From: Xiaojun Liu <xiaojun.liu@silicom.co.il>
To: xiao.w.wang@intel.com, qi.z.zhang@intel.com,
ngai-mint.kwan@intel.com, jacob.e.keller@intel.com
Cc: dev@dpdk.org, Xiaojun Liu <xiaojun.liu@silicom.co.il>
Subject: [dpdk-dev] [PATCH v4 4/5] net/fm10k: add flow interface and switch management
Date: Thu, 9 Apr 2020 14:26:27 +0800 [thread overview]
Message-ID: <1586413588-5759-5-git-send-email-xiaojun.liu@silicom.co.il> (raw)
In-Reply-To: <1586413588-5759-1-git-send-email-xiaojun.liu@silicom.co.il>
Add flow interface to support offload flow into HW.
It supports parse vlan and parse mpls, all these
data will be transferred to ffu data.
Add switch management, includes initialization,
port mapping, epl port link, LED controller, interrupt handler.
It create 3 threads. One for interrupt handler, one for
LED controller, one for statistics.
Signed-off-by: Xiaojun Liu <xiaojun.liu@silicom.co.il>
---
drivers/net/fm10k/Makefile | 2 +
drivers/net/fm10k/base/fm10k_type.h | 1 +
drivers/net/fm10k/switch/fm10k_flow.c | 864 +++++++++++
drivers/net/fm10k/switch/fm10k_flow.h | 26 +
drivers/net/fm10k/switch/fm10k_stats.c | 129 ++
drivers/net/fm10k/switch/fm10k_switch.c | 2362 +++++++++++++++++++++++++++++++
6 files changed, 3384 insertions(+)
create mode 100644 drivers/net/fm10k/switch/fm10k_flow.c
create mode 100644 drivers/net/fm10k/switch/fm10k_flow.h
create mode 100644 drivers/net/fm10k/switch/fm10k_switch.c
diff --git a/drivers/net/fm10k/Makefile b/drivers/net/fm10k/Makefile
index 68a72d5..5a8d227 100644
--- a/drivers/net/fm10k/Makefile
+++ b/drivers/net/fm10k/Makefile
@@ -88,6 +88,8 @@ SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_spico_code.c
SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_config.c
SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_ffu.c
SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_stats.c
+SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_flow.c
+SRCS-$(CONFIG_RTE_LIBRTE_FM10K_PMD) += fm10k_switch.c
endif
ifeq ($(CONFIG_RTE_ARCH_X86), y)
SRCS-$(CONFIG_RTE_LIBRTE_FM10K_INC_VECTOR) += fm10k_rxtx_vec.c
diff --git a/drivers/net/fm10k/base/fm10k_type.h b/drivers/net/fm10k/base/fm10k_type.h
index 84781ba..810e5ee 100644
--- a/drivers/net/fm10k/base/fm10k_type.h
+++ b/drivers/net/fm10k/base/fm10k_type.h
@@ -717,6 +717,7 @@ struct fm10k_hw {
u16 subsystem_vendor_id;
u8 revision_id;
u32 flags;
+ bool switch_ready;
#define FM10K_HW_FLAG_CLOCK_OWNER BIT(0)
};
diff --git a/drivers/net/fm10k/switch/fm10k_flow.c b/drivers/net/fm10k/switch/fm10k_flow.c
new file mode 100644
index 0000000..d9ba2b4
--- /dev/null
+++ b/drivers/net/fm10k/switch/fm10k_flow.c
@@ -0,0 +1,864 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2019 Silicom Ltd. Connectivity Solutions
+ */
+
+#include <sys/queue.h>
+#include <stdio.h>
+#include <errno.h>
+#include <stdint.h>
+#include <string.h>
+#include <unistd.h>
+#include <stdarg.h>
+
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_log.h>
+#include <rte_malloc.h>
+#include <rte_eth_ctrl.h>
+#include <rte_tailq.h>
+#include <rte_flow_driver.h>
+
+#include "fm10k_flow.h"
+#include "fm10k_switch.h"
+#include "fm10k_ffu.h"
+#include "fm10k_config.h"
+
+static int fm10k_flow_validate(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error);
+static struct rte_flow *fm10k_flow_create(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error);
+static int fm10k_flow_destroy(struct rte_eth_dev *dev,
+ struct rte_flow *flow,
+ struct rte_flow_error *error);
+static int fm10k_flow_flush(struct rte_eth_dev *dev,
+ struct rte_flow_error *error);
+static int fm10k_flow_parse_attr(const struct rte_flow_attr *attr,
+ struct rte_flow_error *error);
+
+const struct rte_flow_ops fm10k_flow_ops = {
+ .validate = fm10k_flow_validate,
+ .create = fm10k_flow_create,
+ .destroy = fm10k_flow_destroy,
+ .flush = fm10k_flow_flush,
+};
+
+union fm10k_filter_t cons_filter;
+enum rte_filter_type fm10k_cons_filter_type = RTE_ETH_FILTER_NONE;
+
+/**
+ * MPLS filter configuration.
+ */
+enum fm10k_mpls_type {
+ FM10K_MPLS_TYPE_UNI,
+ FM10K_MPLS_TYPE_MULTI,
+};
+
+enum fm10k_mpls_action {
+ FM10K_MPLS_ACTION_DROP,
+ FM10K_MPLS_ACTION_QUEUE,
+};
+
+struct fm10k_mpls_filter_conf {
+ enum fm10k_mpls_type mpls_type; /**< mandatory for MPLS */
+ uint32_t mpls_header; /**< MPLS header */
+ uint32_t mpls_header_mask; /**< MPLS header mask */
+ enum fm10k_mpls_action mpls_action;
+ uint16_t queue;
+ uint8_t ffu_id;
+ uint8_t ffu_prio;
+};
+
+/**
+ * VLAN filter configuration.
+ */
+struct fm10k_vlan_filter_conf {
+ int ffu_id;
+ uint8_t ffu_prio;
+ uint8_t is_ingress;
+ uint8_t port;
+ uint8_t in_ext_port;
+ uint8_t out_ext_port;
+ uint16_t in_vlan;
+ uint16_t out_vlan;
+};
+
+union fm10k_filter_t {
+ struct fm10k_mpls_filter_conf mpls_filter;
+ struct fm10k_vlan_filter_conf vlan_filter;
+};
+
+typedef int (*parse_filter_t)(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error,
+ union fm10k_filter_t *filter);
+
+struct fm10k_valid_pattern {
+ enum rte_flow_item_type *items;
+ parse_filter_t parse_filter;
+};
+
+static enum rte_flow_item_type pattern_mpls_1[] = {
+ RTE_FLOW_ITEM_TYPE_ETH,
+ RTE_FLOW_ITEM_TYPE_MPLS,
+ RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_vlan_1[] = {
+ RTE_FLOW_ITEM_TYPE_VLAN,
+ RTE_FLOW_ITEM_TYPE_PHY_PORT,
+ RTE_FLOW_ITEM_TYPE_END,
+};
+
+static enum rte_flow_item_type pattern_vlan_2[] = {
+ RTE_FLOW_ITEM_TYPE_VLAN,
+ RTE_FLOW_ITEM_TYPE_PHY_PORT,
+ RTE_FLOW_ITEM_TYPE_PHY_PORT,
+ RTE_FLOW_ITEM_TYPE_END,
+};
+
+static int fm10k_flow_parse_mpls_filter(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error,
+ union fm10k_filter_t *filter);
+static int
+fm10k_flow_parse_vlan_filter(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error,
+ union fm10k_filter_t *filter);
+
+static struct fm10k_valid_pattern fm10k_supported_patterns[] = {
+ /* MPLS */
+ { pattern_mpls_1, fm10k_flow_parse_mpls_filter },
+ /* VLAN */
+ { pattern_vlan_1, fm10k_flow_parse_vlan_filter },
+ /* VLAN */
+ { pattern_vlan_2, fm10k_flow_parse_vlan_filter },
+};
+
+static const struct rte_flow_action *
+fm10k_next_item_of_action(const struct rte_flow_action *actions,
+ uint32_t *index)
+{
+ static const struct rte_flow_action *act;
+
+ act = actions + *index;
+ while (act->type == RTE_FLOW_ACTION_TYPE_VOID) {
+ (*index)++;
+ act = actions + *index;
+ }
+ return act;
+}
+
+/* Find the first VOID or non-VOID item pointer */
+static const struct rte_flow_item *
+fm10k_find_first_item(const struct rte_flow_item *item, bool is_void)
+{
+ bool is_find;
+
+ while (item->type != RTE_FLOW_ITEM_TYPE_END) {
+ if (is_void)
+ is_find = item->type == RTE_FLOW_ITEM_TYPE_VOID;
+ else
+ is_find = item->type != RTE_FLOW_ITEM_TYPE_VOID;
+ if (is_find)
+ break;
+ item++;
+ }
+ return item;
+}
+
+/* Skip all VOID items of the pattern */
+static void
+fm10k_pattern_skip_void_item(struct rte_flow_item *items,
+ const struct rte_flow_item *pattern)
+{
+ uint32_t cpy_count = 0;
+ const struct rte_flow_item *pb = pattern, *pe = pattern;
+
+ for (;;) {
+ /* Find a non-void item first */
+ pb = fm10k_find_first_item(pb, false);
+ if (pb->type == RTE_FLOW_ITEM_TYPE_END) {
+ pe = pb;
+ break;
+ }
+
+ /* Find a void item */
+ pe = fm10k_find_first_item(pb + 1, true);
+
+ cpy_count = pe - pb;
+ rte_memcpy(items, pb, sizeof(struct rte_flow_item) * cpy_count);
+
+ items += cpy_count;
+
+ if (pe->type == RTE_FLOW_ITEM_TYPE_END) {
+ pb = pe;
+ break;
+ }
+
+ pb = pe + 1;
+ }
+ /* Copy the END item. */
+ rte_memcpy(items, pe, sizeof(struct rte_flow_item));
+}
+
+/* Check if the pattern matches a supported item type array */
+static bool
+fm10k_match_pattern(enum rte_flow_item_type *item_array,
+ struct rte_flow_item *pattern)
+{
+ struct rte_flow_item *item = pattern;
+
+ while ((*item_array == item->type) &&
+ (*item_array != RTE_FLOW_ITEM_TYPE_END)) {
+ item_array++;
+ item++;
+ }
+
+ return (*item_array == RTE_FLOW_ITEM_TYPE_END &&
+ item->type == RTE_FLOW_ITEM_TYPE_END);
+}
+
+/* Find if there's parse filter function matched */
+static parse_filter_t
+fm10k_find_parse_filter_func(struct rte_flow_item *pattern, uint32_t *idx)
+{
+ parse_filter_t parse_filter = NULL;
+ uint8_t i = *idx;
+
+ for (; i < RTE_DIM(fm10k_supported_patterns); i++) {
+ if (fm10k_match_pattern(fm10k_supported_patterns[i].items,
+ pattern)) {
+ parse_filter = fm10k_supported_patterns[i].parse_filter;
+ break;
+ }
+ }
+
+ *idx = ++i;
+
+ return parse_filter;
+}
+
+/* Parse attributes */
+static int
+fm10k_flow_parse_attr(const struct rte_flow_attr *attr,
+ struct rte_flow_error *error)
+{
+ /* Not supported */
+ if (attr->group) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ATTR_GROUP,
+ attr, "Not support group.");
+ return -rte_errno;
+ }
+
+ return 0;
+}
+
+/*
+ * MPLS
+ */
+/* 1. Last in item should be NULL as range is not supported.
+ * 2. Supported filter types: MPLS label.
+ * 3. Mask of fields which need to be matched should be
+ * filled with 1.
+ * 4. Mask of fields which needn't to be matched should be
+ * filled with 0.
