git: 3a1f834b5228 - main - pf: Add code to enable filtering for locally delivered packets

From: Doug Rabson <dfr_at_FreeBSD.org>
Date: Tue, 20 Jun 2023 14:35:55 UTC
The branch main has been updated by dfr:

URL: https://cgit.FreeBSD.org/src/commit/?id=3a1f834b5228986a7c14fd60da13cf2700e80996

commit 3a1f834b5228986a7c14fd60da13cf2700e80996
Author:     Doug Rabson <dfr@FreeBSD.org>
AuthorDate: 2023-06-20 13:01:58 +0000
Commit:     Doug Rabson <dfr@FreeBSD.org>
CommitDate: 2023-06-20 14:34:01 +0000

    pf: Add code to enable filtering for locally delivered packets
    
    This is disabled by default since it potentially changes the behavior of
    existing filter rule sets. To enable this extra filter for packets being
    delivered locally, use:
    
            sysctl net.pf.filter_local=1
            service pf restart
    
    PR:             268717
    Reviewed-by:    kp
    MFC-after:      2 weeks
    Differential Revision: https://reviews.freebsd.org/D40373
---
 UPDATING                                     | 12 ++++++++++++
 sys/netpfil/pf/pf_ioctl.c                    | 20 ++++++++++++++++++++
 tests/sys/netpfil/common/utils.subr          |  3 +--
 tests/sys/netpfil/pf/fragmentation_compat.sh |  3 ++-
 tests/sys/netpfil/pf/fragmentation_pass.sh   |  3 ++-
 tests/sys/netpfil/pf/killstate.sh            | 24 ++++++++++++++++--------
 tests/sys/netpfil/pf/map_e.sh                |  3 ++-
 tests/sys/netpfil/pf/pass_block.sh           |  3 ++-
 tests/sys/netpfil/pf/pfsync.sh               |  1 +
 tests/sys/netpfil/pf/route_to.sh             |  3 ++-
 tests/sys/netpfil/pf/set_skip.sh             |  2 +-
 tests/sys/netpfil/pf/table.sh                |  6 ++++--
 12 files changed, 65 insertions(+), 18 deletions(-)

diff --git a/UPDATING b/UPDATING
index 1980411c1853..f4e13d97006d 100644
--- a/UPDATING
+++ b/UPDATING
@@ -27,6 +27,18 @@ NOTE TO PEOPLE WHO THINK THAT FreeBSD 14.x IS SLOW:
 	world, or to merely disable the most expensive debugging functionality
 	at runtime, run "ln -s 'abort:false,junk:false' /etc/malloc.conf".)
 
+20230619:
+	To enable pf rdr rules for connections initiated from the host, pf
+	filter rules can be optionally enabled for packets delivered
+	locally. This can change the behavior of rules which match packets
+	delivered to lo0. To enable this feature:
+
+		sysctl net.pf.filter_local=1
+		service pf restart
+
+	When enabled, its best to ensure that packets delivered locally are not
+	filtered, e.g. by adding a 'skip on lo' rule.
+
 20230613:
 	Improvements to libtacplus(8) mean that tacplus.conf(5) now
 	follows POSIX shell syntax rules. This may cause TACACS+
diff --git a/sys/netpfil/pf/pf_ioctl.c b/sys/netpfil/pf/pf_ioctl.c
index e76a92fb7e7f..b78c30aa4b8c 100644
--- a/sys/netpfil/pf/pf_ioctl.c
+++ b/sys/netpfil/pf/pf_ioctl.c
@@ -195,6 +195,12 @@ static MALLOC_DEFINE(M_PFRULE, "pf_rule", "pf(4) rules");
 #error PF_QNAME_SIZE must be equal to PF_TAG_NAME_SIZE
 #endif
 
