[PATCH v2 3/4] ath10k: Enable SRRI/DRRI support on ddr for WCN3990

[pillair@codeaurora.org]

From: Govind Singh <govinds@codeaurora.org>

SRRI/DRRI are not mapped in the HW Shadow block and can lead
to un-clocked access if common subsystem in the target is
powered down due to idle mode.

To mitigate this problem SRRI/DRRI can be read from
DDR instead of doing an actual hardware read.
Host allocates non cached memory on ddr and configures
the physical address of this memory to the CE hardware.
The hardware updates the RRI on this particular location.
Read SRRI/DRRI from DDR location instead of
direct target read.

Enable retention restore on ddr using hw params to enable
in specific targets.

Signed-off-by: Govind Singh <govinds@codeaurora.org>
Signed-off-by: Rakesh Pillai <pillair@codeaurora.org>
---
 drivers/net/wireless/ath/ath10k/ce.c   | 102 +++++++++++++++++++++++++++++++--
 drivers/net/wireless/ath/ath10k/ce.h   |  10 ++++
 drivers/net/wireless/ath/ath10k/core.c |  14 +++++
 drivers/net/wireless/ath/ath10k/hw.c   |   9 ++-
 drivers/net/wireless/ath/ath10k/hw.h   |  13 ++++-
 drivers/net/wireless/ath/ath10k/snoc.c |   3 +
 6 files changed, 142 insertions(+), 9 deletions(-)

diff --git a/drivers/net/wireless/ath/ath10k/ce.c b/drivers/net/wireless/ath/ath10k/ce.c
index 5053dd92bf01..f6eee5f30a8f 100644
--- a/drivers/net/wireless/ath/ath10k/ce.c
+++ b/drivers/net/wireless/ath/ath10k/ce.c
@@ -185,11 +185,30 @@ static inline u32 ath10k_ce_src_ring_write_index_get(struct ath10k *ar,
 				ar->hw_ce_regs->sr_wr_index_addr);
 }
 
+static inline u32 ath10k_ce_src_ring_read_index_from_ddr(struct ath10k *ar,
+							 u32 ce_id)
+{
+	struct ath10k_ce *ce = ath10k_ce_priv(ar);
+
+	return ce->vaddr_rri[ce_id] & CE_DDR_RRI_MASK;
+}
+
 static inline u32 ath10k_ce_src_ring_read_index_get(struct ath10k *ar,
 						    u32 ce_ctrl_addr)
 {
-	return ath10k_ce_read32(ar, ce_ctrl_addr +
-				ar->hw_ce_regs->current_srri_addr);
+	struct ath10k_ce *ce = ath10k_ce_priv(ar);
+	u32 ce_id = COPY_ENGINE_ID(ce_ctrl_addr);
+	struct ath10k_ce_pipe *ce_state = &ce->ce_states[ce_id];
+	u32 index;
+
+	if (ar->hw_params.rri_on_ddr &&
+	    (ce_state->attr_flags & CE_ATTR_DIS_INTR))
+		index = ath10k_ce_src_ring_read_index_from_ddr(ar, ce_id);
+	else
+		index = ath10k_ce_read32(ar, ce_ctrl_addr +
+					 ar->hw_ce_regs->current_srri_addr);
+
+	return index;
 }
 
 static inline void
@@ -266,11 +285,31 @@ static inline void ath10k_ce_dest_ring_byte_swap_set(struct ath10k *ar,
 			  ath10k_set_ring_byte(n, ctrl_regs->dst_ring));
 }
 
+static inline
+	u32 ath10k_ce_dest_ring_read_index_from_ddr(struct ath10k *ar, u32 ce_id)
+{
+	struct ath10k_ce *ce = ath10k_ce_priv(ar);
+
+	return (ce->vaddr_rri[ce_id] >> CE_DDR_DRRI_SHIFT) &
+		CE_DDR_RRI_MASK;
+}
+
 static inline u32 ath10k_ce_dest_ring_read_index_get(struct ath10k *ar,
 						     u32 ce_ctrl_addr)
 {
-	return ath10k_ce_read32(ar, ce_ctrl_addr +
-				ar->hw_ce_regs->current_drri_addr);
+	struct ath10k_ce *ce = ath10k_ce_priv(ar);
+	u32 ce_id = COPY_ENGINE_ID(ce_ctrl_addr);
+	struct ath10k_ce_pipe *ce_state = &ce->ce_states[ce_id];
+	u32 index;
+
+	if (ar->hw_params.rri_on_ddr &&
+	    (ce_state->attr_flags & CE_ATTR_DIS_INTR))
+		index = ath10k_ce_dest_ring_read_index_from_ddr(ar, ce_id);
+	else
+		index = ath10k_ce_read32(ar, ce_ctrl_addr +
+					 ar->hw_ce_regs->current_drri_addr);
+
+	return index;
 }
 
