Merge tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

Daniel Borkmann says:

====================
pull-request: bpf 2024-07-09

The following pull-request contains BPF updates for your *net* tree.

We've added 3 non-merge commits during the last 1 day(s) which contain
a total of 5 files changed, 81 insertions(+), 11 deletions(-).

The main changes are:

1) Fix a use-after-free in a corner case where tcx_entry got released too
   early. Also add BPF test coverage along with the fix, from Daniel Borkmann.

2) Fix a kernel panic on Loongarch in sk_msg_recvmsg() which got triggered
   by running BPF sockmap selftests, from Geliang Tang.

bpf-for-netdev

* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf:
  skmsg: Skip zero length skb in sk_msg_recvmsg
  selftests/bpf: Extend tcx tests to cover late tcx_entry release
  bpf: Fix too early release of tcx_entry
====================

Link: https://patch.msgid.link/20240709091452.27840-1-daniel@iogearbox.netSigned-off-by: Paolo Abeni <pabeni@redhat.com>

include/net/tcx.h

@@ -13,7 +13,7 @@ struct mini_Qdisc;
 struct tcx_entry {
 	struct mini_Qdisc __rcu *miniq;
 	struct bpf_mprog_bundle bundle;
-	bool miniq_active;
+	u32 miniq_active;
 	struct rcu_head rcu;
 };
 
@@ -125,11 +125,16 @@ static inline void tcx_skeys_dec(bool ingress)
 	tcx_dec();
 }
 
-static inline void tcx_miniq_set_active(struct bpf_mprog_entry *entry,
-					const bool active)
+static inline void tcx_miniq_inc(struct bpf_mprog_entry *entry)
 {
 	ASSERT_RTNL();
-	tcx_entry(entry)->miniq_active = active;
+	tcx_entry(entry)->miniq_active++;
+}
+
+static inline void tcx_miniq_dec(struct bpf_mprog_entry *entry)
+{
+	ASSERT_RTNL();
+	tcx_entry(entry)->miniq_active--;
 }
 
 static inline bool tcx_entry_is_active(struct bpf_mprog_entry *entry)

net/core/skmsg.c

@@ -434,7 +434,8 @@ int sk_msg_recvmsg(struct sock *sk, struct sk_psock *psock, struct msghdr *msg,
 			page = sg_page(sge);
 			if (copied + copy > len)
 				copy = len - copied;
-			copy = copy_page_to_iter(page, sge->offset, copy, iter);
+			if (copy)
+				copy = copy_page_to_iter(page, sge->offset, copy, iter);
 			if (!copy) {
 				copied = copied ? copied : -EFAULT;
 				goto out;

net/sched/sch_ingress.c

@@ -91,7 +91,7 @@ static int ingress_init(struct Qdisc *sch, struct nlattr *opt,
 	entry = tcx_entry_fetch_or_create(dev, true, &created);
 	if (!entry)
 		return -ENOMEM;
-	tcx_miniq_set_active(entry, true);
+	tcx_miniq_inc(entry);
 	mini_qdisc_pair_init(&q->miniqp, sch, &tcx_entry(entry)->miniq);
 	if (created)
 		tcx_entry_update(dev, entry, true);
@@ -121,7 +121,7 @@ static void ingress_destroy(struct Qdisc *sch)
 	tcf_block_put_ext(q->block, sch, &q->block_info);
 
 	if (entry) {
-		tcx_miniq_set_active(entry, false);
+		tcx_miniq_dec(entry);
 		if (!tcx_entry_is_active(entry)) {
 			tcx_entry_update(dev, NULL, true);
 			tcx_entry_free(entry);
@@ -257,7 +257,7 @@ static int clsact_init(struct Qdisc *sch, struct nlattr *opt,
 	entry = tcx_entry_fetch_or_create(dev, true, &created);
 	if (!entry)
 		return -ENOMEM;
-	tcx_miniq_set_active(entry, true);
+	tcx_miniq_inc(entry);
 	mini_qdisc_pair_init(&q->miniqp_ingress, sch, &tcx_entry(entry)->miniq);
 	if (created)
 		tcx_entry_update(dev, entry, true);
@@ -276,7 +276,7 @@ static int clsact_init(struct Qdisc *sch, struct nlattr *opt,
 	entry = tcx_entry_fetch_or_create(dev, false, &created);
 	if (!entry)
 		return -ENOMEM;
-	tcx_miniq_set_active(entry, true);
+	tcx_miniq_inc(entry);
 	mini_qdisc_pair_init(&q->miniqp_egress, sch, &tcx_entry(entry)->miniq);
 	if (created)
 		tcx_entry_update(dev, entry, false);
@@ -302,7 +302,7 @@ static void clsact_destroy(struct Qdisc *sch)
 	tcf_block_put_ext(q->egress_block, sch, &q->egress_block_info);
 
