ath11k: support ARP and NS offload
authorCarl Huang <quic_cjhuang@quicinc.com>
Mon, 14 Mar 2022 05:18:16 +0000 (07:18 +0200)
committerKalle Valo <quic_kvalo@quicinc.com>
Fri, 18 Mar 2022 15:36:02 +0000 (17:36 +0200)
Support ARP and NS offload in WoW state.

Tested this way: put machine A with QCA6390 to WoW state,
ping/ping6 machine A from another machine B, check sniffer to see
any ARP response and Neighbour advertisement from machine A.

Tested-on: QCA6390 hw2.0 PCI WLAN.HST.1.0.1-01740-QCAHSTSWPLZ_V2_TO_X86-1

Signed-off-by: Carl Huang <quic_cjhuang@quicinc.com>
Signed-off-by: Baochen Qiang <quic_bqiang@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/1644308006-22784-6-git-send-email-quic_cjhuang@quicinc.com
drivers/net/wireless/ath/ath11k/core.h
drivers/net/wireless/ath/ath11k/mac.c
drivers/net/wireless/ath/ath11k/wmi.c
drivers/net/wireless/ath/ath11k/wmi.h
drivers/net/wireless/ath/ath11k/wow.c

index 1f5e977..4ba12de 100644 (file)
@@ -213,6 +213,23 @@ enum ath11k_monitor_flags {
        ATH11K_FLAG_MONITOR_VDEV_CREATED,
 };
 
+#define ATH11K_IPV6_UC_TYPE     0
+#define ATH11K_IPV6_AC_TYPE     1
+
+#define ATH11K_IPV6_MAX_COUNT   16
+#define ATH11K_IPV4_MAX_COUNT   2
+
+struct ath11k_arp_ns_offload {
+       u8  ipv4_addr[ATH11K_IPV4_MAX_COUNT][4];
+       u32 ipv4_count;
+       u32 ipv6_count;
+       u8  ipv6_addr[ATH11K_IPV6_MAX_COUNT][16];
+       u8  self_ipv6_addr[ATH11K_IPV6_MAX_COUNT][16];
+       u8  ipv6_type[ATH11K_IPV6_MAX_COUNT];
+       bool ipv6_valid[ATH11K_IPV6_MAX_COUNT];
+       u8  mac_addr[ETH_ALEN];
+};
+
 struct ath11k_vif {
        u32 vdev_id;
        enum wmi_vdev_type vdev_type;
@@ -264,6 +281,8 @@ struct ath11k_vif {
        bool bcca_zero_sent;
        bool do_not_send_tmpl;
        struct ieee80211_chanctx_conf chanctx;
+       struct ath11k_arp_ns_offload arp_ns_offload;
+
 #ifdef CONFIG_ATH11K_DEBUGFS
        struct dentry *debugfs_twt;
 #endif /* CONFIG_ATH11K_DEBUGFS */
index 7a6c758..dee1cb4 100644 (file)
@@ -6,6 +6,11 @@
 
 #include <net/mac80211.h>
 #include <linux/etherdevice.h>
+#include <linux/bitfield.h>
+#include <linux/inetdevice.h>
+#include <net/if_inet6.h>
+#include <net/ipv6.h>
+
 #include "mac.h"
 #include "core.h"
 #include "debug.h"
@@ -3095,6 +3100,7 @@ static void ath11k_mac_op_bss_info_changed(struct ieee80211_hw *hw,
        int ret = 0;
        u8 rateidx;
        u32 rate;
+       u32 ipv4_cnt;
 
        mutex_lock(&ar->conf_mutex);
 
@@ -3387,6 +3393,18 @@ static void ath11k_mac_op_bss_info_changed(struct ieee80211_hw *hw,
            changed & BSS_CHANGED_UNSOL_BCAST_PROBE_RESP)
                ath11k_mac_fils_discovery(arvif, info);
 