 static inline void ath10k_ce_dest_ring_base_addr_set(struct ath10k *ar,
@@ -486,7 +525,7 @@ static int _ath10k_ce_send_nolock_64(struct ath10k_ce_pipe *ce_state,
 	struct ath10k_ce_ring *src_ring = ce_state->src_ring;
 	struct ce_desc_64 *desc, sdesc;
 	unsigned int nentries_mask = src_ring->nentries_mask;
-	unsigned int sw_index = src_ring->sw_index;
+	unsigned int sw_index;
 	unsigned int write_index = src_ring->write_index;
 	u32 ctrl_addr = ce_state->ctrl_addr;
 	__le32 *addr;
@@ -500,6 +539,11 @@ static int _ath10k_ce_send_nolock_64(struct ath10k_ce_pipe *ce_state,
 		ath10k_warn(ar, "%s: send more we can (nbytes: %d, max: %d)\n",
 			    __func__, nbytes, ce_state->src_sz_max);
 
+	if (ar->hw_params.rri_on_ddr)
+		sw_index = ath10k_ce_src_ring_read_index_from_ddr(ar, ce_state->id);
+	else
+		sw_index = src_ring->sw_index;
+
 	if (unlikely(CE_RING_DELTA(nentries_mask,
 				   write_index, sw_index - 1) <= 0)) {
 		ret = -ENOSR;
@@ -1016,7 +1060,10 @@ int ath10k_ce_completed_send_next_nolock(struct ath10k_ce_pipe *ce_state,
 		src_ring->hw_index = read_index;
 	}
 
-	read_index = src_ring->hw_index;
+	if (ar->hw_params.rri_on_ddr)
+		read_index = ath10k_ce_src_ring_read_index_get(ar, ctrl_addr);
+	else
+		read_index = src_ring->hw_index;
 
 	if (read_index == sw_index)
 		return -EIO;
@@ -1841,3 +1888,46 @@ int ath10k_ce_alloc_pipe(struct ath10k *ar, int ce_id,
 	return 0;
 }
 EXPORT_SYMBOL(ath10k_ce_alloc_pipe);
+
+void ath10k_ce_alloc_rri(struct ath10k *ar)
+{
+	int i;
+	u32 value;
+	u32 ctrl1_regs;
+	u32 ce_base_addr;
+	struct ath10k_ce *ce = ath10k_ce_priv(ar);
+
+	ce->vaddr_rri = dma_alloc_coherent(ar->dev,
+					   (CE_COUNT * sizeof(u32)),
+					   &ce->paddr_rri, GFP_KERNEL);
+
+	if (!ce->vaddr_rri)
+		return;
+
+	ath10k_ce_write32(ar, ar->hw_ce_regs->ce_rri_low,
+			  lower_32_bits(ce->paddr_rri));
+	ath10k_ce_write32(ar, ar->hw_ce_regs->ce_rri_high,
+			  (upper_32_bits(ce->paddr_rri) &
+			  CE_DESC_FLAGS_GET_MASK));
+
+	for (i = 0; i < CE_COUNT; i++) {
+		ctrl1_regs = ar->hw_ce_regs->ctrl1_regs->addr;
+		ce_base_addr = ath10k_ce_base_address(ar, i);
+		value = ath10k_ce_read32(ar, ce_base_addr + ctrl1_regs);
+		value |= ar->hw_ce_regs->upd->mask;
+		ath10k_ce_write32(ar, ce_base_addr + ctrl1_regs, value);
+	}
+
+	memset(ce->vaddr_rri, 0, CE_COUNT * sizeof(u32));
+}
+EXPORT_SYMBOL(ath10k_ce_alloc_rri);
+
+void ath10k_ce_free_rri(struct ath10k *ar)
+{
+	struct ath10k_ce *ce = ath10k_ce_priv(ar);
+
+	dma_free_coherent(ar->dev, (CE_COUNT * sizeof(u32)),
+			  ce->vaddr_rri,
+			  ce->paddr_rri);
+}
+EXPORT_SYMBOL(ath10k_ce_free_rri);
diff --git a/drivers/net/wireless/ath/ath10k/ce.h b/drivers/net/wireless/ath/ath10k/ce.h
index 9aea89133209..dbeffaef6024 100644
--- a/drivers/net/wireless/ath/ath10k/ce.h
+++ b/drivers/net/wireless/ath/ath10k/ce.h
@@ -49,6 +49,9 @@ struct ath10k_ce_pipe;
 #define CE_DESC_FLAGS_META_DATA_MASK ar->hw_values->ce_desc_meta_data_mask
 #define CE_DESC_FLAGS_META_DATA_LSB  ar->hw_values->ce_desc_meta_data_lsb
 
