Skip to contentSkip to conversation

Implement a best-effort forward compat solution

PATCH In review Phil Sutter 1 participant 6 messages Aug 10, 2023 – Aug 11, 2023 netfilter

lore
Share & tools
lore

Phil Sutter

author reviewer
#0 lore raw

v2 cover Implement a best-effort forward compat solution

Initial attempts of keeping a compatible version of each rule in the kernel for being dumped so any old user space will be able to parse it despite what conversions to native expressions have taken place have
failed: The dump-only bytecode may contain a lookup expression, therefore requires updating and an extra set and so on. This will be a nightmare to maintain in kernel. Any alternative to this is not transparent to old user space which can't be touched in a scenario of $RANDOM old container has to parse the host's ruleset.

Instead of the above, follow a much simpler route by implementing a compat-mode into current *tables-nft which avoids any of the later internal changes which may prevent an old iptables-nft from parsing a kernel's rule correctly. An up to date host expecting outdated containers accessing its ruleset may create it in a compatible form, trading potential performance regressions in for compatibility.

Patch 1 is just prep work, patch 2 adds the core logic, patch 3 exposes it to CLI and patch 4 finally adds some testing.

This should resolve nfbz#1632[1], albeit requiring adjustments in how users call iptables.

[1] https://bugzilla.netfilter.org/show_bug.cgi?id=1632

Changes since v1:
- Rebase to current HEAD
- Add missing parser and man page adjustments in patch 3

Phil Sutter (4):
nft: Pass nft_handle to add_{target,action}()
nft: Introduce and use bool nft_handle::compat
Add --compat option to *tables-nft and *-nft-restore commands
tests: Test compat mode

 iptables-test.py                              | 19 ++++--
 iptables/arptables-nft-restore.8              | 15 +++--
 iptables/arptables-nft.8                      |  8 +++
 iptables/ebtables-nft.8                       |  6 ++
 iptables/iptables-restore.8.in                | 11 +++-
 iptables/iptables.8.in                        |  7 +++
 iptables/nft-arp.c                            |  2 +-
 iptables/nft-bridge.c                         |  9 +--
 iptables/nft-ipv4.c                           |  2 +-
 iptables/nft-ipv6.c                           |  2 +-
 iptables/nft-shared.c                         |  2 +-
 iptables/nft.c                                | 19 +++---
 iptables/nft.h                                |  7 ++-
 .../testcases/nft-only/0011-compat-mode_0     | 63 +++++++++++++++++++
 iptables/xshared.c                            |  7 ++-
 iptables/xshared.h                            |  1 +
 iptables/xtables-arp.c                        |  1 +
 iptables/xtables-eb.c                         |  7 ++-
 iptables/xtables-restore.c                    | 43 +++++++++++--
 iptables/xtables.c                            |  2 +
 20 files changed, 198 insertions(+), 35 deletions(-)
 create mode 100755 iptables/tests/shell/testcases/nft-only/0011-compat-mode_0
Signature · 1 line
2.40.0
author reviewer
#1 lore raw

PATCH v2 1/4 nft: Pass nft_handle to add_{target,action}()

Prepare for varying rule content based on a global flag.

---
 iptables/nft-arp.c    | 2 +-
 iptables/nft-bridge.c | 9 +++++----
 iptables/nft-ipv4.c   | 2 +-
 iptables/nft-ipv6.c   | 2 +-
 iptables/nft.c        | 9 +++++----
 iptables/nft.h        | 6 ++++--
 6 files changed, 17 insertions(+), 13 deletions(-)
iptables/nft-arp.c +1 -1
index 9868966a03688..14b352cebf9d3 100644
@@ -151,7 +151,7 @@ static int nft_arp_add(struct nft_handle *h, struct nft_rule_ctx *ctx,
 		else if (strcmp(cs->jumpto, XTC_LABEL_RETURN) == 0)
 			ret = add_verdict(r, NFT_RETURN);
 		else
-			ret = add_target(r, cs->target->t);
+			ret = add_target(h, r, cs->target->t);
 	} else if (strlen(cs->jumpto) > 0) {
 		/* No goto in arptables */
 		ret = add_jumpto(r, cs->jumpto, NFT_JUMP);
iptables/nft-bridge.c +5 -4
index 391a8ab723c1c..616ae5a3a2a3c 100644
@@ -117,7 +117,8 @@ static int add_meta_broute(struct nftnl_rule *r)
 	return 0;
 }
 
