Merge branch 'ena-next'

Arthur Kiyanovski says:

====================
Improving performance and reducing latencies, by using latest capabilities exposed in ENA device

This patchset introduces the following:
1. A new placement policy of Tx headers and descriptors, which takes
advantage of an option to place headers + descriptors in device memory
space. This is sometimes referred to as LLQ - low latency queue.
The patch set defines the admin capability, maps the device memory as
write-combined, and adds a mode in transmit datapath to do header +
descriptor placement on the device.
2. Support for RX checksum offloading
3. Miscelaneous small improvements and code cleanups

Note: V1 of this patchset was created as if patches e2a322a0 248ab773
from net were applied to net-next before applying the patchset. This V2
version does not assume this, and should be applyed directly on net-next
without the aformentioned patches.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/amazon/ena/ena_admin_defs.h

@@ -32,71 +32,49 @@
 #ifndef _ENA_ADMIN_H_
 #define _ENA_ADMIN_H_
 
+
 enum ena_admin_aq_opcode {
 	ENA_ADMIN_CREATE_SQ                         = 1,
-
 	ENA_ADMIN_DESTROY_SQ                        = 2,
-
 	ENA_ADMIN_CREATE_CQ                         = 3,
-
 	ENA_ADMIN_DESTROY_CQ                        = 4,
-
 	ENA_ADMIN_GET_FEATURE                       = 8,
-
 	ENA_ADMIN_SET_FEATURE                       = 9,
-
 	ENA_ADMIN_GET_STATS                         = 11,
 };
 
 enum ena_admin_aq_completion_status {
 	ENA_ADMIN_SUCCESS                           = 0,
-
 	ENA_ADMIN_RESOURCE_ALLOCATION_FAILURE       = 1,
-
 	ENA_ADMIN_BAD_OPCODE                        = 2,
-
 	ENA_ADMIN_UNSUPPORTED_OPCODE                = 3,
-
 	ENA_ADMIN_MALFORMED_REQUEST                 = 4,
-
 	/* Additional status is provided in ACQ entry extended_status */
 	ENA_ADMIN_ILLEGAL_PARAMETER                 = 5,
-
 	ENA_ADMIN_UNKNOWN_ERROR                     = 6,
+	ENA_ADMIN_RESOURCE_BUSY                     = 7,
 };
 
 enum ena_admin_aq_feature_id {
 	ENA_ADMIN_DEVICE_ATTRIBUTES                 = 1,
-
 	ENA_ADMIN_MAX_QUEUES_NUM                    = 2,
-
 	ENA_ADMIN_HW_HINTS                          = 3,
-
+	ENA_ADMIN_LLQ                               = 4,
 	ENA_ADMIN_RSS_HASH_FUNCTION                 = 10,
-
 	ENA_ADMIN_STATELESS_OFFLOAD_CONFIG          = 11,
-
 	ENA_ADMIN_RSS_REDIRECTION_TABLE_CONFIG      = 12,
-
 	ENA_ADMIN_MTU                               = 14,
-
 	ENA_ADMIN_RSS_HASH_INPUT                    = 18,
-
 	ENA_ADMIN_INTERRUPT_MODERATION              = 20,
-
 	ENA_ADMIN_AENQ_CONFIG                       = 26,
-
 	ENA_ADMIN_LINK_CONFIG                       = 27,
-
 	ENA_ADMIN_HOST_ATTR_CONFIG                  = 28,
-
 	ENA_ADMIN_FEATURES_OPCODE_NUM               = 32,
 };
 
 enum ena_admin_placement_policy_type {
 	/* descriptors and headers are in host memory */
 	ENA_ADMIN_PLACEMENT_POLICY_HOST             = 1,
-
 	/* descriptors and headers are in device memory (a.k.a Low Latency
 	 * Queue)
 	 */
@@ -105,38 +83,26 @@ enum ena_admin_placement_policy_type {
 
 enum ena_admin_link_types {
 	ENA_ADMIN_LINK_SPEED_1G                     = 0x1,
-
 	ENA_ADMIN_LINK_SPEED_2_HALF_G               = 0x2,
-
 	ENA_ADMIN_LINK_SPEED_5G                     = 0x4,
-
 	ENA_ADMIN_LINK_SPEED_10G                    = 0x8,
-
 	ENA_ADMIN_LINK_SPEED_25G                    = 0x10,
-
 	ENA_ADMIN_LINK_SPEED_40G                    = 0x20,
-
 	ENA_ADMIN_LINK_SPEED_50G                    = 0x40,
-
 	ENA_ADMIN_LINK_SPEED_100G                   = 0x80,
-
 	ENA_ADMIN_LINK_SPEED_200G                   = 0x100,
-
 	ENA_ADMIN_LINK_SPEED_400G                   = 0x200,
 };
 
 enum ena_admin_completion_policy_type {
 	/* completion queue entry for each sq descriptor */
 	ENA_ADMIN_COMPLETION_POLICY_DESC            = 0,
-
 	/* completion queue entry upon request in sq descriptor */
 	ENA_ADMIN_COMPLETION_POLICY_DESC_ON_DEMAND  = 1,
-
 	/* current queue head pointer is updated in OS memory upon sq
 	 * descriptor request
 	 */
 	ENA_ADMIN_COMPLETION_POLICY_HEAD_ON_DEMAND  = 2,
-
 	/* current queue head pointer is updated in OS memory for each sq
 	 * descriptor
 	 */
@@ -149,13 +115,11 @@ enum ena_admin_completion_policy_type {
  */
 enum ena_admin_get_stats_type {
 	ENA_ADMIN_GET_STATS_TYPE_BASIC              = 0,
-
 	ENA_ADMIN_GET_STATS_TYPE_EXTENDED           = 1,
 };
 
 enum ena_admin_get_stats_scope {
 	ENA_ADMIN_SPECIFIC_QUEUE                    = 0,
-
 	ENA_ADMIN_ETH_TRAFFIC                       = 1,
 };
 
@@ -227,7 +191,9 @@ struct ena_admin_acq_common_desc {
 
 	u16 extended_status;
 
-	/* serves as a hint what AQ entries can be revoked */
+	/* indicates to the driver which AQ entry has been consumed by the
+	 *    device and could be reused
+	 */
 	u16 sq_head_indx;
 };
 
@@ -297,7 +263,6 @@ struct ena_admin_aq_create_sq_cmd {
 
 enum ena_admin_sq_direction {
 	ENA_ADMIN_SQ_DIRECTION_TX                   = 1,
-
 	ENA_ADMIN_SQ_DIRECTION_RX                   = 2,
 };
 
@@ -483,8 +448,85 @@ struct ena_admin_device_attr_feature_desc {
 	u32 max_mtu;
 };
 
+enum ena_admin_llq_header_location {
+	/* header is in descriptor list */
+	ENA_ADMIN_INLINE_HEADER                     = 1,
+	/* header in a separate ring, implies 16B descriptor list entry */
+	ENA_ADMIN_HEADER_RING                       = 2,
+};
+
+enum ena_admin_llq_ring_entry_size {
+	ENA_ADMIN_LIST_ENTRY_SIZE_128B              = 1,
+	ENA_ADMIN_LIST_ENTRY_SIZE_192B              = 2,
+	ENA_ADMIN_LIST_ENTRY_SIZE_256B              = 4,
+};
+
+enum ena_admin_llq_num_descs_before_header {
+	ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_0     = 0,
+	ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_1     = 1,
+	ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_2     = 2,
+	ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_4     = 4,
+	ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_8     = 8,
+};
+
+/* packet descriptor list entry always starts with one or more descriptors,
+ * followed by a header. The rest of the descriptors are located in the
+ * beginning of the subsequent entry. Stride refers to how the rest of the
+ * descriptors are placed. This field is relevant only for inline header
+ * mode
+ */
+enum ena_admin_llq_stride_ctrl {
+	ENA_ADMIN_SINGLE_DESC_PER_ENTRY             = 1,
+	ENA_ADMIN_MULTIPLE_DESCS_PER_ENTRY          = 2,
+};
+
+struct ena_admin_feature_llq_desc {
+	u32 max_llq_num;
+
+	u32 max_llq_depth;
+
+	/*  specify the header locations the device supports. bitfield of
+	 *    enum ena_admin_llq_header_location.
+	 */
+	u16 header_location_ctrl_supported;
+
+	/* the header location the driver selected to use. */
+	u16 header_location_ctrl_enabled;
+
+	/* if inline header is specified - this is the size of descriptor
+	 *    list entry. If header in a separate ring is specified - this is
+	 *    the size of header ring entry. bitfield of enum
+	 *    ena_admin_llq_ring_entry_size. specify the entry sizes the device
+	 *    supports
+	 */
+	u16 entry_size_ctrl_supported;
+
+	/* the entry size the driver selected to use. */
+	u16 entry_size_ctrl_enabled;
+
+	/* valid only if inline header is specified. First entry associated
+	 *    with the packet includes descriptors and header. Rest of the
+	 *    entries occupied by descriptors. This parameter defines the max
+	 *    number of descriptors precedding the header in the first entry.
+	 *    The field is bitfield of enum
+	 *    ena_admin_llq_num_descs_before_header and specify the values the
+	 *    device supports
+	 */
+	u16 desc_num_before_header_supported;
+
+	/* the desire field the driver selected to use */
+	u16 desc_num_before_header_enabled;
+
+	/* valid only if inline was chosen. bitfield of enum
+	 *    ena_admin_llq_stride_ctrl
+	 */
+	u16 descriptors_stride_ctrl_supported;
+
+	/* the stride control the driver selected to use */
+	u16 descriptors_stride_ctrl_enabled;
+};
+
 struct ena_admin_queue_feature_desc {
-	/* including LLQs */
 	u32 max_sq_num;
 
 	u32 max_sq_depth;
@@ -493,9 +535,9 @@ struct ena_admin_queue_feature_desc {
 
 	u32 max_cq_depth;
 
-	u32 max_llq_num;
+	u32 max_legacy_llq_num;
 
-	u32 max_llq_depth;
+	u32 max_legacy_llq_depth;
 
 	u32 max_header_size;
 
@@ -584,7 +626,6 @@ struct ena_admin_feature_offload_desc {
 
 enum ena_admin_hash_functions {
 	ENA_ADMIN_TOEPLITZ                          = 1,
-
 	ENA_ADMIN_CRC32                             = 2,
 };
 
@@ -612,27 +653,17 @@ struct ena_admin_feature_rss_flow_hash_function {
 /* RSS flow hash protocols */
 enum ena_admin_flow_hash_proto {
 	ENA_ADMIN_RSS_TCP4                          = 0,
-
 	ENA_ADMIN_RSS_UDP4                          = 1,
-
 	ENA_ADMIN_RSS_TCP6                          = 2,
-
 	ENA_ADMIN_RSS_UDP6                          = 3,
-
 	ENA_ADMIN_RSS_IP4                           = 4,
-
 	ENA_ADMIN_RSS_IP6                           = 5,
-
 	ENA_ADMIN_RSS_IP4_FRAG                      = 6,
-
 	ENA_ADMIN_RSS_NOT_IP                        = 7,
-
 	/* TCPv6 with extension header */
 	ENA_ADMIN_RSS_TCP6_EX                       = 8,
-
 	/* IPv6 with extension header */
 	ENA_ADMIN_RSS_IP6_EX                        = 9,
-
 	ENA_ADMIN_RSS_PROTO_NUM                     = 16,
 };
 
