Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / drivers / net / wireless / ath / ath10k / htt_rx.c
diff --git a/kernel/drivers/net/wireless/ath/ath10k/htt_rx.c b/kernel/drivers/net/wireless/ath/ath10k/htt_rx.c
new file mode 100644 (file)
index 0000000..01a2b38
--- /dev/null
@@ -0,0 +1,2065 @@
+/*
+ * Copyright (c) 2005-2011 Atheros Communications Inc.
+ * Copyright (c) 2011-2013 Qualcomm Atheros, Inc.
+ *
+ * Permission to use, copy, modify, and/or distribute this software for any
+ * purpose with or without fee is hereby granted, provided that the above
+ * copyright notice and this permission notice appear in all copies.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
+ * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
+ * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
+ * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
+ * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
+ * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
+ * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
+ */
+
+#include "core.h"
+#include "htc.h"
+#include "htt.h"
+#include "txrx.h"
+#include "debug.h"
+#include "trace.h"
+#include "mac.h"
+
+#include <linux/log2.h>
+
+#define HTT_RX_RING_SIZE HTT_RX_RING_SIZE_MAX
+#define HTT_RX_RING_FILL_LEVEL (((HTT_RX_RING_SIZE) / 2) - 1)
+
+/* when under memory pressure rx ring refill may fail and needs a retry */
+#define HTT_RX_RING_REFILL_RETRY_MS 50
+
+static int ath10k_htt_rx_get_csum_state(struct sk_buff *skb);
+static void ath10k_htt_txrx_compl_task(unsigned long ptr);
+
+static struct sk_buff *
+ath10k_htt_rx_find_skb_paddr(struct ath10k *ar, u32 paddr)
+{
+       struct ath10k_skb_rxcb *rxcb;
+
+       hash_for_each_possible(ar->htt.rx_ring.skb_table, rxcb, hlist, paddr)
+               if (rxcb->paddr == paddr)
+                       return ATH10K_RXCB_SKB(rxcb);
+
+       WARN_ON_ONCE(1);
+       return NULL;
+}
+
+static void ath10k_htt_rx_ring_free(struct ath10k_htt *htt)
+{
+       struct sk_buff *skb;
+       struct ath10k_skb_rxcb *rxcb;
+       struct hlist_node *n;
+       int i;
+
+       if (htt->rx_ring.in_ord_rx) {
+               hash_for_each_safe(htt->rx_ring.skb_table, i, n, rxcb, hlist) {
+                       skb = ATH10K_RXCB_SKB(rxcb);
+                       dma_unmap_single(htt->ar->dev, rxcb->paddr,
+                                        skb->len + skb_tailroom(skb),
+                                        DMA_FROM_DEVICE);
+                       hash_del(&rxcb->hlist);
+                       dev_kfree_skb_any(skb);
+               }
+       } else {
+               for (i = 0; i < htt->rx_ring.size; i++) {
+                       skb = htt->rx_ring.netbufs_ring[i];
+                       if (!skb)
+                               continue;
+
+                       rxcb = ATH10K_SKB_RXCB(skb);
+                       dma_unmap_single(htt->ar->dev, rxcb->paddr,
+                                        skb->len + skb_tailroom(skb),
+                                        DMA_FROM_DEVICE);
+                       dev_kfree_skb_any(skb);
+               }
+       }
+
+       htt->rx_ring.fill_cnt = 0;
+       hash_init(htt->rx_ring.skb_table);
+       memset(htt->rx_ring.netbufs_ring, 0,
+              htt->rx_ring.size * sizeof(htt->rx_ring.netbufs_ring[0]));
+}
+
+static int __ath10k_htt_rx_ring_fill_n(struct ath10k_htt *htt, int num)
+{
+       struct htt_rx_desc *rx_desc;
+       struct ath10k_skb_rxcb *rxcb;
+       struct sk_buff *skb;
+       dma_addr_t paddr;
+       int ret = 0, idx;
+
+       /* The Full Rx Reorder firmware has no way of telling the host
+        * implicitly when it copied HTT Rx Ring buffers to MAC Rx Ring.
+        * To keep things simple make sure ring is always half empty. This
+        * guarantees there'll be no replenishment overruns possible.
+        */
+       BUILD_BUG_ON(HTT_RX_RING_FILL_LEVEL >= HTT_RX_RING_SIZE / 2);
+
+       idx = __le32_to_cpu(*htt->rx_ring.alloc_idx.vaddr);
+       while (num > 0) {
+               skb = dev_alloc_skb(HTT_RX_BUF_SIZE + HTT_RX_DESC_ALIGN);
+               if (!skb) {
+                       ret = -ENOMEM;
+                       goto fail;
+               }
+
+               if (!IS_ALIGNED((unsigned long)skb->data, HTT_RX_DESC_ALIGN))
+                       skb_pull(skb,
+                                PTR_ALIGN(skb->data, HTT_RX_DESC_ALIGN) -
+                                skb->data);
+
+               /* Clear rx_desc attention word before posting to Rx ring */
+               rx_desc = (struct htt_rx_desc *)skb->data;
+               rx_desc->attention.flags = __cpu_to_le32(0);
+
+               paddr = dma_map_single(htt->ar->dev, skb->data,
+                                      skb->len + skb_tailroom(skb),
+                                      DMA_FROM_DEVICE);
+
+               if (unlikely(dma_mapping_error(htt->ar->dev, paddr))) {
+                       dev_kfree_skb_any(skb);
+                       ret = -ENOMEM;
+                       goto fail;
+               }
+
+               rxcb = ATH10K_SKB_RXCB(skb);
+               rxcb->paddr = paddr;
+               htt->rx_ring.netbufs_ring[idx] = skb;
+               htt->rx_ring.paddrs_ring[idx] = __cpu_to_le32(paddr);
+               htt->rx_ring.fill_cnt++;
+
+               if (htt->rx_ring.in_ord_rx) {
+                       hash_add(htt->rx_ring.skb_table,
+                                &ATH10K_SKB_RXCB(skb)->hlist,
+                                (u32)paddr);
+               }
+
+               num--;
+               idx++;
+               idx &= htt->rx_ring.size_mask;
+       }
+
+fail:
+       /*
+        * Make sure the rx buffer is updated before available buffer
+        * index to avoid any potential rx ring corruption.
+        */
+       mb();
+       *htt->rx_ring.alloc_idx.vaddr = __cpu_to_le32(idx);
+       return ret;
+}
+
+static int ath10k_htt_rx_ring_fill_n(struct ath10k_htt *htt, int num)
+{
+       lockdep_assert_held(&htt->rx_ring.lock);
+       return __ath10k_htt_rx_ring_fill_n(htt, num);
+}
+
+static void ath10k_htt_rx_msdu_buff_replenish(struct ath10k_htt *htt)
+{
+       int ret, num_deficit, num_to_fill;
+
+       /* Refilling the whole RX ring buffer proves to be a bad idea. The
+        * reason is RX may take up significant amount of CPU cycles and starve
+        * other tasks, e.g. TX on an ethernet device while acting as a bridge
+        * with ath10k wlan interface. This ended up with very poor performance
+        * once CPU the host system was overwhelmed with RX on ath10k.
+        *
+        * By limiting the number of refills the replenishing occurs
+        * progressively. This in turns makes use of the fact tasklets are
+        * processed in FIFO order. This means actual RX processing can starve
+        * out refilling. If there's not enough buffers on RX ring FW will not
+        * report RX until it is refilled with enough buffers. This
+        * automatically balances load wrt to CPU power.
+        *
+        * This probably comes at a cost of lower maximum throughput but
+        * improves the average and stability. */
+       spin_lock_bh(&htt->rx_ring.lock);
+       num_deficit = htt->rx_ring.fill_level - htt->rx_ring.fill_cnt;
+       num_to_fill = min(ATH10K_HTT_MAX_NUM_REFILL, num_deficit);
+       num_deficit -= num_to_fill;
+       ret = ath10k_htt_rx_ring_fill_n(htt, num_to_fill);
+       if (ret == -ENOMEM) {
+               /*
+                * Failed to fill it to the desired level -
+                * we'll start a timer and try again next time.
+                * As long as enough buffers are left in the ring for
+                * another A-MPDU rx, no special recovery is needed.
