diff mbox series

[RFC,v1,218/256] cl8k: add tx/tx_queue.c

Message ID 20210617160223.160998-219-viktor.barna@celeno.com (mailing list archive)
State RFC
Delegated to: Kalle Valo
Headers show
Series wireless: cl8k driver for Celeno IEEE 802.11ax devices | expand

Commit Message

Viktor Barna June 17, 2021, 4:01 p.m. UTC
From: Viktor Barna <viktor.barna@celeno.com>

(Part of the split. Please, take a look at the cover letter for more
details).

Signed-off-by: Viktor Barna <viktor.barna@celeno.com>
---
 .../net/wireless/celeno/cl8k/tx/tx_queue.c    | 1620 +++++++++++++++++
 1 file changed, 1620 insertions(+)
 create mode 100644 drivers/net/wireless/celeno/cl8k/tx/tx_queue.c

--
2.30.0
diff mbox series

Patch

diff --git a/drivers/net/wireless/celeno/cl8k/tx/tx_queue.c b/drivers/net/wireless/celeno/cl8k/tx/tx_queue.c
new file mode 100644
index 000000000000..18c5bd2b81f7
--- /dev/null
+++ b/drivers/net/wireless/celeno/cl8k/tx/tx_queue.c
@@ -0,0 +1,1620 @@ 
+// SPDX-License-Identifier: MIT
+/* Copyright(c) 2019-2021, Celeno Communications Ltd. */
+
+#include <net/cfg80211.h>
+#include <net/mac80211.h>
+#include <linux/ieee80211.h>
+#include <linux/types.h>
+
+#include "tx/tx_queue.h"
+#include "tx/tx.h"
+#include "tx/sw_txhdr.h"
+#include "tx/tx_amsdu.h"
+#include "tx/baw.h"
+#ifdef CONFIG_CL_PCIE
+#include "bus/pci/ipc.h"
+#endif
+#include "tx/agg_cfm.h"
+
+const u8 cl_tid2hwq[IEEE80211_NUM_TIDS] = {
+       CL_HWQ_BE,
+       CL_HWQ_BK,
+       CL_HWQ_BK,
+       CL_HWQ_BE,
+       CL_HWQ_VI,
+       CL_HWQ_VI,
+       CL_HWQ_VO,
+       CL_HWQ_VO,
+       /* At the moment, all others TID are mapped to BE */
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+       CL_HWQ_BE,
+};
+
+static u32 cl_txq_total_dump_drv(struct cl_tx_queue *tx_queue)
+{
+       return tx_queue->dump_queue_full + tx_queue->dump_dma_map_fail;
+}
+
+static void cl_txq_sched_list_add(struct cl_tx_queue *tx_queue, struct cl_hw *cl_hw)
+{
+       /* Add to schedule queue */
+       if (tx_queue->sched)
+               return;
+
+       tx_queue->sched = true;
+       if (tx_queue->type == QUEUE_TYPE_AGG)
+               list_add_tail(&tx_queue->sched_list, &cl_hw->list_sched_q_agg);
+       else
+               list_add_tail(&tx_queue->sched_list, &cl_hw->list_sched_q_single);
+}
+
+static void cl_txq_sched_list_remove(struct cl_tx_queue *tx_queue)
+{
+       /* Remove from schedule queue */
+       if (tx_queue->sched) {
+               tx_queue->sched = false;
+               list_del(&tx_queue->sched_list);
+       }
+}
+
+static void cl_txq_sched_list_remove_if_empty(struct cl_tx_queue *tx_queue)
+{
+       /* If queue is empty remove it from schedule list */
+       if (list_empty(&tx_queue->hdrs))
+               cl_txq_sched_list_remove(tx_queue);
+}
+
+static void cl_txq_transfer_single_to_agg(struct cl_hw *cl_hw,
+                                         struct cl_tx_queue *single_queue,
+                                         struct cl_tx_queue *agg_queue, u8 tid)
+{
+       struct cl_sw_txhdr *sw_txhdr, *sw_txhdr_tmp;
+       struct ieee80211_tx_info *tx_info;
+       struct sk_buff *skb;
+       u8 hdr_pads;
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       if (single_queue->num_packets == 0)
+               goto out;
+
+       list_for_each_entry_safe(sw_txhdr, sw_txhdr_tmp, &single_queue->hdrs, tx_queue_list) {
+               if (sw_txhdr->tid != tid)
+                       continue;
+
+               if (!ieee80211_is_data_qos(sw_txhdr->fc))
+                       continue;
+
+               cl_hw->tx_packet_cntr.transfer.single_to_agg++;
+
+               /* Remove from single queue */
+               list_del(&sw_txhdr->tx_queue_list);
+
+               /* Update single queue counters */
+               single_queue->num_packets--;
+               single_queue->total_packets--;
+
+               /* Turn on AMPDU flag */
+               skb = sw_txhdr->skb;
+               tx_info = IEEE80211_SKB_CB(skb);
+               tx_info->flags |= IEEE80211_TX_CTL_AMPDU;
+
+               /* Push skb to agg queue */
+               hdr_pads = CL_SKB_DATA_ALIGN_PADS(skb->data);
+               cl_tx_agg_prep(cl_hw, sw_txhdr, skb->len, hdr_pads, false);
+               agg_queue->total_packets++;
+               sw_txhdr->tx_queue = agg_queue;
+               cl_txq_push(cl_hw, sw_txhdr);
+       }
+
+       /* If single queue is empty remove it from schedule list */
+       cl_txq_sched_list_remove_if_empty(single_queue);
+
+out:
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+static void cl_txq_delete_packets(struct cl_hw *cl_hw, struct cl_tx_queue *tx_queue, u8 sta_idx)
+{
+       struct cl_sw_txhdr *sw_txhdr, *sw_txhdr_tmp;
+
+       list_for_each_entry_safe(sw_txhdr, sw_txhdr_tmp, &tx_queue->hdrs, tx_queue_list) {
+               /*
+                * Brodcast frames do not have cl_sta and should not be
+                * deleted at station remove sequence.
+                */
+               if (!sw_txhdr->cl_sta)
+                       continue;
+
+               if (sw_txhdr->sta_idx != sta_idx)
+                       continue;
+
+               list_del(&sw_txhdr->tx_queue_list);
+               tx_queue->num_packets--;
+
+               cl_tx_single_free_skb(cl_hw, sw_txhdr->skb);
+               cl_sw_txhdr_free(cl_hw, sw_txhdr);
+       }
+
+       /* If queue is empty remove it from schedule list */
+       cl_txq_sched_list_remove_if_empty(tx_queue);
+}
+
+static void cl_txq_reset_counters(struct cl_tx_queue *tx_queue)
+{
+       tx_queue->total_fw_push_desc = 0;
+       tx_queue->total_fw_push_skb = 0;
+       tx_queue->total_fw_cfm = 0;
+       tx_queue->total_packets = 0;
+       tx_queue->dump_queue_full = 0;
+       tx_queue->dump_dma_map_fail = 0;
+
+       memset(tx_queue->stats_hw_amsdu_cnt, 0,
+              sizeof(tx_queue->stats_hw_amsdu_cnt));
+
+       memset(tx_queue->stats_sw_amsdu_cnt, 0,
+              sizeof(tx_queue->stats_sw_amsdu_cnt));
+}
+
+static u16 cl_txq_desc_in_fw(struct cl_tx_queue *tx_queue)
+{
+       return (tx_queue->fw_max_size - tx_queue->fw_free_space);
+}
+
+static void cl_txq_reset_counters_during_traffic(struct cl_tx_queue *tx_queue)
+{
+       /*
+        * This function can be called during traffic, while descriptors
+        * are waiting in firmware. We set total_fw_cfm to minus the number
+        * of descriptors in firmware so that after confirmation arrives
+        * total_fw_cfm will be equal to total_fw_push_desc.
