1 /* SPDX-License-Identifier: BSD-3-Clause
2 * Copyright(c) 2014-2021 Broadcom
10 #include <ethdev_driver.h>
11 #include <ethdev_pci.h>
12 #include <rte_malloc.h>
13 #include <rte_cycles.h>
14 #include <rte_alarm.h>
15 #include <rte_kvargs.h>
19 #include "bnxt_filter.h"
20 #include "bnxt_hwrm.h"
22 #include "bnxt_reps.h"
23 #include "bnxt_ring.h"
26 #include "bnxt_stats.h"
29 #include "bnxt_vnic.h"
30 #include "hsi_struct_def_dpdk.h"
31 #include "bnxt_nvm_defs.h"
32 #include "bnxt_tf_common.h"
33 #include "ulp_flow_db.h"
34 #include "rte_pmd_bnxt.h"
36 #define DRV_MODULE_NAME "bnxt"
37 static const char bnxt_version[] =
38 "Broadcom NetXtreme driver " DRV_MODULE_NAME;
41 * The set of PCI devices this driver supports
43 static const struct rte_pci_id bnxt_pci_id_map[] = {
44 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
45 BROADCOM_DEV_ID_STRATUS_NIC_VF1) },
46 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM,
47 BROADCOM_DEV_ID_STRATUS_NIC_VF2) },
48 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_STRATUS_NIC) },
49 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_VF) },
50 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57304_VF) },
51 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_NS2) },
52 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57406_VF) },
53 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57407_MF) },
54 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5741X_VF) },
55 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_5731X_VF) },
56 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_MF) },
57 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412) },
58 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414) },
59 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_RJ45) },
60 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_RJ45) },
61 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57412_MF) },
62 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_RJ45) },
63 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57417_SFP) },
64 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_SFP) },
65 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57317_SFP) },
66 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57414_MF) },
67 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57416_MF) },
68 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802) },
69 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58804) },
70 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58808) },
71 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58802_VF) },
72 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508) },
73 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504) },
74 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502) },
75 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF1) },
76 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57500_VF2) },
77 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF1) },
78 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF1) },
79 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF1) },
80 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57508_MF2) },
81 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57504_MF2) },
82 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_57502_MF2) },
83 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58812) },
84 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58814) },
85 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818) },
86 { RTE_PCI_DEVICE(PCI_VENDOR_ID_BROADCOM, BROADCOM_DEV_ID_58818_VF) },
87 { .vendor_id = 0, /* sentinel */ },
90 #define BNXT_DEVARG_TRUFLOW "host-based-truflow"
91 #define BNXT_DEVARG_FLOW_XSTAT "flow-xstat"
92 #define BNXT_DEVARG_MAX_NUM_KFLOWS "max-num-kflows"
93 #define BNXT_DEVARG_REPRESENTOR "representor"
94 #define BNXT_DEVARG_REP_BASED_PF "rep-based-pf"
95 #define BNXT_DEVARG_REP_IS_PF "rep-is-pf"
96 #define BNXT_DEVARG_REP_Q_R2F "rep-q-r2f"
97 #define BNXT_DEVARG_REP_Q_F2R "rep-q-f2r"