@@ -640,19 +671,14 @@ enum ena_admin_flow_hash_proto {
 enum ena_admin_flow_hash_fields {
 	/* Ethernet Dest Addr */
 	ENA_ADMIN_RSS_L2_DA                         = BIT(0),
-
 	/* Ethernet Src Addr */
 	ENA_ADMIN_RSS_L2_SA                         = BIT(1),
-
 	/* ipv4/6 Dest Addr */
 	ENA_ADMIN_RSS_L3_DA                         = BIT(2),
-
 	/* ipv4/6 Src Addr */
 	ENA_ADMIN_RSS_L3_SA                         = BIT(3),
-
 	/* tcp/udp Dest Port */
 	ENA_ADMIN_RSS_L4_DP                         = BIT(4),
-
 	/* tcp/udp Src Port */
 	ENA_ADMIN_RSS_L4_SP                         = BIT(5),
 };
@@ -694,14 +720,12 @@ struct ena_admin_feature_rss_flow_hash_input {
 
 enum ena_admin_os_type {
 	ENA_ADMIN_OS_LINUX                          = 1,
-
 	ENA_ADMIN_OS_WIN                            = 2,
-
 	ENA_ADMIN_OS_DPDK                           = 3,
-
 	ENA_ADMIN_OS_FREEBSD                        = 4,
-
 	ENA_ADMIN_OS_IPXE                           = 5,
+	ENA_ADMIN_OS_ESXI			    = 6,
+	ENA_ADMIN_OS_GROUPS_NUM			    = 6,
 };
 
 struct ena_admin_host_info {
@@ -723,11 +747,27 @@ struct ena_admin_host_info {
 	/* 7:0 : major
 	 * 15:8 : minor
 	 * 23:16 : sub_minor
+	 * 31:24 : module_type
 	 */
 	u32 driver_version;
 
 	/* features bitmap */
-	u32 supported_network_features[4];
+	u32 supported_network_features[2];
+
+	/* ENA spec version of driver */
+	u16 ena_spec_version;
+
+	/* ENA device's Bus, Device and Function
+	 * 2:0 : function
+	 * 7:3 : device
+	 * 15:8 : bus
+	 */
+	u16 bdf;
+
+	/* Number of CPUs */
+	u16 num_cpus;
+
+	u16 reserved;
 };
 
 struct ena_admin_rss_ind_table_entry {
@@ -800,6 +840,8 @@ struct ena_admin_get_feat_resp {
 
 		struct ena_admin_device_attr_feature_desc dev_attr;
 
+		struct ena_admin_feature_llq_desc llq;
+
 		struct ena_admin_queue_feature_desc max_queue;
 
 		struct ena_admin_feature_aenq_desc aenq;
@@ -847,6 +889,9 @@ struct ena_admin_set_feat_cmd {
 
 		/* rss indirection table */
 		struct ena_admin_feature_rss_ind_table ind_table;
+
+		/* LLQ configuration */
+		struct ena_admin_feature_llq_desc llq;
 	} u;
 };
 
@@ -876,23 +921,16 @@ struct ena_admin_aenq_common_desc {
 /* asynchronous event notification groups */
 enum ena_admin_aenq_group {
 	ENA_ADMIN_LINK_CHANGE                       = 0,
-
 	ENA_ADMIN_FATAL_ERROR                       = 1,
-
 	ENA_ADMIN_WARNING                           = 2,
-
 	ENA_ADMIN_NOTIFICATION                      = 3,
-
 	ENA_ADMIN_KEEP_ALIVE                        = 4,
-
 	ENA_ADMIN_AENQ_GROUPS_NUM                   = 5,
 };
 
 enum ena_admin_aenq_notification_syndrom {
 	ENA_ADMIN_SUSPEND                           = 0,
-
 	ENA_ADMIN_RESUME                            = 1,
-
 	ENA_ADMIN_UPDATE_HINTS                      = 2,
 };
 
@@ -1008,6 +1046,13 @@ struct ena_admin_ena_mmio_req_read_less_resp {
 #define ENA_ADMIN_HOST_INFO_MINOR_MASK                      GENMASK(15, 8)
 #define ENA_ADMIN_HOST_INFO_SUB_MINOR_SHIFT                 16
 #define ENA_ADMIN_HOST_INFO_SUB_MINOR_MASK                  GENMASK(23, 16)
+#define ENA_ADMIN_HOST_INFO_MODULE_TYPE_SHIFT               24
+#define ENA_ADMIN_HOST_INFO_MODULE_TYPE_MASK                GENMASK(31, 24)
+#define ENA_ADMIN_HOST_INFO_FUNCTION_MASK                   GENMASK(2, 0)
+#define ENA_ADMIN_HOST_INFO_DEVICE_SHIFT                    3
+#define ENA_ADMIN_HOST_INFO_DEVICE_MASK                     GENMASK(7, 3)
+#define ENA_ADMIN_HOST_INFO_BUS_SHIFT                       8
+#define ENA_ADMIN_HOST_INFO_BUS_MASK                        GENMASK(15, 8)
 
 /* aenq_common_desc */
 #define ENA_ADMIN_AENQ_COMMON_DESC_PHASE_MASK               BIT(0)

drivers/net/ethernet/amazon/ena/ena_com.c

@@ -41,9 +41,6 @@
 #define ENA_ASYNC_QUEUE_DEPTH 16
 #define ENA_ADMIN_QUEUE_DEPTH 32
 
-#define MIN_ENA_VER (((ENA_COMMON_SPEC_VERSION_MAJOR) << \
-		ENA_REGS_VERSION_MAJOR_VERSION_SHIFT) \
-		| (ENA_COMMON_SPEC_VERSION_MINOR))
 
 #define ENA_CTRL_MAJOR		0
 #define ENA_CTRL_MINOR		0
@@ -61,6 +58,8 @@
 
 #define ENA_MMIO_READ_TIMEOUT 0xFFFFFFFF
 
+#define ENA_COM_BOUNCE_BUFFER_CNTRL_CNT	4
+
 #define ENA_REGS_ADMIN_INTR_MASK 1
 
 #define ENA_POLL_MS	5
@@ -236,7 +235,7 @@ static struct ena_comp_ctx *__ena_com_submit_admin_cmd(struct ena_com_admin_queu
 	tail_masked = admin_queue->sq.tail & queue_size_mask;
 
 	/* In case of queue FULL */
-	cnt = atomic_read(&admin_queue->outstanding_cmds);
+	cnt = (u16)atomic_read(&admin_queue->outstanding_cmds);
 	if (cnt >= admin_queue->q_depth) {
 		pr_debug("admin queue is full.\n");
 		admin_queue->stats.out_of_space++;
@@ -305,7 +304,7 @@ static struct ena_comp_ctx *ena_com_submit_admin_cmd(struct ena_com_admin_queue
 						     struct ena_admin_acq_entry *comp,
 						     size_t comp_size_in_bytes)
 {
-	unsigned long flags;
+	unsigned long flags = 0;
 	struct ena_comp_ctx *comp_ctx;
 
 	spin_lock_irqsave(&admin_queue->q_lock, flags);
@@ -333,7 +332,7 @@ static int ena_com_init_io_sq(struct ena_com_dev *ena_dev,
 
 	memset(&io_sq->desc_addr, 0x0, sizeof(io_sq->desc_addr));
 
-	io_sq->dma_addr_bits = ena_dev->dma_addr_bits;
+	io_sq->dma_addr_bits = (u8)ena_dev->dma_addr_bits;
 	io_sq->desc_entry_size =
 		(io_sq->direction == ENA_COM_IO_QUEUE_DIRECTION_TX) ?
 		sizeof(struct ena_eth_io_tx_desc) :
@@ -355,23 +354,50 @@ static int ena_com_init_io_sq(struct ena_com_dev *ena_dev,
 						    &io_sq->desc_addr.phys_addr,
 						    GFP_KERNEL);
 		}
-	} else {
+
+		if (!io_sq->desc_addr.virt_addr) {
+			pr_err("memory allocation failed");
+			return -ENOMEM;
+		}
+	}
+
+	if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV) {
+		/* Allocate bounce buffers */
+		io_sq->bounce_buf_ctrl.buffer_size =
+			ena_dev->llq_info.desc_list_entry_size;
+		io_sq->bounce_buf_ctrl.buffers_num =
+			ENA_COM_BOUNCE_BUFFER_CNTRL_CNT;
+		io_sq->bounce_buf_ctrl.next_to_use = 0;
+
+		size = io_sq->bounce_buf_ctrl.buffer_size *
+			 io_sq->bounce_buf_ctrl.buffers_num;
+
 		dev_node = dev_to_node(ena_dev->dmadev);
 		set_dev_node(ena_dev->dmadev, ctx->numa_node);
-		io_sq->desc_addr.virt_addr =
+		io_sq->bounce_buf_ctrl.base_buffer =
 			devm_kzalloc(ena_dev->dmadev, size, GFP_KERNEL);
 		set_dev_node(ena_dev->dmadev, dev_node);
-		if (!io_sq->desc_addr.virt_addr) {
-			io_sq->desc_addr.virt_addr =
+		if (!io_sq->bounce_buf_ctrl.base_buffer)
+			io_sq->bounce_buf_ctrl.base_buffer =
 				devm_kzalloc(ena_dev->dmadev, size, GFP_KERNEL);
-		}
-	}
 
-	if (!io_sq->desc_addr.virt_addr) {
-		pr_err("memory allocation failed");
+		if (!io_sq->bounce_buf_ctrl.base_buffer) {
+			pr_err("bounce buffer memory allocation failed");
 			return -ENOMEM;
 		}
 
+		memcpy(&io_sq->llq_info, &ena_dev->llq_info,
+		       sizeof(io_sq->llq_info));
+
+		/* Initiate the first bounce buffer */
+		io_sq->llq_buf_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, io_sq->llq_info.desc_list_entry_size);
+		io_sq->llq_buf_ctrl.descs_left_in_line =
+			io_sq->llq_info.descs_num_before_header;
+	}
+
 	io_sq->tail = 0;
 	io_sq->next_to_comp = 0;
 	io_sq->phase = 1;
@@ -511,7 +537,8 @@ static int ena_com_comp_status_to_errno(u8 comp_status)
 static int ena_com_wait_and_process_admin_cq_polling(struct ena_comp_ctx *comp_ctx,
 						     struct ena_com_admin_queue *admin_queue)
 {
-	unsigned long flags, timeout;
+	unsigned long flags = 0;
+	unsigned long timeout;
 	int ret;
 
 	timeout = jiffies + usecs_to_jiffies(admin_queue->completion_timeout);
@@ -557,10 +584,160 @@ static int ena_com_wait_and_process_admin_cq_polling(struct ena_comp_ctx *comp_c
 	return ret;
 }
 