+                */
+               mod_timer(&htt->rx_ring.refill_retry_timer, jiffies +
+                         msecs_to_jiffies(HTT_RX_RING_REFILL_RETRY_MS));
+       } else if (num_deficit > 0) {
+               tasklet_schedule(&htt->rx_replenish_task);
+       }
+       spin_unlock_bh(&htt->rx_ring.lock);
+}
+
+static void ath10k_htt_rx_ring_refill_retry(unsigned long arg)
+{
+       struct ath10k_htt *htt = (struct ath10k_htt *)arg;
+
+       ath10k_htt_rx_msdu_buff_replenish(htt);
+}
+
+int ath10k_htt_rx_ring_refill(struct ath10k *ar)
+{
+       struct ath10k_htt *htt = &ar->htt;
+       int ret;
+
+       spin_lock_bh(&htt->rx_ring.lock);
+       ret = ath10k_htt_rx_ring_fill_n(htt, (htt->rx_ring.fill_level -
+                                             htt->rx_ring.fill_cnt));
+       spin_unlock_bh(&htt->rx_ring.lock);
+
+       if (ret)
+               ath10k_htt_rx_ring_free(htt);
+
+       return ret;
+}
+
+void ath10k_htt_rx_free(struct ath10k_htt *htt)
+{
+       del_timer_sync(&htt->rx_ring.refill_retry_timer);
+       tasklet_kill(&htt->rx_replenish_task);
+       tasklet_kill(&htt->txrx_compl_task);
+
+       skb_queue_purge(&htt->tx_compl_q);
+       skb_queue_purge(&htt->rx_compl_q);
+       skb_queue_purge(&htt->rx_in_ord_compl_q);
+
+       ath10k_htt_rx_ring_free(htt);
+
+       dma_free_coherent(htt->ar->dev,
+                         (htt->rx_ring.size *
+                          sizeof(htt->rx_ring.paddrs_ring)),
+                         htt->rx_ring.paddrs_ring,
+                         htt->rx_ring.base_paddr);
+
+       dma_free_coherent(htt->ar->dev,
+                         sizeof(*htt->rx_ring.alloc_idx.vaddr),
+                         htt->rx_ring.alloc_idx.vaddr,
+                         htt->rx_ring.alloc_idx.paddr);
+
+       kfree(htt->rx_ring.netbufs_ring);
+}
+
+static inline struct sk_buff *ath10k_htt_rx_netbuf_pop(struct ath10k_htt *htt)
+{
+       struct ath10k *ar = htt->ar;
+       int idx;
+       struct sk_buff *msdu;
+
+       lockdep_assert_held(&htt->rx_ring.lock);
+
+       if (htt->rx_ring.fill_cnt == 0) {
+               ath10k_warn(ar, "tried to pop sk_buff from an empty rx ring\n");
+               return NULL;
+       }
+
+       idx = htt->rx_ring.sw_rd_idx.msdu_payld;
+       msdu = htt->rx_ring.netbufs_ring[idx];
+       htt->rx_ring.netbufs_ring[idx] = NULL;
+       htt->rx_ring.paddrs_ring[idx] = 0;
+
+       idx++;
+       idx &= htt->rx_ring.size_mask;
+       htt->rx_ring.sw_rd_idx.msdu_payld = idx;
+       htt->rx_ring.fill_cnt--;
+
+       dma_unmap_single(htt->ar->dev,
+                        ATH10K_SKB_RXCB(msdu)->paddr,
+                        msdu->len + skb_tailroom(msdu),
+                        DMA_FROM_DEVICE);
+       ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt rx netbuf pop: ",
+                       msdu->data, msdu->len + skb_tailroom(msdu));
+
+       return msdu;
+}
+
+/* return: < 0 fatal error, 0 - non chained msdu, 1 chained msdu */
+static int ath10k_htt_rx_amsdu_pop(struct ath10k_htt *htt,
+                                  u8 **fw_desc, int *fw_desc_len,
+                                  struct sk_buff_head *amsdu)
+{
+       struct ath10k *ar = htt->ar;
+       int msdu_len, msdu_chaining = 0;
+       struct sk_buff *msdu;
+       struct htt_rx_desc *rx_desc;
+
+       lockdep_assert_held(&htt->rx_ring.lock);
+
+       for (;;) {
+               int last_msdu, msdu_len_invalid, msdu_chained;
+
+               msdu = ath10k_htt_rx_netbuf_pop(htt);
+               if (!msdu) {
+                       __skb_queue_purge(amsdu);
+                       return -ENOENT;
+               }
+
+               __skb_queue_tail(amsdu, msdu);
+
+               rx_desc = (struct htt_rx_desc *)msdu->data;
+
+               /* FIXME: we must report msdu payload since this is what caller
+                *        expects now */
+               skb_put(msdu, offsetof(struct htt_rx_desc, msdu_payload));
+               skb_pull(msdu, offsetof(struct htt_rx_desc, msdu_payload));
+
+               /*
+                * Sanity check - confirm the HW is finished filling in the
+                * rx data.
+                * If the HW and SW are working correctly, then it's guaranteed
+                * that the HW's MAC DMA is done before this point in the SW.
+                * To prevent the case that we handle a stale Rx descriptor,
+                * just assert for now until we have a way to recover.
+                */
+               if (!(__le32_to_cpu(rx_desc->attention.flags)
+                               & RX_ATTENTION_FLAGS_MSDU_DONE)) {
+                       __skb_queue_purge(amsdu);
+                       return -EIO;
+               }
+
+               /*
+                * Copy the FW rx descriptor for this MSDU from the rx
+                * indication message into the MSDU's netbuf. HL uses the
+                * same rx indication message definition as LL, and simply
+                * appends new info (fields from the HW rx desc, and the
+                * MSDU payload itself). So, the offset into the rx
+                * indication message only has to account for the standard
+                * offset of the per-MSDU FW rx desc info within the
+                * message, and how many bytes of the per-MSDU FW rx desc
+                * info have already been consumed. (And the endianness of
+                * the host, since for a big-endian host, the rx ind
+                * message contents, including the per-MSDU rx desc bytes,
+                * were byteswapped during upload.)
+                */
+               if (*fw_desc_len > 0) {
+                       rx_desc->fw_desc.info0 = **fw_desc;
+                       /*
+                        * The target is expected to only provide the basic
+                        * per-MSDU rx descriptors. Just to be sure, verify
+                        * that the target has not attached extension data
+                        * (e.g. LRO flow ID).
+                        */
+
+                       /* or more, if there's extension data */
+                       (*fw_desc)++;
+                       (*fw_desc_len)--;
+               } else {
+                       /*
+                        * When an oversized AMSDU happened, FW will lost
+                        * some of MSDU status - in this case, the FW
+                        * descriptors provided will be less than the
+                        * actual MSDUs inside this MPDU. Mark the FW
+                        * descriptors so that it will still deliver to
+                        * upper stack, if no CRC error for this MPDU.
+                        *
+                        * FIX THIS - the FW descriptors are actually for
+                        * MSDUs in the end of this A-MSDU instead of the
+                        * beginning.
+                        */
+                       rx_desc->fw_desc.info0 = 0;
+               }
+
+               msdu_len_invalid = !!(__le32_to_cpu(rx_desc->attention.flags)
+                                       & (RX_ATTENTION_FLAGS_MPDU_LENGTH_ERR |
+                                          RX_ATTENTION_FLAGS_MSDU_LENGTH_ERR));
+               msdu_len = MS(__le32_to_cpu(rx_desc->msdu_start.info0),
+                             RX_MSDU_START_INFO0_MSDU_LENGTH);
+               msdu_chained = rx_desc->frag_info.ring2_more_count;
+
+               if (msdu_len_invalid)
+                       msdu_len = 0;
+
+               skb_trim(msdu, 0);
+               skb_put(msdu, min(msdu_len, HTT_RX_MSDU_SIZE));
+               msdu_len -= msdu->len;
+
+               /* Note: Chained buffers do not contain rx descriptor */
+               while (msdu_chained--) {
+                       msdu = ath10k_htt_rx_netbuf_pop(htt);
+                       if (!msdu) {
+                               __skb_queue_purge(amsdu);
+                               return -ENOENT;
+                       }
+
+                       __skb_queue_tail(amsdu, msdu);
+                       skb_trim(msdu, 0);
+                       skb_put(msdu, min(msdu_len, HTT_RX_BUF_SIZE));
+                       msdu_len -= msdu->len;
+                       msdu_chaining = 1;
+               }
+
+               last_msdu = __le32_to_cpu(rx_desc->msdu_end.info0) &
+                               RX_MSDU_END_INFO0_LAST_MSDU;
+
+               trace_ath10k_htt_rx_desc(ar, &rx_desc->attention,
+                                        sizeof(*rx_desc) - sizeof(u32));
+
+               if (last_msdu)
+                       break;
+       }
+
+       if (skb_queue_empty(amsdu))
+               msdu_chaining = -1;
+
+       /*
+        * Don't refill the ring yet.
+        *
+        * First, the elements popped here are still in use - it is not
+        * safe to overwrite them until the matching call to
+        * mpdu_desc_list_next. Second, for efficiency it is preferable to
+        * refill the rx ring with 1 PPDU's worth of rx buffers (something
+        * like 32 x 3 buffers), rather than one MPDU's worth of rx buffers
+        * (something like 3 buffers). Consequently, we'll rely on the txrx
+        * SW to tell us when it is done pulling all the PPDU's rx buffers
+        * out of the rx ring, and then refill it just once.
+        */
+
+       return msdu_chaining;
+}
+
+static void ath10k_htt_rx_replenish_task(unsigned long ptr)
+{
+       struct ath10k_htt *htt = (struct ath10k_htt *)ptr;
+
+       ath10k_htt_rx_msdu_buff_replenish(htt);
+}
+
+static struct sk_buff *ath10k_htt_rx_pop_paddr(struct ath10k_htt *htt,
+                                              u32 paddr)
+{
+       struct ath10k *ar = htt->ar;
+       struct ath10k_skb_rxcb *rxcb;
+       struct sk_buff *msdu;
+
+       lockdep_assert_held(&htt->rx_ring.lock);
+
+       msdu = ath10k_htt_rx_find_skb_paddr(ar, paddr);
+       if (!msdu)
+               return NULL;
+
+       rxcb = ATH10K_SKB_RXCB(msdu);
+       hash_del(&rxcb->hlist);
+       htt->rx_ring.fill_cnt--;
+
+       dma_unmap_single(htt->ar->dev, rxcb->paddr,
+                        msdu->len + skb_tailroom(msdu),
+                        DMA_FROM_DEVICE);
+       ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt rx netbuf pop: ",
+                       msdu->data, msdu->len + skb_tailroom(msdu));
+
+       return msdu;
+}
+
+static int ath10k_htt_rx_pop_paddr_list(struct ath10k_htt *htt,
+                                       struct htt_rx_in_ord_ind *ev,
+                                       struct sk_buff_head *list)
+{
+       struct ath10k *ar = htt->ar;
+       struct htt_rx_in_ord_msdu_desc *msdu_desc = ev->msdu_descs;
+       struct htt_rx_desc *rxd;
+       struct sk_buff *msdu;
+       int msdu_count;
+       bool is_offload;
+       u32 paddr;
+
+       lockdep_assert_held(&htt->rx_ring.lock);
+
+       msdu_count = __le16_to_cpu(ev->msdu_count);
+       is_offload = !!(ev->info & HTT_RX_IN_ORD_IND_INFO_OFFLOAD_MASK);
+
+       while (msdu_count--) {
+               paddr = __le32_to_cpu(msdu_desc->msdu_paddr);
+
+               msdu = ath10k_htt_rx_pop_paddr(htt, paddr);
+               if (!msdu) {
+                       __skb_queue_purge(list);
+                       return -ENOENT;
+               }
+
+               __skb_queue_tail(list, msdu);
+
+               if (!is_offload) {
+                       rxd = (void *)msdu->data;
+
+                       trace_ath10k_htt_rx_desc(ar, rxd, sizeof(*rxd));
+
+                       skb_put(msdu, sizeof(*rxd));
+                       skb_pull(msdu, sizeof(*rxd));
+                       skb_put(msdu, __le16_to_cpu(msdu_desc->msdu_len));
+
+                       if (!(__le32_to_cpu(rxd->attention.flags) &
+                             RX_ATTENTION_FLAGS_MSDU_DONE)) {
+                               ath10k_warn(htt->ar, "tried to pop an incomplete frame, oops!\n");
+                               return -EIO;
+                       }
+               }
+
+               msdu_desc++;
+       }
+
+       return 0;
+}
+
+int ath10k_htt_rx_alloc(struct ath10k_htt *htt)
+{
+       struct ath10k *ar = htt->ar;
+       dma_addr_t paddr;
+       void *vaddr;
+       size_t size;
+       struct timer_list *timer = &htt->rx_ring.refill_retry_timer;
+
+       htt->rx_confused = false;
+
+       /* XXX: The fill level could be changed during runtime in response to
+        * the host processing latency. Is this really worth it?
