[PATCH v2] fragment: add fast path

[PATCH v2] fragment: add fast path

[Changli Gao]

add fast path

As the fragments are sent in order in most of OSes, such as Windows, Darwin and
FreeBSD, it is likely the new fragments are at the end of the inet_frag_queue.
In the fast path, we check if the skb at the end of the inet_frag_queue is the
prev we expect.

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
----
 include/net/inet_frag.h |    1 +
 net/ipv4/ip_fragment.c  |   12 ++++++++++++
 net/ipv6/reassembly.c   |   11 +++++++++++
 3 files changed, 24 insertions(+)
diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
index 39f2dc9..16ff29a 100644
--- a/include/net/inet_frag.h
+++ b/include/net/inet_frag.h
@@ -20,6 +20,7 @@ struct inet_frag_queue {
 	atomic_t		refcnt;
 	struct timer_list	timer;      /* when will this queue expire? */
 	struct sk_buff		*fragments; /* list of received fragments */
+	struct sk_buff		*fragments_tail;
 	ktime_t			stamp;
 	int			len;        /* total length of orig datagram */
 	int			meat;
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index 858d346..dbe8999 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -314,6 +314,7 @@ static int ip_frag_reinit(struct ipq *qp)
 	qp->q.len = 0;
 	qp->q.meat = 0;
 	qp->q.fragments = NULL;
+	qp->q.fragments_tail = NULL;
 	qp->iif = 0;
 
 	return 0;
@@ -386,6 +387,11 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 	 * in the chain of fragments so far.  We must know where to put
 	 * this fragment, right?
 	 */
+	prev = qp->q.fragments_tail;
+	if (!prev || FRAG_CB(prev)->offset < offset) {
+			next = NULL;
+			goto found;
+	}
 	prev = NULL;
 	for (next = qp->q.fragments; next != NULL; next = next->next) {
 		if (FRAG_CB(next)->offset >= offset)
@@ -393,6 +399,7 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 		prev = next;
 	}
 
+found:
 	/* We found where to put this one.  Check for overlap with
 	 * preceding fragment, and, if needed, align things so that
 	 * any overlaps are eliminated.
@@ -451,6 +458,8 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 
 	/* Insert this fragment in the chain of fragments. */
 	skb->next = next;
+	if (!next)
+		qp->q.fragments_tail = skb;
 	if (prev)
 		prev->next = skb;
 	else
@@ -504,6 +513,8 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
 			goto out_nomem;
 
 		fp->next = head->next;
+		if (!fp->next)
+			qp->q.fragments_tail = fp;
 		prev->next = fp;
 
 		skb_morph(head, qp->q.fragments);
@@ -574,6 +585,7 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
 	iph->tot_len = htons(len);
 	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMOKS);
 	qp->q.fragments = NULL;
+	qp->q.fragments_tail = NULL;
 	return 0;
 
 out_nomem:
diff --git a/net/ipv6/reassembly.c b/net/ipv6/reassembly.c
index 0b97230..b832f7b 100644
--- a/net/ipv6/reassembly.c
+++ b/net/ipv6/reassembly.c
@@ -333,6 +333,11 @@ static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
 	 * in the chain of fragments so far.  We must know where to put
 	 * this fragment, right?
 	 */
+	prev = fq->q.fragments_tail;
+	if (!prev || FRAG6_CB(prev)->offset < offset) {
+			next = NULL;
+			goto found;
+	}
 	prev = NULL;
 	for(next = fq->q.fragments; next != NULL; next = next->next) {
 		if (FRAG6_CB(next)->offset >= offset)
@@ -340,6 +345,7 @@ static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
 		prev = next;
 	}
 
+found:
 	/* We found where to put this one.  Check for overlap with
 	 * preceding fragment, and, if needed, align things so that
 	 * any overlaps are eliminated.
@@ -397,6 +403,8 @@ static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
 
