qed: Add support for tlv request processing.
authorSudarsana Reddy Kalluru <sudarsana.kalluru@cavium.com>
Tue, 22 May 2018 07:28:38 +0000 (00:28 -0700)
committerDavid S. Miller <davem@davemloft.net>
Wed, 23 May 2018 03:29:53 +0000 (23:29 -0400)
The patch adds driver support for processing TLV requests/repsonses
from the mfw and upper driver layers respectively. The implementation
reads the requested TLVs from the shared memory, requests the values
from upper layer drivers, populates this info (TLVs) shared memory and
notifies MFW about the TLV values.

Signed-off-by: Sudarsana Reddy Kalluru <Sudarsana.Kalluru@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/qlogic/qed/Makefile
drivers/net/ethernet/qlogic/qed/qed.h
drivers/net/ethernet/qlogic/qed/qed_main.c
drivers/net/ethernet/qlogic/qed/qed_mcp.h
drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c [new file with mode: 0644]
include/linux/qed/qed_if.h

index c70cf2ad81c0af027b8e251f0cadae320c88871b..a0acb94d65f08b7210d29fba60db290e7289972c 100644 (file)
@@ -3,7 +3,7 @@ obj-$(CONFIG_QED) := qed.o
 
 qed-y := qed_cxt.o qed_dev.o qed_hw.o qed_init_fw_funcs.o qed_init_ops.o \
         qed_int.o qed_main.o qed_mcp.o qed_sp_commands.o qed_spq.o qed_l2.o \
-        qed_selftest.o qed_dcbx.o qed_debug.o qed_ptp.o
+        qed_selftest.o qed_dcbx.o qed_debug.o qed_ptp.o qed_mng_tlv.o
 qed-$(CONFIG_QED_SRIOV) += qed_sriov.o qed_vf.o
 qed-$(CONFIG_QED_LL2) += qed_ll2.o
 qed-$(CONFIG_QED_RDMA) += qed_roce.o qed_rdma.o qed_iwarp.o
index adcff495466eb0e0f0e2c53a366137aaa8631196..dfdbe529e2bf0fbe3a6112f91e19c07d5596b274 100644 (file)
@@ -92,6 +92,8 @@ struct qed_eth_cb_ops;
 struct qed_dev_info;
 union qed_mcp_protocol_stats;
 enum qed_mcp_protocol_type;
+enum qed_mfw_tlv_type;
+union qed_mfw_tlv_data;
 
 /* helpers */
 #define QED_MFW_GET_FIELD(name, field) \
@@ -907,4 +909,7 @@ void qed_get_protocol_stats(struct qed_dev *cdev,
 int qed_slowpath_irq_req(struct qed_hwfn *hwfn);
 void qed_slowpath_irq_sync(struct qed_hwfn *p_hwfn);
 
+int qed_mfw_fill_tlv_data(struct qed_hwfn *hwfn,
+                         enum qed_mfw_tlv_type type,
+                         union qed_mfw_tlv_data *tlv_data);
 #endif /* _QED_H */
index 9feed3b79cd47585609b1e6d89b29fbd66358bcb..cbf0ea97fc8e940f79f84cc38164694bfec297da 100644 (file)
@@ -2088,3 +2088,9 @@ void qed_get_protocol_stats(struct qed_dev *cdev,
                return;
        }
 }
+
+int qed_mfw_fill_tlv_data(struct qed_hwfn *hwfn, enum qed_mfw_tlv_type type,
+                         union qed_mfw_tlv_data *tlv_buf)
+{
+       return -EINVAL;
+}
index 250579ba632bde6bebeb90eaca8a55f7e97bbd6a..591877fca67a7bedd35a63f09fc3a877ee75c84f 100644 (file)
@@ -213,6 +213,40 @@ enum qed_ov_wol {
        QED_OV_WOL_ENABLED
 };
 
+enum qed_mfw_tlv_type {
+       QED_MFW_TLV_GENERIC = 0x1,      /* Core driver TLVs */
+       QED_MFW_TLV_ETH = 0x2,          /* L2 driver TLVs */
+       QED_MFW_TLV_MAX = 0x4,
+};
+
+struct qed_mfw_tlv_generic {
+#define QED_MFW_TLV_FLAGS_SIZE 2
+       struct {
+               u8 ipv4_csum_offload;
+               u8 lso_supported;
+               bool b_set;
+       } flags;
+
+#define QED_MFW_TLV_MAC_COUNT 3
+       /* First entry for primary MAC, 2 secondary MACs possible */
+       u8 mac[QED_MFW_TLV_MAC_COUNT][6];
+       bool mac_set[QED_MFW_TLV_MAC_COUNT];
+
+       u64 rx_frames;
+       bool rx_frames_set;
+       u64 rx_bytes;
+       bool rx_bytes_set;
+       u64 tx_frames;
+       bool tx_frames_set;
+       u64 tx_bytes;
+       bool tx_bytes_set;
+};
+
+union qed_mfw_tlv_data {
+       struct qed_mfw_tlv_generic generic;
+       struct qed_mfw_tlv_eth eth;
+};
+
 /**
  * @brief - returns the link params of the hw function
  *
@@ -561,6 +595,17 @@ int qed_mcp_bist_nvm_get_image_att(struct qed_hwfn *p_hwfn,
                                   struct bist_nvm_image_att *p_image_att,
                                   u32 image_index);
 
+/**
+ * @brief - Processes the TLV request from MFW i.e., get the required TLV info
+ *          from the qed client and send it to the MFW.