+ */
+static int
+fm10k_flow_parse_mpls_pattern(__rte_unused struct rte_eth_dev *dev,
+ const struct rte_flow_item *pattern,
+ struct rte_flow_error *error,
+ struct fm10k_mpls_filter_conf *filter)
+{
+ const struct rte_flow_item *item = pattern;
+ const struct rte_flow_item_mpls *mpls_spec;
+ const struct rte_flow_item_mpls *mpls_mask;
+ const struct rte_flow_item_eth *eth_spec;
+ enum rte_flow_item_type item_type;
+ const uint8_t label_mask[3] = {0xFF, 0xFF, 0xF0};
+ uint32_t label_be = 0;
+ uint32_t be_mask = 0;
+
+ for (; item->type != RTE_FLOW_ITEM_TYPE_END; item++) {
+ if (item->last) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Not support range");
+ return -rte_errno;
+ }
+ item_type = item->type;
+ switch (item_type) {
+ case RTE_FLOW_ITEM_TYPE_ETH:
+ if (!item->spec || item->mask) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Invalid ETH item");
+ return -rte_errno;
+ }
+ eth_spec =
+ (const struct rte_flow_item_eth *)item->spec;
+
+ if (rte_be_to_cpu_16(eth_spec->type) == 0x8847) {
+ filter->mpls_type = FM10K_MPLS_TYPE_UNI;
+ } else if (rte_be_to_cpu_16(eth_spec->type) == 0x8848) {
+ filter->mpls_type = FM10K_MPLS_TYPE_MULTI;
+ } else {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item, "Invalid ETH item");
+ return -rte_errno;
+ }
+ break;
+ case RTE_FLOW_ITEM_TYPE_MPLS:
+ mpls_spec =
+ (const struct rte_flow_item_mpls *)item->spec;
+ mpls_mask =
+ (const struct rte_flow_item_mpls *)item->mask;
+
+ if (!mpls_spec || !mpls_mask) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Invalid MPLS item");
+ return -rte_errno;
+ }
+
+ if (memcmp(mpls_mask->label_tc_s, label_mask, 3)) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Invalid MPLS label mask");
+ return -rte_errno;
+ }
+ rte_memcpy(((uint8_t *)&label_be + 1),
+ mpls_spec->label_tc_s, 3);
+ rte_memcpy(((uint8_t *)&be_mask + 1),
+ mpls_mask->label_tc_s, 3);
+ filter->mpls_header =
+ rte_be_to_cpu_32(label_be) >> 4;
+ filter->mpls_header_mask =
+ rte_be_to_cpu_32(be_mask) >> 4;
+
+ fm10k_cons_filter_type = RTE_ETH_FILTER_TUNNEL;
+ break;
+ default:
+ break;
+ }
+ }
+
+ return 0;
+}
+
+/* MPLS action only supports QUEUE or DROP. */
+static int
+fm10k_flow_parse_mpls_action(const struct rte_flow_action *actions,
+ struct rte_flow_error *error,
+ struct fm10k_mpls_filter_conf *filter)
+{
+ const struct rte_flow_action *act;
+ const struct rte_flow_action_queue *act_q;
+ uint32_t index = 0;
+
+ /* Check if the first non-void action is QUEUE or DROP. */
+ act = fm10k_next_item_of_action(actions, &index);
+ if (act->type != RTE_FLOW_ACTION_TYPE_QUEUE &&
+ act->type != RTE_FLOW_ACTION_TYPE_DROP) {
+ rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
+ act, "Not supported action.");
+ return -rte_errno;
+ }
+
+ if (act->type == RTE_FLOW_ACTION_TYPE_QUEUE) {
+ act_q = (const struct rte_flow_action_queue *)act->conf;
+ filter->mpls_action = FM10K_MPLS_ACTION_QUEUE;
+ filter->queue = act_q->index;
+ } else {
+ filter->mpls_action = FM10K_MPLS_ACTION_DROP;
+ }
+
+ /* Check if the next non-void item is END */
+ index++;
+ act = fm10k_next_item_of_action(actions, &index);
+ if (act->type != RTE_FLOW_ACTION_TYPE_END) {
+ rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
+ act, "Not supported action.");
+ return -rte_errno;
+ }
+
+ return 0;
+}
+
+static int
+fm10k_flow_parse_mpls_filter(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error,
+ union fm10k_filter_t *filter)
+{
+ struct fm10k_mpls_filter_conf *mpls_filter =
+ &filter->mpls_filter;
+ int ret;
+
+ ret = fm10k_flow_parse_mpls_pattern(dev, pattern,
+ error, mpls_filter);
+ if (ret)
+ return ret;
+
+ ret = fm10k_flow_parse_mpls_action(actions, error, mpls_filter);
+ if (ret)
+ return ret;
+
+ ret = fm10k_flow_parse_attr(attr, error);
+ return ret;
+}
+
+/*
+ * VLAN
+ */
+static int
+fm10k_flow_parse_vlan_pattern(__rte_unused struct rte_eth_dev *dev,
+ const struct rte_flow_item *pattern,
+ struct rte_flow_error *error,
+ struct fm10k_vlan_filter_conf *filter)
+{
+ const struct rte_flow_item *item = pattern;
+ const struct rte_flow_item_vlan *vlan_spec;
+ const struct rte_flow_item_phy_port *pp_spec;
+ enum rte_flow_item_type item_type;
+
+ PMD_INIT_LOG(DEBUG, "Parse vlan pattern ffu id %d", filter->ffu_id);
+
+ for (; item->type != RTE_FLOW_ITEM_TYPE_END; item++) {
+ if (item->last) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Not support range");
+ return -rte_errno;
+ }
+ item_type = item->type;
+ switch (item_type) {
+ case RTE_FLOW_ITEM_TYPE_VLAN:
+ vlan_spec =
+ (const struct rte_flow_item_vlan *)item->spec;
+
+ if (!vlan_spec) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Invalid VLAN item");
+ return -rte_errno;
+ }
+ break;
+
+ case RTE_FLOW_ITEM_TYPE_PHY_PORT:
+ pp_spec =
+ (const struct rte_flow_item_phy_port *)item->spec;
+ if (!pp_spec) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ item,
+ "Invalid PHY PORT item");
+ return -rte_errno;
+ }
+ break;
+
+ default:
+ break;
+ }
+ }
+
+ return 0;
+}
+
+static int
+fm10k_flow_parse_vlan_action(struct rte_eth_dev *dev,
+ const struct rte_flow_action *actions,
+ struct rte_flow_error *error,
+ struct fm10k_vlan_filter_conf *filter)
+{
+ const struct rte_flow_action *act;
+ uint32_t index = 0;
+
+ PMD_INIT_LOG(DEBUG, "Parse vlan action name %s ffu id %d",
+ dev->device->name, filter->ffu_id);
+
+ /* Check if the first non-void action is QUEUE or DROP. */
+ act = fm10k_next_item_of_action(actions, &index);
+ if (act->type != RTE_FLOW_ACTION_TYPE_MARK) {
+ rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
+ act, "Not supported action.");
+ return -rte_errno;
+ }
+
+ index++;
+ act = fm10k_next_item_of_action(actions, &index);
+ if (act->type != RTE_FLOW_ACTION_TYPE_MARK &&
+ act->type != RTE_FLOW_ACTION_TYPE_END) {
+ rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
+ act, "Not supported action.");
+ return -rte_errno;
+ }
+ if (act->type == RTE_FLOW_ACTION_TYPE_END)
+ return 0;
+
+ index++;
+ /* Check if the next non-void item is END */
+ act = fm10k_next_item_of_action(actions, &index);
+ if (act->type != RTE_FLOW_ACTION_TYPE_END) {
+ rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ACTION,
+ act, "Not supported action.");
+ return -rte_errno;
+ }
+
+ return 0;
+}
+
+static int
+fm10k_flow_parse_vlan_filter(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error,
+ union fm10k_filter_t *filter)
+{
+ int ret;
+ struct fm10k_vlan_filter_conf *vlan_filter =
+ &filter->vlan_filter;
+
+ ret = fm10k_flow_parse_vlan_pattern(dev, pattern,
+ error, vlan_filter);
+ if (ret)
+ return ret;
+
+ ret = fm10k_flow_parse_vlan_action(dev, actions, error, vlan_filter);
+ if (ret)
+ return ret;
+
+ if (attr->ingress)
+ vlan_filter->is_ingress = 1;
+ else if (attr->egress)
+ vlan_filter->is_ingress = 0;
+ vlan_filter->ffu_prio = attr->priority;
+
+ ret = fm10k_flow_parse_attr(attr, error);
+ return ret;
+}
+
+/*
+ *
+ */
+static int
+fm10k_flow_validate(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error)
+{
+ struct rte_flow_item *items; /* internal pattern w/o VOID items */
+ parse_filter_t parse_filter;
+ uint32_t item_num = 0; /* non-void item number of pattern*/
+ uint32_t i = 0;
+ bool flag = false;
+ int ret = -1;
+
+ if (!pattern) {
+ rte_flow_error_set(error, EINVAL, RTE_FLOW_ERROR_TYPE_ITEM_NUM,
+ NULL, "NULL pattern.");
+ return -rte_errno;
+ }
+
+ if (!actions) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ACTION_NUM,
+ NULL, "NULL action.");
+ return -rte_errno;
+ }
+
+ if (!attr) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ATTR,
+ NULL, "NULL attribute.");
+ return -rte_errno;
+ }
+
+ /* Get the non-void item number of pattern */
+ while ((pattern + i)->type != RTE_FLOW_ITEM_TYPE_END) {
+ if ((pattern + i)->type != RTE_FLOW_ITEM_TYPE_VOID)
+ item_num++;
+ i++;
+ }
+ item_num++;
+
+ items = rte_zmalloc("fm10k_pattern",
+ item_num * sizeof(struct rte_flow_item), 0);
+ if (!items) {
+ rte_flow_error_set(error, ENOMEM, RTE_FLOW_ERROR_TYPE_ITEM_NUM,
+ NULL, "No memory for PMD internal items.");
+ return -ENOMEM;
+ }
+
+ fm10k_pattern_skip_void_item(items, pattern);
+
+ i = 0;
+ do {
+ parse_filter = fm10k_find_parse_filter_func(items, &i);
+ if (!parse_filter && !flag) {
+ rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ pattern, "Unsupported pattern");
+ rte_free(items);
+ return -rte_errno;
+ }
+ if (parse_filter)
+ ret = parse_filter(dev, attr, items, actions,
+ error, &cons_filter);
+ flag = true;
+ } while ((ret < 0) && (i < RTE_DIM(fm10k_supported_patterns)));
+
+ rte_free(items);
+
+ return ret;
+}
+
+static struct fm10k_cfg_flow *
+fm10k_flow_cfg_transfer(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error)
+{
+ int i;
+ u8 port_id;
+ int set_port_num = 0, set_vlan_num = 0;
+ u16 fw_port_id = 0, bp_port_id = 0;
+ u16 filter_vlan_id = 0, fw_vlan_id = 0, bp_vlan_id = 0;
+ struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ struct fm10k_cfg_flow *cf;
+
+ cf = rte_zmalloc("fm10k_rule", sizeof(struct fm10k_cfg_flow), 0);
+ if (!cf) {
+ rte_flow_error_set(error, ENOMEM,
+ RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Failed to allocate memory");
+ return NULL;
+ }
+ memset(cf, 0, sizeof(struct fm10k_cfg_flow));
+
+ port_id = fm10k_switch_dpdk_port_no_get(hw);
+ for (i = 0; i < 4; i++) {
+ if (pattern[i].type == RTE_FLOW_ITEM_TYPE_VLAN) {
+ filter_vlan_id =
+ rte_be_to_cpu_16
+ (((const struct rte_flow_item_vlan *)
+ pattern[i].spec)->tci);
+ } else if (pattern[i].type == RTE_FLOW_ITEM_TYPE_PHY_PORT) {
+ if (set_port_num)
+ bp_port_id =
+ ((const struct rte_flow_item_phy_port *)
+ pattern[i].spec)->index;
+ else
+ fw_port_id =
+ ((const struct rte_flow_item_phy_port *)
+ pattern[i].spec)->index;
+ set_port_num++;
+ } else if (pattern[i].type == RTE_FLOW_ITEM_TYPE_END) {
+ break;
+ }
+ }
+
+ for (i = 0; i < 3; i++) {
+ if (actions[i].type == RTE_FLOW_ACTION_TYPE_MARK) {
+ if (set_vlan_num)
+ bp_vlan_id =
+ ((const struct rte_flow_action_mark *)
+ actions[i].conf)->id;
+ else
+ fw_vlan_id =
+ ((const struct rte_flow_action_mark *)
+ actions[i].conf)->id;
+ set_vlan_num++;
+ } else if (actions[i].type == RTE_FLOW_ACTION_TYPE_END) {
+ break;
+ }
+ }
+
+ if (attr->ingress && !attr->egress) {
+ /* this port is DPDK port and it is destination port */
+ cf->src_port.port_type = FM10K_CONFIG_FLOW_EXT_PORT;
+ cf->src_port.port_no = fw_port_id;
+ cf->src_port.vlan_id = filter_vlan_id;
+ cf->fw_port[0].port_type = FM10K_CONFIG_FLOW_DPDK_PORT;
+ cf->fw_port[0].port_no = port_id;
+ cf->fw_port[0].vlan_id = fw_vlan_id;
+ } else if (!attr->ingress && attr->egress) {
+ /* this port is DPDK port and it is source port */
+ cf->src_port.port_type = FM10K_CONFIG_FLOW_DPDK_PORT;
+ cf->src_port.port_no = port_id;
+ cf->src_port.vlan_id = filter_vlan_id;
+ cf->fw_port[0].port_type = FM10K_CONFIG_FLOW_EXT_PORT;
+ cf->fw_port[0].port_no = fw_port_id;
+ cf->fw_port[0].vlan_id = fw_vlan_id;
+ } else if (!attr->ingress && !attr->egress) {
+ /* two ports are external port */
+ cf->src_port.port_type = FM10K_CONFIG_FLOW_EXT_PORT;
+ cf->src_port.port_no = port_id;
+ cf->src_port.vlan_id = filter_vlan_id;
+ cf->fw_port[0].port_type = FM10K_CONFIG_FLOW_EXT_PORT;
+ cf->fw_port[0].port_no = fw_port_id;
+ cf->fw_port[0].vlan_id = fw_vlan_id;
+ } else {
+ /* two ports are DPDK port */
+ cf->src_port.port_type = FM10K_CONFIG_FLOW_DPDK_PORT;
+ cf->src_port.port_no = port_id;
+ cf->src_port.vlan_id = filter_vlan_id;
+ cf->fw_port[0].port_type = FM10K_CONFIG_FLOW_DPDK_PORT;
+ cf->fw_port[0].port_no = fw_port_id;
+ cf->fw_port[0].vlan_id = fw_vlan_id;
+ }
+
+ if (set_port_num == 2 && set_vlan_num == 2) {
+ cf->fw_port[1].port_type = FM10K_CONFIG_FLOW_EXT_PORT;
+ cf->fw_port[1].port_no = bp_port_id;
+ cf->fw_port[1].vlan_id = bp_vlan_id;
+ }
+
+ return cf;
+}
+
+static struct rte_flow *
+fm10k_flow_create(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr,
+ const struct rte_flow_item pattern[],
+ const struct rte_flow_action actions[],
+ struct rte_flow_error *error)
+{
+ struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ struct rte_flow *flow;
+ struct fm10k_switch *sw = fm10k_switch_get();
+ struct fm10k_cfg_flow *cf;
+ int ret;
+
+ flow = rte_zmalloc("fm10k_flow", sizeof(struct rte_flow), 0);
+ if (!flow) {
+ rte_flow_error_set(error, ENOMEM,
+ RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Failed to allocate memory");
+ return flow;
+ }
+
+ ret = fm10k_flow_validate(dev, attr, pattern, actions, error);
+ if (ret < 0)
+ return NULL;
+
+ cf = fm10k_flow_cfg_transfer(dev, attr, pattern, actions, error);
+ if (!cf)
+ goto free_flow;
+
+ flow->rule = cf;
+ fm10k_ffu_flow_enable(sw, cf);
+ fm10k_config_flow_list_add_tail(fm10k_config_flowset_current_get(), cf);
+
+ TAILQ_INSERT_TAIL((struct fm10k_flow_list *)
+ fm10k_switch_dpdk_port_flow_list_get(hw), flow, node);
+ return flow;
+
+free_flow:
+ rte_flow_error_set(error, -ret,
+ RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Failed to create flow.");
+ rte_free(flow);
+ return NULL;
+}
+
+static int
+fm10k_flow_destroy(struct rte_eth_dev *dev,
+ struct rte_flow *flow,
+ struct rte_flow_error *error)
+{
+ struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ int ret = 0;
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (flow->rule) {
+ fm10k_config_flow_list_delete
+ ((struct fm10k_cfg_flow *)flow->rule);
+ fm10k_ffu_flow_disable(sw,
+ (struct fm10k_cfg_flow *)flow->rule);
+ }
+
+ if (!ret) {
+ TAILQ_REMOVE((struct fm10k_flow_list *)
+ fm10k_switch_dpdk_port_flow_list_get(hw),
+ flow, node);
+ rte_free(flow->rule);
+ rte_free(flow);
+ } else {
+ rte_flow_error_set(error, -ret,
+ RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "Failed to destroy flow.");
+ }
+ return ret;
+}
+
+static int
+fm10k_flow_flush(struct rte_eth_dev *dev, struct rte_flow_error *error)
+{
+ struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
+ struct rte_flow *flow;
+ void *temp;
+ int ret = 0;
+ struct fm10k_cfg_flow *cf;
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ /* Delete flows in flow list. */
+ TAILQ_FOREACH_SAFE(flow,
+ (struct fm10k_flow_list *)
+ fm10k_switch_dpdk_port_flow_list_get(hw),
+ node, temp) {
+ cf = flow->rule;
+ if (cf) {
+ fm10k_config_flow_list_delete(cf);
+ fm10k_ffu_flow_disable(sw, cf);
+ } else {
+ rte_flow_error_set(error, -ret,
+ RTE_FLOW_ERROR_TYPE_HANDLE, NULL,
+ "No such rule in flow.");
+ }
+ TAILQ_REMOVE((struct fm10k_flow_list *)
+ fm10k_switch_dpdk_port_flow_list_get(hw),
+ flow, node);
+ rte_free(flow);
+ }
+
+ return ret;
+}
+
+void
+fm10k_flow_list_init(void *flow_list)
+{
+ TAILQ_INIT((struct fm10k_flow_list *)flow_list);
+}
+
+/* Flow operations */
+const struct rte_flow_ops *
+fm10k_flow_ops_get(void)
+{
+ return &fm10k_flow_ops;
+}
diff --git a/drivers/net/fm10k/switch/fm10k_flow.h b/drivers/net/fm10k/switch/fm10k_flow.h
new file mode 100644
index 0000000..c538544
--- /dev/null
+++ b/drivers/net/fm10k/switch/fm10k_flow.h
@@ -0,0 +1,26 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2019 Silicom Ltd. Connectivity Solutions
+ */
+
+#ifndef _FM10K_SW_FLOW_H_
+#define _FM10K_SW_FLOW_H_
+
+#include <rte_time.h>
+#include <rte_kvargs.h>
+#include <rte_hash.h>
+#include <rte_flow.h>
+#include <rte_flow_driver.h>
+#include <rte_tm_driver.h>
+
+/*
+ * Struct to store flow created.