+/**
+ * Set the LLQ configurations of the firmware
+ *
+ * The driver provides only the enabled feature values to the device,
+ * which in turn, checks if they are supported.
+ */
+static int ena_com_set_llq(struct ena_com_dev *ena_dev)
+{
+	struct ena_com_admin_queue *admin_queue;
+	struct ena_admin_set_feat_cmd cmd;
+	struct ena_admin_set_feat_resp resp;
+	struct ena_com_llq_info *llq_info = &ena_dev->llq_info;
+	int ret;
+
+	memset(&cmd, 0x0, sizeof(cmd));
+	admin_queue = &ena_dev->admin_queue;
+
+	cmd.aq_common_descriptor.opcode = ENA_ADMIN_SET_FEATURE;
+	cmd.feat_common.feature_id = ENA_ADMIN_LLQ;
+
+	cmd.u.llq.header_location_ctrl_enabled = llq_info->header_location_ctrl;
+	cmd.u.llq.entry_size_ctrl_enabled = llq_info->desc_list_entry_size_ctrl;
+	cmd.u.llq.desc_num_before_header_enabled = llq_info->descs_num_before_header;
+	cmd.u.llq.descriptors_stride_ctrl_enabled = llq_info->desc_stride_ctrl;
+
+	ret = ena_com_execute_admin_command(admin_queue,
+					    (struct ena_admin_aq_entry *)&cmd,
+					    sizeof(cmd),
+					    (struct ena_admin_acq_entry *)&resp,
+					    sizeof(resp));
+
+	if (unlikely(ret))
+		pr_err("Failed to set LLQ configurations: %d\n", ret);
+
+	return ret;
+}
+
+static int ena_com_config_llq_info(struct ena_com_dev *ena_dev,
+				   struct ena_admin_feature_llq_desc *llq_features,
+				   struct ena_llq_configurations *llq_default_cfg)
+{
+	struct ena_com_llq_info *llq_info = &ena_dev->llq_info;
+	u16 supported_feat;
+	int rc;
+
+	memset(llq_info, 0, sizeof(*llq_info));
+
+	supported_feat = llq_features->header_location_ctrl_supported;
+
+	if (likely(supported_feat & llq_default_cfg->llq_header_location)) {
+		llq_info->header_location_ctrl =
+			llq_default_cfg->llq_header_location;
+	} else {
+		pr_err("Invalid header location control, supported: 0x%x\n",
+		       supported_feat);
+		return -EINVAL;
+	}
+
+	if (likely(llq_info->header_location_ctrl == ENA_ADMIN_INLINE_HEADER)) {
+		supported_feat = llq_features->descriptors_stride_ctrl_supported;
+		if (likely(supported_feat & llq_default_cfg->llq_stride_ctrl)) {
+			llq_info->desc_stride_ctrl = llq_default_cfg->llq_stride_ctrl;
+		} else	{
+			if (supported_feat & ENA_ADMIN_MULTIPLE_DESCS_PER_ENTRY) {
+				llq_info->desc_stride_ctrl = ENA_ADMIN_MULTIPLE_DESCS_PER_ENTRY;
+			} else if (supported_feat & ENA_ADMIN_SINGLE_DESC_PER_ENTRY) {
+				llq_info->desc_stride_ctrl = ENA_ADMIN_SINGLE_DESC_PER_ENTRY;
+			} else {
+				pr_err("Invalid desc_stride_ctrl, supported: 0x%x\n",
+				       supported_feat);
+				return -EINVAL;
+			}
+
+			pr_err("Default llq stride ctrl is not supported, performing fallback, default: 0x%x, supported: 0x%x, used: 0x%x\n",
+			       llq_default_cfg->llq_stride_ctrl, supported_feat,
+			       llq_info->desc_stride_ctrl);
+		}
+	} else {
+		llq_info->desc_stride_ctrl = 0;
+	}
+
+	supported_feat = llq_features->entry_size_ctrl_supported;
+	if (likely(supported_feat & llq_default_cfg->llq_ring_entry_size)) {
+		llq_info->desc_list_entry_size_ctrl = llq_default_cfg->llq_ring_entry_size;
+		llq_info->desc_list_entry_size = llq_default_cfg->llq_ring_entry_size_value;
+	} else {
+		if (supported_feat & ENA_ADMIN_LIST_ENTRY_SIZE_128B) {
+			llq_info->desc_list_entry_size_ctrl = ENA_ADMIN_LIST_ENTRY_SIZE_128B;
+			llq_info->desc_list_entry_size = 128;
+		} else if (supported_feat & ENA_ADMIN_LIST_ENTRY_SIZE_192B) {
+			llq_info->desc_list_entry_size_ctrl = ENA_ADMIN_LIST_ENTRY_SIZE_192B;
+			llq_info->desc_list_entry_size = 192;
+		} else if (supported_feat & ENA_ADMIN_LIST_ENTRY_SIZE_256B) {
+			llq_info->desc_list_entry_size_ctrl = ENA_ADMIN_LIST_ENTRY_SIZE_256B;
+			llq_info->desc_list_entry_size = 256;
+		} else {
+			pr_err("Invalid entry_size_ctrl, supported: 0x%x\n",
+			       supported_feat);
+			return -EINVAL;
+		}
+
+		pr_err("Default llq ring entry size is not supported, performing fallback, default: 0x%x, supported: 0x%x, used: 0x%x\n",
+		       llq_default_cfg->llq_ring_entry_size, supported_feat,
+		       llq_info->desc_list_entry_size);
+	}
+	if (unlikely(llq_info->desc_list_entry_size & 0x7)) {
+		/* The desc list entry size should be whole multiply of 8
+		 * This requirement comes from __iowrite64_copy()
+		 */
+		pr_err("illegal entry size %d\n",
+		       llq_info->desc_list_entry_size);
+		return -EINVAL;
+	}
+
+	if (llq_info->desc_stride_ctrl == ENA_ADMIN_MULTIPLE_DESCS_PER_ENTRY)
+		llq_info->descs_per_entry = llq_info->desc_list_entry_size /
+			sizeof(struct ena_eth_io_tx_desc);
+	else
+		llq_info->descs_per_entry = 1;
+
+	supported_feat = llq_features->desc_num_before_header_supported;
+	if (likely(supported_feat & llq_default_cfg->llq_num_decs_before_header)) {
+		llq_info->descs_num_before_header = llq_default_cfg->llq_num_decs_before_header;
+	} else {
+		if (supported_feat & ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_2) {
+			llq_info->descs_num_before_header = ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_2;
+		} else if (supported_feat & ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_1) {
+			llq_info->descs_num_before_header = ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_1;
+		} else if (supported_feat & ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_4) {
+			llq_info->descs_num_before_header = ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_4;
+		} else if (supported_feat & ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_8) {
+			llq_info->descs_num_before_header = ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_8;
+		} else {
+			pr_err("Invalid descs_num_before_header, supported: 0x%x\n",
+			       supported_feat);
+			return -EINVAL;
+		}
+
+		pr_err("Default llq num descs before header is not supported, performing fallback, default: 0x%x, supported: 0x%x, used: 0x%x\n",
+		       llq_default_cfg->llq_num_decs_before_header,
+		       supported_feat, llq_info->descs_num_before_header);
+	}
+
+	rc = ena_com_set_llq(ena_dev);
+	if (rc)
+		pr_err("Cannot set LLQ configuration: %d\n", rc);
+
+	return 0;
+}
+
 static int ena_com_wait_and_process_admin_cq_interrupts(struct ena_comp_ctx *comp_ctx,
 							struct ena_com_admin_queue *admin_queue)
 {
-	unsigned long flags;
+	unsigned long flags = 0;
 	int ret;
 
 	wait_for_completion_timeout(&comp_ctx->wait_event,
@@ -606,7 +783,7 @@ static u32 ena_com_reg_bar_read32(struct ena_com_dev *ena_dev, u16 offset)
 	volatile struct ena_admin_ena_mmio_req_read_less_resp *read_resp =
 		mmio_read->read_resp;
 	u32 mmio_read_reg, ret, i;
-	unsigned long flags;
+	unsigned long flags = 0;
 	u32 timeout = mmio_read->reg_read_to;
 
 	might_sleep();
@@ -728,15 +905,17 @@ static void ena_com_io_queue_free(struct ena_com_dev *ena_dev,
 	if (io_sq->desc_addr.virt_addr) {
 		size = io_sq->desc_entry_size * io_sq->q_depth;
 
-		if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
 		dma_free_coherent(ena_dev->dmadev, size,
 				  io_sq->desc_addr.virt_addr,
 				  io_sq->desc_addr.phys_addr);
-		else
-			devm_kfree(ena_dev->dmadev, io_sq->desc_addr.virt_addr);
 
 		io_sq->desc_addr.virt_addr = NULL;
 	}
+
+	if (io_sq->bounce_buf_ctrl.base_buffer) {
+		devm_kfree(ena_dev->dmadev, io_sq->bounce_buf_ctrl.base_buffer);
+		io_sq->bounce_buf_ctrl.base_buffer = NULL;
+	}
 }
 
 static int wait_for_reset_state(struct ena_com_dev *ena_dev, u32 timeout,
@@ -1248,7 +1427,7 @@ void ena_com_abort_admin_commands(struct ena_com_dev *ena_dev)
 void ena_com_wait_for_abort_completion(struct ena_com_dev *ena_dev)
 {
 	struct ena_com_admin_queue *admin_queue = &ena_dev->admin_queue;
-	unsigned long flags;
+	unsigned long flags = 0;
 
 	spin_lock_irqsave(&admin_queue->q_lock, flags);
 	while (atomic_read(&admin_queue->outstanding_cmds) != 0) {
@@ -1292,7 +1471,7 @@ bool ena_com_get_admin_running_state(struct ena_com_dev *ena_dev)
 void ena_com_set_admin_running_state(struct ena_com_dev *ena_dev, bool state)
 {
 	struct ena_com_admin_queue *admin_queue = &ena_dev->admin_queue;
-	unsigned long flags;
+	unsigned long flags = 0;
 
 	spin_lock_irqsave(&admin_queue->q_lock, flags);
 	ena_dev->admin_queue.running_state = state;
@@ -1326,7 +1505,7 @@ int ena_com_set_aenq_config(struct ena_com_dev *ena_dev, u32 groups_flag)
 	}
 
 	if ((get_resp.u.aenq.supported_groups & groups_flag) != groups_flag) {
-		pr_warn("Trying to set unsupported aenq events. supported flag: %x asked flag: %x\n",
+		pr_warn("Trying to set unsupported aenq events. supported flag: 0x%x asked flag: 0x%x\n",
 			get_resp.u.aenq.supported_groups, groups_flag);
 		return -EOPNOTSUPP;
 	}
@@ -1400,11 +1579,6 @@ int ena_com_validate_version(struct ena_com_dev *ena_dev)
 			ENA_REGS_VERSION_MAJOR_VERSION_SHIFT,
 		ver & ENA_REGS_VERSION_MINOR_VERSION_MASK);
 
-	if (ver < MIN_ENA_VER) {
-		pr_err("ENA version is lower than the minimal version the driver supports\n");
-		return -1;
-	}
-
 	pr_info("ena controller version: %d.%d.%d implementation version %d\n",
 		(ctrl_ver & ENA_REGS_CONTROLLER_VERSION_MAJOR_VERSION_MASK) >>
 			ENA_REGS_CONTROLLER_VERSION_MAJOR_VERSION_SHIFT,
@@ -1479,7 +1653,7 @@ int ena_com_mmio_reg_read_request_init(struct ena_com_dev *ena_dev)
 				    sizeof(*mmio_read->read_resp),
 				    &mmio_read->read_resp_dma_addr, GFP_KERNEL);
 	if (unlikely(!mmio_read->read_resp))
-		return -ENOMEM;
+		goto err;
 
 	ena_com_mmio_reg_read_request_write_dev_addr(ena_dev);
 
@@ -1488,6 +1662,10 @@ int ena_com_mmio_reg_read_request_init(struct ena_com_dev *ena_dev)
 	mmio_read->readless_supported = true;
 
 	return 0;
+
+err:
+
+	return -ENOMEM;
 }
 
 void ena_com_set_mmio_read_mode(struct ena_com_dev *ena_dev, bool readless_supported)
@@ -1523,8 +1701,7 @@ void ena_com_mmio_reg_read_request_write_dev_addr(struct ena_com_dev *ena_dev)
 }
 
 int ena_com_admin_init(struct ena_com_dev *ena_dev,
-		       struct ena_aenq_handlers *aenq_handlers,
-		       bool init_spinlock)
+		       struct ena_aenq_handlers *aenq_handlers)
 {
 	struct ena_com_admin_queue *admin_queue = &ena_dev->admin_queue;
 	u32 aq_caps, acq_caps, dev_sts, addr_low, addr_high;
@@ -1550,7 +1727,6 @@ int ena_com_admin_init(struct ena_com_dev *ena_dev,
 
 	atomic_set(&admin_queue->outstanding_cmds, 0);
 
-	if (init_spinlock)
 	spin_lock_init(&admin_queue->q_lock);
 
 	ret = ena_com_init_comp_ctxt(admin_queue);
@@ -1748,6 +1924,15 @@ int ena_com_get_dev_attr_feat(struct ena_com_dev *ena_dev,
 	else
 		return rc;
 
+	rc = ena_com_get_feature(ena_dev, &get_resp, ENA_ADMIN_LLQ);
+	if (!rc)
+		memcpy(&get_feat_ctx->llq, &get_resp.u.llq,
+		       sizeof(get_resp.u.llq));
+	else if (rc == -EOPNOTSUPP)
+		memset(&get_feat_ctx->llq, 0x0, sizeof(get_feat_ctx->llq));
+	else
+		return rc;
+
 	return 0;
 }
 
@@ -1779,6 +1964,7 @@ void ena_com_aenq_intr_handler(struct ena_com_dev *dev, void *data)
 	struct ena_admin_aenq_entry *aenq_e;
 	struct ena_admin_aenq_common_desc *aenq_common;
 	struct ena_com_aenq *aenq  = &dev->aenq;
+	unsigned long long timestamp;
 	ena_aenq_handler handler_cb;
 	u16 masked_head, processed = 0;
 	u8 phase;
@@ -1796,10 +1982,11 @@ void ena_com_aenq_intr_handler(struct ena_com_dev *dev, void *data)
 		 */
 		dma_rmb();
 
