Merge tag 'net-5.13-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Networking fixes, including fixes from bpf, wireless, netfilter and
  wireguard trees.

  The bpf vs lockdown+audit fix is the most notable.

  Things haven't slowed down just yet, both in terms of regressions in
  current release and largish fixes for older code, but we usually see a
  slowdown only after -rc5.

  Current release - regressions:

   - virtio-net: fix page faults and crashes when XDP is enabled

   - mlx5e: fix HW timestamping with CQE compression, and make sure they
     are only allowed to coexist with capable devices

   - stmmac:
      - fix kernel panic due to NULL pointer dereference of
        mdio_bus_data
      - fix double clk unprepare when no PHY device is connected

  Current release - new code bugs:

   - mt76: a few fixes for the recent MT7921 devices and runtime power
     management

  Previous releases - regressions:

   - ice:
      - track AF_XDP ZC enabled queues in bitmap to fix copy mode Tx
      - fix allowing VF to request more/less queues via virtchnl
      - correct supported and advertised autoneg by using PHY
        capabilities
      - allow all LLDP packets from PF to Tx

   - kbuild: quote OBJCOPY var to avoid a pahole call break the build

  Previous releases - always broken:

   - bpf, lockdown, audit: fix buggy SELinux lockdown permission checks

   - mt76: address the recent FragAttack vulnerabilities not covered by
     generic fixes

   - ipv6: fix KASAN: slab-out-of-bounds Read in
     fib6_nh_flush_exceptions

   - Bluetooth:
      - fix the erroneous flush_work() order, to avoid double free
      - use correct lock to prevent UAF of hdev object

   - nfc: fix NULL ptr dereference in llcp_sock_getname() after failed
     connect

   - ieee802154: multiple fixes to error checking and return values

   - igb: fix XDP with PTP enabled

   - intel: add correct exception tracing for XDP

   - tls: fix use-after-free when TLS offload device goes down and back
     up

   - ipvs: ignore IP_VS_SVC_F_HASHED flag when adding service

   - netfilter: nft_ct: skip expectations for confirmed conntrack

   - mptcp: fix falling back to TCP in presence of out of order packets
     early in connection lifetime

   - wireguard: switch from O(n) to a O(1) algorithm for maintaining
     peers, fixing stalls and a large memory leak in the process

  Misc:

   - devlink: correct VIRTUAL port to not have phys_port attributes

   - Bluetooth: fix VIRTIO_ID_BT assigned number

   - net: return the correct errno code ENOBUF -> ENOMEM

   - wireguard:
      - peer: allocate in kmem_cache saving 25% on peer memory
      - do not use -O3"

* tag 'net-5.13-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (91 commits)
  cxgb4: avoid link re-train during TC-MQPRIO configuration
  sch_htb: fix refcount leak in htb_parent_to_leaf_offload
  wireguard: allowedips: free empty intermediate nodes when removing single node
  wireguard: allowedips: allocate nodes in kmem_cache
  wireguard: allowedips: remove nodes in O(1)
  wireguard: allowedips: initialize list head in selftest
  wireguard: peer: allocate in kmem_cache
  wireguard: use synchronize_net rather than synchronize_rcu
  wireguard: do not use -O3
  wireguard: selftests: make sure rp_filter is disabled on vethc
  wireguard: selftests: remove old conntrack kconfig value
  virtchnl: Add missing padding to virtchnl_proto_hdrs
  ice: Allow all LLDP packets from PF to Tx
  ice: report supported and advertised autoneg using PHY capabilities
  ice: handle the VF VSI rebuild failure
  ice: Fix VFR issues for AVF drivers that expect ATQLEN cleared
  ice: Fix allowing VF to request more/less queues via virtchnl
  virtio-net: fix for skb_over_panic inside big mode
  ipv6: Fix KASAN: slab-out-of-bounds Read in fib6_nh_flush_exceptions
  fib: Return the correct errno code
  ...

1 files changed, 49 insertions, 11 deletions

diff --git a/net/tls/tls_device.c b/net/tls/tls_device.c
index 76a6f8c2eec4..bd9f1567aa39 100644
--- a/net/tls/tls_device.c
+++ b/net/tls/tls_device.c
@@ -50,6 +50,7 @@ static void tls_device_gc_task(struct work_struct *work);
 static DECLARE_WORK(tls_device_gc_work, tls_device_gc_task);
 static LIST_HEAD(tls_device_gc_list);
 static LIST_HEAD(tls_device_list);
+static LIST_HEAD(tls_device_down_list);
 static DEFINE_SPINLOCK(tls_device_lock);
 
 static void tls_device_free_ctx(struct tls_context *ctx)
@@ -680,15 +681,13 @@ static void tls_device_resync_rx(struct tls_context *tls_ctx,
 	struct tls_offload_context_rx *rx_ctx = tls_offload_ctx_rx(tls_ctx);
 	struct net_device *netdev;
 