+ */
+struct rte_flow {
+ TAILQ_ENTRY(rte_flow) node;
+ enum rte_filter_type filter_type;
+ void *rule;
+};
+
+TAILQ_HEAD(fm10k_flow_list, rte_flow);
+
+#endif /* _FM10K_SW_FLOW_H_ */
diff --git a/drivers/net/fm10k/switch/fm10k_stats.c b/drivers/net/fm10k/switch/fm10k_stats.c
index 2e8d549..4f03bf7 100644
--- a/drivers/net/fm10k/switch/fm10k_stats.c
+++ b/drivers/net/fm10k/switch/fm10k_stats.c
@@ -998,6 +998,134 @@ struct fm10k_stats_data {
}
}
+void
+fm10k_stats_dpdk_port_print(struct fm10k_switch *sw)
+{
+ int i, j, lport, pf_no;
+ struct fm10k_port_counters counters;
+ struct fm10k_ext_port ext_port;
+ static struct fm10k_stats_data last_data[FM10K_SW_PEPS_MAX];
+ static struct fm10k_stats_data last_queue_data[FM10K_SW_PEPS_MAX][4];
+ struct fm10k_stats_data data, mydata;
+ char pf_ports[10];
+
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ if (sw->dpdk_cfg->ports[i].hw == NULL)
+ continue;
+ if (sw->dpdk_cfg->ports[i].type != FM10K_CONFIG_DPDK_PF)
+ continue;
+ if (sw->dpdk_cfg->dpdk_port_map[i].type ==
+ FM10K_CONFIG_PORT_MAP_NULL)
+ continue;
+
+ memset(&mydata, 0, sizeof(mydata));
+ if (sw->dpdk_cfg->dpdk_port_map[i].type ==
+ FM10K_CONFIG_PORT_MAP_PFS) {
+ for (j = 0; j < 2; j++) {
+ pf_no =
+ sw->dpdk_cfg->dpdk_port_map[i].map_no[j];
+ lport = sw->pep_map[pf_no].logical_port;
+ memset(&ext_port, 0, sizeof(ext_port));
+ ext_port.portno = lport;
+ fm10k_get_port_counters
+ (sw, &ext_port, &counters);
+ data.rx_pkts = counters.cnt_rx_bcst_pkts +
+ counters.cnt_rx_ucst_pkts;
+ data.tx_pkts = counters.cnt_tx_bcst_pkts +
+ counters.cnt_tx_ucst_pkts;
+ data.rx_drop = counters.cnt_cmpriv_drop_pkts;
+ mydata.rx_pkts += data.rx_pkts -
+ last_data[pf_no].rx_pkts;
+ mydata.tx_pkts += data.tx_pkts -
+ last_data[pf_no].tx_pkts;
+ mydata.rx_drop += data.rx_drop -
+ last_data[pf_no].rx_drop;
+ last_data[pf_no] = data;
+ }
+ } else if (sw->dpdk_cfg->dpdk_port_map[i].type ==
+ FM10K_CONFIG_PORT_MAP_PF) {
+ pf_no = sw->dpdk_cfg->dpdk_port_map[i].map_no[0];
+ lport = sw->pep_map[pf_no].logical_port;
+ memset(&ext_port, 0, sizeof(ext_port));
+ ext_port.portno = lport;
+ fm10k_get_port_counters(sw, &ext_port, &counters);
+ data.rx_pkts = counters.cnt_rx_bcst_pkts +
+ counters.cnt_rx_ucst_pkts;
+ data.tx_pkts = counters.cnt_tx_bcst_pkts +
+ counters.cnt_tx_ucst_pkts;
+ data.rx_drop = counters.cnt_cmpriv_drop_pkts;
+ mydata.rx_pkts += data.rx_pkts -
+ last_data[pf_no].rx_pkts;
+ mydata.tx_pkts += data.tx_pkts -
+ last_data[pf_no].tx_pkts;
+ mydata.rx_drop += data.rx_drop -
+ last_data[pf_no].rx_drop;
+ last_data[pf_no] = data;
+ } else {
+ continue;
+ }
+
+ if (sw->dpdk_cfg->dpdk_port_map[i].type ==
+ FM10K_CONFIG_PORT_MAP_PF)
+ sprintf(pf_ports, "%d",
+ sw->dpdk_cfg->dpdk_port_map[i].map_no[0]);
+ else
+ sprintf(pf_ports, "%d/%d",
+ sw->dpdk_cfg->dpdk_port_map[i].map_no[0],
+ sw->dpdk_cfg->dpdk_port_map[i].map_no[1]);
+
+ FM10K_SW_INFO("DPDK port %-2d pf %-5s tx_pkt %-12llu "
+ "rx_pkt %-12llu drop_pkt %-12llu\n",
+ i, pf_ports,
+ (unsigned long long)mydata.tx_pkts,
+ (unsigned long long)mydata.rx_pkts,
+ (unsigned long long)mydata.rx_drop);
+
+ if (!fm10k_config_check_debug(sw->dpdk_cfg,
+ FM10K_CONFIG_DEBUG_STATS_QUEUE))
+ continue;
+ memset(&mydata, 0, sizeof(mydata));
+ for (j = 0;
+ j < sw->dpdk_cfg->ports[i].tx_queue_num;
+ j++) {
+ struct fm10k_hw *hw = sw->dpdk_cfg->ports[i].hw;
+ uint16_t queue_id = j, pf_no;
+
+ fm10k_switch_dpdk_hw_queue_map(hw, queue_id,
+ sw->dpdk_cfg->ports[i].tx_queue_num,
+ &hw, &queue_id);
+ pf_no = fm10k_switch_dpdk_pf_no_get(hw);
+ data.tx_pkts =
+ FM10K_READ_REG(hw, FM10K_QPTC(queue_id));
+ data.rx_pkts =
+ FM10K_READ_REG(hw, FM10K_QPRC(queue_id));
+ data.rx_drop =
+ FM10K_READ_REG(hw, FM10K_QPRDC(queue_id));
+ mydata.tx_pkts += data.tx_pkts -
+ last_queue_data[pf_no][queue_id].tx_pkts;
+ mydata.rx_pkts += data.rx_pkts -
+ last_queue_data[pf_no][queue_id].rx_pkts;
+ mydata.rx_drop += data.rx_drop -
+ last_queue_data[pf_no][queue_id].rx_drop;
+ FM10K_SW_INFO(" queue %d(%d:%d) tx_pkt %-12llu "
+ "rx_pkt %-12llu drop_pkt %-12llu\n", j,
+ pf_no, queue_id,
+ (unsigned long long)(data.tx_pkts -
+ last_queue_data[pf_no][queue_id].tx_pkts),
+ (unsigned long long)(data.rx_pkts -
+ last_queue_data[pf_no][queue_id].rx_pkts),
+ (unsigned long long)(data.rx_drop -
+ last_queue_data[pf_no][queue_id].rx_drop));
+ last_queue_data[pf_no][queue_id] = data;
+ }
+ FM10K_SW_INFO(" total tx_pkt %-12llu "
+ "rx_pkt %-12llu drop_pkt %-12llu\n",
+ (unsigned long long)mydata.tx_pkts,
+ (unsigned long long)mydata.rx_pkts,
+ (unsigned long long)mydata.rx_drop);
+ }
+}
+
static void
fm10k_stats_port_counter_print(struct fm10k_switch *sw, int lport)
{
@@ -1080,6 +1208,7 @@ struct fm10k_stats_data {
FM10K_CONFIG_DEBUG_STATS_PORT)) {
FM10K_SW_INFO("--- port statistic ---\n");
fm10k_stats_epl_port_print(sw);
+ fm10k_stats_dpdk_port_print(sw);
}
if (fm10k_config_check_debug(sw->dpdk_cfg,
FM10K_CONFIG_DEBUG_STATS_FFU)) {
diff --git a/drivers/net/fm10k/switch/fm10k_switch.c b/drivers/net/fm10k/switch/fm10k_switch.c
new file mode 100644
index 0000000..ec36fa9
--- /dev/null
+++ b/drivers/net/fm10k/switch/fm10k_switch.c
@@ -0,0 +1,2362 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright 2019 Silicom Ltd. Connectivity Solutions
+ */
+
+#include <rte_ethdev.h>
+#include <rte_ethdev_pci.h>
+#include <rte_malloc.h>
+
+#include "../base/fm10k_type.h"
+#include "../base/fm10k_osdep.h"
+#include "../fm10k.h"
+#include "../fm10k_logs.h"
+#include "fm10k_debug.h"
+#include "fm10k_regs.h"
+#include "fm10k_switch.h"
+#include "fm10k_ext_port.h"
+#include "fm10k_serdes.h"
+#include "fm10k_i2c.h"
+#include "fm10k_ffu.h"
+#include "fm10k_stats.h"
+#include "fm10k_config.h"
+
+static struct fm10k_switch fm10k_sw;
+
+#define FM10K_AM_TIMEOUT 16384
+#define FM10K_COMP_PPM_SCALE 1000000
+#define FM10K_LED_POLL_INTERVAL_MS 500
+#define FM10K_LED_BLINKS_PER_SECOND 2
+
+/* Max try times to acquire switch status */
+#define FM10K_QUERY_SWITCH_STATE_TIMES 10
+/* Wait interval to get switch status */
+#define FM10K_WAIT_SWITCH_MSG_US 100000
+
+/*
+ * use epl as external port map, only support QUAD_ON
+ */
+struct fm10k_sched_prog {
+ uint8_t idle;
+ uint8_t phys; /* physical port */
+ uint8_t log; /* logical port */
+ uint8_t quad; /* now, only support QUAD_ON */
+/* convert entry to idle if EPL in quad port mode */
+#define FM10K_SW_QUAD_OFF 0
+/* always use quad port mode */
+#define FM10K_SW_QUAD_ON 1
+/* use quad port mode if link speed is 40/100G */
+#define FM10K_SW_QUAD_40_100 2
+};
+
+static struct fm10k_sched_prog
+ fm10k_sched_prog[FM10K_SW_PEPS_SUPPORTED + FM10K_SW_EPLS_SUPPORTED + 1];
+
+/*
+ * Note that for physical port numbers, the initial values used for
+ * the EPL entries assume EPL[A] is EPL[0] and EPL[B] is EPL[1].
+ * fm10k_switch_determine_epls() will update these physical port
+ * numbers based on the actual A and B indices.
+ */
+static void
+fm10k_switch_set_sched_prog(void)
+{
+ int i;
+ int start = 0;
+
+ for (i = 0; i < FM10K_SW_EPLS_SUPPORTED; i++) {
+ fm10k_sched_prog[i].idle = 0;
+ fm10k_sched_prog[i].phys = fm10k_epl_port_map[i].physical_port;
+ fm10k_sched_prog[i].log = fm10k_epl_port_map[i].logical_port;
+ fm10k_sched_prog[i].quad = FM10K_SW_QUAD_ON;
+ }
+ start += FM10K_SW_EPLS_SUPPORTED;
+ for (i = 0; i < FM10K_SW_PEPS_SUPPORTED; i++) {
+ fm10k_sched_prog[start + i].idle = 0;
+ fm10k_sched_prog[start + i].phys =
+ fm10k_pep_port_map[i].physical_port;
+ fm10k_sched_prog[start + i].log =
+ fm10k_pep_port_map[i].logical_port;
+ fm10k_sched_prog[start + i].quad = FM10K_SW_QUAD_40_100;
+ }
+ start += FM10K_SW_PEPS_SUPPORTED;
+ fm10k_sched_prog[start].idle = 1;
+}
+
+static void
+fm10k_switch_determine_epls(struct fm10k_switch *sw)
+{
+ struct fm10k_device_info *cfg = sw->info;
+ unsigned int i;
+ uint8_t phys;
+
+ sw->epla_no = 0;
+ sw->eplb_no = 6;
+
+ switch (FM10K_SW_CARD_ID(cfg->subvendor, cfg->subdevice)) {
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4_REV_2):
+ sw->epla_no = 1;
+ sw->eplb_no = 6;
+ break;
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRL_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRM_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE325G2DSIR):
+ sw->epla_no = 1;
+ sw->eplb_no = 7;
+ break;
+ }
+
+ for (i = 0; i < sizeof(fm10k_sched_prog) /
+ sizeof(fm10k_sched_prog[0]); i++) {
+ if (fm10k_sched_prog[i].idle)
+ continue;
+
+ phys = fm10k_sched_prog[i].phys;
+ if (phys <= 3) {
+ /* Substitute actual epla phys port number */
+ fm10k_sched_prog[i].phys = sw->epla_no * 4 + phys;
+ } else if (phys >= 4 && phys <= 7) {
+ /* Substitute actual eplb phys port number */
+ fm10k_sched_prog[i].phys = sw->eplb_no * 4 + (phys - 4);
+ }
+ }
+}
+
+static void
+fm10k_hw_eicr_disable_source(struct fm10k_switch *sw, unsigned int source)
+{
+ unsigned int shift;
+
+ switch (source) {
+ case FM10K_SW_EICR_PCA_FAULT:
+ shift = FM10K_SW_EICR_PCA_FAULT_SHIFT;
+ break;
+ case FM10K_SW_EICR_THI_FAULT:
+ shift = FM10K_SW_EICR_THI_FAULT_SHIFT;
+ break;
+ case FM10K_SW_EICR_FUM_FAULT:
+ shift = FM10K_SW_EICR_FUM_FAULT_SHIFT;
+ break;
+ case FM10K_SW_EICR_MAILBOX:
+ shift = FM10K_SW_EICR_MAILBOX_SHIFT;
+ break;
+ case FM10K_SW_EICR_SWITCH_READY:
+ shift = FM10K_SW_EICR_SWITCH_READY_SHIFT;
+ break;
+ case FM10K_SW_EICR_SWITCH_NREADY:
+ shift = FM10K_SW_EICR_SWITCH_NREADY_SHIFT;
+ break;
+ case FM10K_SW_EICR_SWITCH_INT:
+ shift = FM10K_SW_EICR_SWITCH_INT_SHIFT;
+ break;
+ case FM10K_SW_EICR_SRAM_ERROR:
+ shift = FM10K_SW_EICR_SRAM_ERROR_SHIFT;
+ break;
+ case FM10K_SW_EICR_VFLR:
+ shift = FM10K_SW_EICR_VFLR_SHIFT;
+ break;
+ case FM10K_SW_EICR_MAX_HOLD_TIME:
+ shift = FM10K_SW_EICR_MAX_HOLD_TIME_SHIFT;
+ break;
+ default:
+ return;
+ }
+
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_EIMR, FM10K_SW_EIMR_DISABLE << (shift * 2));
+ fm10k_write_flush(sw);
+}
+
+unsigned int
+fm10k_switch_eplidx_to_eplno(struct fm10k_switch *sw, unsigned int eplidx)
+{
+ return eplidx ? sw->eplb_no : sw->epla_no;
+}
+
+static uint16_t
+fm10k_switch_ppm_to_tx_clk_compensation_timeout(uint32_t pcs, uint32_t num_ppm)
+{
+ unsigned int scale;
+ unsigned int ppm;
+ unsigned int timeout;
+ unsigned int am_ppm;
+
+ if (num_ppm == 0 || pcs == FM10K_SW_EPL_PCS_SEL_DISABLE) {
+ if (pcs == FM10K_SW_EPL_PCS_SEL_40GBASER ||
+ pcs == FM10K_SW_EPL_PCS_SEL_100GBASER)
+ return (FM10K_AM_TIMEOUT / 2);
+ else
+ return 0;
+ }
+
+ if (pcs == FM10K_SW_EPL_PCS_SEL_40GBASER ||
+ pcs == FM10K_SW_EPL_PCS_SEL_100GBASER) {
+ am_ppm = 1000000 / FM10K_AM_TIMEOUT;
+ scale = FM10K_COMP_PPM_SCALE / 2;
+ } else {
+ am_ppm = 0;
+ scale = FM10K_COMP_PPM_SCALE;
+ }
+
+ ppm = num_ppm + am_ppm;
+ timeout = scale / ppm;
+
+ if (timeout >= 0xffff)
+ return 0xffff;
+ else
+ return timeout;
+}
+
+static int
+fm10k_switch_configure_epls(struct fm10k_hw *hw, struct fm10k_switch *sw)
+{
+ struct fm10k_ext_ports *ext_ports = sw->ext_ports;
+ struct fm10k_ext_port *port;
+ struct fm10k_device_info *cfg = sw->info;
+ u32 mac_cfg[FM10K_SW_MAC_CFG_ARRAY_SIZE];
+ u32 data, pcs, qpl;
+ u32 pcstmp;
+ unsigned int i, j;
+ unsigned int dic_enable;
+ unsigned int anti_bubble_wm;
+ unsigned int rate_fifo_wm;
+ unsigned int rate_fifo_slow_inc, rate_fifo_fast_inc;
+ int error;
+ u16 timeout;
+
+ cfg = fm10k_get_device_info(hw);
+ if (cfg == NULL)
+ return -1;
+
+ /*
+ * Assumptions:
+ * - All external interfaces are the same speed
+ * - 1G/10G ports are packed into the minimum number of EPLs
+ * - quad-mode EPLs use lane 0 as master
+ * - The lowest numbered PEPs are used
+ * - PEPs are always used in x8 mode.