-static int _add_action(struct nftnl_rule *r, struct iptables_command_state *cs)
+static int _add_action(struct nft_handle *h, struct nftnl_rule *r,
+		       struct iptables_command_state *cs)
 {
 	const char *table = nftnl_rule_get_str(r, NFTNL_RULE_TABLE);
 
@@ -133,7 +134,7 @@ static int _add_action(struct nftnl_rule *r, struct iptables_command_state *cs)
 		}
 	}
 
-	return add_action(r, cs, false);
+	return add_action(h, r, cs, false);
 }
 
 static int
@@ -221,7 +222,7 @@ static int nft_bridge_add(struct nft_handle *h, struct nft_rule_ctx *ctx,
 			if (nft_bridge_add_match(h, fw, ctx, r, iter->u.match->m))
 				break;
 		} else {
-			if (add_target(r, iter->u.watcher->t))
+			if (add_target(h, r, iter->u.watcher->t))
 				break;
 		}
 	}
@@ -229,7 +230,7 @@ static int nft_bridge_add(struct nft_handle *h, struct nft_rule_ctx *ctx,
 	if (add_counters(r, cs->counters.pcnt, cs->counters.bcnt) < 0)
 		return -1;
 
-	return _add_action(r, cs);
+	return _add_action(h, r, cs);
 }
 
 static bool nft_rule_to_ebtables_command_state(struct nft_handle *h,
iptables/nft-ipv4.c +1 -1
index 2f10220edd509..663052fc57f0a 100644
@@ -95,7 +95,7 @@ static int nft_ipv4_add(struct nft_handle *h, struct nft_rule_ctx *ctx,
 	if (add_counters(r, cs->counters.pcnt, cs->counters.bcnt) < 0)
 		return -1;
 
-	return add_action(r, cs, !!(cs->fw.ip.flags & IPT_F_GOTO));
+	return add_action(h, r, cs, !!(cs->fw.ip.flags & IPT_F_GOTO));
 }
 
 static bool nft_ipv4_is_same(const struct iptables_command_state *a,
iptables/nft-ipv6.c +1 -1
index d53f87c1d26e3..8bc633df0e93a 100644
@@ -81,7 +81,7 @@ static int nft_ipv6_add(struct nft_handle *h, struct nft_rule_ctx *ctx,
 	if (add_counters(r, cs->counters.pcnt, cs->counters.bcnt) < 0)
 		return -1;
 
-	return add_action(r, cs, !!(cs->fw6.ipv6.flags & IP6T_F_GOTO));
+	return add_action(h, r, cs, !!(cs->fw6.ipv6.flags & IP6T_F_GOTO));
 }
 
 static bool nft_ipv6_is_same(const struct iptables_command_state *a,
iptables/nft.c +5 -4
index 97fd4f49fdb4c..1fc12b0c659c7 100644
@@ -1538,7 +1538,8 @@ static int add_meta_nftrace(struct nftnl_rule *r)
 	return 0;
 }
 
-int add_target(struct nftnl_rule *r, struct xt_entry_target *t)
+int add_target(struct nft_handle *h, struct nftnl_rule *r,
+	       struct xt_entry_target *t)
 {
 	struct nftnl_expr *expr;
 	int ret;
@@ -1587,8 +1588,8 @@ int add_verdict(struct nftnl_rule *r, int verdict)
 	return 0;
 }
 
-int add_action(struct nftnl_rule *r, struct iptables_command_state *cs,
-	       bool goto_set)
+int add_action(struct nft_handle *h, struct nftnl_rule *r,
+	       struct iptables_command_state *cs, bool goto_set)
 {
 	int ret = 0;
 