+        */
+       u32 desc_in_fw = cl_txq_desc_in_fw(tx_queue);
+
+       cl_txq_reset_counters(tx_queue);
+       tx_queue->total_fw_cfm = -desc_in_fw;
+}
+
+static void cl_txq_agg_size_set(struct cl_hw *cl_hw)
+{
+       struct cl_tx_queue *tx_queue = NULL;
+       u16 new_size = 0;
+       u16 drv_max_size = 0;
+       int i = 0;
+       int j = 0;
+
+       if (!cl_hw->used_agg_queues || !cl_hw->conf->ci_tx_packet_limit)
+               return;
+
+       new_size = cl_hw->conf->ci_tx_packet_limit / cl_hw->used_agg_queues;
+       drv_max_size = max(new_size, cl_hw->conf->ci_tx_queue_size_agg);
+
+       for (i = 0; i < IPC_MAX_BA_SESSIONS; i++) {
+               tx_queue = &cl_hw->tx_queues.agg[i];
+
+               if (!tx_queue->cl_sta)
+                       continue;
+
+               tx_queue->max_packets = drv_max_size;
+
+               j++;
+               if (j == cl_hw->used_agg_queues)
+                       break;
+       }
+
+       cl_dbg_trace(cl_hw, "drv_max_size = %u\n", drv_max_size);
+}
+
+static int cl_txq_request_find(struct cl_hw *cl_hw, u8 sta_idx, u8 tid)
+{
+       int i = 0;
+       struct cl_req_agg_db *req_agg_db = NULL;
+       u8 req_agg_queues = 0;
+
+       for (i = 0; (i < IPC_MAX_BA_SESSIONS) && (req_agg_queues < cl_hw->req_agg_queues); i++) {
+               req_agg_db = &cl_hw->req_agg_db[i];
+
+               if (!req_agg_db->is_used)
+                       continue;
+
+               req_agg_queues++;
+
+               if (sta_idx == req_agg_db->sta_idx && tid == req_agg_db->tid)
+                       return i;
+       }
+
+       return -1;
+}
+
+static void cl_txq_traffic_counters_print_bcmc(struct cl_hw *cl_hw,
+                                              char **buf, int *len, ssize_t *buf_size)
+{
+       struct cl_tx_queue *tx_queue = &cl_hw->tx_queues.bcmc;
+       unsigned long flags;
+       u8 hw_index;
+       u32 total_packets;
+       u16 fw_curr;
+       u32 total_push;
+       u32 total_cfm;
+       u32 dump;
+
+       spin_lock_irqsave(&cl_hw->tx_lock_bcmc, flags);
+
+       hw_index = tx_queue->hw_index;
+       total_packets = tx_queue->total_packets;
+       fw_curr = cl_txq_desc_in_fw(tx_queue);
+       total_push = tx_queue->total_fw_push_skb;
+       total_cfm = tx_queue->total_fw_cfm;
+       dump = cl_txq_total_dump_drv(tx_queue);
+
+       spin_unlock_irqrestore(&cl_hw->tx_lock_bcmc, flags);
+
+       if (total_packets == 0)
+               return;
+
+       cl_snprintf(buf, len, buf_size,
+                   "\nTX MULTICAST AND BOROADCAST QUEUE (MAX 1):\n"
+                   "|-----------------------------------------------------------|\n"
+                   "| hw  | driver   | fw      | fw total | fw total | dump     |\n"
+                   "| idx | total    | current | push     | cfm      |          |\n"
+                   "|-----+----------+---------+----------+----------+----------|\n"
+                   "| %3u |%10u|%9u|%10u|%10u|%10u|\n",
+                   hw_index, total_packets, fw_curr, total_push, total_cfm, dump);
+       cl_snprintf(buf, len, buf_size,
+                   "|-----------------------------------------------------------|\n");
+}
+
+static void cl_txq_traffic_counters_print_single(struct cl_hw *cl_hw,
+                                                char **buf, int *len, ssize_t *buf_size)
+{
+       u16 queue_idx = 0;
+       u32 sta_idx = 0, ac = 0;
+       struct cl_tx_queue *tx_queue;
+
+       cl_snprintf(buf, len, buf_size,
+                   "\nTX SINGLE QUEUES (MAX %d):\n", MAX_SINGLE_QUEUES);
+       cl_snprintf(buf, len, buf_size,
+                   "|----------------------------------------------------------------------"
+                   "----------|\n"
+                   "| idx | sta | ac | driver   | driver  | fw      | fw total | fw total |"
+                   " dump     |\n"
+                   "|     |     |    | total    | current | current | push     | cfm      |"
+                   "          |\n"
+                   "|-----+-----+----+----------+---------+---------+----------+----------+"
+                   "----------|\n");
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       for (sta_idx = 0; sta_idx < FW_MAX_NUM_STA; sta_idx++) {
+               for (ac = 0; ac < AC_MAX; ac++) {
+                       queue_idx = QUEUE_IDX(sta_idx, ac);
+                       tx_queue = &cl_hw->tx_queues.single[queue_idx];
+
+                       if (tx_queue->total_packets == 0)
+                               continue;
+
+                       if (tx_queue->index == HIGH_PRIORITY_QUEUE)
+                               cl_snprintf(buf, len, buf_size,
+                                           "|-----+-----+----+----------+---------+---------+"
+                                           "----------+----------+----------|\n");
+
+                       cl_snprintf(buf, len, buf_size,
+                                   "| %3u | %3u | %2u |%10u|%9u|%9u|%10u|%10u|%10u|\n",
+                                   tx_queue->index,
+                                   sta_idx,
+                                   tx_queue->hw_index,
+                                   tx_queue->total_packets,
+                                   tx_queue->num_packets,
+                                   cl_txq_desc_in_fw(tx_queue),
+                                   tx_queue->total_fw_push_skb,
+                                   tx_queue->total_fw_cfm,
+                                   cl_txq_total_dump_drv(tx_queue));
+               }
+       }
+
+       cl_snprintf(buf, len, buf_size,
+                   "|------------------------------------------------------------------------"
+                   "--------|\n");
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+static void cl_txq_traffic_counters_print_agg(struct cl_hw *cl_hw,
+                                             char **buf, int *len, ssize_t *buf_size)
+{
+       u32 ba_idx = 0;
+       struct cl_tx_queue *tx_queue;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       if (cl_hw->used_agg_queues == 0)
+               goto out;
+
+       cl_snprintf(buf, len, buf_size,
+                   "\nTX AGGREGATION QUEUES (MAX %d):\n", IPC_MAX_BA_SESSIONS);
+       cl_snprintf(buf, len, buf_size,
+                   "|-----------------------------------------------------------------------"
+                   "---------------------|\n"
+                   "| idx | sta | tid | driver   | driver  | fw      | fw total | fw total |"
+                   " fw total | dump     |\n"
+                   "|     | idx |     | total    | current | current | push skb | push desc|"
+                   " cfm      |          |\n"
+                   "|-----+-----+-----+----------+---------+---------+----------+----------+"
+                   "----------+----------|\n");
+
+       for (ba_idx = 0; ba_idx < IPC_MAX_BA_SESSIONS; ba_idx++) {
+               tx_queue = &cl_hw->tx_queues.agg[ba_idx];
+
+               if (!cl_hw->tx_queues.agg[ba_idx].cl_sta)
+                       continue;
+
+               if (tx_queue->total_packets == 0)
+                       continue;
+
+               cl_snprintf(buf, len, buf_size,
+                           "| %3u | %3u | %3u |%10u|%9u|%9u|%10u|%10u|%10u|%10u|\n",
+                           tx_queue->index,
+                           tx_queue->cl_sta->sta_idx,
+                           tx_queue->tid,
+                           tx_queue->total_packets,
+                           tx_queue->num_packets,
+                           cl_txq_desc_in_fw(tx_queue),
