mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Rishikesh Jethwani <rjethwani@purestorage.com>
To: netdev@vger.kernel.org
Cc: Rishikesh Jethwani <rjethwani@purestorage.com>,
	stable@vger.kernel.org, John Fastabend <john.fastabend@gmail.com>,
	Jakub Kicinski <kuba@kernel.org>,
	Sabrina Dubroca <sd@queasysnail.net>,
	"David S. Miller" <davem@davemloft.net>,
	Eric Dumazet <edumazet@kernel.org>,
	Paolo Abeni <pabeni@redhat.com>, Simon Horman <horms@kernel.org>,
	Tariq Toukan <tariqt@nvidia.com>,
	Alice Mikityanska <alice.kernel@fastmail.im>,
	Boris Pismenny <borisp@mellanox.com>,
	linux-kernel@vger.kernel.org
Subject: [PATCH net v4 1/2] tls: restrict device-down TX fallback to hardware TX sockets
Date: Wed, 11 Nov 2026 16:29:28 -0700	[thread overview]
Message-ID: <20261111232930.2532497-1-rjethwani@purestorage.com> (raw)

tls_device_down() installs the hardware TX fallback on RX-only and
software-TX sockets. The fallback expects a hardware context in
priv_ctx_tx, so an RX-only socket crashes on its next transmit:

  BUG: kernel NULL pointer dereference, address: 0000000000000008
  tls_sw_fallback+0x11b/0x450
  validate_xmit_skb.isra.0+0x14f/0x420
  __dev_queue_xmit+0x5e8/0xf30

Switch only tls_validate_xmit_skb() to tls_validate_xmit_skb_sw(), and use
the cmpxchg result to decide whether to delete driver TX state. Publish
the hardware validator before releasing device_offload_lock.

Driver deletion cannot rely on tx_conf or rx_conf: they are published
after setup releases the lock, so device-down can miss a completed
attachment and leak driver state. Track RX state with TLS_RX_DEV_CLOSED:
set it at context creation and after deletion, and clear it after a
successful driver add under device_offload_lock. This avoids deleting
absent RX state and reading configuration bitfields while their shared
byte is updated.

Found by code inspection. The reproduced RX-only crash is fixed: the
socket survives interface down/up and continues receiving TLS records.
Tested on Mellanox ConnectX-6 Dx (Crypto Enabled).

Fixes: c55dcdd435aa ("net/tls: Fix use-after-free after the TLS device goes down and up")
Fixes: 4799ac81e52a ("tls: Add rx inline crypto offload")
Cc: stable@vger.kernel.org
Signed-off-by: Rishikesh Jethwani <rjethwani@purestorage.com>
---
 include/net/tls.h    |  8 ++++----
 net/tls/tls_device.c | 19 ++++++++++++-------
 net/tls/tls_main.c   |  1 +
 3 files changed, 17 insertions(+), 11 deletions(-)

diff --git a/include/net/tls.h b/include/net/tls.h
index e57bef58851e..915f39bddc83 100644
--- a/include/net/tls.h
+++ b/include/net/tls.h
@@ -187,10 +187,10 @@ enum tls_context_flags {
 	 * to be atomic.
 	 */
 	TLS_TX_SYNC_SCHED = 1,
-	/* tls_dev_del was called for the RX side, device state was released,
-	 * but tls_ctx->netdev might still be kept, because TX-side driver
-	 * resources might not be released yet. Used to prevent the second
-	 * tls_dev_del call in tls_device_down if it happens simultaneously.
+	/* No RX device state is installed. Set at context creation and after
+	 * RX tls_dev_del, cleared after a successful RX tls_dev_add. Prevents
+	 * deleting absent or already released RX state while ctx->netdev may
+	 * still be kept for TX.
 	 */
 	TLS_RX_DEV_CLOSED = 2,
 };
diff --git a/net/tls/tls_device.c b/net/tls/tls_device.c
index 4cdfbb57ecff..a42cbc89db2b 100644
--- a/net/tls/tls_device.c
+++ b/net/tls/tls_device.c
@@ -1195,13 +1195,13 @@ int tls_set_device_offload(struct sock *sk)
 		goto release_lock;
 
 	tls_device_attach(ctx, sk, netdev);
-	up_read(&device_offload_lock);
 