+       if (changed & BSS_CHANGED_ARP_FILTER) {
+               ipv4_cnt = min(info->arp_addr_cnt, ATH11K_IPV4_MAX_COUNT);
+               memcpy(arvif->arp_ns_offload.ipv4_addr, info->arp_addr_list,
+                      ipv4_cnt * sizeof(u32));
+               memcpy(arvif->arp_ns_offload.mac_addr, vif->addr, ETH_ALEN);
+               arvif->arp_ns_offload.ipv4_count = ipv4_cnt;
+
+               ath11k_dbg(ar->ab, ATH11K_DBG_MAC, "mac arp_addr_cnt %d vif->addr %pM, offload_addr %pI4\n",
+                          info->arp_addr_cnt,
+                          vif->addr, arvif->arp_ns_offload.ipv4_addr);
+       }
+
        mutex_unlock(&ar->conf_mutex);
 }
 
@@ -8087,6 +8105,101 @@ static void ath11k_mac_op_sta_statistics(struct ieee80211_hw *hw,
        }
 }
 
+static void ath11k_generate_ns_mc_addr(struct ath11k *ar,
+                                      struct ath11k_arp_ns_offload *offload)
+{
+       int i;
+
+       for (i = 0; i < offload->ipv6_count; i++) {
+               offload->self_ipv6_addr[i][0] = 0xff;
+               offload->self_ipv6_addr[i][1] = 0x02;
+               offload->self_ipv6_addr[i][11] = 0x01;
+               offload->self_ipv6_addr[i][12] = 0xff;
+               offload->self_ipv6_addr[i][13] =
+                                       offload->ipv6_addr[i][13];
+               offload->self_ipv6_addr[i][14] =
+                                       offload->ipv6_addr[i][14];
+               offload->self_ipv6_addr[i][15] =
+                                       offload->ipv6_addr[i][15];
+               ath11k_dbg(ar->ab, ATH11K_DBG_MAC, "NS solicited addr %pI6\n",
+                          offload->self_ipv6_addr[i]);
+       }
+}
+
+static void ath11k_mac_op_ipv6_changed(struct ieee80211_hw *hw,
+                                      struct ieee80211_vif *vif,
+                                      struct inet6_dev *idev)
+{
+       struct ath11k *ar = hw->priv;
+       struct ath11k_arp_ns_offload *offload;
+       struct ath11k_vif *arvif = ath11k_vif_to_arvif(vif);
+       struct inet6_ifaddr *ifa6;
+       struct ifacaddr6 *ifaca6;
+       struct list_head *p;
+       u32 count, scope;
+
+       ath11k_dbg(ar->ab, ATH11K_DBG_MAC, "mac op ipv6 changed\n");
+
+       offload = &arvif->arp_ns_offload;
+       count = 0;
+
+       read_lock_bh(&idev->lock);
+
+       memset(offload->ipv6_addr, 0, sizeof(offload->ipv6_addr));
+       memset(offload->self_ipv6_addr, 0, sizeof(offload->self_ipv6_addr));
+       memcpy(offload->mac_addr, vif->addr, ETH_ALEN);
+
+       /* get unicast address */
+       list_for_each(p, &idev->addr_list) {