+                           tx_queue->total_fw_push_skb,
+                           tx_queue->total_fw_push_desc,
+                           tx_queue->total_fw_cfm,
+                           cl_txq_total_dump_drv(tx_queue));
+       }
+
+       cl_snprintf(buf, len, buf_size,
+                   "|------------------------------------------------------------------------"
+                   "--------------------|\n");
+
+out:
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+}
+
+static void cl_txq_traffic_counters_print_mac(struct cl_hw *cl_hw,
+                                             char **buf, int *len, ssize_t *buf_size)
+{
+       struct ieee80211_local *local = hw_to_local(cl_hw->hw);
+       u32 i = 0, total_len = 0, q_len[IEEE80211_MAX_QUEUES] = {0};
+       unsigned long flags;
+
+       spin_lock_irqsave(&local->queue_stop_reason_lock, flags);
+
+       for (i = 0; i < IEEE80211_MAX_QUEUES; i++) {
+               q_len[i] = skb_queue_len(&local->pending[i]);
+               total_len += q_len[i];
+       }
+
+       spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
+
+       if (total_len == 0)
+               return;
+
+       cl_snprintf(buf, len, buf_size,
+                   "\nMAC80211 PENDING QUEUES (MAX %d):\n", IEEE80211_MAX_QUEUES);
+       cl_snprintf(buf, len, buf_size,
+                   "|--------------------|\n"
+                   "| queue |  current   |\n"
+                   "|-------+------------|\n");
+
+       for (i = 0; i < IEEE80211_MAX_QUEUES; i++)
+               if (q_len[i] > 0)
+                       cl_snprintf(buf, len, buf_size, "| %5u | %10u |\n", i, q_len[i]);
+
+       cl_snprintf(buf, len, buf_size, "|--------------------|\n");
+}
+
+static int cl_txq_traffic_counters_print(struct cl_hw *cl_hw)
+{
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       cl_txq_traffic_counters_print_bcmc(cl_hw, &buf, &len, &buf_size);
+       cl_txq_traffic_counters_print_single(cl_hw, &buf, &len, &buf_size);
+       cl_txq_traffic_counters_print_agg(cl_hw, &buf, &len, &buf_size);
+       cl_txq_traffic_counters_print_mac(cl_hw, &buf, &len, &buf_size);
+
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static void cl_txq_drop_reasons_print_bcmc(struct cl_hw *cl_hw,
+                                          char **buf, int *len, ssize_t *buf_size)
+{
+       u32 total = 0;
+       u32 dump_queue_full = 0;
+       u32 dump_dma_map_fail = 0;
+       struct cl_tx_queue *tx_queue = &cl_hw->tx_queues.bcmc;
+       unsigned long flags;
+
+       spin_lock_irqsave(&cl_hw->tx_lock_bcmc, flags);
+
+       total = cl_txq_total_dump_drv(tx_queue);
+       dump_queue_full = tx_queue->dump_queue_full;
+       dump_dma_map_fail = tx_queue->dump_dma_map_fail;
+
+       spin_unlock_irqrestore(&cl_hw->tx_lock_bcmc, flags);
+
+       if (total > 0)
+               cl_snprintf(buf, len, buf_size,
+                           "|bcmc  |     |%10u|%10u|%10u|\n",
+                           dump_queue_full, dump_dma_map_fail, total);
+}
+
+static void cl_txq_drop_reasons_print_single(struct cl_hw *cl_hw,
+                                            char **buf, int *len, ssize_t *buf_size)
+{
+       u32 i = 0, total = 0;
+       struct cl_tx_queue *tx_queue;
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       for (i = 0; i < MAX_SINGLE_QUEUES; i++) {
+               tx_queue = &cl_hw->tx_queues.single[i];
+               total = cl_txq_total_dump_drv(tx_queue);
+
+               if (total == 0)
+                       continue;
+
+               cl_snprintf(buf, len, buf_size,
+                           "|single|%5u|%10u|%10u|%10u|\n",
+                           tx_queue->index,
+                           tx_queue->dump_queue_full,
+                           tx_queue->dump_dma_map_fail,
+                           total);
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+static void cl_txq_drop_reasons_print_agg(struct cl_hw *cl_hw,
+                                         char **buf, int *len, ssize_t *buf_size)
+{
+       u32 i = 0, total = 0;
+       struct cl_tx_queue *tx_queue;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       for (i = 0; i < IPC_MAX_BA_SESSIONS; i++) {
+               tx_queue = &cl_hw->tx_queues.agg[i];
+               total = cl_txq_total_dump_drv(tx_queue);
+
+               if (total == 0)
+                       continue;
+
+               cl_snprintf(buf, len, buf_size,
+                           "|agg   |%5u|%10u|%10u|%10u|\n",
+                           tx_queue->index,
+                           tx_queue->dump_queue_full,
+                           tx_queue->dump_dma_map_fail,
+                           total);
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+}
+
+static int cl_txq_drop_reasons_print(struct cl_hw *cl_hw)
+{
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       cl_snprintf(&buf, &len, &buf_size,
+                   "-----------------------------------------------\n"
+                   "| type | idx | queue    | dma map  | total    |\n"
+                   "|      |     | full     | fail     | dump     |\n"
+                   "|------+-----+----------+----------+----------|\n");
+
+       cl_txq_drop_reasons_print_bcmc(cl_hw, &buf, &len, &buf_size);
+       cl_txq_drop_reasons_print_single(cl_hw, &buf, &len, &buf_size);
+       cl_txq_drop_reasons_print_agg(cl_hw, &buf, &len, &buf_size);
+
+       cl_snprintf(&buf, &len, &buf_size,
+                   "-----------------------------------------------\n");
+
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static int cl_txq_global_counters_print(struct cl_hw *cl_hw)
+{
+       struct cl_tx_forward_cntr *forward = &cl_hw->tx_packet_cntr.forward;
+       struct cl_tx_drop_cntr *drop = &cl_hw->tx_packet_cntr.drop;
+       struct cl_tx_transfer_cntr *transfer = &cl_hw->tx_packet_cntr.transfer;
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       cl_snprintf(&buf, &len, &buf_size,
+                   "FORWARD\n"
+                   "----------------------------\n"
+                   "tx_start          = %u\n", forward->tx_start);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "drv_fast_agg      = %u\n", forward->drv_fast_agg);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "drv_fast_single   = %u\n", forward->drv_fast_single);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "to_mac            = %u\n", forward->to_mac);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "from_mac_single   = %u\n", forward->from_mac_single);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "from_mac_agg      = %u\n", forward->from_mac_agg);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "DROP\n"
+                   "----------------------------\n"
+                   "wd_restart        = %u\n", drop->wd_restart);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "radio_off         = %u\n", drop->radio_off);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "in_recovery       = %u\n", drop->in_recovery);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "short_length      = %u\n", drop->short_length);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "pending_full      = %u\n", drop->pending_full);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "packet_limit      = %u\n", drop->packet_limit);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "dev_flags         = %u\n", drop->dev_flags);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "length_limit      = %u\n", drop->length_limit);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "txhdr_alloc_fail  = %u\n", drop->txhdr_alloc_fail);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "queue_null        = %u\n", drop->queue_null);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "amsdu_alloc_fail  = %u\n", drop->amsdu_alloc_fail);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "amsdu_dma_map_err = %u\n", drop->amsdu_dma_map_err);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "build_hdr_fail    = %u\n", drop->build_hdr_fail);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "key_disable       = %u\n", drop->key_disable);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "queue_flush       = %u\n", drop->queue_flush);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "sta_null_in_agg   = %u\n", drop->sta_null_in_agg);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "TRANSFER\n"
