net/sched: Handle TC_ACT_REDIRECT from qdisc filter chains

When a TC filter attached to a qdisc filter chain returns TC_ACT_REDIRECT (ex: via an eBPF program calling bpf_redirect() or an act_bpf action), the redirect was silently lost i.e no qdisc classify function handled TC_ACT_REDIRECT, so the packet fell through the switch and was enqueued normally instead of being redirected. This has been broken since bpf_redirect() was introduced for TC in commit 27b29f63058d ("bpf: add bpf_redirect() helper"). We got lucky for a long time because bpf_net_context was a per-CPU variable that was always available. commit 401cb7dae813 ("net: Reference bpf_redirect_info via task_struct on PREEMPT_RT.") turned bpf_net_context into a task_struct member that is only set up by explicit callers. Without a caller setting it up, bpf_redirect() itself crashes with a NULL pointer dereference in bpf_net_ctx_get_ri(). However, even with bpf_net_context available, TC_ACT_REDIRECT from qdisc filter chains cannot be honored without adding skb_do_redirect() calls to every qdisc classify function, which would require changes across net/sched/. Isolate it to ebpf core where it belongs. Instead, add a tcf_classify_qdisc() inline helper in pkt_cls.h, as a wrapper around tcf_classify() for use by qdisc classify functions and tcf_qevent_handle(). When the classify verdict is TC_ACT_REDIRECT, the wrapper converts it to TC_ACT_SHOT, dropping the packet rather than letting it continue silently. Dropping is preferred over letting the packet through because the user immediately sees packet loss. Silently passing the packet through would hide the problem and leave the user wondering why their redirect is not working. The clsact fast path, tc_run() continues to call tcf_classify() directly and is unaffected: TC_ACT_REDIRECT is returned as-is and handled by sch_handle_egress/ingress() calling skb_do_redirect() as before.

Jamal Hadi Salim committed 4 months ago

Commit ec48b3b1 parent 3f4920d15 files · +27 −17Browse files

Unified · Split

15 files

include/net/pkt_cls.h +13 −1
net/sched/cls_api.c +1 −3
net/sched/sch_cake.c +1 −1
net/sched/sch_drr.c +1 −1
net/sched/sch_dualpi2.c +1 −1
net/sched/sch_ets.c +1 −1
net/sched/sch_fq_codel.c +1 −1
net/sched/sch_fq_pie.c +1 −1
net/sched/sch_hfsc.c +1 −1
net/sched/sch_htb.c +1 −1
net/sched/sch_multiq.c +1 −1
net/sched/sch_prio.c +1 −1
net/sched/sch_qfq.c +1 −1
net/sched/sch_sfb.c +1 −1
net/sched/sch_sfq.c +1 −1

include/net/pkt_cls.h+13−1

@@ -156,8 +156,20 @@ static inline int tcf_classify(struct sk_buff *skb, · 155 unmodified lines
156	156	{
157	157	return TC_ACT_UNSPEC;
158	158	}
159		−
160	159	#endif
	160	+static inline int tcf_classify_qdisc(struct sk_buff *skb,
	161	+ const struct tcf_proto *tp,
	162	+ struct tcf_result *res, bool compat_mode)
	163	+{
	164	+ int ret = tcf_classify(skb, NULL, tp, res, compat_mode);
	165	+
	166	+ /* TC_ACT_REDIRECT from qdisc filter chains is not supported.
	167	+ * Use BPF via tcx or mirred redirect instead.
	168	+ */
	169	+ if (unlikely(ret == TC_ACT_REDIRECT))
	170	+ ret = TC_ACT_SHOT;
	171	+ return ret;
	172	+}
161	173
162	174	static inline unsigned long
163	175	__cls_set_class(unsigned long *clp, unsigned long cl)

