Skip to contentSkip to conversation

tc: Ensure we have enough buffer space when sending filter netlink notifications

PATCH Applied Toke Høiland-Jørgensen 5 participants 8 messages Apr 7, 2025 – Apr 14, 2025

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tc: Ensure we have enough buffer space when sending filter netlink notifications

The tfilter_notify() and tfilter_del_notify() functions assume that NLMSG_GOODSIZE is always enough to dump the filter chain. This is not always the case, which can lead to silent notify failures (because the return code of tfilter_notify() is not always checked). In particular, this can lead to NLM_F_ECHO not being honoured even though an action succeeds, which forces userspace to create workarounds[0].

Fix this by increasing the message size if dumping the filter chain into the allocated skb fails. Use the size of the incoming skb as a size hint if set, so we can start at a larger value when appropriate.

To trigger this, run the following commands:

 # ip link add type veth
 # tc qdisc replace dev veth0 root handle 1: fq_codel
 # tc -echo filter add dev veth0 parent 1: u32 match u32 0 0 $(for i in $(seq 32); do echo action pedit munge ip dport set 22; done)

Before this fix, tc just returns:

Not a filter(cmd 2)

After the fix, we get the correct echo:

added filter dev veth0 parent 1: protocol all pref 49152 u32 chain 0 fh 800::800 order 2048 key ht 800 bkt 0 terminal flowid not_in_hw
match 00000000/00000000 at 0
action order 1: pedit action pass keys 1
index 1 ref 1 bind 1

	key #0  at 20: val 00000016 mask ffff0000

[repeated 32 times]

[0] https://github.com/openvswitch/ovs/commit/106ef21860c935e5e0017a88bf42b94025c4e511

---
 net/sched/cls_api.c | 66 ++++++++++++++++++++++++++++++---------------
 1 file changed, 45 insertions(+), 21 deletions(-)
net/sched/cls_api.c +45 -21
index 4f648af8cfaa..ecec0a1e1c1a 100644
@@ -2057,6 +2057,7 @@ static int tcf_fill_node(struct net *net, struct sk_buff *skb,
 	struct tcmsg *tcm;
 	struct nlmsghdr  *nlh;
 	unsigned char *b = skb_tail_pointer(skb);
+	int ret = -EMSGSIZE;
 
 	nlh = nlmsg_put(skb, portid, seq, event, sizeof(*tcm), flags);
 	if (!nlh)
@@ -2101,11 +2102,45 @@ static int tcf_fill_node(struct net *net, struct sk_buff *skb,
 
 	return skb->len;
 
+cls_op_not_supp:
+	ret = -EOPNOTSUPP;
 out_nlmsg_trim:
 nla_put_failure:
-cls_op_not_supp:
 	nlmsg_trim(skb, b);
-	return -1;
+	return ret;
+}
+
+static struct sk_buff *tfilter_notify_prep(struct net *net,
+					   struct sk_buff *oskb,
+					   struct nlmsghdr *n,
+					   struct tcf_proto *tp,
+					   struct tcf_block *block,
+					   struct Qdisc *q, u32 parent,
+					   void *fh, int event,
+					   u32 portid, bool rtnl_held,
+					   struct netlink_ext_ack *extack)
+{
+	unsigned int size = oskb ? max(NLMSG_GOODSIZE, oskb->len) : NLMSG_GOODSIZE;
+	struct sk_buff *skb;
+	int ret;
+
+retry:
+	skb = alloc_skb(size, GFP_KERNEL);
+	if (!skb)
+		return ERR_PTR(-ENOBUFS);
+
+	ret = tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
+			    n->nlmsg_seq, n->nlmsg_flags, event, false,
+			    rtnl_held, extack);
+	if (ret <= 0) {
+		kfree_skb(skb);
+		if (ret == -EMSGSIZE) {
+			size += NLMSG_GOODSIZE;
+			goto retry;
+		}
+		return ERR_PTR(-EINVAL);
+	}
+	return skb;
 }
 
 static int tfilter_notify(struct net *net, struct sk_buff *oskb,
@@ -2121,16 +2156,10 @@ static int tfilter_notify(struct net *net, struct sk_buff *oskb,
 	if (!unicast && !rtnl_notify_needed(net, n->nlmsg_flags, RTNLGRP_TC))
 		return 0;
 