@@ -1604,7 +1605,7 @@ int add_action(struct nftnl_rule *r, struct iptables_command_state *cs,
 		else if (strcmp(cs->jumpto, "NFLOG") == 0)
 			ret = add_log(r, cs);
 		else
-			ret = add_target(r, cs->target->t);
+			ret = add_target(h, r, cs->target->t);
 	} else if (strlen(cs->jumpto) > 0) {
 		/* Not standard, then it's a go / jump to chain */
 		if (goto_set)
iptables/nft.h +4 -2
index 5acbbf82e2c29..a89aff0af68d0 100644
@@ -192,9 +192,11 @@ int add_counters(struct nftnl_rule *r, uint64_t packets, uint64_t bytes);
 int add_verdict(struct nftnl_rule *r, int verdict);
 int add_match(struct nft_handle *h, struct nft_rule_ctx *ctx,
 	      struct nftnl_rule *r, struct xt_entry_match *m);
-int add_target(struct nftnl_rule *r, struct xt_entry_target *t);
+int add_target(struct nft_handle *h, struct nftnl_rule *r,
+	       struct xt_entry_target *t);
 int add_jumpto(struct nftnl_rule *r, const char *name, int verdict);
-int add_action(struct nftnl_rule *r, struct iptables_command_state *cs, bool goto_set);
+int add_action(struct nft_handle *h, struct nftnl_rule *r,
+	       struct iptables_command_state *cs, bool goto_set);
 int add_log(struct nftnl_rule *r, struct iptables_command_state *cs);
 char *get_comment(const void *data, uint32_t data_len);
 
Signature · 1 line
2.40.0
author reviewer
#2 lore raw

PATCH v2 2/4 nft: Introduce and use bool nft_handle::compat

In reply to Phil Sutter

If set, create rules using compat expressions where possible and disable the bitwise expression avoidance introduced in 323259001d617 ("nft:
Optimize class-based IP prefix matches").

---
 iptables/nft-shared.c |  2 +-
 iptables/nft.c        | 10 ++++++----
 iptables/nft.h        |  1 +
 3 files changed, 8 insertions(+), 5 deletions(-)
iptables/nft-shared.c +1 -1
index 34ca9d16569d0..5e0ca00e7dd36 100644
@@ -198,7 +198,7 @@ void add_addr(struct nft_handle *h, struct nftnl_rule *r,
 
 	for (i = 0; i < len; i++) {
 		if (m[i] != 0xff) {
-			bitwise = m[i] != 0;
+			bitwise = h->compat || m[i] != 0;
 			break;
 		}
 	}
iptables/nft.c +6 -4
index 1fc12b0c659c7..09ff9cf11e195 100644
@@ -1476,10 +1476,12 @@ int add_match(struct nft_handle *h, struct nft_rule_ctx *ctx,
 	case NFT_COMPAT_RULE_APPEND:
 	case NFT_COMPAT_RULE_INSERT:
 	case NFT_COMPAT_RULE_REPLACE:
-		if (!strcmp(m->u.user.name, "limit"))
-			return add_nft_limit(r, m);
-		else if (!strcmp(m->u.user.name, "among"))
+		if (!strcmp(m->u.user.name, "among"))
 			return add_nft_among(h, r, m);
+		else if (h->compat)
+			break;
+		else if (!strcmp(m->u.user.name, "limit"))
+			return add_nft_limit(r, m);
 		else if (!strcmp(m->u.user.name, "udp"))
 			return add_nft_udp(h, r, m);
 		else if (!strcmp(m->u.user.name, "tcp"))
@@ -1544,7 +1546,7 @@ int add_target(struct nft_handle *h, struct nftnl_rule *r,
 	struct nftnl_expr *expr;
 	int ret;
 