net/sched/cls_api.c+1−3

@@ -4045,9 +4045,7 @@ struct sk_buff tcf_qevent_handle(struct tcf_qevent qe, struct Qdisc *sch, stru · 4044 unmodified lines
4045	4045
4046	4046	fl = rcu_dereference_bh(qe->filter_chain);
4047	4047
4048		− switch (tcf_classify(skb, NULL, fl, &cl_res, false)) {
4049		− case TC_ACT_REDIRECT:
4050		− fallthrough;
	4048	+ switch (tcf_classify_qdisc(skb, fl, &cl_res, false)) {
4051	4049	case TC_ACT_SHOT:
4052	4050	qdisc_qstats_drop(sch);
4053	4051	__qdisc_drop(skb, to_free);

net/sched/sch_cake.c+1−1

@@ -1727,7 +1727,7 @@ static u32 cake_classify(struct Qdisc sch, struct cake_tin_data *t, · 1726 unmodified lines
1727	1727	goto hash;
1728	1728
1729	1729	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
1730		− result = tcf_classify(skb, NULL, filter, &res, false);
	1730	+ result = tcf_classify_qdisc(skb, filter, &res, false);
1731	1731
1732	1732	if (result >= 0) {
1733	1733	#ifdef CONFIG_NET_CLS_ACT

net/sched/sch_drr.c+1−1

@@ -312,7 +312,7 @@ static struct drr_class drr_classify(struct sk_buff skb, struct Qdisc *sch, · 311 unmodified lines
312	312
313	313	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
314	314	fl = rcu_dereference_bh(q->filter_list);
315		− result = tcf_classify(skb, NULL, fl, &res, false);
	315	+ result = tcf_classify_qdisc(skb, fl, &res, false);
316	316	if (result >= 0) {
317	317	#ifdef CONFIG_NET_CLS_ACT
318	318	switch (result) {

net/sched/sch_dualpi2.c+1−1

@@ -364,7 +364,7 @@ static int dualpi2_skb_classify(struct dualpi2_sched_data *q, · 363 unmodified lines
364	364	if (!fl)
365	365	return NET_XMIT_SUCCESS;
366	366
367		− result = tcf_classify(skb, NULL, fl, &res, false);
	367	+ result = tcf_classify_qdisc(skb, fl, &res, false);
368	368	if (result >= 0) {
369	369	#ifdef CONFIG_NET_CLS_ACT
370	370	switch (result) {

net/sched/sch_ets.c+1−1

@@ -391,7 +391,7 @@ static struct ets_class ets_classify(struct sk_buff skb, struct Qdisc *sch, · 390 unmodified lines
391	391	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
392	392	if (TC_H_MAJ(skb->priority) != sch->handle) {
393	393	fl = rcu_dereference_bh(q->filter_list);
394		− err = tcf_classify(skb, NULL, fl, &res, false);
	394	+ err = tcf_classify_qdisc(skb, fl, &res, false);
395	395	#ifdef CONFIG_NET_CLS_ACT
396	396	switch (err) {
397	397	case TC_ACT_STOLEN:

net/sched/sch_fq_codel.c+1−1

@@ -91,7 +91,7 @@ static unsigned int fq_codel_classify(struct sk_buff skb, struct Qdisc sch, · 90 unmodified lines
91	91	return fq_codel_hash(q, skb) + 1;
92	92
93	93	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
94		− result = tcf_classify(skb, NULL, filter, &res, false);
	94	+ result = tcf_classify_qdisc(skb, filter, &res, false);
95	95	if (result >= 0) {
96	96	#ifdef CONFIG_NET_CLS_ACT
97	97	switch (result) {

net/sched/sch_fq_pie.c+1−1

@@ -96,7 +96,7 @@ static unsigned int fq_pie_classify(struct sk_buff skb, struct Qdisc sch, · 95 unmodified lines
96	96	return fq_pie_hash(q, skb) + 1;
97	97
98	98	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
99		− result = tcf_classify(skb, NULL, filter, &res, false);
	99	+ result = tcf_classify_qdisc(skb, filter, &res, false);
100	100	if (result >= 0) {
101	101	#ifdef CONFIG_NET_CLS_ACT
102	102	switch (result) {

