From: Viktor Barna <[email protected]>
(Part of the split. Please, take a look at the cover letter for more
details).
Signed-off-by: Viktor Barna <[email protected]>
---
drivers/net/wireless/celeno/cl8k/ampdu.c | 331 +++++++++++++++++++++++
1 file changed, 331 insertions(+)
create mode 100644 drivers/net/wireless/celeno/cl8k/ampdu.c
diff --git a/drivers/net/wireless/celeno/cl8k/ampdu.c b/drivers/net/wireless/celeno/cl8k/ampdu.c
new file mode 100644
index 000000000000..6609bc62d340
--- /dev/null
+++ b/drivers/net/wireless/celeno/cl8k/ampdu.c
@@ -0,0 +1,331 @@
+// SPDX-License-Identifier: GPL-2.0 OR BSD-2-Clause
+/* Copyright(c) 2019-2022, Celeno Communications Ltd. */
+
+#include "hw.h"
+#include "recovery.h"
+#include "utils.h"
+#include "ampdu.h"
+
+int cl_ampdu_rx_start(struct cl_hw *cl_hw,
+ struct cl_sta *cl_sta,
+ u16 tid,
+ u16 ssn,
+ u16 buf_size)
+{
+ /* @IEEE80211_AMPDU_RX_START: start RX aggregation */
+ if (!cl_hw->conf->ci_agg_rx)
+ return -EOPNOTSUPP;
+
+ cl_dbg_trace(cl_hw, "sta_idx [%u] tid [%u]\n", cl_sta->sta_idx, tid);
+
+ buf_size = min(buf_size, cl_hw->conf->ce_max_agg_size_rx);
+
+ if (cl_hw->conf->ci_fast_rx_en)
+ cl_rx_reorder_init(cl_hw, cl_sta, tid, buf_size);
+
+ cl_msg_tx_ba_add(cl_hw, BA_AGMT_RX, cl_sta->sta_idx, tid, buf_size, ssn);
+
+ return 0;
+}
+
+void cl_ampdu_rx_stop(struct cl_hw *cl_hw,
+ struct cl_sta *cl_sta,
+ u16 tid)
+{
+ /* @IEEE80211_AMPDU_RX_STOP: stop RX aggregation */
+ cl_dbg_trace(cl_hw, "sta_idx [%u] tid [%u]\n", cl_sta->sta_idx, tid);
+
+ if (cl_hw->conf->ci_fast_rx_en)
+ cl_rx_reorder_close(cl_sta, tid);
+}
+
+int cl_ampdu_tx_start(struct cl_hw *cl_hw,
+ struct ieee80211_vif *vif,
+ struct cl_sta *cl_sta,
+ u16 tid,
+ u16 ssn)
+{
+ /* @IEEE80211_AMPDU_TX_START: start TX aggregation */
+ struct mm_available_ba_txq_cfm *cfm = NULL;
+ int ret = 0;
+
+ if (!ieee80211_hw_check(cl_hw->hw, AMPDU_AGGREGATION) || !cl_hw->conf->ci_agg_tx)
+ return -ENOTSUPP;
+
+ if (!cl_txq_is_agg_available(cl_hw)) {
+ cl_dbg_warn(cl_hw, "No free aggregation queue for sta_idx [%u] tid [%u]\n",
+ cl_sta->sta_idx, tid);
+ return -1;
+ }
+
+ ret = cl_msg_tx_available_ba_txq(cl_hw, cl_sta->sta_idx, tid);
+ if (ret)
+ return ret;
+
+ /* Read FW confirm message */
+ cfm = (struct mm_available_ba_txq_cfm *)(cl_hw->msg_cfm_params[MM_AVAILABLE_BA_TXQ_CFM]);
+ if (!cfm)
+ return -ENOMSG;
+
+ /* Check if status is valid */
+ if (cfm->status != BA_TXQUEUE_INVALID && cfm->status != BA_TXQUEUE_VALID) {
+ cl_dbg_verbose(cl_hw, "Status Error (%u)\n", cfm->status);
+ cl_msg_tx_free_cfm_params(cl_hw, MM_AVAILABLE_BA_TXQ_CFM);
+ return -EIO;
+ }
+
+ if (cfm->status == BA_TXQUEUE_INVALID) {
+ cl_dbg_warn(cl_hw, "BA_TXQUEUE_INVALID - sta_idx [%u] tid [%u]\n",
+ cfm->sta_idx, cfm->tid);
+ cl_msg_tx_free_cfm_params(cl_hw, MM_AVAILABLE_BA_TXQ_CFM);
+ return -1;
+ }
+
+ cl_msg_tx_free_cfm_params(cl_hw, MM_AVAILABLE_BA_TXQ_CFM);
+ cl_txq_agg_request_add(cl_hw, cl_sta->sta_idx, tid);
+ cl_baw_start(&cl_sta->baws[tid], ssn);
+
+ /* Mandatory callback once setup preparations are done at lower level */