+        */
+       htt->rx_ring.size = HTT_RX_RING_SIZE;
+       htt->rx_ring.size_mask = htt->rx_ring.size - 1;
+       htt->rx_ring.fill_level = HTT_RX_RING_FILL_LEVEL;
+
+       if (!is_power_of_2(htt->rx_ring.size)) {
+               ath10k_warn(ar, "htt rx ring size is not power of 2\n");
+               return -EINVAL;
+       }
+
+       htt->rx_ring.netbufs_ring =
+               kzalloc(htt->rx_ring.size * sizeof(struct sk_buff *),
+                       GFP_KERNEL);
+       if (!htt->rx_ring.netbufs_ring)
+               goto err_netbuf;
+
+       size = htt->rx_ring.size * sizeof(htt->rx_ring.paddrs_ring);
+
+       vaddr = dma_alloc_coherent(htt->ar->dev, size, &paddr, GFP_DMA);
+       if (!vaddr)
+               goto err_dma_ring;
+
+       htt->rx_ring.paddrs_ring = vaddr;
+       htt->rx_ring.base_paddr = paddr;
+
+       vaddr = dma_alloc_coherent(htt->ar->dev,
+                                  sizeof(*htt->rx_ring.alloc_idx.vaddr),
+                                  &paddr, GFP_DMA);
+       if (!vaddr)
+               goto err_dma_idx;
+
+       htt->rx_ring.alloc_idx.vaddr = vaddr;
+       htt->rx_ring.alloc_idx.paddr = paddr;
+       htt->rx_ring.sw_rd_idx.msdu_payld = htt->rx_ring.size_mask;
+       *htt->rx_ring.alloc_idx.vaddr = 0;
+
+       /* Initialize the Rx refill retry timer */
+       setup_timer(timer, ath10k_htt_rx_ring_refill_retry, (unsigned long)htt);
+
+       spin_lock_init(&htt->rx_ring.lock);
+
+       htt->rx_ring.fill_cnt = 0;
+       htt->rx_ring.sw_rd_idx.msdu_payld = 0;
+       hash_init(htt->rx_ring.skb_table);
+
+       tasklet_init(&htt->rx_replenish_task, ath10k_htt_rx_replenish_task,
+                    (unsigned long)htt);
+
+       skb_queue_head_init(&htt->tx_compl_q);
+       skb_queue_head_init(&htt->rx_compl_q);
+       skb_queue_head_init(&htt->rx_in_ord_compl_q);
+
+       tasklet_init(&htt->txrx_compl_task, ath10k_htt_txrx_compl_task,
+                    (unsigned long)htt);
+
+       ath10k_dbg(ar, ATH10K_DBG_BOOT, "htt rx ring size %d fill_level %d\n",
+                  htt->rx_ring.size, htt->rx_ring.fill_level);
+       return 0;
+
+err_dma_idx:
+       dma_free_coherent(htt->ar->dev,
+                         (htt->rx_ring.size *
+                          sizeof(htt->rx_ring.paddrs_ring)),
+                         htt->rx_ring.paddrs_ring,
+                         htt->rx_ring.base_paddr);
+err_dma_ring:
+       kfree(htt->rx_ring.netbufs_ring);
+err_netbuf:
+       return -ENOMEM;
+}
+
+static int ath10k_htt_rx_crypto_param_len(struct ath10k *ar,
+                                         enum htt_rx_mpdu_encrypt_type type)
+{
+       switch (type) {
+       case HTT_RX_MPDU_ENCRYPT_NONE:
+               return 0;
+       case HTT_RX_MPDU_ENCRYPT_WEP40:
+       case HTT_RX_MPDU_ENCRYPT_WEP104:
+               return IEEE80211_WEP_IV_LEN;
+       case HTT_RX_MPDU_ENCRYPT_TKIP_WITHOUT_MIC:
+       case HTT_RX_MPDU_ENCRYPT_TKIP_WPA:
+               return IEEE80211_TKIP_IV_LEN;
+       case HTT_RX_MPDU_ENCRYPT_AES_CCM_WPA2:
+               return IEEE80211_CCMP_HDR_LEN;
+       case HTT_RX_MPDU_ENCRYPT_WEP128:
+       case HTT_RX_MPDU_ENCRYPT_WAPI:
+               break;
+       }
+
+       ath10k_warn(ar, "unsupported encryption type %d\n", type);
+       return 0;
+}
+
+#define MICHAEL_MIC_LEN 8
+
+static int ath10k_htt_rx_crypto_tail_len(struct ath10k *ar,
+                                        enum htt_rx_mpdu_encrypt_type type)
+{
+       switch (type) {
+       case HTT_RX_MPDU_ENCRYPT_NONE:
+               return 0;
+       case HTT_RX_MPDU_ENCRYPT_WEP40:
+       case HTT_RX_MPDU_ENCRYPT_WEP104:
+               return IEEE80211_WEP_ICV_LEN;
+       case HTT_RX_MPDU_ENCRYPT_TKIP_WITHOUT_MIC:
+       case HTT_RX_MPDU_ENCRYPT_TKIP_WPA:
+               return IEEE80211_TKIP_ICV_LEN;
+       case HTT_RX_MPDU_ENCRYPT_AES_CCM_WPA2:
+               return IEEE80211_CCMP_MIC_LEN;
+       case HTT_RX_MPDU_ENCRYPT_WEP128:
+       case HTT_RX_MPDU_ENCRYPT_WAPI:
+               break;
+       }
+
+       ath10k_warn(ar, "unsupported encryption type %d\n", type);
+       return 0;
+}
+
+struct rfc1042_hdr {
+       u8 llc_dsap;
+       u8 llc_ssap;
+       u8 llc_ctrl;
+       u8 snap_oui[3];
+       __be16 snap_type;
+} __packed;
+
+struct amsdu_subframe_hdr {
+       u8 dst[ETH_ALEN];
+       u8 src[ETH_ALEN];
+       __be16 len;
+} __packed;
+
+static const u8 rx_legacy_rate_idx[] = {
+       3,      /* 0x00  - 11Mbps  */
+       2,      /* 0x01  - 5.5Mbps */
+       1,      /* 0x02  - 2Mbps   */
+       0,      /* 0x03  - 1Mbps   */
+       3,      /* 0x04  - 11Mbps  */
+       2,      /* 0x05  - 5.5Mbps */
+       1,      /* 0x06  - 2Mbps   */
+       0,      /* 0x07  - 1Mbps   */
+       10,     /* 0x08  - 48Mbps  */
+       8,      /* 0x09  - 24Mbps  */
+       6,      /* 0x0A  - 12Mbps  */
+       4,      /* 0x0B  - 6Mbps   */
+       11,     /* 0x0C  - 54Mbps  */
+       9,      /* 0x0D  - 36Mbps  */
+       7,      /* 0x0E  - 18Mbps  */
+       5,      /* 0x0F  - 9Mbps   */
+};
+
+static void ath10k_htt_rx_h_rates(struct ath10k *ar,
+                                 struct ieee80211_rx_status *status,
+                                 struct htt_rx_desc *rxd)
+{
+       enum ieee80211_band band;
+       u8 cck, rate, rate_idx, bw, sgi, mcs, nss;
+       u8 preamble = 0;
+       u32 info1, info2, info3;
+
+       /* Band value can't be set as undefined but freq can be 0 - use that to
+        * determine whether band is provided.
+        *
+        * FIXME: Perhaps this can go away if CCK rate reporting is a little
+        * reworked?
+        */
+       if (!status->freq)
+               return;
+
+       band = status->band;
+       info1 = __le32_to_cpu(rxd->ppdu_start.info1);
+       info2 = __le32_to_cpu(rxd->ppdu_start.info2);
+       info3 = __le32_to_cpu(rxd->ppdu_start.info3);
+
+       preamble = MS(info1, RX_PPDU_START_INFO1_PREAMBLE_TYPE);
+
+       switch (preamble) {
+       case HTT_RX_LEGACY:
+               cck = info1 & RX_PPDU_START_INFO1_L_SIG_RATE_SELECT;
+               rate = MS(info1, RX_PPDU_START_INFO1_L_SIG_RATE);
+               rate_idx = 0;
+
+               if (rate < 0x08 || rate > 0x0F)
+                       break;
+
+               switch (band) {
+               case IEEE80211_BAND_2GHZ:
+                       if (cck)
+                               rate &= ~BIT(3);
+                       rate_idx = rx_legacy_rate_idx[rate];
+                       break;
+               case IEEE80211_BAND_5GHZ:
+                       rate_idx = rx_legacy_rate_idx[rate];
+                       /* We are using same rate table registering
+                          HW - ath10k_rates[]. In case of 5GHz skip
+                          CCK rates, so -4 here */
+                       rate_idx -= 4;
+                       break;
+               default:
+                       break;
+               }
+
+               status->rate_idx = rate_idx;
+               break;
+       case HTT_RX_HT:
+       case HTT_RX_HT_WITH_TXBF:
+               /* HT-SIG - Table 20-11 in info2 and info3 */
+               mcs = info2 & 0x1F;
+               nss = mcs >> 3;
+               bw = (info2 >> 7) & 1;
+               sgi = (info3 >> 7) & 1;
+
+               status->rate_idx = mcs;
+               status->flag |= RX_FLAG_HT;
+               if (sgi)
+                       status->flag |= RX_FLAG_SHORT_GI;
+               if (bw)
+                       status->flag |= RX_FLAG_40MHZ;
+               break;
+       case HTT_RX_VHT:
+       case HTT_RX_VHT_WITH_TXBF:
+               /* VHT-SIG-A1 in info2, VHT-SIG-A2 in info3
+                  TODO check this */
+               mcs = (info3 >> 4) & 0x0F;
+               nss = ((info2 >> 10) & 0x07) + 1;
+               bw = info2 & 3;
+               sgi = info3 & 1;
+
+               status->rate_idx = mcs;
+               status->vht_nss = nss;
+
+               if (sgi)
+                       status->flag |= RX_FLAG_SHORT_GI;
+
+               switch (bw) {
+               /* 20MHZ */
+               case 0:
+                       break;
+               /* 40MHZ */
+               case 1:
+                       status->flag |= RX_FLAG_40MHZ;
+                       break;
+               /* 80MHZ */
+               case 2:
+                       status->vht_flag |= RX_VHT_FLAG_80MHZ;
+               }
+
+               status->flag |= RX_FLAG_VHT;
+               break;
+       default:
+               break;
+       }
+}
+
+static bool ath10k_htt_rx_h_channel(struct ath10k *ar,
+                                   struct ieee80211_rx_status *status)
+{
+       struct ieee80211_channel *ch;
+
+       spin_lock_bh(&ar->data_lock);
+       ch = ar->scan_channel;
+       if (!ch)
+               ch = ar->rx_channel;
+       spin_unlock_bh(&ar->data_lock);
+
+       if (!ch)
+               return false;
+
+       status->band = ch->band;
+       status->freq = ch->center_freq;
+
+       return true;
+}
+
+static void ath10k_htt_rx_h_signal(struct ath10k *ar,
+                                  struct ieee80211_rx_status *status,
+                                  struct htt_rx_desc *rxd)
+{
+       /* FIXME: Get real NF */
+       status->signal = ATH10K_DEFAULT_NOISE_FLOOR +
+                        rxd->ppdu_start.rssi_comb;
+       status->flag &= ~RX_FLAG_NO_SIGNAL_VAL;
+}
+
+static void ath10k_htt_rx_h_mactime(struct ath10k *ar,
+                                   struct ieee80211_rx_status *status,
+                                   struct htt_rx_desc *rxd)
+{
+       /* FIXME: TSF is known only at the end of PPDU, in the last MPDU. This
+        * means all prior MSDUs in a PPDU are reported to mac80211 without the
+        * TSF. Is it worth holding frames until end of PPDU is known?