+		timestamp =
+			(unsigned long long)aenq_common->timestamp_low |
+			((unsigned long long)aenq_common->timestamp_high << 32);
 		pr_debug("AENQ! Group[%x] Syndrom[%x] timestamp: [%llus]\n",
-			 aenq_common->group, aenq_common->syndrom,
-			 (u64)aenq_common->timestamp_low +
-				 ((u64)aenq_common->timestamp_high << 32));
+			 aenq_common->group, aenq_common->syndrom, timestamp);
 
 		/* Handle specific event*/
 		handler_cb = ena_com_get_specific_aenq_cb(dev,
@@ -2441,6 +2628,10 @@ int ena_com_allocate_host_info(struct ena_com_dev *ena_dev)
 	if (unlikely(!host_attr->host_info))
 		return -ENOMEM;
 
+	host_attr->host_info->ena_spec_version = ((ENA_COMMON_SPEC_VERSION_MAJOR <<
+		ENA_REGS_VERSION_MAJOR_VERSION_SHIFT) |
+		(ENA_COMMON_SPEC_VERSION_MINOR));
+
 	return 0;
 }
 
@@ -2712,3 +2903,34 @@ void ena_com_get_intr_moderation_entry(struct ena_com_dev *ena_dev,
 	intr_moder_tbl[level].pkts_per_interval;
 	entry->bytes_per_interval = intr_moder_tbl[level].bytes_per_interval;
 }
+
+int ena_com_config_dev_mode(struct ena_com_dev *ena_dev,
+			    struct ena_admin_feature_llq_desc *llq_features,
+			    struct ena_llq_configurations *llq_default_cfg)
+{
+	int rc;
+	int size;
+
+	if (!llq_features->max_llq_num) {
+		ena_dev->tx_mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_HOST;
+		return 0;
+	}
+
+	rc = ena_com_config_llq_info(ena_dev, llq_features, llq_default_cfg);
+	if (rc)
+		return rc;
+
+	/* Validate the descriptor is not too big */
+	size = ena_dev->tx_max_header_size;
+	size += ena_dev->llq_info.descs_num_before_header *
+		sizeof(struct ena_eth_io_tx_desc);
+
+	if (unlikely(ena_dev->llq_info.desc_list_entry_size < size)) {
+		pr_err("the size of the LLQ entry is smaller than needed\n");
+		return -EINVAL;
+	}
+
+	ena_dev->tx_mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_DEV;
+
+	return 0;
+}

drivers/net/ethernet/amazon/ena/ena_com.h

@@ -37,6 +37,7 @@
 #include <linux/delay.h>
 #include <linux/dma-mapping.h>
 #include <linux/gfp.h>
+#include <linux/io.h>
 #include <linux/sched.h>
 #include <linux/sizes.h>
 #include <linux/spinlock.h>
@@ -108,6 +109,14 @@ enum ena_intr_moder_level {
 	ENA_INTR_MAX_NUM_OF_LEVELS,
 };
 
+struct ena_llq_configurations {
+	enum ena_admin_llq_header_location llq_header_location;
+	enum ena_admin_llq_ring_entry_size llq_ring_entry_size;
+	enum ena_admin_llq_stride_ctrl  llq_stride_ctrl;
+	enum ena_admin_llq_num_descs_before_header llq_num_decs_before_header;
+	u16 llq_ring_entry_size_value;
+};
+
 struct ena_intr_moder_entry {
 	unsigned int intr_moder_interval;
 	unsigned int pkts_per_interval;
@@ -142,6 +151,15 @@ struct ena_com_tx_meta {
 	u16 l4_hdr_len; /* In words */
 };
 
+struct ena_com_llq_info {
+	u16 header_location_ctrl;
+	u16 desc_stride_ctrl;
+	u16 desc_list_entry_size_ctrl;
+	u16 desc_list_entry_size;
+	u16 descs_num_before_header;
+	u16 descs_per_entry;
+};
+
 struct ena_com_io_cq {
 	struct ena_com_io_desc_addr cdesc_addr;
 
@@ -179,6 +197,20 @@ struct ena_com_io_cq {
 
 } ____cacheline_aligned;
 
+struct ena_com_io_bounce_buffer_control {
+	u8 *base_buffer;
+	u16 next_to_use;
+	u16 buffer_size;
+	u16 buffers_num;  /* Must be a power of 2 */
+};
+
+/* This struct is to keep tracking the current location of the next llq entry */
+struct ena_com_llq_pkt_ctrl {
+	u8 *curr_bounce_buf;
+	u16 idx;
+	u16 descs_left_in_line;
+};
+
 struct ena_com_io_sq {
 	struct ena_com_io_desc_addr desc_addr;
 
@@ -190,6 +222,9 @@ struct ena_com_io_sq {
 
 	u32 msix_vector;
 	struct ena_com_tx_meta cached_tx_meta;
+	struct ena_com_llq_info llq_info;
+	struct ena_com_llq_pkt_ctrl llq_buf_ctrl;
+	struct ena_com_io_bounce_buffer_control bounce_buf_ctrl;
 
 	u16 q_depth;
 	u16 qid;
@@ -197,6 +232,7 @@ struct ena_com_io_sq {
 	u16 idx;
 	u16 tail;
 	u16 next_to_comp;
+	u16 llq_last_copy_tail;
 	u32 tx_max_header_size;
 	u8 phase;
 	u8 desc_entry_size;
@@ -334,6 +370,8 @@ struct ena_com_dev {
 	u16 intr_delay_resolution;
 	u32 intr_moder_tx_interval;
 	struct ena_intr_moder_entry *intr_moder_tbl;
+
+	struct ena_com_llq_info llq_info;
 };
 
 struct ena_com_dev_get_features_ctx {
@@ -342,6 +380,7 @@ struct ena_com_dev_get_features_ctx {
 	struct ena_admin_feature_aenq_desc aenq;
 	struct ena_admin_feature_offload_desc offload;
 	struct ena_admin_ena_hw_hints hw_hints;
+	struct ena_admin_feature_llq_desc llq;
 };
 
 struct ena_com_create_io_ctx {
@@ -397,8 +436,6 @@ void ena_com_mmio_reg_read_request_destroy(struct ena_com_dev *ena_dev);
 /* ena_com_admin_init - Init the admin and the async queues
  * @ena_dev: ENA communication layer struct
  * @aenq_handlers: Those handlers to be called upon event.
- * @init_spinlock: Indicate if this method should init the admin spinlock or
- * the spinlock was init before (for example, in a case of FLR).
  *
  * Initialize the admin submission and completion queues.
  * Initialize the asynchronous events notification queues.
@@ -406,8 +443,7 @@ void ena_com_mmio_reg_read_request_destroy(struct ena_com_dev *ena_dev);
  * @return - 0 on success, negative value on failure.
  */
 int ena_com_admin_init(struct ena_com_dev *ena_dev,
-		       struct ena_aenq_handlers *aenq_handlers,
-		       bool init_spinlock);
+		       struct ena_aenq_handlers *aenq_handlers);
 
 /* ena_com_admin_destroy - Destroy the admin and the async events queues.
  * @ena_dev: ENA communication layer struct
@@ -935,6 +971,16 @@ void ena_com_get_intr_moderation_entry(struct ena_com_dev *ena_dev,
 				       enum ena_intr_moder_level level,
 				       struct ena_intr_moder_entry *entry);
 
+/* ena_com_config_dev_mode - Configure the placement policy of the device.
+ * @ena_dev: ENA communication layer struct
+ * @llq_features: LLQ feature descriptor, retrieve via
+ *                ena_com_get_dev_attr_feat.
+ * @ena_llq_config: The default driver LLQ parameters configurations
+ */
+int ena_com_config_dev_mode(struct ena_com_dev *ena_dev,
+			    struct ena_admin_feature_llq_desc *llq_features,
+			    struct ena_llq_configurations *llq_default_config);
+
 static inline bool ena_com_get_adaptive_moderation_enabled(struct ena_com_dev *ena_dev)
 {
 	return ena_dev->adaptive_coalescing;
@@ -1044,4 +1090,21 @@ static inline void ena_com_update_intr_reg(struct ena_eth_io_intr_reg *intr_reg,
 		intr_reg->intr_control |= ENA_ETH_IO_INTR_REG_INTR_UNMASK_MASK;
 }
 
+static inline u8 *ena_com_get_next_bounce_buffer(struct ena_com_io_bounce_buffer_control *bounce_buf_ctrl)
+{
+	u16 size, buffers_num;
+	u8 *buf;
+
+	size = bounce_buf_ctrl->buffer_size;
+	buffers_num = bounce_buf_ctrl->buffers_num;
+
+	buf = bounce_buf_ctrl->base_buffer +
+		(bounce_buf_ctrl->next_to_use++ & (buffers_num - 1)) * size;
+
+	prefetchw(bounce_buf_ctrl->base_buffer +
+		(bounce_buf_ctrl->next_to_use & (buffers_num - 1)) * size);
+
+	return buf;
+}
+
 #endif /* !(ENA_COM) */

drivers/net/ethernet/amazon/ena/ena_common_defs.h

@@ -32,8 +32,8 @@
 #ifndef _ENA_COMMON_H_
 #define _ENA_COMMON_H_
 
-#define ENA_COMMON_SPEC_VERSION_MAJOR	0 /*  */
-#define ENA_COMMON_SPEC_VERSION_MINOR	10 /*  */
+#define ENA_COMMON_SPEC_VERSION_MAJOR        2
+#define ENA_COMMON_SPEC_VERSION_MINOR        0
 
 /* ENA operates with 48-bit memory addresses. ena_mem_addr_t */
 struct ena_common_mem_addr {

drivers/net/ethernet/amazon/ena/ena_eth_com.c

@@ -59,16 +59,7 @@ static inline struct ena_eth_io_rx_cdesc_base *ena_com_get_next_rx_cdesc(
 	return cdesc;
 }
 
-static inline void ena_com_cq_inc_head(struct ena_com_io_cq *io_cq)
-{
-	io_cq->head++;
-
-	/* Switch phase bit in case of wrap around */
-	if (unlikely((io_cq->head & (io_cq->q_depth - 1)) == 0))
-		io_cq->phase ^= 1;
-}
-
-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;
@@ -80,45 +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;
 }
@@ -186,7 +291,7 @@ 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,
+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;
@@ -232,8 +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,
@@ -250,6 +354,9 @@ static inline void ena_com_rx_set_flags(struct ena_com_rx_ctx *ena_rx_ctx,
 	ena_rx_ctx->l4_csum_err =
 		(cdesc->status & ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_ERR_MASK) >>
 		ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_ERR_SHIFT;
+	ena_rx_ctx->l4_csum_checked =
+		!!((cdesc->status & ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_CHECKED_MASK) >>
+		ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_CHECKED_SHIFT);
 	ena_rx_ctx->hash = cdesc->hash;
 	ena_rx_ctx->frag =
 		(cdesc->status & ENA_ETH_IO_RX_CDESC_BASE_IPV4_FRAG_MASK) >>
@@ -271,18 +378,19 @@ 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;
 	}
 
