From: Jesper Dangaard Brouer <brouer@redhat.com>
To: Eric Dumazet <eric.dumazet@gmail.com>,
"David S. Miller" <davem@davemloft.net>,
Florian Westphal <fw@strlen.de>
Cc: Jesper Dangaard Brouer <brouer@redhat.com>,
netdev@vger.kernel.org, Pablo Neira Ayuso <pablo@netfilter.org>,
Thomas Graf <tgraf@suug.ch>, Cong Wang <amwang@redhat.com>,
"Patrick McHardy" <kaber@trash.net>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
Herbert Xu <herbert@gondor.hengli.com.au>
Subject: [RFC net-next PATCH V1 9/9] net: frag remove readers-writer lock (hack)
Date: Fri, 23 Nov 2012 14:08:47 +0100 [thread overview]
Message-ID: <20121123130847.18764.87682.stgit@dragon> (raw)
In-Reply-To: <20121123130749.18764.25962.stgit@dragon>
Do NOT APPLY this patch.
After all the other patches, the rw_lock is now the contention point.
This is a quick hack, that remove the readers-writer lock, by
disabling/breaking hash rebuilding. Just to see how big the
performance gain would be.
2x10G size(4416) result: 6481+6764 = 13245 Mbit/s (gen: 7652+8077 Mbit/s)
4x10G size(4416) result:(5610+6283+5735+5238)=22866 Mbit/s
(gen: 6530+7860+5967+5238 =25595 Mbit/s)
And the results show, that its a big win. With 4x10G size(4416)
before: 17923 Mbit/s -> now: 22866 Mbit/s increase 4943 Mbit/s.
With 2x10G size(4416) before 10689 Mbit/s -> 13245 Mbit/s
increase 2556 Mbit/s.
I'll work on a real solution for removing the rw_lock while still
supporting hash rebuilding. Suggestions and ideas are welcome.
NOT-signed-off
---
include/net/inet_frag.h | 2 +-
net/ipv4/inet_fragment.c | 23 +++++++++++++----------
net/ipv4/ip_fragment.c | 2 +-
net/ipv6/netfilter/nf_conntrack_reasm.c | 2 +-
net/ipv6/reassembly.c | 2 +-
5 files changed, 17 insertions(+), 14 deletions(-)
diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
index 5054228..2fb8578 100644
--- a/include/net/inet_frag.h
+++ b/include/net/inet_frag.h
@@ -58,7 +58,7 @@ struct inet_frag_bucket {
struct inet_frags {
struct inet_frag_bucket hash[INETFRAGS_HASHSZ];
- rwlock_t lock; /* Rebuild lock */
+// rwlock_t lock; /* Rebuild lock */
u32 rnd;
int qsize;
int secret_interval;
diff --git a/net/ipv4/inet_fragment.c b/net/ipv4/inet_fragment.c
index 447423f..63227d6 100644
--- a/net/ipv4/inet_fragment.c
+++ b/net/ipv4/inet_fragment.c
@@ -35,8 +35,11 @@ static void inet_frag_secret_rebuild(unsigned long dummy)
unsigned long now = jiffies;
int i;
+ //HACK don't rebuild
+ return;
+
/* Per bucket lock NOT needed here, due to write lock protection */
- write_lock(&f->lock);
+// write_lock(&f->lock);
get_random_bytes(&f->rnd, sizeof(u32));
for (i = 0; i < INETFRAGS_HASHSZ; i++) {
@@ -59,7 +62,7 @@ static void inet_frag_secret_rebuild(unsigned long dummy)
}
}
}
- write_unlock(&f->lock);
+// write_unlock(&f->lock);
mod_timer(&f->secret_timer, now + f->secret_interval);
}
@@ -74,7 +77,7 @@ void inet_frags_init(struct inet_frags *f)
spin_lock_init(&hb->chain_lock);
INIT_HLIST_HEAD(&hb->chain);
}
- rwlock_init(&f->lock);
+// rwlock_init(&f->lock);
f->rnd = (u32) ((num_physpages ^ (num_physpages>>7)) ^
(jiffies ^ (jiffies >> 6)));
@@ -115,14 +118,14 @@ static inline void fq_unlink(struct inet_frag_queue *fq, struct inet_frags *f)
struct inet_frag_bucket *hb;
unsigned int hash;
- read_lock(&f->lock);
+ //read_lock(&f->lock);
hash = f->hashfn(fq);
hb = &f->hash[hash];
spin_lock_bh(&hb->chain_lock);
hlist_del(&fq->list);
spin_unlock_bh(&hb->chain_lock);
- read_unlock(&f->lock);
+ //read_unlock(&f->lock);
inet_frag_lru_del(fq);
}
@@ -249,7 +252,7 @@ static struct inet_frag_queue *inet_frag_intern(struct netns_frags *nf,
#endif
unsigned int hash;
- read_lock(&f->lock); /* Protects against hash rebuild */
+ //read_lock(&f->lock); /* Protects against hash rebuild */
/*
* While we stayed w/o the lock other CPU could update
* the rnd seed, so we need to re-calculate the hash
@@ -268,7 +271,7 @@ static struct inet_frag_queue *inet_frag_intern(struct netns_frags *nf,
if (qp->net == nf && f->match(qp, arg)) {
atomic_inc(&qp->refcnt);
spin_unlock_bh(&hb->chain_lock);
- read_unlock(&f->lock);
+ //read_unlock(&f->lock);
qp_in->last_in |= INET_FRAG_COMPLETE;
inet_frag_put(qp_in, f);
return qp;