+ *
+ * @param p_hwfn
+ * @param p_ptt
+ *
+ * @param return 0 upon success.
+ */
+int qed_mfw_process_tlv_req(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt);
+
 /* Using hwfn number (and not pf_num) is required since in CMT mode,
  * same pf_num may be used by two different hwfn
  * TODO - this shouldn't really be in .h file, but until all fields
@@ -621,6 +666,14 @@ struct qed_mcp_mb_params {
        u32                     mcp_param;
 };
 
+struct qed_drv_tlv_hdr {
+       u8 tlv_type;
+       u8 tlv_length;  /* In dwords - not including this header */
+       u8 tlv_reserved;
+#define QED_DRV_TLV_FLAGS_CHANGED 0x01
+       u8 tlv_flags;
+};
+
 /**
  * @brief Initialize the interface with the MCP
  *
diff --git a/drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c b/drivers/net/ethernet/qlogic/qed/qed_mng_tlv.c
new file mode 100644 (file)
index 0000000..d58a714
--- /dev/null
@@ -0,0 +1,378 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <linux/types.h>
+#include <asm/byteorder.h>
+#include <linux/bug.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/vmalloc.h>
+#include "qed.h"
+#include "qed_hw.h"
+#include "qed_mcp.h"
+#include "qed_reg_addr.h"
+
+#define TLV_TYPE(p)     (p[0])
+#define TLV_LENGTH(p)   (p[1])
+#define TLV_FLAGS(p)    (p[3])
+
+#define QED_TLV_DATA_MAX (14)
+struct qed_tlv_parsed_buf {
+       /* To be filled with the address to set in Value field */
+       void *p_val;
+
+       /* To be used internally in case the value has to be modified */
+       u8 data[QED_TLV_DATA_MAX];
+};
+
+static int qed_mfw_get_tlv_group(u8 tlv_type, u8 *tlv_group)
+{
+       switch (tlv_type) {
+       case DRV_TLV_FEATURE_FLAGS:
+       case DRV_TLV_LOCAL_ADMIN_ADDR:
+       case DRV_TLV_ADDITIONAL_MAC_ADDR_1:
+       case DRV_TLV_ADDITIONAL_MAC_ADDR_2:
+       case DRV_TLV_OS_DRIVER_STATES:
+       case DRV_TLV_PXE_BOOT_PROGRESS:
+       case DRV_TLV_RX_FRAMES_RECEIVED:
+       case DRV_TLV_RX_BYTES_RECEIVED:
+       case DRV_TLV_TX_FRAMES_SENT:
+       case DRV_TLV_TX_BYTES_SENT:
+       case DRV_TLV_NPIV_ENABLED:
+       case DRV_TLV_PCIE_BUS_RX_UTILIZATION:
+       case DRV_TLV_PCIE_BUS_TX_UTILIZATION:
+       case DRV_TLV_DEVICE_CPU_CORES_UTILIZATION:
+       case DRV_TLV_LAST_VALID_DCC_TLV_RECEIVED:
+       case DRV_TLV_NCSI_RX_BYTES_RECEIVED:
+       case DRV_TLV_NCSI_TX_BYTES_SENT:
+               *tlv_group |= QED_MFW_TLV_GENERIC;
+               break;
+       case DRV_TLV_LSO_MAX_OFFLOAD_SIZE:
+       case DRV_TLV_LSO_MIN_SEGMENT_COUNT:
+       case DRV_TLV_PROMISCUOUS_MODE:
+       case DRV_TLV_TX_DESCRIPTORS_QUEUE_SIZE:
+       case DRV_TLV_RX_DESCRIPTORS_QUEUE_SIZE:
+       case DRV_TLV_NUM_OF_NET_QUEUE_VMQ_CFG:
+       case DRV_TLV_NUM_OFFLOADED_CONNECTIONS_TCP_IPV4:
+       case DRV_TLV_NUM_OFFLOADED_CONNECTIONS_TCP_IPV6:
+       case DRV_TLV_TX_DESCRIPTOR_QUEUE_AVG_DEPTH:
+       case DRV_TLV_RX_DESCRIPTORS_QUEUE_AVG_DEPTH:
+       case DRV_TLV_IOV_OFFLOAD:
+       case DRV_TLV_TX_QUEUES_EMPTY:
+       case DRV_TLV_RX_QUEUES_EMPTY:
+       case DRV_TLV_TX_QUEUES_FULL:
+       case DRV_TLV_RX_QUEUES_FULL:
+               *tlv_group |= QED_MFW_TLV_ETH;
+               break;
+       default:
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+/* Returns size of the data buffer or, -1 in case TLV data is not available. */
+static int
+qed_mfw_get_gen_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