@@ -292,23 +400,32 @@ 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 */
@@ -360,10 +477,14 @@ 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 <<
@@ -386,15 +507,14 @@ 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,
@@ -453,15 +573,18 @@ 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;
@@ -472,43 +595,7 @@ 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;
-}
-
-int ena_com_tx_comp_req_id_get(struct ena_com_io_cq *io_cq, u16 *req_id)
-{
-	u8 expected_phase, cdesc_phase;
-	struct ena_eth_io_tx_cdesc *cdesc;
-	u16 masked_head;
-
-	masked_head = io_cq->head & (io_cq->q_depth - 1);
-	expected_phase = io_cq->phase;
-
-	cdesc = (struct ena_eth_io_tx_cdesc *)
-		((uintptr_t)io_cq->cdesc_addr.virt_addr +
-		(masked_head * io_cq->cdesc_entry_size_in_bytes));
-
-	/* When the current completion descriptor phase isn't the same as the
-	 * expected, it mean that the device still didn't update
-	 * this completion.
-	 */
-	cdesc_phase = READ_ONCE(cdesc->flags) & ENA_ETH_IO_TX_CDESC_PHASE_MASK;
-	if (cdesc_phase != expected_phase)
-		return -EAGAIN;
-
-	dma_rmb();
-	if (unlikely(cdesc->req_id >= io_cq->q_depth)) {
-		pr_err("Invalid req id %d\n", cdesc->req_id);
-		return -EINVAL;
-	}
-
-	ena_com_cq_inc_head(io_cq);
-
-	*req_id = READ_ONCE(cdesc->req_id);
-
-	return 0;
+	return ena_com_sq_update_tail(io_sq);
 }
 
 bool ena_com_cq_empty(struct ena_com_io_cq *io_cq)

drivers/net/ethernet/amazon/ena/ena_eth_com.h

@@ -67,6 +67,7 @@ struct ena_com_rx_ctx {
 	enum ena_eth_io_l4_proto_index l4_proto;
 	bool l3_csum_err;
 	bool l4_csum_err;
+	u8 l4_csum_checked;
 	/* fragmented packet */
 	bool frag;
 	u32 hash;
@@ -86,8 +87,6 @@ int ena_com_add_single_rx_desc(struct ena_com_io_sq *io_sq,
 			       struct ena_com_buf *ena_buf,
 			       u16 req_id);
 
-int ena_com_tx_comp_req_id_get(struct ena_com_io_cq *io_cq, u16 *req_id);
-
 bool ena_com_cq_empty(struct ena_com_io_cq *io_cq);
 
 static inline void ena_com_unmask_intr(struct ena_com_io_cq *io_cq,
@@ -96,7 +95,7 @@ static inline void ena_com_unmask_intr(struct ena_com_io_cq *io_cq,
 	writel(intr_reg->intr_control, io_cq->unmask_reg);
 }
 
-static inline int ena_com_sq_empty_space(struct ena_com_io_sq *io_sq)
+static inline int ena_com_free_desc(struct ena_com_io_sq *io_sq)
 {
 	u16 tail, next_to_comp, cnt;
 
@@ -107,11 +106,28 @@ static inline int ena_com_sq_empty_space(struct ena_com_io_sq *io_sq)
 	return io_sq->q_depth - 1 - cnt;
 }
 
-static inline int ena_com_write_sq_doorbell(struct ena_com_io_sq *io_sq)
+/* Check if the submission queue has enough space to hold required_buffers */
+static inline bool ena_com_sq_have_enough_space(struct ena_com_io_sq *io_sq,
+						u16 required_buffers)
 {
-	u16 tail;
+	int temp;
 
-	tail = io_sq->tail;
+	if (io_sq->mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
+		return ena_com_free_desc(io_sq) >= required_buffers;
+
+	/* This calculation doesn't need to be 100% accurate. So to reduce
+	 * the calculation overhead just Subtract 2 lines from the free descs
+	 * (one for the header line and one to compensate the devision
+	 * down calculation.
+	 */
+	temp = required_buffers / io_sq->llq_info.descs_per_entry + 2;
+
+	return ena_com_free_desc(io_sq) > temp;
+}
+
+static inline int ena_com_write_sq_doorbell(struct ena_com_io_sq *io_sq)
+{
+	u16 tail = io_sq->tail;
 
 	pr_debug("write submission queue doorbell for queue: %d tail: %d\n",
 		 io_sq->qid, tail);
@@ -159,4 +175,48 @@ static inline void ena_com_comp_ack(struct ena_com_io_sq *io_sq, u16 elem)
 	io_sq->next_to_comp += elem;
 }
 
+static inline void ena_com_cq_inc_head(struct ena_com_io_cq *io_cq)
+{
+	io_cq->head++;
+
+	/* Switch phase bit in case of wrap around */
+	if (unlikely((io_cq->head & (io_cq->q_depth - 1)) == 0))
+		io_cq->phase ^= 1;
+}
+
+static inline int ena_com_tx_comp_req_id_get(struct ena_com_io_cq *io_cq,
+					     u16 *req_id)
+{
+	u8 expected_phase, cdesc_phase;
+	struct ena_eth_io_tx_cdesc *cdesc;
+	u16 masked_head;
+
+	masked_head = io_cq->head & (io_cq->q_depth - 1);
+	expected_phase = io_cq->phase;
+
+	cdesc = (struct ena_eth_io_tx_cdesc *)
+		((uintptr_t)io_cq->cdesc_addr.virt_addr +
+		(masked_head * io_cq->cdesc_entry_size_in_bytes));
+
+	/* When the current completion descriptor phase isn't the same as the
+	 * expected, it mean that the device still didn't update
+	 * this completion.
+	 */
+	cdesc_phase = READ_ONCE(cdesc->flags) & ENA_ETH_IO_TX_CDESC_PHASE_MASK;
+	if (cdesc_phase != expected_phase)
+		return -EAGAIN;
+
+	dma_rmb();
+
+	*req_id = READ_ONCE(cdesc->req_id);
+	if (unlikely(*req_id >= io_cq->q_depth)) {
+		pr_err("Invalid req id %d\n", cdesc->req_id);
+		return -EINVAL;
+	}
+
+	ena_com_cq_inc_head(io_cq);
+
+	return 0;
+}
+
 #endif /* ENA_ETH_COM_H_ */

drivers/net/ethernet/amazon/ena/ena_eth_io_defs.h

@@ -34,23 +34,16 @@
 
 enum ena_eth_io_l3_proto_index {
 	ENA_ETH_IO_L3_PROTO_UNKNOWN                 = 0,
-
 	ENA_ETH_IO_L3_PROTO_IPV4                    = 8,
-
 	ENA_ETH_IO_L3_PROTO_IPV6                    = 11,
-
 	ENA_ETH_IO_L3_PROTO_FCOE                    = 21,
-
 	ENA_ETH_IO_L3_PROTO_ROCE                    = 22,
 };
 
 enum ena_eth_io_l4_proto_index {
 	ENA_ETH_IO_L4_PROTO_UNKNOWN                 = 0,
-
 	ENA_ETH_IO_L4_PROTO_TCP                     = 12,
-
 	ENA_ETH_IO_L4_PROTO_UDP                     = 13,
-
 	ENA_ETH_IO_L4_PROTO_ROUTEABLE_ROCE          = 23,
 };
 
@@ -242,9 +235,13 @@ struct ena_eth_io_rx_cdesc_base {
 	 *    checksum error detected, or, the controller didn't
 	 *    validate the checksum. This bit is valid only when
 	 *    l4_proto_idx indicates TCP/UDP packet, and,
-	 *    ipv4_frag is not set
+	 *    ipv4_frag is not set. This bit is valid only when
+	 *    l4_csum_checked below is set.
 	 * 15 : ipv4_frag - Indicates IPv4 fragmented packet
-	 * 23:16 : reserved16
+	 * 16 : l4_csum_checked - L4 checksum was verified
+	 *    (could be OK or error), when cleared the status of
+	 *    checksum is unknown
+	 * 23:17 : reserved17 - MBZ
 	 * 24 : phase
 	 * 25 : l3_csum2 - second checksum engine result
 	 * 26 : first - Indicates first descriptor in
@@ -390,6 +387,8 @@ struct ena_eth_io_numa_node_cfg_reg {
 #define ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_ERR_MASK           BIT(14)
 #define ENA_ETH_IO_RX_CDESC_BASE_IPV4_FRAG_SHIFT            15
 #define ENA_ETH_IO_RX_CDESC_BASE_IPV4_FRAG_MASK             BIT(15)
+#define ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_CHECKED_SHIFT      16
+#define ENA_ETH_IO_RX_CDESC_BASE_L4_CSUM_CHECKED_MASK       BIT(16)
 #define ENA_ETH_IO_RX_CDESC_BASE_PHASE_SHIFT                24
 #define ENA_ETH_IO_RX_CDESC_BASE_PHASE_MASK                 BIT(24)
 #define ENA_ETH_IO_RX_CDESC_BASE_L3_CSUM2_SHIFT             25

drivers/net/ethernet/amazon/ena/ena_ethtool.c

@@ -81,6 +81,7 @@ static const struct ena_stats ena_stats_tx_strings[] = {
 	ENA_STAT_TX_ENTRY(doorbells),
 	ENA_STAT_TX_ENTRY(prepare_ctx_err),
 	ENA_STAT_TX_ENTRY(bad_req_id),
+	ENA_STAT_TX_ENTRY(llq_buffer_copy),
 	ENA_STAT_TX_ENTRY(missed_tx),
 };
 
@@ -96,6 +97,7 @@ static const struct ena_stats ena_stats_rx_strings[] = {
 	ENA_STAT_RX_ENTRY(rx_copybreak_pkt),
 	ENA_STAT_RX_ENTRY(bad_req_id),
 	ENA_STAT_RX_ENTRY(empty_rx_ring),
+	ENA_STAT_RX_ENTRY(csum_unchecked),
 };
 
 static const struct ena_stats ena_stats_ena_com_strings[] = {

drivers/net/ethernet/amazon/ena/ena_netdev.c

@@ -237,6 +237,17 @@ static int ena_setup_tx_resources(struct ena_adapter *adapter, int qid)
 		}
 	}
 
+	size = tx_ring->tx_max_header_size;
+	tx_ring->push_buf_intermediate_buf = vzalloc_node(size, node);
+	if (!tx_ring->push_buf_intermediate_buf) {
+		tx_ring->push_buf_intermediate_buf = vzalloc(size);
+		if (!tx_ring->push_buf_intermediate_buf) {
+			vfree(tx_ring->tx_buffer_info);
+			vfree(tx_ring->free_tx_ids);
+			return -ENOMEM;
+		}
+	}
+
 	/* Req id ring for TX out of order completions */
 	for (i = 0; i < tx_ring->ring_size; i++)
 		tx_ring->free_tx_ids[i] = i;
@@ -265,6 +276,9 @@ static void ena_free_tx_resources(struct ena_adapter *adapter, int qid)
 
 	vfree(tx_ring->free_tx_ids);
 	tx_ring->free_tx_ids = NULL;
+
+	vfree(tx_ring->push_buf_intermediate_buf);
+	tx_ring->push_buf_intermediate_buf = NULL;
 }
 
 /* ena_setup_all_tx_resources - allocate I/O Tx queues resources for All queues
@@ -602,6 +616,36 @@ static void ena_free_all_rx_bufs(struct ena_adapter *adapter)
 		ena_free_rx_bufs(adapter, i);
 }
 
+static inline void ena_unmap_tx_skb(struct ena_ring *tx_ring,
+				    struct ena_tx_buffer *tx_info)
+{
+	struct ena_com_buf *ena_buf;
+	u32 cnt;
+	int i;
+
+	ena_buf = tx_info->bufs;
+	cnt = tx_info->num_of_bufs;
+
+	if (unlikely(!cnt))
+		return;
+
+	if (tx_info->map_linear_data) {
+		dma_unmap_single(tx_ring->dev,
+				 dma_unmap_addr(ena_buf, paddr),
+				 dma_unmap_len(ena_buf, len),
+				 DMA_TO_DEVICE);
+		ena_buf++;
+		cnt--;
+	}
+
+	/* unmap remaining mapped pages */
+	for (i = 0; i < cnt; i++) {
+		dma_unmap_page(tx_ring->dev, dma_unmap_addr(ena_buf, paddr),
+			       dma_unmap_len(ena_buf, len), DMA_TO_DEVICE);
+		ena_buf++;
+	}
+}
+
 /* ena_free_tx_bufs - Free Tx Buffers per Queue
  * @tx_ring: TX ring for which buffers be freed
  */
@@ -612,9 +656,6 @@ static void ena_free_tx_bufs(struct ena_ring *tx_ring)
 
 	for (i = 0; i < tx_ring->ring_size; i++) {
 		struct ena_tx_buffer *tx_info = &tx_ring->tx_buffer_info[i];
-		struct ena_com_buf *ena_buf;
-		int nr_frags;
-		int j;
 
 		if (!tx_info->skb)
 			continue;
@@ -630,21 +671,7 @@ static void ena_free_tx_bufs(struct ena_ring *tx_ring)
 				   tx_ring->qid, i);
 		}
 
