[PATCH net 0/3] netfilter updates for net

[PATCH net 0/3] netfilter updates for net

[Florian Westphal]

Three late fixes for netfilter:

1) If nf_queue user requests packet truncation below size of l3 header,
   we corrupt the skb, then crash.  Reject such requests.

2) add cond_resched() calls when doing cycle detection in the
   nf_tables graph.  This avoids softlockup warning with certain
   rulesets.

3) Reject rulesets that use nftables 'queue' expression in family/chain
   combinations other than those that are supported.  Currently the ruleset
   will load, but when userspace attempts to reinject you get WARN splat +
   packet drops.

The following changes since commit 9b134b1694ec8926926ba6b7b80884ea829245a0:

  bridge: Do not send empty IFLA_AF_SPEC attribute (2022-07-26 15:35:53 +0200)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf.git

for you to fetch changes up to 47f4f510ad586032b85c89a0773fbb011d412425:

  netfilter: nft_queue: only allow supported familes and hooks (2022-07-26 21:12:42 +0200)

----------------------------------------------------------------
Florian Westphal (3):
  netfilter: nf_queue: do not allow packet truncation below transport header offset
  netfilter: nf_tables: add rescheduling points during loop detection walks
  netfilter: nft_queue: only allow supported familes and hooks

 net/netfilter/nf_tables_api.c   |  6 ++++++
 net/netfilter/nfnetlink_queue.c |  7 ++++++-
 net/netfilter/nft_queue.c       | 27 +++++++++++++++++++++++++++
 3 files changed, 39 insertions(+), 1 deletion(-)
-- 
2.35.1

[PATCH net 1/3] netfilter: nf_queue: do not allow packet truncation below transport header offset

[Florian Westphal]

Domingo Dirutigliano and Nicola Guerrera report kernel panic when
sending nf_queue verdict with 1-byte nfta_payload attribute.

The IP/IPv6 stack pulls the IP(v6) header from the packet after the
input hook.

If user truncates the packet below the header size, this skb_pull() will
result in a malformed skb (skb->len < 0).

Fixes: 7af4cc3fa158 ("[NETFILTER]: Add "nfnetlink_queue" netfilter queue handler over nfnetlink")
Reported-by: Domingo Dirutigliano <pwnzer0tt1@proton.me>
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 net/netfilter/nfnetlink_queue.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/net/netfilter/nfnetlink_queue.c b/net/netfilter/nfnetlink_queue.c
index a364f8e5e698..87a9009d5234 100644
--- a/net/netfilter/nfnetlink_queue.c
+++ b/net/netfilter/nfnetlink_queue.c
@@ -843,11 +843,16 @@ nfqnl_enqueue_packet(struct nf_queue_entry *entry, unsigned int queuenum)
 }
 
 static int
-nfqnl_mangle(void *data, int data_len, struct nf_queue_entry *e, int diff)
+nfqnl_mangle(void *data, unsigned int data_len, struct nf_queue_entry *e, int diff)
 {
 	struct sk_buff *nskb;
 
 	if (diff < 0) {
+		unsigned int min_len = skb_transport_offset(e->skb);
+
+		if (data_len < min_len)
+			return -EINVAL;
+
 		if (pskb_trim(e->skb, data_len))
 			return -ENOMEM;
 	} else if (diff > 0) {
-- 
2.35.1

[PATCH net 2/3] netfilter: nf_tables: add rescheduling points during loop detection walks

[Florian Westphal]

Add explicit rescheduling points during ruleset walk.

Switching to a faster algorithm is possible but this is a much
smaller change, suitable for nf tree.

Link: https://bugzilla.netfilter.org/show_bug.cgi?id=1460
Signed-off-by: Florian Westphal <fw@strlen.de>
Acked-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 net/netfilter/nf_tables_api.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/net/netfilter/nf_tables_api.c b/net/netfilter/nf_tables_api.c
index 646d5fd53604..9f976b11d896 100644
--- a/net/netfilter/nf_tables_api.c
+++ b/net/netfilter/nf_tables_api.c
@@ -3340,6 +3340,8 @@ int nft_chain_validate(const struct nft_ctx *ctx, const struct nft_chain *chain)
 			if (err < 0)
 				return err;
 		}
+
+		cond_resched();
 	}
 
 	return 0;
@@ -9367,9 +9369,13 @@ static int nf_tables_check_loops(const struct nft_ctx *ctx,
 				break;
 			}
 		}
+
+		cond_resched();
 	}
 
 	list_for_each_entry(set, &ctx->table->sets, list) {
+		cond_resched();
+
 		if (!nft_is_active_next(ctx->net, set))
 			continue;
 		if (!(set->flags & NFT_SET_MAP) ||
-- 
2.35.1

[PATCH net 3/3] netfilter: nft_queue: only allow supported familes and hooks

[Florian Westphal]

Trying to use 'queue' statement in ingress (for example)
triggers a splat on reinject:

WARNING: CPU: 3 PID: 1345 at net/netfilter/nf_queue.c:291

... because nf_reinject cannot find the ruleset head.