net/sched/sch_hfsc.c+1−1

@@ -1143,7 +1143,7 @@ hfsc_classify(struct sk_buff skb, struct Qdisc sch, int *qerr) · 1142 unmodified lines
1143	1143	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
1144	1144	head = &q->root;
1145	1145	tcf = rcu_dereference_bh(q->root.filter_list);
1146		− while (tcf && (result = tcf_classify(skb, NULL, tcf, &res, false)) >= 0) {
	1146	+ while (tcf && (result = tcf_classify_qdisc(skb, tcf, &res, false)) >= 0) {
1147	1147	#ifdef CONFIG_NET_CLS_ACT
1148	1148	switch (result) {
1149	1149	case TC_ACT_QUEUED:

net/sched/sch_htb.c+1−1

@@ -243,7 +243,7 @@ static struct htb_class htb_classify(struct sk_buff skb, struct Qdisc *sch, · 242 unmodified lines
243	243	}
244	244
245	245	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
246		− while (tcf && (result = tcf_classify(skb, NULL, tcf, &res, false)) >= 0) {
	246	+ while (tcf && (result = tcf_classify_qdisc(skb, tcf, &res, false)) >= 0) {
247	247	#ifdef CONFIG_NET_CLS_ACT
248	248	switch (result) {
249	249	case TC_ACT_QUEUED:

net/sched/sch_multiq.c+1−1

@@ -36,7 +36,7 @@ multiq_classify(struct sk_buff skb, struct Qdisc sch, int *qerr) · 35 unmodified lines
36	36	int err;
37	37
38	38	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
39		− err = tcf_classify(skb, NULL, fl, &res, false);
	39	+ err = tcf_classify_qdisc(skb, fl, &res, false);
40	40	#ifdef CONFIG_NET_CLS_ACT
41	41	switch (err) {
42	42	case TC_ACT_STOLEN:

net/sched/sch_prio.c+1−1

@@ -39,7 +39,7 @@ prio_classify(struct sk_buff skb, struct Qdisc sch, int *qerr) · 38 unmodified lines
39	39	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
40	40	if (TC_H_MAJ(skb->priority) != sch->handle) {
41	41	fl = rcu_dereference_bh(q->filter_list);
42		− err = tcf_classify(skb, NULL, fl, &res, false);
	42	+ err = tcf_classify_qdisc(skb, fl, &res, false);
43	43	#ifdef CONFIG_NET_CLS_ACT
44	44	switch (err) {
45	45	case TC_ACT_STOLEN:

net/sched/sch_qfq.c+1−1

@@ -709,7 +709,7 @@ static struct qfq_class qfq_classify(struct sk_buff skb, struct Qdisc *sch, · 708 unmodified lines
709	709
710	710	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
711	711	fl = rcu_dereference_bh(q->filter_list);
712		− result = tcf_classify(skb, NULL, fl, &res, false);
	712	+ result = tcf_classify_qdisc(skb, fl, &res, false);
713	713	if (result >= 0) {
714	714	#ifdef CONFIG_NET_CLS_ACT
715	715	switch (result) {

net/sched/sch_sfb.c+1−1

@@ -260,7 +260,7 @@ static bool sfb_classify(struct sk_buff skb, struct tcf_proto fl, · 259 unmodified lines
260	260	struct tcf_result res;
261	261	int result;
262	262
263		− result = tcf_classify(skb, NULL, fl, &res, false);
	263	+ result = tcf_classify_qdisc(skb, fl, &res, false);
264	264	if (result >= 0) {
265	265	#ifdef CONFIG_NET_CLS_ACT
266	266	switch (result) {

net/sched/sch_sfq.c+1−1

@@ -171,7 +171,7 @@ static unsigned int sfq_classify(struct sk_buff skb, struct Qdisc sch, · 170 unmodified lines
171	171	return sfq_hash(q, skb) + 1;
172	172
173	173	*qerr = NET_XMIT_SUCCESS \| __NET_XMIT_BYPASS;
174		− result = tcf_classify(skb, NULL, fl, &res, false);
	174	+ result = tcf_classify_qdisc(skb, fl, &res, false);
175	175	if (result >= 0) {
176	176	#ifdef CONFIG_NET_CLS_ACT
177	177	switch (result) {

net/sched: Handle TC_ACT_REDIRECT from qdisc filter chains

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code