+VNET_DEFINE_STATIC(bool, pf_filter_local) = false;
+#define V_pf_filter_local	VNET(pf_filter_local)
+SYSCTL_BOOL(_net_pf, OID_AUTO, filter_local, CTLFLAG_VNET | CTLFLAG_RW,
+    &VNET_NAME(pf_filter_local), false,
+    "Enable filtering for packets delivered to local network stack");
+
 static void		 pf_init_tagset(struct pf_tagset *, unsigned int *,
 			    unsigned int);
 static void		 pf_cleanup_tagset(struct pf_tagset *);
@@ -6682,6 +6688,13 @@ hook_pf(void)
 	pla.pa_hook = V_pf_ip4_out_hook;
 	ret = pfil_link(&pla);
 	MPASS(ret == 0);
+	if (V_pf_filter_local) {
+		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
+		pla.pa_head = V_inet_local_pfil_head;
+		pla.pa_hook = V_pf_ip4_out_hook;
+		ret = pfil_link(&pla);
+		MPASS(ret == 0);
+	}
 #endif
 #ifdef INET6
 	pha.pa_type = PFIL_TYPE_IP6;
@@ -6703,6 +6716,13 @@ hook_pf(void)
 	pla.pa_hook = V_pf_ip6_out_hook;
 	ret = pfil_link(&pla);
 	MPASS(ret == 0);
+	if (V_pf_filter_local) {
+		pla.pa_flags = PFIL_OUT | PFIL_HEADPTR | PFIL_HOOKPTR;
+		pla.pa_head = V_inet6_local_pfil_head;
+		pla.pa_hook = V_pf_ip6_out_hook;
+		ret = pfil_link(&pla);
+		MPASS(ret == 0);
+	}
 #endif
 
 	atomic_store_bool(&V_pf_pfil_hooked, true);
diff --git a/tests/sys/netpfil/common/utils.subr b/tests/sys/netpfil/common/utils.subr
index f4eec24618a7..e354f6638b87 100644
--- a/tests/sys/netpfil/common/utils.subr
+++ b/tests/sys/netpfil/common/utils.subr
@@ -55,11 +55,10 @@ firewall_config()
 		jexec ${jname} ipfw -q -f flush
 		jexec ${jname} /bin/sh $cwd/ipfw.rule
 	elif [ ${fw} == "pf" ]; then
+		jexec ${jname} sysctl net.pf.filter_local=1
 		jexec ${jname} pfctl -e
 		jexec ${jname} pfctl -F all
 		jexec ${jname} pfctl -f $cwd/pf.rule
-		jexec ${jname} pfilctl link -o pf:default-out inet-local
-		jexec ${jname} pfilctl link -o pf:default-out6 inet6-local
 	elif [ ${fw} == "ipf" ]; then
 		jexec ${jname} ipf -E
 		jexec ${jname} ipf -Fa -f $cwd/ipf.rule
diff --git a/tests/sys/netpfil/pf/fragmentation_compat.sh b/tests/sys/netpfil/pf/fragmentation_compat.sh
index a783755e4592..3e559a216b54 100644
--- a/tests/sys/netpfil/pf/fragmentation_compat.sh
+++ b/tests/sys/netpfil/pf/fragmentation_compat.sh
@@ -112,7 +112,8 @@ v6_body()
 		"scrub fragment reassemble" \
 		"block in" \
 		"pass in inet6 proto icmp6 icmp6-type { neighbrsol, neighbradv }" \
-		"pass in inet6 proto icmp6 icmp6-type { echoreq, echorep }"
+		"pass in inet6 proto icmp6 icmp6-type { echoreq, echorep }" \
+		"set skip on lo"
 
 	# Host test
 	atf_check -s exit:0 -o ignore \
diff --git a/tests/sys/netpfil/pf/fragmentation_pass.sh b/tests/sys/netpfil/pf/fragmentation_pass.sh
index d257de730d2d..e2d28c307502 100644
--- a/tests/sys/netpfil/pf/fragmentation_pass.sh
+++ b/tests/sys/netpfil/pf/fragmentation_pass.sh
@@ -116,7 +116,8 @@ v6_body()
 		"pass keep state" \
 		"block in" \
 		"pass in inet6 proto icmp6 icmp6-type { neighbrsol, neighbradv }" \
-		"pass in inet6 proto icmp6 icmp6-type { echoreq, echorep }"
+		"pass in inet6 proto icmp6 icmp6-type { echoreq, echorep }" \
+		"set skip on lo"
 