-	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
-	if (!skb)
-		return -ENOBUFS;
-
-	if (tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
-			  n->nlmsg_seq, n->nlmsg_flags, event,
-			  false, rtnl_held, extack) <= 0) {
-		kfree_skb(skb);
-		return -EINVAL;
-	}
+	skb = tfilter_notify_prep(net, oskb, n, tp, block, q, parent, fh, event,
+				  portid, rtnl_held, extack);
+	if (IS_ERR(skb))
+		return PTR_ERR(skb);
 
 	if (unicast)
 		err = rtnl_unicast(skb, net, portid);
@@ -2153,16 +2182,11 @@ static int tfilter_del_notify(struct net *net, struct sk_buff *oskb,
 	if (!rtnl_notify_needed(net, n->nlmsg_flags, RTNLGRP_TC))
 		return tp->ops->delete(tp, fh, last, rtnl_held, extack);
 
-	skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
-	if (!skb)
-		return -ENOBUFS;
-
-	if (tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
-			  n->nlmsg_seq, n->nlmsg_flags, RTM_DELTFILTER,
-			  false, rtnl_held, extack) <= 0) {
+	skb = tfilter_notify_prep(net, oskb, n, tp, block, q, parent, fh,
+				  RTM_DELTFILTER, portid, rtnl_held, extack);
+	if (IS_ERR(skb)) {
 		NL_SET_ERR_MSG(extack, "Failed to build del event notification");
-		kfree_skb(skb);
-		return -EINVAL;
+		return PTR_ERR(skb);
 	}
 
 	err = tp->ops->delete(tp, fh, last, rtnl_held, extack);
Signature · 1 line
2.49.0
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 30 lines · Mon, Apr 07, 2025 at 12:55:34PM +0200, [email protected] wrote:

The tfilter_notify() and tfilter_del_notify() functions assume that NLMSG_GOODSIZE is always enough to dump the filter chain. This is not always the case, which can lead to silent notify failures (because the return code of tfilter_notify() is not always checked). In particular, this can lead to NLM_F_ECHO not being honoured even though an action succeeds, which forces userspace to create workarounds[0].

Fix this by increasing the message size if dumping the filter chain into the allocated skb fails. Use the size of the incoming skb as a size hint if set, so we can start at a larger value when appropriate.

To trigger this, run the following commands:

# ip link add type veth
# tc qdisc replace dev veth0 root handle 1: fq_codel
# tc -echo filter add dev veth0 parent 1: u32 match u32 0 0 $(for i in $(seq 32); do echo action pedit munge ip dport set 22; done)

Before this fix, tc just returns:

Not a filter(cmd 2)

After the fix, we get the correct echo:

added filter dev veth0 parent 1: protocol all pref 49152 u32 chain 0 fh 800::800 order 2048 key ht 800 bkt 0 terminal flowid not_in_hw
match 00000000/00000000 at 0
action order 1: pedit action pass keys 1
index 1 ref 1 bind 1
key #0 at 20: val 00000016 mask ffff0000
[repeated 32 times]

Quoted 6 lines

[0] https://github.com/openvswitch/ovs/commit/106ef21860c935e5e0017a88bf42b94025c4e511

maintainer
#2 lore raw
In reply to Toke Høiland-Jørgensen
Quoted 10 lines · Toke Høiland-Jørgensen

To trigger this, run the following commands:

 # ip link add type veth
 # tc qdisc replace dev veth0 root handle 1: fq_codel
 # tc -echo filter add dev veth0 parent 1: u32 match u32 0 0 $(for i in $(seq 32); do echo action pedit munge ip dport set 22; done)

Before this fix, tc just returns:

Not a filter(cmd 2)

Quoted 8 lines

After the fix, we get the correct echo:

added filter dev veth0 parent 1: protocol all pref 49152 u32 chain 0 fh 800::800 order 2048 key ht 800 bkt 0 terminal flowid not_in_hw
match 00000000/00000000 at 0
action order 1: pedit action pass keys 1
index 1 ref 1 bind 1

	key #0  at 20: val 00000016 mask ffff0000

[repeated 32 times]

I think it would be great if you could follow-up capturing the above in a self-test.