-		ena_buf = tx_info->bufs;
-		dma_unmap_single(tx_ring->dev,
-				 ena_buf->paddr,
-				 ena_buf->len,
-				 DMA_TO_DEVICE);
-
-		/* unmap remaining mapped pages */
-		nr_frags = tx_info->num_of_bufs - 1;
-		for (j = 0; j < nr_frags; j++) {
-			ena_buf++;
-			dma_unmap_page(tx_ring->dev,
-				       ena_buf->paddr,
-				       ena_buf->len,
-				       DMA_TO_DEVICE);
-		}
+		ena_unmap_tx_skb(tx_ring, tx_info);
 
 		dev_kfree_skb_any(tx_info->skb);
 	}
@@ -735,8 +762,6 @@ static int ena_clean_tx_irq(struct ena_ring *tx_ring, u32 budget)
 	while (tx_pkts < budget) {
 		struct ena_tx_buffer *tx_info;
 		struct sk_buff *skb;
-		struct ena_com_buf *ena_buf;
-		int i, nr_frags;
 
 		rc = ena_com_tx_comp_req_id_get(tx_ring->ena_com_io_cq,
 						&req_id);
@@ -756,24 +781,7 @@ static int ena_clean_tx_irq(struct ena_ring *tx_ring, u32 budget)
 		tx_info->skb = NULL;
 		tx_info->last_jiffies = 0;
 
-		if (likely(tx_info->num_of_bufs != 0)) {
-			ena_buf = tx_info->bufs;
-
-			dma_unmap_single(tx_ring->dev,
-					 dma_unmap_addr(ena_buf, paddr),
-					 dma_unmap_len(ena_buf, len),
-					 DMA_TO_DEVICE);
-
-			/* unmap remaining mapped pages */
-			nr_frags = tx_info->num_of_bufs - 1;
-			for (i = 0; i < nr_frags; i++) {
-				ena_buf++;
-				dma_unmap_page(tx_ring->dev,
-					       dma_unmap_addr(ena_buf, paddr),
-					       dma_unmap_len(ena_buf, len),
-					       DMA_TO_DEVICE);
-			}
-		}
+		ena_unmap_tx_skb(tx_ring, tx_info);
 
 		netif_dbg(tx_ring->adapter, tx_done, tx_ring->netdev,
 			  "tx_poll: q %d skb %p completed\n", tx_ring->qid,
@@ -804,12 +812,13 @@ static int ena_clean_tx_irq(struct ena_ring *tx_ring, u32 budget)
 	 */
 	smp_mb();
 
-	above_thresh = ena_com_sq_empty_space(tx_ring->ena_com_io_sq) >
-		ENA_TX_WAKEUP_THRESH;
+	above_thresh = ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
+						    ENA_TX_WAKEUP_THRESH);
 	if (unlikely(netif_tx_queue_stopped(txq) && above_thresh)) {
 		__netif_tx_lock(txq, smp_processor_id());
-		above_thresh = ena_com_sq_empty_space(tx_ring->ena_com_io_sq) >
-			ENA_TX_WAKEUP_THRESH;
+		above_thresh =
+			ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
+						     ENA_TX_WAKEUP_THRESH);
 		if (netif_tx_queue_stopped(txq) && above_thresh) {
 			netif_tx_wake_queue(txq);
 			u64_stats_update_begin(&tx_ring->syncp);
@@ -985,8 +994,19 @@ static inline void ena_rx_checksum(struct ena_ring *rx_ring,
 			return;
 		}
 
+		if (likely(ena_rx_ctx->l4_csum_checked)) {
 			skb->ip_summed = CHECKSUM_UNNECESSARY;
+		} else {
+			u64_stats_update_begin(&rx_ring->syncp);
+			rx_ring->rx_stats.csum_unchecked++;
+			u64_stats_update_end(&rx_ring->syncp);
+			skb->ip_summed = CHECKSUM_NONE;
 		}
+	} else {
+		skb->ip_summed = CHECKSUM_NONE;
+		return;
+	}
+
 }
 
 static void ena_set_rx_hash(struct ena_ring *rx_ring,
@@ -1101,8 +1121,10 @@ static int ena_clean_rx_irq(struct ena_ring *rx_ring, struct napi_struct *napi,
 
 	rx_ring->next_to_clean = next_to_clean;
 
-	refill_required = ena_com_sq_empty_space(rx_ring->ena_com_io_sq);
-	refill_threshold = rx_ring->ring_size / ENA_RX_REFILL_THRESH_DIVIDER;
+	refill_required = ena_com_free_desc(rx_ring->ena_com_io_sq);
+	refill_threshold =
+		min_t(int, rx_ring->ring_size / ENA_RX_REFILL_THRESH_DIVIDER,
+		      ENA_RX_REFILL_THRESH_PACKET);
 
 	/* Optimization, try to batch new rx buffers */
 	if (refill_required > refill_threshold) {
@@ -1299,7 +1321,6 @@ static int ena_enable_msix(struct ena_adapter *adapter, int num_queues)
 
 	/* Reserved the max msix vectors we might need */
 	msix_vecs = ENA_MAX_MSIX_VEC(num_queues);
-
 	netif_dbg(adapter, probe, adapter->netdev,
 		  "trying to enable MSI-X, vectors %d\n", msix_vecs);
 
@@ -1592,7 +1613,7 @@ static int ena_up_complete(struct ena_adapter *adapter)
 
 static int ena_create_io_tx_queue(struct ena_adapter *adapter, int qid)
 {
-	struct ena_com_create_io_ctx ctx = { 0 };
+	struct ena_com_create_io_ctx ctx;
 	struct ena_com_dev *ena_dev;
 	struct ena_ring *tx_ring;
 	u32 msix_vector;
@@ -1605,6 +1626,8 @@ static int ena_create_io_tx_queue(struct ena_adapter *adapter, int qid)
 	msix_vector = ENA_IO_IRQ_IDX(qid);
 	ena_qid = ENA_IO_TXQ_IDX(qid);
 
+	memset(&ctx, 0x0, sizeof(ctx));
+
 	ctx.direction = ENA_COM_IO_QUEUE_DIRECTION_TX;
 	ctx.qid = ena_qid;
 	ctx.mem_queue_type = ena_dev->tx_mem_queue_type;
@@ -1658,7 +1681,7 @@ static int ena_create_all_io_tx_queues(struct ena_adapter *adapter)
 static int ena_create_io_rx_queue(struct ena_adapter *adapter, int qid)
 {
 	struct ena_com_dev *ena_dev;
-	struct ena_com_create_io_ctx ctx = { 0 };
+	struct ena_com_create_io_ctx ctx;
 	struct ena_ring *rx_ring;
 	u32 msix_vector;
 	u16 ena_qid;
@@ -1670,6 +1693,8 @@ static int ena_create_io_rx_queue(struct ena_adapter *adapter, int qid)
 	msix_vector = ENA_IO_IRQ_IDX(qid);
 	ena_qid = ENA_IO_RXQ_IDX(qid);
 
+	memset(&ctx, 0x0, sizeof(ctx));
+
 	ctx.qid = ena_qid;
 	ctx.direction = ENA_COM_IO_QUEUE_DIRECTION_RX;
 	ctx.mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_HOST;
@@ -1980,73 +2005,70 @@ static int ena_check_and_linearize_skb(struct ena_ring *tx_ring,
 	return rc;
 }
 
-/* Called with netif_tx_lock. */
-static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
+static int ena_tx_map_skb(struct ena_ring *tx_ring,
+			  struct ena_tx_buffer *tx_info,
+			  struct sk_buff *skb,
+			  void **push_hdr,
+			  u16 *header_len)
 {
-	struct ena_adapter *adapter = netdev_priv(dev);
-	struct ena_tx_buffer *tx_info;
-	struct ena_com_tx_ctx ena_tx_ctx;
-	struct ena_ring *tx_ring;
-	struct netdev_queue *txq;
+	struct ena_adapter *adapter = tx_ring->adapter;
 	struct ena_com_buf *ena_buf;
-	void *push_hdr;
-	u32 len, last_frag;
-	u16 next_to_use;
-	u16 req_id;
-	u16 push_len;
-	u16 header_len;
 	dma_addr_t dma;
-	int qid, rc, nb_hw_desc;
-	int i = -1;
-
-	netif_dbg(adapter, tx_queued, dev, "%s skb %p\n", __func__, skb);
-	/*  Determine which tx ring we will be placed on */
-	qid = skb_get_queue_mapping(skb);
-	tx_ring = &adapter->tx_ring[qid];
-	txq = netdev_get_tx_queue(dev, qid);
+	u32 skb_head_len, frag_len, last_frag;
+	u16 push_len = 0;
+	u16 delta = 0;
+	int i = 0;
 
-	rc = ena_check_and_linearize_skb(tx_ring, skb);
-	if (unlikely(rc))
-		goto error_drop_packet;
-
-	skb_tx_timestamp(skb);
-	len = skb_headlen(skb);
-
-	next_to_use = tx_ring->next_to_use;
-	req_id = tx_ring->free_tx_ids[next_to_use];
-	tx_info = &tx_ring->tx_buffer_info[req_id];
-	tx_info->num_of_bufs = 0;
-
-	WARN(tx_info->skb, "SKB isn't NULL req_id %d\n", req_id);
-	ena_buf = tx_info->bufs;
+	skb_head_len = skb_headlen(skb);
 	tx_info->skb = skb;
+	ena_buf = tx_info->bufs;
 
 	if (tx_ring->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV) {
-		/* prepared the push buffer */
-		push_len = min_t(u32, len, tx_ring->tx_max_header_size);
-		header_len = push_len;
-		push_hdr = skb->data;
+		/* When the device is LLQ mode, the driver will copy
+		 * the header into the device memory space.
+		 * the ena_com layer assume the header is in a linear
+		 * memory space.
+		 * This assumption might be wrong since part of the header
+		 * can be in the fragmented buffers.
+		 * Use skb_header_pointer to make sure the header is in a
+		 * linear memory space.
+		 */
+
+		push_len = min_t(u32, skb->len, tx_ring->tx_max_header_size);
+		*push_hdr = skb_header_pointer(skb, 0, push_len,
+					       tx_ring->push_buf_intermediate_buf);
+		*header_len = push_len;
+		if (unlikely(skb->data != *push_hdr)) {
+			u64_stats_update_begin(&tx_ring->syncp);
+			tx_ring->tx_stats.llq_buffer_copy++;
+			u64_stats_update_end(&tx_ring->syncp);
+
+			delta = push_len - skb_head_len;
+		}
 	} else {
-		push_len = 0;
-		header_len = min_t(u32, len, tx_ring->tx_max_header_size);
-		push_hdr = NULL;
+		*push_hdr = NULL;
+		*header_len = min_t(u32, skb_head_len,
+				    tx_ring->tx_max_header_size);
 	}
 
-	netif_dbg(adapter, tx_queued, dev,
+	netif_dbg(adapter, tx_queued, adapter->netdev,
 		  "skb: %p header_buf->vaddr: %p push_len: %d\n", skb,
-		  push_hdr, push_len);
+		  *push_hdr, push_len);
 
-	if (len > push_len) {
+	if (skb_head_len > push_len) {
 		dma = dma_map_single(tx_ring->dev, skb->data + push_len,
-				     len - push_len, DMA_TO_DEVICE);
-		if (dma_mapping_error(tx_ring->dev, dma))
+				     skb_head_len - push_len, DMA_TO_DEVICE);
+		if (unlikely(dma_mapping_error(tx_ring->dev, dma)))
 			goto error_report_dma_error;
 
 		ena_buf->paddr = dma;
-		ena_buf->len = len - push_len;
+		ena_buf->len = skb_head_len - push_len;
 
 		ena_buf++;
 		tx_info->num_of_bufs++;
+		tx_info->map_linear_data = 1;
+	} else {
+		tx_info->map_linear_data = 0;
 	}
 
 	last_frag = skb_shinfo(skb)->nr_frags;
@@ -2054,18 +2076,75 @@ static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	for (i = 0; i < last_frag; i++) {
 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
 