 	# Host test
 	atf_check -s exit:0 -o ignore \
diff --git a/tests/sys/netpfil/pf/killstate.sh b/tests/sys/netpfil/pf/killstate.sh
index 4263938e26be..cd4eeee05a10 100644
--- a/tests/sys/netpfil/pf/killstate.sh
+++ b/tests/sys/netpfil/pf/killstate.sh
@@ -60,7 +60,8 @@ v4_body()
 	jexec alcatraz pfctl -e
 
 	pft_set_rules alcatraz "block all" \
-		"pass in proto icmp"
+		"pass in proto icmp" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
@@ -126,7 +127,8 @@ v6_body()
 	jexec alcatraz pfctl -e
 
 	pft_set_rules alcatraz "block all" \
-		"pass in proto icmp6"
+		"pass in proto icmp6" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
@@ -189,7 +191,8 @@ label_body()
 
 	pft_set_rules alcatraz "block all" \
 		"pass in proto tcp label bar" \
-		"pass in proto icmp label foo"
+		"pass in proto icmp label foo" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
@@ -251,7 +254,8 @@ multilabel_body()
 	jexec alcatraz pfctl -e
 
 	pft_set_rules alcatraz "block all" \
-		"pass in proto icmp label foo label bar"
+		"pass in proto icmp label foo label bar" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
@@ -281,7 +285,8 @@ multilabel_body()
 	fi
 
 	pft_set_rules alcatraz "block all" \
-		"pass in proto icmp label foo label bar"
+		"pass in proto icmp label foo label bar" \
+		"set skip on lo"
 
 	# Reestablish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
@@ -329,7 +334,8 @@ gateway_body()
 	jexec alcatraz pfctl -e
 
 	pft_set_rules alcatraz "block all" \
-		"pass in reply-to (${epair}b 192.0.2.1) proto icmp"
+		"pass in reply-to (${epair}b 192.0.2.1) proto icmp" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	# Note: use pft_ping so we always use the same ID, so pf considers all
@@ -469,7 +475,8 @@ interface_body()
 	jexec alcatraz pfctl -e
 
 	pft_set_rules alcatraz "block all" \
-		"pass in proto icmp"
+		"pass in proto icmp" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
@@ -525,7 +532,8 @@ id_body()
 
 	pft_set_rules alcatraz "block all" \
 		"pass in proto tcp" \
-		"pass in proto icmp"
+		"pass in proto icmp" \
+		"set skip on lo"
 
 	# Sanity check & establish state
 	atf_check -s exit:0 -o ignore ${common_dir}/pft_ping.py \
diff --git a/tests/sys/netpfil/pf/map_e.sh b/tests/sys/netpfil/pf/map_e.sh
index ce0e567ae3c8..ea8ce33bf323 100644
--- a/tests/sys/netpfil/pf/map_e.sh
+++ b/tests/sys/netpfil/pf/map_e.sh
@@ -66,7 +66,8 @@ map_e_body()
 	pft_set_rules echo "block return all" \
 		"pass in on ${epair_echo}b inet proto tcp from 198.51.100.1 port 19720:19723 to (${epair_echo}b) port 7" \
 		"pass in on ${epair_echo}b inet proto tcp from 198.51.100.1 port 36104:36107 to (${epair_echo}b) port 7" \
-		"pass in on ${epair_echo}b inet proto tcp from 198.51.100.1 port 52488:52491 to (${epair_echo}b) port 7"
+		"pass in on ${epair_echo}b inet proto tcp from 198.51.100.1 port 52488:52491 to (${epair_echo}b) port 7" \
+		"set skip on lo"
 