Thanks!

Paolo

#3 lore raw
Quoted 20 lines · Paolo Abeni
Quoted 10 lines · Toke Høiland-Jørgensen

To trigger this, run the following commands:

 # ip link add type veth
 # tc qdisc replace dev veth0 root handle 1: fq_codel
 # tc -echo filter add dev veth0 parent 1: u32 match u32 0 0 $(for i in $(seq 32); do echo action pedit munge ip dport set 22; done)

Before this fix, tc just returns:

Not a filter(cmd 2)

Quoted 8 lines

After the fix, we get the correct echo:

added filter dev veth0 parent 1: protocol all pref 49152 u32 chain 0 fh 800::800 order 2048 key ht 800 bkt 0 terminal flowid not_in_hw
match 00000000/00000000 at 0
action order 1: pedit action pass keys 1
index 1 ref 1 bind 1

	key #0  at 20: val 00000016 mask ffff0000

[repeated 32 times]

Quoted 2 lines

I think it would be great if you could follow-up capturing the above in a self-test.

Sure, can do!

-Toke

#4 lore raw
In reply to Toke Høiland-Jørgensen
This patch was applied to netdev/net.git (main) by Paolo Abeni <[email protected]>:

Hello:

This patch was applied to netdev/net.git (main)
by Paolo Abeni <[email protected]>:

Quoted 8 lines · On Mon, 7 Apr 2025 12:55:34 +0200 you wrote:

The tfilter_notify() and tfilter_del_notify() functions assume that NLMSG_GOODSIZE is always enough to dump the filter chain. This is not always the case, which can lead to silent notify failures (because the return code of tfilter_notify() is not always checked). In particular, this can lead to NLM_F_ECHO not being honoured even though an action succeeds, which forces userspace to create workarounds[0].

[...]

Here is the summary with links:
- [net] tc: Ensure we have enough buffer space when sending filter netlink notifications

    https://git.kernel.org/netdev/net/c/369609fc6272

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html
#5 lore raw
In reply to Toke Høiland-Jørgensen
Quoted 17 lines · Toke Høiland-Jørgensen
net/sched/cls_api.c +17 -0
+static struct sk_buff *tfilter_notify_prep(struct net *net,
+					   struct sk_buff *oskb,
+					   struct nlmsghdr *n,
+					   struct tcf_proto *tp,
+					   struct tcf_block *block,
+					   struct Qdisc *q, u32 parent,
+					   void *fh, int event,
+					   u32 portid, bool rtnl_held,
+					   struct netlink_ext_ack *extack)
+{
+	unsigned int size = oskb ? max(NLMSG_GOODSIZE, oskb->len) : NLMSG_GOODSIZE;
+	struct sk_buff *skb;
+	int ret;
+
+retry:
+	skb = alloc_skb(size, GFP_KERNEL);
+	if (!skb)

net/sched/cls_api.c static int tcf_fill_node(struct net *net, struct sk_buff *skb, · line 2139

Quoted 10 lines
net/sched/cls_api.c +10 -0
+		return ERR_PTR(-ENOBUFS);
+
+	ret = tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
+			    n->nlmsg_seq, n->nlmsg_flags, event, false,
+			    rtnl_held, extack);
+	if (ret <= 0) {
+		kfree_skb(skb);
+		if (ret == -EMSGSIZE) {
+			size += NLMSG_GOODSIZE;
+			goto retry;

It is a bit concerning to see this technically unbound loop.

Review

Quoted 2 lines
(quoted diff) +2 -0
+		}
+		return ERR_PTR(-EINVAL);

I think you probably want to propagate the error code from tcf_fill_node() here.

Thanks.