-	if (strcmp(t->u.user.name, "TRACE") == 0)
+	if (!h->compat && strcmp(t->u.user.name, "TRACE") == 0)
 		return add_meta_nftrace(r);
 
 	expr = nftnl_expr_alloc("target");
iptables/nft.h +1 -0
index a89aff0af68d0..fb9fc81ea2704 100644
@@ -111,6 +111,7 @@ struct nft_handle {
 	struct list_head	cmd_list;
 	bool			cache_init;
 	int			verbose;
+	bool			compat;
 
 	/* meta data, for error reporting */
 	struct {
Signature · 1 line
2.40.0
author reviewer
#3 lore raw

PATCH v2 3/4 Add --compat option to *tables-nft and *-nft-restore commands

In reply to Phil Sutter

The flag sets nft_handle::compat boolean, indicating a compatible rule implementation is wanted. Users expecting their created rules to be fetched from kernel by an older version of *tables-nft may use this to avoid potential compatibility issues.

---
Changes since v1:
- Expect short option '-C' in {ip,ip6,eb}tables-nft-restore command line parser
- Support -C/--compat in arptables-nft-restore, too
- Update man pages with the new flag

---
 iptables/arptables-nft-restore.8 | 15 +++++++----
 iptables/arptables-nft.8         |  8 ++++++
 iptables/ebtables-nft.8          |  6 +++++
 iptables/iptables-restore.8.in   | 11 ++++++--
 iptables/iptables.8.in           |  7 ++++++
 iptables/xshared.c               |  7 +++++-
 iptables/xshared.h               |  1 +
 iptables/xtables-arp.c           |  1 +
 iptables/xtables-eb.c            |  7 +++++-
 iptables/xtables-restore.c       | 43 +++++++++++++++++++++++++++++---
 iptables/xtables.c               |  2 ++
 11 files changed, 95 insertions(+), 13 deletions(-)
iptables/arptables-nft-restore.8 +10 -5
index 09d9082cf9fd3..12ac9ebd5062d 100644
@@ -22,18 +22,23 @@
 .SH NAME
 arptables-restore \- Restore ARP Tables (nft-based)
 .SH SYNOPSIS
-\fBarptables\-restore
+.BR arptables\-restore " [" --compat ]
 .SH DESCRIPTION
-.PP
 .B arptables-restore
 is used to restore ARP Tables from data specified on STDIN or
 via a file as first argument.
-Use I/O redirection provided by your shell to read from a file
-.TP
+Use I/O redirection provided by your shell to read from a file.
+.P
 .B arptables-restore
 flushes (deletes) all previous contents of the respective ARP Table.
+.TP
+.BR -C , " --compat"
+Create rules in a mostly compatible way, enabling older versions of
+\fBarptables\-nft\fP to correctly parse the rules received from kernel. This
+mode is only useful in very specific situations and will likely impact packet
+filtering performance.
+
 .SH AUTHOR
 Jesper Dangaard Brouer <[email protected]>
 .SH SEE ALSO
 \fBarptables\-save\fP(8), \fBarptables\fP(8)
-.PP
iptables/arptables-nft.8 +8 -0
index ea31e0842acd4..673a7bd58e9cd 100644
@@ -220,6 +220,14 @@ counters of a rule (during
 .B APPEND,
 .B REPLACE
 operations).
+.SS "OTHER OPTIONS"
+The following additional options can be specified:
+.TP
+\fB\-\-compat\fP
+Create rules in a mostly compatible way, enabling older versions of
+\fBarptables\-nft\fP to correctly parse the rules received from kernel. This
+mode is only useful in very specific situations and will likely impact packet
+filtering performance.
 
 .SS RULE-SPECIFICATIONS
 The following command line arguments make up a rule specification (as used 
iptables/ebtables-nft.8 +6 -0
index 0304b5088cd8c..baada6c67437f 100644
@@ -359,6 +359,12 @@ to try to automatically load missing kernel modules.
 .TP
 .B --concurrent
 Use a file lock to support concurrent scripts updating the ebtables kernel tables.
+.TP
+.B --compat
+Create rules in a mostly compatible way, enabling older versions of
+\fBebtables\-nft\fP to correctly parse the rules received from kernel. This
+mode is only useful in very specific situations and will likely impact packet
+filtering performance.
 
