From: Florian Westphal <fw@strlen.de>
To: Florian Westphal <fw@strlen.de>
Cc: Eric Dumazet <eric.dumazet@gmail.com>, netfilter-devel@vger.kernel.org
Subject: Re: [PATCH 7/7] netfilter: connlimit: use rbtree for per-host conntrack obj storage
Date: Sun, 9 Mar 2014 22:45:13 +0100 [thread overview]
Message-ID: <20140309214513.GD14012@breakpoint.cc> (raw)
In-Reply-To: <20140309184357.GC14012@breakpoint.cc>
Florian Westphal <fw@strlen.de> wrote:
> Eric Dumazet <eric.dumazet@gmail.com> wrote:
> > > Hmm, that would be nice. I need to think about it again,
> > > problem is that moving it at this time could result in
> > > freeing the would-be parent of the new node.
> >
> > Yeah, thats why fq_gc() is followed by a full lookup.
> >
> > In practice, the lookup done in fq_gc() brings in cpu cache all the
> > cache lines, and second lookup is very fast.
>
> I had wondered about this. Ok, that makes sense.
> I'll change it to be more like fq.
>
> Thanks for explaining this.
Not exactly pretty, but in most cases the restart won't be needed
(either because we found node-to-add-to or no stale node to remove).
I'll fold the following into patch #7, but will wait until Tuesday
before resend in order to give others a chance to comment.
diff --git a/net/netfilter/xt_connlimit.c b/net/netfilter/xt_connlimit.c
--- a/net/netfilter/xt_connlimit.c
+++ b/net/netfilter/xt_connlimit.c
@@ -202,7 +202,9 @@ count_tree(struct net *net, struct rb_root *root,
struct xt_connlimit_conn *conn;
unsigned int count = 0;
unsigned int gc_count = 0;
+ bool no_gc = false;
+ restart:
rbnode = &(root->rb_node);
while (*rbnode) {
int diff;
@@ -230,7 +232,7 @@ count_tree(struct net *net, struct rb_root *root,
return count + 1;
}
- if (gc_count >= ARRAY_SIZE(gc_nodes))
+ if (no_gc || gc_count >= ARRAY_SIZE(gc_nodes))
continue;
/* only used for GC on hhead, retval and 'addit' ignored */
@@ -239,15 +241,22 @@ count_tree(struct net *net, struct rb_root *root,
gc_nodes[gc_count++] = rbconn;
}
+ if (gc_count) {
+ no_gc = true;
+ tree_nodes_free(root, gc_nodes, gc_count);
+ gc_count = 0;
+ goto restart;
+ }
+
/* no match, need to insert new node */
rbconn = kmem_cache_alloc(connlimit_rb_cachep, GFP_ATOMIC);
if (rbconn == NULL)
- goto out;
+ return 0;
conn = kmem_cache_alloc(connlimit_conn_cachep, GFP_ATOMIC);
if (conn == NULL) {
kmem_cache_free(connlimit_rb_cachep, rbconn);
- goto out;
+ return 0;
}
conn->tuple = *tuple;
@@ -259,10 +268,7 @@ count_tree(struct net *net, struct rb_root *root,
rb_link_node(&rbconn->node, parent, rbnode);
rb_insert_color(&rbconn->node, root);
- count = 1;
- out:
- tree_nodes_free(root, gc_nodes, gc_count);
- return count;
+ return 1;
}
static int count_them(struct net *net,
next prev parent reply other threads:[~2014-03-09 21:45 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-03-07 13:37 [PATCH 0/7] netfilter: connlimit: scalability improvements Florian Westphal
2014-03-07 13:37 ` [PATCH 1/7] netfilter: connlimit: factor hlist search into new function Florian Westphal
2014-03-07 13:37 ` [PATCH 2/7] netfilter: connlimit: improve packet-to-closed-connection logic Florian Westphal
2014-03-07 13:37 ` [PATCH 3/7] netfilter: connlimit: move insertion of new element out of count function Florian Westphal
2014-03-07 13:37 ` [PATCH 4/7] netfilter: connlimit: use kmem_cache for conn objects Florian Westphal
2014-03-07 13:37 ` [PATCH 5/7] netfilter: connlimit: use keyed locks Florian Westphal
2014-03-09 17:13 ` Jan Engelhardt
2014-03-09 18:31 ` Florian Westphal
2014-03-07 13:37 ` [PATCH 6/7] netfilter: connlimit: make same_source_net signed Florian Westphal
2014-03-07 13:37 ` [PATCH 7/7] netfilter: connlimit: use rbtree for per-host conntrack obj storage Florian Westphal
2014-03-07 14:47 ` Eric Dumazet
2014-03-07 16:15 ` Florian Westphal
2014-03-09 18:42 ` Eric Dumazet
2014-03-09 18:43 ` Florian Westphal
2014-03-09 21:45 ` Florian Westphal [this message]
2014-03-12 12:58 ` [PATCH 0/7] netfilter: connlimit: scalability improvements Pablo Neira Ayuso
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=20140309214513.GD14012@breakpoint.cc \
--to=fw@strlen.de \
--cc=eric.dumazet@gmail.com \
--cc=netfilter-devel@vger.kernel.org \
/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).