+        *
+        * FIXME: Can we get/compute 64bit TSF?
+        */
+       status->mactime = __le32_to_cpu(rxd->ppdu_end.common.tsf_timestamp);
+       status->flag |= RX_FLAG_MACTIME_END;
+}
+
+static void ath10k_htt_rx_h_ppdu(struct ath10k *ar,
+                                struct sk_buff_head *amsdu,
+                                struct ieee80211_rx_status *status)
+{
+       struct sk_buff *first;
+       struct htt_rx_desc *rxd;
+       bool is_first_ppdu;
+       bool is_last_ppdu;
+
+       if (skb_queue_empty(amsdu))
+               return;
+
+       first = skb_peek(amsdu);
+       rxd = (void *)first->data - sizeof(*rxd);
+
+       is_first_ppdu = !!(rxd->attention.flags &
+                          __cpu_to_le32(RX_ATTENTION_FLAGS_FIRST_MPDU));
+       is_last_ppdu = !!(rxd->attention.flags &
+                         __cpu_to_le32(RX_ATTENTION_FLAGS_LAST_MPDU));
+
+       if (is_first_ppdu) {
+               /* New PPDU starts so clear out the old per-PPDU status. */
+               status->freq = 0;
+               status->rate_idx = 0;
+               status->vht_nss = 0;
+               status->vht_flag &= ~RX_VHT_FLAG_80MHZ;
+               status->flag &= ~(RX_FLAG_HT |
+                                 RX_FLAG_VHT |
+                                 RX_FLAG_SHORT_GI |
+                                 RX_FLAG_40MHZ |
+                                 RX_FLAG_MACTIME_END);
+               status->flag |= RX_FLAG_NO_SIGNAL_VAL;
+
+               ath10k_htt_rx_h_signal(ar, status, rxd);
+               ath10k_htt_rx_h_channel(ar, status);
+               ath10k_htt_rx_h_rates(ar, status, rxd);
+       }
+
+       if (is_last_ppdu)
+               ath10k_htt_rx_h_mactime(ar, status, rxd);
+}
+
+static const char * const tid_to_ac[] = {
+       "BE",
+       "BK",
+       "BK",
+       "BE",
+       "VI",
+       "VI",
+       "VO",
+       "VO",
+};
+
+static char *ath10k_get_tid(struct ieee80211_hdr *hdr, char *out, size_t size)
+{
+       u8 *qc;
+       int tid;
+
+       if (!ieee80211_is_data_qos(hdr->frame_control))
+               return "";
+
+       qc = ieee80211_get_qos_ctl(hdr);
+       tid = *qc & IEEE80211_QOS_CTL_TID_MASK;
+       if (tid < 8)
+               snprintf(out, size, "tid %d (%s)", tid, tid_to_ac[tid]);
+       else
+               snprintf(out, size, "tid %d", tid);
+
+       return out;
+}
+
+static void ath10k_process_rx(struct ath10k *ar,
+                             struct ieee80211_rx_status *rx_status,
+                             struct sk_buff *skb)
+{
+       struct ieee80211_rx_status *status;
+       struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
+       char tid[32];
+
+       status = IEEE80211_SKB_RXCB(skb);
+       *status = *rx_status;
+
+       ath10k_dbg(ar, ATH10K_DBG_DATA,
+                  "rx skb %p len %u peer %pM %s %s sn %u %s%s%s%s%s %srate_idx %u vht_nss %u freq %u band %u flag 0x%x fcs-err %i mic-err %i amsdu-more %i\n",
+                  skb,
+                  skb->len,
+                  ieee80211_get_SA(hdr),
+                  ath10k_get_tid(hdr, tid, sizeof(tid)),
+                  is_multicast_ether_addr(ieee80211_get_DA(hdr)) ?
+                                                       "mcast" : "ucast",
+                  (__le16_to_cpu(hdr->seq_ctrl) & IEEE80211_SCTL_SEQ) >> 4,
+                  status->flag == 0 ? "legacy" : "",
+                  status->flag & RX_FLAG_HT ? "ht" : "",
+                  status->flag & RX_FLAG_VHT ? "vht" : "",
+                  status->flag & RX_FLAG_40MHZ ? "40" : "",
+                  status->vht_flag & RX_VHT_FLAG_80MHZ ? "80" : "",
+                  status->flag & RX_FLAG_SHORT_GI ? "sgi " : "",
+                  status->rate_idx,
+                  status->vht_nss,
+                  status->freq,
+                  status->band, status->flag,
+                  !!(status->flag & RX_FLAG_FAILED_FCS_CRC),
+                  !!(status->flag & RX_FLAG_MMIC_ERROR),
+                  !!(status->flag & RX_FLAG_AMSDU_MORE));
+       ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "rx skb: ",
+                       skb->data, skb->len);
+       trace_ath10k_rx_hdr(ar, skb->data, skb->len);
+       trace_ath10k_rx_payload(ar, skb->data, skb->len);
+
+       ieee80211_rx(ar->hw, skb);
+}
+
+static int ath10k_htt_rx_nwifi_hdrlen(struct ieee80211_hdr *hdr)
+{
+       /* nwifi header is padded to 4 bytes. this fixes 4addr rx */
+       return round_up(ieee80211_hdrlen(hdr->frame_control), 4);
+}
+
+static void ath10k_htt_rx_h_undecap_raw(struct ath10k *ar,
+                                       struct sk_buff *msdu,
+                                       struct ieee80211_rx_status *status,
+                                       enum htt_rx_mpdu_encrypt_type enctype,
+                                       bool is_decrypted)
+{
+       struct ieee80211_hdr *hdr;
+       struct htt_rx_desc *rxd;
+       size_t hdr_len;
+       size_t crypto_len;
+       bool is_first;
+       bool is_last;
+
+       rxd = (void *)msdu->data - sizeof(*rxd);
+       is_first = !!(rxd->msdu_end.info0 &
+                     __cpu_to_le32(RX_MSDU_END_INFO0_FIRST_MSDU));
+       is_last = !!(rxd->msdu_end.info0 &
+                    __cpu_to_le32(RX_MSDU_END_INFO0_LAST_MSDU));
+
+       /* Delivered decapped frame:
+        * [802.11 header]
+        * [crypto param] <-- can be trimmed if !fcs_err &&
+        *                    !decrypt_err && !peer_idx_invalid
+        * [amsdu header] <-- only if A-MSDU
+        * [rfc1042/llc]
+        * [payload]
+        * [FCS] <-- at end, needs to be trimmed
+        */
+
+       /* This probably shouldn't happen but warn just in case */
+       if (unlikely(WARN_ON_ONCE(!is_first)))
+               return;
+
+       /* This probably shouldn't happen but warn just in case */
+       if (unlikely(WARN_ON_ONCE(!(is_first && is_last))))
+               return;
+
+       skb_trim(msdu, msdu->len - FCS_LEN);
+
+       /* In most cases this will be true for sniffed frames. It makes sense
+        * to deliver them as-is without stripping the crypto param. This would
+        * also make sense for software based decryption (which is not
+        * implemented in ath10k).
+        *
+        * If there's no error then the frame is decrypted. At least that is
+        * the case for frames that come in via fragmented rx indication.
+        */
+       if (!is_decrypted)
+               return;
+
+       /* The payload is decrypted so strip crypto params. Start from tail
+        * since hdr is used to compute some stuff.
+        */
+
+       hdr = (void *)msdu->data;
+
+       /* Tail */
+       skb_trim(msdu, msdu->len - ath10k_htt_rx_crypto_tail_len(ar, enctype));
+
+       /* MMIC */
+       if (!ieee80211_has_morefrags(hdr->frame_control) &&
+           enctype == HTT_RX_MPDU_ENCRYPT_TKIP_WPA)
+               skb_trim(msdu, msdu->len - 8);
+
+       /* Head */
+       hdr_len = ieee80211_hdrlen(hdr->frame_control);
+       crypto_len = ath10k_htt_rx_crypto_param_len(ar, enctype);
+
+       memmove((void *)msdu->data + crypto_len,
+               (void *)msdu->data, hdr_len);
+       skb_pull(msdu, crypto_len);
+}
+
+static void ath10k_htt_rx_h_undecap_nwifi(struct ath10k *ar,
+                                         struct sk_buff *msdu,
+                                         struct ieee80211_rx_status *status,
+                                         const u8 first_hdr[64])
+{
+       struct ieee80211_hdr *hdr;
+       size_t hdr_len;
+       u8 da[ETH_ALEN];
+       u8 sa[ETH_ALEN];
+
+       /* Delivered decapped frame:
+        * [nwifi 802.11 header] <-- replaced with 802.11 hdr
+        * [rfc1042/llc]
+        *
+        * Note: The nwifi header doesn't have QoS Control and is
+        * (always?) a 3addr frame.
+        *
+        * Note2: There's no A-MSDU subframe header. Even if it's part
+        * of an A-MSDU.