+#define CE_DDR_RRI_MASK			GENMASK(15, 0)
+#define CE_DDR_DRRI_SHIFT		16
+
 struct ce_desc {
 	__le32 addr;
 	__le16 nbytes;
@@ -157,6 +160,8 @@ struct ath10k_ce {
 	spinlock_t ce_lock;
 	const struct ath10k_bus_ops *bus_ops;
 	struct ath10k_ce_pipe ce_states[CE_COUNT_MAX];
+	u32 *vaddr_rri;
+	dma_addr_t paddr_rri;
 };
 
 /*==================Send====================*/
@@ -265,6 +270,8 @@ int ath10k_ce_disable_interrupts(struct ath10k *ar);
 void ath10k_ce_enable_interrupts(struct ath10k *ar);
 void ath10k_ce_dump_registers(struct ath10k *ar,
 			      struct ath10k_fw_crash_data *crash_data);
+void ath10k_ce_alloc_rri(struct ath10k *ar);
+void ath10k_ce_free_rri(struct ath10k *ar);
 
 /* ce_attr.flags values */
 /* Use NonSnooping PCIe accesses? */
@@ -331,6 +338,9 @@ static inline u32 ath10k_ce_base_address(struct ath10k *ar, unsigned int ce_id)
 	return CE0_BASE_ADDRESS + (CE1_BASE_ADDRESS - CE0_BASE_ADDRESS) * ce_id;
 }
 
+#define COPY_ENGINE_ID(COPY_ENGINE_BASE_ADDRESS) (((COPY_ENGINE_BASE_ADDRESS) \
+		- CE0_BASE_ADDRESS) / (CE1_BASE_ADDRESS - CE0_BASE_ADDRESS))
+
 #define CE_SRC_RING_TO_DESC(baddr, idx) \
 	(&(((struct ce_desc *)baddr)[idx]))
 
diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
index 5a9d222acfe6..6da6b8c6ef6a 100644
--- a/drivers/net/wireless/ath/ath10k/core.c
+++ b/drivers/net/wireless/ath/ath10k/core.c
@@ -91,6 +91,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.target_64bit = false,
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA988X_HW_2_0_VERSION,
@@ -122,6 +123,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA9887_HW_1_0_VERSION,
@@ -153,6 +155,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA6174_HW_2_1_VERSION,
@@ -183,6 +186,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA6174_HW_2_1_VERSION,
@@ -213,6 +217,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA6174_HW_3_0_VERSION,
@@ -243,6 +248,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA6174_HW_3_2_VERSION,
@@ -276,6 +282,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA99X0_HW_2_0_DEV_VERSION,
@@ -312,6 +319,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA9984_HW_1_0_DEV_VERSION,
@@ -353,6 +361,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA9888_HW_2_0_DEV_VERSION,
@@ -393,6 +402,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA9377_HW_1_0_DEV_VERSION,
@@ -423,6 +433,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA9377_HW_1_1_DEV_VERSION,
@@ -455,6 +466,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = QCA4019_HW_1_0_DEV_VERSION,
@@ -492,6 +504,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL,
 		.per_ce_irq = false,
 		.shadow_reg_support = false,
+		.rri_on_ddr = false,
 	},
 	{
 		.id = WCN3990_HW_1_0_DEV_VERSION,
@@ -514,6 +527,7 @@ static const struct ath10k_hw_params ath10k_hw_params_list[] = {
 		.rx_ring_fill_level = HTT_RX_RING_FILL_LEVEL_DUAL_MAC,
 		.per_ce_irq = true,
 		.shadow_reg_support = true,
+		.rri_on_ddr = true,
 	},
 };
 
diff --git a/drivers/net/wireless/ath/ath10k/hw.c b/drivers/net/wireless/ath/ath10k/hw.c
index 497ac33e0fbf..677535b3d207 100644
--- a/drivers/net/wireless/ath/ath10k/hw.c
+++ b/drivers/net/wireless/ath/ath10k/hw.c
@@ -310,6 +310,12 @@ static struct ath10k_hw_ce_dst_src_wm_regs wcn3990_wm_dst_ring = {
 	.wm_high	= &wcn3990_dst_wm_high,
 };
 