 .SS
 RULE SPECIFICATIONS
iptables/iptables-restore.8.in +9 -2
index aa816f794d6f3..383099929e3bd 100644
@@ -23,11 +23,11 @@ iptables-restore \(em Restore IP Tables
 .P
 ip6tables-restore \(em Restore IPv6 Tables
 .SH SYNOPSIS
-\fBiptables\-restore\fP [\fB\-chntvV\fP] [\fB\-w\fP \fIseconds\fP]
+\fBiptables\-restore\fP [\fB\-cChntvV\fP] [\fB\-w\fP \fIseconds\fP]
 [\fB\-M\fP \fImodprobe\fP] [\fB\-T\fP \fIname\fP]
 [\fIfile\fP]
 .P
-\fBip6tables\-restore\fP [\fB\-chntvV\fP] [\fB\-w\fP \fIseconds\fP]
+\fBip6tables\-restore\fP [\fB\-cChntvV\fP] [\fB\-w\fP \fIseconds\fP]
 [\fB\-M\fP \fImodprobe\fP] [\fB\-T\fP \fIname\fP]
 [\fIfile\fP]
 .SH DESCRIPTION
@@ -74,6 +74,13 @@ determine the executable's path.
 .TP
 \fB\-T\fP, \fB\-\-table\fP \fIname\fP
 Restore only the named table even if the input stream contains other ones.
+.TP
+\fB\-C\fP, \fB\-\-compat\fP
+This flag is only relevant with \fBnft\fP-variants and ignored otherwise. If
+set, rules will be created in a mostly compatible way, enabling older versions
+of \fBiptables\-nft\fP to correctly parse the rules received from kernel. This
+mode is only useful in very specific situations and will likely impact packet
+filtering performance.
 .SH BUGS
 None known as of iptables-1.2.1 release
 .SH AUTHORS
iptables/iptables.8.in +7 -0
index ecaa5553942df..c0e92f27dc722 100644
@@ -397,6 +397,13 @@ corresponding to that rule's position in the chain.
 \fB\-\-modprobe=\fP\fIcommand\fP
 When adding or inserting rules into a chain, use \fIcommand\fP
 to load any necessary modules (targets, match extensions, etc).
+.TP
+\fB\-\-compat\fP
+This flag is only relevant with \fBnft\fP-variants and ignored otherwise. If
+set, rules will be created in a mostly compatible way, enabling older versions
+of \fBiptables\-nft\fP to correctly parse the rules received from kernel. This
+mode is only useful in very specific situations and will likely impact packet
+filtering performance.
 
 .SH LOCK FILE
 iptables uses the \fI@XT_LOCK_NAME@\fP file to take an exclusive lock at
iptables/xshared.c +6 -1
index 5f75a0a57a023..74b7a041516df 100644
@@ -1263,7 +1263,8 @@ xtables_printhelp(const struct xtables_rule_match *matches)
 	printf(
 "  --modprobe=<command>		try to insert modules using this command\n"
 "  --set-counters -c PKTS BYTES	set the counter during insert/append\n"
-"[!] --version	-V		print package version.\n");
+"[!] --version	-V		print package version\n"
+"  --compat			create rules compatible for parsing with old binaries\n");
 
 	if (afinfo->family == NFPROTO_ARP) {
 		int i;
@@ -1787,6 +1788,10 @@ void do_parse(int argc, char *argv[],
 
 			exit_tryhelp(2, p->line);
 