+               if (count >= ATH11K_IPV6_MAX_COUNT)
+                       goto generate;
+
+               ifa6 = list_entry(p, struct inet6_ifaddr, if_list);
+               if (ifa6->flags & IFA_F_DADFAILED)
+                       continue;
+               scope = ipv6_addr_src_scope(&ifa6->addr);
+               if (scope == IPV6_ADDR_SCOPE_LINKLOCAL ||
+                   scope == IPV6_ADDR_SCOPE_GLOBAL) {
+                       memcpy(offload->ipv6_addr[count], &ifa6->addr.s6_addr,
+                              sizeof(ifa6->addr.s6_addr));
+                       offload->ipv6_type[count] = ATH11K_IPV6_UC_TYPE;
+                       ath11k_dbg(ar->ab, ATH11K_DBG_MAC, "mac count %d ipv6 uc %pI6 scope %d\n",
+                                  count, offload->ipv6_addr[count],
+                                  scope);
+                       count++;
+               } else {
+                       ath11k_warn(ar->ab, "Unsupported ipv6 scope: %d\n", scope);
+               }
+       }
+
+       /* get anycast address */
+       for (ifaca6 = idev->ac_list; ifaca6; ifaca6 = ifaca6->aca_next) {
+               if (count >= ATH11K_IPV6_MAX_COUNT)
+                       goto generate;
+
+               scope = ipv6_addr_src_scope(&ifaca6->aca_addr);
+               if (scope == IPV6_ADDR_SCOPE_LINKLOCAL ||
+                   scope == IPV6_ADDR_SCOPE_GLOBAL) {
+                       memcpy(offload->ipv6_addr[count], &ifaca6->aca_addr,
+                              sizeof(ifaca6->aca_addr));
+                       offload->ipv6_type[count] = ATH11K_IPV6_AC_TYPE;
+                       ath11k_dbg(ar->ab, ATH11K_DBG_MAC, "mac count %d ipv6 ac %pI6 scope %d\n",
+                                  count, offload->ipv6_addr[count],
+                                  scope);
+                       count++;
+               } else {
+                       ath11k_warn(ar->ab, "Unsupported ipv scope: %d\n", scope);
+               }
+       }
+
+generate:
+       offload->ipv6_count = count;
+       read_unlock_bh(&idev->lock);
+
+       /* generate ns multicast address */
+       ath11k_generate_ns_mc_addr(ar, offload);
+}
+
 static const struct ieee80211_ops ath11k_ops = {
        .tx                             = ath11k_mac_op_tx,
        .start                          = ath11k_mac_op_start,
@@ -8132,6 +8245,11 @@ static const struct ieee80211_ops ath11k_ops = {
 #ifdef CONFIG_ATH11K_DEBUGFS
        .sta_add_debugfs                = ath11k_debugfs_sta_op_add,
 #endif
+
+#if IS_ENABLED(CONFIG_IPV6)
+       .ipv6_addr_change = ath11k_mac_op_ipv6_changed,
+#endif
+
 };
 