+                   "----------------------------\n"
+                   "single_to_agg     = %u\n", transfer->single_to_agg);
+       cl_snprintf(&buf, &len, &buf_size,
+                   "agg_to_single     = %u\n", transfer->agg_to_single);
+
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static int cl_txq_stop_reasons_print(struct cl_hw *cl_hw)
+{
+       struct ieee80211_local *local = hw_to_local(cl_hw->hw);
+       unsigned long queue_stop_reasons;
+       unsigned long flags;
+       u8 i = 0;
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       cl_snprintf(&buf, &len, &buf_size,
+                   "|------------------------|\n"
+                   "|queue|queue_stop_reasons|\n"
+                   "|-----+------------------|\n");
+
+       for (i = 0; i < IEEE80211_MAX_QUEUES; i++) {
+               spin_lock_irqsave(&local->queue_stop_reason_lock, flags);
+               queue_stop_reasons = local->queue_stop_reasons[i];
+               spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
+
+               if (queue_stop_reasons)
+                       cl_snprintf(&buf, &len, &buf_size, "|%5u|0x%-16lx|\n",
+                                   i, queue_stop_reasons);
+       }
+
+       cl_snprintf(&buf, &len, &buf_size, "|------------------------|\n");
+
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static int cl_txq_requested_agg_print(struct cl_hw *cl_hw)
+{
+       u8 i = 0;
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       cl_snprintf(&buf, &len, &buf_size,
+                   "## used_agg_queues = %u\n", cl_hw->used_agg_queues);
+
+       if (cl_hw->used_agg_queues) {
+               for (i = 0; i < IPC_MAX_BA_SESSIONS; i++) {
+                       if (!cl_hw->tx_queues.agg[i].cl_sta)
+                               continue;
+
+                       cl_snprintf(&buf, &len, &buf_size,
+                                   "%u) sta_idx = %u, tid = %u\n", i + 1,
+                                   cl_hw->tx_queues.agg[i].cl_sta->sta_idx,
+                                   cl_hw->tx_queues.agg[i].tid);
+               }
+       }
+
+       cl_snprintf(&buf, &len, &buf_size, "## req_agg_queues = %u\n", cl_hw->req_agg_queues);
+
+       if (cl_hw->req_agg_queues) {
+               for (i = 0; i < IPC_MAX_BA_SESSIONS; i++) {
+                       if (!cl_hw->req_agg_db[i].is_used)
+                               continue;
+
+                       cl_snprintf(&buf, &len, &buf_size, "%u) sta_idx = %u, tid = %u\n",
+                                   i + 1, cl_hw->req_agg_db[i].sta_idx,
+                                   cl_hw->req_agg_db[i].tid);
+               }
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static void cl_txq_hw_amsdu_stats_print(struct cl_hw *cl_hw,
+                                       char **buf, int *len, ssize_t *buf_size)
+{
+       u32 i, j;
+       struct cl_tx_queue *tx_queue;
+
+       cl_snprintf(buf, len, buf_size, "HW TX-AMSDU STATS:\n");
+
+       cl_snprintf(buf, len, buf_size, "|-----------");
+       for (i = 0; i < CL_AMSDU_TX_PAYLOAD_MAX; i++)
+               cl_snprintf(buf, len, buf_size, "-----------");
+
+       cl_snprintf(buf, len, buf_size, "|\n|agg|sta|tid");
+       for (i = 0; i < CL_AMSDU_TX_PAYLOAD_MAX; i++)
+               cl_snprintf(buf, len, buf_size, "| amsdu #%u ", i + 1);
+
+       cl_snprintf(buf, len, buf_size, "|\n|---+---+---");
+       for (i = 0; i < CL_AMSDU_TX_PAYLOAD_MAX; i++)
+               cl_snprintf(buf, len, buf_size, "+----------");
+
+       cl_snprintf(buf, len, buf_size, "|\n");
+
+       for (j = 0; j < IPC_MAX_BA_SESSIONS; j++) {
+               tx_queue = &cl_hw->tx_queues.agg[j];
+
+               if (!tx_queue->cl_sta)
+                       continue;
+
+               if (tx_queue->total_packets == 0)
+                       continue;
+
+               cl_snprintf(buf, len, buf_size, "|%3u|%3u|%3u",
+                           tx_queue->index,
+                           tx_queue->cl_sta->sta_idx,
+                           tx_queue->tid);
+
+               for (i = 0; i < CL_AMSDU_TX_PAYLOAD_MAX; i++)
+                       cl_snprintf(buf, len, buf_size, "|%10u", tx_queue->stats_hw_amsdu_cnt[i]);
+
+               cl_snprintf(buf, len, buf_size, "|\n");
+       }
+
+       cl_snprintf(buf, len, buf_size,  "|-----------");
+       for (i = 0; i < CL_AMSDU_TX_PAYLOAD_MAX; i++)
+               cl_snprintf(buf, len, buf_size, "-----------");
+
+       cl_snprintf(buf, len, buf_size, "|\n");
+}
+
+static void cl_txq_sw_amsdu_stats_print(struct cl_hw *cl_hw,
+                                       char **buf, int *len, ssize_t *buf_size)
+{
+       u32 i, j;
+       struct cl_tx_queue *tx_queue;
+
+       if (cl_hw->conf->ci_tx_sw_amsdu_max_packets < 2)
+               return;
+
+       cl_snprintf(buf, len, buf_size, "SW TX-AMSDU STATS:\n");
+
+       for (j = 0; j < IPC_MAX_BA_SESSIONS; j++) {
+               tx_queue = &cl_hw->tx_queues.agg[j];
+
+               if (!tx_queue->cl_sta)
+                       continue;
+
+               if (tx_queue->total_packets == 0)
+                       continue;
+
+               cl_snprintf(buf, len, buf_size,
+                           "\nagg idx %u, sta %u, tid %u :\n",
+                           tx_queue->index,
+                           tx_queue->cl_sta->sta_idx,
+                           tx_queue->tid);
+               cl_snprintf(buf, len, buf_size, "----------------------------\n");
+
+               for (i = 0; i < cl_hw->conf->ci_tx_sw_amsdu_max_packets; i++)
+                       if (tx_queue->stats_sw_amsdu_cnt[i] > 0)
+                               cl_snprintf(buf, len, buf_size,
+                                           "amsdu #%u = %u\n", i + 1,
+                                           tx_queue->stats_sw_amsdu_cnt[i]);
+       }
+}
+
+static int cl_txq_amsdu_stats_print(struct cl_hw *cl_hw)
+{
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       if (cl_hw->used_agg_queues == 0)
+               goto out;
+
+       cl_txq_hw_amsdu_stats_print(cl_hw, &buf, &len, &buf_size);
+       cl_txq_sw_amsdu_stats_print(cl_hw, &buf, &len, &buf_size);
+out:
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static void cl_txq_max_size_print_single(struct cl_hw *cl_hw,
+                                        char **buf, int *len, ssize_t *buf_size)
+{
+       u8 ac = 0;
+       u8 sta_idx = 0;
+       u16 queue_idx = 0;
+       struct cl_tx_queue *tx_queue;
+
+       cl_snprintf(buf, len, buf_size,
+                   "MAX SIZE SINGLE QUEUES:\n"
+                   "|----------------------|\n"
+                   "|idx|sta|ac| drv | fw  |\n"
+                   "|---+---+--+-----+-----|\n");
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       for (sta_idx = 0; sta_idx < FW_MAX_NUM_STA; sta_idx++) {