 	if (ingress_entry) {
-		tcx_miniq_set_active(ingress_entry, false);
+		tcx_miniq_dec(ingress_entry);
 		if (!tcx_entry_is_active(ingress_entry)) {
 			tcx_entry_update(dev, NULL, true);
 			tcx_entry_free(ingress_entry);
@@ -310,7 +310,7 @@ static void clsact_destroy(struct Qdisc *sch)
 	}
 
 	if (egress_entry) {
-		tcx_miniq_set_active(egress_entry, false);
+		tcx_miniq_dec(egress_entry);
 		if (!tcx_entry_is_active(egress_entry)) {
 			tcx_entry_update(dev, NULL, false);
 			tcx_entry_free(egress_entry);

tools/testing/selftests/bpf/config

@@ -58,9 +58,12 @@ CONFIG_MPLS=y
 CONFIG_MPLS_IPTUNNEL=y
 CONFIG_MPLS_ROUTING=y
 CONFIG_MPTCP=y
+CONFIG_NET_ACT_SKBMOD=y
+CONFIG_NET_CLS=y
 CONFIG_NET_CLS_ACT=y
 CONFIG_NET_CLS_BPF=y
 CONFIG_NET_CLS_FLOWER=y
+CONFIG_NET_CLS_MATCHALL=y
 CONFIG_NET_FOU=y
 CONFIG_NET_FOU_IP_TUNNELS=y
 CONFIG_NET_IPGRE=y

tools/testing/selftests/bpf/prog_tests/tc_links.c

@@ -9,6 +9,8 @@
 #define ping_cmd "ping -q -c1 -w1 127.0.0.1 > /dev/null"
 
 #include "test_tc_link.skel.h"
+
+#include "netlink_helpers.h"
 #include "tc_helpers.h"
 
 void serial_test_tc_links_basic(void)
@@ -1787,6 +1789,65 @@ void serial_test_tc_links_ingress(void)
 	test_tc_links_ingress(BPF_TCX_INGRESS, false, false);
 }
 
+struct qdisc_req {
+	struct nlmsghdr  n;
+	struct tcmsg     t;
+	char             buf[1024];
+};
+
+static int qdisc_replace(int ifindex, const char *kind, bool block)
+{
+	struct rtnl_handle rth = { .fd = -1 };
+	struct qdisc_req req;
+	int err;
+
+	err = rtnl_open(&rth, 0);
+	if (!ASSERT_OK(err, "open_rtnetlink"))
+		return err;
+
+	memset(&req, 0, sizeof(req));
+	req.n.nlmsg_len = NLMSG_LENGTH(sizeof(struct tcmsg));
+	req.n.nlmsg_flags = NLM_F_CREATE | NLM_F_REPLACE | NLM_F_REQUEST;
+	req.n.nlmsg_type = RTM_NEWQDISC;
+	req.t.tcm_family = AF_UNSPEC;
+	req.t.tcm_ifindex = ifindex;
+	req.t.tcm_parent = 0xfffffff1;
+
+	addattr_l(&req.n, sizeof(req), TCA_KIND, kind, strlen(kind) + 1);
+	if (block)
+		addattr32(&req.n, sizeof(req), TCA_INGRESS_BLOCK, 1);
+
+	err = rtnl_talk(&rth, &req.n, NULL);
+	ASSERT_OK(err, "talk_rtnetlink");
+	rtnl_close(&rth);
+	return err;
+}
+
+void serial_test_tc_links_dev_chain0(void)
+{
+	int err, ifindex;
+
+	ASSERT_OK(system("ip link add dev foo type veth peer name bar"), "add veth");
+	ifindex = if_nametoindex("foo");
+	ASSERT_NEQ(ifindex, 0, "non_zero_ifindex");
+	err = qdisc_replace(ifindex, "ingress", true);
+	if (!ASSERT_OK(err, "attaching ingress"))
+		goto cleanup;
+	ASSERT_OK(system("tc filter add block 1 matchall action skbmod swap mac"), "add block");
+	err = qdisc_replace(ifindex, "clsact", false);
+	if (!ASSERT_OK(err, "attaching clsact"))
+		goto cleanup;
+	/* Heuristic: kern_sync_rcu() alone does not work; a wait-time of ~5s
+	 * triggered the issue without the fix reliably 100% of the time.
+	 */
+	sleep(5);
+	ASSERT_OK(system("tc filter add dev foo ingress matchall action skbmod swap mac"), "add filter");
+cleanup:
+	ASSERT_OK(system("ip link del dev foo"), "del veth");
+	ASSERT_EQ(if_nametoindex("foo"), 0, "foo removed");
+	ASSERT_EQ(if_nametoindex("bar"), 0, "bar removed");
+}
+
 static void test_tc_links_dev_mixed(int target)
 {
 	LIBBPF_OPTS(bpf_tc_opts, tc_opts, .handle = 1, .priority = 1);