+		case 15: /* --compat */
+			p->compat = true;
+			break;
+
 		case 1: /* non option */
 			if (optarg[0] == '!' && optarg[1] == '\0') {
 				if (invert)
iptables/xshared.h +1 -0
index a200e0d620ad3..f69a7b432d33f 100644
@@ -283,6 +283,7 @@ struct xt_cmd_parse {
 	int				line;
 	int				verbose;
 	bool				xlate;
+	bool				compat;
 	struct xt_cmd_parse_ops		*ops;
 };
 
iptables/xtables-arp.c +1 -0
index 71518a9cbdb6a..c6a9c6d68cb10 100644
@@ -78,6 +78,7 @@ static struct option original_opts[] = {
 	{ "line-numbers", 0, 0, '0' },
 	{ "modprobe", 1, 0, 'M' },
 	{ "set-counters", 1, 0, 'c' },
+	{ "compat", 0, 0, 15 },
 	{ 0 }
 };
 
iptables/xtables-eb.c +6 -1
index 08eec79d80400..ffd51efaaf0f0 100644
@@ -223,6 +223,7 @@ struct option ebt_original_options[] =
 	{ "init-table"     , no_argument      , 0, 11  },
 	{ "concurrent"     , no_argument      , 0, 13  },
 	{ "check"          , required_argument, 0, 14  },
+	{ "compat"         , no_argument      , 0, 15  },
 	{ 0 }
 };
 
@@ -335,7 +336,8 @@ static void print_help(const struct xtables_target *t,
 "--modprobe -M program         : try to insert modules using this program\n"
 "--concurrent                  : use a file lock to support concurrent scripts\n"
 "--verbose -v                  : verbose mode\n"
-"--version -V                  : print package version\n\n"
+"--version -V                  : print package version\n"
+"--compat                      : create rules compatible for parsing with old binaries\n\n"
 "Environment variable:\n"
 /*ATOMIC_ENV_VARIABLE "          : if set <FILE> (see above) will equal its value"*/
 "\n\n");
@@ -1097,6 +1099,9 @@ int do_commandeb(struct nft_handle *h, int argc, char *argv[], char **table,
 			return 1;
 		case 13 :
 			break;
+		case 15:
+			h->compat = true;
+			break;
 		case 1 :
 			if (!strcmp(optarg, "!"))
 				ebt_check_inverse2(optarg, argc, argv);
iptables/xtables-restore.c +39 -4

124 lines not on this page · raw message

iptables/xtables.c +2 -0
index 22d6ea58376fc..25b4dbc6b8475 100644
@@ -82,6 +82,7 @@ static struct option original_opts[] = {
 	{.name = "goto",	  .has_arg = 1, .val = 'g'},
 	{.name = "ipv4",	  .has_arg = 0, .val = '4'},
 	{.name = "ipv6",	  .has_arg = 0, .val = '6'},
+	{.name = "compat",        .has_arg = 0, .val = 15 },
 	{NULL},
 };
 
@@ -161,6 +162,7 @@ int do_commandx(struct nft_handle *h, int argc, char *argv[], char **table,
 
 	do_parse(argc, argv, &p, &cs, &args);
 	h->verbose = p.verbose;
+	h->compat = p.compat;
 
 	if (!nft_table_builtin_find(h, p.table))
 		xtables_error(VERSION_PROBLEM,
Signature · 1 line
2.40.0
author reviewer
#4 lore raw

PATCH v2 4/4 tests: Test compat mode

In reply to Phil Sutter

Extend iptables-test.py by a third mode, which is using xtables-nft-multi and passing --compat to all calls creating rules.

Also add a shell testcase asserting the effectiveness of --compat by comparing debug (-vv) output.

---
 iptables-test.py                              | 19 ++++--
 .../testcases/nft-only/0011-compat-mode_0     | 63 +++++++++++++++++++
 2 files changed, 78 insertions(+), 4 deletions(-)
 create mode 100755 iptables/tests/shell/testcases/nft-only/0011-compat-mode_0
iptables-test.py +15 -4
index 6f63cdbeda9af..22b445df00b9c 100755
@@ -28,6 +28,8 @@ EBTABLES_SAVE = "ebtables-save"
 #IPTABLES_SAVE = ['xtables-save','-4']
 #IP6TABLES_SAVE = ['xtables-save','-6']
 