#6 lore raw
Quoted 22 lines · Cong Wang
Quoted 21 lines · Toke Høiland-Jørgensen
net/sched/cls_api.c +21 -0
+static struct sk_buff *tfilter_notify_prep(struct net *net,
+					   struct sk_buff *oskb,
+					   struct nlmsghdr *n,
+					   struct tcf_proto *tp,
+					   struct tcf_block *block,
+					   struct Qdisc *q, u32 parent,
+					   void *fh, int event,
+					   u32 portid, bool rtnl_held,
+					   struct netlink_ext_ack *extack)
+{
+	unsigned int size = oskb ? max(NLMSG_GOODSIZE, oskb->len) : NLMSG_GOODSIZE;
+	struct sk_buff *skb;
+	int ret;
+
+retry:
+	skb = alloc_skb(size, GFP_KERNEL);
+	if (!skb)
+		return ERR_PTR(-ENOBUFS);
+
+	ret = tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
+			    n->nlmsg_seq, n->nlmsg_flags, event, false,
Quoted 8 lines

net/sched/cls_api.c static int tcf_fill_node(struct net *net, struct sk_buff *skb, · line 2139

Quoted 6 lines
net/sched/cls_api.c +6 -0
+			    rtnl_held, extack);
+	if (ret <= 0) {
+		kfree_skb(skb);
+		if (ret == -EMSGSIZE) {
+			size += NLMSG_GOODSIZE;
+			goto retry;

It is a bit concerning to see this technically unbound loop.

Well, I did think about that. The loop will terminate eventually by either succeeding, or failing the allocation. Most likely the former, since this is only called after a filter has been successfully installed. I.e., it's not like the amount of data being put into the skb is unbounded.

Quoted 5 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		}
+		return ERR_PTR(-EINVAL);

I think you probably want to propagate the error code from tcf_fill_node() here.

I just kept the existing return value (of tfilter_notify()) for the same error case. tcf_fill_node() always returns -1 on error, so I think it makes more sense to keep this?

Paolo already merged the patch, and I don't think it's worth it to follow up with any fixes, cf the above. WDYT?

-Toke

#7 lore raw
Quoted 33 lines · Toke Høiland-Jørgensen
Quoted 22 lines · Cong Wang
Quoted 21 lines · Toke Høiland-Jørgensen
net/sched/cls_api.c +21 -0
+static struct sk_buff *tfilter_notify_prep(struct net *net,
+					   struct sk_buff *oskb,
+					   struct nlmsghdr *n,
+					   struct tcf_proto *tp,
+					   struct tcf_block *block,
+					   struct Qdisc *q, u32 parent,
+					   void *fh, int event,
+					   u32 portid, bool rtnl_held,
+					   struct netlink_ext_ack *extack)
+{
+	unsigned int size = oskb ? max(NLMSG_GOODSIZE, oskb->len) : NLMSG_GOODSIZE;
+	struct sk_buff *skb;
+	int ret;
+
+retry:
+	skb = alloc_skb(size, GFP_KERNEL);
+	if (!skb)
+		return ERR_PTR(-ENOBUFS);
+
+	ret = tcf_fill_node(net, skb, tp, block, q, parent, fh, portid,
+			    n->nlmsg_seq, n->nlmsg_flags, event, false,
Quoted 8 lines

net/sched/cls_api.c static int tcf_fill_node(struct net *net, struct sk_buff *skb, · line 2139

Quoted 6 lines
net/sched/cls_api.c +6 -0
+			    rtnl_held, extack);
+	if (ret <= 0) {
+		kfree_skb(skb);
+		if (ret == -EMSGSIZE) {
+			size += NLMSG_GOODSIZE;
+			goto retry;

It is a bit concerning to see this technically unbound loop.

Quoted 5 lines

Well, I did think about that. The loop will terminate eventually by either succeeding, or failing the allocation. Most likely the former, since this is only called after a filter has been successfully installed. I.e., it's not like the amount of data being put into the skb is unbounded.

Yeah, I totally agree, it is probably just a theoretical problem.

Quoted 7 lines
Quoted 6 lines

Review

Quoted 3 lines
(quoted diff) +2 -0
+		}
+		return ERR_PTR(-EINVAL);

I think you probably want to propagate the error code from tcf_fill_node() here.

Quoted 6 lines

I just kept the existing return value (of tfilter_notify()) for the same error case. tcf_fill_node() always returns -1 on error, so I think it makes more sense to keep this?

Paolo already merged the patch, and I don't think it's worth it to follow up with any fixes, cf the above. WDYT?

Not a big deal, we can keep as it is.

Thanks.