From: Florian Westphal <fw@strlen.de>
To: <netfilter-devel@vger.kernel.org>
Subject: [PATCH v2 nf-next 0/9] netfilter: don't copy initns hooks to new namespaces
Date: Fri, 23 Oct 2015 12:43:17 +0200 [thread overview]
Message-ID: <1445597006-11264-1-git-send-email-fw@strlen.de> (raw)
Historically, a particular table or netfilter feature (defrag, iptables
filter table ...) was registered with the netfilter core hook mechanism
on module load.
When netns support was added to iptables only the ip/ip6tables ruleset
was made namespace aware, not the actual hook points.
This has changed -- after Eric Biedermans recent work we now have
per net namespace hooks.
When a new namespace is created, all the hooks registered 'globally'
(i.e. via nf_register_hook() instead of a particular namespace via
nf_register_net_hook api) get copied to the new netns.
This means f.e. that when ipt_filter table/module is loaded on a system,
then each namespace on that system has an (empty) iptables filter ruleset.
This work aims to change all major hook users to nf_register_net_hook
so that when a new netns is created it has no hooks at all, even when the
initial namespace uses conntrack, iptables and bridge netfilter.
To keep behaviour somewhat compatible, xtable hooks are registered once a
iptables set/getsockopt call is made within a net namespace.
This also means that e.g. conntrack behaviour is not yet optimal, we
still create all the data structures and only skip hook registration
at this time.
Caveats:
- conntrack is no longer active just by loading nf_conntrack module -- at
least one (x)tables rule that requires conntrack has to be added, e.g.
conntrack match or S/DNAT target.
Loading the nat table is *not* sufficient.
Changes since v1:
- Don't add a dependency on conntrack in the nat table.
It causes hard to resolve dependency problems on initcall ordering, f.e.
ip6tables nat will fail if link order caues nat table initcall before
conntrack_ipv6 initcall, and so on (also makes the patch set pretty much
useless with MODULES=n builds).
- change all targets and matches that need conntrack, including MASQUERADE,
REDIRECT, SNAT/DNAT to register w. conntrack.
- get rid of a couple of refcount bugs.
- reduce amount of copy&pastry in iptable_foo modules. For this to work
all the XXX_register_table() functions had to be extended with the location of
the table pointer storage, problem is that this must be setup by the time the
hook is registered since we can see packets right away.
Ads section:
conntrack+filter + nat table used in init namespace, single TCP_STREAM lo netperf:
87380 16384 16384 30.00 14348.66
with patch set, netperf running in net namespace without rules:
87380 16384 16384 30.00 15683.97
routing from ns3 -> ns2, filter + nat table & conntrack in all namespaces:
87380 16384 16384 30.00 5664.46
without conntrack+any tables in those namespaces:
87380 16384 16384 30.00 7336.54
Comments welcome.
include/linux/netfilter.h | 29 ++++------
include/linux/netfilter/x_tables.h | 6 +-
include/linux/netfilter_arp/arp_tables.h | 9 +--
include/linux/netfilter_ingress.h | 9 ++-
include/linux/netfilter_ipv4/ip_tables.h | 9 +--
include/linux/netfilter_ipv6/ip6_tables.h | 9 +--
include/net/netfilter/ipv4/nf_defrag_ipv4.h | 3 -
include/net/netfilter/ipv6/nf_defrag_ipv6.h | 3 -
include/net/netfilter/nf_conntrack.h | 4 +
include/net/netfilter/nf_conntrack_l3proto.h | 4 +
net/bridge/br_netfilter_hooks.c | 68 +++++++++++++++++++++++--
net/ipv4/netfilter/arp_tables.c | 66 +++++++++++++++---------
net/ipv4/netfilter/arptable_filter.c | 39 ++++++++------
net/ipv4/netfilter/ip_tables.c | 63 +++++++++++++----------
net/ipv4/netfilter/ipt_CLUSTERIP.c | 4 -
net/ipv4/netfilter/ipt_MASQUERADE.c | 8 ++
net/ipv4/netfilter/ipt_SYNPROXY.c | 4 -