+ */
+ switch (cfg->ext_port_speed) {
+ case 10:
+ pcs = FM10K_SW_EPL_PCS_SEL_10GBASER;
+ qpl = FM10K_SW_EPL_QPL_MODE_L1_L1_L1_L1;
+ dic_enable = 1;
+ anti_bubble_wm = 5;
+ rate_fifo_wm = 3;
+ rate_fifo_slow_inc = 12;
+ rate_fifo_fast_inc = 13;
+ break;
+ case 25:
+ pcs = FM10K_SW_EPL_PCS_SEL_100GBASER;
+ qpl = FM10K_SW_EPL_QPL_MODE_L1_L1_L1_L1;
+ dic_enable = 0;
+ anti_bubble_wm = 6;
+ rate_fifo_wm = 3;
+ rate_fifo_slow_inc = 32;
+ rate_fifo_fast_inc = 33;
+ break;
+ case 40:
+ pcs = FM10K_SW_EPL_PCS_SEL_40GBASER;
+ qpl = FM10K_SW_EPL_QPL_MODE_L4_XX_XX_XX;
+ dic_enable = 1;
+ anti_bubble_wm = 4;
+ rate_fifo_wm = 5;
+ rate_fifo_slow_inc = 51;
+ rate_fifo_fast_inc = 52;
+ break;
+ case 100:
+ pcs = FM10K_SW_EPL_PCS_SEL_100GBASER;
+ qpl = FM10K_SW_EPL_QPL_MODE_L4_XX_XX_XX;
+ dic_enable = 1;
+ anti_bubble_wm = 4;
+ rate_fifo_wm = 3;
+ rate_fifo_slow_inc = 129;
+ rate_fifo_fast_inc = 130;
+ break;
+ default:
+ error = -1;
+ goto done;
+ }
+
+ /*
+ * EPL_CFG_A
+ *
+ * Mark all used lanes as active and all unused lanes as
+ * inactive. Adjust timeout and skew tolerance values for EPLs that
+ * are used.
+ */
+ for (i = 0; i < FM10K_SW_EPLS_MAX; i++) {
+ data = fm10k_read_switch_reg(sw, FM10K_SW_EPL_CFG_A(i));
+ data &= ~FM10K_SW_EPL_CFG_A_ACTIVE_QUAD;
+ if (FM10K_SW_EXT_PORTS_EPL_USED(ext_ports, i)) {
+ for (j = 0; j < ext_ports->num_ports; j++) {
+ port = &ext_ports->ports[j];
+ if (port->eplno == i)
+ data |= port->is_quad ?
+ FM10K_SW_EPL_CFG_A_ACTIVE_QUAD :
+ FM10K_SW_EPL_CFG_A_ACTIVE
+ (port->first_lane);
+ }
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ EPL_CFG_A_TIMEOUT, 19, data);
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ EPL_CFG_A_SKEW_TOLERANCE, 38, data);
+ }
+ fm10k_write_switch_reg(sw, FM10K_SW_EPL_CFG_A(i), data);
+ }
+
+ /*
+ * EPL_CFG_B
+ *
+ * Disable all unused lanes and configure all used ones
+ * appropriately. For EPLs used in quad port mode, the master lane
+ * is the only one that gets configured.
+ */
+ data = 0;
+ for (i = 0; i < FM10K_SW_EPLS_MAX; i++) {
+ if (FM10K_SW_EXT_PORTS_EPL_USED(ext_ports, i)) {
+ for (j = 0; j < FM10K_SW_EPL_LANES; j++) {
+ if (j < ext_ports->ports_per_epl)
+ pcstmp = pcs;
+ else
+ pcstmp = FM10K_SW_EPL_PCS_SEL_DISABLE;
+ data |= FM10K_SW_MAKE_REG_FIELD_IDX
+ (EPL_CFG_B_PCS_SEL, j, pcstmp, j, j);
+ }
+ data |= FM10K_SW_MAKE_REG_FIELD
+ (EPL_CFG_B_QPL_MODE, qpl);
+ } else {
+ for (j = 0; j < FM10K_SW_EPL_LANES; j++)
+ data |= FM10K_SW_MAKE_REG_FIELD_IDX
+ (EPL_CFG_B_PCS_SEL, j,
+ FM10K_SW_EPL_PCS_SEL_DISABLE, j, j);
+ data |= FM10K_SW_MAKE_REG_FIELD
+ (EPL_CFG_B_QPL_MODE,
+ FM10K_SW_EPL_QPL_MODE_XX_XX_XX_XX);
+ }
+ fm10k_write_switch_reg(sw, FM10K_SW_EPL_CFG_B(i), data);
+ }
+
+ /*
+ * MAC_CFG, LINK_RULES and PCS_ML_BASER_CFG
+ *
+ * Only EPLs/lanes that are being used are initialized. All others
+ * are left at their defaults.
+ */
+ for (i = 0; i < FM10K_SW_EPLS_MAX; i++) {
+ if (!FM10K_SW_EXT_PORTS_EPL_USED(ext_ports, i))
+ continue;
+ for (j = 0; j < ext_ports->ports_per_epl; j++) {
+ fm10k_read_switch_array(sw, FM10K_SW_MAC_CFG(i, j),
+ mac_cfg, FM10K_SW_MAC_CFG_ARRAY_SIZE);
+
+ /* dic enable */
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_TX_IDLE_ENABLE_DIC,
+ dic_enable, mac_cfg);
+
+ /* ifg */
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_IDLE_MIN_IFG_BYTES,
+ 12, mac_cfg);
+
+ /* tx pad size: (64 bytes + 8 preamble) / 4 */
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_MIN_COLUMNS,
+ 18, mac_cfg);
+
+ /* tx clock compensation */
+ timeout =
+ fm10k_switch_ppm_to_tx_clk_compensation_timeout(pcs,
+ 100);
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_TX_CLOCK_COMPENSATION_ENABLE,
+ (timeout > 0), mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_CLOCK_COMPENSATION_TIMEOUT,
+ timeout, mac_cfg);
+
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_PREAMBLE_MODE,
+ 0, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_RX_MIN_FRAME_LENGTH,
+ 64, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_RX_MAX_FRAME_LENGTH,
+ FM10K_SW_PACKET_SIZE_MAX, mac_cfg);
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_RX_IGNORE_IFG_ERRORS,
+ 0, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_FCS_MODE,
+ FM10K_SW_TX_MAX_FCS_MODE_REPLACE_NORMAL, mac_cfg);
+
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_IEEE_1588_ENABLE,
+ 0, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_DRAIN_MODE,
+ FM10K_SW_TX_MAC_DRAIN_MODE_DRAIN_NORMAL, mac_cfg);
+
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_PC_ACT_TIMEOUT,
+ 100, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_PC_ACT_TIME_SCALE,
+ 3, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_LPI_TIMEOUT,
+ 180, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_LPI_TIME_SCALE,
+ 1, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_LPI_HOLD_TIMEOUT,
+ 20, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_LPI_HOLD_TIME_SCALE,
+ 1, mac_cfg);
+
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_TX_LPI_AUTOMATIC,
+ 1, mac_cfg);
+ FM10K_SW_SET_ARRAY_BIT(mac_cfg,
+ MAC_CFG_TX_LP_IDLE_REQUEST,
+ 0, mac_cfg);
+
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_FAULT_MODE,
+ FM10K_SW_TX_MAC_FAULT_MODE_NORMAL, mac_cfg);
+
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_ANTI_BUBBLE_WATERMARK,
+ anti_bubble_wm, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_RATE_FIFO_WATERMARK,
+ rate_fifo_wm, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_RATE_FIFO_FAST_INC,
+ rate_fifo_fast_inc, mac_cfg);
+ FM10K_SW_REPLACE_ARRAY_FIELD(mac_cfg,
+ MAC_CFG_TX_RATE_FIFO_SLOW_INC,
+ rate_fifo_slow_inc, mac_cfg);
+
+ fm10k_write_switch_array(sw, FM10K_SW_MAC_CFG(i, j),
+ mac_cfg, FM10K_SW_MAC_CFG_ARRAY_SIZE);
+
+ data =
+ fm10k_read_switch_reg(sw,
+ FM10K_SW_LINK_RULES(i, j));
+
+ /* link-up debounce params */
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ LINK_RULES_FAULT_TIME_SCALE_UP,
+ 4, data);
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ LINK_RULES_FAULT_TICKS_UP,
+ 30, data);
+
+ /* link-down debounce params */
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ LINK_RULES_FAULT_TIME_SCALE_DOWN,
+ 4, data);
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ LINK_RULES_FAULT_TICKS_DOWN,
+ 5, data);
+
+ FM10K_SW_REPLACE_REG_FIELD(data,
+ LINK_RULES_HEARTBEAT_TIME_SCALE,
+ cfg->ext_port_speed == 10 ? 4 : 0, data);
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_LINK_RULES(i, j), data);
+ }
+
+ if (cfg->ext_port_speed != 10)
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_PCS_ML_BASER_CFG(i), 0x00003fff);
+ }
+
+ /*
+ * LANE_CFG, LANE_SERDES_CFG, LANE_ENERGY_DETECT_CFG,
+ * LANE_SIGNAL_DETECT_CFG, and EPL_FIFO_ERROR_STATUS
+ *
+ * Only EPLs/lanes that are being used are initialized. All others
+ * are left at their defaults.
+ */
+ for (i = 0; i < FM10K_SW_EPLS_MAX; i++) {
+ if (!FM10K_SW_EXT_PORTS_EPL_USED(ext_ports, i))
+ continue;
+ for (j = 0; j < ext_ports->ports_per_epl; j++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_LANE_CFG(i, j), 0);
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_LANE_SERDES_CFG(i, j), 0x1106);
+
+ data = fm10k_read_switch_reg(sw,
+ FM10K_SW_LANE_ENERGY_DETECT_CFG(i, j));
+
+ data |=
+ FM10K_SW_LANE_ENERGY_DETECT_CFG_ED_MASK_RX_SIGNAL_OK |
+ FM10K_SW_LANE_ENERGY_DETECT_CFG_ED_MASK_RX_RDY |
+ FM10K_SW_LANE_ENERGY_DETECT_CFG_ED_MASK_RX_ACTIVITY;
+
+ data &=
+ ~FM10K_SW_LANE_ENERGY_DETECT_CFG_ED_MASK_ENERGY_DETECT;
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_LANE_ENERGY_DETECT_CFG(i, j), data);
+
+ data = fm10k_read_switch_reg(sw,
+ FM10K_SW_LANE_SIGNAL_DETECT_CFG(i, j));
+
+ data &=
+ ~(FM10K_SW_LANE_SIGNAL_DETECT_CFG_SD_MASK_RX_SIGNAL_OK |
+ FM10K_SW_LANE_SIGNAL_DETECT_CFG_SD_MASK_RX_RDY);
+
+ data |=
+ FM10K_SW_LANE_SIGNAL_DETECT_CFG_SD_MASK_RX_ACTIVITY |
+ FM10K_SW_LANE_SIGNAL_DETECT_CFG_SD_MASK_ENERGY_DETECT;
+
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_LANE_SIGNAL_DETECT_CFG(i, j), data);
+
+ data = 0;
+ data |= ((1 << j) << 4) | (1 << j);
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_EPL_FIFO_ERROR_STATUS(i), data);
+ }
+ }
+
+ error = fm10k_epl_serdes_reset_and_load_all(sw);
+ if (error)
+ goto done;
+
+ /*
+ * EPL_FIFO_ERROR_STATUS LINK_IP
+ */
+ for (i = 0; i < FM10K_SW_EPLS_MAX; i++) {
+ if (!FM10K_SW_EXT_PORTS_EPL_USED(ext_ports, i))
+ continue;
+ for (j = 0; j < ext_ports->ports_per_epl; j++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_LINK_IP(i, j), FM10K_SW_MASK32(31, 0));
+ }
+
+ data = 0;
+ data |= ((1 << j) << 4) | (1 << j);
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_EPL_FIFO_ERROR_STATUS(i), data);
+ }
+done:
+ return (error);
+}
+
+int fm10k_switch_dpdk_port_no_get(struct fm10k_hw *hw);
+
+int
+fm10k_switch_mirror_set(struct fm10k_hw *hw, u16 dest_port, u16 vlan)
+{
+ struct fm10k_switch *sw = &fm10k_sw;
+ int src_port = fm10k_switch_dpdk_port_no_get(hw);
+
+ /* source port is external port number */
+ if (src_port < 0 || src_port == dest_port)
+ return -1;
+
+ return fm10k_ffu_mirror_set(sw, src_port, dest_port, vlan);
+}
+
+int fm10k_switch_mirror_reset(struct fm10k_hw *hw)
+{
+ struct fm10k_switch *sw = &fm10k_sw;
+ int src_port = fm10k_switch_dpdk_port_no_get(hw);
+
+ if (src_port < 0)
+ return -1;
+
+ return fm10k_ffu_mirror_reset(sw, src_port);
+}
+
+typedef struct {
+ u8 lport;
+ u8 has_ftag;
+} fm10k_sw_lport;
+
+static int
+fm10k_switch_init(struct fm10k_hw *hw, struct fm10k_switch *sw)
+{
+ u32 data;
+ unsigned int num_lports = sw->info->num_peps + FM10K_SW_EPLS_SUPPORTED;
+ fm10k_sw_lport all_lports[num_lports];
+ struct fm10k_device_info *cfg = sw->info;
+ unsigned int i;
+ unsigned int is_quad;
+ unsigned int table_idx;
+ int error;
+ u32 watermark;
+ u64 data64, data64_2;
+
+ /*
+ * Build list of all logical ports that might appear in the
+ * scheduler program. Note that for any particular card
+ * configuration, not all of these logical ports may be used.