+                         struct qed_mfw_tlv_generic *p_drv_buf,
+                         struct qed_tlv_parsed_buf *p_buf)
+{
+       switch (p_tlv->tlv_type) {
+       case DRV_TLV_FEATURE_FLAGS:
+               if (p_drv_buf->flags.b_set) {
+                       memset(p_buf->data, 0, sizeof(u8) * QED_TLV_DATA_MAX);
+                       p_buf->data[0] = p_drv_buf->flags.ipv4_csum_offload ?
+                           1 : 0;
+                       p_buf->data[0] |= (p_drv_buf->flags.lso_supported ?
+                                          1 : 0) << 1;
+                       p_buf->p_val = p_buf->data;
+                       return QED_MFW_TLV_FLAGS_SIZE;
+               }
+               break;
+
+       case DRV_TLV_LOCAL_ADMIN_ADDR:
+       case DRV_TLV_ADDITIONAL_MAC_ADDR_1:
+       case DRV_TLV_ADDITIONAL_MAC_ADDR_2:
+               {
+                       int idx = p_tlv->tlv_type - DRV_TLV_LOCAL_ADMIN_ADDR;
+
+                       if (p_drv_buf->mac_set[idx]) {
+                               p_buf->p_val = p_drv_buf->mac[idx];
+                               return ETH_ALEN;
+                       }
+                       break;
+               }
+
+       case DRV_TLV_RX_FRAMES_RECEIVED:
+               if (p_drv_buf->rx_frames_set) {
+                       p_buf->p_val = &p_drv_buf->rx_frames;
+                       return sizeof(p_drv_buf->rx_frames);
+               }
+               break;
+       case DRV_TLV_RX_BYTES_RECEIVED:
+               if (p_drv_buf->rx_bytes_set) {
+                       p_buf->p_val = &p_drv_buf->rx_bytes;
+                       return sizeof(p_drv_buf->rx_bytes);
+               }
+               break;
+       case DRV_TLV_TX_FRAMES_SENT:
+               if (p_drv_buf->tx_frames_set) {
+                       p_buf->p_val = &p_drv_buf->tx_frames;
+                       return sizeof(p_drv_buf->tx_frames);
+               }
+               break;
+       case DRV_TLV_TX_BYTES_SENT:
+               if (p_drv_buf->tx_bytes_set) {
+                       p_buf->p_val = &p_drv_buf->tx_bytes;
+                       return sizeof(p_drv_buf->tx_bytes);
+               }
+               break;
+       default:
+               break;
+       }
+
+       return -1;
+}
+
+static int
+qed_mfw_get_eth_tlv_value(struct qed_drv_tlv_hdr *p_tlv,
+                         struct qed_mfw_tlv_eth *p_drv_buf,
+                         struct qed_tlv_parsed_buf *p_buf)
+{
+       switch (p_tlv->tlv_type) {
+       case DRV_TLV_LSO_MAX_OFFLOAD_SIZE:
+               if (p_drv_buf->lso_maxoff_size_set) {
+                       p_buf->p_val = &p_drv_buf->lso_maxoff_size;
+                       return sizeof(p_drv_buf->lso_maxoff_size);
+               }
+               break;
+       case DRV_TLV_LSO_MIN_SEGMENT_COUNT:
+               if (p_drv_buf->lso_minseg_size_set) {
+                       p_buf->p_val = &p_drv_buf->lso_minseg_size;
+                       return sizeof(p_drv_buf->lso_minseg_size);
+               }
+               break;
+       case DRV_TLV_PROMISCUOUS_MODE:
+               if (p_drv_buf->prom_mode_set) {
+                       p_buf->p_val = &p_drv_buf->prom_mode;
+                       return sizeof(p_drv_buf->prom_mode);
+               }
+               break;
+       case DRV_TLV_TX_DESCRIPTORS_QUEUE_SIZE:
+               if (p_drv_buf->tx_descr_size_set) {
