net: ena: add functions for handling Low Latency Queues in ena_com

This patch introduces APIs for detection, initialization, configuration
and actual usage of low latency queues(LLQ). It extends transmit API with
creation of LLQ descriptors in device memory (which include host buffers
descriptors as well as packet header)

Signed-off-by: Arthur Kiyanovski <akiyano@amazon.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 179 insertions, 52 deletions

diff --git a/drivers/net/ethernet/amazon/ena/ena_eth_com.c b/drivers/net/ethernet/amazon/ena/ena_eth_com.c
index 2fa032bab0c3..2a53b549e55a 100644
--- a/drivers/net/ethernet/amazon/ena/ena_eth_com.c
+++ b/drivers/net/ethernet/amazon/ena/ena_eth_com.c
@@ -59,7 +59,7 @@ static inline struct ena_eth_io_rx_cdesc_base *ena_com_get_next_rx_cdesc(
 	return cdesc;
 }
 
-static inline void *get_sq_desc(struct ena_com_io_sq *io_sq)
+static inline void *get_sq_desc_regular_queue(struct ena_com_io_sq *io_sq)
 {
 	u16 tail_masked;
 	u32 offset;
@@ -71,45 +71,159 @@ static inline void *get_sq_desc(struct ena_com_io_sq *io_sq)
 	return (void *)((uintptr_t)io_sq->desc_addr.virt_addr + offset);
 }
 
-static inline void ena_com_copy_curr_sq_desc_to_dev(struct ena_com_io_sq *io_sq)
+static inline int ena_com_write_bounce_buffer_to_dev(struct ena_com_io_sq *io_sq,
+						     u8 *bounce_buffer)
 {
-	u16 tail_masked = io_sq->tail & (io_sq->q_depth - 1);
-	u32 offset = tail_masked * io_sq->desc_entry_size;
+	struct ena_com_llq_info *llq_info = &io_sq->llq_info;
 
-	/* In case this queue isn't a LLQ */
-	if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
-		return;
+	u16 dst_tail_mask;
+	u32 dst_offset;
 
-	memcpy_toio(io_sq->desc_addr.pbuf_dev_addr + offset,
-		    io_sq->desc_addr.virt_addr + offset,
-		    io_sq->desc_entry_size);
-}
+	dst_tail_mask = io_sq->tail & (io_sq->q_depth - 1);
+	dst_offset = dst_tail_mask * llq_info->desc_list_entry_size;
+
+	/* Make sure everything was written into the bounce buffer before
+	 * writing the bounce buffer to the device
+	 */
+	wmb();
+
+	/* The line is completed. Copy it to dev */
+	__iowrite64_copy(io_sq->desc_addr.pbuf_dev_addr + dst_offset,
+			 bounce_buffer, (llq_info->desc_list_entry_size) / 8);
 
-static inline void ena_com_sq_update_tail(struct ena_com_io_sq *io_sq)
-{
 	io_sq->tail++;
 
 	/* Switch phase bit in case of wrap around */
 	if (unlikely((io_sq->tail & (io_sq->q_depth - 1)) == 0))
 		io_sq->phase ^= 1;
+
+	return 0;
 }
 
-static inline int ena_com_write_header(struct ena_com_io_sq *io_sq,
-				       u8 *head_src, u16 header_len)
+static inline int ena_com_write_header_to_bounce(struct ena_com_io_sq *io_sq,
+						 u8 *header_src,
+						 u16 header_len)
 {
-	u16 tail_masked = io_sq->tail & (io_sq->q_depth - 1);
-	u8 __iomem *dev_head_addr =
-		io_sq->header_addr + (tail_masked * io_sq->tx_max_header_size);
+	struct ena_com_llq_pkt_ctrl *pkt_ctrl = &io_sq->llq_buf_ctrl;
+	struct ena_com_llq_info *llq_info = &io_sq->llq_info;
+	u8 *bounce_buffer = pkt_ctrl->curr_bounce_buf;
+	u16 header_offset;
 
-	if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
+	if (unlikely(io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST))
 		return 0;
 
