tc: Ensure we have enough buffer space when sending filter netlink notifications
Subsystems (3)
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
@@ -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
Quoted 30 lines · Mon, Apr 07, 2025 at 12:55:34PM +0200, [email protected] wrote:
The
tfilter_notify()andtfilter_del_notify()functions assume thatNLMSG_GOODSIZEis always enough to dump the filter chain. This is not always the case, which can lead to silent notify failures (because the return code oftfilter_notify()is not always checked). In particular, this can lead toNLM_F_ECHOnot 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 maskffff0000
[repeated 32 times]
Quoted 6 lines
[0] https://github.com/openvswitch/ovs/commit/106ef21860c935e5e0017a88bf42b94025c4e511
Fixes 1da177e4c3f4 ("Linux-2.6.12-rc2")Reported-by Frode Nordahl <[email protected]>Signed-off-by Toke Høiland-Jørgensen <[email protected]>
Paolo Abeni
maintainerQuoted 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 1key #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
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 1key #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
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()andtfilter_del_notify()functions assume thatNLMSG_GOODSIZEis always enough to dump the filter chain. This is not always the case, which can lead to silent notify failures (because the return code oftfilter_notify()is not always checked). In particular, this can lead toNLM_F_ECHOnot 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
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.
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
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.