+                       p_buf->p_val = &p_drv_buf->tx_descr_size;
+                       return sizeof(p_drv_buf->tx_descr_size);
+               }
+               break;
+       case DRV_TLV_RX_DESCRIPTORS_QUEUE_SIZE:
+               if (p_drv_buf->rx_descr_size_set) {
+                       p_buf->p_val = &p_drv_buf->rx_descr_size;
+                       return sizeof(p_drv_buf->rx_descr_size);
+               }
+               break;
+       case DRV_TLV_NUM_OF_NET_QUEUE_VMQ_CFG:
+               if (p_drv_buf->netq_count_set) {
+                       p_buf->p_val = &p_drv_buf->netq_count;
+                       return sizeof(p_drv_buf->netq_count);
+               }
+               break;
+       case DRV_TLV_NUM_OFFLOADED_CONNECTIONS_TCP_IPV4:
+               if (p_drv_buf->tcp4_offloads_set) {
+                       p_buf->p_val = &p_drv_buf->tcp4_offloads;
+                       return sizeof(p_drv_buf->tcp4_offloads);
+               }
+               break;
+       case DRV_TLV_NUM_OFFLOADED_CONNECTIONS_TCP_IPV6:
+               if (p_drv_buf->tcp6_offloads_set) {
+                       p_buf->p_val = &p_drv_buf->tcp6_offloads;
+                       return sizeof(p_drv_buf->tcp6_offloads);
+               }
+               break;
+       case DRV_TLV_TX_DESCRIPTOR_QUEUE_AVG_DEPTH:
+               if (p_drv_buf->tx_descr_qdepth_set) {
+                       p_buf->p_val = &p_drv_buf->tx_descr_qdepth;
+                       return sizeof(p_drv_buf->tx_descr_qdepth);
+               }
+               break;
+       case DRV_TLV_RX_DESCRIPTORS_QUEUE_AVG_DEPTH:
+               if (p_drv_buf->rx_descr_qdepth_set) {
+                       p_buf->p_val = &p_drv_buf->rx_descr_qdepth;
+                       return sizeof(p_drv_buf->rx_descr_qdepth);
+               }
+               break;
+       case DRV_TLV_IOV_OFFLOAD:
+               if (p_drv_buf->iov_offload_set) {
+                       p_buf->p_val = &p_drv_buf->iov_offload;
+                       return sizeof(p_drv_buf->iov_offload);
+               }
+               break;
+       case DRV_TLV_TX_QUEUES_EMPTY:
+               if (p_drv_buf->txqs_empty_set) {
+                       p_buf->p_val = &p_drv_buf->txqs_empty;
+                       return sizeof(p_drv_buf->txqs_empty);
+               }
+               break;
+       case DRV_TLV_RX_QUEUES_EMPTY:
+               if (p_drv_buf->rxqs_empty_set) {
+                       p_buf->p_val = &p_drv_buf->rxqs_empty;
+                       return sizeof(p_drv_buf->rxqs_empty);
+               }
+               break;
+       case DRV_TLV_TX_QUEUES_FULL:
+               if (p_drv_buf->num_txqs_full_set) {
+                       p_buf->p_val = &p_drv_buf->num_txqs_full;
+                       return sizeof(p_drv_buf->num_txqs_full);
+               }
+               break;
+       case DRV_TLV_RX_QUEUES_FULL:
+               if (p_drv_buf->num_rxqs_full_set) {
+                       p_buf->p_val = &p_drv_buf->num_rxqs_full;
+                       return sizeof(p_drv_buf->num_rxqs_full);
+               }
+               break;
+       default:
+               break;
+       }
+
+       return -1;
+}
+
+static int qed_mfw_update_tlvs(struct qed_hwfn *p_hwfn,
+                              u8 tlv_group, u8 *p_mfw_buf, u32 size)
+{
+       union qed_mfw_tlv_data *p_tlv_data;
+       struct qed_tlv_parsed_buf buffer;