98 #define BNXT_DEVARG_REP_FC_R2F "rep-fc-r2f"
99 #define BNXT_DEVARG_REP_FC_F2R "rep-fc-f2r"
101 static const char *const bnxt_dev_args[] = {
102 BNXT_DEVARG_REPRESENTOR,
104 BNXT_DEVARG_FLOW_XSTAT,
105 BNXT_DEVARG_MAX_NUM_KFLOWS,
106 BNXT_DEVARG_REP_BASED_PF,
107 BNXT_DEVARG_REP_IS_PF,
108 BNXT_DEVARG_REP_Q_R2F,
109 BNXT_DEVARG_REP_Q_F2R,
110 BNXT_DEVARG_REP_FC_R2F,
111 BNXT_DEVARG_REP_FC_F2R,
116 * truflow == false to disable the feature
117 * truflow == true to enable the feature
119 #define BNXT_DEVARG_TRUFLOW_INVALID(truflow) ((truflow) > 1)
122 * flow_xstat == false to disable the feature
123 * flow_xstat == true to enable the feature
125 #define BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat) ((flow_xstat) > 1)
128 * rep_is_pf == false to indicate VF representor
129 * rep_is_pf == true to indicate PF representor
131 #define BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf) ((rep_is_pf) > 1)
134 * rep_based_pf == Physical index of the PF
136 #define BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf) ((rep_based_pf) > 15)
138 * rep_q_r2f == Logical COS Queue index for the rep to endpoint direction
140 #define BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f) ((rep_q_r2f) > 3)
143 * rep_q_f2r == Logical COS Queue index for the endpoint to rep direction
145 #define BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r) ((rep_q_f2r) > 3)
148 * rep_fc_r2f == Flow control for the representor to endpoint direction
150 #define BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f) ((rep_fc_r2f) > 1)
153 * rep_fc_f2r == Flow control for the endpoint to representor direction
155 #define BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r) ((rep_fc_f2r) > 1)
157 int bnxt_cfa_code_dynfield_offset = -1;
160 * max_num_kflows must be >= 32
161 * and must be a power-of-2 supported value
162 * return: 1 -> invalid
165 static int bnxt_devarg_max_num_kflow_invalid(uint16_t max_num_kflows)
167 if (max_num_kflows < 32 || !rte_is_power_of_2(max_num_kflows))
172 static int bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask);
173 static int bnxt_dev_uninit(struct rte_eth_dev *eth_dev);
174 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev);
175 static int bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev);
176 static void bnxt_cancel_fw_health_check(struct bnxt *bp);
177 static int bnxt_restore_vlan_filters(struct bnxt *bp);
178 static void bnxt_dev_recover(void *arg);
179 static void bnxt_free_error_recovery_info(struct bnxt *bp);
180 static void bnxt_free_rep_info(struct bnxt *bp);
182 int is_bnxt_in_error(struct bnxt *bp)
184 if (bp->flags & BNXT_FLAG_FATAL_ERROR)
186 if (bp->flags & BNXT_FLAG_FW_RESET)
192 /***********************/
195 * High level utility functions
198 static uint16_t bnxt_rss_ctxts(const struct bnxt *bp)
200 unsigned int num_rss_rings = RTE_MIN(bp->rx_nr_rings,
201 BNXT_RSS_TBL_SIZE_P5);
203 if (!BNXT_CHIP_P5(bp))
206 return RTE_ALIGN_MUL_CEIL(num_rss_rings,
207 BNXT_RSS_ENTRIES_PER_CTX_P5) /
208 BNXT_RSS_ENTRIES_PER_CTX_P5;
211 uint16_t bnxt_rss_hash_tbl_size(const struct bnxt *bp)
213 if (!BNXT_CHIP_P5(bp))
214 return HW_HASH_INDEX_SIZE;
216 return bnxt_rss_ctxts(bp) * BNXT_RSS_ENTRIES_PER_CTX_P5;
219 static void bnxt_free_parent_info(struct bnxt *bp)
221 rte_free(bp->parent);
225 static void bnxt_free_pf_info(struct bnxt *bp)
231 static void bnxt_free_link_info(struct bnxt *bp)
233 rte_free(bp->link_info);
234 bp->link_info = NULL;
237 static void bnxt_free_leds_info(struct bnxt *bp)
246 static void bnxt_free_flow_stats_info(struct bnxt *bp)
248 rte_free(bp->flow_stat);
249 bp->flow_stat = NULL;
252 static void bnxt_free_cos_queues(struct bnxt *bp)
254 rte_free(bp->rx_cos_queue);
255 bp->rx_cos_queue = NULL;
256 rte_free(bp->tx_cos_queue);
257 bp->tx_cos_queue = NULL;
260 static void bnxt_free_mem(struct bnxt *bp, bool reconfig)
262 bnxt_free_filter_mem(bp);
263 bnxt_free_vnic_attributes(bp);
264 bnxt_free_vnic_mem(bp);
266 /* tx/rx rings are configured as part of *_queue_setup callbacks.