+COMPAT_ARG = ""
+
 EXTENSIONS_PATH = "extensions"
 LOGFILE="/tmp/iptables-test.log"
 log_file = None
@@ -83,7 +85,7 @@ STDERR_IS_TTY = sys.stderr.isatty()
     '''
     ret = 0
 
-    cmd = iptables + " -A " + rule
+    cmd = iptables + COMPAT_ARG + " -A " + rule
     ret = execute_cmd(cmd, filename, lineno, netns)
 
     #
@@ -318,7 +320,7 @@ STDERR_IS_TTY = sys.stderr.isatty()
 
     # load all rules via iptables_restore
 
-    command = EXECUTABLE + " " + iptables + "-restore"
+    command = EXECUTABLE + " " + iptables + "-restore" + COMPAT_ARG
     if netns:
         command = "ip netns exec " + netns + " " + command
 
@@ -558,6 +560,8 @@ STDERR_IS_TTY = sys.stderr.isatty()
                         help='Check for missing tests')
     parser.add_argument('-n', '--nftables', action='store_true',
                         help='Test iptables-over-nftables')
+    parser.add_argument('-c', '--nft-compat', action='store_true',
+                        help='Test iptables-over-nftables in compat mode')
     parser.add_argument('-N', '--netns', action='store_const',
                         const='____iptables-container-test',
                         help='Test netnamespace path')
@@ -577,8 +581,10 @@ STDERR_IS_TTY = sys.stderr.isatty()
         variants.append("legacy")
     if args.nftables:
         variants.append("nft")
+    if args.nft_compat:
+        variants.append("nft_compat")
     if len(variants) == 0:
-        variants = [ "legacy", "nft" ]
+        variants = [ "legacy", "nft", "nft_compat" ]
 
     if os.getuid() != 0:
         print("You need to be root to run this, sorry", file=sys.stderr)
@@ -598,7 +604,12 @@ STDERR_IS_TTY = sys.stderr.isatty()
     total_tests = 0
     for variant in variants:
         global EXECUTABLE
-        EXECUTABLE = "xtables-" + variant + "-multi"
+        global COMPAT_ARG
+        if variant == "nft_compat":
+            EXECUTABLE = "xtables-nft-multi"
+            COMPAT_ARG = " --compat"
+        else:
+            EXECUTABLE = "xtables-" + variant + "-multi"
 
         test_files = 0
         tests = 0
iptables/tests/shell/testcases/nft-only/0011-compat-mode_0 new file +63 -0

63 lines not on this page · raw message

Signature · 1 line
2.40.0
author reviewer
#5 lore raw
In reply to Phil Sutter
Quoted 28 lines · Phil Sutter

Initial attempts of keeping a compatible version of each rule in the kernel for being dumped so any old user space will be able to parse it despite what conversions to native expressions have taken place have
failed: The dump-only bytecode may contain a lookup expression, therefore requires updating and an extra set and so on. This will be a nightmare to maintain in kernel. Any alternative to this is not transparent to old user space which can't be touched in a scenario of $RANDOM old container has to parse the host's ruleset.

Instead of the above, follow a much simpler route by implementing a compat-mode into current *tables-nft which avoids any of the later internal changes which may prevent an old iptables-nft from parsing a kernel's rule correctly. An up to date host expecting outdated containers accessing its ruleset may create it in a compatible form, trading potential performance regressions in for compatibility.

Patch 1 is just prep work, patch 2 adds the core logic, patch 3 exposes it to CLI and patch 4 finally adds some testing.

This should resolve nfbz#1632[1], albeit requiring adjustments in how users call iptables.

[1] https://bugzilla.netfilter.org/show_bug.cgi?id=1632

Changes since v1:
- Rebase to current HEAD
- Add missing parser and man page adjustments in patch 3

Quoted 5 lines

Phil Sutter (4):
nft: Pass nft_handle to add_{target,action}()
nft: Introduce and use bool nft_handle::compat
Add --compat option to *tables-nft and *-nft-restore commands
tests: Test compat mode

Series applied.