wifi: ath12k: subscribe required word mask from rx tlv

Most of the RX descriptors fields are currently not used in the
ath12k driver. Hence add support to selectively subscribe to the
required quad words (64 bits) within msdu_end and mpdu_start of
rx_desc.

Add compact rx_desc structures and configure the bit mask for Rx TLVs
(msdu_end, mpdu_start, mpdu_end) via registers. With these registers
SW can configure to DMA the partial TLV struct to Rx buffer.

Each TLV type has its own register to configure the mask value.
The mask value configured in register will indicate if a particular
QWORD has to be written to rx buffer or not i.e., if Nth bit is enabled
in the mask Nth QWORD will be written and it will not be written if the
bit is disabled in mask. While 0th bit indicates whether TLV tag will be
written or not.

Advantages of Qword subscription of TLVs
- Avoid multiple cache-line misses as the all the required fields
of the TLV are within 128 bytes.
- Memory optimization as TLVs + DATA + SHINFO can fit in 2k buffer
even for 64 bit kernel.

Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.1.1-00188-QCAHKSWPL_SILICONZ-1
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.0.c5-00481-QCAHMTSWPL_V1.0_V2.0_SILICONZ-3
Signed-off-by: Karthikeyan Kathirvel <quic_kathirve@quicinc.com>
Co-developed-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Signed-off-by: Raj Kumar Bhagat <quic_rajkbhag@quicinc.com>
Acked-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://msgid.link/20240129065724.2310207-10-quic_rajkbhag@quicinc.com

drivers/net/wireless/ath/ath12k/dp.c

@@ -997,8 +997,25 @@ void ath12k_dp_pdev_pre_alloc(struct ath12k_base *ab)
 	}
 }
 
+bool ath12k_dp_wmask_compaction_rx_tlv_supported(struct ath12k_base *ab)
+{
+	if (test_bit(WMI_TLV_SERVICE_WMSK_COMPACTION_RX_TLVS, ab->wmi_ab.svc_map) &&
+	    ab->hw_params->hal_ops->rxdma_ring_wmask_rx_mpdu_start &&
+	    ab->hw_params->hal_ops->rxdma_ring_wmask_rx_msdu_end &&
+	    ab->hw_params->hal_ops->get_hal_rx_compact_ops) {
+		return true;
+	}
+	return false;
+}
+
 void ath12k_dp_hal_rx_desc_init(struct ath12k_base *ab)
 {
+	if (ath12k_dp_wmask_compaction_rx_tlv_supported(ab)) {
+		/* RX TLVS compaction is supported, hence change the hal_rx_ops
+		 * to compact hal_rx_ops.
+		 */
+		ab->hal_rx_ops = ab->hw_params->hal_ops->get_hal_rx_compact_ops();
+	}
 	ab->hal.hal_desc_sz =
 		ab->hal_rx_ops->rx_desc_get_desc_size();
 }

drivers/net/wireless/ath/ath12k/dp.h

@@ -766,6 +766,11 @@ enum htt_stats_internal_ppdu_frametype {
 #define HTT_RX_RING_SELECTION_CFG_RX_MSDU_START_OFFSET  GENMASK(31, 16)
 #define HTT_RX_RING_SELECTION_CFG_RX_ATTENTION_OFFSET   GENMASK(15, 0)
 
+#define HTT_RX_RING_SELECTION_CFG_WORD_MASK_COMPACT_SET	BIT(23)
+#define HTT_RX_RING_SELECTION_CFG_RX_MPDU_START_MASK	GENMASK(15, 0)
+#define HTT_RX_RING_SELECTION_CFG_RX_MPDU_END_MASK	GENMASK(18, 16)
+#define HTT_RX_RING_SELECTION_CFG_RX_MSDU_END_MASK	GENMASK(16, 0)
+
 enum htt_rx_filter_tlv_flags {
 	HTT_RX_FILTER_TLV_FLAGS_MPDU_START		= BIT(0),
 	HTT_RX_FILTER_TLV_FLAGS_MSDU_START		= BIT(1),
@@ -1089,6 +1094,11 @@ struct htt_rx_ring_selection_cfg_cmd {
 	__le32 rx_mpdu_offset;
 	__le32 rx_msdu_offset;
 	__le32 rx_attn_offset;
+	__le32 info2;
+	__le32 reserved[2];
+	__le32 rx_mpdu_start_end_mask;
+	__le32 rx_msdu_end_word_mask;
+	__le32 info3;
 } __packed;
 
