[PATCH net v3] net-timestamp: make sk_tskey more predictable in error path

[PATCH net v3] net-timestamp: make sk_tskey more predictable in error path

[Vadim Fedorenko]

When SOF_TIMESTAMPING_OPT_ID is used to ambiguate timestamped datagrams,
the sk_tskey can become unpredictable in case of any error happened
during sendmsg(). Move increment later in the code and make decrement of
sk_tskey in error path. This solution is still racy in case of multiple
threads doing snedmsg() over the very same socket in parallel, but still
makes error path much more predictable.

Fixes: 09c2d251b707 ("net-timestamp: add key to disambiguate concurrent datagrams")
Reported-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Vadim Fedorenko <vadfed@meta.com>
---
v1 -> v2: wrong submission
v1 -> v3: 
 - use local boolean variable instead of checking the same conditions
   (suggested by Willem)

 net/ipv4/ip_output.c  | 13 ++++++++-----
 net/ipv6/ip6_output.c | 13 ++++++++-----
 2 files changed, 16 insertions(+), 10 deletions(-)

diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index 41537d18eecf..67d846622365 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -972,8 +972,8 @@ static int __ip_append_data(struct sock *sk,
 	unsigned int maxfraglen, fragheaderlen, maxnonfragsize;
 	int csummode = CHECKSUM_NONE;
 	struct rtable *rt = (struct rtable *)cork->dst;
+	bool paged, hold_tskey, extra_uref = false;
 	unsigned int wmem_alloc_delta = 0;
-	bool paged, extra_uref = false;
 	u32 tskey = 0;
 
 	skb = skb_peek_tail(queue);
@@ -982,10 +982,6 @@ static int __ip_append_data(struct sock *sk,
 	mtu = cork->gso_size ? IP_MAX_MTU : cork->fragsize;
 	paged = !!cork->gso_size;
 
-	if (cork->tx_flags & SKBTX_ANY_TSTAMP &&
-	    READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_OPT_ID)
-		tskey = atomic_inc_return(&sk->sk_tskey) - 1;
-
 	hh_len = LL_RESERVED_SPACE(rt->dst.dev);
 
 	fragheaderlen = sizeof(struct iphdr) + (opt ? opt->optlen : 0);
@@ -1052,6 +1048,11 @@ static int __ip_append_data(struct sock *sk,
 
 	cork->length += length;
 
+	hold_tskey = cork->tx_flags & SKBTX_ANY_TSTAMP &&
+		     READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_OPT_ID;
+	if (hold_tskey)
+		tskey = atomic_inc_return(&sk->sk_tskey) - 1;
+
 	/* So, what's going on in the loop below?
 	 *
 	 * We use calculated fragment length to generate chained skb,
@@ -1274,6 +1275,8 @@ static int __ip_append_data(struct sock *sk,
 	cork->length -= length;
 	IP_INC_STATS(sock_net(sk), IPSTATS_MIB_OUTDISCARDS);
 	refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc);
+	if (hold_tskey)
+		atomic_dec(&sk->sk_tskey);
 	return err;
 }
 
diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index a722a43dd668..31b86fe661aa 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -1424,11 +1424,11 @@ static int __ip6_append_data(struct sock *sk,
 	bool zc = false;
 	u32 tskey = 0;
 	struct rt6_info *rt = (struct rt6_info *)cork->dst;
+	bool paged, hold_tskey, extra_uref = false;
 	struct ipv6_txoptions *opt = v6_cork->opt;
 	int csummode = CHECKSUM_NONE;
 	unsigned int maxnonfragsize, headersize;
 	unsigned int wmem_alloc_delta = 0;
-	bool paged, extra_uref = false;
 
 	skb = skb_peek_tail(queue);
 	if (!skb) {
@@ -1440,10 +1440,6 @@ static int __ip6_append_data(struct sock *sk,
 	mtu = cork->gso_size ? IP6_MAX_MTU : cork->fragsize;
 	orig_mtu = mtu;
 
-	if (cork->tx_flags & SKBTX_ANY_TSTAMP &&
-	    READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_OPT_ID)
-		tskey = atomic_inc_return(&sk->sk_tskey) - 1;
-
 	hh_len = LL_RESERVED_SPACE(rt->dst.dev);
 
 	fragheaderlen = sizeof(struct ipv6hdr) + rt->rt6i_nfheader_len +
@@ -1538,6 +1534,11 @@ static int __ip6_append_data(struct sock *sk,
 			flags &= ~MSG_SPLICE_PAGES;
 	}
 
+	hold_tskey = cork->tx_flags & SKBTX_ANY_TSTAMP &&
+		     READ_ONCE(sk->sk_tsflags) & SOF_TIMESTAMPING_OPT_ID;
+	if (hold_tskey)
+		tskey = atomic_inc_return(&sk->sk_tskey) - 1;
+
 	/*
 	 * Let's try using as much space as possible.
 	 * Use MTU if total length of the message fits into the MTU.
@@ -1794,6 +1795,8 @@ static int __ip6_append_data(struct sock *sk,
 	cork->length -= length;
 	IP6_INC_STATS(sock_net(sk), rt->rt6i_idev, IPSTATS_MIB_OUTDISCARDS);
 	refcount_add(wmem_alloc_delta, &sk->sk_wmem_alloc);
+	if (hold_tskey)
+		atomic_dec(&sk->sk_tskey);
 	return err;
 }
 
-- 
2.39.3

Re: [PATCH net v3] net-timestamp: make sk_tskey more predictable in error path

[Willem de Bruijn]

Vadim Fedorenko wrote:
> When SOF_TIMESTAMPING_OPT_ID is used to ambiguate timestamped datagrams,
> the sk_tskey can become unpredictable in case of any error happened
> during sendmsg(). Move increment later in the code and make decrement of
> sk_tskey in error path. This solution is still racy in case of multiple
> threads doing snedmsg() over the very same socket in parallel, but still
> makes error path much more predictable.
> 
> Fixes: 09c2d251b707 ("net-timestamp: add key to disambiguate concurrent datagrams")
> Reported-by: Andy Lutomirski <luto@amacapital.net>
> Signed-off-by: Vadim Fedorenko <vadfed@meta.com>

Reviewed-by: Willem de Bruijn <willemb@google.com>

Re: [PATCH net v3] net-timestamp: make sk_tskey more predictable in error path

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Paolo Abeni <pabeni@redhat.com>:

On Tue, 13 Feb 2024 03:04:28 -0800 you wrote:
> When SOF_TIMESTAMPING_OPT_ID is used to ambiguate timestamped datagrams,
> the sk_tskey can become unpredictable in case of any error happened
> during sendmsg(). Move increment later in the code and make decrement of
> sk_tskey in error path. This solution is still racy in case of multiple
> threads doing snedmsg() over the very same socket in parallel, but still
> makes error path much more predictable.
> 
> [...]

Here is the summary with links:
  - [net,v3] net-timestamp: make sk_tskey more predictable in error path
    https://git.kernel.org/netdev/net/c/488b6d91b071

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