public inbox for netdev@vger.kernel.org
 help / color / mirror / Atom feed
From: Stephen Hemminger <stephen@networkplumber.org>
To: netdev@vger.kernel.org
Cc: Stephen Hemminger <stephen@networkplumber.org>,
	stable@vger.kernel.org, William Liu <will@willsroot.io>,
	Savino Dicanosa <savy@syst3mfailure.io>
Subject: [PATCH net v2 02/10] net/sched: netem: add per-CPU recursion guard for duplication
Date: Sat, 14 Mar 2026 17:14:06 -0700	[thread overview]
Message-ID: <20260315001649.23931-3-stephen@networkplumber.org> (raw)
In-Reply-To: <20260315001649.23931-1-stephen@networkplumber.org>

Add a per-CPU recursion depth counter to netem_enqueue(). When netem
duplicates a packet, the clone is re-enqueued at the root qdisc. If
the tree contains other netem instances, this can recurse without
bound, causing soft lockups and OOM.

This approach was previously considered but rejected on the grounds
that netem_dequeue calling enqueue on a child netem could bypass the
depth check. That concern does not apply: the child netem's
netem_enqueue() increments the same per-CPU counter, so the total
nesting depth across all netem instances in the call chain is tracked
correctly.

A depth limit of 4 is generous for any legitimate configuration.

Fixes: 0afb51e72855 ("[PKT_SCHED]: netem: reinsert for duplication")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=220774
Cc: stable@vger.kernel.org
Reported-by: William Liu <will@willsroot.io>
Reported-by: Savino Dicanosa <savy@syst3mfailure.io>

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 net/sched/sch_netem.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/net/sched/sch_netem.c b/net/sched/sch_netem.c
index 0ccf74a9cb82..085fa3ad6f83 100644
--- a/net/sched/sch_netem.c
+++ b/net/sched/sch_netem.c
@@ -21,6 +21,7 @@
 #include <linux/rtnetlink.h>
 #include <linux/reciprocal_div.h>
 #include <linux/rbtree.h>
+#include <linux/percpu.h>
 
 #include <net/gso.h>
 #include <net/netlink.h>
@@ -29,6 +30,15 @@
 
 #define VERSION "1.3"
 
+/*
+ * Limit for recursion from duplication.
+ * Duplicated packets are re-enqueued at the root qdisc, which may
+ * reach this or another netem instance, causing nested calls to
+ * netem_enqueue(). This per-CPU counter limits the total depth.
+ */
+static DEFINE_PER_CPU(unsigned int, netem_enqueue_depth);
+#define NETEM_RECURSION_LIMIT	4
+
 /*	Network Emulation Queuing algorithm.
 	====================================
 
@@ -460,6 +470,14 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	/* Do not fool qdisc_drop_all() */
 	skb->prev = NULL;
 
+	/* Guard against recursion from duplication re-injection. */
+	if (unlikely(this_cpu_inc_return(netem_enqueue_depth) >
+		     NETEM_RECURSION_LIMIT)) {
+		this_cpu_dec(netem_enqueue_depth);
+		qdisc_drop(skb, sch, to_free);
+		return NET_XMIT_DROP;
+	}
+
 	/* Random duplication */
 	if (q->duplicate && q->duplicate >= get_crandom(&q->dup_cor, &q->prng))
 		++count;
@@ -474,6 +492,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 	if (count == 0) {
 		qdisc_qstats_drop(sch);
 		__qdisc_drop(skb, to_free);
+		this_cpu_dec(netem_enqueue_depth);
 		return NET_XMIT_SUCCESS | __NET_XMIT_BYPASS;
 	}
 
@@ -529,6 +548,7 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		qdisc_drop_all(skb, sch, to_free);
 		if (skb2)
 			__qdisc_drop(skb2, to_free);
+		this_cpu_dec(netem_enqueue_depth);
 		return NET_XMIT_DROP;
 	}
 
@@ -643,8 +663,10 @@ static int netem_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 		/* Parent qdiscs accounted for 1 skb of size @prev_len */
 		qdisc_tree_reduce_backlog(sch, -(nb - 1), -(len - prev_len));
 	} else if (!skb) {
+		this_cpu_dec(netem_enqueue_depth);
 		return NET_XMIT_DROP;
 	}
+	this_cpu_dec(netem_enqueue_depth);
 	return NET_XMIT_SUCCESS;
 }
 
-- 
2.51.0


  parent reply	other threads:[~2026-03-15  0:17 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-03-15  0:14 [PATCH net v2 00/10] netem: fixes and selftests Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 01/10] Revert "net/sched: Restrict conditions for adding duplicating netems to qdisc tree" Stephen Hemminger
2026-03-15  0:14 ` Stephen Hemminger [this message]
2026-03-15  0:14 ` [PATCH net v2 03/10] selftests/tc-testing: update netem tests after check_netem_in_tree revert Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 04/10] net/sched: netem: restructure dequeue to avoid re-entrancy with child qdisc Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 05/10] net/sched: netem: fix probability gaps in 4-state loss model Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 06/10] net/sched: netem: fix slot delay calculation overflow Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 07/10] net/sched: netem: fix queue limit check to include reordered packets Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 08/10] net/sched: netem: null-terminate tfifo linear queue tail Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 09/10] net/sched: netem: only reseed PRNG when seed is explicitly provided Stephen Hemminger
2026-03-15  0:14 ` [PATCH net v2 10/10] selftests/tc-testing: add netem configuration and traffic tests Stephen Hemminger
2026-03-15 15:19 ` [PATCH net v2 00/10] netem: fixes and selftests Jakub Kicinski
2026-03-15 16:09   ` Stephen Hemminger

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=20260315001649.23931-3-stephen@networkplumber.org \
    --to=stephen@networkplumber.org \
    --cc=netdev@vger.kernel.org \
    --cc=savy@syst3mfailure.io \
    --cc=stable@vger.kernel.org \
    --cc=will@willsroot.io \
    /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