+       struct qed_drv_tlv_hdr tlv;
+       int len = 0;
+       u32 offset;
+       u8 *p_tlv;
+
+       p_tlv_data = vzalloc(sizeof(*p_tlv_data));
+       if (!p_tlv_data)
+               return -ENOMEM;
+
+       if (qed_mfw_fill_tlv_data(p_hwfn, tlv_group, p_tlv_data)) {
+               vfree(p_tlv_data);
+               return -EINVAL;
+       }
+
+       memset(&tlv, 0, sizeof(tlv));
+       for (offset = 0; offset < size;
+            offset += sizeof(tlv) + sizeof(u32) * tlv.tlv_length) {
+               p_tlv = &p_mfw_buf[offset];
+               tlv.tlv_type = TLV_TYPE(p_tlv);
+               tlv.tlv_length = TLV_LENGTH(p_tlv);
+               tlv.tlv_flags = TLV_FLAGS(p_tlv);
+
+               DP_VERBOSE(p_hwfn, QED_MSG_SP,
+                          "Type %d length = %d flags = 0x%x\n", tlv.tlv_type,
+                          tlv.tlv_length, tlv.tlv_flags);
+
+               if (tlv_group == QED_MFW_TLV_GENERIC)
+                       len = qed_mfw_get_gen_tlv_value(&tlv,
+                                                       &p_tlv_data->generic,
+                                                       &buffer);
+               else if (tlv_group == QED_MFW_TLV_ETH)
+                       len = qed_mfw_get_eth_tlv_value(&tlv,
+                                                       &p_tlv_data->eth,
+                                                       &buffer);
+
+               if (len > 0) {
+                       WARN(len > 4 * tlv.tlv_length,
+                            "Incorrect MFW TLV length %d, it shouldn't be greater than %d\n",
+                            len, 4 * tlv.tlv_length);
+                       len = min_t(int, len, 4 * tlv.tlv_length);
+                       tlv.tlv_flags |= QED_DRV_TLV_FLAGS_CHANGED;
+                       TLV_FLAGS(p_tlv) = tlv.tlv_flags;
+                       memcpy(p_mfw_buf + offset + sizeof(tlv),
+                              buffer.p_val, len);
+               }
+       }
+
+       vfree(p_tlv_data);
+
+       return 0;
+}
+
+int qed_mfw_process_tlv_req(struct qed_hwfn *p_hwfn, struct qed_ptt *p_ptt)
+{
+       u32 addr, size, offset, resp, param, val, global_offsize, global_addr;
+       u8 tlv_group = 0, id, *p_mfw_buf = NULL, *p_temp;
+       struct qed_drv_tlv_hdr tlv;
+       int rc;
+
+       addr = SECTION_OFFSIZE_ADDR(p_hwfn->mcp_info->public_base,
+                                   PUBLIC_GLOBAL);
+       global_offsize = qed_rd(p_hwfn, p_ptt, addr);
+       global_addr = SECTION_ADDR(global_offsize, 0);
+       addr = global_addr + offsetof(struct public_global, data_ptr);
+       addr = qed_rd(p_hwfn, p_ptt, addr);
+       size = qed_rd(p_hwfn, p_ptt, global_addr +
+                     offsetof(struct public_global, data_size));
+
+       if (!size) {
+               DP_NOTICE(p_hwfn, "Invalid TLV req size = %d\n", size);
+               goto drv_done;
+       }
+
+       p_mfw_buf = vzalloc(size);
+       if (!p_mfw_buf) {
+               DP_NOTICE(p_hwfn, "Failed allocate memory for p_mfw_buf\n");
+               goto drv_done;
+       }
+
+       /* Read the TLV request to local buffer. MFW represents the TLV in
+        * little endian format and mcp returns it bigendian format. Hence
+        * driver need to convert data to little endian first and then do the
+        * memcpy (casting) to preserve the MFW TLV format in the driver buffer.