+        */
+
+       /* pull decapped header and copy SA & DA */
+       hdr = (struct ieee80211_hdr *)msdu->data;
+       hdr_len = ath10k_htt_rx_nwifi_hdrlen(hdr);
+       ether_addr_copy(da, ieee80211_get_DA(hdr));
+       ether_addr_copy(sa, ieee80211_get_SA(hdr));
+       skb_pull(msdu, hdr_len);
+
+       /* push original 802.11 header */
+       hdr = (struct ieee80211_hdr *)first_hdr;
+       hdr_len = ieee80211_hdrlen(hdr->frame_control);
+       memcpy(skb_push(msdu, hdr_len), hdr, hdr_len);
+
+       /* original 802.11 header has a different DA and in
+        * case of 4addr it may also have different SA
+        */
+       hdr = (struct ieee80211_hdr *)msdu->data;
+       ether_addr_copy(ieee80211_get_DA(hdr), da);
+       ether_addr_copy(ieee80211_get_SA(hdr), sa);
+}
+
+static void *ath10k_htt_rx_h_find_rfc1042(struct ath10k *ar,
+                                         struct sk_buff *msdu,
+                                         enum htt_rx_mpdu_encrypt_type enctype)
+{
+       struct ieee80211_hdr *hdr;
+       struct htt_rx_desc *rxd;
+       size_t hdr_len, crypto_len;
+       void *rfc1042;
+       bool is_first, is_last, is_amsdu;
+
+       rxd = (void *)msdu->data - sizeof(*rxd);
+       hdr = (void *)rxd->rx_hdr_status;
+
+       is_first = !!(rxd->msdu_end.info0 &
+                     __cpu_to_le32(RX_MSDU_END_INFO0_FIRST_MSDU));
+       is_last = !!(rxd->msdu_end.info0 &
+                    __cpu_to_le32(RX_MSDU_END_INFO0_LAST_MSDU));
+       is_amsdu = !(is_first && is_last);
+
+       rfc1042 = hdr;
+
+       if (is_first) {
+               hdr_len = ieee80211_hdrlen(hdr->frame_control);
+               crypto_len = ath10k_htt_rx_crypto_param_len(ar, enctype);
+
+               rfc1042 += round_up(hdr_len, 4) +
+                          round_up(crypto_len, 4);
+       }
+
+       if (is_amsdu)
+               rfc1042 += sizeof(struct amsdu_subframe_hdr);
+
+       return rfc1042;
+}
+
+static void ath10k_htt_rx_h_undecap_eth(struct ath10k *ar,
+                                       struct sk_buff *msdu,
+                                       struct ieee80211_rx_status *status,
+                                       const u8 first_hdr[64],
+                                       enum htt_rx_mpdu_encrypt_type enctype)
+{
+       struct ieee80211_hdr *hdr;
+       struct ethhdr *eth;
+       size_t hdr_len;
+       void *rfc1042;
+       u8 da[ETH_ALEN];
+       u8 sa[ETH_ALEN];
+
+       /* Delivered decapped frame:
+        * [eth header] <-- replaced with 802.11 hdr & rfc1042/llc
+        * [payload]
+        */
+
+       rfc1042 = ath10k_htt_rx_h_find_rfc1042(ar, msdu, enctype);
+       if (WARN_ON_ONCE(!rfc1042))
+               return;
+
+       /* pull decapped header and copy SA & DA */
+       eth = (struct ethhdr *)msdu->data;
+       ether_addr_copy(da, eth->h_dest);
+       ether_addr_copy(sa, eth->h_source);
+       skb_pull(msdu, sizeof(struct ethhdr));
+
+       /* push rfc1042/llc/snap */
+       memcpy(skb_push(msdu, sizeof(struct rfc1042_hdr)), rfc1042,
+              sizeof(struct rfc1042_hdr));
+
+       /* push original 802.11 header */
+       hdr = (struct ieee80211_hdr *)first_hdr;
+       hdr_len = ieee80211_hdrlen(hdr->frame_control);
+       memcpy(skb_push(msdu, hdr_len), hdr, hdr_len);
+
+       /* original 802.11 header has a different DA and in
+        * case of 4addr it may also have different SA
+        */
+       hdr = (struct ieee80211_hdr *)msdu->data;
+       ether_addr_copy(ieee80211_get_DA(hdr), da);
+       ether_addr_copy(ieee80211_get_SA(hdr), sa);
+}
+
+static void ath10k_htt_rx_h_undecap_snap(struct ath10k *ar,
+                                        struct sk_buff *msdu,
+                                        struct ieee80211_rx_status *status,
+                                        const u8 first_hdr[64])
+{
+       struct ieee80211_hdr *hdr;
+       size_t hdr_len;
+
+       /* Delivered decapped frame:
+        * [amsdu header] <-- replaced with 802.11 hdr
+        * [rfc1042/llc]
+        * [payload]
+        */
+
+       skb_pull(msdu, sizeof(struct amsdu_subframe_hdr));
+
+       hdr = (struct ieee80211_hdr *)first_hdr;
+       hdr_len = ieee80211_hdrlen(hdr->frame_control);
+       memcpy(skb_push(msdu, hdr_len), hdr, hdr_len);
+}
+
+static void ath10k_htt_rx_h_undecap(struct ath10k *ar,
+                                   struct sk_buff *msdu,
+                                   struct ieee80211_rx_status *status,
+                                   u8 first_hdr[64],
+                                   enum htt_rx_mpdu_encrypt_type enctype,
+                                   bool is_decrypted)
+{
+       struct htt_rx_desc *rxd;
+       enum rx_msdu_decap_format decap;
+       struct ieee80211_hdr *hdr;
+
+       /* First msdu's decapped header:
+        * [802.11 header] <-- padded to 4 bytes long
+        * [crypto param] <-- padded to 4 bytes long
+        * [amsdu header] <-- only if A-MSDU
+        * [rfc1042/llc]
+        *
+        * Other (2nd, 3rd, ..) msdu's decapped header:
+        * [amsdu header] <-- only if A-MSDU
+        * [rfc1042/llc]
+        */
+
+       rxd = (void *)msdu->data - sizeof(*rxd);
+       hdr = (void *)rxd->rx_hdr_status;
+       decap = MS(__le32_to_cpu(rxd->msdu_start.info1),
+                  RX_MSDU_START_INFO1_DECAP_FORMAT);
+
+       switch (decap) {
+       case RX_MSDU_DECAP_RAW:
+               ath10k_htt_rx_h_undecap_raw(ar, msdu, status, enctype,
+                                           is_decrypted);
+               break;
+       case RX_MSDU_DECAP_NATIVE_WIFI:
+               ath10k_htt_rx_h_undecap_nwifi(ar, msdu, status, first_hdr);
+               break;
+       case RX_MSDU_DECAP_ETHERNET2_DIX:
+               ath10k_htt_rx_h_undecap_eth(ar, msdu, status, first_hdr, enctype);
+               break;
+       case RX_MSDU_DECAP_8023_SNAP_LLC:
+               ath10k_htt_rx_h_undecap_snap(ar, msdu, status, first_hdr);
+               break;
+       }
+}
+
+static int ath10k_htt_rx_get_csum_state(struct sk_buff *skb)
+{
+       struct htt_rx_desc *rxd;
+       u32 flags, info;
+       bool is_ip4, is_ip6;
+       bool is_tcp, is_udp;
+       bool ip_csum_ok, tcpudp_csum_ok;
+
+       rxd = (void *)skb->data - sizeof(*rxd);
+       flags = __le32_to_cpu(rxd->attention.flags);
+       info = __le32_to_cpu(rxd->msdu_start.info1);
+
+       is_ip4 = !!(info & RX_MSDU_START_INFO1_IPV4_PROTO);
+       is_ip6 = !!(info & RX_MSDU_START_INFO1_IPV6_PROTO);
+       is_tcp = !!(info & RX_MSDU_START_INFO1_TCP_PROTO);
+       is_udp = !!(info & RX_MSDU_START_INFO1_UDP_PROTO);
+       ip_csum_ok = !(flags & RX_ATTENTION_FLAGS_IP_CHKSUM_FAIL);
+       tcpudp_csum_ok = !(flags & RX_ATTENTION_FLAGS_TCP_UDP_CHKSUM_FAIL);
+
+       if (!is_ip4 && !is_ip6)
+               return CHECKSUM_NONE;
+       if (!is_tcp && !is_udp)
+               return CHECKSUM_NONE;
+       if (!ip_csum_ok)
+               return CHECKSUM_NONE;
+       if (!tcpudp_csum_ok)
+               return CHECKSUM_NONE;
+
+       return CHECKSUM_UNNECESSARY;
+}
+
+static void ath10k_htt_rx_h_csum_offload(struct sk_buff *msdu)
+{
+       msdu->ip_summed = ath10k_htt_rx_get_csum_state(msdu);
+}
+
+static void ath10k_htt_rx_h_mpdu(struct ath10k *ar,
+                                struct sk_buff_head *amsdu,
+                                struct ieee80211_rx_status *status)
+{
+       struct sk_buff *first;
+       struct sk_buff *last;
+       struct sk_buff *msdu;
+       struct htt_rx_desc *rxd;
+       struct ieee80211_hdr *hdr;
+       enum htt_rx_mpdu_encrypt_type enctype;
+       u8 first_hdr[64];
+       u8 *qos;
+       size_t hdr_len;
+       bool has_fcs_err;
+       bool has_crypto_err;
+       bool has_tkip_err;
+       bool has_peer_idx_invalid;
+       bool is_decrypted;
+       u32 attention;
+
+       if (skb_queue_empty(amsdu))
+               return;
+
+       first = skb_peek(amsdu);
+       rxd = (void *)first->data - sizeof(*rxd);
+
+       enctype = MS(__le32_to_cpu(rxd->mpdu_start.info0),
+                    RX_MPDU_START_INFO0_ENCRYPT_TYPE);
+
+       /* First MSDU's Rx descriptor in an A-MSDU contains full 802.11
+        * decapped header. It'll be used for undecapping of each MSDU.
+        */
+       hdr = (void *)rxd->rx_hdr_status;
+       hdr_len = ieee80211_hdrlen(hdr->frame_control);
+       memcpy(first_hdr, hdr, hdr_len);
+
+       /* Each A-MSDU subframe will use the original header as the base and be
+        * reported as a separate MSDU so strip the A-MSDU bit from QoS Ctl.