267 * If the number of rings change across fw update,
268 * we don't have much choice except to warn the user.
272 bnxt_free_tx_rings(bp);
273 bnxt_free_rx_rings(bp);
275 bnxt_free_async_cp_ring(bp);
276 bnxt_free_rxtx_nq_ring(bp);
278 rte_free(bp->grp_info);
282 static int bnxt_alloc_parent_info(struct bnxt *bp)
284 bp->parent = rte_zmalloc("bnxt_parent_info",
285 sizeof(struct bnxt_parent_info), 0);
286 if (bp->parent == NULL)
292 static int bnxt_alloc_pf_info(struct bnxt *bp)
294 bp->pf = rte_zmalloc("bnxt_pf_info", sizeof(struct bnxt_pf_info), 0);
301 static int bnxt_alloc_link_info(struct bnxt *bp)
304 rte_zmalloc("bnxt_link_info", sizeof(struct bnxt_link_info), 0);
305 if (bp->link_info == NULL)
311 static int bnxt_alloc_leds_info(struct bnxt *bp)
316 bp->leds = rte_zmalloc("bnxt_leds",
317 BNXT_MAX_LED * sizeof(struct bnxt_led_info),
319 if (bp->leds == NULL)
325 static int bnxt_alloc_cos_queues(struct bnxt *bp)
328 rte_zmalloc("bnxt_rx_cosq",
329 BNXT_COS_QUEUE_COUNT *
330 sizeof(struct bnxt_cos_queue_info),
332 if (bp->rx_cos_queue == NULL)
336 rte_zmalloc("bnxt_tx_cosq",
337 BNXT_COS_QUEUE_COUNT *
338 sizeof(struct bnxt_cos_queue_info),
340 if (bp->tx_cos_queue == NULL)
346 static int bnxt_alloc_flow_stats_info(struct bnxt *bp)
348 bp->flow_stat = rte_zmalloc("bnxt_flow_xstat",
349 sizeof(struct bnxt_flow_stat_info), 0);
350 if (bp->flow_stat == NULL)
356 static int bnxt_alloc_mem(struct bnxt *bp, bool reconfig)
360 rc = bnxt_alloc_ring_grps(bp);
364 rc = bnxt_alloc_async_ring_struct(bp);
368 rc = bnxt_alloc_vnic_mem(bp);
372 rc = bnxt_alloc_vnic_attributes(bp);
376 rc = bnxt_alloc_filter_mem(bp);
380 rc = bnxt_alloc_async_cp_ring(bp);
384 rc = bnxt_alloc_rxtx_nq_ring(bp);
388 if (BNXT_FLOW_XSTATS_EN(bp)) {
389 rc = bnxt_alloc_flow_stats_info(bp);
397 bnxt_free_mem(bp, reconfig);
401 static int bnxt_setup_one_vnic(struct bnxt *bp, uint16_t vnic_id)
403 struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
404 struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
405 uint64_t rx_offloads = dev_conf->rxmode.offloads;
406 struct bnxt_rx_queue *rxq;
410 rc = bnxt_vnic_grp_alloc(bp, vnic);
414 PMD_DRV_LOG(DEBUG, "vnic[%d] = %p vnic->fw_grp_ids = %p\n",
415 vnic_id, vnic, vnic->fw_grp_ids);
417 rc = bnxt_hwrm_vnic_alloc(bp, vnic);
421 /* Alloc RSS context only if RSS mode is enabled */
422 if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS) {
423 int j, nr_ctxs = bnxt_rss_ctxts(bp);
425 /* RSS table size in Thor is 512.