-	if (unlikely(!io_sq->header_addr)) {
-		pr_err("Push buffer header ptr is NULL\n");
-		return -EINVAL;
+	header_offset =
+		llq_info->descs_num_before_header * io_sq->desc_entry_size;
+
+	if (unlikely((header_offset + header_len) >
+		     llq_info->desc_list_entry_size)) {
+		pr_err("trying to write header larger than llq entry can accommodate\n");
+		return -EFAULT;
+	}
+
+	if (unlikely(!bounce_buffer)) {
+		pr_err("bounce buffer is NULL\n");
+		return -EFAULT;
+	}
+
+	memcpy(bounce_buffer + header_offset, header_src, header_len);
+
+	return 0;
+}
+
+static inline void *get_sq_desc_llq(struct ena_com_io_sq *io_sq)
+{
+	struct ena_com_llq_pkt_ctrl *pkt_ctrl = &io_sq->llq_buf_ctrl;
+	u8 *bounce_buffer;
+	void *sq_desc;
+
+	bounce_buffer = pkt_ctrl->curr_bounce_buf;
+
+	if (unlikely(!bounce_buffer)) {
+		pr_err("bounce buffer is NULL\n");
+		return NULL;
+	}
+
+	sq_desc = bounce_buffer + pkt_ctrl->idx * io_sq->desc_entry_size;
+	pkt_ctrl->idx++;
+	pkt_ctrl->descs_left_in_line--;
+
+	return sq_desc;
+}
+
+static inline int ena_com_close_bounce_buffer(struct ena_com_io_sq *io_sq)
+{
+	struct ena_com_llq_pkt_ctrl *pkt_ctrl = &io_sq->llq_buf_ctrl;
+	struct ena_com_llq_info *llq_info = &io_sq->llq_info;
+	int rc;
+
+	if (unlikely(io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST))
+		return 0;
+
+	/* bounce buffer was used, so write it and get a new one */
+	if (pkt_ctrl->idx) {
+		rc = ena_com_write_bounce_buffer_to_dev(io_sq,
+							pkt_ctrl->curr_bounce_buf);
+		if (unlikely(rc))
+			return rc;
+
+		pkt_ctrl->curr_bounce_buf =
+			ena_com_get_next_bounce_buffer(&io_sq->bounce_buf_ctrl);
+		memset(io_sq->llq_buf_ctrl.curr_bounce_buf,
+		       0x0, llq_info->desc_list_entry_size);
+	}
+
+	pkt_ctrl->idx = 0;
+	pkt_ctrl->descs_left_in_line = llq_info->descs_num_before_header;
+	return 0;
+}
+
+static inline void *get_sq_desc(struct ena_com_io_sq *io_sq)
+{
+	if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV)
+		return get_sq_desc_llq(io_sq);
+
+	return get_sq_desc_regular_queue(io_sq);
+}
+
+static inline int ena_com_sq_update_llq_tail(struct ena_com_io_sq *io_sq)
+{
+	struct ena_com_llq_pkt_ctrl *pkt_ctrl = &io_sq->llq_buf_ctrl;
+	struct ena_com_llq_info *llq_info = &io_sq->llq_info;
+	int rc;
+
+	if (!pkt_ctrl->descs_left_in_line) {
+		rc = ena_com_write_bounce_buffer_to_dev(io_sq,
+							pkt_ctrl->curr_bounce_buf);
+		if (unlikely(rc))
+			return rc;
+
+		pkt_ctrl->curr_bounce_buf =
+			ena_com_get_next_bounce_buffer(&io_sq->bounce_buf_ctrl);
+			memset(io_sq->llq_buf_ctrl.curr_bounce_buf,
+			       0x0, llq_info->desc_list_entry_size);
+
+		pkt_ctrl->idx = 0;
+		if (unlikely(llq_info->desc_stride_ctrl == ENA_ADMIN_SINGLE_DESC_PER_ENTRY))
+			pkt_ctrl->descs_left_in_line = 1;
+		else
+			pkt_ctrl->descs_left_in_line =
+			llq_info->desc_list_entry_size / io_sq->desc_entry_size;
 	}
 