+ */
+ table_idx = 0;
+ for (i = 0; i < sw->info->num_peps; i++, table_idx++) {
+ all_lports[table_idx].lport = sw->pep_map[i].logical_port;
+ all_lports[table_idx].has_ftag = 1;
+ }
+ for (i = 0; i < FM10K_SW_EPLS_SUPPORTED; i++, table_idx++) {
+ all_lports[table_idx].lport = sw->epl_map[i].logical_port;
+ all_lports[table_idx].has_ftag = 0;
+ }
+
+ if (table_idx != num_lports) {
+ FM10K_SW_ERR("fm10k switch lport table construction error");
+ return -1;
+ }
+
+ /*
+ * Reset the switch to get to the default state
+ */
+ data = fm10k_read_switch_reg(sw, FM10K_SW_SOFT_RESET);
+ data &= ~FM10K_SW_SOFT_RESET_SWITCH_READY;
+ fm10k_write_switch_reg(sw, FM10K_SW_SOFT_RESET, data);
+ fm10k_write_flush(sw);
+
+ usec_delay(100);
+ data = fm10k_read_switch_reg(sw, FM10K_SW_SOFT_RESET);
+ data |= FM10K_SW_SOFT_RESET_SWITCH_RESET |
+ FM10K_SW_SOFT_RESET_EPL_RESET;
+ fm10k_write_switch_reg(sw, FM10K_SW_SOFT_RESET, data);
+ fm10k_write_flush(sw);
+ usec_delay(1000);
+
+ /* Clear memories */
+ fm10k_write_switch_reg64(sw, FM10K_SW_BIST_CTRL,
+ FM10K_SW_BIST_CTRL_BIST_MODE_FABRIC |
+ FM10K_SW_BIST_CTRL_BIST_MODE_TUNNEL |
+ FM10K_SW_BIST_CTRL_BIST_MODE_EPL);
+ fm10k_write_flush(sw);
+
+ fm10k_write_switch_reg64(sw, FM10K_SW_BIST_CTRL,
+ FM10K_SW_BIST_CTRL_BIST_MODE_FABRIC |
+ FM10K_SW_BIST_CTRL_BIST_MODE_TUNNEL |
+ FM10K_SW_BIST_CTRL_BIST_MODE_EPL |
+ FM10K_SW_BIST_CTRL_BIST_RUN_FABRIC |
+ FM10K_SW_BIST_CTRL_BIST_RUN_TUNNEL |
+ FM10K_SW_BIST_CTRL_BIST_RUN_EPL);
+ fm10k_write_flush(sw);
+ usec_delay(800);
+
+ fm10k_write_switch_reg64(sw, FM10K_SW_BIST_CTRL, 0);
+ fm10k_write_flush(sw);
+
+ data = fm10k_read_switch_reg(sw, FM10K_SW_SOFT_RESET);
+ data &= ~(FM10K_SW_SOFT_RESET_SWITCH_RESET |
+ FM10K_SW_SOFT_RESET_EPL_RESET);
+ fm10k_write_switch_reg(sw, FM10K_SW_SOFT_RESET, data);
+ fm10k_write_flush(sw);
+ /* ensure switch reset is deasserted for at least 100ns */
+ usec_delay(1);
+
+ sw->epl_sbus = fm10k_sbus_attach(sw, "EPL", FM10K_SW_SBUS_EPL_CFG);
+ if (sw->epl_sbus == NULL) {
+ error = -1;
+ goto done;
+ }
+
+ /* Clear non-BIST accessible pause state memories */
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_CM_EGRESS_PAUSE_COUNT(i, 0), 0);
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_CM_EGRESS_PAUSE_COUNT(i, 1), 0);
+ }
+
+ /* Initialize RXQ_MCAST list */
+ for (i = 0; i < FM10K_SW_SCHED_RXQ_STORAGE_POINTERS_ENTRIES; i++) {
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_SCHED_RXQ_STORAGE_POINTERS(i),
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_RXQ_STORAGE_POINTERS_HEAD_PAGE, i) |
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_RXQ_STORAGE_POINTERS_TAIL_PAGE, i));
+ }
+ for (i = 0;
+ i < FM10K_SW_SCHED_RXQ_FREELIST_INIT_ENTRIES -
+ FM10K_SW_SCHED_RXQ_STORAGE_POINTERS_ENTRIES; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_RXQ_FREELIST_INIT,
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_RXQ_FREELIST_INIT_ADDRESS,
+ i + FM10K_SW_SCHED_RXQ_STORAGE_POINTERS_ENTRIES));
+ }
+ /* Initialize TXQ list */
+ for (i = 0; i < FM10K_SW_SCHED_TXQ_HEAD_PERQ_ENTRIES; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_TXQ_HEAD_PERQ(i),
+ FM10K_SW_MAKE_REG_FIELD(SCHED_TXQ_HEAD_PERQ_HEAD, i));
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_TXQ_TAIL0_PERQ(i),
+ FM10K_SW_MAKE_REG_FIELD(SCHED_TXQ_TAIL0_PERQ_TAIL, i));
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_TXQ_TAIL1_PERQ(i),
+ FM10K_SW_MAKE_REG_FIELD(SCHED_TXQ_TAIL1_PERQ_TAIL, i));
+ }
+ for (i = 0;
+ i < FM10K_SW_SCHED_TXQ_FREELIST_INIT_ENTRIES -
+ FM10K_SW_SCHED_TXQ_HEAD_PERQ_ENTRIES; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_TXQ_FREELIST_INIT,
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_TXQ_FREELIST_INIT_ADDRESS,
+ i + FM10K_SW_SCHED_TXQ_HEAD_PERQ_ENTRIES));
+ }
+ /* Initialize free segment list */
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_SSCHED_RX_PERPORT(i),
+ FM10K_SW_MAKE_REG_FIELD(SCHED_SSCHED_RX_PERPORT_NEXT, i));
+ }
+ for (i = 0;
+ i < FM10K_SW_SCHED_FREELIST_INIT_ENTRIES -
+ FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_FREELIST_INIT,
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_FREELIST_INIT_ADDRESS,
+ i + FM10K_SW_LOGICAL_PORTS_MAX));
+ }
+ /* Disable switch scan chain */
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCAN_DATA_IN,
+ FM10K_SW_SCAN_DATA_IN_UPDATE_NODES |
+ FM10K_SW_SCAN_DATA_IN_PASSTHRU);
+
+ error = fm10k_switch_configure_epls(hw, sw);
+ if (error)
+ goto done;
+
+ /*
+ * for now configure store-and-forward between PEPs and external
+ * ports regardless of relative speeds
+ */
+ for (i = 0; i < num_lports; i++) {
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_SAF_MATRIX(all_lports[i].lport),
+ FM10K_SW_SAF_MATRIX_ENABLE_SNF_ALL_PORTS);
+ }
+
+ /* Disable MTU violation trap */
+ data = fm10k_read_switch_reg(sw, FM10K_SW_SYS_CFG_1);
+ data &= ~FM10K_SW_SYS_CFG_1_TRAP_MTU_VIOLATIONS;
+ fm10k_write_switch_reg(sw, FM10K_SW_SYS_CFG_1, data);
+
+ /* Disable ingress VLAN filtering and learning */
+ for (i = 0; i < num_lports; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_PORT_CFG_3(all_lports[i].lport), 0);
+ }
+ /*
+ * Make all ports members of every VLAN, and configure every VLAN to
+ * use MST instance 0.
+ */
+ data64 = 0;
+ for (i = 0; i < num_lports; i++)
+ data64 |=
+ FM10K_SW_INGRESS_VID_TABLE_MEMBERSHIP
+ (all_lports[i].lport);
+ for (i = 0; i < FM10K_SW_INGRESS_VID_TABLE_ENTRIES; i++) {
+ fm10k_write_switch_reg128(sw, FM10K_SW_INGRESS_VID_TABLE(i),
+ FM10K_SW_INGRESS_VID_TABLE_REFLECT, data64);
+ }
+ data64 = 0;
+ for (i = 0; i < num_lports; i++)
+ data64 |=
+ FM10K_SW_EGRESS_VID_TABLE_MEMBERSHIP
+ (all_lports[i].lport);
+ for (i = 0; i < FM10K_SW_EGRESS_VID_TABLE_ENTRIES; i++)
+ fm10k_write_switch_reg128(sw,
+ FM10K_SW_EGRESS_VID_TABLE(i), 0, data64);
+
+ /* Init MOD_VLAN_TAG_VID1_MAP */
+ for (i = 0; i < FM10K_SW_INGRESS_VID_TABLE_ENTRIES; i++) {
+ data64 = i;
+ data64 = data64 << 48;
+ fm10k_write_switch_reg64(sw,
+ 0xE80000 + 0x2 * i + 0x20000, data64);
+ }
+
+ /* Configure MST instance 0 to forward for all ports */
+ data64 = 0;
+ for (i = 0; i < num_lports; i++)
+ data64 |=
+ FM10K_SW_EGRESS_MST_TABLE_FORWARDING
+ (all_lports[i].lport);
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_EGRESS_MST_TABLE(0), data64);
+
+ data64 = 0;
+ data64_2 = 0;
+ for (i = 0; i < num_lports; i++) {
+ if (all_lports[i].lport <
+ FM10K_SW_INGRESS_MST_TABLE_PORTS_PER_TABLE) {
+ data64 |=
+ FM10K_SW_MAKE_REG_FIELD_IDX64
+ (INGRESS_MST_TABLE_STP_STATE,
+ all_lports[i].lport,
+ FM10K_SW_INGRESS_MST_TABLE_STP_STATE_FORWARD,
+ all_lports[i].lport, all_lports[i].lport);
+ } else {
+ data64_2 |=
+ FM10K_SW_MAKE_REG_FIELD_IDX64
+ (INGRESS_MST_TABLE_STP_STATE,
+ all_lports[i].lport,
+ FM10K_SW_INGRESS_MST_TABLE_STP_STATE_FORWARD,
+ all_lports[i].lport, all_lports[i].lport);
+ }
+ }
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_INGRESS_MST_TABLE(0, 0), data64);
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_INGRESS_MST_TABLE(1, 0), data64_2);
+
+ for (i = 0; i < num_lports; i++) {
+ data64 = fm10k_read_switch_reg64(sw,
+ FM10K_SW_PARSER_PORT_CFG_1(all_lports[i].lport));
+ data64 |= FM10K_SW_PARSER_PORT_CFG_1_VLAN1_TAG(0) |
+ FM10K_SW_PARSER_PORT_CFG_1_VLAN2_TAG(0);
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_PARSER_PORT_CFG_1(all_lports[i].lport), data64);
+
+ /*
+ * Configure tags for f-tagged lports
+ */
+ if (all_lports[i].has_ftag) {
+ data64 = fm10k_read_switch_reg64(sw,
+ FM10K_SW_PARSER_PORT_CFG_1(all_lports[i].lport));
+ data64 |= FM10K_SW_PARSER_PORT_CFG_1_FTAG;
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_PARSER_PORT_CFG_1(all_lports[i].lport),
+ data64);
+
+ data64 = fm10k_read_switch_reg64(sw,
+ FM10K_SW_MOD_PER_PORT_CFG_2(all_lports[i].lport));
+ data64 |= FM10K_SW_MOD_PER_PORT_CFG_2_FTAG;
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_MOD_PER_PORT_CFG_2(all_lports[i].lport),
+ data64);
+ }
+ data64 = fm10k_read_switch_reg64(sw,
+ FM10K_SW_PARSER_PORT_CFG_2(all_lports[i].lport));
+ data64 |= FM10K_SW_PARSER_PORT_CFG_2_PARSE_L3;
+ data64 |= FM10K_SW_PARSER_PORT_CFG_2_PARSE_L4;
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_PARSER_PORT_CFG_2(all_lports[i].lport), data64);
+ }
+
+ /*
+ * Assign default SGLORTs for the EPL ports in the PARSER config.
+ * This isn't necessary for the PEPs as for those, as all frames
+ * ingressing from PEPs are tagged with an SGLORT that is derived
+ * from a per-tx-queue setting.
+ *
+ * The PORT_CFG_ISL register offset is determined by logical port
+ * number and the register contents determined by the corresponding
+ * SGLORT.
+ */
+ for (i = 0; i < FM10K_SW_EPLS_SUPPORTED; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_PORT_CFG_ISL
+ (fm10k_sw.epl_map[i].logical_port),
+ FM10K_SW_MAKE_REG_FIELD
+ (PORT_CFG_ISL_SGLORT,
+ fm10k_sw.epl_map[i].glort));
+ }
+
+ /*
+ * FFU counter, 11.10.3.5 POLICER_CFG[0..3]
+ */
+ fm10k_write_switch_reg64(sw,
+ 0xE40000 + 0x2 * FM10K_SW_FFU_CNT_BANK + 0x11000,
+ FM10K_SW_POLICER_LAST);
+
+ /*
+ * 11.10.3.1 POLICER_CFG_4K[0..1][0..4095]
+ * 11.10.3.2 POLICER_CFG_512[0..1][0..511]
+ */
+ for (i = 0; i < FM10K_SW_FFU_CNT_MAX; i++) {
+ if (FM10K_SW_FFU_CNT_BANK < 2) {
+ fm10k_read_switch_reg64(sw,
+ 0xE40000 + 0x2000 * FM10K_SW_FFU_CNT_BANK +
+ 0x2 * (i + FM10K_SW_FFU_CNT_START) + 0x0);
+ } else {
+ fm10k_read_switch_reg64(sw,
+ 0xE40000 + 0x400 * (FM10K_SW_FFU_CNT_BANK - 2) +
+ 0x2 * (i + FM10K_SW_FFU_CNT_START) + 0x4000);
+ }
+ }
+
+ /*
+ * FFU
+ *
+ * The FFU is programmed to match on SGLORT and to set the DGLORT to
+ * a unique value corresponding to the given SGLORT. One TCAM entry
+ * is required per host port per PEP and one per external interface.
+ * Only slice 0 is used.
+ */
+ if (fm10k_ffu_init(sw, sw->dpdk_cfg) < 0)
+ return -1;
+
+ /*
+ * Program the segment scheduler (see tables at top)
+ *
+ * The TX and RX schedules are the same. Page 0 is used.
+ */
+ if (cfg->ext_port_speed == 40 || cfg->ext_port_speed == 100)
+ is_quad = 1;
+ else
+ is_quad = 0;
+ for (i = 0; i < sizeof(fm10k_sched_prog) /
+ sizeof(fm10k_sched_prog[0]); i++) {
+ /*
+ * In addition to explicit idle cycles, non-quad port
+ * entries are converted to idle cycles if the interfaces
+ * are configured in quad-port mode.
+ */
+ if (fm10k_sched_prog[i].idle ||
+ (!fm10k_sched_prog[i].quad && is_quad))
+ data = FM10K_SW_SCHED_SCHEDULE_IDLE;
+ else
+ data = FM10K_SW_SCHED_SCHEDULE_ENTRY
+ (fm10k_sched_prog[i].phys,
+ fm10k_sched_prog[i].log,
+ (fm10k_sched_prog[i].quad == FM10K_SW_QUAD_40_100) ?
+ is_quad : fm10k_sched_prog[i].quad);
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_RX_SCHEDULE(0, i), data);
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_SCHED_TX_SCHEDULE(0, i), data);
+ }
+
+ fm10k_write_switch_reg(sw, FM10K_SW_SCHED_SCHEDULE_CTRL,
+ FM10K_SW_SCHED_SCHEDULE_CTRL_RX_ENABLE |
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_SCHEDULE_CTRL_RX_MAX_INDEX,
+ (sizeof(fm10k_sched_prog) /
+ sizeof(fm10k_sched_prog[0])) - 1) |
+ FM10K_SW_SCHED_SCHEDULE_CTRL_TX_ENABLE |
+ FM10K_SW_MAKE_REG_FIELD
+ (SCHED_SCHEDULE_CTRL_TX_MAX_INDEX,
+ (sizeof(fm10k_sched_prog) /
+ sizeof(fm10k_sched_prog[0])) - 1));
+
+ /* Per 5.7.10.4 */
+ watermark = FM10K_SW_MEM_POOL_SEGS_MAX -
+ ((sw->info->num_peps +
+ FM10K_SW_EPLS_SUPPORTED * FM10K_SW_EPL_LANES) *
+ FM10K_SW_HOWMANY(FM10K_SW_PACKET_SIZE_MAX,
+ FM10K_SW_MEM_POOL_SEG_SIZE, FM10K_SW_MEM_POOL_SEG_SIZE)) -
+ FM10K_SW_MEM_POOL_SEGS_RSVD;
+ fm10k_write_switch_reg(sw, FM10K_SW_CM_GLOBAL_WM,
+ FM10K_SW_MAKE_REG_FIELD(CM_GLOBAL_WM_WATERMARK, watermark));
+ fm10k_write_switch_reg(sw, FM10K_SW_CM_GLOBAL_CFG,
+ FM10K_SW_CM_GLOBAL_CFG_WM_SWEEP_EN |
+ FM10K_SW_CM_GLOBAL_CFG_PAUSE_GEN_SWEEP_EN |
+ FM10K_SW_CM_GLOBAL_CFG_PAUSE_REC_SWEEP_EN |
+ FM10K_SW_MAKE_REG_FIELD
+ (CM_GLOBAL_CFG_NUM_SWEEPER_PORTS,
+ FM10K_SW_LOGICAL_PORTS_MAX));
+
+ /* Configure stats counters */
+ data = FM10K_SW_RX_STATS_CFG_ENABLE_ALL_BANKS;
+ data64 =
+ FM10K_SW_MOD_STATS_CFG_ENABLE_GROUP_7 |
+ FM10K_SW_MOD_STATS_CFG_ENABLE_GROUP_8;
+ for (i = 0; i < num_lports; i++) {
+ fm10k_write_switch_reg(sw,
+ FM10K_SW_RX_STATS_CFG(all_lports[i].lport),
+ data | ((all_lports[i].has_ftag) ?
+ FM10K_SW_MAKE_REG_FIELD
+ (RX_STATS_CFG_PER_FRAME_ADJUSTMENT,
+ FM10K_SW_FTAG_SIZE) :
+ 0));
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_MOD_STATS_CFG(all_lports[i].lport),
+ data64);
+ }
+
+ /* Transition switch to ready */
+ data = fm10k_read_switch_reg(sw, FM10K_SW_SOFT_RESET);
+ data |= FM10K_SW_SOFT_RESET_SWITCH_READY;
+ fm10k_write_switch_reg(sw, FM10K_SW_SOFT_RESET, data);
+
+ done:
+ return (error);
+}
+
+static void
+fm10k_switch_leds_init(struct fm10k_switch *sw)
+{
+ struct fm10k_device_info *cfg = sw->info;
+ unsigned int i;
+ uint8_t addr;
+ uint32_t data;
+ int max;
+
+ switch (FM10K_SW_CARD_ID(cfg->subvendor, cfg->subdevice)) {
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QS41):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QS41):
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QS43):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QS43):
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QL4):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * Set up the first PCA9545 mux so we can get at the PCA9635
+ * that the LED control lines are connected to.
+ */
+ fm10k_i2c_write8(sw->i2c, 0x70, 0x04);
+
+ /*
+ * PCA9635 initialization
+ * only differences from defaults are noted
+ */
+
+ /* MODE1 - put into sleep mode to ensure it is brought out
+ * of sleep without violating the oscillator startup wait
+ */
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x00, 0x10);
+ fm10k_udelay(10);
+
+ /* MODE1 - normal mode, disable all call */
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x00, 0x00);
+
+ /* Wait for oscillator to stabilize after coming out of sleep */
+ fm10k_udelay(500);
+
+ /* MODE2 - group control is blinking, open drain outputs,
+ * OE high -> LEDn = 0
+ */
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x01, 0x20);
+
+ /* PWM0 - 100% duty cycle */
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x02, 0xff);
+
+ /* PWM1 - 100% duty cycle */
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x03, 0xff);
+
+ /* GRPPWM - 50% blink duty cycle */
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x12, 0x80);
+
+ /* GRPFREQ - FM10K_LED_BLINKS_PER_SECOND */
+ if (24 / FM10K_LED_BLINKS_PER_SECOND > 1)
+ max = 24 / FM10K_LED_BLINKS_PER_SECOND;
+ else
+ max = 1;
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x13, max - 1);
+
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4_REV_2):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * Set up the first PCA9545 mux so we can get at the PCA9635s
+ * that the LED control lines are connected to.