 struct htt_rx_ring_tlv_filter {
@@ -1105,6 +1115,9 @@ struct htt_rx_ring_tlv_filter {
 	u16 rx_msdu_end_offset;
 	u16 rx_msdu_start_offset;
 	u16 rx_attn_offset;
+	u16 rx_mpdu_start_wmask;
+	u16 rx_mpdu_end_wmask;
+	u32 rx_msdu_end_wmask;
 };
 
 #define HTT_STATS_FRAME_CTRL_TYPE_MGMT  0x0
@@ -1821,5 +1834,6 @@ struct ath12k_rx_desc_info *ath12k_dp_get_rx_desc(struct ath12k_base *ab,
 						  u32 cookie);
 struct ath12k_tx_desc_info *ath12k_dp_get_tx_desc(struct ath12k_base *ab,
 						  u32 desc_id);
+bool ath12k_dp_wmask_compaction_rx_tlv_supported(struct ath12k_base *ab);
 void ath12k_dp_hal_rx_desc_init(struct ath12k_base *ab);
 #endif

drivers/net/wireless/ath/ath12k/dp_rx.c

@@ -3939,10 +3939,16 @@ int ath12k_dp_rxdma_ring_sel_config_qcn9274(struct ath12k_base *ab)
 	tlv_filter.rx_msdu_end_offset =
 		ab->hal_rx_ops->rx_desc_get_msdu_end_offset();
 
-	/* TODO: Selectively subscribe to required qwords within msdu_end
-	 * and mpdu_start and setup the mask in below msg
-	 * and modify the rx_desc struct
-	 */
+	if (ath12k_dp_wmask_compaction_rx_tlv_supported(ab)) {
+		tlv_filter.rx_mpdu_start_wmask =
+			ab->hw_params->hal_ops->rxdma_ring_wmask_rx_mpdu_start();
+		tlv_filter.rx_msdu_end_wmask =
+			ab->hw_params->hal_ops->rxdma_ring_wmask_rx_msdu_end();
+		ath12k_dbg(ab, ATH12K_DBG_DATA,
+			   "Configuring compact tlv masks rx_mpdu_start_wmask 0x%x rx_msdu_end_wmask 0x%x\n",
+			   tlv_filter.rx_mpdu_start_wmask, tlv_filter.rx_msdu_end_wmask);
+	}
+
 	ret = ath12k_dp_tx_htt_rx_filter_setup(ab, ring_id, 0,
 					       HAL_RXDMA_BUF,
 					       DP_RXDMA_REFILL_RING_SIZE,

drivers/net/wireless/ath/ath12k/dp_tx.c

@@ -964,6 +964,26 @@ int ath12k_dp_tx_htt_rx_filter_setup(struct ath12k_base *ab, u32 ring_id,
 					 HTT_RX_RING_SELECTION_CFG_RX_ATTENTION_OFFSET);
 	}
 