+               for (ac = 0; ac < AC_MAX; ac++) {
+                       queue_idx = QUEUE_IDX(sta_idx, ac);
+                       tx_queue = &cl_hw->tx_queues.single[queue_idx];
+
+                       if (tx_queue->total_fw_push_skb == 0)
+                               continue;
+
+                       cl_snprintf(buf, len, buf_size,
+                                   "|%3u|%3u|%2u|%5u|%5u|\n",
+                                   tx_queue->index,
+                                   sta_idx,
+                                   tx_queue->hw_index,
+                                   tx_queue->max_packets,
+                                   tx_queue->fw_max_size);
+               }
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+
+       cl_snprintf(buf, len, buf_size, "|----------------------|\n");
+}
+
+static void cl_txq_max_size_print_agg(struct cl_hw *cl_hw,
+                                     char **buf, int *len, ssize_t *buf_size)
+{
+       u8 agg_idx = 0;
+       struct cl_tx_queue *tx_queue;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       if (cl_hw->used_agg_queues == 0)
+               goto out;
+
+       cl_snprintf(buf, len, buf_size,
+                   "MAX SIZE AGGREGATION QUEUES:\n"
+                   "|-----------------------|\n"
+                   "|idx|sta|tid| drv | fw  |\n"
+                   "|---+---+---+-----+-----|\n");
+
+       for (agg_idx = 0; agg_idx < IPC_MAX_BA_SESSIONS; agg_idx++) {
+               tx_queue = &cl_hw->tx_queues.agg[agg_idx];
+
+               if (!tx_queue->cl_sta)
+                       continue;
+
+               cl_snprintf(buf, len, buf_size,
+                           "|%3u|%3u|%3u|%5u|%5u|\n",
+                           tx_queue->index,
+                           tx_queue->cl_sta->sta_idx,
+                           tx_queue->tid,
+                           tx_queue->max_packets,
+                           tx_queue->fw_max_size);
+       }
+
+       cl_snprintf(buf, len, buf_size, "|----------------------|\n");
+
+out:
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+}
+
+static int cl_txq_max_size_print(struct cl_hw *cl_hw)
+{
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       cl_txq_max_size_print_single(cl_hw, &buf, &len, &buf_size);
+       cl_txq_max_size_print_agg(cl_hw, &buf, &len, &buf_size);
+
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static void cl_txq_stats_reset_bcmc(struct cl_hw *cl_hw)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&cl_hw->tx_lock_bcmc, flags);
+       cl_txq_reset_counters_during_traffic(&cl_hw->tx_queues.bcmc);
+       spin_unlock_irqrestore(&cl_hw->tx_lock_bcmc, flags);
+}
+
+static void cl_txq_stats_reset_single(struct cl_hw *cl_hw)
+{
+       u16 i = 0;
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       for (i = 0; i < MAX_SINGLE_QUEUES; i++)
+               cl_txq_reset_counters_during_traffic(&cl_hw->tx_queues.single[i]);
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+static void cl_txq_stats_reset_agg(struct cl_hw *cl_hw)
+{
+       u16 i = 0;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       for (i = 0; i < IPC_MAX_BA_SESSIONS; i++)
+               cl_txq_reset_counters_during_traffic(&cl_hw->tx_queues.agg[i]);
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+}
+
+static void cl_txq_stats_reset(struct cl_hw *cl_hw)
+{
+       cl_txq_stats_reset_bcmc(cl_hw);
+       cl_txq_stats_reset_single(cl_hw);
+       cl_txq_stats_reset_agg(cl_hw);
+
+       memset(&cl_hw->tx_packet_cntr, 0, sizeof(struct cl_tx_packet_cntr));
+
+       pr_debug("Reset queue stats\n");
+}
+
+static void cl_txq_sched_list_print_single(struct cl_hw *cl_hw,
+                                          char **buf, int *len, ssize_t *buf_size)
+{
+       struct cl_tx_queue *tx_queue;
+       u32 num_queues = 0;
+
+       cl_snprintf(buf, len, buf_size, "TX single sched list:\n");
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       list_for_each_entry(tx_queue, &cl_hw->list_sched_q_single, sched_list) {
+               num_queues++;
+               cl_snprintf(buf, len, buf_size, "%u) Index = %u\n", num_queues, tx_queue->index);
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+
+       if (num_queues == 0)
+               cl_snprintf(buf, len, buf_size, "empty\n");
+}
+
+static void cl_txq_sched_list_print_agg(struct cl_hw *cl_hw,
+                                       char **buf, int *len, ssize_t *buf_size)
+{
+       struct cl_tx_queue *tx_queue;
+       u32 num_queues = 0;
+
+       cl_snprintf(buf, len, buf_size, "\nTX agg sched list:\n");
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       list_for_each_entry(tx_queue, &cl_hw->list_sched_q_agg, sched_list) {
+               num_queues++;
+               cl_snprintf(buf, len, buf_size, "%u) Index = %u\n", num_queues, tx_queue->index);
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+
+       if (num_queues == 0)
+               cl_snprintf(buf, len, buf_size, "empty\n");
+}
+
+static int cl_txq_sched_list_print(struct cl_hw *cl_hw)
+{
+       char *buf = NULL;
+       ssize_t buf_size;
+       int err = 0;
+       int len = 0;
+
+       cl_txq_sched_list_print_single(cl_hw, &buf, &len, &buf_size);
+       cl_txq_sched_list_print_agg(cl_hw, &buf, &len, &buf_size);
+
+       err = cl_vendor_reply(cl_hw, buf, len);
+       kfree(buf);
+
+       return err;
+}
+
+static int cl_txq_cli_help(struct cl_hw *cl_hw)
+{
+       char *buf = kzalloc(PAGE_SIZE, GFP_KERNEL);
+       int err = 0;
+
+       if (!buf)
+               return -ENOMEM;
+
+       snprintf(buf, PAGE_SIZE,
+                "txq usage:\n"
+                "-a : Print traffic counters\n"
+                "-b : Print drop counters\n"
+                "-c : Print global counters\n"
+                "-d : Print stop reasons\n"
+                "-e : Print requested aggregations\n"
+                "-f : Print AMSDU statistics\n"
+                "-m : Print maximum queues size\n"
+                "-r : Reset queue stats\n"
+                "-s : Print schedule list\n");
+
+       err = cl_vendor_reply(cl_hw, buf, strlen(buf));
+       kfree(buf);
+
+       return err;
+}
+
+static void cl_txq_task_single(struct cl_hw *cl_hw)
+{
+       /* Schedule single queues */
+       struct cl_tx_queue *tx_queue, *tx_queue_tmp;
+
+       spin_lock(&cl_hw->tx_lock_single);
+
+       list_for_each_entry_safe(tx_queue, tx_queue_tmp, &cl_hw->list_sched_q_single, sched_list)
+               cl_txq_sched(cl_hw, tx_queue);
+
+       /* Rotate the queue so next schedule will start with a different queue */
+       list_rotate_left(&cl_hw->list_sched_q_single);
+
+       spin_unlock(&cl_hw->tx_lock_single);
+}
+
+static void cl_txq_task_agg(struct cl_hw *cl_hw)
+{
+       /* Schedule agg queueus */
+       struct cl_tx_queue *tx_queue, *tx_queue_tmp;
+
+       spin_lock(&cl_hw->tx_lock_agg);
+
+       list_for_each_entry_safe(tx_queue, tx_queue_tmp, &cl_hw->list_sched_q_agg, sched_list)
+               cl_txq_sched(cl_hw, tx_queue);
+
+       /* Rotate the queue so next schedule will start with a different queue */
+       list_rotate_left(&cl_hw->list_sched_q_agg);
+
+       spin_unlock(&cl_hw->tx_lock_agg);
+}
+
+static void cl_txq_task(unsigned long data)
+{
+       struct cl_hw *cl_hw = (struct cl_hw *)data;
+
+       cl_txq_task_single(cl_hw);
+       cl_txq_task_agg(cl_hw);
+}
+
+static void cl_txq_agg_inc_usage_cntr(struct cl_hw *cl_hw)
+{
+       /* Should be called in cl_hw->tx_lock_agg context */
+       cl_hw->used_agg_queues++;
+       cl_txq_agg_size_set(cl_hw);
+}
+
+static void cl_txq_agg_dec_usage_cntr(struct cl_hw *cl_hw)