+ ieee80211_start_tx_ba_cb_irqsafe(vif, cl_sta->addr, tid);
+
+ return 0;
+}
+
+int cl_ampdu_tx_operational(struct cl_hw *cl_hw,
+ struct cl_sta *cl_sta,
+ u16 tid,
+ u16 buf_size,
+ bool amsdu_supported)
+{
+ /* @IEEE80211_AMPDU_TX_OPERATIONAL: TX aggregation has become operational */
+ struct mm_ba_add_cfm *cfm = NULL;
+ struct cl_baw *baw = &cl_sta->baws[tid];
+ u16 ssn = baw->ssn;
+ int ret = 0;
+
+ buf_size = min(buf_size, cl_hw->conf->ce_max_agg_size_tx);
+
+ /* Send MM_BA_ADD_TX_REQ message to firmware */
+ ret = cl_msg_tx_ba_add(cl_hw, BA_AGMT_TX, cl_sta->sta_idx, tid, buf_size, ssn);
+ if (ret)
+ return ret;
+
+ /* Handle message confirmation */
+ cfm = (struct mm_ba_add_cfm *)(cl_hw->msg_cfm_params[MM_BA_ADD_TX_CFM]);
+ if (!cfm)
+ return -ENOMSG;
+
+ if (cfm->status != BA_AGMT_ESTABLISHED) {
+ cl_dbg_verbose(cl_hw, "Status Error (%u)\n", cfm->status);
+ cl_msg_tx_free_cfm_params(cl_hw, MM_BA_ADD_TX_CFM);
+ cl_txq_agg_request_del(cl_hw, cl_sta->sta_idx, tid);
+ return -EIO;
+ }
+
+ cl_baw_operational(cl_hw, baw, cfm->agg_idx, amsdu_supported);
+ cl_agg_cfm_set_ssn(cl_hw, ssn, cfm->agg_idx);
+ cl_hw->ipc_env->ring_indices_elem->indices->new_ssn_idx[cfm->agg_idx] = cpu_to_le32(ssn);
+
+ if (amsdu_supported)
+ cl_tx_amsdu_set_max_len(cl_hw, cl_sta, tid);
+ else
+ cl_dbg_trace(cl_hw, "AMSDU not supported - sta_idx=%u\n", cl_sta->sta_idx);
+
+ cl_txq_agg_alloc(cl_hw, cl_sta, cfm, buf_size);
+ cl_msg_tx_free_cfm_params(cl_hw, MM_BA_ADD_TX_CFM);
+
+ return 0;
+}
+
+void _cl_ampdu_tx_stop(struct cl_hw *cl_hw,
+ struct cl_tx_queue *tx_queue,
+ struct cl_sta *cl_sta,
+ u8 tid)
+{
+ struct mm_ba_del_cfm *cfm = NULL;
+ u8 fw_agg_idx = tx_queue->index;
+
+ if (cl_recovery_in_progress(cl_hw))
+ goto out;
+
+ /*
+ * TX stop flow:
+ * 1) Flush TX queues - done in cl_ampdu_tx_stop()
+ * 2) Poll confirmation queue and clear enhanced TIM
+ * 3) Send MM_STA_DEL_REQ message to firmware
+ * 4) Poll again confirmation and flush confirmation queue
+ * 5) Reset write index
+ */
+ cl_agg_cfm_poll_empty(cl_hw, fw_agg_idx, false);
+
+ /* Send MM_BA_DEL_REQ message to firmware */
+ if (cl_msg_tx_ba_del(cl_hw, cl_sta->sta_idx, tid))
+ goto out;
+
+ cfm = (struct mm_ba_del_cfm *)(cl_hw->msg_cfm_params[MM_BA_DEL_CFM]);
+ if (!cfm) {
+ cl_dbg_err(cl_hw, "Unable to fetch CFM\n");
+ goto out;
+ }
+
+ /* Check confirmation status */
+ if (cfm->status != BA_AGMT_DELETED && cfm->status != BA_AGMT_DOES_NOT_EXIST)
+ cl_dbg_verbose(cl_hw, "Status Error (%u)\n", cfm->status);
+
+ cl_msg_tx_free_cfm_params(cl_hw, MM_BA_DEL_CFM);
+
+out:
+ cl_agg_cfm_poll_empty(cl_hw, fw_agg_idx, true);
+ cl_txq_agg_free(cl_hw, tx_queue, cl_sta, tid);
+
+ /* 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;
+}
+
+int cl_ampdu_tx_stop(struct cl_hw *cl_hw,
+ struct ieee80211_vif *vif,
+ enum ieee80211_ampdu_mlme_action action,
+ struct cl_sta *cl_sta,
+ u16 tid)
+{
+ /*
+ * @IEEE80211_AMPDU_TX_STOP_CONT: stop TX aggregation but continue transmitting
+ * queued packets, now unaggregated. After all packets are transmitted the
+ * driver has to call ieee80211_stop_tx_ba_cb_irqsafe().