 	/* Insert this fragment in the chain of fragments. */
 	skb->next = next;
+	if (!next)
+		fq->q.fragments_tail = skb;
 	if (prev)
 		prev->next = skb;
 	else
@@ -463,6 +471,8 @@ static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
 			goto out_oom;
 
 		fp->next = head->next;
+		if (!fp->next)
+			fq->q.fragments_tail = fp;
 		prev->next = fp;
 
 		skb_morph(head, fq->q.fragments);
@@ -549,6 +559,7 @@ static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
 	IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
 	rcu_read_unlock();
 	fq->q.fragments = NULL;
+	fq->q.fragments_tail = NULL;
 	return 1;
 
 out_oversize:

Re: [PATCH v2] fragment: add fast path

[Eric Dumazet]

Le vendredi 25 juin 2010 à 10:54 +0800, Changli Gao a écrit :
> add fast path
> 
> As the fragments are sent in order in most of OSes, such as Windows, Darwin and
> FreeBSD, it is likely the new fragments are at the end of the inet_frag_queue.
> In the fast path, we check if the skb at the end of the inet_frag_queue is the
> prev we expect.
> 
> Signed-off-by: Changli Gao <xiaosuo@gmail.com>

This patch is fine, but they are two indentation glitches.

> ----
>  include/net/inet_frag.h |    1 +
>  net/ipv4/ip_fragment.c  |   12 ++++++++++++
>  net/ipv6/reassembly.c   |   11 +++++++++++
>  3 files changed, 24 insertions(+)
> diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
> index 39f2dc9..16ff29a 100644
> --- a/include/net/inet_frag.h
> +++ b/include/net/inet_frag.h
> @@ -20,6 +20,7 @@ struct inet_frag_queue {
>  	atomic_t		refcnt;
>  	struct timer_list	timer;      /* when will this queue expire? */
>  	struct sk_buff		*fragments; /* list of received fragments */
> +	struct sk_buff		*fragments_tail;
>  	ktime_t			stamp;
>  	int			len;        /* total length of orig datagram */
>  	int			meat;
> diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
> index 858d346..dbe8999 100644
> --- a/net/ipv4/ip_fragment.c
> +++ b/net/ipv4/ip_fragment.c
> @@ -314,6 +314,7 @@ static int ip_frag_reinit(struct ipq *qp)
>  	qp->q.len = 0;
>  	qp->q.meat = 0;
>  	qp->q.fragments = NULL;
> +	qp->q.fragments_tail = NULL;
>  	qp->iif = 0;
>  
>  	return 0;
> @@ -386,6 +387,11 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
>  	 * in the chain of fragments so far.  We must know where to put
>  	 * this fragment, right?
>  	 */
> +	prev = qp->q.fragments_tail;
> +	if (!prev || FRAG_CB(prev)->offset < offset) {

strange indentation : one tab in excess

> +			next = NULL;
> +			goto found;
> +	}
>  	prev = NULL;
>  	for (next = qp->q.fragments; next != NULL; next = next->next) {
>  		if (FRAG_CB(next)->offset >= offset)
> @@ -393,6 +399,7 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
>  		prev = next;
>  	}
>  
> +found:
>  	/* We found where to put this one.  Check for overlap with
>  	 * preceding fragment, and, if needed, align things so that
>  	 * any overlaps are eliminated.
> @@ -451,6 +458,8 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
>  
>  	/* Insert this fragment in the chain of fragments. */
>  	skb->next = next;
> +	if (!next)
> +		qp->q.fragments_tail = skb;
>  	if (prev)
>  		prev->next = skb;
>  	else
> @@ -504,6 +513,8 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
>  			goto out_nomem;
>  
>  		fp->next = head->next;
> +		if (!fp->next)
> +			qp->q.fragments_tail = fp;
>  		prev->next = fp;
>  
>  		skb_morph(head, qp->q.fragments);
> @@ -574,6 +585,7 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
>  	iph->tot_len = htons(len);
>  	IP_INC_STATS_BH(net, IPSTATS_MIB_REASMOKS);
>  	qp->q.fragments = NULL;
> +	qp->q.fragments_tail = NULL;
>  	return 0;
>  
>  out_nomem:
> diff --git a/net/ipv6/reassembly.c b/net/ipv6/reassembly.c
> index 0b97230..b832f7b 100644
> --- a/net/ipv6/reassembly.c
> +++ b/net/ipv6/reassembly.c
> @@ -333,6 +333,11 @@ static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
>  	 * in the chain of fragments so far.  We must know where to put
>  	 * this fragment, right?
>  	 */
> +	prev = fq->q.fragments_tail;
> +	if (!prev || FRAG6_CB(prev)->offset < offset) {