+{
+       /* Should be called in cl_hw->tx_lock_agg context */
+       WARN_ON_ONCE(cl_hw->used_agg_queues == 0);
+
+       cl_hw->used_agg_queues--;
+       cl_txq_agg_size_set(cl_hw);
+}
+
+static void cl_txq_init_single(struct cl_hw *cl_hw)
+{
+       struct cl_tx_queue *tx_queue;
+       u32 i;
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       INIT_LIST_HEAD(&cl_hw->list_sched_q_single);
+
+       for (i = 0; i < MAX_SINGLE_QUEUES; i++) {
+               tx_queue = &cl_hw->tx_queues.single[i];
+               memset(tx_queue, 0, sizeof(struct cl_tx_queue));
+               INIT_LIST_HEAD(&tx_queue->hdrs);
+               tx_queue->hw_index = i / FW_MAX_NUM_STA;
+               tx_queue->fw_max_size = IPC_TXDESC_CNT_SINGLE;
+               tx_queue->fw_free_space = IPC_TXDESC_CNT_SINGLE;
+               tx_queue->index = i;
+               tx_queue->max_packets = cl_hw->conf->ci_tx_queue_size_single;
+               tx_queue->type = QUEUE_TYPE_SINGLE;
+       }
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+static void cl_txq_init_bcmc(struct cl_hw *cl_hw)
+{
+       unsigned long flags;
+       struct cl_tx_queue *tx_queue;
+
+       spin_lock_irqsave(&cl_hw->tx_lock_bcmc, flags);
+
+       tx_queue = &cl_hw->tx_queues.bcmc;
+       memset(tx_queue, 0, sizeof(struct cl_tx_queue));
+       INIT_LIST_HEAD(&tx_queue->hdrs);
+       tx_queue->hw_index = CL_HWQ_BCN;
+       tx_queue->fw_max_size = IPC_TXDESC_CNT_BCMC;
+       tx_queue->fw_free_space = IPC_TXDESC_CNT_BCMC;
+       tx_queue->index = 0;
+       tx_queue->max_packets = 0;
+       tx_queue->type = QUEUE_TYPE_BCMC;
+
+       spin_unlock_irqrestore(&cl_hw->tx_lock_bcmc, flags);
+}
+
+static void cl_txq_init_agg(struct cl_hw *cl_hw)
+{
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+       INIT_LIST_HEAD(&cl_hw->list_sched_q_agg);
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+}
+
+static void cl_txq_agg_request_reset(struct cl_hw *cl_hw)
+{
+       cl_hw->req_agg_queues = 0;
+       memset(cl_hw->req_agg_db, 0, sizeof(cl_hw->req_agg_db));
+}
+
+void cl_txq_init(struct cl_hw *cl_hw)
+{
+       tasklet_init(&cl_hw->tx_task, cl_txq_task, (unsigned long)cl_hw);
+
+       cl_txq_agg_request_reset(cl_hw);
+       cl_txq_init_single(cl_hw);
+       cl_txq_init_bcmc(cl_hw);
+       cl_txq_init_agg(cl_hw);
+}
+
+void cl_txq_stop(struct cl_hw *cl_hw)
+{
+       tasklet_kill(&cl_hw->tx_task);
+}
+
+struct cl_tx_queue *cl_txq_get(struct cl_hw *cl_hw, struct cl_sw_txhdr *sw_txhdr)
+{
+       struct cl_sta *cl_sta = sw_txhdr->cl_sta;
+       struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(sw_txhdr->skb);
+       u8 hw_queue = sw_txhdr->hw_queue;
+
+       if (!cl_sta &&
+           hw_queue == CL_HWQ_VO &&
+           is_multicast_ether_addr(sw_txhdr->hdr80211->addr1)) {
+               /*
+                * If HW queue is VO and packet is multicast, it was not buffered
+                * by mac80211, and it should be pushed to the high-priority queue
+                * and not to the bcmc queue.
+                */
+               return &cl_hw->tx_queues.single[HIGH_PRIORITY_QUEUE];
+       } else if (!cl_sta &&
+                  (hw_queue != CL_HWQ_BCN) &&
+                  !(tx_info->flags & IEEE80211_TX_CTL_NO_PS_BUFFER)) {
+               /*
+                * If station is NULL, but HW queue is not BCN,
+                * it most go to the high-priority queue.
+                */
+               tx_info->flags |= IEEE80211_TX_CTL_NO_PS_BUFFER;
+               sw_txhdr->hw_queue = CL_HWQ_VO;
+               return &cl_hw->tx_queues.single[HIGH_PRIORITY_QUEUE];
+       } else if (cl_sta && (tx_info->flags & IEEE80211_TX_CTL_AMPDU)) {
+               /* Agg packet */
+               return cl_sta->agg_tx_queues[sw_txhdr->tid];
+       } else if (hw_queue == CL_HWQ_BCN) {
+               return &cl_hw->tx_queues.bcmc;
+       } else if (tx_info->flags & IEEE80211_TX_CTL_NO_PS_BUFFER) {
+               /*
+                * Only frames that are power save response or non-bufferable MMPDU
+                * will have this flag set our driver will push those frame to the
+                * highiest priority queue.
+                */
+               return &cl_hw->tx_queues.single[HIGH_PRIORITY_QUEUE];
+       }
+
+       return &cl_hw->tx_queues.single[QUEUE_IDX(sw_txhdr->sta_idx, hw_queue)];
+}
+
+void cl_txq_push(struct cl_hw *cl_hw, struct cl_sw_txhdr *sw_txhdr)
+{
+       struct cl_tx_queue *tx_queue = sw_txhdr->tx_queue;
+
+       if (tx_queue->num_packets < tx_queue->max_packets) {
+               tx_queue->num_packets++;
+
+               /*
+                * This prioritization of action frames helps Samsung Galaxy Note 8 to
+                * open BA session more easily, when phy dev is PHY_DEV_OLYMPUS
+                */
+               if (ieee80211_is_action(sw_txhdr->fc))
+                       list_add(&sw_txhdr->tx_queue_list, &tx_queue->hdrs);
+               else
+                       list_add_tail(&sw_txhdr->tx_queue_list, &tx_queue->hdrs);
+
+               /* If it is the first packet in the queue, add the queue to the sched list */
+               cl_txq_sched_list_add(tx_queue, cl_hw);
+       } else {
+               struct cl_sta *cl_sta = sw_txhdr->cl_sta;
+               u8 tid = sw_txhdr->tid;
+
+               /* If the SW queue full, release the packet */
+               tx_queue->dump_queue_full++;
+
+               if (cl_sta && cl_sta->amsdu_anchor[tid].sw_txhdr) {
+                       if (cl_sta->amsdu_anchor[tid].sw_txhdr == sw_txhdr) {
+                               cl_sta->amsdu_anchor[tid].sw_txhdr = NULL;
+                               cl_sta->amsdu_anchor[tid].packet_cnt = 0;
+                       }
+               }
+
+               dev_kfree_skb_any(sw_txhdr->skb);
+               cl_sw_txhdr_free(cl_hw, sw_txhdr);
+
+               /* Schedule tasklet to try and empty the queue */
+               tasklet_schedule(&cl_hw->tx_task);
+       }
+}
+
+void cl_txq_sched(struct cl_hw *cl_hw, struct cl_tx_queue *tx_queue)
+{
+       struct cl_sw_txhdr *sw_txhdr, *sw_txhdr_tmp;
+
+       if (!test_bit(CL_DEV_STARTED, &cl_hw->drv_flags) ||
+           cl_hw->tx_disable_flags ||
+           cl_txq_is_fw_full(tx_queue))
+               return;
+
+       /* Go over all descriptors in queue */
+       list_for_each_entry_safe(sw_txhdr, sw_txhdr_tmp, &tx_queue->hdrs, tx_queue_list) {
+               list_del(&sw_txhdr->tx_queue_list);
+               tx_queue->num_packets--;
+
+               cl_tx_push(cl_hw, sw_txhdr, tx_queue);
+
+               if (cl_txq_is_fw_full(tx_queue))
+                       break;
+       }
+
+       /* If queue is empty remove it from schedule list */
+       cl_txq_sched_list_remove_if_empty(tx_queue);
+}
+
+void cl_txq_agg_alloc(struct cl_hw *cl_hw, struct cl_sta *cl_sta,
+                     struct mm_ba_add_cfm *ba_add_cfm, u16 buf_size)
+{
+       u8 tid = ba_add_cfm->tid;
+       u8 fw_agg_idx = ba_add_cfm->agg_idx;
+       u8 sta_idx = cl_sta->sta_idx;
+       u8 ac = cl_tid2hwq[tid];
+       u16 single_queue_idx = QUEUE_IDX(sta_idx, ac);
+       struct cl_tx_queue *tx_queue = &cl_hw->tx_queues.agg[fw_agg_idx];
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       /* Init aggregated queue struct */
+       memset(tx_queue, 0, sizeof(struct cl_tx_queue));
+       INIT_LIST_HEAD(&tx_queue->hdrs);
+
+       /*
+        * Firmware agg queues size is static and set to 512, so that for the worst
+        * case of HE stations,that support AMPDU of 256, it has room for two full
+        * aggregation.