426 * Cap max Rx rings to same value
428 if (bp->rx_nr_rings > BNXT_RSS_TBL_SIZE_P5) {
429 PMD_DRV_LOG(ERR, "RxQ cnt %d > reta_size %d\n",
430 bp->rx_nr_rings, BNXT_RSS_TBL_SIZE_P5);
435 for (j = 0; j < nr_ctxs; j++) {
436 rc = bnxt_hwrm_vnic_ctx_alloc(bp, vnic, j);
442 "HWRM vnic %d ctx %d alloc failure rc: %x\n",
446 vnic->num_lb_ctxts = nr_ctxs;
450 * Firmware sets pf pair in default vnic cfg. If the VLAN strip
451 * setting is not available at this time, it will not be
452 * configured correctly in the CFA.
454 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
455 vnic->vlan_strip = true;
457 vnic->vlan_strip = false;
459 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
463 rc = bnxt_set_hwrm_vnic_filters(bp, vnic);
467 for (j = 0; j < bp->rx_num_qs_per_vnic; j++) {
468 rxq = bp->eth_dev->data->rx_queues[j];
471 "rxq[%d]->vnic=%p vnic->fw_grp_ids=%p\n",
472 j, rxq->vnic, rxq->vnic->fw_grp_ids);
474 if (BNXT_HAS_RING_GRPS(bp) && rxq->rx_deferred_start)
475 rxq->vnic->fw_grp_ids[j] = INVALID_HW_RING_ID;
477 vnic->rx_queue_cnt++;
480 PMD_DRV_LOG(DEBUG, "vnic->rx_queue_cnt = %d\n", vnic->rx_queue_cnt);
482 rc = bnxt_vnic_rss_configure(bp, vnic);
486 bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
488 rc = bnxt_hwrm_vnic_tpa_cfg(bp, vnic,
489 (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) ?
496 PMD_DRV_LOG(ERR, "HWRM vnic %d cfg failure rc: %x\n",
501 static int bnxt_register_fc_ctx_mem(struct bnxt *bp)
505 rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_in_tbl.dma,
506 &bp->flow_stat->rx_fc_in_tbl.ctx_id);
511 "rx_fc_in_tbl.va = %p rx_fc_in_tbl.dma = %p"
512 " rx_fc_in_tbl.ctx_id = %d\n",
513 bp->flow_stat->rx_fc_in_tbl.va,
514 (void *)((uintptr_t)bp->flow_stat->rx_fc_in_tbl.dma),
515 bp->flow_stat->rx_fc_in_tbl.ctx_id);
517 rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->rx_fc_out_tbl.dma,
518 &bp->flow_stat->rx_fc_out_tbl.ctx_id);
523 "rx_fc_out_tbl.va = %p rx_fc_out_tbl.dma = %p"
524 " rx_fc_out_tbl.ctx_id = %d\n",
525 bp->flow_stat->rx_fc_out_tbl.va,
526 (void *)((uintptr_t)bp->flow_stat->rx_fc_out_tbl.dma),
527 bp->flow_stat->rx_fc_out_tbl.ctx_id);
529 rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_in_tbl.dma,
530 &bp->flow_stat->tx_fc_in_tbl.ctx_id);
535 "tx_fc_in_tbl.va = %p tx_fc_in_tbl.dma = %p"
536 " tx_fc_in_tbl.ctx_id = %d\n",
537 bp->flow_stat->tx_fc_in_tbl.va,
538 (void *)((uintptr_t)bp->flow_stat->tx_fc_in_tbl.dma),
539 bp->flow_stat->tx_fc_in_tbl.ctx_id);
541 rc = bnxt_hwrm_ctx_rgtr(bp, bp->flow_stat->tx_fc_out_tbl.dma,
542 &bp->flow_stat->tx_fc_out_tbl.ctx_id);
547 "tx_fc_out_tbl.va = %p tx_fc_out_tbl.dma = %p"
548 " tx_fc_out_tbl.ctx_id = %d\n",
549 bp->flow_stat->tx_fc_out_tbl.va,
550 (void *)((uintptr_t)bp->flow_stat->tx_fc_out_tbl.dma),