+        *
+        */
+       for (offset = 0; offset < size; offset += sizeof(u32)) {
+               val = qed_rd(p_hwfn, p_ptt, addr + offset);
+               val = be32_to_cpu(val);
+               memcpy(&p_mfw_buf[offset], &val, sizeof(u32));
+       }
+
+       /* Parse the headers to enumerate the requested TLV groups */
+       for (offset = 0; offset < size;
+            offset += sizeof(tlv) + sizeof(u32) * tlv.tlv_length) {
+               p_temp = &p_mfw_buf[offset];
+               tlv.tlv_type = TLV_TYPE(p_temp);
+               tlv.tlv_length = TLV_LENGTH(p_temp);
+               if (qed_mfw_get_tlv_group(tlv.tlv_type, &tlv_group))
+                       DP_VERBOSE(p_hwfn, NETIF_MSG_DRV,
+                                  "Un recognized TLV %d\n", tlv.tlv_type);
+       }
+
+       /* Sanitize the TLV groups according to personality */
+       if ((tlv_group & QED_MFW_TLV_ETH) && !QED_IS_L2_PERSONALITY(p_hwfn)) {
+               DP_VERBOSE(p_hwfn, QED_MSG_SP,
+                          "Skipping L2 TLVs for non-L2 function\n");
+               tlv_group &= ~QED_MFW_TLV_ETH;
+       }
+
+       /* Update the TLV values in the local buffer */
+       for (id = QED_MFW_TLV_GENERIC; id < QED_MFW_TLV_MAX; id <<= 1) {
+               if (tlv_group & id)
+                       if (qed_mfw_update_tlvs(p_hwfn, id, p_mfw_buf, size))
+                               goto drv_done;
+       }
+
+       /* Write the TLV data to shared memory. The stream of 4 bytes first need
+        * to be mem-copied to u32 element to make it as LSB format. And then
+        * converted to big endian as required by mcp-write.
+        */
+       for (offset = 0; offset < size; offset += sizeof(u32)) {
+               memcpy(&val, &p_mfw_buf[offset], sizeof(u32));
+               val = cpu_to_be32(val);
+               qed_wr(p_hwfn, p_ptt, addr + offset, val);
+       }
+
+drv_done:
+       rc = qed_mcp_cmd(p_hwfn, p_ptt, DRV_MSG_CODE_GET_TLV_DONE, 0, &resp,
+                        &param);
+
+       vfree(p_mfw_buf);
+
+       return rc;
+}
index 907976fd56f7dedb7795b0a48adef5d184977c7e..8e4fad427ac2cc522f7abb09665919de308dbe6b 100644 (file)
@@ -182,6 +182,43 @@ enum qed_led_mode {
        QED_LED_MODE_RESTORE
 };
 
+struct qed_mfw_tlv_eth {
+       u16 lso_maxoff_size;
+       bool lso_maxoff_size_set;
+       u16 lso_minseg_size;
+       bool lso_minseg_size_set;
+       u8 prom_mode;
+       bool prom_mode_set;
+       u16 tx_descr_size;
+       bool tx_descr_size_set;
+       u16 rx_descr_size;
+       bool rx_descr_size_set;
+       u16 netq_count;
+       bool netq_count_set;
+       u32 tcp4_offloads;
+       bool tcp4_offloads_set;
+       u32 tcp6_offloads;
+       bool tcp6_offloads_set;
+       u16 tx_descr_qdepth;
+       bool tx_descr_qdepth_set;
+       u16 rx_descr_qdepth;
+       bool rx_descr_qdepth_set;
+       u8 iov_offload;
+#define QED_MFW_TLV_IOV_OFFLOAD_NONE            (0)
+#define QED_MFW_TLV_IOV_OFFLOAD_MULTIQUEUE      (1)
+#define QED_MFW_TLV_IOV_OFFLOAD_VEB             (2)
+#define QED_MFW_TLV_IOV_OFFLOAD_VEPA            (3)
+       bool iov_offload_set;
+       u8 txqs_empty;
+       bool txqs_empty_set;
+       u8 rxqs_empty;
+       bool rxqs_empty_set;
+       u8 num_txqs_full;
+       bool num_txqs_full_set;
+       u8 num_rxqs_full;
+       bool num_rxqs_full_set;
+};
+
 #define DIRECT_REG_WR(reg_addr, val) writel((u32)val, \
                                            (void __iomem *)(reg_addr))