+	if (tlv_filter->rx_mpdu_start_wmask > 0 &&
+	    tlv_filter->rx_msdu_end_wmask > 0) {
+		cmd->info2 |=
+			le32_encode_bits(true,
+					 HTT_RX_RING_SELECTION_CFG_WORD_MASK_COMPACT_SET);
+		cmd->rx_mpdu_start_end_mask =
+			le32_encode_bits(tlv_filter->rx_mpdu_start_wmask,
+					 HTT_RX_RING_SELECTION_CFG_RX_MPDU_START_MASK);
+		/* mpdu_end is not used for any hardwares so far
+		 * please assign it in future if any chip is
+		 * using through hal ops
+		 */
+		cmd->rx_mpdu_start_end_mask |=
+			le32_encode_bits(tlv_filter->rx_mpdu_end_wmask,
+					 HTT_RX_RING_SELECTION_CFG_RX_MPDU_END_MASK);
+		cmd->rx_msdu_end_word_mask =
+			le32_encode_bits(tlv_filter->rx_msdu_end_wmask,
+					 HTT_RX_RING_SELECTION_CFG_RX_MSDU_END_MASK);
+	}
+
 	ret = ath12k_htc_send(&ab->htc, ab->dp.eid, skb);
 	if (ret)
 		goto err_free;

drivers/net/wireless/ath/ath12k/hal.h

@@ -1082,6 +1082,9 @@ struct hal_rx_ops {
 
 struct hal_ops {
 	int (*create_srng_config)(struct ath12k_base *ab);
+	u16 (*rxdma_ring_wmask_rx_mpdu_start)(void);
+	u32 (*rxdma_ring_wmask_rx_msdu_end)(void);
+	const struct hal_rx_ops *(*get_hal_rx_compact_ops)(void);
 	const struct ath12k_hal_tcl_to_wbm_rbm_map *tcl_to_wbm_rbm_map;
 };
 
@@ -1089,6 +1092,7 @@ extern const struct hal_ops hal_qcn9274_ops;
 extern const struct hal_ops hal_wcn7850_ops;
 
 extern const struct hal_rx_ops hal_rx_qcn9274_ops;
+extern const struct hal_rx_ops hal_rx_qcn9274_compact_ops;
 extern const struct hal_rx_ops hal_rx_wcn7850_ops;
 
 u32 ath12k_hal_reo_qdesc_size(u32 ba_window_size, u8 tid);

drivers/net/wireless/ath/ath12k/rx_desc.h

 /* SPDX-License-Identifier: BSD-3-Clause-Clear */
 /*
  * Copyright (c) 2018-2021 The Linux Foundation. All rights reserved.
- * Copyright (c) 2021-2023 Qualcomm Innovation Center, Inc. All rights reserved.
+ * Copyright (c) 2021-2024 Qualcomm Innovation Center, Inc. All rights reserved.
  */
 #ifndef ATH12K_RX_DESC_H
 #define ATH12K_RX_DESC_H
@@ -147,6 +147,61 @@ struct rx_mpdu_start_qcn9274 {
 	__le32 res1;
 } __packed;
 