551 bp->flow_stat->tx_fc_out_tbl.ctx_id);
553 memset(bp->flow_stat->rx_fc_out_tbl.va,
555 bp->flow_stat->rx_fc_out_tbl.size);
556 rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
557 CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
558 bp->flow_stat->rx_fc_out_tbl.ctx_id,
559 bp->flow_stat->max_fc,
564 memset(bp->flow_stat->tx_fc_out_tbl.va,
566 bp->flow_stat->tx_fc_out_tbl.size);
567 rc = bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
568 CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
569 bp->flow_stat->tx_fc_out_tbl.ctx_id,
570 bp->flow_stat->max_fc,
576 static int bnxt_alloc_ctx_mem_buf(char *type, size_t size,
577 struct bnxt_ctx_mem_buf_info *ctx)
582 ctx->va = rte_zmalloc(type, size, 0);
585 rte_mem_lock_page(ctx->va);
587 ctx->dma = rte_mem_virt2iova(ctx->va);
588 if (ctx->dma == RTE_BAD_IOVA)
594 static int bnxt_init_fc_ctx_mem(struct bnxt *bp)
596 struct rte_pci_device *pdev = bp->pdev;
597 char type[RTE_MEMZONE_NAMESIZE];
601 max_fc = bp->flow_stat->max_fc;
603 sprintf(type, "bnxt_rx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
604 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
605 /* 4 bytes for each counter-id */
606 rc = bnxt_alloc_ctx_mem_buf(type,
608 &bp->flow_stat->rx_fc_in_tbl);
612 sprintf(type, "bnxt_rx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
613 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
614 /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
615 rc = bnxt_alloc_ctx_mem_buf(type,
617 &bp->flow_stat->rx_fc_out_tbl);
621 sprintf(type, "bnxt_tx_fc_in_" PCI_PRI_FMT, pdev->addr.domain,
622 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
623 /* 4 bytes for each counter-id */
624 rc = bnxt_alloc_ctx_mem_buf(type,
626 &bp->flow_stat->tx_fc_in_tbl);
630 sprintf(type, "bnxt_tx_fc_out_" PCI_PRI_FMT, pdev->addr.domain,
631 pdev->addr.bus, pdev->addr.devid, pdev->addr.function);
632 /* 16 bytes for each counter - 8 bytes pkt_count, 8 bytes byte_count */
633 rc = bnxt_alloc_ctx_mem_buf(type,
635 &bp->flow_stat->tx_fc_out_tbl);
639 rc = bnxt_register_fc_ctx_mem(bp);
644 static int bnxt_init_ctx_mem(struct bnxt *bp)
648 if (!(bp->fw_cap & BNXT_FW_CAP_ADV_FLOW_COUNTERS) ||
649 !(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) ||
650 !BNXT_FLOW_XSTATS_EN(bp))
653 rc = bnxt_hwrm_cfa_counter_qcaps(bp, &bp->flow_stat->max_fc);
657 rc = bnxt_init_fc_ctx_mem(bp);
662 static int bnxt_update_phy_setting(struct bnxt *bp)
664 struct rte_eth_link new;
667 rc = bnxt_get_hwrm_link_config(bp, &new);
669 PMD_DRV_LOG(ERR, "Failed to get link settings\n");
674 * On BCM957508-N2100 adapters, FW will not allow any user other
675 * than BMC to shutdown the port. bnxt_get_hwrm_link_config() call
676 * always returns link up. Force phy update always in that case.