+        * To keep this logic, of room for two aggregations, for non-HE stations, or
+        * for HE stations that do not support AMPDU of 256, we initialize fw_max_size
+        to twice the buffer size supported by the station.
+        */
+       tx_queue->fw_max_size = min_t(u16, cl_hw->max_agg_tx_q_size, buf_size * 2);
+       tx_queue->fw_free_space = tx_queue->fw_max_size;
+
+       tx_queue->max_packets = cl_hw->conf->ci_tx_queue_size_agg;
+       tx_queue->hw_index = ac;
+       tx_queue->cl_sta = cl_sta;
+       tx_queue->type = QUEUE_TYPE_AGG;
+       tx_queue->tid = tid;
+       tx_queue->index = fw_agg_idx;
+
+#ifdef CONFIG_CL_PCIE
+       /* Reset the synchronization counters between the fw and the IPC layer */
+       cl_hw->ipc_env->ring_indices_elem->indices->txdesc_write_idx.agg[fw_agg_idx] = 0;
+#endif
+
+       /* Attach the cl_hw chosen queue to the station and agg queues DB */
+       cl_sta->agg_tx_queues[tid] = tx_queue;
+       cl_agg_cfm_set_tx_queue(cl_hw, tx_queue, fw_agg_idx);
+
+       /* Notify upper mac80211 layer of queues resources status */
+       cl_txq_agg_inc_usage_cntr(cl_hw);
+       cl_txq_agg_request_del(cl_hw, sta_idx, tid);
+
+       /*
+        * Move the qos descriptors to the new allocated aggregated queues,
+        * otherwise we might reorder packets)
+        */
+       cl_txq_transfer_single_to_agg(cl_hw, &cl_hw->tx_queues.single[single_queue_idx],
+                                     tx_queue, tid);
+       /* Move the BA window pending packets to agg path */
+       cl_baw_pending_to_agg(cl_hw, cl_sta, tid);
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+
+       cl_dbg_trace(cl_hw, "Allocate queue [%u] to station [%u] tid [%u]\n",
+                    fw_agg_idx, sta_idx, tid);
+}
+
+void cl_txq_agg_free(struct cl_hw *cl_hw, struct cl_tx_queue *tx_queue,
+                    struct cl_sta *cl_sta, u8 tid)
+{
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       cl_dbg_trace(cl_hw, "Free queue [%u] of station [%u] tid [%u]\n",
+                    tx_queue->index, cl_sta->sta_idx, tid);
+
+       memset(tx_queue, 0, sizeof(struct cl_tx_queue));
+
+       cl_txq_agg_dec_usage_cntr(cl_hw);
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+}
+
+void cl_txq_agg_stop(struct cl_sta *cl_sta, u8 tid)
+{
+       cl_sta->agg_tx_queues[tid] = NULL;
+}
+
+void cl_txq_sta_add(struct cl_hw *cl_hw, struct cl_sta *cl_sta)
+{
+       /* Set cl_sta field for all single queues of this station */
+       u8 ac;
+       u16 queue_idx;
+
+       for (ac = 0; ac < AC_MAX; ac++) {
+               queue_idx = QUEUE_IDX(cl_sta->sta_idx, ac);
+               cl_hw->tx_queues.single[queue_idx].cl_sta = cl_sta;
+       }
+
+       /* Reset pointers to TX agg queues */
+       memset(cl_sta->agg_tx_queues, 0, sizeof(cl_sta->agg_tx_queues));
+}
+
+void cl_txq_sta_remove(struct cl_hw *cl_hw, u8 sta_idx)
+{
+       /* Clear cl_sta field for all single queues of this station */
+       u8 ac;
+       u16 queue_idx;
+
+       for (ac = 0; ac < AC_MAX; ac++) {
+               queue_idx = QUEUE_IDX(sta_idx, ac);
+               cl_hw->tx_queues.single[queue_idx].cl_sta = NULL;
+       }
+}
+
+void cl_txq_transfer_agg_to_single(struct cl_hw *cl_hw, struct cl_tx_queue *agg_queue)
+{
+       /*
+        * 1) Remove from aggregation queue
+        * 2) Free sw_txhdr
+        * 3) Push to single queue
+        */
+       struct cl_sw_txhdr *sw_txhdr, *sw_txhdr_tmp;
+       struct sk_buff *skb;
+       struct ieee80211_tx_info *tx_info;
+       struct cl_tx_queue *single_queue;
+       struct cl_sta *cl_sta = agg_queue->cl_sta;
+       u16 single_queue_idx = 0;
+
+       if (agg_queue->num_packets == 0)
+               return;
+
+       single_queue_idx = QUEUE_IDX(cl_sta->sta_idx, agg_queue->hw_index);
+       single_queue = &cl_hw->tx_queues.single[single_queue_idx];
+
+       list_for_each_entry_safe(sw_txhdr, sw_txhdr_tmp, &agg_queue->hdrs, tx_queue_list) {
+               list_del(&sw_txhdr->tx_queue_list);
+               agg_queue->num_packets--;
+
+               skb = sw_txhdr->skb;
+               tx_info = IEEE80211_SKB_CB(skb);
+
+               if (cl_tx_ctrl_is_amsdu(tx_info)) {
+                       cl_tx_amsdu_transfer_single(cl_hw, sw_txhdr);
+               } else {
+                       tx_info->flags &= ~IEEE80211_TX_CTL_AMPDU;
+
+                       if (cl_tx_8023_to_wlan(cl_hw, skb, cl_sta, sw_txhdr->tid) == 0) {
+                               cl_hw->tx_packet_cntr.transfer.agg_to_single++;
+                               cl_tx_single(cl_hw, cl_sta, skb, false, false);
+                       }
+               }
+
+               cl_sw_txhdr_free(cl_hw, sw_txhdr);
+       }
+
+       /* If queue is empty remove it from schedule list */
+       cl_txq_sched_list_remove_if_empty(agg_queue);
+}
+
+void cl_txq_flush(struct cl_hw *cl_hw, struct cl_tx_queue *tx_queue)
+{
+       struct cl_sw_txhdr *sw_txhdr, *sw_txhdr_tmp;
+       struct ieee80211_tx_info *tx_info;
+
+       if (tx_queue->num_packets == 0)
+               return;
+
+       list_for_each_entry_safe(sw_txhdr, sw_txhdr_tmp, &tx_queue->hdrs, tx_queue_list) {
+               list_del(&sw_txhdr->tx_queue_list);
+               tx_queue->num_packets--;
+
+               /* Can not send AMSDU frames as singles */
+               tx_info = IEEE80211_SKB_CB(sw_txhdr->skb);
+
+               /* Free mid & last AMSDU sub frames */
+               if (cl_tx_ctrl_is_amsdu(tx_info)) {
+                       cl_tx_amsdu_flush_sub_frames(cl_hw, sw_txhdr);
+               } else {
+                       if (tx_queue->type == QUEUE_TYPE_SINGLE)
+                               cl_tx_single_free_skb(cl_hw, sw_txhdr->skb);
+                       else
+                               kfree_skb(sw_txhdr->skb);
+
+                       cl_sw_txhdr_free(cl_hw, sw_txhdr);
+                       cl_hw->tx_packet_cntr.drop.queue_flush++;
+               }
+       }
+
+       /* Remove from schedule queue */
+       cl_txq_sched_list_remove(tx_queue);