+ */
+ fm10k_i2c_write8(sw->i2c, 0x70, 0x01);
+
+ /*
+ * PCA9635 initialization
+ * only differences from defaults are noted
+ */
+
+ addr = 0x6a;
+ for (i = 0; i < 2; i++) {
+ /* MODE1 - put into sleep mode to ensure it is
+ * brought out of sleep without violating the
+ * oscillator startup wait
+ */
+ fm10k_i2c_write16(sw->i2c, addr, 0x00, 0x10);
+ fm10k_udelay(10);
+
+ /* MODE1 - normal mode, disable all call */
+ fm10k_i2c_write16(sw->i2c, addr, 0x00, 0x00);
+
+ /* MODE2 - group control is blinking, open drain
+ * outputs, OE high -> LEDn = 0
+ */
+ fm10k_i2c_write16(sw->i2c, addr, 0x01, 0x20);
+
+ /* Wait for oscillator to stabilize
+ * after coming out of sleep
+ */
+ fm10k_udelay(500);
+
+ /* PWM0 - 100% duty cycle */
+ fm10k_i2c_write16(sw->i2c, addr, 0x02, 0xff);
+
+ /* PWM3 - 100% duty cycle */
+ fm10k_i2c_write16(sw->i2c, addr, 0x05, 0xff);
+
+ /* GRPPWM - 50% blink duty cycle */
+ fm10k_i2c_write16(sw->i2c, addr, 0x12, 0x80);
+
+ /* GRPFREQ - FM10K_LED_BLINKS_PER_SECOND */
+ if (24 / FM10K_LED_BLINKS_PER_SECOND > 1)
+ max = 24 / FM10K_LED_BLINKS_PER_SECOND;
+ else
+ max = 1;
+ fm10k_i2c_write16(sw->i2c, addr, 0x13, max - 1);
+
+ addr = 0x69;
+ }
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRL_QXSL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ addr = 0x62;
+ fm10k_i2c_write16(sw->i2c, addr, 0x03, 0x88);
+ fm10k_i2c_write16(sw->i2c, addr, 0x01, 0x0);
+
+ data = fm10k_read_switch_reg(sw, 0xc2b);
+ data |= 1 << 24;
+ fm10k_write_switch_reg(sw, 0xc2b, data);
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRM_QXSL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * PCA9538 initialization
+ * only differences from defaults are noted
+ */
+ /*
+ * 00000000: No link
+ * 00000001: 10G Port 0
+ * 00000010: 40G Port 0
+ * 00000100: 25G Port 0
+ * 00001000: 100G Port 0
+ */
+ addr = 0x62;
+ fm10k_i2c_write16(sw->i2c, addr, 0x03, 0x0);
+ fm10k_i2c_write16(sw->i2c, addr, 0x01, 0x0);
+
+ addr = 0x65;
+ fm10k_i2c_write16(sw->i2c, addr, 0x03, 0xf0);
+ fm10k_i2c_write16(sw->i2c, addr, 0x01, 0x0);
+
+ addr = 0x66;
+ fm10k_i2c_write16(sw->i2c, addr, 0x03, 0x0);
+ fm10k_i2c_write16(sw->i2c, addr, 0x01, 0x0);
+
+ /* set LEC_CFG */
+ data = fm10k_read_switch_reg(sw, 0xc2b);
+ data |= 1 << 24;
+ fm10k_write_switch_reg(sw, 0xc2b, data);
+
+ /* port from rdifd, LED */
+ fm10k_gpio_output_set(sw, 4, 0);
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ default:
+ FM10K_SW_ERR("don't know how to operate LEDs for this card "
+ "(subvendor=0x%04x subdevice=0x%04x)",
+ cfg->subvendor, cfg->subdevice);
+ break;
+ }
+}
+
+static unsigned int
+fm10k_switch_pca9635_led_bits(uint8_t led_flags)
+{
+ unsigned int bits;
+
+ if (led_flags & FM10K_SW_EXT_PORT_LED_FLAG_UP) {
+ if (led_flags & FM10K_SW_EXT_PORT_LED_FLAG_ACTIVE)
+ bits = 0x3; /* group blink */
+ else
+ bits = 0x1; /* full on */
+ } else {
+ bits = 0; /* off */
+ }
+ return (bits);
+}
+
+static void
+fm10k_switch_process_leds(void *ctx)
+{
+ struct fm10k_switch *sw = ctx;
+ struct fm10k_device_info *cfg = sw->info;
+ struct fm10k_ext_ports *ports = sw->ext_ports;
+ struct fm10k_ext_port *port;
+ unsigned int i;
+ unsigned int num_ports = ports->num_ports;
+ uint32_t data;
+ uint8_t update_port[num_ports];
+ uint8_t led_flags = 0, read;
+ uint8_t addr;
+
+ FM10K_SW_SWITCH_LOCK(sw);
+
+ if (sw->master_hw->sw_addr == NULL) {
+ FM10K_SW_SWITCH_UNLOCK(sw);
+ return;
+ }
+
+ for (i = 0; i < num_ports; i++) {
+ port = &ports->ports[i];
+ data = fm10k_read_switch_reg(sw,
+ FM10K_SW_EPL_LED_STATUS(port->eplno));
+ led_flags =
+ ((data & FM10K_SW_EPL_LED_STATUS_PORT_LINK_UP
+ (port->first_lane)) ?
+ FM10K_SW_EXT_PORT_LED_FLAG_UP : 0) |
+ ((data &
+ (FM10K_SW_EPL_LED_STATUS_PORT_TRANSMITTING
+ (port->first_lane) |
+ FM10K_SW_EPL_LED_STATUS_PORT_RECEIVING
+ (port->first_lane))) ?
+ FM10K_SW_EXT_PORT_LED_FLAG_ACTIVE : 0);
+ update_port[i] = (led_flags != port->last_led_flags);
+ port->last_led_flags = led_flags;
+ }
+ FM10K_SW_SWITCH_UNLOCK(sw);
+
+ switch (FM10K_SW_CARD_ID(cfg->subvendor, cfg->subdevice)) {
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QS41):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QS41):
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QS43):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QS43):
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QL4):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * Set up the first PCA9545 mux so we can get at the PCA9635
+ * that the LED control lines are connected to.
+ */
+ fm10k_i2c_write8(sw->i2c, 0x70, 0x04);
+
+ if (!(update_port[0] || update_port[1])) {
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+ }
+
+ led_flags =
+ fm10k_switch_pca9635_led_bits
+ (ports->ports[0].last_led_flags) |
+ (fm10k_switch_pca9635_led_bits
+ (ports->ports[1].last_led_flags) << 2);
+
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x14, led_flags);
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4_REV_2):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * Set up the first PCA9545 mux so we can get at the PCA9635s
+ * that the LED control lines are connected to.
+ */
+ fm10k_i2c_write8(sw->i2c, 0x70, 0x01);
+
+ /* will need to update for QSFPs operating
+ * as four independent lanes/ports
+ */
+ for (i = 0; i < 2; i++) {
+ if (update_port[i] == 0)
+ continue;
+
+ addr = (i == 0) ? 0x6a : 0x69;
+
+ port = &ports->ports[i];
+ led_flags =
+ fm10k_switch_pca9635_led_bits
+ (port->last_led_flags);
+
+ switch (port->lane_speed * port->num_lanes) {
+ case 100:
+ fm10k_i2c_write16(sw->i2c,
+ addr, 0x14, led_flags);
+ break;
+ case 40:
+ fm10k_i2c_write16(sw->i2c,
+ addr, 0x14, led_flags << 6);
+ break;
+ }
+ }
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRL_QXSL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ led_flags = 0;
+ addr = 0x62;
+ for (i = 0; i < 2; i++) {
+ if (!update_port[i])
+ continue;
+ port = &ports->ports[i];
+ if (port->last_led_flags &
+ FM10K_SW_EXT_PORT_LED_FLAG_UP) {
+ switch (port->lane_speed * port->num_lanes) {
+ case 100:
+ case 25:
+ led_flags |= 0x6 << (4 * i); /* 100G */
+ break;
+ case 40:
+ case 10:
+ led_flags |= 0x4 << (4 * i); /* 40G */
+ break;
+ default:
+ led_flags = 0;
+ }
+ } else {
+ led_flags = 0; /* off */
+ }
+ }
+
+ if (update_port[0] || update_port[1]) {
+ fm10k_i2c_read8_ext(sw->i2c, addr, 0x1, &read);
+ if (update_port[0])
+ led_flags |= read & 0xf0;
+ else
+ led_flags |= read & 0xf;
+ fm10k_i2c_write16(sw->i2c, addr, 0x1, led_flags);
+ fm10k_i2c_read8_ext(sw->i2c, addr, 0x1, &read);
+ }
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRM_QXSL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /* will need to update for QSFPs
+ * operating as four independent lanes/ports
+ */
+ for (i = 0; i < 2; i++) {
+ if (!update_port[i])
+ continue;
+ /*
+ * 00000000: No link
+ * 00000001: 10G Port 0
+ * 00000010: 40G Port 0
+ * 00000100: 25G Port 0
+ * 00001000: 100G Port 0
+ */
+ addr = (i == 0) ? 0x62 : 0x66;
+ port = &ports->ports[i];
+ if (port->last_led_flags &
+ FM10K_SW_EXT_PORT_LED_FLAG_UP) {
+ switch (port->lane_speed * port->num_lanes) {
+ case 100:
+ led_flags = 0x08; /* 100G */
+ break;
+ case 40:
+ led_flags = 0x02; /* 40G */
+ break;
+ }
+ } else {
+ led_flags = 0; /* off */
+ }
+ fm10k_i2c_write16(sw->i2c,
+ addr, 0x1, led_flags);
+ }
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+ }
+}
+
+static void *
+fm10k_switch_leds_update(void *ctx)
+{
+ struct fm10k_switch *sw = ctx;
+
+ while (sw->detaching == 0) {
+ fm10k_switch_process_leds(ctx);
+ usec_delay(FM10K_LED_POLL_INTERVAL_MS * 1000);
+ }
+ return NULL;
+}
+
+static void
+fm10k_switch_leds_off(struct fm10k_switch *sw, struct fm10k_ext_ports *ports)
+{
+ struct fm10k_device_info *cfg = sw->info;
+ struct fm10k_ext_port *port;
+ unsigned int i;
+ uint8_t led_flags;
+ uint8_t addr;
+
+ switch (FM10K_SW_CARD_ID(cfg->subvendor, cfg->subdevice)) {
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QS41):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QS41):
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QS43):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QS43):
+ case FM10K_SW_CARD(SILICOM, PE340G2DBIR_QL4):
+ case FM10K_SW_CARD(SILICOM_RB, PE340G2DBIR_QL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * Set up the first PCA9545 mux so we can get at the PCA9635
+ * that the LED control lines are connected to.
+ */
+ fm10k_i2c_write8(sw->i2c, 0x70, 0x04);
+
+ led_flags =
+ fm10k_switch_pca9635_led_bits(0) |
+ (fm10k_switch_pca9635_led_bits(0) << 2);
+
+ fm10k_i2c_write16(sw->i2c, 0x6a, 0x14, led_flags);
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIR_QXSL4_REV_2):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRL_QXSL4):
+ case FM10K_SW_CARD(SILICOM, PE3100G2DQIRM_QXSL4):
+ FM10K_SW_I2C_REQ_LOCK(sw->i2c);
+ /*
+ * Set up the first PCA9545 mux so we can get at the PCA9635s
+ * that the LED control lines are connected to.
+ */
+ fm10k_i2c_write8(sw->i2c, 0x70, 0x01);
+
+ /* will need to update for QSFPs
+ * operating as four independent lanes/ports
+ */
+ addr = 0x6a;
+ led_flags = fm10k_switch_pca9635_led_bits(0);
+ for (i = 0; i < 2; i++) {
+ port = &ports->ports[i];
+ switch (port->lane_speed * port->num_lanes) {
+ case 100:
+ fm10k_i2c_write16(sw->i2c,
+ addr, 0x14, led_flags);
+ break;
+ case 40:
+ fm10k_i2c_write16(sw->i2c,
+ addr, 0x14, led_flags << 6);
+ break;
+ }
+
+ addr = 0x69;
+ }
+ FM10K_SW_I2C_REQ_UNLOCK(sw->i2c);
+ break;
+ }
+}
+
+static void *
+fm10k_switch_process_intr(void *ctx)
+{
+ struct fm10k_switch *sw = ctx;
+ struct fm10k_ext_ports *ports;
+ uint64_t gid;
+ uint64_t reenable_mask;
+ uint16_t epl_mask;
+
+ while (sw->detaching == 0) {
+ sem_wait(&sw->intr_tq);
+ /*
+ * Mask off all global interrupt detect interrupts
+ * toward PCIe to ensure that the PEP sees the
+ * interrupt condition clear so that all subsequent
+ * events will be recognized. The same result could
+ * be achieved by masking off all block-internal
+ * interrupt sources in all blocks prior to handling
+ * any global interrupt detect interrupts, but the
+ * approach taken here makes the interrupt processing
+ * easier to decompose and increases the likelihood that
+ * more work will be coalesced into fewer interrupts.
+ */
+ FM10K_SW_TRACE("masking off PCIe gid interrupts");
+ FM10K_SW_SWITCH_LOCK(sw);
+ /*
+ * sw->ext_ports will be set to NULL during detach
+ * to indicate that we are to no longer process EPL
+ * interrupts.
+ */
+ ports = sw->ext_ports;
+ /*
+ * sw->sm_mailbox_enabled will be set to 0 during
+ * detach to indicate that we are to no longer process
+ * PEP interrupts.
+ */
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_INTERRUPT_MASK_PCIE(sw->pepno),
+ FM10K_SW_INTERRUPT_MASK_PCIE_ALL);
+ gid = fm10k_read_switch_reg64(sw,
+ FM10K_SW_GLOBAL_INTERRUPT_DETECT);
+ FM10K_SW_SWITCH_UNLOCK(sw);
+
+ reenable_mask = 0;
+ if (ports) {
+ epl_mask = fm10k_ext_ports_epl_intrs(ports, gid);
+ reenable_mask |=
+ FM10K_SW_MAKE_REG_FIELD64
+ (INTERRUPT_MASK_PCIE_EPL, epl_mask);
+ }
+
+ if (gid & FM10K_SW_GLOBAL_INTERRUPT_DETECT_I2C)
+ fm10k_i2c_intr(sw->i2c);
+
+ reenable_mask |= FM10K_SW_INTERRUPT_MASK_PCIE_I2C;
+
+ FM10K_SW_TRACE("re-enabling PCIe gid interrupts");
+ FM10K_SW_SWITCH_LOCK(sw);
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_INTERRUPT_MASK_PCIE(sw->pepno),
+ ~reenable_mask);
+ FM10K_SW_SWITCH_UNLOCK(sw);
+ }
+ return NULL;
+}
+
+void
+fm10k_switch_intr(struct fm10k_hw *hw)
+{
+ if (hw)
+ sem_post(&fm10k_sw.intr_tq);
+}
+
+static void
+fm10k_switch_enable_interrupts(struct fm10k_switch *sw)
+{
+ uint64_t data64;
+ unsigned int i;
+
+ data64 = fm10k_read_switch_reg64(sw,
+ FM10K_SW_INTERRUPT_MASK_PCIE(sw->pepno));
+
+ /* enable interrupts from all EPLs in use */
+ FM10K_SW_REPLACE_REG_FIELD64(data64, INTERRUPT_MASK_PCIE_EPL,
+ ~sw->ext_ports->epl_mask, data64);
+
+ /* enable interrupts from all non-master PEPs in use */
+ FM10K_SW_REPLACE_REG_FIELD64(data64, INTERRUPT_MASK_PCIE_PCIE,
+ ~sw->pep_mask, data64);
+
+ /* enable I2C interrupt */
+ data64 &= ~FM10K_SW_INTERRUPT_MASK_PCIE_I2C;
+ fm10k_write_switch_reg64(sw,
+ FM10K_SW_INTERRUPT_MASK_PCIE(sw->pepno), data64);
+
+ /* enable outbound mailbox interrupts from all non-master PEPs */
+ for (i = 0; i < FM10K_SW_PEPS_MAX; i++) {
+ if (sw->pep_mask & (1 << i))
+ fm10k_write_switch_reg(sw, FM10K_SW_PCIE_GLOBAL(i, IM),
+ ~FM10K_SW_IM_MAILBOX);
+ }
+ FM10K_WRITE_REG(sw->master_hw,
+ FM10K_SW_EIMR, FM10K_SW_EIMR_FIELD(SWITCH_INT, ENABLE));
+}
+
+static void
+fm10k_switch_detach(struct fm10k_hw *hw)
+{
+ struct fm10k_ext_ports *ports;
+ struct fm10k_switch *sw = &fm10k_sw;
+
+ if (hw == NULL || sw == NULL)
+ return;
+
+ FM10K_SW_SWITCH_LOCK(sw);
+ sw->detaching = 1;
+ FM10K_SW_SWITCH_UNLOCK(sw);
+
+ if (sw->ext_ports) {
+ ports = sw->ext_ports;
+ /*
+ * Ensure that any further switch interrupts will not
+ * process sw->ext_ports before detaching them.