 static void ath11k_mac_update_ch_list(struct ath11k *ar,
index 62b13f5..828f9ca 100644 (file)
@@ -8580,3 +8580,158 @@ int ath11k_wmi_wow_config_pno(struct ath11k *ar, u32 vdev_id,
        return ath11k_wmi_cmd_send(ar->wmi, skb, WMI_NETWORK_LIST_OFFLOAD_CONFIG_CMDID);
 }
 
+static void ath11k_wmi_fill_ns_offload(struct ath11k *ar,
+                                      struct ath11k_arp_ns_offload *offload,
+                                      u8 **ptr,
+                                      bool enable,
+                                      bool ext)
+{
+       struct wmi_ns_offload_tuple *ns;
+       struct wmi_tlv *tlv;
+       u8 *buf_ptr = *ptr;
+       u32 ns_cnt, ns_ext_tuples;
+       int i, max_offloads;
+
+       ns_cnt = offload->ipv6_count;
+
+       tlv  = (struct wmi_tlv *)buf_ptr;
+
+       if (ext) {
+               ns_ext_tuples = offload->ipv6_count - WMI_MAX_NS_OFFLOADS;
+               tlv->header = FIELD_PREP(WMI_TLV_TAG, WMI_TAG_ARRAY_STRUCT) |
+                             FIELD_PREP(WMI_TLV_LEN, ns_ext_tuples * sizeof(*ns));
+               i = WMI_MAX_NS_OFFLOADS;
+               max_offloads = offload->ipv6_count;
+       } else {
+               tlv->header = FIELD_PREP(WMI_TLV_TAG, WMI_TAG_ARRAY_STRUCT) |
+                             FIELD_PREP(WMI_TLV_LEN, WMI_MAX_NS_OFFLOADS * sizeof(*ns));
+               i = 0;
+               max_offloads = WMI_MAX_NS_OFFLOADS;
+       }
+
+       buf_ptr += sizeof(*tlv);
+
+       for (; i < max_offloads; i++) {
+               ns = (struct wmi_ns_offload_tuple *)buf_ptr;
+               ns->tlv_header = FIELD_PREP(WMI_TLV_TAG, WMI_TAG_NS_OFFLOAD_TUPLE) |
+                                FIELD_PREP(WMI_TLV_LEN, sizeof(*ns) - TLV_HDR_SIZE);
+
+               if (enable) {
+                       if (i < ns_cnt)
+                               ns->flags |= WMI_NSOL_FLAGS_VALID;
+
+                       memcpy(ns->target_ipaddr[0], offload->ipv6_addr[i], 16);
+                       memcpy(ns->solicitation_ipaddr, offload->self_ipv6_addr[i], 16);
+                       ath11k_ce_byte_swap(ns->target_ipaddr[0], 16);
+                       ath11k_ce_byte_swap(ns->solicitation_ipaddr, 16);
+
+                       if (offload->ipv6_type[i])
+                               ns->flags |= WMI_NSOL_FLAGS_IS_IPV6_ANYCAST;
+
+                       memcpy(ns->target_mac.addr, offload->mac_addr, ETH_ALEN);
+                       ath11k_ce_byte_swap(ns->target_mac.addr, 8);
+
+                       if (ns->target_mac.word0 != 0 ||
+                           ns->target_mac.word1 != 0) {
+                               ns->flags |= WMI_NSOL_FLAGS_MAC_VALID;
+                       }
+
+                       ath11k_dbg(ar->ab, ATH11K_DBG_WMI,
+                                  "wmi index %d ns_solicited %pI6 target %pI6",
+                                  i, ns->solicitation_ipaddr,
+                                  ns->target_ipaddr[0]);
+               }
+
+               buf_ptr += sizeof(*ns);
+       }
+
+       *ptr = buf_ptr;
+}
+
+static void ath11k_wmi_fill_arp_offload(struct ath11k *ar,
+                                       struct ath11k_arp_ns_offload *offload,
+                                       u8 **ptr,