-	memcpy_toio(dev_head_addr, head_src, header_len);
+	return 0;
+}
+
+static inline int ena_com_sq_update_tail(struct ena_com_io_sq *io_sq)
+{
+	if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV)
+		return ena_com_sq_update_llq_tail(io_sq);
+
+	io_sq->tail++;
+
+	/* Switch phase bit in case of wrap around */
+	if (unlikely((io_sq->tail & (io_sq->q_depth - 1)) == 0))
+		io_sq->phase ^= 1;
 
 	return 0;
 }
@@ -177,8 +291,8 @@ static inline bool ena_com_meta_desc_changed(struct ena_com_io_sq *io_sq,
 	return false;
 }
 
-static inline void ena_com_create_and_store_tx_meta_desc(struct ena_com_io_sq *io_sq,
-							 struct ena_com_tx_ctx *ena_tx_ctx)
+static inline int ena_com_create_and_store_tx_meta_desc(struct ena_com_io_sq *io_sq,
+							struct ena_com_tx_ctx *ena_tx_ctx)
 {
 	struct ena_eth_io_tx_meta_desc *meta_desc = NULL;
 	struct ena_com_tx_meta *ena_meta = &ena_tx_ctx->ena_meta;
@@ -223,8 +337,7 @@ static inline void ena_com_create_and_store_tx_meta_desc(struct ena_com_io_sq *i
 	memcpy(&io_sq->cached_tx_meta, ena_meta,
 	       sizeof(struct ena_com_tx_meta));
 
-	ena_com_copy_curr_sq_desc_to_dev(io_sq);
-	ena_com_sq_update_tail(io_sq);
+	return ena_com_sq_update_tail(io_sq);
 }
 
 static inline void ena_com_rx_set_flags(struct ena_com_rx_ctx *ena_rx_ctx,
@@ -262,18 +375,19 @@ int ena_com_prepare_tx(struct ena_com_io_sq *io_sq,
 {
 	struct ena_eth_io_tx_desc *desc = NULL;
 	struct ena_com_buf *ena_bufs = ena_tx_ctx->ena_bufs;
-	void *push_header = ena_tx_ctx->push_header;
+	void *buffer_to_push = ena_tx_ctx->push_header;
 	u16 header_len = ena_tx_ctx->header_len;
 	u16 num_bufs = ena_tx_ctx->num_bufs;
-	int total_desc, i, rc;
+	u16 start_tail = io_sq->tail;
+	int i, rc;
 	bool have_meta;
 	u64 addr_hi;
 
 	WARN(io_sq->direction != ENA_COM_IO_QUEUE_DIRECTION_TX, "wrong Q type");
 
 	/* num_bufs +1 for potential meta desc */
-	if (ena_com_sq_empty_space(io_sq) < (num_bufs + 1)) {
-		pr_err("Not enough space in the tx queue\n");
+	if (unlikely(!ena_com_sq_have_enough_space(io_sq, num_bufs + 1))) {
+		pr_debug("Not enough space in the tx queue\n");
 		return -ENOMEM;
 	}
 
@@ -283,23 +397,32 @@ int ena_com_prepare_tx(struct ena_com_io_sq *io_sq,
 		return -EINVAL;
 	}
 
-	/* start with pushing the header (if needed) */
-	rc = ena_com_write_header(io_sq, push_header, header_len);
+	if (unlikely(io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV &&
+		     !buffer_to_push))
+		return -EINVAL;
+
+	rc = ena_com_write_header_to_bounce(io_sq, buffer_to_push, header_len);
 	if (unlikely(rc))
 		return rc;
 
 	have_meta = ena_tx_ctx->meta_valid && ena_com_meta_desc_changed(io_sq,
 			ena_tx_ctx);
-	if (have_meta)
-		ena_com_create_and_store_tx_meta_desc(io_sq, ena_tx_ctx);
+	if (have_meta) {
+		rc = ena_com_create_and_store_tx_meta_desc(io_sq, ena_tx_ctx);
+		if (unlikely(rc))
+			return rc;
+	}
 