678 if (!new.link_status || IS_BNXT_DEV_957508_N2100(bp)) {
679 rc = bnxt_set_hwrm_link_config(bp, true);
681 PMD_DRV_LOG(ERR, "Failed to update PHY settings\n");
689 static int bnxt_start_nic(struct bnxt *bp)
691 struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(bp->eth_dev);
692 struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
693 uint32_t intr_vector = 0;
694 uint32_t queue_id, base = BNXT_MISC_VEC_ID;
695 uint32_t vec = BNXT_MISC_VEC_ID;
699 if (bp->eth_dev->data->mtu > RTE_ETHER_MTU) {
700 bp->eth_dev->data->dev_conf.rxmode.offloads |=
701 DEV_RX_OFFLOAD_JUMBO_FRAME;
702 bp->flags |= BNXT_FLAG_JUMBO;
704 bp->eth_dev->data->dev_conf.rxmode.offloads &=
705 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
706 bp->flags &= ~BNXT_FLAG_JUMBO;
709 /* THOR does not support ring groups.
710 * But we will use the array to save RSS context IDs.
712 if (BNXT_CHIP_P5(bp))
713 bp->max_ring_grps = BNXT_MAX_RSS_CTXTS_P5;
715 rc = bnxt_alloc_all_hwrm_stat_ctxs(bp);
717 PMD_DRV_LOG(ERR, "HWRM stat ctx alloc failure rc: %x\n", rc);
721 rc = bnxt_alloc_hwrm_rings(bp);
723 PMD_DRV_LOG(ERR, "HWRM ring alloc failure rc: %x\n", rc);
727 rc = bnxt_alloc_all_hwrm_ring_grps(bp);
729 PMD_DRV_LOG(ERR, "HWRM ring grp alloc failure: %x\n", rc);
733 if (!(bp->vnic_cap_flags & BNXT_VNIC_CAP_COS_CLASSIFY))
736 for (j = 0, i = 0; i < BNXT_COS_QUEUE_COUNT; i++) {
737 if (bp->rx_cos_queue[i].id != 0xff) {
738 struct bnxt_vnic_info *vnic = &bp->vnic_info[j++];
742 "Num pools more than FW profile\n");
746 vnic->cos_queue_id = bp->rx_cos_queue[i].id;
752 rc = bnxt_mq_rx_configure(bp);
754 PMD_DRV_LOG(ERR, "MQ mode configure failure rc: %x\n", rc);
759 rc = bnxt_setup_one_vnic(bp, 0);
762 /* VNIC configuration */
763 if (BNXT_RFS_NEEDS_VNIC(bp)) {
764 for (i = 1; i < bp->nr_vnics; i++) {
765 rc = bnxt_setup_one_vnic(bp, i);
771 rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, &bp->vnic_info[0], 0, NULL);
774 "HWRM cfa l2 rx mask failure rc: %x\n", rc);
778 /* check and configure queue intr-vector mapping */
779 if ((rte_intr_cap_multiple(intr_handle) ||
780 !RTE_ETH_DEV_SRIOV(bp->eth_dev).active) &&
781 bp->eth_dev->data->dev_conf.intr_conf.rxq != 0) {
782 intr_vector = bp->eth_dev->data->nb_rx_queues;
783 PMD_DRV_LOG(DEBUG, "intr_vector = %d\n", intr_vector);
784 if (intr_vector > bp->rx_cp_nr_rings) {
785 PMD_DRV_LOG(ERR, "At most %d intr queues supported",
789 rc = rte_intr_efd_enable(intr_handle, intr_vector);
794 if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
795 intr_handle->intr_vec =
796 rte_zmalloc("intr_vec",
797 bp->eth_dev->data->nb_rx_queues *
799 if (intr_handle->intr_vec == NULL) {
800 PMD_DRV_LOG(ERR, "Failed to allocate %d rx_queues"
801 " intr_vec", bp->eth_dev->data->nb_rx_queues);
805 PMD_DRV_LOG(DEBUG, "intr_handle->intr_vec = %p "
806 "intr_handle->nb_efd = %d intr_handle->max_intr = %d\n",
807 intr_handle->intr_vec, intr_handle->nb_efd,