+                                       bool enable)
+{
+       struct wmi_arp_offload_tuple *arp;
+       struct wmi_tlv *tlv;
+       u8 *buf_ptr = *ptr;
+       int i;
+
+       /* fill arp tuple */
+       tlv = (struct wmi_tlv *)buf_ptr;
+       tlv->header = FIELD_PREP(WMI_TLV_TAG, WMI_TAG_ARRAY_STRUCT) |
+                     FIELD_PREP(WMI_TLV_LEN, WMI_MAX_ARP_OFFLOADS * sizeof(*arp));
+       buf_ptr += sizeof(*tlv);
+
+       for (i = 0; i < WMI_MAX_ARP_OFFLOADS; i++) {
+               arp = (struct wmi_arp_offload_tuple *)buf_ptr;
+               arp->tlv_header = FIELD_PREP(WMI_TLV_TAG, WMI_TAG_ARP_OFFLOAD_TUPLE) |
+                                 FIELD_PREP(WMI_TLV_LEN, sizeof(*arp) - TLV_HDR_SIZE);
+
+               if (enable && i < offload->ipv4_count) {
+                       /* Copy the target ip addr and flags */
+                       arp->flags = WMI_ARPOL_FLAGS_VALID;
+                       memcpy(arp->target_ipaddr, offload->ipv4_addr[i], 4);
+                       ath11k_ce_byte_swap(arp->target_ipaddr, 4);
+
+                       ath11k_dbg(ar->ab, ATH11K_DBG_WMI, "wmi arp offload address %pI4",
+                                  arp->target_ipaddr);
+               }
+
+               buf_ptr += sizeof(*arp);
+       }
+
+       *ptr = buf_ptr;
+}
+
+int ath11k_wmi_arp_ns_offload(struct ath11k *ar,
+                             struct ath11k_vif *arvif, bool enable)
+{
+       struct ath11k_arp_ns_offload *offload;
+       struct wmi_set_arp_ns_offload_cmd *cmd;
+       struct wmi_tlv *tlv;
+       struct sk_buff *skb;
+       u8 *buf_ptr;
+       size_t len;
+       u8 ns_cnt, ns_ext_tuples = 0;
+
+       offload = &arvif->arp_ns_offload;
+       ns_cnt = offload->ipv6_count;
+
+       len = sizeof(*cmd) +
+             sizeof(*tlv) +
+             WMI_MAX_NS_OFFLOADS * sizeof(struct wmi_ns_offload_tuple) +
+             sizeof(*tlv) +
+             WMI_MAX_ARP_OFFLOADS * sizeof(struct wmi_arp_offload_tuple);
+
+       if (ns_cnt > WMI_MAX_NS_OFFLOADS) {
+               ns_ext_tuples = ns_cnt - WMI_MAX_NS_OFFLOADS;
+               len += sizeof(*tlv) +
+                      ns_ext_tuples * sizeof(struct wmi_ns_offload_tuple);
+       }
+
+       skb = ath11k_wmi_alloc_skb(ar->wmi->wmi_ab, len);
+       if (!skb)
+               return -ENOMEM;
+
+       buf_ptr = skb->data;
+       cmd = (struct wmi_set_arp_ns_offload_cmd *)buf_ptr;
+       cmd->tlv_header = FIELD_PREP(WMI_TLV_TAG,
+                                    WMI_TAG_SET_ARP_NS_OFFLOAD_CMD) |
+                         FIELD_PREP(WMI_TLV_LEN, sizeof(*cmd) - TLV_HDR_SIZE);
+
+       cmd->flags = 0;
+       cmd->vdev_id = arvif->vdev_id;
+       cmd->num_ns_ext_tuples = ns_ext_tuples;
+
+       buf_ptr += sizeof(*cmd);
+
+       ath11k_wmi_fill_ns_offload(ar, offload, &buf_ptr, enable, 0);
+       ath11k_wmi_fill_arp_offload(ar, offload, &buf_ptr, enable);
+
+       if (ns_ext_tuples)
+               ath11k_wmi_fill_ns_offload(ar, offload, &buf_ptr, enable, 1);
+
+       return ath11k_wmi_cmd_send(ar->wmi, skb, WMI_SET_ARP_NS_OFFLOAD_CMDID);
+}
index 17e5098..e2d03c3 100644 (file)
@@ -13,6 +13,7 @@ struct ath11k_base;
 struct ath11k;
 struct ath11k_fw_stats;
 struct ath11k_fw_dbglog;
+struct ath11k_vif;
 