+static struct ath10k_hw_ce_ctrl1_upd wcn3990_ctrl1_upd = {
+	.shift = 19,
+	.mask = 0x00080000,
+	.enable = 0x00000000,
+};
+
 const struct ath10k_hw_ce_regs wcn3990_ce_regs = {
 	.sr_base_addr		= 0x00000000,
 	.sr_size_addr		= 0x00000008,
@@ -320,8 +326,6 @@ const struct ath10k_hw_ce_regs wcn3990_ce_regs = {
 	.dst_wr_index_addr	= 0x00000040,
 	.current_srri_addr	= 0x00000044,
 	.current_drri_addr	= 0x00000048,
-	.ddr_addr_for_rri_low	= 0x00000004,
-	.ddr_addr_for_rri_high	= 0x00000008,
 	.ce_rri_low		= 0x0024C004,
 	.ce_rri_high		= 0x0024C008,
 	.host_ie_addr		= 0x0000002c,
@@ -331,6 +335,7 @@ const struct ath10k_hw_ce_regs wcn3990_ce_regs = {
 	.misc_regs		= &wcn3990_misc_reg,
 	.wm_srcr		= &wcn3990_wm_src_ring,
 	.wm_dstr		= &wcn3990_wm_dst_ring,
+	.upd			= &wcn3990_ctrl1_upd,
 };
 
 const struct ath10k_hw_values wcn3990_values = {
diff --git a/drivers/net/wireless/ath/ath10k/hw.h b/drivers/net/wireless/ath/ath10k/hw.h
index 74faee5a2578..c6240a28c026 100644
--- a/drivers/net/wireless/ath/ath10k/hw.h
+++ b/drivers/net/wireless/ath/ath10k/hw.h
@@ -336,6 +336,12 @@ struct ath10k_hw_ce_dst_src_wm_regs {
 	struct ath10k_hw_ce_regs_addr_map *wm_low;
 	struct ath10k_hw_ce_regs_addr_map *wm_high; };
 
+struct ath10k_hw_ce_ctrl1_upd {
+	u32 shift;
+	u32 mask;
+	u32 enable;
+};
+
 struct ath10k_hw_ce_regs {
 	u32 sr_base_addr;
 	u32 sr_size_addr;
@@ -358,7 +364,9 @@ struct ath10k_hw_ce_regs {
 	struct ath10k_hw_ce_cmd_halt *cmd_halt;
 	struct ath10k_hw_ce_host_ie *host_ie;
 	struct ath10k_hw_ce_dst_src_wm_regs *wm_srcr;
-	struct ath10k_hw_ce_dst_src_wm_regs *wm_dstr; };
+	struct ath10k_hw_ce_dst_src_wm_regs *wm_dstr;
+	struct ath10k_hw_ce_ctrl1_upd *upd;
+};
 
 struct ath10k_hw_values {
 	u32 rtc_state_val_on;
@@ -575,6 +583,9 @@ struct ath10k_hw_params {
 
 	/* target supporting shadow register for ce write */
 	bool shadow_reg_support;
+
+	/* target supporting retention restore on ddr */
+	bool rri_on_ddr;
 };
 
 struct htt_rx_desc;
diff --git a/drivers/net/wireless/ath/ath10k/snoc.c b/drivers/net/wireless/ath/ath10k/snoc.c
index 2e490ff124f1..47a4d2a5bd4c 100644
--- a/drivers/net/wireless/ath/ath10k/snoc.c
+++ b/drivers/net/wireless/ath/ath10k/snoc.c
@@ -767,6 +767,7 @@ static void ath10k_snoc_hif_power_down(struct ath10k *ar)
 	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot hif power down\n");
 
 	ath10k_snoc_wlan_disable(ar);
+	ath10k_ce_free_rri(ar);
 }
 
 static int ath10k_snoc_hif_power_up(struct ath10k *ar)
@@ -782,6 +783,8 @@ static int ath10k_snoc_hif_power_up(struct ath10k *ar)
 		return ret;
 	}
 
+	ath10k_ce_alloc_rri(ar);
+
 	ret = ath10k_snoc_init_pipes(ar);
 	if (ret) {
 		ath10k_err(ar, "failed to initialize CE: %d\n", ret);
-- 
2.14.1