The netdev family doesn't support async resume at the moment anyway,
so disallow loading such rulesets with a more appropriate
error message.

v2: add 'validate' callback and also check hook points, v1 did
allow ingress use in 'table inet', but that doesn't work either. (Pablo)

Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
 net/netfilter/nft_queue.c | 27 +++++++++++++++++++++++++++
 1 file changed, 27 insertions(+)

diff --git a/net/netfilter/nft_queue.c b/net/netfilter/nft_queue.c
index 15e4b7640dc0..da29e92c03e2 100644
--- a/net/netfilter/nft_queue.c
+++ b/net/netfilter/nft_queue.c
@@ -68,6 +68,31 @@ static void nft_queue_sreg_eval(const struct nft_expr *expr,
 	regs->verdict.code = ret;
 }
 
+static int nft_queue_validate(const struct nft_ctx *ctx,
+			      const struct nft_expr *expr,
+			      const struct nft_data **data)
+{
+	static const unsigned int supported_hooks = ((1 << NF_INET_PRE_ROUTING) |
+						     (1 << NF_INET_LOCAL_IN) |
+						     (1 << NF_INET_FORWARD) |
+						     (1 << NF_INET_LOCAL_OUT) |
+						     (1 << NF_INET_POST_ROUTING));
+
+	switch (ctx->family) {
+	case NFPROTO_IPV4:
+	case NFPROTO_IPV6:
+	case NFPROTO_INET:
+	case NFPROTO_BRIDGE:
+		break;
+	case NFPROTO_NETDEV: /* lacks okfn */
+		fallthrough;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	return nft_chain_validate_hooks(ctx->chain, supported_hooks);
+}
+
 static const struct nla_policy nft_queue_policy[NFTA_QUEUE_MAX + 1] = {
 	[NFTA_QUEUE_NUM]	= { .type = NLA_U16 },
 	[NFTA_QUEUE_TOTAL]	= { .type = NLA_U16 },
@@ -164,6 +189,7 @@ static const struct nft_expr_ops nft_queue_ops = {
 	.eval		= nft_queue_eval,
 	.init		= nft_queue_init,
 	.dump		= nft_queue_dump,
+	.validate	= nft_queue_validate,
 	.reduce		= NFT_REDUCE_READONLY,
 };
 
@@ -173,6 +199,7 @@ static const struct nft_expr_ops nft_queue_sreg_ops = {
 	.eval		= nft_queue_sreg_eval,
 	.init		= nft_queue_sreg_init,
 	.dump		= nft_queue_sreg_dump,
+	.validate	= nft_queue_validate,
 	.reduce		= NFT_REDUCE_READONLY,
 };
 
-- 
2.35.1

Re: [PATCH net 1/3] netfilter: nf_queue: do not allow packet truncation below transport header offset

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (master)
by Florian Westphal <fw@strlen.de>:

On Tue, 26 Jul 2022 21:20:54 +0200 you wrote:
> Domingo Dirutigliano and Nicola Guerrera report kernel panic when
> sending nf_queue verdict with 1-byte nfta_payload attribute.
> 
> The IP/IPv6 stack pulls the IP(v6) header from the packet after the
> input hook.
> 
> If user truncates the packet below the header size, this skb_pull() will
> result in a malformed skb (skb->len < 0).
> 
> [...]

Here is the summary with links:
  - [net,1/3] netfilter: nf_queue: do not allow packet truncation below transport header offset
    https://git.kernel.org/netdev/net/c/99a63d36cb3e
  - [net,2/3] netfilter: nf_tables: add rescheduling points during loop detection walks
    https://git.kernel.org/netdev/net/c/81ea01066741
  - [net,3/3] netfilter: nft_queue: only allow supported familes and hooks
    https://git.kernel.org/netdev/net/c/47f4f510ad58

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html