net/ipv4/netfilter/iptable_filter.c | 55 ++++++++++++--------
net/ipv4/netfilter/iptable_mangle.c | 41 ++++++++++-----
net/ipv4/netfilter/iptable_nat.c | 41 ++++++++-------
net/ipv4/netfilter/iptable_raw.c | 42 ++++++++++-----
net/ipv4/netfilter/iptable_security.c | 44 ++++++++++------
net/ipv4/netfilter/nf_conntrack_l3proto_ipv4.c | 62 ++++++++++++++++++----
net/ipv4/netfilter/nf_defrag_ipv4.c | 49 ++++++++++++++++--
net/ipv4/netfilter/nft_masq_ipv4.c | 7 ++
net/ipv4/netfilter/nft_redir_ipv4.c | 7 ++
net/ipv6/netfilter/ip6_tables.c | 65 ++++++++++++++---------
net/ipv6/netfilter/ip6t_SYNPROXY.c | 4 -
net/ipv6/netfilter/ip6table_filter.c | 47 ++++++++++-------
net/ipv6/netfilter/ip6table_mangle.c | 45 +++++++++-------
net/ipv6/netfilter/ip6table_nat.c | 41 ++++++++-------
net/ipv6/netfilter/ip6table_raw.c | 46 ++++++++++------
net/ipv6/netfilter/ip6table_security.c | 44 +++++++++-------
net/ipv6/netfilter/nf_conntrack_l3proto_ipv6.c | 61 +++++++++++++++++-----
net/ipv6/netfilter/nf_defrag_ipv6_hooks.c | 50 ++++++++++++++++--
net/ipv6/netfilter/nft_masq_ipv6.c | 7 ++
net/ipv6/netfilter/nft_redir_ipv6.c | 7 ++
net/netfilter/nf_conntrack_proto.c | 48 +++++++++++++++++
net/netfilter/nft_ct.c | 24 ++++----
net/netfilter/nft_masq.c | 5 +
net/netfilter/nft_nat.c | 11 +++-
net/netfilter/nft_redir.c | 2
net/netfilter/x_tables.c | 65 ++++++++++++++---------
net/netfilter/xt_CONNSECMARK.c | 4 -
net/netfilter/xt_CT.c | 6 +-
net/netfilter/xt_NETMAP.c | 11 +++-
net/netfilter/xt_REDIRECT.c | 12 +++-
net/netfilter/xt_TPROXY.c | 15 +++--
net/netfilter/xt_connbytes.c | 4 -
net/netfilter/xt_connlabel.c | 6 +-
net/netfilter/xt_connlimit.c | 6 +-
net/netfilter/xt_connmark.c | 8 +-
net/netfilter/xt_conntrack.c | 4 -
net/netfilter/xt_helper.c | 4 -
net/netfilter/xt_nat.c | 18 ++++++
net/netfilter/xt_socket.c | 33 ++++++++++--
net/netfilter/xt_state.c | 4 -
57 files changed, 970 insertions(+), 422 deletions(-)
next reply other threads:[~2015-10-23 10:43 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-10-23 10:43 Florian Westphal [this message]
2015-10-23 10:43 ` [PATCH v2 nf-next 1/9] netfilter: ingress: don't use nf_hook_list_active Florian Westphal
2015-11-06 18:33 ` Pablo Neira Ayuso
2015-10-23 10:43 ` [PATCH v2 nf-next 2/9] netfilter: add and use nf_ct_netns_get/put Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 3/9] netfilter: conntrack: register hooks in netns when needed by ruleset Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 4/9] netfilter: xtables: don't register xt hooks in namespace at init time Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 5/9] netfilter: defrag: only register defrag functionality if needed Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 6/9] netfilter: nat: add dependencies on conntrack module Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 7/9] netfilter: bridge: register hooks only when bridge is added Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 8/9] netfilter: don't call nf_hook_state_init/_hook_slow unless needed Florian Westphal
2015-10-23 10:43 ` [PATCH v2 nf-next 9/9] nftables: add conntrack dependencies for nat/masq/redir expressions Florian Westphal
2015-10-26 22:55 ` [PATCH v2 nf-next 0/9] netfilter: don't copy initns hooks to new namespaces Pablo Neira Ayuso
2015-10-26 23:09 ` Florian Westphal
2015-10-27 16:35 ` Florian Westphal
2015-10-28 12:39 ` Jan Engelhardt
2015-10-29 20:35 ` Pablo Neira Ayuso
2015-10-29 22:13 ` Florian Westphal
2015-11-02 13:00 ` Florian Westphal
2015-11-24 10:27 ` Pablo Neira Ayuso
2015-11-24 10:59 ` Florian Westphal
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=1445597006-11264-1-git-send-email-fw@strlen.de \
--to=fw@strlen.de \
--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).