+        */
+       hdr = (void *)first_hdr;
+       qos = ieee80211_get_qos_ctl(hdr);
+       qos[0] &= ~IEEE80211_QOS_CTL_A_MSDU_PRESENT;
+
+       /* Some attention flags are valid only in the last MSDU. */
+       last = skb_peek_tail(amsdu);
+       rxd = (void *)last->data - sizeof(*rxd);
+       attention = __le32_to_cpu(rxd->attention.flags);
+
+       has_fcs_err = !!(attention & RX_ATTENTION_FLAGS_FCS_ERR);
+       has_crypto_err = !!(attention & RX_ATTENTION_FLAGS_DECRYPT_ERR);
+       has_tkip_err = !!(attention & RX_ATTENTION_FLAGS_TKIP_MIC_ERR);
+       has_peer_idx_invalid = !!(attention & RX_ATTENTION_FLAGS_PEER_IDX_INVALID);
+
+       /* Note: If hardware captures an encrypted frame that it can't decrypt,
+        * e.g. due to fcs error, missing peer or invalid key data it will
+        * report the frame as raw.
+        */
+       is_decrypted = (enctype != HTT_RX_MPDU_ENCRYPT_NONE &&
+                       !has_fcs_err &&
+                       !has_crypto_err &&
+                       !has_peer_idx_invalid);
+
+       /* Clear per-MPDU flags while leaving per-PPDU flags intact. */
+       status->flag &= ~(RX_FLAG_FAILED_FCS_CRC |
+                         RX_FLAG_MMIC_ERROR |
+                         RX_FLAG_DECRYPTED |
+                         RX_FLAG_IV_STRIPPED |
+                         RX_FLAG_MMIC_STRIPPED);
+
+       if (has_fcs_err)
+               status->flag |= RX_FLAG_FAILED_FCS_CRC;
+
+       if (has_tkip_err)
+               status->flag |= RX_FLAG_MMIC_ERROR;
+
+       if (is_decrypted)
+               status->flag |= RX_FLAG_DECRYPTED |
+                               RX_FLAG_IV_STRIPPED |
+                               RX_FLAG_MMIC_STRIPPED;
+
+       skb_queue_walk(amsdu, msdu) {
+               ath10k_htt_rx_h_csum_offload(msdu);
+               ath10k_htt_rx_h_undecap(ar, msdu, status, first_hdr, enctype,
+                                       is_decrypted);
+
+               /* Undecapping involves copying the original 802.11 header back
+                * to sk_buff. If frame is protected and hardware has decrypted
+                * it then remove the protected bit.
+                */
+               if (!is_decrypted)
+                       continue;
+
+               hdr = (void *)msdu->data;
+               hdr->frame_control &= ~__cpu_to_le16(IEEE80211_FCTL_PROTECTED);
+       }
+}
+
+static void ath10k_htt_rx_h_deliver(struct ath10k *ar,
+                                   struct sk_buff_head *amsdu,
+                                   struct ieee80211_rx_status *status)
+{
+       struct sk_buff *msdu;
+
+       while ((msdu = __skb_dequeue(amsdu))) {
+               /* Setup per-MSDU flags */
+               if (skb_queue_empty(amsdu))
+                       status->flag &= ~RX_FLAG_AMSDU_MORE;
+               else
+                       status->flag |= RX_FLAG_AMSDU_MORE;
+
+               ath10k_process_rx(ar, status, msdu);
+       }
+}
+
+static int ath10k_unchain_msdu(struct sk_buff_head *amsdu)
+{
+       struct sk_buff *skb, *first;
+       int space;
+       int total_len = 0;
+
+       /* TODO:  Might could optimize this by using
+        * skb_try_coalesce or similar method to
+        * decrease copying, or maybe get mac80211 to
+        * provide a way to just receive a list of
+        * skb?
+        */
+
+       first = __skb_dequeue(amsdu);
+
+       /* Allocate total length all at once. */
+       skb_queue_walk(amsdu, skb)
+               total_len += skb->len;
+
+       space = total_len - skb_tailroom(first);
+       if ((space > 0) &&
+           (pskb_expand_head(first, 0, space, GFP_ATOMIC) < 0)) {
+               /* TODO:  bump some rx-oom error stat */
+               /* put it back together so we can free the
+                * whole list at once.
+                */
+               __skb_queue_head(amsdu, first);
+               return -1;
+       }
+
+       /* Walk list again, copying contents into
+        * msdu_head
+        */
+       while ((skb = __skb_dequeue(amsdu))) {
+               skb_copy_from_linear_data(skb, skb_put(first, skb->len),
+                                         skb->len);
+               dev_kfree_skb_any(skb);
+       }
+
+       __skb_queue_head(amsdu, first);
+       return 0;
+}
+
+static void ath10k_htt_rx_h_unchain(struct ath10k *ar,
+                                   struct sk_buff_head *amsdu,
+                                   bool chained)
+{
+       struct sk_buff *first;
+       struct htt_rx_desc *rxd;
+       enum rx_msdu_decap_format decap;
+
+       first = skb_peek(amsdu);
+       rxd = (void *)first->data - sizeof(*rxd);
+       decap = MS(__le32_to_cpu(rxd->msdu_start.info1),
+                  RX_MSDU_START_INFO1_DECAP_FORMAT);
+
+       if (!chained)
+               return;
+
+       /* FIXME: Current unchaining logic can only handle simple case of raw
+        * msdu chaining. If decapping is other than raw the chaining may be
+        * more complex and this isn't handled by the current code. Don't even
+        * try re-constructing such frames - it'll be pretty much garbage.
+        */
+       if (decap != RX_MSDU_DECAP_RAW ||
+           skb_queue_len(amsdu) != 1 + rxd->frag_info.ring2_more_count) {
+               __skb_queue_purge(amsdu);
+               return;
+       }
+
+       ath10k_unchain_msdu(amsdu);
+}
+
+static bool ath10k_htt_rx_amsdu_allowed(struct ath10k *ar,
+                                       struct sk_buff_head *amsdu,
+                                       struct ieee80211_rx_status *rx_status)
+{
+       struct sk_buff *msdu;
+       struct htt_rx_desc *rxd;
+       bool is_mgmt;
+       bool has_fcs_err;
+
+       msdu = skb_peek(amsdu);
+       rxd = (void *)msdu->data - sizeof(*rxd);
+
+       /* FIXME: It might be a good idea to do some fuzzy-testing to drop
+        * invalid/dangerous frames.
+        */
+
+       if (!rx_status->freq) {
+               ath10k_warn(ar, "no channel configured; ignoring frame(s)!\n");
+               return false;
+       }
+
+       is_mgmt = !!(rxd->attention.flags &
+                    __cpu_to_le32(RX_ATTENTION_FLAGS_MGMT_TYPE));
+       has_fcs_err = !!(rxd->attention.flags &
+                        __cpu_to_le32(RX_ATTENTION_FLAGS_FCS_ERR));
+
+       /* Management frames are handled via WMI events. The pros of such
+        * approach is that channel is explicitly provided in WMI events
+        * whereas HTT doesn't provide channel information for Rxed frames.
+        *
+        * However some firmware revisions don't report corrupted frames via
+        * WMI so don't drop them.
+        */
+       if (is_mgmt && !has_fcs_err) {
+               ath10k_dbg(ar, ATH10K_DBG_HTT, "htt rx mgmt ctrl\n");
+               return false;
+       }
+
+       if (test_bit(ATH10K_CAC_RUNNING, &ar->dev_flags)) {
+               ath10k_dbg(ar, ATH10K_DBG_HTT, "htt rx cac running\n");
+               return false;
+       }
+
+       return true;
+}
+
+static void ath10k_htt_rx_h_filter(struct ath10k *ar,
+                                  struct sk_buff_head *amsdu,
+                                  struct ieee80211_rx_status *rx_status)
+{
+       if (skb_queue_empty(amsdu))
+               return;
+
+       if (ath10k_htt_rx_amsdu_allowed(ar, amsdu, rx_status))
+               return;
+
+       __skb_queue_purge(amsdu);
+}
+
+static void ath10k_htt_rx_handler(struct ath10k_htt *htt,
+                                 struct htt_rx_indication *rx)
+{
+       struct ath10k *ar = htt->ar;
+       struct ieee80211_rx_status *rx_status = &htt->rx_status;
+       struct htt_rx_indication_mpdu_range *mpdu_ranges;
+       struct sk_buff_head amsdu;
+       int num_mpdu_ranges;
+       int fw_desc_len;
+       u8 *fw_desc;
+       int i, ret, mpdu_count = 0;
+
+       lockdep_assert_held(&htt->rx_ring.lock);
+
+       if (htt->rx_confused)
+               return;
+
+       fw_desc_len = __le16_to_cpu(rx->prefix.fw_rx_desc_bytes);
+       fw_desc = (u8 *)&rx->fw_desc;
+
+       num_mpdu_ranges = MS(__le32_to_cpu(rx->hdr.info1),
+                            HTT_RX_INDICATION_INFO1_NUM_MPDU_RANGES);
+       mpdu_ranges = htt_rx_ind_get_mpdu_ranges(rx);
+
+       ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt rx ind: ",
+                       rx, sizeof(*rx) +
+                       (sizeof(struct htt_rx_indication_mpdu_range) *
+                               num_mpdu_ranges));
+
+       for (i = 0; i < num_mpdu_ranges; i++)
+               mpdu_count += mpdu_ranges[i].mpdu_count;
+
+       while (mpdu_count--) {
+               __skb_queue_head_init(&amsdu);
+               ret = ath10k_htt_rx_amsdu_pop(htt, &fw_desc,
+                                             &fw_desc_len, &amsdu);
+               if (ret < 0) {
+                       ath10k_warn(ar, "rx ring became corrupted: %d\n", ret);
+                       __skb_queue_purge(&amsdu);
+                       /* FIXME: It's probably a good idea to reboot the
+                        * device instead of leaving it inoperable.
