Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net

Pull networking fixes and more updates from David Miller:

 1) Tunneling fixes from Tom Herbert and Alexander Duyck.

 2) AF_UNIX updates some struct sock bit fields with the socket lock,
    whereas setsockopt() sets overlapping ones with locking.  Seperate
    out the synchronized vs.  the AF_UNIX unsynchronized ones to avoid
    corruption.  From Andrey Ryabinin.

 3) Mount BPF filesystem with mount_nodev rather than mount_ns, from
    Eric Biederman.

 4) A couple kmemdup conversions, from Muhammad Falak R Wani.

 5) BPF verifier fixes from Alexei Starovoitov.

 6) Don't let tunneled UDP packets get stuck in socket queues, if
    something goes wrong during the encapsulation just drop the packet
    rather than signalling an error up the call stack.  From Hannes
    Frederic Sowa.

 7) SKB ref after free in batman-adv, from Florian Westphal.

 8) TCP iSCSI, ocfs2, rds, and tipc have to disable BH in it's TCP
    callbacks since the TCP stack runs pre-emptibly now.  From Eric
    Dumazet.

 9) Fix crash in fixed_phy_add, from Rabin Vincent.

10) Fix length checks in xen-netback, from Paul Durrant.

11) Fix mixup in KEY vs KEYID macsec attributes, from Sabrina Dubroca.

12) RDS connection spamming bug fixes from Sowmini Varadhan

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (152 commits)
  net: suppress warnings on dev_alloc_skb
  uapi glibc compat: fix compilation when !__USE_MISC in glibc
  udp: prevent skbs lingering in tunnel socket queues
  bpf: teach verifier to recognize imm += ptr pattern
  bpf: support decreasing order in direct packet access
  net: usb: ch9200: use kmemdup
  ps3_gelic: use kmemdup
  net:liquidio: use kmemdup
  bpf: Use mount_nodev not mount_ns to mount the bpf filesystem
  net: cdc_ncm: update datagram size after changing mtu
  tuntap: correctly wake up process during uninit
  intel: Add support for IPv6 IP-in-IP offload
  ip6_gre: Do not allow segmentation offloads GRE_CSUM is enabled with FOU/GUE
  RDS: TCP: Avoid rds connection churn from rogue SYNs
  RDS: TCP: rds_tcp_accept_worker() must exit gracefully when terminating rds-tcp
  net: sock: move ->sk_shutdown out of bitfields.
  ipv6: Don't reset inner headers in ip6_tnl_xmit
  ip4ip6: Support for GSO/GRO
  ip6ip6: Support for GSO/GRO
  ipv6: Set features for IPv6 tunnels
  ...

1 files changed, 23 insertions, 3 deletions

diff --git a/drivers/net/wireless/intel/iwlwifi/mvm/rx.c b/drivers/net/wireless/intel/iwlwifi/mvm/rx.c
index 263e8a8576b7..ab7f7eda9c13 100644
--- a/drivers/net/wireless/intel/iwlwifi/mvm/rx.c
+++ b/drivers/net/wireless/intel/iwlwifi/mvm/rx.c
@@ -97,6 +97,7 @@ void iwl_mvm_rx_rx_phy_cmd(struct iwl_mvm *mvm, struct iwl_rx_cmd_buffer *rxb)
  * Adds the rxb to a new skb and give it to mac80211
  */
 static void iwl_mvm_pass_packet_to_mac80211(struct iwl_mvm *mvm,
+					    struct ieee80211_sta *sta,
 					    struct napi_struct *napi,
 					    struct sk_buff *skb,
 					    struct ieee80211_hdr *hdr, u16 len,
@@ -131,7 +132,7 @@ static void iwl_mvm_pass_packet_to_mac80211(struct iwl_mvm *mvm,
 				fraglen, rxb->truesize);
 	}
 
-	ieee80211_rx_napi(mvm->hw, NULL, skb, napi);
+	ieee80211_rx_napi(mvm->hw, sta, skb, napi);
 }
 
 /*
@@ -271,6 +272,7 @@ void iwl_mvm_rx_rx_mpdu(struct iwl_mvm *mvm, struct napi_struct *napi,
 	u32 rate_n_flags;
 	u32 rx_pkt_status;
 	u8 crypt_len = 0;
+	bool take_ref;
 
 	phy_info = &mvm->last_phy_info;
 	rx_res = (struct iwl_rx_mpdu_res_start *)pkt->data;
@@ -453,8 +455,26 @@ void iwl_mvm_rx_rx_mpdu(struct iwl_mvm *mvm, struct napi_struct *napi,
 		     mvm->sched_scan_pass_all == SCHED_SCAN_PASS_ALL_ENABLED))
 		mvm->sched_scan_pass_all = SCHED_SCAN_PASS_ALL_FOUND;
 
-	iwl_mvm_pass_packet_to_mac80211(mvm, napi, skb, hdr, len, ampdu_status,
-					crypt_len, rxb);
+	if (unlikely(ieee80211_is_beacon(hdr->frame_control) ||
+		     ieee80211_is_probe_resp(hdr->frame_control)))
+		rx_status->boottime_ns = ktime_get_boot_ns();
+
+	/* Take a reference briefly to kick off a d0i3 entry delay so
+	 * we can handle bursts of RX packets without toggling the
+	 * state too often.  But don't do this for beacons if we are
+	 * going to idle because the beacon filtering changes we make
+	 * cause the firmware to send us collateral beacons. */
+	take_ref = !(test_bit(STATUS_TRANS_GOING_IDLE, &mvm->trans->status) &&
+		     ieee80211_is_beacon(hdr->frame_control));
+
+	if (take_ref)
+		iwl_mvm_ref(mvm, IWL_MVM_REF_RX);
+
+	iwl_mvm_pass_packet_to_mac80211(mvm, sta, napi, skb, hdr, len,
+					ampdu_status, crypt_len, rxb);
+
+	if (take_ref)
+		iwl_mvm_unref(mvm, IWL_MVM_REF_RX);
 }
 
 static void iwl_mvm_update_rx_statistics(struct iwl_mvm *mvm,