed = CHECKSUM_NONE; 不计算校验和
}
if (end > qp->len) { 该片段比队列内其它成员位置要大
/* Some bits beyond end -> corruption. */
if (qp->last_in & LAST_IN)
goto err;
qp->len = end;
}
}
if (end == offset) 片段的数据区长度为零
goto err;
/* Point into the IP datagram 'data' part. */
skb_pull(skb, (skb->nh.raw+ihl) - skb->data); 删除IP包的头部
skb_trim(skb, end - offset); 去队尾部可能的衬垫
/* Find out which fragments are in front and at the back of us
* in the chain of fragments so far. We must know where to put
* this fragment, right?
*/
prev = NULL; 扫描重组队列中的包片段,
取偏移大于或等于当前包偏移的前一成员作为插入位置
for(next = qp->fragments; next != NULL; next = next->next) {
if (FRAG_CB(next)->offset >= offset)
break; /* bingo! */
prev = next;
}
当前偏移的包
/* We found where to put this one. Check for overlap with
* preceding fragment, and, if needed, align things so that
* any overlaps are eliminated.
*/
if (prev) {
int i = (FRAG_CB(prev)->offset + prev->len) - offset; 求prev尾部与当前偏移之差
if (i > 0) { 插入点成员尾部大于当前包开始, 说明当前包与前一包重叠
offset += i; 当前包起点后移
if (end <= offset)
goto err;
skb_pull(skb, i); 删除当前包前部i字节.
if (skb->ip_summed != CHECKSUM_UNNECESSARY)
skb->ip_summed = CHECKSUM_NONE;
}
}
; next是当前包的后一包
while (next && FRAG_CB(next)->offset < end) { 后一包与当前包有重叠
int i = end - FRAG_CB(next)->offset; /* overlap is 'i' bytes */
if (i < next->len) { 当前包尾部小于后一包尾部
/* Eat head of the next overlapped fragment
* and leave the loop. The next ones cannot overlap.
*/
FRAG_CB(next)->offset += i; 后一包起点后移
skb_pull(next, i); 删除后一包i字节
qp->meat -= i; meat为队列已容纳的总字节数
if (next->ip_summed != CHECKSUM_UNNECESSARY)
next->ip_summed = CHECKSUM_NONE;
break;
} else { 当前包尾部大于或等于后一包尾部, 则删除后一包
struct sk_buff *free_it = next;
/* Old fragmnet is completely overridden with
* new one drop it.
*/
next = next->next;
if (prev)
prev->next = next;
else 说明next包是队列首包
qp->fragments = next;
qp->meat -= free_it->len;
frag_kfree_skb(free_it);
}
}
FRAG_CB(skb)->offset = offset; 在skb的cb[]块上记录当前包代表的数据位移
/* Insert this fragment in the chain of fragments. */
skb->next = next;
if (prev)
prev->next = skb;
else
qp->fragments = skb; 作为首包
if (skb->dev)
qp->iif = skb->dev->ifindex; 取包的输入设号号
skb->dev = NULL;
qp->meat += skb->len;
atomic_add(skb->truesize, &ip_frag_mem); truesize为包描述结构与数据区总长
if (offset == 0) 首包
qp->last_in |= FIRST_IN;
return;
err:
kfree_skb(skb);
}
/* Build a new IP datagram from all its fragments.
*
* FIXME: We copy here because we lack an effective way of handling lists
* of bits on input. Until the new skb data handling is in I'm not going
* to touch this with a bargepole.
*/
static inline unsigned int csum_add(unsigned int csum, unsigned int addend)
{
csum += addend;
return csum + (csum < addend);
}
static struct sk_buff *ip_frag_reasm(struct ipq *qp, struct net_device *dev)
{
struct sk_buff *skb;
struct iphdr *iph;
struct sk_buff *fp, *head = qp->fragments;
int len;
int ihlen;
ipq_kill(qp); 将队列从散列中删除
BUG_TRAP(head != NULL);
BUG_TRAP(FRAG_CB(head)->offset == 0);
/* Allocate a new buffer for the datagram. */
ihlen = head->nh.iph->ihl*4; 取队列头IP包头长度
len = ihlen + qp->len; 总长度
if(len > 65535)
goto out_oversize;
skb = dev_alloc_skb(len);
if (!skb)
goto out_nomem;
/* Fill in the basic details. */
skb->mac.raw = skb->data;
skb->nh.raw = skb->data;
FRAG_CB(skb)->h = FRAG_CB(head)->h; 复制IP选项信息
skb->ip_summed = head->ip_summed;
skb->csum = 0;
/* Copy the original IP headers into the new buffer. */
memcpy(skb_put(skb, ihlen), head->nh.iph, ihlen);
/* Copy the data portions of all fragments into the new buffer. */
for (fp=head; fp; fp = fp->next) {
memcpy(skb_put(skb, fp->len), fp->data, fp->len);
if (skb->ip_summed != fp->ip_summed)
skb->ip_summed = CHECKSUM_NONE;
else if (skb->ip_summed == CHECKSUM_HW)
skb->csum = csum_add(skb->csum, fp->csum);
}
skb->dst = dst_clone(head->dst);
skb->pkt_type = head->pkt_type;
skb->protocol = head->protocol;
skb->dev = dev;
/*
* Clearly bogus, because security markings of the individual
* fragments should have been checked for consistency before
* gluing, and intermediate coalescing of fragments may have
* taken place in ip_defrag() before ip_glue() ever got called.
* If we're not going to do the consistency checking, we might
* as well take the value associated with the first fragment.
* --rct
*/
skb->security = head->security;
#ifdef CONFIG_NETFILTER
/* Connection association is same as fragment (if any). */
skb->nfct = head->nfct;
nf_conntrack_get(skb->nfct);
#ifdef CONFIG_NETFILTER_DEBUG
skb->nf_debug = head->nf_debug;
#endif
#endif
/* Done with all fragments. Fixup the new IP header. */
iph = skb->nh.iph;
iph->frag_off = 0;
iph->tot_len = htons(len);
IP_INC_STATS_BH(IpReasmOKs);
return skb;
out_nomem:
NETDEBUG(printk(KERN_ERR
"IP: queue_glue: no memory for gluing queue %p
",
qp));
goto out_fail;
out_oversize:
if (net_ratelimit())
printk(KERN_INFO
"Oversized IP packet from %d.%d.%d.%d.
",
NIPQUAD(qp->saddr));
out_fail:
IP_INC_STATS_BH(IpReasmFails);
return NULL;
}
Edited by lucian_yao on 05/17/01 10:40
AM. |