808 intr_handle->max_intr);
809 for (queue_id = 0; queue_id < bp->eth_dev->data->nb_rx_queues;
811 intr_handle->intr_vec[queue_id] =
812 vec + BNXT_RX_VEC_START;
813 if (vec < base + intr_handle->nb_efd - 1)
818 /* enable uio/vfio intr/eventfd mapping */
819 rc = rte_intr_enable(intr_handle);
820 #ifndef RTE_EXEC_ENV_FREEBSD
821 /* In FreeBSD OS, nic_uio driver does not support interrupts */
826 rc = bnxt_update_phy_setting(bp);
830 bp->mark_table = rte_zmalloc("bnxt_mark_table", BNXT_MARK_TABLE_SZ, 0);
832 PMD_DRV_LOG(ERR, "Allocation of mark table failed\n");
837 /* Some of the error status returned by FW may not be from errno.h */
844 static int bnxt_shutdown_nic(struct bnxt *bp)
846 bnxt_free_all_hwrm_resources(bp);
847 bnxt_free_all_filters(bp);
848 bnxt_free_all_vnics(bp);
853 * Device configuration and status function
856 uint32_t bnxt_get_speed_capabilities(struct bnxt *bp)
858 uint32_t link_speed = 0;
859 uint32_t speed_capa = 0;
861 if (bp->link_info == NULL)
864 link_speed = bp->link_info->support_speeds;
866 /* If PAM4 is configured, use PAM4 supported speed */
867 if (link_speed == 0 && bp->link_info->support_pam4_speeds > 0)
868 link_speed = bp->link_info->support_pam4_speeds;
870 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_LINK_SPEED_100MB)
871 speed_capa |= ETH_LINK_SPEED_100M;
872 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100MBHD)
873 speed_capa |= ETH_LINK_SPEED_100M_HD;
874 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_1GB)
875 speed_capa |= ETH_LINK_SPEED_1G;
876 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_2_5GB)
877 speed_capa |= ETH_LINK_SPEED_2_5G;
878 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_10GB)
879 speed_capa |= ETH_LINK_SPEED_10G;
880 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_20GB)
881 speed_capa |= ETH_LINK_SPEED_20G;
882 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_25GB)
883 speed_capa |= ETH_LINK_SPEED_25G;
884 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_40GB)
885 speed_capa |= ETH_LINK_SPEED_40G;
886 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_50GB)
887 speed_capa |= ETH_LINK_SPEED_50G;
888 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_SPEEDS_100GB)
889 speed_capa |= ETH_LINK_SPEED_100G;
890 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_50G)
891 speed_capa |= ETH_LINK_SPEED_50G;
892 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_100G)
893 speed_capa |= ETH_LINK_SPEED_100G;
894 if (link_speed & HWRM_PORT_PHY_QCFG_OUTPUT_SUPPORT_PAM4_SPEEDS_200G)
895 speed_capa |= ETH_LINK_SPEED_200G;
897 if (bp->link_info->auto_mode ==
898 HWRM_PORT_PHY_QCFG_OUTPUT_AUTO_MODE_NONE)
899 speed_capa |= ETH_LINK_SPEED_FIXED;
904 static int bnxt_dev_info_get_op(struct rte_eth_dev *eth_dev,
905 struct rte_eth_dev_info *dev_info)
907 struct rte_pci_device *pdev = RTE_DEV_TO_PCI(eth_dev->device);
908 struct bnxt *bp = eth_dev->data->dev_private;