-		len = skb_frag_size(frag);
-		dma = skb_frag_dma_map(tx_ring->dev, frag, 0, len,
-				       DMA_TO_DEVICE);
-		if (dma_mapping_error(tx_ring->dev, dma))
+		frag_len = skb_frag_size(frag);
+
+		if (unlikely(delta >= frag_len)) {
+			delta -= frag_len;
+			continue;
+		}
+
+		dma = skb_frag_dma_map(tx_ring->dev, frag, delta,
+				       frag_len - delta, DMA_TO_DEVICE);
+		if (unlikely(dma_mapping_error(tx_ring->dev, dma)))
 			goto error_report_dma_error;
 
 		ena_buf->paddr = dma;
-		ena_buf->len = len;
+		ena_buf->len = frag_len - delta;
 		ena_buf++;
+		tx_info->num_of_bufs++;
+		delta = 0;
 	}
 
-	tx_info->num_of_bufs += last_frag;
+	return 0;
+
+error_report_dma_error:
+	u64_stats_update_begin(&tx_ring->syncp);
+	tx_ring->tx_stats.dma_mapping_err++;
+	u64_stats_update_end(&tx_ring->syncp);
+	netdev_warn(adapter->netdev, "failed to map skb\n");
+
+	tx_info->skb = NULL;
+
+	tx_info->num_of_bufs += i;
+	ena_unmap_tx_skb(tx_ring, tx_info);
+
+	return -EINVAL;
+}
+
+/* Called with netif_tx_lock. */
+static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	struct ena_adapter *adapter = netdev_priv(dev);
+	struct ena_tx_buffer *tx_info;
+	struct ena_com_tx_ctx ena_tx_ctx;
+	struct ena_ring *tx_ring;
+	struct netdev_queue *txq;
+	void *push_hdr;
+	u16 next_to_use, req_id, header_len;
+	int qid, rc, nb_hw_desc;
+
+	netif_dbg(adapter, tx_queued, dev, "%s skb %p\n", __func__, skb);
+	/*  Determine which tx ring we will be placed on */
+	qid = skb_get_queue_mapping(skb);
+	tx_ring = &adapter->tx_ring[qid];
+	txq = netdev_get_tx_queue(dev, qid);
+
+	rc = ena_check_and_linearize_skb(tx_ring, skb);
+	if (unlikely(rc))
+		goto error_drop_packet;
+
+	skb_tx_timestamp(skb);
+
+	next_to_use = tx_ring->next_to_use;
+	req_id = tx_ring->free_tx_ids[next_to_use];
+	tx_info = &tx_ring->tx_buffer_info[req_id];
+	tx_info->num_of_bufs = 0;
+
+	WARN(tx_info->skb, "SKB isn't NULL req_id %d\n", req_id);
+
+	rc = ena_tx_map_skb(tx_ring, tx_info, skb, &push_hdr, &header_len);
+	if (unlikely(rc))
+		goto error_drop_packet;
 
 	memset(&ena_tx_ctx, 0x0, sizeof(struct ena_com_tx_ctx));
 	ena_tx_ctx.ena_bufs = tx_info->bufs;
@@ -2081,14 +2160,22 @@ static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	rc = ena_com_prepare_tx(tx_ring->ena_com_io_sq, &ena_tx_ctx,
 				&nb_hw_desc);
 
+	/* ena_com_prepare_tx() can't fail due to overflow of tx queue,
+	 * since the number of free descriptors in the queue is checked
+	 * after sending the previous packet. In case there isn't enough
+	 * space in the queue for the next packet, it is stopped
+	 * until there is again enough available space in the queue.
+	 * All other failure reasons of ena_com_prepare_tx() are fatal
+	 * and therefore require a device reset.
+	 */
 	if (unlikely(rc)) {
 		netif_err(adapter, tx_queued, dev,
 			  "failed to prepare tx bufs\n");
 		u64_stats_update_begin(&tx_ring->syncp);
-		tx_ring->tx_stats.queue_stop++;
 		tx_ring->tx_stats.prepare_ctx_err++;
 		u64_stats_update_end(&tx_ring->syncp);
-		netif_tx_stop_queue(txq);
+		adapter->reset_reason = ENA_REGS_RESET_DRIVER_INVALID_STATE;
+		set_bit(ENA_FLAG_TRIGGER_RESET, &adapter->flags);
 		goto error_unmap_dma;
 	}
 
@@ -2110,8 +2197,8 @@ static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	 * to sgl_size + 2. one for the meta descriptor and one for header
 	 * (if the header is larger than tx_max_header_size).
 	 */
-	if (unlikely(ena_com_sq_empty_space(tx_ring->ena_com_io_sq) <
-		     (tx_ring->sgl_size + 2))) {
+	if (unlikely(!ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
+						   tx_ring->sgl_size + 2))) {
 		netif_dbg(adapter, tx_queued, dev, "%s stop queue %d\n",
 			  __func__, qid);
 
@@ -2130,8 +2217,8 @@ static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		 */
 		smp_mb();
 
-		if (ena_com_sq_empty_space(tx_ring->ena_com_io_sq)
-				> ENA_TX_WAKEUP_THRESH) {
+		if (ena_com_sq_have_enough_space(tx_ring->ena_com_io_sq,
+						 ENA_TX_WAKEUP_THRESH)) {
 			netif_tx_wake_queue(txq);
 			u64_stats_update_begin(&tx_ring->syncp);
 			tx_ring->tx_stats.queue_wakeup++;
@@ -2151,35 +2238,11 @@ static netdev_tx_t ena_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	return NETDEV_TX_OK;
 
-error_report_dma_error:
-	u64_stats_update_begin(&tx_ring->syncp);
-	tx_ring->tx_stats.dma_mapping_err++;
-	u64_stats_update_end(&tx_ring->syncp);
-	netdev_warn(adapter->netdev, "failed to map skb\n");
-
-	tx_info->skb = NULL;
-
 error_unmap_dma:
-	if (i >= 0) {
-		/* save value of frag that failed */
-		last_frag = i;
-
-		/* start back at beginning and unmap skb */
+	ena_unmap_tx_skb(tx_ring, tx_info);
 	tx_info->skb = NULL;
-		ena_buf = tx_info->bufs;
-		dma_unmap_single(tx_ring->dev, dma_unmap_addr(ena_buf, paddr),
-				 dma_unmap_len(ena_buf, len), DMA_TO_DEVICE);
-
-		/* unmap remaining mapped pages */
-		for (i = 0; i < last_frag; i++) {
-			ena_buf++;
-			dma_unmap_page(tx_ring->dev, dma_unmap_addr(ena_buf, paddr),
-				       dma_unmap_len(ena_buf, len), DMA_TO_DEVICE);
-		}
-	}
 
 error_drop_packet:
-
 	dev_kfree_skb(skb);
 	return NETDEV_TX_OK;
 }
@@ -2201,7 +2264,8 @@ static u16 ena_select_queue(struct net_device *dev, struct sk_buff *skb,
 	return qid;
 }
 
-static void ena_config_host_info(struct ena_com_dev *ena_dev)
+static void ena_config_host_info(struct ena_com_dev *ena_dev,
+				 struct pci_dev *pdev)
 {
 	struct ena_admin_host_info *host_info;
 	int rc;
@@ -2215,6 +2279,7 @@ static void ena_config_host_info(struct ena_com_dev *ena_dev)
 
 	host_info = ena_dev->host_attr.host_info;
 
+	host_info->bdf = (pdev->bus->number << 8) | pdev->devfn;
 	host_info->os_type = ENA_ADMIN_OS_LINUX;
 	host_info->kernel_ver = LINUX_VERSION_CODE;
 	strncpy(host_info->kernel_ver_str, utsname()->version,
@@ -2225,7 +2290,9 @@ static void ena_config_host_info(struct ena_com_dev *ena_dev)
 	host_info->driver_version =
 		(DRV_MODULE_VER_MAJOR) |
 		(DRV_MODULE_VER_MINOR << ENA_ADMIN_HOST_INFO_MINOR_SHIFT) |
-		(DRV_MODULE_VER_SUBMINOR << ENA_ADMIN_HOST_INFO_SUB_MINOR_SHIFT);
+		(DRV_MODULE_VER_SUBMINOR << ENA_ADMIN_HOST_INFO_SUB_MINOR_SHIFT) |
+		("K"[0] << ENA_ADMIN_HOST_INFO_MODULE_TYPE_SHIFT);
+	host_info->num_cpus = num_online_cpus();
 
 	rc = ena_com_set_host_attributes(ena_dev);
 	if (rc) {
@@ -2436,7 +2503,7 @@ static int ena_device_init(struct ena_com_dev *ena_dev, struct pci_dev *pdev,
 	}
 
 	/* ENA admin level init */
-	rc = ena_com_admin_init(ena_dev, &aenq_handlers, true);
+	rc = ena_com_admin_init(ena_dev, &aenq_handlers);
 	if (rc) {
 		dev_err(dev,
 			"Can not initialize ena admin queue with device\n");
@@ -2449,7 +2516,7 @@ static int ena_device_init(struct ena_com_dev *ena_dev, struct pci_dev *pdev,
 	 */
 	ena_com_set_admin_polling_mode(ena_dev, true);
 
-	ena_config_host_info(ena_dev);
+	ena_config_host_info(ena_dev, pdev);
 
 	/* Get Device Attributes*/
 	rc = ena_com_get_dev_attr_feat(ena_dev, get_feat_ctx);
@@ -2534,15 +2601,14 @@ static void ena_destroy_device(struct ena_adapter *adapter, bool graceful)
 
 	dev_up = test_bit(ENA_FLAG_DEV_UP, &adapter->flags);
 	adapter->dev_up_before_reset = dev_up;
-
 	if (!graceful)
 		ena_com_set_admin_running_state(ena_dev, false);
 
 	if (test_bit(ENA_FLAG_DEV_UP, &adapter->flags))
 		ena_down(adapter);
 
-	/* Before releasing the ENA resources, a device reset is required.
-	 * (to prevent the device from accessing them).
+	/* Stop the device from sending AENQ events (in case reset flag is set
+	 *  and device is up, ena_close already reset the device
 	 * In case the reset flag is set and the device is up, ena_down()
 	 * already perform the reset, so it can be skipped.
 	 */
@@ -2611,7 +2677,9 @@ static int ena_restore_device(struct ena_adapter *adapter)
 
 	set_bit(ENA_FLAG_DEVICE_RUNNING, &adapter->flags);
 	mod_timer(&adapter->timer_service, round_jiffies(jiffies + HZ));
-	dev_err(&pdev->dev, "Device reset completed successfully\n");
+	dev_err(&pdev->dev,
+		"Device reset completed successfully, Driver info: %s\n",
+		version);
 
 	return rc;
 err_disable_msix:
@@ -2800,7 +2868,7 @@ static void check_for_empty_rx_ring(struct ena_adapter *adapter)
 		rx_ring = &adapter->rx_ring[i];
 
 		refill_required =
-			ena_com_sq_empty_space(rx_ring->ena_com_io_sq);
+			ena_com_free_desc(rx_ring->ena_com_io_sq);
 		if (unlikely(refill_required == (rx_ring->ring_size - 1))) {
 			rx_ring->empty_rx_queue++;
 
@@ -2946,7 +3014,7 @@ static int ena_calc_io_queue_num(struct pci_dev *pdev,
 
 	/* In case of LLQ use the llq number in the get feature cmd */
 	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV) {
-		io_sq_num = get_feat_ctx->max_queues.max_llq_num;
+		io_sq_num = get_feat_ctx->max_queues.max_legacy_llq_num;
 
 		if (io_sq_num == 0) {
 			dev_err(&pdev->dev,
@@ -2974,18 +3042,52 @@ static int ena_calc_io_queue_num(struct pci_dev *pdev,
 	return io_queue_num;
 }
 