 	i=0
 	while [ ${i} -lt ${NC_TRY_COUNT} ]
diff --git a/tests/sys/netpfil/pf/pass_block.sh b/tests/sys/netpfil/pf/pass_block.sh
index 0f034b23a730..2a226f5c9651 100644
--- a/tests/sys/netpfil/pf/pass_block.sh
+++ b/tests/sys/netpfil/pf/pass_block.sh
@@ -230,7 +230,8 @@ urpf_body()
 		--replyif ${epair_one}a
 
 	pft_set_rules alcatraz \
-		"block quick from urpf-failed"
+		"block quick from urpf-failed" \
+		"set skip on lo"
 	jexec alcatraz pfctl -e
 
 	# Correct source still works
diff --git a/tests/sys/netpfil/pf/pfsync.sh b/tests/sys/netpfil/pf/pfsync.sh
index 75788eed4bbe..1b61ec4f03a0 100644
--- a/tests/sys/netpfil/pf/pfsync.sh
+++ b/tests/sys/netpfil/pf/pfsync.sh
@@ -152,6 +152,7 @@ defer_body()
 	route add -net 203.0.113.0/24 198.51.100.1
 
 	# Enable pf
+	jexec alcatraz sysctl net.pf.filter_local=0
 	jexec alcatraz pfctl -e
 	pft_set_rules alcatraz \
 		"set skip on ${epair_sync}a" \
diff --git a/tests/sys/netpfil/pf/route_to.sh b/tests/sys/netpfil/pf/route_to.sh
index 203d0a944a5b..18e0e02db65e 100644
--- a/tests/sys/netpfil/pf/route_to.sh
+++ b/tests/sys/netpfil/pf/route_to.sh
@@ -230,7 +230,8 @@ multiwanlocal_body()
 		"block in"	\
 		"block out"	\
 		"pass out quick route-to (${epair_cl_two}a 203.0.113.129) inet proto tcp from 203.0.113.128 to any port 7" \
-		"pass out on ${epair_cl_one}a inet proto tcp from any to any port 7"
+		"pass out on ${epair_cl_one}a inet proto tcp from any to any port 7" \
+		"set skip on lo"
 
 	# This should work
 	result=$(jexec client nc -N -w 1 192.0.2.2 7 | wc -c)
diff --git a/tests/sys/netpfil/pf/set_skip.sh b/tests/sys/netpfil/pf/set_skip.sh
index 9b3d655a6d1d..c666622e3d15 100644
--- a/tests/sys/netpfil/pf/set_skip.sh
+++ b/tests/sys/netpfil/pf/set_skip.sh
@@ -101,7 +101,7 @@ set_skip_dynamic_body()
 	vnet_mkjail alcatraz
 	jexec alcatraz pfctl -e
 	pft_set_rules alcatraz "set skip on epair" \
-		"block"
+		"block on ! lo"
 
 	epair=$(vnet_mkepair)
 	ifconfig ${epair}a 192.0.2.2/24 up
diff --git a/tests/sys/netpfil/pf/table.sh b/tests/sys/netpfil/pf/table.sh
index 64dbd3a36201..b820d0c11e75 100644
--- a/tests/sys/netpfil/pf/table.sh
+++ b/tests/sys/netpfil/pf/table.sh
@@ -52,7 +52,8 @@ v4_counters_body()
 	    "table <foo> counters { 192.0.2.1 }" \
 	    "block all" \
 	    "pass in from <foo> to any" \
-	    "pass out from any to <foo>"
+	    "pass out from any to <foo>" \
+	    "set skip on lo"
 
 	atf_check -s exit:0 -o ignore ping -c 3 192.0.2.2
 
@@ -91,7 +92,8 @@ v6_counters_body()
 	    "table <foo6> counters { 2001:db8:42::1 }" \
 	    "block all" \
 	    "pass in from <foo6> to any" \
-	    "pass out from any to <foo6>"
+	    "pass out from any to <foo6>" \
+	    "set skip on lo"
 
 	atf_check -s exit:0 -o ignore ping -6 -c 3 2001:db8:42::2