@@ -282,7 +285,7 @@ static struct inet_frag_queue *inet_frag_intern(struct netns_frags *nf,
atomic_inc(&qp->refcnt);
hlist_add_head(&qp->list, &hb->chain);
spin_unlock_bh(&hb->chain_lock);
- read_unlock(&f->lock);
+ //read_unlock(&f->lock);
inet_frag_lru_add(nf, qp);
return qp;
}
@@ -342,12 +345,12 @@ struct inet_frag_queue *inet_frag_find(struct netns_frags *nf,
if (q->net == nf && f->match(q, key)) {
atomic_inc(&q->refcnt);
spin_unlock_bh(&hb->chain_lock);
- read_unlock(&f->lock);
+ //read_unlock(&f->lock);
return q;
}
}
spin_unlock_bh(&hb->chain_lock);
- read_unlock(&f->lock);
+ //read_unlock(&f->lock);
return inet_frag_create(nf, f, key);
}
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index 7b1cf51..b2cb05f 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -289,7 +289,7 @@ static inline struct ipq *ip_find(struct net *net, struct iphdr *iph, u32 user)
arg.iph = iph;
arg.user = user;
- read_lock(&ip4_frags.lock);
+ //read_lock(&ip4_frags.lock);
hash = ipqhashfn(iph->id, iph->saddr, iph->daddr, iph->protocol);
q = inet_frag_find(&net->ipv4.frags, &ip4_frags, &arg, hash);
diff --git a/net/ipv6/netfilter/nf_conntrack_reasm.c b/net/ipv6/netfilter/nf_conntrack_reasm.c
index c088831..5b57e03 100644
--- a/net/ipv6/netfilter/nf_conntrack_reasm.c
+++ b/net/ipv6/netfilter/nf_conntrack_reasm.c
@@ -175,7 +175,7 @@ static inline struct frag_queue *fq_find(struct net *net, __be32 id,
arg.src = src;
arg.dst = dst;
- read_lock_bh(&nf_frags.lock);
+ //read_lock_bh(&nf_frags.lock);
hash = inet6_hash_frag(id, src, dst, nf_frags.rnd);
q = inet_frag_find(&net->nf_frag.frags, &nf_frags, &arg, hash);
diff --git a/net/ipv6/reassembly.c b/net/ipv6/reassembly.c
index 9cfe047..2c74394 100644
--- a/net/ipv6/reassembly.c
+++ b/net/ipv6/reassembly.c
@@ -193,7 +193,7 @@ fq_find(struct net *net, __be32 id, const struct in6_addr *src, const struct in6
arg.src = src;
arg.dst = dst;
- read_lock(&ip6_frags.lock);
+ //read_lock(&ip6_frags.lock);
hash = inet6_hash_frag(id, src, dst, ip6_frags.rnd);
q = inet_frag_find(&net->ipv6.frags, &ip6_frags, &arg, hash);
next prev parent reply other threads:[~2012-11-23 13:10 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-11-23 13:08 [RFC net-next PATCH V1 0/9] net: fragmentation performance scalability on NUMA/SMP systems Jesper Dangaard Brouer
2012-11-23 13:08 ` [RFC net-next PATCH V1 2/9] net: frag cache line adjust inet_frag_queue.net Jesper Dangaard Brouer
2012-11-23 13:08 ` [RFC net-next PATCH V1 4/9] net: frag helper functions for mem limit tracking Jesper Dangaard Brouer
2012-11-23 13:08 ` [RFC net-next PATCH V1 7/9] net: frag queue locking per hash bucket Jesper Dangaard Brouer
2012-11-27 9:07 ` Jesper Dangaard Brouer
2012-11-27 15:00 ` Jesper Dangaard Brouer
2012-11-23 13:08 ` [RFC net-next PATCH V1 8/9] net: increase frag queue hash size and cache-line Jesper Dangaard Brouer
2012-11-23 13:08 ` Jesper Dangaard Brouer [this message]
2012-11-26 6:03 ` [RFC net-next PATCH V1 9/9] net: frag remove readers-writer lock (hack) Stephen Hemminger
2012-11-26 9:18 ` Florian Westphal
[not found] ` <20121123130806.18764.41854.stgit@dragon>
2012-11-23 19:58 ` [RFC net-next PATCH V1 1/9] net: frag evictor, avoid killing warm frag queues Florian Westphal
2012-11-24 11:36 ` Jesper Dangaard Brouer
2012-11-25 2:31 ` [RFC net-next PATCH V1 0/9] net: fragmentation performance scalability on NUMA/SMP systems Eric Dumazet
2012-11-25 8:53 ` Jesper Dangaard Brouer
2012-11-25 16:11 ` Eric Dumazet
2012-11-26 14:42 ` Jesper Dangaard Brouer
2012-11-26 15:15 ` Eric Dumazet
2012-11-26 15:29 ` Jesper Dangaard Brouer
[not found] ` <20121123130826.18764.66507.stgit@dragon>
2012-11-26 2:54 ` [RFC net-next PATCH V1 5/9] net: frag per CPU mem limit and LRU list accounting Cong Wang
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20121123130847.18764.87682.stgit@dragon \
--to=brouer@redhat.com \
--cc=amwang@redhat.com \
--cc=davem@davemloft.net \
--cc=eric.dumazet@gmail.com \
--cc=fw@strlen.de \
--cc=herbert@gondor.hengli.com.au \
--cc=kaber@trash.net \
--cc=netdev@vger.kernel.org \
--cc=pablo@netfilter.org \
--cc=paulmck@linux.vnet.ibm.com \
--cc=tgraf@suug.ch \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).