-	/* If the caller doesn't want send packets */
+	/* If the caller doesn't want to send packets */
 	if (unlikely(!num_bufs && !header_len)) {
-		*nb_hw_desc = have_meta ? 0 : 1;
-		return 0;
+		rc = ena_com_close_bounce_buffer(io_sq);
+		*nb_hw_desc = io_sq->tail - start_tail;
+		return rc;
 	}
 
 	desc = get_sq_desc(io_sq);
+	if (unlikely(!desc))
+		return -EFAULT;
 	memset(desc, 0x0, sizeof(struct ena_eth_io_tx_desc));
 
 	/* Set first desc when we don't have meta descriptor */
@@ -351,10 +474,14 @@ int ena_com_prepare_tx(struct ena_com_io_sq *io_sq,
 	for (i = 0; i < num_bufs; i++) {
 		/* The first desc share the same desc as the header */
 		if (likely(i != 0)) {
-			ena_com_copy_curr_sq_desc_to_dev(io_sq);
-			ena_com_sq_update_tail(io_sq);
+			rc = ena_com_sq_update_tail(io_sq);
+			if (unlikely(rc))
+				return rc;
 
 			desc = get_sq_desc(io_sq);
+			if (unlikely(!desc))
+				return -EFAULT;
+
 			memset(desc, 0x0, sizeof(struct ena_eth_io_tx_desc));
 
 			desc->len_ctrl |= (io_sq->phase <<
@@ -377,15 +504,14 @@ int ena_com_prepare_tx(struct ena_com_io_sq *io_sq,
 	/* set the last desc indicator */
 	desc->len_ctrl |= ENA_ETH_IO_TX_DESC_LAST_MASK;
 
-	ena_com_copy_curr_sq_desc_to_dev(io_sq);
-
-	ena_com_sq_update_tail(io_sq);
+	rc = ena_com_sq_update_tail(io_sq);
+	if (unlikely(rc))
+		return rc;
 
-	total_desc = max_t(u16, num_bufs, 1);
-	total_desc += have_meta ? 1 : 0;
+	rc = ena_com_close_bounce_buffer(io_sq);
 
-	*nb_hw_desc = total_desc;
-	return 0;
+	*nb_hw_desc = io_sq->tail - start_tail;
+	return rc;
 }
 
 int ena_com_rx_pkt(struct ena_com_io_cq *io_cq,
@@ -444,15 +570,18 @@ int ena_com_add_single_rx_desc(struct ena_com_io_sq *io_sq,
 
 	WARN(io_sq->direction != ENA_COM_IO_QUEUE_DIRECTION_RX, "wrong Q type");
 
-	if (unlikely(ena_com_sq_empty_space(io_sq) == 0))
+	if (unlikely(!ena_com_sq_have_enough_space(io_sq, 1)))
 		return -ENOSPC;
 
 	desc = get_sq_desc(io_sq);
+	if (unlikely(!desc))
+		return -EFAULT;
+
 	memset(desc, 0x0, sizeof(struct ena_eth_io_rx_desc));
 
 	desc->length = ena_buf->len;
 
-	desc->ctrl |= ENA_ETH_IO_RX_DESC_FIRST_MASK;
+	desc->ctrl = ENA_ETH_IO_RX_DESC_FIRST_MASK;
 	desc->ctrl |= ENA_ETH_IO_RX_DESC_LAST_MASK;
 	desc->ctrl |= io_sq->phase & ENA_ETH_IO_RX_DESC_PHASE_MASK;
 	desc->ctrl |= ENA_ETH_IO_RX_DESC_COMP_REQ_MASK;
@@ -463,9 +592,7 @@ int ena_com_add_single_rx_desc(struct ena_com_io_sq *io_sq,
 	desc->buff_addr_hi =
 		((ena_buf->paddr & GENMASK_ULL(io_sq->dma_addr_bits - 1, 32)) >> 32);
 
-	ena_com_sq_update_tail(io_sq);
-
-	return 0;
+	return ena_com_sq_update_tail(io_sq);
 }
 
 bool ena_com_cq_empty(struct ena_com_io_cq *io_cq)