-static void ena_set_push_mode(struct pci_dev *pdev, struct ena_com_dev *ena_dev,
-			      struct ena_com_dev_get_features_ctx *get_feat_ctx)
+static int ena_set_queues_placement_policy(struct pci_dev *pdev,
+					   struct ena_com_dev *ena_dev,
+					   struct ena_admin_feature_llq_desc *llq,
+					   struct ena_llq_configurations *llq_default_configurations)
 {
 	bool has_mem_bar;
+	int rc;
+	u32 llq_feature_mask;
+
+	llq_feature_mask = 1 << ENA_ADMIN_LLQ;
+	if (!(ena_dev->supported_features & llq_feature_mask)) {
+		dev_err(&pdev->dev,
+			"LLQ is not supported Fallback to host mode policy.\n");
+		ena_dev->tx_mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_HOST;
+		return 0;
+	}
 
 	has_mem_bar = pci_select_bars(pdev, IORESOURCE_MEM) & BIT(ENA_MEM_BAR);
 
-	/* Enable push mode if device supports LLQ */
-	if (has_mem_bar && (get_feat_ctx->max_queues.max_llq_num > 0))
-		ena_dev->tx_mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_DEV;
-	else
+	rc = ena_com_config_dev_mode(ena_dev, llq, llq_default_configurations);
+	if (unlikely(rc)) {
+		dev_err(&pdev->dev,
+			"Failed to configure the device mode.  Fallback to host mode policy.\n");
 		ena_dev->tx_mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_HOST;
+		return 0;
+	}
+
+	/* Nothing to config, exit */
+	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
+		return 0;
+
+	if (!has_mem_bar) {
+		dev_err(&pdev->dev,
+			"ENA device does not expose LLQ bar. Fallback to host mode policy.\n");
+		ena_dev->tx_mem_queue_type = ENA_ADMIN_PLACEMENT_POLICY_HOST;
+		return 0;
+	}
+
+	ena_dev->mem_bar = devm_ioremap_wc(&pdev->dev,
+					   pci_resource_start(pdev, ENA_MEM_BAR),
+					   pci_resource_len(pdev, ENA_MEM_BAR));
+
+	if (!ena_dev->mem_bar)
+		return -EFAULT;
+
+	return 0;
 }
 
 static void ena_set_dev_offloads(struct ena_com_dev_get_features_ctx *feat,
@@ -3110,6 +3212,15 @@ static void ena_release_bars(struct ena_com_dev *ena_dev, struct pci_dev *pdev)
 	pci_release_selected_regions(pdev, release_bars);
 }
 
+static inline void set_default_llq_configurations(struct ena_llq_configurations *llq_config)
+{
+	llq_config->llq_header_location = ENA_ADMIN_INLINE_HEADER;
+	llq_config->llq_ring_entry_size = ENA_ADMIN_LIST_ENTRY_SIZE_128B;
+	llq_config->llq_stride_ctrl = ENA_ADMIN_MULTIPLE_DESCS_PER_ENTRY;
+	llq_config->llq_num_decs_before_header = ENA_ADMIN_LLQ_NUM_DESCS_BEFORE_HEADER_2;
+	llq_config->llq_ring_entry_size_value = 128;
+}
+
 static int ena_calc_queue_size(struct pci_dev *pdev,
 			       struct ena_com_dev *ena_dev,
 			       u16 *max_tx_sgl_size,
@@ -3125,7 +3236,7 @@ static int ena_calc_queue_size(struct pci_dev *pdev,
 
 	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV)
 		queue_size = min_t(u32, queue_size,
-				   get_feat_ctx->max_queues.max_llq_depth);
+				   get_feat_ctx->max_queues.max_legacy_llq_depth);
 
 	queue_size = rounddown_pow_of_two(queue_size);
 
@@ -3158,7 +3269,9 @@ static int ena_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	static int version_printed;
 	struct net_device *netdev;
 	struct ena_adapter *adapter;
+	struct ena_llq_configurations llq_config;
 	struct ena_com_dev *ena_dev = NULL;
+	char *queue_type_str;
 	static int adapters_found;
 	int io_queue_num, bars, rc;
 	int queue_size;
@@ -3212,17 +3325,14 @@ static int ena_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 		goto err_free_region;
 	}
 
-	ena_set_push_mode(pdev, ena_dev, &get_feat_ctx);
+	set_default_llq_configurations(&llq_config);
 
-	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV) {
-		ena_dev->mem_bar = devm_ioremap_wc(&pdev->dev,
-						   pci_resource_start(pdev, ENA_MEM_BAR),
-						   pci_resource_len(pdev, ENA_MEM_BAR));
-		if (!ena_dev->mem_bar) {
-			rc = -EFAULT;
+	rc = ena_set_queues_placement_policy(pdev, ena_dev, &get_feat_ctx.llq,
+					     &llq_config);
+	if (rc) {
+		dev_err(&pdev->dev, "ena device init failed\n");
 		goto err_device_destroy;
 	}
-	}
 
 	/* initial Tx interrupt delay, Assumes 1 usec granularity.
 	* Updated during device initialization with the real granularity
@@ -3236,8 +3346,10 @@ static int ena_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 		goto err_device_destroy;
 	}
 
-	dev_info(&pdev->dev, "creating %d io queues. queue size: %d\n",
-		 io_queue_num, queue_size);
+	dev_info(&pdev->dev, "creating %d io queues. queue size: %d. LLQ is %s\n",
+		 io_queue_num, queue_size,
+		 (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_DEV) ?
+		 "ENABLED" : "DISABLED");
 
 	/* dev zeroed in init_etherdev */
 	netdev = alloc_etherdev_mq(sizeof(struct ena_adapter), io_queue_num);
@@ -3327,9 +3439,15 @@ static int ena_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
 	timer_setup(&adapter->timer_service, ena_timer_service, 0);
 	mod_timer(&adapter->timer_service, round_jiffies(jiffies + HZ));
 
-	dev_info(&pdev->dev, "%s found at mem %lx, mac addr %pM Queues %d\n",
+	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
+		queue_type_str = "Regular";
+	else
+		queue_type_str = "Low Latency";
+
+	dev_info(&pdev->dev,
+		 "%s found at mem %lx, mac addr %pM Queues %d, Placement policy: %s\n",
 		 DEVICE_NAME, (long)pci_resource_start(pdev, 0),
-		 netdev->dev_addr, io_queue_num);
+		 netdev->dev_addr, io_queue_num, queue_type_str);
 
 	set_bit(ENA_FLAG_DEVICE_RUNNING, &adapter->flags);
 

drivers/net/ethernet/amazon/ena/ena_netdev.h

@@ -43,9 +43,9 @@
 #include "ena_com.h"
 #include "ena_eth_com.h"
 
-#define DRV_MODULE_VER_MAJOR	1
-#define DRV_MODULE_VER_MINOR	5
-#define DRV_MODULE_VER_SUBMINOR 0
+#define DRV_MODULE_VER_MAJOR	2
+#define DRV_MODULE_VER_MINOR	0
+#define DRV_MODULE_VER_SUBMINOR 1
 
 #define DRV_MODULE_NAME		"ena"
 #ifndef DRV_MODULE_VERSION
@@ -61,6 +61,17 @@
 #define ENA_ADMIN_MSIX_VEC		1
 #define ENA_MAX_MSIX_VEC(io_queues)	(ENA_ADMIN_MSIX_VEC + (io_queues))
 
+/* The ENA buffer length fields is 16 bit long. So when PAGE_SIZE == 64kB the
+ * driver passes 0.
+ * Since the max packet size the ENA handles is ~9kB limit the buffer length to
+ * 16kB.
+ */
+#if PAGE_SIZE > SZ_16K
+#define ENA_PAGE_SIZE SZ_16K
+#else
+#define ENA_PAGE_SIZE PAGE_SIZE
+#endif
+
 #define ENA_MIN_MSIX_VEC		2
 
 #define ENA_REG_BAR			0
@@ -70,7 +81,7 @@
 #define ENA_DEFAULT_RING_SIZE	(1024)
 
 #define ENA_TX_WAKEUP_THRESH		(MAX_SKB_FRAGS + 2)
-#define ENA_DEFAULT_RX_COPYBREAK	(128 - NET_IP_ALIGN)
+#define ENA_DEFAULT_RX_COPYBREAK	(256 - NET_IP_ALIGN)
 
 /* limit the buffer size to 600 bytes to handle MTU changes from very
  * small to very large, in which case the number of buffers per packet
@@ -95,10 +106,11 @@
  */
 #define ENA_TX_POLL_BUDGET_DIVIDER	4
 
-/* Refill Rx queue when number of available descriptors is below
- * QUEUE_SIZE / ENA_RX_REFILL_THRESH_DIVIDER
+/* Refill Rx queue when number of required descriptors is above
+ * QUEUE_SIZE / ENA_RX_REFILL_THRESH_DIVIDER or ENA_RX_REFILL_THRESH_PACKET
  */
 #define ENA_RX_REFILL_THRESH_DIVIDER	8
+#define ENA_RX_REFILL_THRESH_PACKET	256
 
 /* Number of queues to check for missing queues per timer service */
 #define ENA_MONITORED_TX_QUEUES	4
@@ -151,6 +163,9 @@ struct ena_tx_buffer {
 	/* num of buffers used by this skb */
 	u32 num_of_bufs;
 
+	/* Indicate if bufs[0] map the linear data of the skb. */
+	u8 map_linear_data;
+
 	/* Used for detect missing tx packets to limit the number of prints */
 	u32 print_once;
 	/* Save the last jiffies to detect missing tx packets
@@ -186,6 +201,7 @@ struct ena_stats_tx {
 	u64 tx_poll;
 	u64 doorbells;
 	u64 bad_req_id;
+	u64 llq_buffer_copy;
 	u64 missed_tx;
 };
 
@@ -201,6 +217,7 @@ struct ena_stats_rx {
 	u64 rx_copybreak_pkt;
 	u64 bad_req_id;
 	u64 empty_rx_ring;
+	u64 csum_unchecked;
 };
 
 struct ena_ring {
@@ -257,6 +274,8 @@ struct ena_ring {
 		struct ena_stats_tx tx_stats;
 		struct ena_stats_rx rx_stats;
 	};
+
+	u8 *push_buf_intermediate_buf;
 	int empty_rx_queue;
 } ____cacheline_aligned;
 
@@ -355,15 +374,4 @@ void ena_dump_stats_to_buf(struct ena_adapter *adapter, u8 *buf);
 
 int ena_get_sset_count(struct net_device *netdev, int sset);
 
-/* The ENA buffer length fields is 16 bit long. So when PAGE_SIZE == 64kB the
- * driver passas 0.
- * Since the max packet size the ENA handles is ~9kB limit the buffer length to
- * 16kB.
- */
-#if PAGE_SIZE > SZ_16K
-#define ENA_PAGE_SIZE SZ_16K
-#else
-#define ENA_PAGE_SIZE PAGE_SIZE
-#endif
-
 #endif /* !(ENA_H) */

drivers/net/ethernet/amazon/ena/ena_regs_defs.h

@@ -34,37 +34,25 @@
 
 enum ena_regs_reset_reason_types {
 	ENA_REGS_RESET_NORMAL                       = 0,
-
 	ENA_REGS_RESET_KEEP_ALIVE_TO                = 1,
-
 	ENA_REGS_RESET_ADMIN_TO                     = 2,
-
 	ENA_REGS_RESET_MISS_TX_CMPL                 = 3,
-
 	ENA_REGS_RESET_INV_RX_REQ_ID                = 4,
-
 	ENA_REGS_RESET_INV_TX_REQ_ID                = 5,
-
 	ENA_REGS_RESET_TOO_MANY_RX_DESCS            = 6,
-
 	ENA_REGS_RESET_INIT_ERR                     = 7,
-
 	ENA_REGS_RESET_DRIVER_INVALID_STATE         = 8,
-
 	ENA_REGS_RESET_OS_TRIGGER                   = 9,
-
 	ENA_REGS_RESET_OS_NETDEV_WD                 = 10,
-
 	ENA_REGS_RESET_SHUTDOWN                     = 11,
-
 	ENA_REGS_RESET_USER_TRIGGER                 = 12,
-
 	ENA_REGS_RESET_GENERIC                      = 13,
-
 	ENA_REGS_RESET_MISS_INTERRUPT               = 14,
 };
 
 /* ena_registers offsets */
+
+/* 0 base */
 #define ENA_REGS_VERSION_OFF                                0x0
 #define ENA_REGS_CONTROLLER_VERSION_OFF                     0x4
 #define ENA_REGS_CAPS_OFF                                   0x8