+ */
+ FM10K_SW_SWITCH_LOCK(sw);
+ sw->ext_ports = NULL;
+ FM10K_SW_SWITCH_UNLOCK(sw);
+ fm10k_switch_leds_off(sw, ports);
+ fm10k_ext_ports_detach(ports);
+ }
+
+ if (sw->epl_sbus)
+ fm10k_sbus_detach(sw->epl_sbus);
+
+ /*
+ * Detach i2c after ext_ports so ext_ports can use i2c to disable
+ * phys.
+ */
+ if (sw->i2c)
+ fm10k_i2c_detach(sw->i2c);
+
+ fm10k_hw_eicr_disable_source(sw, FM10K_SW_EICR_SWITCH_INT);
+
+ pthread_join(sw->intr_task, NULL);
+ pthread_join(sw->led_task, NULL);
+ pthread_join(sw->stats_task, NULL);
+}
+
+static unsigned int
+fm10k_switch_get_fabric_clock(struct fm10k_switch *sw)
+{
+ uint32_t pll_fabric;
+ unsigned int freq;
+
+ pll_fabric = fm10k_read_switch_reg(sw, FM10K_SW_PLL_FABRIC_LOCK);
+
+ switch (FM10K_SW_REG_FIELD(pll_fabric, PLL_FABRIC_FREQSEL)) {
+ case FM10K_SW_PLL_FABRIC_FREQSEL_CTRL: {
+ uint32_t ctrl;
+ unsigned int refdiv, fbdiv4, fbdiv255;
+
+ ctrl = fm10k_read_switch_reg(sw, FM10K_SW_PLL_FABRIC_CTRL);
+ refdiv = FM10K_SW_REG_FIELD(ctrl, PLL_FABRIC_REFDIV);
+ fbdiv4 = FM10K_SW_REG_FIELD(ctrl, PLL_FABRIC_FBDIV4);
+ fbdiv255 = FM10K_SW_REG_FIELD(ctrl, PLL_FABRIC_FBDIV255);
+
+ freq = (15625 * 4 * fbdiv255 * (1 + fbdiv4)) / (refdiv * 100);
+ break;
+ }
+ case FM10K_SW_PLL_FABRIC_FREQSEL_F600:
+ freq = 600;
+ break;
+ case FM10K_SW_PLL_FABRIC_FREQSEL_F500:
+ freq = 500;
+ break;
+ case FM10K_SW_PLL_FABRIC_FREQSEL_F400:
+ freq = 400;
+ break;
+ case FM10K_SW_PLL_FABRIC_FREQSEL_F300:
+ freq = 300;
+ break;
+ default:
+ freq = 0;
+ break;
+ }
+
+ return freq;
+}
+
+static unsigned int
+fm10k_switch_get_sku(struct fm10k_switch *sw)
+{
+ uint32_t fuse_data;
+
+ fuse_data = fm10k_read_switch_reg(sw, FM10K_SW_FUSE_DATA_0);
+ return FM10K_SW_REG_FIELD(fuse_data, FUSE_SKU);
+}
+
+static const char *
+fm10k_switch_get_sku_name(unsigned int sku)
+{
+ const char *name;
+
+ switch (sku) {
+ case FM10K_SW_FUSE_SKU_FM10840:
+ name = "FM10840";
+ break;
+ case FM10K_SW_FUSE_SKU_FM10420:
+ name = "FM10420";
+ break;
+ case FM10K_SW_FUSE_SKU_FM10064:
+ name = "FM10064";
+ break;
+ default:
+ name = "FM10xxx-unknown";
+ break;
+ }
+
+ return name;
+}
+
+static void
+fm10k_switch_describe(struct fm10k_switch *sw)
+{
+ unsigned int i, pair;
+ unsigned int reset_state, active_state;
+ uint32_t device_cfg, resets;
+
+ if (!fm10k_config_check_debug(sw->dpdk_cfg, FM10K_CONFIG_DEBUG_CONFIG))
+ return;
+
+ FM10K_SW_INFO("switch: device is %s, fabric clock %u MHz",
+ fm10k_switch_get_sku_name(fm10k_switch_get_sku(sw)),
+ fm10k_switch_get_fabric_clock(sw));
+
+ device_cfg = fm10k_read_switch_reg(sw, FM10K_SW_DEVICE_CFG);
+ FM10K_SW_INFO("switch: 100G Ethernet is %s",
+ (device_cfg & FM10K_SW_DEVICE_CFG_PCIE_100G_DIS) ?
+ "disabled" : "enabled");
+ switch (FM10K_SW_REG_FIELD(device_cfg, DEVICE_CFG_FEATURE)) {
+ case FM10K_SW_DEVICE_CFG_PCIE_FULL:
+ FM10K_SW_INFO("switch: PEPs 0-7 support 2x4 and 1x8 "
+ "modes, PEP 8 is x1");
+ break;
+ case FM10K_SW_DEVICE_CFG_PCIE_HALF:
+ FM10K_SW_INFO("switch: PEPs 0-3 support 2x4 and 1x8 "
+ "modes, PEPs 4 and 6 are 1x4, PEP 8 is x1");
+ break;
+ case FM10K_SW_DEVICE_CFG_PCIE_BASIC:
+ FM10K_SW_INFO("switch: PEP 0 supports, PEP 8 is x1, "
+ "only one can be used");
+ break;
+ }
+
+ resets = fm10k_read_switch_reg(sw, FM10K_SW_SOFT_RESET);
+ for (i = 0; i < FM10K_SW_PEPS_MAX; i++) {
+ if (!(device_cfg & FM10K_SW_DEVICE_CFG_PCIE_EN(i)))
+ continue;
+
+ pair = i / 2;
+ reset_state = !!(resets & FM10K_SW_SOFT_RESET_PCIE_RESET(i));
+ active_state = !!(resets & FM10K_SW_SOFT_RESET_PCIE_ACTIVE(i));
+ if (pair < FM10K_SW_DEVICE_CFG_PCIE_MODE_PAIRS &&
+ !(device_cfg & FM10K_SW_DEVICE_CFG_PCIE_MODE_2X4(pair))) {
+ if (i % 2 == 0)
+ FM10K_SW_INFO("switch: PEP[%u,%u] is enabled in 1x8 "
+ "mode (Reset=%u, Active=%u)", i, i + 1,
+ reset_state, active_state);
+ else
+ FM10K_SW_INFO("switch: PEP[%u] unexpectedly enabled when "
+ "PEP[%u,%u] is in 1x8 mode (Reset=%u, "
+ "Active=%u)", i, i - 1, i, reset_state,
+ active_state);
+ } else {
+ FM10K_SW_INFO("switch: PEP[%u] is enabled in 1x4 "
+ "mode (Reset=%u, Active=%u)", i, reset_state,
+ active_state);
+ }
+ }
+}
+
+static struct fm10k_switch *
+fm10k_switch_attach(struct fm10k_hw *hw, struct fm10k_switch *sw)
+{
+ int i;
+ int error = 0;
+
+ /*
+ * apparently no way to determine one's own PEP number.
+ */
+ switch (sw->info->num_peps) {
+ case 1:
+ sw->pepno = 0;
+ break;
+
+ case 2:
+ /*
+ * assumption is using even numbered PEPs
+ * starting with 0, and the highest numbered PEP
+ * is the master
+ */
+ sw->pepno = 2;
+ for (i = 0; i < sw->info->num_peps - 1; i++)
+ sw->pep_mask |= (1 << (i * 2));
+ break;
+
+ case 4:
+ sw->pepno = 6;
+ sw->pep_mask = 0x7f;
+ break;
+
+ default:
+ sw->pepno = 0;
+ }
+
+ /* When not zero, initialize serdes in 'near loopback' mode */
+ sw->serdes_loopback = 0;
+
+ pthread_mutex_init(&sw->lock, NULL);
+
+ fm10k_switch_determine_epls(sw);
+
+ sw->ext_ports = fm10k_ext_ports_attach(sw);
+ if (sw->ext_ports == NULL)
+ goto fail;
+
+ /* label all external ports with their logical port numbers */
+ for (i = 0; i < sw->ext_ports->num_ports; i++)
+ sw->ext_ports->ports[i].lport = sw->epl_map[i].logical_port;
+
+ /* interrupt */
+ sem_init(&sw->intr_tq, 0, 0);
+ pthread_create(&sw->intr_task, NULL, fm10k_switch_process_intr, sw);
+
+ fm10k_switch_enable_interrupts(sw);
+ fm10k_switch_describe(sw);
+
+ sw->i2c = fm10k_i2c_attach(sw);
+ if (sw->i2c == NULL)
+ goto fail;
+
+ error = fm10k_switch_init(hw, sw);
+ if (error)
+ goto fail;
+
+ for (i = 0; i < sw->ext_ports->num_ports; i++)
+ fm10k_ext_port_up(&sw->ext_ports->ports[i]);
+
+ usec_delay(10000);
+ fm10k_switch_leds_init(sw);
+
+ pthread_create(&sw->led_task, NULL, &fm10k_switch_leds_update, sw);
+ pthread_create(&sw->stats_task, NULL, &fm10k_switch_process_stats, sw);
+ sw->inited = 1;
+ return sw;
+fail:
+ if (sw != NULL)
+ fm10k_switch_detach(hw);
+ return NULL;
+}
+
+struct fm10k_switch*
+fm10k_switch_get(void)
+{
+ return &fm10k_sw;
+}
+
+static int
+fm10k_switch_dpdk_port_get(struct fm10k_switch *sw, int pf_no)
+{
+ int i;
+
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ if (sw->dpdk_cfg->ports[i].pf_no == pf_no)
+ return i;
+ }
+ return -1;
+}
+
+static int
+fm10k_switch_mapped_ext_port_get(struct fm10k_switch *sw, int pf_no)
+{
+ int i;
+ struct fm10k_cfg_port_pf_map *map;
+
+ for (i = 0; i < FM10K_SW_EXT_PORTS_MAX; i++) {
+ map = &sw->dpdk_cfg->ext_port_map[i];
+ if (map->type == FM10K_CONFIG_PORT_MAP_PF) {
+ if (map->map_no[0] == pf_no)
+ return i;
+ } else if (map->type == FM10K_CONFIG_PORT_MAP_PFS) {
+ if (map->map_no[0] == pf_no || map->map_no[1] == pf_no)
+ return i;
+ }
+ }
+ return -1;
+}
+
+static int
+fm10k_switch_start(struct fm10k_switch *sw,
+ struct fm10k_hw *master_hw, eth_fm10k_dev_init_half_func *func)
+{
+ int i, j;
+ struct fm10k_dpdk_port *port;
+ struct fm10k_cfg_port_pf_map *map;
+ struct fm10k_cfg_port_pf_map *dpdk_map;
+ int dpdk_port_no, ext_port_no = 0;
+ int pf_no;
+
+ sw->info = fm10k_get_device_info(master_hw);
+ sw->master_hw = master_hw;
+ sw->pep_map = fm10k_pep_port_map;
+ sw->epl_map = fm10k_epl_port_map;
+
+ sw->info->ext_port_speed = sw->dpdk_cfg->ext_port_speed;
+ fm10k_switch_set_sched_prog();
+ fm10k_switch_attach(master_hw, &fm10k_sw);
+
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ port = &sw->dpdk_cfg->ports[i];
+ map = &sw->dpdk_cfg->dpdk_port_map[i];
+ if (port->type != FM10K_CONFIG_DPDK_PF ||
+ map->type != FM10K_CONFIG_PORT_MAP_PFS)
+ continue;
+
+ for (j = 0; j < 2; j++) {
+ pf_no = map->map_no[j];
+ dpdk_port_no =
+ fm10k_switch_dpdk_port_get(sw, pf_no);
+ dpdk_map =
+ &sw->dpdk_cfg->dpdk_port_map[dpdk_port_no];
+ if (map->type == FM10K_CONFIG_PORT_MAP_PF) {
+ dpdk_map->type = FM10K_CONFIG_PORT_MAP_PFSS;
+ dpdk_map->map_no[0] =
+ sw->dpdk_cfg->dpdk_port_map[i].map_no[0];
+ dpdk_map->map_no[1] =
+ sw->dpdk_cfg->dpdk_port_map[i].map_no[1];
+ }
+ }
+ }
+
+ /* Make sure Switch Manager is ready before going forward. */
+ for (i = 0; i < FM10K_QUERY_SWITCH_STATE_TIMES; i++) {
+ if (master_hw->switch_ready)
+ break;
+ /* Delay some time to acquire async LPORT_MAP info. */
+ rte_delay_us(FM10K_WAIT_SWITCH_MSG_US);
+ }
+
+ if (!master_hw->switch_ready) {
+ PMD_INIT_LOG(ERR, "switch is not ready");
+ return -1;
+ }
+
+ /* do initialize all ports, after switch is ready */
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ uint32_t sglort;
+ struct fm10k_dpdk_port *port = &sw->dpdk_cfg->ports[i];
+ struct fm10k_hw *hw = port->hw;
+
+ if (hw == NULL)
+ break;
+
+ map = &sw->dpdk_cfg->ext_port_map[ext_port_no];
+ if (port->type == FM10K_CONFIG_DPDK_PF) {
+ pf_no = fm10k_switch_dpdk_pf_no_get(hw);
+ if (map->type == FM10K_CONFIG_PORT_MAP_PF) {
+ sglort = fm10k_switch_pf_glort_get(pf_no);
+ } else if (map->type == FM10K_CONFIG_PORT_MAP_PFS) {
+ ext_port_no =
+ fm10k_switch_mapped_ext_port_get
+ (sw, pf_no);
+ sglort =
+ fm10k_switch_pfs_glort_get
+ (map->map_no[0], map->map_no[1]);
+ } else {
+ FM10K_SW_ERR("Unknown mapped port type %d!",
+ port->type);
+ return -1;
+ }
+ hw->mac.dglort_map = sglort | 0xffff0000;
+ hw->switch_ready = master_hw->switch_ready;
+ func(hw);
+ }
+ }
+ return 0;
+}
+
+static int
+fm10k_switch_dpdk_port_reg(struct fm10k_switch *sw,
+ struct fm10k_hw *hw, void *rte_dev, uint8_t is_pf, bool master)
+{
+ int i;
+ struct fm10k_dpdk_port *port;
+ struct fm10k_cfg_port_pf_map *map;
+
+ if (sw->dpdk_cfg == NULL && is_pf) {
+ if (fm10k_config_init(sw, hw) < 0)
+ return -1;
+ }
+
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ port = &sw->dpdk_cfg->ports[i];
+ if (master && port->hw &&
+ port->type == FM10K_CONFIG_DPDK_PF) {
+ port->pf_no =
+ sw->dpdk_cfg->pf_max - 1 -
+ (hw->mac.addr[5] - port->hw->mac.addr[5]);
+ sw->dpdk_cfg->pf_hw[port->pf_no] = port->hw;
+ }
+
+ if (port->hw == NULL) {
+ port->hw = hw;
+ port->rte_dev = rte_dev;
+ fm10k_flow_list_init(&port->flow_list);
+ map = &sw->dpdk_cfg->dpdk_port_map[i];
+ if (is_pf) {
+ sw->dpdk_cfg->pf_bind++;
+ port->type = FM10K_CONFIG_DPDK_PF;
+ if (map->type == FM10K_CONFIG_PORT_MAP_NULL) {
+ map->type = FM10K_CONFIG_PORT_MAP_PF;
+ map->map_no[0] = i;
+ }
+ } else {
+ port->type = FM10K_CONFIG_DPDK_VF;
+ }
+ if (master)
+ sw->dpdk_cfg->master_hw = hw;
+ if (sw->dpdk_cfg->master_hw) {
+ port->pf_no =
+ sw->dpdk_cfg->pf_max - 1 -
+ (sw->dpdk_cfg->master_hw->mac.addr[5] -
+ hw->mac.addr[5]);
+ sw->dpdk_cfg->pf_hw[port->pf_no] = port->hw;
+ }
+
+ return 0;
+ }
+ }
+ return -1;
+}
+
+int
+fm10k_switch_dpdk_port_no_get(struct fm10k_hw *hw)
+{
+ int i;
+
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ if (fm10k_sw.dpdk_cfg->ports[i].hw == NULL)
+ break;
+
+ if (fm10k_sw.dpdk_cfg->ports[i].hw == hw)