same here : one tab in excess

> +			next = NULL;
> +			goto found;
> +	}
>  	prev = NULL;
>  	for(next = fq->q.fragments; next != NULL; next = next->next) {
>  		if (FRAG6_CB(next)->offset >= offset)
> @@ -340,6 +345,7 @@ static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
>  		prev = next;
>  	}
>  
> +found:
>  	/* We found where to put this one.  Check for overlap with
>  	 * preceding fragment, and, if needed, align things so that
>  	 * any overlaps are eliminated.
> @@ -397,6 +403,8 @@ static int ip6_frag_queue(struct frag_queue *fq, struct sk_buff *skb,
>  
>  	/* Insert this fragment in the chain of fragments. */
>  	skb->next = next;
> +	if (!next)
> +		fq->q.fragments_tail = skb;
>  	if (prev)
>  		prev->next = skb;
>  	else
> @@ -463,6 +471,8 @@ static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
>  			goto out_oom;
>  
>  		fp->next = head->next;
> +		if (!fp->next)
> +			fq->q.fragments_tail = fp;
>  		prev->next = fp;
>  
>  		skb_morph(head, fq->q.fragments);
> @@ -549,6 +559,7 @@ static int ip6_frag_reasm(struct frag_queue *fq, struct sk_buff *prev,
>  	IP6_INC_STATS_BH(net, __in6_dev_get(dev), IPSTATS_MIB_REASMOKS);
>  	rcu_read_unlock();
>  	fq->q.fragments = NULL;
> +	fq->q.fragments_tail = NULL;
>  	return 1;
>  
>  out_oversize:

Re: [PATCH v2] fragment: add fast path

[Changli Gao]

On Tue, Jun 29, 2010 at 9:54 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
> Le vendredi 25 juin 2010 à 10:54 +0800, Changli Gao a écrit :
>> add fast path
>>
>> As the fragments are sent in order in most of OSes, such as Windows, Darwin and
>> FreeBSD, it is likely the new fragments are at the end of the inet_frag_queue.
>> In the fast path, we check if the skb at the end of the inet_frag_queue is the
>> prev we expect.
>>
>> Signed-off-by: Changli Gao <xiaosuo@gmail.com>
>
> This patch is fine, but they are two indentation glitches.
>

Oh, Thanks. I'll fix them.

-- 
Regards,
Changli Gao(xiaosuo@gmail.com)

Re: [PATCH v2] fragment: add fast path

[YOSHIFUJI Hideaki]

Changli Gao wrote:
> On Tue, Jun 29, 2010 at 9:54 PM, Eric Dumazet <eric.dumazet@gmail.com> wrote:
>> Le vendredi 25 juin 2010 à 10:54 +0800, Changli Gao a écrit :
>>> add fast path
>>>
>>> As the fragments are sent in order in most of OSes, such as Windows, Darwin and
>>> FreeBSD, it is likely the new fragments are at the end of the inet_frag_queue.
>>> In the fast path, we check if the skb at the end of the inet_frag_queue is the
>>> prev we expect.
>>>
>>> Signed-off-by: Changli Gao <xiaosuo@gmail.com>
>> This patch is fine, but they are two indentation glitches.
>>
> 
> Oh, Thanks. I'll fix them.

And, I think it is better not to just say it as "fast path"
because it does not sufficient.  Probably "fast path for
in-order fragments" or something like that.

Regards,

--yoshfuji