 #define PSOC_HOST_MAX_NUM_SS (8)
 
@@ -5796,6 +5797,49 @@ struct wmi_wow_nlo_config_cmd {
         */
 } __packed;
 
+#define WMI_MAX_NS_OFFLOADS           2
+#define WMI_MAX_ARP_OFFLOADS          2
+
+#define WMI_ARPOL_FLAGS_VALID              BIT(0)
+#define WMI_ARPOL_FLAGS_MAC_VALID          BIT(1)
+#define WMI_ARPOL_FLAGS_REMOTE_IP_VALID    BIT(2)
+
+struct wmi_arp_offload_tuple {
+       u32 tlv_header;
+       u32 flags;
+       u8 target_ipaddr[4];
+       u8 remote_ipaddr[4];
+       struct wmi_mac_addr target_mac;
+} __packed;
+
+#define WMI_NSOL_FLAGS_VALID               BIT(0)
+#define WMI_NSOL_FLAGS_MAC_VALID           BIT(1)
+#define WMI_NSOL_FLAGS_REMOTE_IP_VALID     BIT(2)
+#define WMI_NSOL_FLAGS_IS_IPV6_ANYCAST     BIT(3)
+
+#define WMI_NSOL_MAX_TARGET_IPS    2
+
+struct wmi_ns_offload_tuple {
+       u32 tlv_header;
+       u32 flags;
+       u8 target_ipaddr[WMI_NSOL_MAX_TARGET_IPS][16];
+       u8 solicitation_ipaddr[16];
+       u8 remote_ipaddr[16];
+       struct wmi_mac_addr target_mac;
+} __packed;
+
+struct wmi_set_arp_ns_offload_cmd {
+       u32 tlv_header;
+       u32 flags;
+       u32 vdev_id;
+       u32 num_ns_ext_tuples;
+       /* The TLVs follow:
+        * wmi_ns_offload_tuple  ns_tuples[WMI_MAX_NS_OFFLOADS];
+        * wmi_arp_offload_tuple arp_tuples[WMI_MAX_ARP_OFFLOADS];
+        * wmi_ns_offload_tuple  ns_ext_tuples[num_ns_ext_tuples];
+        */
+} __packed;
+
 int ath11k_wmi_cmd_send(struct ath11k_pdev_wmi *wmi, struct sk_buff *skb,
                        u32 cmd_id);
 struct sk_buff *ath11k_wmi_alloc_skb(struct ath11k_wmi_base *wmi_sc, u32 len);
@@ -5968,4 +6012,7 @@ int ath11k_wmi_wow_add_wakeup_event(struct ath11k *ar, u32 vdev_id,
                                    u32 enable);
 int ath11k_wmi_hw_data_filter_cmd(struct ath11k *ar, u32 vdev_id,
                                  u32 filter_bitmap, bool enable);
+int ath11k_wmi_arp_ns_offload(struct ath11k *ar,
+                             struct ath11k_vif *arvif, bool enable);
+
 #endif
index 602b8f4..bc6a4d3 100644 (file)
@@ -559,6 +559,43 @@ static int ath11k_wow_clear_hw_filter(struct ath11k *ar)
        return 0;
 }
 
+static int ath11k_wow_arp_ns_offload(struct ath11k *ar, bool enable)
+{
+       struct ath11k_vif *arvif;
+       int ret;
+
+       lockdep_assert_held(&ar->conf_mutex);
+
+       list_for_each_entry(arvif, &ar->arvifs, list) {
+               if (arvif->vdev_type != WMI_VDEV_TYPE_STA)
+                       continue;
+
+               ret = ath11k_wmi_arp_ns_offload(ar, arvif, enable);
+
+               if (ret) {
+                       ath11k_warn(ar->ab, "failed to set arp ns offload vdev %i: enable %d, ret %d\n",
+                                   arvif->vdev_id, enable, ret);
+                       return ret;
+               }
+       }
+
+       return 0;
+}
+
+static int ath11k_wow_protocol_offload(struct ath11k *ar, bool enable)
+{
+       int ret;
+
+       ret = ath11k_wow_arp_ns_offload(ar, enable);
+       if (ret) {
+               ath11k_warn(ar->ab, "failed to offload ARP and NS %d %d\n",
+                           enable, ret);
+               return ret;
+       }
+
+       return 0;
+}
+
 int ath11k_wow_op_suspend(struct ieee80211_hw *hw,
                          struct cfg80211_wowlan *wowlan)
 {
@@ -589,6 +626,14 @@ int ath11k_wow_op_suspend(struct ieee80211_hw *hw,
                goto cleanup;
        }
 
+       ret = ath11k_wow_protocol_offload(ar, true);
+       if (ret) {
+               ath11k_warn(ar->ab, "failed to set wow protocol offload events: %d\n",
+                           ret);
+               goto cleanup;
+       }
+
+       ath11k_mac_drain_tx(ar);
        ret = ath11k_mac_wait_tx_complete(ar);
        if (ret) {
                ath11k_warn(ar->ab, "failed to wait tx complete: %d\n", ret);
@@ -690,6 +735,13 @@ int ath11k_wow_op_resume(struct ieee80211_hw *hw)
                goto exit;
        }
 
+       ret = ath11k_wow_protocol_offload(ar, false);
+       if (ret) {
+               ath11k_warn(ar->ab, "failed to clear wow protocol offload events: %d\n",
+                           ret);
+               goto exit;
+       }
+
 exit:
        if (ret) {
                switch (ar->state) {