-	if (WARN_ON(test_and_set_bit(TLS_RX_SYNC_RUNNING, &tls_ctx->flags)))
-		return;
-
 	trace_tls_device_rx_resync_send(sk, seq, rcd_sn, rx_ctx->resync_type);
+	rcu_read_lock();
 	netdev = READ_ONCE(tls_ctx->netdev);
 	if (netdev)
 		netdev->tlsdev_ops->tls_dev_resync(netdev, sk, seq, rcd_sn,
 						   TLS_OFFLOAD_CTX_DIR_RX);
-	clear_bit_unlock(TLS_RX_SYNC_RUNNING, &tls_ctx->flags);
+	rcu_read_unlock();
 	TLS_INC_STATS(sock_net(sk), LINUX_MIB_TLSRXDEVICERESYNC);
 }
 
@@ -761,6 +760,8 @@ void tls_device_rx_resync_new_rec(struct sock *sk, u32 rcd_len, u32 seq)
 
 	if (tls_ctx->rx_conf != TLS_HW)
 		return;
+	if (unlikely(test_bit(TLS_RX_DEV_DEGRADED, &tls_ctx->flags)))
+		return;
 
 	prot = &tls_ctx->prot_info;
 	rx_ctx = tls_offload_ctx_rx(tls_ctx);
@@ -963,6 +964,17 @@ int tls_device_decrypted(struct sock *sk, struct tls_context *tls_ctx,
 
 	ctx->sw.decrypted |= is_decrypted;
 
+	if (unlikely(test_bit(TLS_RX_DEV_DEGRADED, &tls_ctx->flags))) {
+		if (likely(is_encrypted || is_decrypted))
+			return 0;
+
+		/* After tls_device_down disables the offload, the next SKB will
+		 * likely have initial fragments decrypted, and final ones not
+		 * decrypted. We need to reencrypt that single SKB.
+		 */
+		return tls_device_reencrypt(sk, skb);
+	}
+
 	/* Return immediately if the record is either entirely plaintext or
 	 * entirely ciphertext. Otherwise handle reencrypt partially decrypted
 	 * record.
@@ -1292,6 +1304,26 @@ 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)	{
+		/* Stop offloaded TX and switch to the fallback.
+		 * tls_is_sk_tx_device_offloaded will return false.
+		 */
+		WRITE_ONCE(ctx->sk->sk_validate_xmit_skb, tls_validate_xmit_skb_sw);
+
+		/* Stop the RX and TX resync.
+		 * tls_dev_resync must not be called after tls_dev_del.
+		 */
+		WRITE_ONCE(ctx->netdev, NULL);
+
+		/* Start skipping the RX resync logic completely. */
+		set_bit(TLS_RX_DEV_DEGRADED, &ctx->flags);
+
+		/* Sync with inflight packets. After this point:
+		 * TX: no non-encrypted packets will be passed to the driver.
+		 * RX: resync requests from the driver will be ignored.
+		 */
+		synchronize_net();
+
+		/* Release the offload context on the driver side. */
 		if (ctx->tx_conf == TLS_HW)
 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
 							TLS_OFFLOAD_CTX_DIR_TX);
@@ -1299,15 +1331,21 @@ static int tls_device_down(struct net_device *netdev)
 		    !test_bit(TLS_RX_DEV_CLOSED, &ctx->flags))
 			netdev->tlsdev_ops->tls_dev_del(netdev, ctx,
 							TLS_OFFLOAD_CTX_DIR_RX);
-		WRITE_ONCE(ctx->netdev, NULL);
-		smp_mb__before_atomic(); /* pairs with test_and_set_bit() */
-		while (test_bit(TLS_RX_SYNC_RUNNING, &ctx->flags))
-			usleep_range(10, 200);
+
 		dev_put(netdev);
-		list_del_init(&ctx->list);
 
-		if (refcount_dec_and_test(&ctx->refcount))
-			tls_device_free_ctx(ctx);
+		/* Move the context to a separate list for two reasons:
+		 * 1. When the context is deallocated, list_del is called.
+		 * 2. It's no longer an offloaded context, so we don't want to
+		 *    run offload-specific code on this context.
+		 */
+		spin_lock_irqsave(&tls_device_lock, flags);
+		list_move_tail(&ctx->list, &tls_device_down_list);
+		spin_unlock_irqrestore(&tls_device_lock, flags);
+
+		/* Device contexts for RX and TX will be freed in on sk_destruct
+		 * by tls_device_free_ctx. rx_conf and tx_conf stay in TLS_HW.
+		 */
 	}
 
 	up_write(&device_offload_lock);