+ * @IEEE80211_AMPDU_TX_STOP_FLUSH: stop TX aggregation and flush all packets,
+ * called when the station is removed. There's no need or reason to call
+ * ieee80211_stop_tx_ba_cb_irqsafe() in this case as mac80211 assumes the
+ * session is gone and removes the station.
+ * @IEEE80211_AMPDU_TX_STOP_FLUSH_CONT: called when TX aggregation is stopped
+ * but the driver hasn't called ieee80211_stop_tx_ba_cb_irqsafe() yet and
+ * now the connection is dropped and the station will be removed. Drivers
+ * should clean up and drop remaining packets when this is called.
+ */
+
+ /* !!! Comment in agg-tx.c, ___ieee80211_stop_tx_ba_session(): !!!
+ * !!! HW shall not deny going back to legacy !!!
+ * !!! Therefore cl_ampdu_tx_stop() always returns 0 !!!
+ */
+
+ struct cl_tx_queue *tx_queue = cl_sta->agg_tx_queues[tid];
+ struct cl_baw *baw = &cl_sta->baws[tid];
+
+ spin_lock_bh(&cl_hw->tx_lock_agg);
+
+ cl_baw_stop(baw);
+ cl_txq_agg_request_del(cl_hw, cl_sta->sta_idx, tid);
+
+ /* Check if BA session exist */
+ if (!tx_queue) {
+ spin_unlock_bh(&cl_hw->tx_lock_agg);
+
+ if (!cl_recovery_in_progress(cl_hw))
+ cl_dbg_warn(cl_hw, "Queue doesn't exist - sta_idx [%u] tid [%u]\n",
+ cl_sta->sta_idx, tid);
+
+ goto out;
+ }
+
+ if (action == IEEE80211_AMPDU_TX_STOP_CONT) {
+ /*
+ * The order of flow here is very important here to avoid reorder problem!
+ * 1) Take single lock to block single traffic
+ * 2) Stop agg traffic.
+ * 3) Transfer agg-to-single and push all skbs from agg queue to single queue.
+ * 4) Transfer BA window pending queue to single queue.
+ * 5) Release single lock
+ */
+ spin_lock_bh(&cl_hw->tx_lock_single);
+ cl_txq_agg_stop(cl_sta, tid);
+ cl_txq_transfer_agg_to_single(cl_hw, tx_queue);
+ cl_baw_pending_to_single(cl_hw, cl_sta, baw);
+ spin_unlock_bh(&cl_hw->tx_lock_single);
+ } else {
+ cl_txq_agg_stop(cl_sta, tid);
+ cl_txq_flush_agg(cl_hw, tx_queue, false);
+ cl_baw_pending_purge(baw);
+ }
+
+ cl_tx_amsdu_anchor_reset(&cl_sta->amsdu_anchor[tid]);
+
+ spin_unlock_bh(&cl_hw->tx_lock_agg);
+
+ _cl_ampdu_tx_stop(cl_hw, tx_queue, cl_sta, tid);
+
+out:
+ /* Mandatory callback once we've made our own tear down ops */
+ if (action != IEEE80211_AMPDU_TX_STOP_FLUSH)
+ ieee80211_stop_tx_ba_cb_irqsafe(vif, cl_sta->addr, tid);
+
+ return 0;
+}
+
+#define HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_OFFSET 3
+#define HE_EXP_MAX 22 /* 2 ^ 22 = 4194304 < 6500631 */
+
+static void _cl_ampdu_size_exp(struct ieee80211_sta *sta,
+ u8 *ampdu_exp_he,
+ u8 *ampdu_exp_vht,
+ u8 *ampdu_exp_ht)
+{
+ struct ieee80211_sta_he_cap *he_cap = &sta->he_cap;
+ u8 mac_cap_info3 = he_cap->he_cap_elem.mac_cap_info[3];
+ u8 he_exp;
+
+ if (sta->ht_cap.ht_supported)
+ *ampdu_exp_ht = IEEE80211_HT_MAX_AMPDU_FACTOR + sta->ht_cap.ampdu_factor;
+
+ if (sta->vht_cap.vht_supported) {
+ u32 vht_exp = (sta->vht_cap.cap &
+ IEEE80211_VHT_CAP_MAX_A_MPDU_LENGTH_EXPONENT_MASK) >>
+ IEEE80211_VHT_CAP_MAX_A_MPDU_LENGTH_EXPONENT_SHIFT;
+
+ *ampdu_exp_vht = IEEE80211_HT_MAX_AMPDU_FACTOR + vht_exp;