+#define QCN9274_MPDU_START_SELECT_MPDU_START_TAG			BIT(0)
+#define QCN9274_MPDU_START_SELECT_INFO0_REO_QUEUE_DESC_LO		BIT(1)
+#define QCN9274_MPDU_START_SELECT_INFO1_PN_31_0				BIT(2)
+#define QCN9274_MPDU_START_SELECT_PN_95_32				BIT(3)
+#define QCN9274_MPDU_START_SELECT_PN_127_96_INFO2			BIT(4)
+#define QCN9274_MPDU_START_SELECT_PEER_MDATA_INFO3_PHY_PPDU_ID		BIT(5)
+#define QCN9274_MPDU_START_SELECT_AST_IDX_SW_PEER_ID_INFO4		BIT(6)
+#define QCN9274_MPDU_START_SELECT_INFO5_INFO6				BIT(7)
+#define QCN9274_MPDU_START_SELECT_FRAME_CTRL_DURATION_ADDR1_31_0	BIT(8)
+#define QCN9274_MPDU_START_SELECT_ADDR2_47_0_ADDR1_47_32		BIT(9)
+#define QCN9274_MPDU_START_SELECT_ADDR3_47_0_SEQ_CTRL			BIT(10)
+#define QCN9274_MPDU_START_SELECT_ADDR4_47_0_QOS_CTRL			BIT(11)
+#define QCN9274_MPDU_START_SELECT_HT_CTRL_INFO7				BIT(12)
+#define QCN9274_MPDU_START_SELECT_ML_ADDR1_47_0_ML_ADDR2_15_0		BIT(13)
+#define QCN9274_MPDU_START_SELECT_ML_ADDR2_47_16_INFO8			BIT(14)
+#define QCN9274_MPDU_START_SELECT_RES_0_RES_1				BIT(15)
+
+#define QCN9274_MPDU_START_WMASK (QCN9274_MPDU_START_SELECT_INFO1_PN_31_0 |	\
+		QCN9274_MPDU_START_SELECT_PN_95_32 |				\
+		QCN9274_MPDU_START_SELECT_PN_127_96_INFO2 |			\
+		QCN9274_MPDU_START_SELECT_PEER_MDATA_INFO3_PHY_PPDU_ID |	\
+		QCN9274_MPDU_START_SELECT_AST_IDX_SW_PEER_ID_INFO4 |		\
+		QCN9274_MPDU_START_SELECT_INFO5_INFO6 |				\
+		QCN9274_MPDU_START_SELECT_FRAME_CTRL_DURATION_ADDR1_31_0 |	\
+		QCN9274_MPDU_START_SELECT_ADDR2_47_0_ADDR1_47_32 |		\
+		QCN9274_MPDU_START_SELECT_ADDR3_47_0_SEQ_CTRL |			\
+		QCN9274_MPDU_START_SELECT_ADDR4_47_0_QOS_CTRL)
+
+/* The below rx_mpdu_start_qcn9274_compact structure is tied with the mask
+ * value QCN9274_MPDU_START_WMASK. If the mask value changes the structure
+ * will also change.
+ */
+
+struct rx_mpdu_start_qcn9274_compact {
+	__le32 info1;
+	__le32 pn[4];
+	__le32 info2;
+	__le32 peer_meta_data;
+	__le16 info3;
+	__le16 phy_ppdu_id;
+	__le16 ast_index;
+	__le16 sw_peer_id;
+	__le32 info4;
+	__le32 info5;
+	__le32 info6;
+	__le16 frame_ctrl;
+	__le16 duration;
+	u8 addr1[ETH_ALEN];
+	u8 addr2[ETH_ALEN];
+	u8 addr3[ETH_ALEN];
+	__le16 seq_ctrl;
+	u8 addr4[ETH_ALEN];
+	__le16 qos_ctrl;
+} __packed;
+
 /* rx_mpdu_start
  *
  * reo_destination_indication
@@ -786,6 +841,52 @@ struct rx_msdu_end_qcn9274 {
 	__le32 info14;
 } __packed;
 