 	/* following this assignment tls_is_skb_tx_device_offloaded
 	 * will return true and the context might be accessed
 	 * by the netdev's xmit function.
 	 */
 	smp_store_release(&sk->sk_validate_xmit_skb, tls_validate_xmit_skb);
+	up_read(&device_offload_lock);
 	dev_put(netdev);
 
 	return 0;
@@ -1282,6 +1282,7 @@ int tls_set_device_offload_rx(struct sock *sk, struct tls_context *ctx)
 		goto free_sw_resources;
 
 	tls_device_attach(ctx, sk, netdev);
+	clear_bit(TLS_RX_DEV_CLOSED, &ctx->flags);
 	up_read(&device_offload_lock);
 
 	dev_put(netdev);
@@ -1314,12 +1315,11 @@ void tls_device_offload_cleanup_rx(struct sock *sk)
 
 	netdev->tlsdev_ops->tls_dev_del(netdev, tls_ctx,
 					TLS_OFFLOAD_CTX_DIR_RX);
+	set_bit(TLS_RX_DEV_CLOSED, &tls_ctx->flags);
 
 	if (tls_ctx->tx_conf != TLS_HW) {
 		dev_put(netdev);
 		rcu_assign_pointer(tls_ctx->netdev, NULL);
-	} else {
-		set_bit(TLS_RX_DEV_CLOSED, &tls_ctx->flags);
 	}
 out:
 	up_read(&device_offload_lock);
@@ -1350,10 +1350,14 @@ static int tls_device_down(struct net_device *netdev)
 	spin_unlock_irqrestore(&tls_device_lock, flags);
 
 	list_for_each_entry_safe(ctx, tmp, &list, list)	{
+		bool tx_hw;
+
 		/* Stop offloaded TX and switch to the fallback.
 		 * tls_is_skb_tx_device_offloaded will return false.
 		 */
-		WRITE_ONCE(ctx->sk->sk_validate_xmit_skb, tls_validate_xmit_skb_sw);
+		tx_hw = cmpxchg(&ctx->sk->sk_validate_xmit_skb,
+				tls_validate_xmit_skb,
+				tls_validate_xmit_skb_sw) == tls_validate_xmit_skb;
 
 		/* Stop the RX and TX resync.
 		 * tls_dev_resync must not be called after tls_dev_del.
@@ -1370,13 +1374,14 @@ static int tls_device_down(struct net_device *netdev)
 		synchronize_net();
 
 		/* Release the offload context on the driver side. */
-		if (ctx->tx_conf == TLS_HW)
+		if (tx_hw)
 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
 							TLS_OFFLOAD_CTX_DIR_TX);
-		if (ctx->rx_conf == TLS_HW &&
-		    !test_bit(TLS_RX_DEV_CLOSED, &ctx->flags))
+		if (!test_bit(TLS_RX_DEV_CLOSED, &ctx->flags)) {
 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
 							TLS_OFFLOAD_CTX_DIR_RX);
+			set_bit(TLS_RX_DEV_CLOSED, &ctx->flags);
+		}
 
 		dev_put(netdev);
 
diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c
index b8d5a843f761..9b06d0f035c5 100644
--- a/net/tls/tls_main.c
+++ b/net/tls/tls_main.c
@@ -935,6 +935,7 @@ struct tls_context *tls_ctx_create(struct sock *sk)
 		return NULL;
 
 	mutex_init(&ctx->tx_lock);
+	set_bit(TLS_RX_DEV_CLOSED, &ctx->flags);
 	ctx->sk_proto = READ_ONCE(sk->sk_prot);
 	ctx->sk = sk;
 	/* Release semantic of rcu_assign_pointer() ensures that
-- 
2.50.1


             reply	other threads:[~2026-11-11 23:29 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-11 23:29 Rishikesh Jethwani [this message]
2026-11-11 23:29 ` [PATCH net v4 2/2] tls: fix socket use-after-free during device-down Rishikesh Jethwani
2026-11-16  0:02 ` [PATCH net v4 1/2] tls: restrict device-down TX fallback to hardware TX sockets netdev-bot+sashiko

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=20261111232930.2532497-1-rjethwani@purestorage.com \
    --to=rjethwani@purestorage.com \
    --cc=alice.kernel@fastmail.im \
    --cc=borisp@mellanox.com \
    --cc=davem@davemloft.net \
    --cc=edumazet@kernel.org \
    --cc=horms@kernel.org \
    --cc=john.fastabend@gmail.com \
    --cc=kuba@kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=sd@queasysnail.net \
    --cc=stable@vger.kernel.org \
    --cc=tariqt@nvidia.com \
    /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

all inboxes | Powered by JetHome®