+                        */
+                       htt->rx_confused = true;
+                       break;
+               }
+
+               ath10k_htt_rx_h_ppdu(ar, &amsdu, rx_status);
+               ath10k_htt_rx_h_unchain(ar, &amsdu, ret > 0);
+               ath10k_htt_rx_h_filter(ar, &amsdu, rx_status);
+               ath10k_htt_rx_h_mpdu(ar, &amsdu, rx_status);
+               ath10k_htt_rx_h_deliver(ar, &amsdu, rx_status);
+       }
+
+       tasklet_schedule(&htt->rx_replenish_task);
+}
+
+static void ath10k_htt_rx_frag_handler(struct ath10k_htt *htt,
+                                      struct htt_rx_fragment_indication *frag)
+{
+       struct ath10k *ar = htt->ar;
+       struct ieee80211_rx_status *rx_status = &htt->rx_status;
+       struct sk_buff_head amsdu;
+       int ret;
+       u8 *fw_desc;
+       int fw_desc_len;
+
+       fw_desc_len = __le16_to_cpu(frag->fw_rx_desc_bytes);
+       fw_desc = (u8 *)frag->fw_msdu_rx_desc;
+
+       __skb_queue_head_init(&amsdu);
+
+       spin_lock_bh(&htt->rx_ring.lock);
+       ret = ath10k_htt_rx_amsdu_pop(htt, &fw_desc, &fw_desc_len,
+                                     &amsdu);
+       spin_unlock_bh(&htt->rx_ring.lock);
+
+       tasklet_schedule(&htt->rx_replenish_task);
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT_DUMP, "htt rx frag ahead\n");
+
+       if (ret) {
+               ath10k_warn(ar, "failed to pop amsdu from httr rx ring for fragmented rx %d\n",
+                           ret);
+               __skb_queue_purge(&amsdu);
+               return;
+       }
+
+       if (skb_queue_len(&amsdu) != 1) {
+               ath10k_warn(ar, "failed to pop frag amsdu: too many msdus\n");
+               __skb_queue_purge(&amsdu);
+               return;
+       }
+
+       ath10k_htt_rx_h_ppdu(ar, &amsdu, rx_status);
+       ath10k_htt_rx_h_filter(ar, &amsdu, rx_status);
+       ath10k_htt_rx_h_mpdu(ar, &amsdu, rx_status);
+       ath10k_htt_rx_h_deliver(ar, &amsdu, rx_status);
+
+       if (fw_desc_len > 0) {
+               ath10k_dbg(ar, ATH10K_DBG_HTT,
+                          "expecting more fragmented rx in one indication %d\n",
+                          fw_desc_len);
+       }
+}
+
+static void ath10k_htt_rx_frm_tx_compl(struct ath10k *ar,
+                                      struct sk_buff *skb)
+{
+       struct ath10k_htt *htt = &ar->htt;
+       struct htt_resp *resp = (struct htt_resp *)skb->data;
+       struct htt_tx_done tx_done = {};
+       int status = MS(resp->data_tx_completion.flags, HTT_DATA_TX_STATUS);
+       __le16 msdu_id;
+       int i;
+
+       lockdep_assert_held(&htt->tx_lock);
+
+       switch (status) {
+       case HTT_DATA_TX_STATUS_NO_ACK:
+               tx_done.no_ack = true;
+               break;
+       case HTT_DATA_TX_STATUS_OK:
+               break;
+       case HTT_DATA_TX_STATUS_DISCARD:
+       case HTT_DATA_TX_STATUS_POSTPONE:
+       case HTT_DATA_TX_STATUS_DOWNLOAD_FAIL:
+               tx_done.discard = true;
+               break;
+       default:
+               ath10k_warn(ar, "unhandled tx completion status %d\n", status);
+               tx_done.discard = true;
+               break;
+       }
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT, "htt tx completion num_msdus %d\n",
+                  resp->data_tx_completion.num_msdus);
+
+       for (i = 0; i < resp->data_tx_completion.num_msdus; i++) {
+               msdu_id = resp->data_tx_completion.msdus[i];
+               tx_done.msdu_id = __le16_to_cpu(msdu_id);
+               ath10k_txrx_tx_unref(htt, &tx_done);
+       }
+}
+
+static void ath10k_htt_rx_addba(struct ath10k *ar, struct htt_resp *resp)
+{
+       struct htt_rx_addba *ev = &resp->rx_addba;
+       struct ath10k_peer *peer;
+       struct ath10k_vif *arvif;
+       u16 info0, tid, peer_id;
+
+       info0 = __le16_to_cpu(ev->info0);
+       tid = MS(info0, HTT_RX_BA_INFO0_TID);
+       peer_id = MS(info0, HTT_RX_BA_INFO0_PEER_ID);
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT,
+                  "htt rx addba tid %hu peer_id %hu size %hhu\n",
+                  tid, peer_id, ev->window_size);
+
+       spin_lock_bh(&ar->data_lock);
+       peer = ath10k_peer_find_by_id(ar, peer_id);
+       if (!peer) {
+               ath10k_warn(ar, "received addba event for invalid peer_id: %hu\n",
+                           peer_id);
+               spin_unlock_bh(&ar->data_lock);
+               return;
+       }
+
+       arvif = ath10k_get_arvif(ar, peer->vdev_id);
+       if (!arvif) {
+               ath10k_warn(ar, "received addba event for invalid vdev_id: %u\n",
+                           peer->vdev_id);
+               spin_unlock_bh(&ar->data_lock);
+               return;
+       }
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT,
+                  "htt rx start rx ba session sta %pM tid %hu size %hhu\n",
+                  peer->addr, tid, ev->window_size);
+
+       ieee80211_start_rx_ba_session_offl(arvif->vif, peer->addr, tid);
+       spin_unlock_bh(&ar->data_lock);
+}
+
+static void ath10k_htt_rx_delba(struct ath10k *ar, struct htt_resp *resp)
+{
+       struct htt_rx_delba *ev = &resp->rx_delba;
+       struct ath10k_peer *peer;
+       struct ath10k_vif *arvif;
+       u16 info0, tid, peer_id;
+
+       info0 = __le16_to_cpu(ev->info0);
+       tid = MS(info0, HTT_RX_BA_INFO0_TID);
+       peer_id = MS(info0, HTT_RX_BA_INFO0_PEER_ID);
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT,
+                  "htt rx delba tid %hu peer_id %hu\n",
+                  tid, peer_id);
+
+       spin_lock_bh(&ar->data_lock);
+       peer = ath10k_peer_find_by_id(ar, peer_id);
+       if (!peer) {
+               ath10k_warn(ar, "received addba event for invalid peer_id: %hu\n",
+                           peer_id);
+               spin_unlock_bh(&ar->data_lock);
+               return;
+       }
+
+       arvif = ath10k_get_arvif(ar, peer->vdev_id);
+       if (!arvif) {
+               ath10k_warn(ar, "received addba event for invalid vdev_id: %u\n",
+                           peer->vdev_id);
+               spin_unlock_bh(&ar->data_lock);
+               return;
+       }
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT,
+                  "htt rx stop rx ba session sta %pM tid %hu\n",
+                  peer->addr, tid);
+
+       ieee80211_stop_rx_ba_session_offl(arvif->vif, peer->addr, tid);
+       spin_unlock_bh(&ar->data_lock);
+}
+
+static int ath10k_htt_rx_extract_amsdu(struct sk_buff_head *list,
+                                      struct sk_buff_head *amsdu)
+{
+       struct sk_buff *msdu;
+       struct htt_rx_desc *rxd;
+
+       if (skb_queue_empty(list))
+               return -ENOBUFS;
+
+       if (WARN_ON(!skb_queue_empty(amsdu)))
+               return -EINVAL;
+
+       while ((msdu = __skb_dequeue(list))) {
+               __skb_queue_tail(amsdu, msdu);
+
+               rxd = (void *)msdu->data - sizeof(*rxd);
+               if (rxd->msdu_end.info0 &
+                   __cpu_to_le32(RX_MSDU_END_INFO0_LAST_MSDU))
+                       break;
+       }
+
+       msdu = skb_peek_tail(amsdu);
+       rxd = (void *)msdu->data - sizeof(*rxd);
+       if (!(rxd->msdu_end.info0 &
+             __cpu_to_le32(RX_MSDU_END_INFO0_LAST_MSDU))) {
+               skb_queue_splice_init(amsdu, list);
+               return -EAGAIN;
+       }
+
+       return 0;
+}
+
+static void ath10k_htt_rx_h_rx_offload_prot(struct ieee80211_rx_status *status,
+                                           struct sk_buff *skb)
+{
+       struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
+
+       if (!ieee80211_has_protected(hdr->frame_control))
+               return;
+
+       /* Offloaded frames are already decrypted but firmware insists they are
+        * protected in the 802.11 header. Strip the flag.  Otherwise mac80211
+        * will drop the frame.
+        */
+
+       hdr->frame_control &= ~__cpu_to_le16(IEEE80211_FCTL_PROTECTED);
+       status->flag |= RX_FLAG_DECRYPTED |
+                       RX_FLAG_IV_STRIPPED |
+                       RX_FLAG_MMIC_STRIPPED;
+}
+
+static void ath10k_htt_rx_h_rx_offload(struct ath10k *ar,
+                                      struct sk_buff_head *list)
+{
+       struct ath10k_htt *htt = &ar->htt;
+       struct ieee80211_rx_status *status = &htt->rx_status;
+       struct htt_rx_offload_msdu *rx;
+       struct sk_buff *msdu;
+       size_t offset;
+
+       while ((msdu = __skb_dequeue(list))) {
+               /* Offloaded frames don't have Rx descriptor. Instead they have
+                * a short meta information header.
+                */
+
+               rx = (void *)msdu->data;
+
+               skb_put(msdu, sizeof(*rx));
+               skb_pull(msdu, sizeof(*rx));
+
+               if (skb_tailroom(msdu) < __le16_to_cpu(rx->msdu_len)) {
+                       ath10k_warn(ar, "dropping frame: offloaded rx msdu is too long!\n");
+                       dev_kfree_skb_any(msdu);
+                       continue;
+               }
+
+               skb_put(msdu, __le16_to_cpu(rx->msdu_len));
+
+               /* Offloaded rx header length isn't multiple of 2 nor 4 so the
+                * actual payload is unaligned. Align the frame.  Otherwise
+                * mac80211 complains.  This shouldn't reduce performance much
+                * because these offloaded frames are rare.
+                */
+               offset = 4 - ((unsigned long)msdu->data & 3);
+               skb_put(msdu, offset);
+               memmove(msdu->data + offset, msdu->data, msdu->len);
+               skb_pull(msdu, offset);
+
+               /* FIXME: The frame is NWifi. Re-construct QoS Control
+                * if possible later.