+#define QCN9274_MSDU_END_SELECT_MSDU_END_TAG				BIT(0)
+#define QCN9274_MSDU_END_SELECT_INFO0_PHY_PPDUID_IP_HDR_CSUM_INFO1	BIT(1)
+#define QCN9274_MSDU_END_SELECT_INFO2_CUMULATIVE_CSUM_RULE_IND_0	BIT(2)
+#define QCN9274_MSDU_END_SELECT_IPV6_OP_CRC_INFO3_TYPE13		BIT(3)
+#define QCN9274_MSDU_END_SELECT_RULE_IND_1_TCP_SEQ_NUM			BIT(4)
+#define QCN9274_MSDU_END_SELECT_TCP_ACK_NUM_INFO4_WINDOW_SIZE		BIT(5)
+#define QCN9274_MSDU_END_SELECT_SA_SW_PER_ID_INFO5_SA_DA_ID		BIT(6)
+#define QCN9274_MSDU_END_SELECT_INFO6_FSE_METADATA			BIT(7)
+#define QCN9274_MSDU_END_SELECT_CCE_MDATA_TCP_UDP_CSUM_INFO7_IP_LEN	BIT(8)
+#define QCN9274_MSDU_END_SELECT_INFO8_INFO9				BIT(9)
+#define QCN9274_MSDU_END_SELECT_INFO10_INFO11				BIT(10)
+#define QCN9274_MSDU_END_SELECT_VLAN_CTAG_STAG_CI_PEER_MDATA		BIT(11)
+#define QCN9274_MSDU_END_SELECT_INFO12_AND_FLOW_ID_TOEPLITZ		BIT(12)
+#define QCN9274_MSDU_END_SELECT_PPDU_START_TS_63_32_PHY_MDATA		BIT(13)
+#define QCN9274_MSDU_END_SELECT_PPDU_START_TS_31_0_TOEPLITZ_HASH_2_4	BIT(14)
+#define QCN9274_MSDU_END_SELECT_RES0_SA_47_0				BIT(15)
+#define QCN9274_MSDU_END_SELECT_INFO13_INFO14				BIT(16)
+
+#define QCN9274_MSDU_END_WMASK (QCN9274_MSDU_END_SELECT_MSDU_END_TAG |	\
+		QCN9274_MSDU_END_SELECT_SA_SW_PER_ID_INFO5_SA_DA_ID |	\
+		QCN9274_MSDU_END_SELECT_INFO10_INFO11 |			\
+		QCN9274_MSDU_END_SELECT_INFO12_AND_FLOW_ID_TOEPLITZ |	\
+		QCN9274_MSDU_END_SELECT_PPDU_START_TS_63_32_PHY_MDATA |	\
+		QCN9274_MSDU_END_SELECT_INFO13_INFO14)
+
+/* The below rx_msdu_end_qcn9274_compact structure is tied with the mask value
+ * QCN9274_MSDU_END_WMASK. If the mask value changes the structure will also
+ * change.
+ */
+
+struct rx_msdu_end_qcn9274_compact {
+	__le64 msdu_end_tag;
+	__le16 sa_sw_peer_id;
+	__le16 info5;
+	__le16 sa_idx;
+	__le16 da_idx_or_sw_peer_id;
+	__le32 info10;
+	__le32 info11;
+	__le32 info12;
+	__le32 flow_id_toeplitz;
+	__le32 ppdu_start_timestamp_63_32;
+	__le32 phy_meta_data;
+	__le32 info13;
+	__le32 info14;
+} __packed;
+
 /* These macro definitions are only used for WCN7850 */
 #define RX_MSDU_END_WCN7850_INFO2_KEY_ID			BIT(7, 0)
 
@@ -1450,16 +1551,18 @@ struct rx_msdu_end_wcn7850 {
  *
  */
 
-/* TODO: Move to compact TLV approach
- * By default these tlv's are not aligned to 128b boundary
- * Need to remove unused qwords and make them compact/aligned
- */
 struct hal_rx_desc_qcn9274 {
 	struct rx_msdu_end_qcn9274 msdu_end;
 	struct rx_mpdu_start_qcn9274 mpdu_start;
 	u8 msdu_payload[];
 } __packed;
 
+struct hal_rx_desc_qcn9274_compact {
+	struct rx_msdu_end_qcn9274_compact msdu_end;
+	struct rx_mpdu_start_qcn9274_compact mpdu_start;
+	u8 msdu_payload[];
+} __packed;
+
 #define RX_BE_PADDING0_BYTES 8
 #define RX_BE_PADDING1_BYTES 8
 
@@ -1484,6 +1587,7 @@ struct hal_rx_desc_wcn7850 {
 struct hal_rx_desc {
 	union {
 		struct hal_rx_desc_qcn9274 qcn9274;
+		struct hal_rx_desc_qcn9274_compact qcn9274_compact;
 		struct hal_rx_desc_wcn7850 wcn7850;
 	} u;
 } __packed;

drivers/net/wireless/ath/ath12k/wmi.h

@@ -2163,6 +2163,8 @@ enum wmi_tlv_service {
 
 	WMI_TLV_SERVICE_11BE = 289,
 
+	WMI_TLV_SERVICE_WMSK_COMPACTION_RX_TLVS = 361,
+
 	WMI_TLV_SERVICE_PEER_METADATA_V1A_V1B_SUPPORT = 365,
 
 	WMI_MAX_EXT2_SERVICE,