netfilter-devel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Phil Sutter <phil@nwl.cc>
To: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: netfilter-devel@vger.kernel.org
Subject: [iptables RFC 2/2] libxtables: Boost rule target checks by announcing chain names
Date: Fri,  4 Mar 2022 14:19:44 +0100	[thread overview]
Message-ID: <20220304131944.30801-3-phil@nwl.cc> (raw)
In-Reply-To: <20220304131944.30801-1-phil@nwl.cc>

When restoring a ruleset, feed libxtables with chain names from
respective lines to avoid extension searches for them when parsing rules
jumping to them later.

This is kind of a double-edged blade: the obvious downside is that
*tables-restore won't detect user-defined chain name and extension
clashes anymore. The upside is a tremendous performance improvement
restoring large rulesets. The same crooked ruleset as mentioned in
earlier patches (50k chains, 130k rules of which 90k jump to a chain)
yields these numbers:

variant	 unoptimized	non-targets cache	announced chains
----------------------------------------------------------------
legacy   1m12s		37s			2.5s
nft      1m35s		53s			8s

Note that iptables-legacy-restore allows the clashes already as long as
the name does not match a standard target, but with this patch it stops
warning about it. iptables-nft-restore does not care at all, even allows
adding a chain named 'ACCEPT' (and rules can't reach it because '-j
ACCEPT' translates to a native nftables verdict). The latter is a bug by
itself.

Signed-off-by: Phil Sutter <phil@nwl.cc>
---
 include/xtables.h           | 3 +++
 iptables/iptables-restore.c | 1 +
 iptables/xtables-restore.c  | 1 +
 libxtables/xtables.c        | 6 ++++++
 4 files changed, 11 insertions(+)

diff --git a/include/xtables.h b/include/xtables.h
index ca674c2663eb4..816a157d5577d 100644
--- a/include/xtables.h
+++ b/include/xtables.h
@@ -645,6 +645,9 @@ const char *xt_xlate_get(struct xt_xlate *xl);
 #define xt_xlate_rule_get xt_xlate_get
 const char *xt_xlate_set_get(struct xt_xlate *xl);
 
+/* informed target lookups */
+void xtables_announce_chain(const char *name);
+
 #ifdef XTABLES_INTERNAL
 
 /* Shipped modules rely on this... */
diff --git a/iptables/iptables-restore.c b/iptables/iptables-restore.c
index 1917fb2315665..4cf0d3dadead9 100644
--- a/iptables/iptables-restore.c
+++ b/iptables/iptables-restore.c
@@ -308,6 +308,7 @@ ip46tables_restore_main(const struct iptables_restore_cb *cb,
 						cb->ops->strerror(errno));
 			}
 
+			xtables_announce_chain(chain);
 			ret = 1;
 
 		} else if (in_table) {
diff --git a/iptables/xtables-restore.c b/iptables/xtables-restore.c
index 81b25a43c9a04..4f9ffefdfab22 100644
--- a/iptables/xtables-restore.c
+++ b/iptables/xtables-restore.c
@@ -201,6 +201,7 @@ static void xtables_restore_parse_line(struct nft_handle *h,
 				      policy, chain, line,
 				      strerror(errno));
 		}
+		xtables_announce_chain(chain);
 		ret = 1;
 	} else if (state->in_table) {
 		char *pcnt = NULL;
diff --git a/libxtables/xtables.c b/libxtables/xtables.c
index 3cb9a87c9406d..96ba16014af46 100644
--- a/libxtables/xtables.c
+++ b/libxtables/xtables.c
@@ -322,6 +322,12 @@ static void notargets_hlist_insert(const char *name)
 	hlist_add_head(&cur->node, &notargets[djb_hash(name) % NOTARGET_HSIZE]);
 }
 
+void xtables_announce_chain(const char *name)
+{
+	if (!notargets_hlist_lookup(name))
+		notargets_hlist_insert(name);
+}
+
 void xtables_init(void)
 {
 	/* xtables cannot be used with setuid in a safe way. */
-- 
2.34.1


  parent reply	other threads:[~2022-03-04 13:19 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-04 13:19 [iptables RFC 0/2] Speed up restoring huge rulesets Phil Sutter
2022-03-04 13:19 ` [iptables RFC 1/2] libxtables: Implement notargets hash table Phil Sutter
2022-03-10 12:17   ` Florian Westphal
2022-03-10 13:04     ` Phil Sutter
2022-03-04 13:19 ` Phil Sutter [this message]
2022-03-10 12:21   ` [iptables RFC 2/2] libxtables: Boost rule target checks by announcing chain names Florian Westphal
2022-03-10 13:57     ` Phil Sutter

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=20220304131944.30801-3-phil@nwl.cc \
    --to=phil@nwl.cc \
    --cc=netfilter-devel@vger.kernel.org \
    --cc=pablo@netfilter.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).