+ return i;
+ }
+ return -1;
+}
+
+void *
+fm10k_switch_dpdk_port_rte_dev_get(struct fm10k_hw *hw)
+{
+ int port_no = fm10k_switch_dpdk_port_no_get(hw);
+
+ if (port_no < 0)
+ return NULL;
+
+ return fm10k_switch_get()->dpdk_cfg->ports[port_no].rte_dev;
+}
+
+struct fm10k_flow_list *
+fm10k_switch_dpdk_port_flow_list_get(struct fm10k_hw *hw)
+{
+ int port_no = fm10k_switch_dpdk_port_no_get(hw);
+
+ if (port_no < 0)
+ return NULL;
+
+ return fm10k_switch_get()->dpdk_cfg->ports[port_no].flow_list;
+}
+
+static int
+fm10k_switch_dpdk_cfg_check(struct fm10k_switch *sw)
+{
+ int i;
+ bool need_default = true;
+ struct fm10k_dpdk_port *port;
+ struct fm10k_cfg_port_pf_map *map;
+
+ if (sw->dpdk_cfg->master_hw == NULL) {
+ FM10K_SW_ERR("Master PF is not bound!!!");
+ return -1;
+ }
+
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ port = &sw->dpdk_cfg->ports[i];
+ map = &sw->dpdk_cfg->dpdk_port_map[i];
+ if (port->type == FM10K_CONFIG_DPDK_PF &&
+ map->type != FM10K_CONFIG_PORT_MAP_NULL) {
+ need_default = false;
+ break;
+ }
+ }
+
+ if (need_default) {
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ map = &sw->dpdk_cfg->dpdk_port_map[i];
+ if (port->type == FM10K_CONFIG_DPDK_PF) {
+ map->type = FM10K_CONFIG_PORT_MAP_PF;
+ map->map_no[0] = i;
+ }
+ }
+ }
+
+ return 0;
+}
+
+static void
+fm10k_switch_dpdk_cfg_describe(struct fm10k_switch *sw)
+{
+ int i;
+ struct fm10k_cfg_port_pf_map *map;
+
+ if (!fm10k_config_check_debug(sw->dpdk_cfg, FM10K_CONFIG_DEBUG_CONFIG))
+ return;
+
+ FM10K_SW_INFO("--- FM10K DYNAMIC CONFIG ---\n");
+ FM10K_SW_INFO(" PF Bind : %d\n", sw->dpdk_cfg->pf_bind);
+
+ FM10K_SW_INFO("--- PF ---\n");
+ for (i = 0; i < FM10K_SW_PEP_PORTS_MAX; i++) {
+ uint8_t *mac;
+ struct rte_eth_dev *dev;
+ struct rte_pci_device *pdev;
+
+ if (sw->dpdk_cfg->pf_hw[i] == NULL)
+ continue;
+ dev = (struct rte_eth_dev *)
+ fm10k_switch_dpdk_port_rte_dev_get
+ (sw->dpdk_cfg->pf_hw[i]);
+ pdev = RTE_ETH_DEV_TO_PCI(dev);
+ mac = sw->dpdk_cfg->pf_hw[i]->mac.addr;
+ FM10K_SW_INFO(" PF%d : Logical %d Glort %#x "
+ "PCI Addr %02x:%02x.%x "
+ "MAC Addr %02x:%02x:%02x:%02x:%02x:%02x\n",
+ i, fm10k_pep_port_map[i].logical_port,
+ fm10k_switch_pf_glort_get(i), pdev->addr.bus,
+ pdev->addr.devid,
+ pdev->addr.function,
+ mac[0], mac[1], mac[2], mac[3], mac[4], mac[5]);
+ }
+
+ FM10K_SW_INFO("--- DPDK PORT ---\n");
+ for (i = 0; i < FM10K_SW_LOGICAL_PORTS_MAX; i++) {
+ if (sw->dpdk_cfg->ports[i].type == FM10K_CONFIG_DPDK_NULL)
+ break;
+ map = &sw->dpdk_cfg->dpdk_port_map[i];
+ if (map->type == FM10K_CONFIG_PORT_MAP_NULL)
+ break;
+ if (map->type == FM10K_CONFIG_PORT_MAP_PF) {
+ FM10K_SW_INFO(" DPDK PORT%d : Map to PF%d\n",
+ i, map->map_no[0]);
+ } else {
+ FM10K_SW_INFO(" DPDK PORT%d : Map to PF%d&PF%d Glort %#x\n",
+ i,
+ map->map_no[0], map->map_no[1],
+ fm10k_switch_pfs_glort_get
+ (map->map_no[0], map->map_no[1]));
+ }
+ }
+
+ FM10K_SW_INFO("--- EXT PORT ---\n");
+ for (i = 0; i < sw->dpdk_cfg->ext_port_num; i++) {
+ map = &sw->dpdk_cfg->ext_port_map[i];
+ FM10K_SW_INFO(" EXT PORT%d : Logical %d Glort %#x", i,
+ fm10k_epl_port_map[i].logical_port,
+ fm10k_switch_epl_glort_get(i));
+
+ if (map->type == FM10K_CONFIG_PORT_MAP_NULL)
+ FM10K_SW_INFO("\n");
+ else if (map->type == FM10K_CONFIG_PORT_MAP_PF)
+ FM10K_SW_INFO(" Map to PF%d\n", map->map_no[0]);
+ else
+ FM10K_SW_INFO(" Map to PF%d&PF%d Glort %#x\n",
+ map->map_no[0], map->map_no[1],
+ fm10k_switch_pfs_glort_get
+ (map->map_no[0], map->map_no[1]));
+ }
+}
+
+int
+fm10k_switch_dpdk_port_start(struct fm10k_hw *hw, void *rte_dev,
+ uint8_t is_pf, bool master, eth_fm10k_dev_init_half_func *func)
+{
+ int ret;
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (fm10k_switch_dpdk_port_reg(sw, hw, rte_dev, is_pf, master) != 0) {
+ FM10K_SW_ERR("Register ports failed!!!");
+ return -1;
+ }
+
+ /*
+ * After all pfs are started
+ * start switch here
+ */
+ if (fm10k_sw.dpdk_cfg->pf_max != 0 &&
+ fm10k_sw.dpdk_cfg->pf_bind == fm10k_sw.dpdk_cfg->pf_num) {
+ if (fm10k_switch_dpdk_cfg_check(&fm10k_sw) != 0)
+ return -1;
+
+ ret = fm10k_switch_start(&fm10k_sw,
+ fm10k_sw.dpdk_cfg->master_hw, func);
+ fm10k_switch_dpdk_cfg_describe(&fm10k_sw);
+ return ret;
+ }
+ return 0;
+}
+
+void fm10k_switch_dpdk_port_stop(struct fm10k_hw *hw)
+{
+ if (hw)
+ return;
+}
+
+/*
+ * for multi-host, only dpdk port 0 and 1 are real rx/tx packets
+ * so we need init/setup/start dpdk port queue for them.
+ * for dpdk port 2/3, we need init/setup except start.
+ * we map the pf queue to 0/1 dpdk queue first, then map
+ * the other pf queue to 2/3 dpdk queue.
+ */
+int
+fm10k_switch_dpdk_hw_queue_map(struct fm10k_hw *hw,
+ uint16_t queue, uint16_t max_queue,
+ struct fm10k_hw **map_hw, uint16_t *map_queue)
+{
+ int idx, pf_no;
+ int dpdk_port_no = fm10k_switch_dpdk_port_no_get(hw);
+ struct fm10k_dpdk_port *port;
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (dpdk_port_no < 0) {
+ FM10K_SW_ERR("Can not find the dpdk port!!!");
+ return -1;
+ }
+
+ port = &sw->dpdk_cfg->ports[dpdk_port_no];
+
+ if (port->type == FM10K_CONFIG_DPDK_VF) {
+ FM10K_SW_ERR("Not support yet!!!");
+ return -1;
+ }
+
+ if (port->type != FM10K_CONFIG_DPDK_PF) {
+ FM10K_SW_ERR("Can not be here!!!");
+ return -1;
+ }
+
+ if (sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].type ==
+ FM10K_CONFIG_PORT_MAP_PF) {
+ pf_no = sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].map_no[0];
+ *map_hw = sw->dpdk_cfg->pf_hw[pf_no];
+ *map_queue = queue;
+ return 1;
+ } else if (sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].type ==
+ FM10K_CONFIG_PORT_MAP_PFS) {
+ idx = queue % 2;
+ pf_no = sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].map_no[idx];
+ *map_hw = sw->dpdk_cfg->pf_hw[pf_no];
+ *map_queue = queue / 2;
+ return 1;
+ } else if (sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].type ==
+ FM10K_CONFIG_PORT_MAP_PFSS) {
+ idx = queue % 2;
+ pf_no = sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].map_no[idx];
+ *map_hw = sw->dpdk_cfg->pf_hw[pf_no];
+ *map_queue = (max_queue + 1) / 2 + queue / 2;
+ return 0;
+ } else if (sw->dpdk_cfg->dpdk_port_map[dpdk_port_no].type ==
+ FM10K_CONFIG_PORT_MAP_NULL) {
+ FM10K_SW_ERR("Unmapped dpdk port %d!!!", dpdk_port_no);
+ return -1;
+ }
+ FM10K_SW_ERR("Unknown mapped type!!!");
+ return -1;
+}
+
+int
+fm10k_switch_dpdk_mapped_hw_get(struct fm10k_hw *hw, struct fm10k_hw *hw_list[])
+{
+ int pf_no, pf_no_ext;
+ int dpdk_port_no = fm10k_switch_dpdk_port_no_get(hw);
+ struct fm10k_dpdk_port *port;
+ struct fm10k_cfg_port_pf_map *map;
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (dpdk_port_no < 0) {
+ FM10K_SW_ERR("Can not find the dpdk port!!!");
+ return -1;
+ }
+
+ port = &sw->dpdk_cfg->ports[dpdk_port_no];
+ map = &sw->dpdk_cfg->dpdk_port_map[dpdk_port_no];
+
+ if (port->type == FM10K_CONFIG_DPDK_VF) {
+ hw_list[0] = hw;
+ hw_list[1] = NULL;
+ return 1;
+ }
+
+ if (port->type != FM10K_CONFIG_DPDK_PF) {
+ FM10K_SW_ERR("Can not be here!!!");
+ return -1;
+ } else if (map->type == FM10K_CONFIG_PORT_MAP_PF) {
+ pf_no = map->map_no[0];
+ hw_list[0] = sw->dpdk_cfg->pf_hw[pf_no];
+ hw_list[1] = NULL;
+ return 1;
+ } else if (map->type == FM10K_CONFIG_PORT_MAP_PFS) {
+ pf_no = map->map_no[0];
+ hw_list[0] = sw->dpdk_cfg->pf_hw[pf_no];
+ pf_no_ext = map->map_no[1];
+ hw_list[1] = sw->dpdk_cfg->pf_hw[pf_no_ext];
+ return 2;
+ } else if (map->type == FM10K_CONFIG_PORT_MAP_PFSS) {
+ pf_no = map->map_no[0];
+ hw_list[0] = sw->dpdk_cfg->pf_hw[pf_no];
+ pf_no_ext = map->map_no[1];
+ hw_list[1] = sw->dpdk_cfg->pf_hw[pf_no_ext];
+ return 0;
+ }
+
+ hw_list[0] = NULL;
+ hw_list[1] = NULL;
+ return 0;
+}
+
+void
+fm10k_switch_dpdk_tx_queue_num_set(struct fm10k_hw *hw, uint8_t num)
+{
+ int port_no = fm10k_switch_dpdk_port_no_get(hw);
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (port_no < 0) {
+ FM10K_SW_ERR("Can not find the dpdk port!!!");
+ return;
+ }
+
+ sw->dpdk_cfg->ports[port_no].tx_queue_num = num;
+}
+
+void
+fm10k_switch_dpdk_rx_queue_num_set(struct fm10k_hw *hw, uint8_t num)
+{
+ int port_no = fm10k_switch_dpdk_port_no_get(hw);
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (port_no < 0) {
+ FM10K_SW_ERR("Can not find the dpdk port!!!");
+ return;
+ }
+
+ sw->dpdk_cfg->ports[port_no].rx_queue_num = num;
+}
+
+int
+fm10k_switch_dpdk_pf_no_get(struct fm10k_hw *hw)
+{
+ int port_no = fm10k_switch_dpdk_port_no_get(hw);
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ if (port_no < 0) {
+ FM10K_SW_ERR("Can not find the dpdk port!!!");
+ return -1;
+ }
+
+ return sw->dpdk_cfg->ports[port_no].pf_no;
+}
+
+
+void
+fm10k_switch_flowset_switchto(const char *name)
+{
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ fm10k_ffu_flowset_switch(sw, name);
+}
+
+void
+fm10k_switch_show_port(void)
+{
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ fm10k_stats_epl_port_print(sw);
+ fm10k_stats_dpdk_port_print(sw);
+}
+
+void
+fm10k_switch_show_ffu(void)
+{
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ fm10k_stats_ffu_count_print(sw);
+}
+
+void
+fm10k_switch_show_bank(void)
+{
+ struct fm10k_switch *sw = fm10k_switch_get();
+
+ fm10k_stats_port_bank_print(sw);
+}
--
1.8.3.1
next prev parent reply other threads:[~2020-04-09 6:28 UTC|newest]
Thread overview: 17+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <1584687199-19198-1-git-send-email-xiaojun.liu@silicom.co.il>
2020-04-09 6:26 ` [dpdk-dev] [PATCH v4 0/5] support fm10k " Xiaojun Liu
2020-04-09 6:26 ` [dpdk-dev] [PATCH v4 1/5] net/fm10k: add basic functions for " Xiaojun Liu
2020-04-17 13:56 ` Ferruh Yigit
2020-04-09 6:26 ` [dpdk-dev] [PATCH v4 2/5] net/fm10k: add epl serdes and port control functions Xiaojun Liu
2020-04-17 13:57 ` Ferruh Yigit
2020-04-09 6:26 ` [dpdk-dev] [PATCH v4 3/5] net/fm10k: add ffu and statistics and config file functions Xiaojun Liu
2020-04-09 6:26 ` Xiaojun Liu [this message]
2020-04-09 6:26 ` [dpdk-dev] [PATCH v4 5/5] net/fm10k: add switch management support Xiaojun Liu
2020-04-20 5:09 ` Ye Xiaolong
2020-04-16 8:33 ` [dpdk-dev] [PATCH v4 0/5] support fm10k switch management Wang, Xiao W
2020-04-17 4:02 ` Ye Xiaolong
2020-04-17 14:18 ` Ferruh Yigit
2020-04-20 4:55 ` Ye Xiaolong
2020-04-20 6:37 ` Xiaojun Liu
2020-04-20 8:14 ` Ye Xiaolong
2020-04-20 8:24 ` Xiaojun Liu
2020-04-20 13:05 ` Ferruh Yigit
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1586413588-5759-5-git-send-email-xiaojun.liu@silicom.co.il \
--to=xiaojun.liu@silicom.co.il \
--cc=dev@dpdk.org \
--cc=jacob.e.keller@intel.com \
--cc=ngai-mint.kwan@intel.com \
--cc=qi.z.zhang@intel.com \
--cc=xiao.w.wang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).