+
+       /* Sanity check that queue is empty */
+       WARN_ON(tx_queue->num_packets > 0);
+}
+
+void cl_txq_flush_single(struct cl_hw *cl_hw, u16 idx)
+{
+       spin_lock_bh(&cl_hw->tx_lock_single);
+       cl_txq_flush(cl_hw, &cl_hw->tx_queues.single[idx]);
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+void cl_txq_flush_all_agg(struct cl_hw *cl_hw)
+{
+       int i = 0;
+
+       for (i = 0; i < IPC_MAX_BA_SESSIONS; i++)
+               cl_txq_flush(cl_hw, &cl_hw->tx_queues.agg[i]);
+}
+
+void cl_txq_flush_all_single(struct cl_hw *cl_hw)
+{
+       int i = 0;
+
+       for (i = 0; i < MAX_SINGLE_QUEUES; i++)
+               cl_txq_flush(cl_hw, &cl_hw->tx_queues.single[i]);
+}
+
+void cl_txq_flush_sta(struct cl_hw *cl_hw, struct cl_sta *cl_sta)
+{
+       int i = 0;
+       u8 sta_idx = cl_sta->sta_idx;
+       u32 queue_idx = 0;
+       struct cl_tx_queue *tx_queue = NULL;
+
+       spin_lock_bh(&cl_hw->tx_lock_agg);
+
+       /* Flush all aggregation queues for this station */
+       for (i = 0; i < IEEE80211_NUM_TIDS; i++)
+               if (cl_sta->agg_tx_queues[i])
+                       cl_txq_flush(cl_hw, cl_sta->agg_tx_queues[i]);
+
+       spin_unlock_bh(&cl_hw->tx_lock_agg);
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+
+       /* Flush all single queues for this station */
+       for (i = 0; i < AC_MAX; i++) {
+               queue_idx = QUEUE_IDX(sta_idx, i);
+               tx_queue = &cl_hw->tx_queues.single[queue_idx];
+               cl_txq_flush(cl_hw, tx_queue);
+               cl_txq_reset_counters(tx_queue);
+       }
+
+       /* Go over high prioirty queue and delete packets belonging to this station */
+       cl_txq_delete_packets(cl_hw, &cl_hw->tx_queues.single[HIGH_PRIORITY_QUEUE], sta_idx);
+
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+}
+
+void cl_txq_agg_request_add(struct cl_hw *cl_hw, u8 sta_idx, u8 tid)
+{
+       int i = cl_txq_request_find(cl_hw, sta_idx, tid);
+       struct cl_req_agg_db *req_agg_db = NULL;
+
+       if (i != -1) {
+               cl_dbg_trace(cl_hw, "ALREADY_ADDED - entry = %d, sta_idx = %u, tid = %u\n",
+                            i, sta_idx, tid);
+               return;
+       }
+
+       for (i = 0; i < IPC_MAX_BA_SESSIONS; i++) {
+               req_agg_db = &cl_hw->req_agg_db[i];
+
+               if (!req_agg_db->is_used) {
+                       cl_dbg_trace(cl_hw, "ADD - entry = %d, sta_idx = %u, tid = %u\n",
+                                    i, sta_idx, tid);
+                       req_agg_db->is_used = true;
+                       req_agg_db->sta_idx = sta_idx;
+                       req_agg_db->tid = tid;
+                       cl_hw->req_agg_queues++;
+                       return;
+               }
+       }
+}
+
+void cl_txq_agg_request_del(struct cl_hw *cl_hw, u8 sta_idx, u8 tid)
+{
+       int i = cl_txq_request_find(cl_hw, sta_idx, tid);
+
+       if (i != -1) {
+               cl_dbg_trace(cl_hw, "DEL - entry = %d, sta_idx = %u, tid = %u\n",
+                            i, sta_idx, tid);
+               cl_hw->req_agg_db[i].is_used = false;
+               cl_hw->req_agg_queues--;
+       }
+}
+
+bool cl_txq_is_agg_available(struct cl_hw *cl_hw)
+{
+       u8 total_agg_queues = cl_hw->used_agg_queues + cl_hw->req_agg_queues;
+
+       return (total_agg_queues < IPC_MAX_BA_SESSIONS);
+}
+
+bool cl_txq_single_is_full(struct cl_hw *cl_hw, u16 idx)
+{
+       struct cl_tx_queue *tx_queue = &cl_hw->tx_queues.single[idx];
+       bool is_full = 0;
+
+       spin_lock_bh(&cl_hw->tx_lock_single);
+       is_full = (tx_queue->max_packets == tx_queue->num_packets);
+       spin_unlock_bh(&cl_hw->tx_lock_single);
+
+       return is_full;
+}
+
+void cl_txq_single_sched(struct cl_hw *cl_hw, u16 idx)
+{
+       /*
+        * Don't take lock because it is already taken by
+        * all functions that call cl_txq_single_sched().
+        */
+       struct cl_tx_queue *tx_queue = &cl_hw->tx_queues.single[idx];
+
+       if (tx_queue->num_packets)
+               cl_txq_sched(cl_hw, tx_queue);
+}
+
+bool cl_txq_is_fw_empty(struct cl_tx_queue *tx_queue)
+{
+       return (tx_queue->fw_free_space == tx_queue->fw_max_size);
+}
+
+bool cl_txq_is_fw_full(struct cl_tx_queue *tx_queue)
+{
+       return (tx_queue->fw_free_space == 0);
+}
+
+bool cl_txq_frames_pending(struct cl_hw *cl_hw)
+{
+       int i = 0;
+
+       /* Check if we have multicast/bradcast frame in FW queues */
+       if (!cl_txq_is_fw_empty(&cl_hw->tx_queues.bcmc))
+               return true;
+
+       /* Check if we have singles frame in FW queues */
+       for (i = 0; i < MAX_SINGLE_QUEUES; i++)
+               if (!cl_txq_is_fw_empty(&cl_hw->tx_queues.single[i]))
+                       return true;
+
+       /* Check if we have aggregation frame in FW queues */
+       for (i = 0; i < IPC_MAX_BA_SESSIONS; i++)
+               if (!cl_txq_is_fw_empty(&cl_hw->tx_queues.agg[i]))
+                       return true;
+
+       return false;
+}
+
+int cl_txq_cli(struct cl_hw *cl_hw, struct cli_params *cli_params)
+{
+       switch (cli_params->option) {
+       case 'a':
+               return cl_txq_traffic_counters_print(cl_hw);
+       case 'b':
+               return cl_txq_drop_reasons_print(cl_hw);
+       case 'c':
+               return cl_txq_global_counters_print(cl_hw);
+       case 'd':
+               return cl_txq_stop_reasons_print(cl_hw);
+       case 'e':
+               return cl_txq_requested_agg_print(cl_hw);
+       case 'f':
+               return cl_txq_amsdu_stats_print(cl_hw);
+       case 'm':
+               return cl_txq_max_size_print(cl_hw);
+       case 'r':
+               cl_txq_stats_reset(cl_hw);
+               break;
+       case 's':
+               return cl_txq_sched_list_print(cl_hw);
+       case '?':
+               return cl_txq_cli_help(cl_hw);
+       default:
+               cl_dbg_err(cl_hw, "Illegal option (%c) - try '?' for help\n", cli_params->option);
+               break;
+       }
+
+       return 0;
+}