+                */
+
+               memset(status, 0, sizeof(*status));
+               status->flag |= RX_FLAG_NO_SIGNAL_VAL;
+
+               ath10k_htt_rx_h_rx_offload_prot(status, msdu);
+               ath10k_htt_rx_h_channel(ar, status);
+               ath10k_process_rx(ar, status, msdu);
+       }
+}
+
+static void ath10k_htt_rx_in_ord_ind(struct ath10k *ar, struct sk_buff *skb)
+{
+       struct ath10k_htt *htt = &ar->htt;
+       struct htt_resp *resp = (void *)skb->data;
+       struct ieee80211_rx_status *status = &htt->rx_status;
+       struct sk_buff_head list;
+       struct sk_buff_head amsdu;
+       u16 peer_id;
+       u16 msdu_count;
+       u8 vdev_id;
+       u8 tid;
+       bool offload;
+       bool frag;
+       int ret;
+
+       lockdep_assert_held(&htt->rx_ring.lock);
+
+       if (htt->rx_confused)
+               return;
+
+       skb_pull(skb, sizeof(resp->hdr));
+       skb_pull(skb, sizeof(resp->rx_in_ord_ind));
+
+       peer_id = __le16_to_cpu(resp->rx_in_ord_ind.peer_id);
+       msdu_count = __le16_to_cpu(resp->rx_in_ord_ind.msdu_count);
+       vdev_id = resp->rx_in_ord_ind.vdev_id;
+       tid = SM(resp->rx_in_ord_ind.info, HTT_RX_IN_ORD_IND_INFO_TID);
+       offload = !!(resp->rx_in_ord_ind.info &
+                       HTT_RX_IN_ORD_IND_INFO_OFFLOAD_MASK);
+       frag = !!(resp->rx_in_ord_ind.info & HTT_RX_IN_ORD_IND_INFO_FRAG_MASK);
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT,
+                  "htt rx in ord vdev %i peer %i tid %i offload %i frag %i msdu count %i\n",
+                  vdev_id, peer_id, tid, offload, frag, msdu_count);
+
+       if (skb->len < msdu_count * sizeof(*resp->rx_in_ord_ind.msdu_descs)) {
+               ath10k_warn(ar, "dropping invalid in order rx indication\n");
+               return;
+       }
+
+       /* The event can deliver more than 1 A-MSDU. Each A-MSDU is later
+        * extracted and processed.
+        */
+       __skb_queue_head_init(&list);
+       ret = ath10k_htt_rx_pop_paddr_list(htt, &resp->rx_in_ord_ind, &list);
+       if (ret < 0) {
+               ath10k_warn(ar, "failed to pop paddr list: %d\n", ret);
+               htt->rx_confused = true;
+               return;
+       }
+
+       /* Offloaded frames are very different and need to be handled
+        * separately.
+        */
+       if (offload)
+               ath10k_htt_rx_h_rx_offload(ar, &list);
+
+       while (!skb_queue_empty(&list)) {
+               __skb_queue_head_init(&amsdu);
+               ret = ath10k_htt_rx_extract_amsdu(&list, &amsdu);
+               switch (ret) {
+               case 0:
+                       /* Note: The in-order indication may report interleaved
+                        * frames from different PPDUs meaning reported rx rate
+                        * to mac80211 isn't accurate/reliable. It's still
+                        * better to report something than nothing though. This
+                        * should still give an idea about rx rate to the user.
+                        */
+                       ath10k_htt_rx_h_ppdu(ar, &amsdu, status);
+                       ath10k_htt_rx_h_filter(ar, &amsdu, status);
+                       ath10k_htt_rx_h_mpdu(ar, &amsdu, status);
+                       ath10k_htt_rx_h_deliver(ar, &amsdu, status);
+                       break;
+               case -EAGAIN:
+                       /* fall through */
+               default:
+                       /* Should not happen. */
+                       ath10k_warn(ar, "failed to extract amsdu: %d\n", ret);
+                       htt->rx_confused = true;
+                       __skb_queue_purge(&list);
+                       return;
+               }
+       }
+
+       tasklet_schedule(&htt->rx_replenish_task);
+}
+
+void ath10k_htt_t2h_msg_handler(struct ath10k *ar, struct sk_buff *skb)
+{
+       struct ath10k_htt *htt = &ar->htt;
+       struct htt_resp *resp = (struct htt_resp *)skb->data;
+
+       /* confirm alignment */
+       if (!IS_ALIGNED((unsigned long)skb->data, 4))
+               ath10k_warn(ar, "unaligned htt message, expect trouble\n");
+
+       ath10k_dbg(ar, ATH10K_DBG_HTT, "htt rx, msg_type: 0x%0X\n",
+                  resp->hdr.msg_type);
+       switch (resp->hdr.msg_type) {
+       case HTT_T2H_MSG_TYPE_VERSION_CONF: {
+               htt->target_version_major = resp->ver_resp.major;
+               htt->target_version_minor = resp->ver_resp.minor;
+               complete(&htt->target_version_received);
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_RX_IND:
+               spin_lock_bh(&htt->rx_ring.lock);
+               __skb_queue_tail(&htt->rx_compl_q, skb);
+               spin_unlock_bh(&htt->rx_ring.lock);
+               tasklet_schedule(&htt->txrx_compl_task);
+               return;
+       case HTT_T2H_MSG_TYPE_PEER_MAP: {
+               struct htt_peer_map_event ev = {
+                       .vdev_id = resp->peer_map.vdev_id,
+                       .peer_id = __le16_to_cpu(resp->peer_map.peer_id),
+               };
+               memcpy(ev.addr, resp->peer_map.addr, sizeof(ev.addr));
+               ath10k_peer_map_event(htt, &ev);
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_PEER_UNMAP: {
+               struct htt_peer_unmap_event ev = {
+                       .peer_id = __le16_to_cpu(resp->peer_unmap.peer_id),
+               };
+               ath10k_peer_unmap_event(htt, &ev);
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_MGMT_TX_COMPLETION: {
+               struct htt_tx_done tx_done = {};
+               int status = __le32_to_cpu(resp->mgmt_tx_completion.status);
+
+               tx_done.msdu_id =
+                       __le32_to_cpu(resp->mgmt_tx_completion.desc_id);
+
+               switch (status) {
+               case HTT_MGMT_TX_STATUS_OK:
+                       break;
+               case HTT_MGMT_TX_STATUS_RETRY:
+                       tx_done.no_ack = true;
+                       break;
+               case HTT_MGMT_TX_STATUS_DROP:
+                       tx_done.discard = true;
+                       break;
+               }
+
+               spin_lock_bh(&htt->tx_lock);
+               ath10k_txrx_tx_unref(htt, &tx_done);
+               spin_unlock_bh(&htt->tx_lock);
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_TX_COMPL_IND:
+               spin_lock_bh(&htt->tx_lock);
+               __skb_queue_tail(&htt->tx_compl_q, skb);
+               spin_unlock_bh(&htt->tx_lock);
+               tasklet_schedule(&htt->txrx_compl_task);
+               return;
+       case HTT_T2H_MSG_TYPE_SEC_IND: {
+               struct ath10k *ar = htt->ar;
+               struct htt_security_indication *ev = &resp->security_indication;
+
+               ath10k_dbg(ar, ATH10K_DBG_HTT,
+                          "sec ind peer_id %d unicast %d type %d\n",
+                         __le16_to_cpu(ev->peer_id),
+                         !!(ev->flags & HTT_SECURITY_IS_UNICAST),
+                         MS(ev->flags, HTT_SECURITY_TYPE));
+               complete(&ar->install_key_done);
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_RX_FRAG_IND: {
+               ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt event: ",
+                               skb->data, skb->len);
+               ath10k_htt_rx_frag_handler(htt, &resp->rx_frag_ind);
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_TEST:
+               /* FIX THIS */
+               break;
+       case HTT_T2H_MSG_TYPE_STATS_CONF:
+               trace_ath10k_htt_stats(ar, skb->data, skb->len);
+               break;
+       case HTT_T2H_MSG_TYPE_TX_INSPECT_IND:
+               /* Firmware can return tx frames if it's unable to fully
+                * process them and suspects host may be able to fix it. ath10k
+                * sends all tx frames as already inspected so this shouldn't
+                * happen unless fw has a bug.
+                */
+               ath10k_warn(ar, "received an unexpected htt tx inspect event\n");
+               break;
+       case HTT_T2H_MSG_TYPE_RX_ADDBA:
+               ath10k_htt_rx_addba(ar, resp);
+               break;
+       case HTT_T2H_MSG_TYPE_RX_DELBA:
+               ath10k_htt_rx_delba(ar, resp);
+               break;
+       case HTT_T2H_MSG_TYPE_PKTLOG: {
+               struct ath10k_pktlog_hdr *hdr =
+                       (struct ath10k_pktlog_hdr *)resp->pktlog_msg.payload;
+
+               trace_ath10k_htt_pktlog(ar, resp->pktlog_msg.payload,
+                                       sizeof(*hdr) +
+                                       __le16_to_cpu(hdr->size));
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_RX_FLUSH: {
+               /* Ignore this event because mac80211 takes care of Rx
+                * aggregation reordering.
+                */
+               break;
+       }
+       case HTT_T2H_MSG_TYPE_RX_IN_ORD_PADDR_IND: {
+               spin_lock_bh(&htt->rx_ring.lock);
+               __skb_queue_tail(&htt->rx_in_ord_compl_q, skb);
+               spin_unlock_bh(&htt->rx_ring.lock);
+               tasklet_schedule(&htt->txrx_compl_task);
+               return;
+       }
+       case HTT_T2H_MSG_TYPE_TX_CREDIT_UPDATE_IND:
+               /* FIXME: This WMI-TLV event is overlapping with 10.2
+                * CHAN_CHANGE - both being 0xF. Neither is being used in
+                * practice so no immediate action is necessary. Nevertheless
+                * HTT may need an abstraction layer like WMI has one day.
+                */
+               break;
+       default:
+               ath10k_warn(ar, "htt event (%d) not handled\n",
+                           resp->hdr.msg_type);
+               ath10k_dbg_dump(ar, ATH10K_DBG_HTT_DUMP, NULL, "htt event: ",
+                               skb->data, skb->len);
+               break;
+       };
+
+       /* Free the indication buffer */
+       dev_kfree_skb_any(skb);
+}
+
+static void ath10k_htt_txrx_compl_task(unsigned long ptr)
+{
+       struct ath10k_htt *htt = (struct ath10k_htt *)ptr;
+       struct ath10k *ar = htt->ar;
+       struct htt_resp *resp;
+       struct sk_buff *skb;
+
+       spin_lock_bh(&htt->tx_lock);
+       while ((skb = __skb_dequeue(&htt->tx_compl_q))) {
+               ath10k_htt_rx_frm_tx_compl(htt->ar, skb);
+               dev_kfree_skb_any(skb);
+       }
+       spin_unlock_bh(&htt->tx_lock);
+
+       spin_lock_bh(&htt->rx_ring.lock);
+       while ((skb = __skb_dequeue(&htt->rx_compl_q))) {
+               resp = (struct htt_resp *)skb->data;
+               ath10k_htt_rx_handler(htt, &resp->rx_ind);
+               dev_kfree_skb_any(skb);
+       }
+
+       while ((skb = __skb_dequeue(&htt->rx_in_ord_compl_q))) {
+               ath10k_htt_rx_in_ord_ind(ar, skb);
+               dev_kfree_skb_any(skb);
+       }
+       spin_unlock_bh(&htt->rx_ring.lock);
+}