+ }
+
+ he_exp = (mac_cap_info3 & IEEE80211_HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_MASK) >>
+ HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_OFFSET;
+
+ if (sta->vht_cap.vht_supported) {
+ if (he_exp)
+ *ampdu_exp_he = min(IEEE80211_HE_VHT_MAX_AMPDU_FACTOR + he_exp, HE_EXP_MAX);
+ else
+ *ampdu_exp_he = *ampdu_exp_vht;
+ } else if (sta->ht_cap.ht_supported) {
+ if (he_exp)
+ *ampdu_exp_he = IEEE80211_HE_HT_MAX_AMPDU_FACTOR + he_exp;
+ else
+ *ampdu_exp_he = *ampdu_exp_ht;
+ }
+}
+
+static void _cl_ampdu_size_exp_6g(struct ieee80211_sta *sta, u8 *ampdu_exp_he)
+{
+ u8 mac_cap_info3 = sta->he_cap.he_cap_elem.mac_cap_info[3];
+ u8 he_exp_ext = (mac_cap_info3 & IEEE80211_HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_MASK) >>
+ HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_OFFSET;
+
+ if (he_exp_ext) {
+ *ampdu_exp_he = min(IEEE80211_HE_VHT_MAX_AMPDU_FACTOR + he_exp_ext, HE_EXP_MAX);
+ } else {
+ struct ieee80211_he_6ghz_capa *he_6g_cap = &sta->he_6ghz_capa;
+ u8 he_exp_6ghz = (le16_to_cpu(he_6g_cap->capa) &
+ HE_6GHZ_CAP_MAX_AMPDU_LEN_EXP_MASK) >>
+ HE_6GHZ_CAP_MAX_AMPDU_LEN_EXP_OFFSET;
+
+ *ampdu_exp_he = min(HE_6GHZ_CAP_MAX_AMPDU_LEN_FACTOR + he_exp_6ghz, HE_EXP_MAX);
+ }
+}
+
+void cl_ampdu_size_exp(struct cl_hw *cl_hw, struct ieee80211_sta *sta,
+ u8 *ampdu_exp_he, u8 *ampdu_exp_vht, u8 *ampdu_exp_ht)
+{
+ if (cl_band_is_6g(cl_hw))
+ _cl_ampdu_size_exp_6g(sta, ampdu_exp_he);
+ else
+ _cl_ampdu_size_exp(sta, ampdu_exp_he, ampdu_exp_vht, ampdu_exp_ht);
+
+ cl_dbg_info(cl_hw, "ampdu_size_exp: he = %u, vht = %u, ht = %u\n",
+ *ampdu_exp_he, *ampdu_exp_vht, *ampdu_exp_ht);
+}
+
--
2.36.1
On Tue, 2022-05-24 at 14:33 +0300, [email protected] wrote:
>
> + if (sta->vht_cap.vht_supported) {
> + u32 vht_exp = (sta->vht_cap.cap &
> + IEEE80211_VHT_CAP_MAX_A_MPDU_LENGTH_EXPONENT_MASK) >>
> + IEEE80211_VHT_CAP_MAX_A_MPDU_LENGTH_EXPONENT_SHIFT;
You can (often) simplify that kind of thing using u32_get_bits()
> + he_exp = (mac_cap_info3 & IEEE80211_HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_MASK) >>
> + HE_MAC_CAP3_MAX_AMPDU_LEN_EXP_OFFSET;
or u8_get_bits(), etc.
There are even le/be versions.
> + struct ieee80211_he_6ghz_capa *he_6g_cap = &sta->he_6ghz_capa;
> + u8 he_exp_6ghz = (le16_to_cpu(he_6g_cap->capa) &
> + HE_6GHZ_CAP_MAX_AMPDU_LEN_EXP_MASK) >>
> + HE_6GHZ_CAP_MAX_AMPDU_LEN_EXP_OFFSET;
which you could use here.
I know we don't use it everywhere yet, but IMHO it's ultimately nicer
code, and I definitely try to use it in new code.
johannes
On Tue, 2022-05-24 at 14:33 +0300, [email protected] wrote:
>
> + if (!ieee80211_hw_check(cl_hw->hw, AMPDU_AGGREGATION) || !cl_hw->conf->ci_agg_tx)
> + return -ENOTSUPP;
> +
> + if (!cl_txq_is_agg_available(cl_hw)) {
> + cl_dbg_warn(cl_hw, "No free aggregation queue for sta_idx [%u] tid [%u]\n",
> + cl_sta->sta_idx, tid);
> + return -1;
>
That'd be -EPERM, which is really not what you want to return - please
use symbolic errors throughout. Maybe -ENOSPC in this case.
johannes