909 uint16_t max_vnics, i, j, vpool, vrxq;
910 unsigned int max_rx_rings;
913 rc = is_bnxt_in_error(bp);
918 dev_info->max_mac_addrs = bp->max_l2_ctx;
919 dev_info->max_hash_mac_addrs = 0;
921 /* PF/VF specifics */
923 dev_info->max_vfs = pdev->max_vfs;
925 max_rx_rings = bnxt_max_rings(bp);
926 /* For the sake of symmetry, max_rx_queues = max_tx_queues */
927 dev_info->max_rx_queues = max_rx_rings;
928 dev_info->max_tx_queues = max_rx_rings;
929 dev_info->reta_size = bnxt_rss_hash_tbl_size(bp);
930 dev_info->hash_key_size = 40;
931 max_vnics = bp->max_vnics;
934 dev_info->min_mtu = RTE_ETHER_MIN_MTU;
935 dev_info->max_mtu = BNXT_MAX_MTU;
937 /* Fast path specifics */
938 dev_info->min_rx_bufsize = 1;
939 dev_info->max_rx_pktlen = BNXT_MAX_PKT_LEN;
941 dev_info->rx_offload_capa = BNXT_DEV_RX_OFFLOAD_SUPPORT;
942 if (bp->flags & BNXT_FLAG_PTP_SUPPORTED)
943 dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_TIMESTAMP;
944 dev_info->tx_queue_offload_capa = DEV_TX_OFFLOAD_MBUF_FAST_FREE;
945 dev_info->tx_offload_capa = BNXT_DEV_TX_OFFLOAD_SUPPORT |
946 dev_info->tx_queue_offload_capa;
947 dev_info->flow_type_rss_offloads = BNXT_ETH_RSS_SUPPORT;
949 dev_info->speed_capa = bnxt_get_speed_capabilities(bp);
952 dev_info->default_rxconf = (struct rte_eth_rxconf) {
958 .rx_free_thresh = 32,
959 .rx_drop_en = BNXT_DEFAULT_RX_DROP_EN,
962 dev_info->default_txconf = (struct rte_eth_txconf) {
968 .tx_free_thresh = 32,
971 eth_dev->data->dev_conf.intr_conf.lsc = 1;
973 eth_dev->data->dev_conf.intr_conf.rxq = 1;
974 dev_info->rx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
975 dev_info->rx_desc_lim.nb_max = BNXT_MAX_RX_RING_DESC;
976 dev_info->tx_desc_lim.nb_min = BNXT_MIN_RING_DESC;
977 dev_info->tx_desc_lim.nb_max = BNXT_MAX_TX_RING_DESC;
979 if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
980 dev_info->switch_info.name = eth_dev->device->name;
981 dev_info->switch_info.domain_id = bp->switch_domain_id;
982 dev_info->switch_info.port_id =
983 BNXT_PF(bp) ? BNXT_SWITCH_PORT_ID_PF :
984 BNXT_SWITCH_PORT_ID_TRUSTED_VF;
990 * TODO: default_rxconf, default_txconf, rx_desc_lim, and tx_desc_lim
991 * need further investigation.
995 vpool = 64; /* ETH_64_POOLS */
996 vrxq = 128; /* ETH_VMDQ_DCB_NUM_QUEUES */
997 for (i = 0; i < 4; vpool >>= 1, i++) {
998 if (max_vnics > vpool) {
999 for (j = 0; j < 5; vrxq >>= 1, j++) {
1000 if (dev_info->max_rx_queues > vrxq) {
1006 /* Not enough resources to support VMDq */
1010 /* Not enough resources to support VMDq */
1014 dev_info->max_vmdq_pools = vpool;
1015 dev_info->vmdq_queue_num = vrxq;
1017 dev_info->vmdq_pool_base = 0;
1018 dev_info->vmdq_queue_base = 0;
1023 /* Configure the device based on the configuration provided */
1024 static int bnxt_dev_configure_op(struct rte_eth_dev *eth_dev)
1026 struct bnxt *bp = eth_dev->data->dev_private;
1027 uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1030 bp->rx_queues = (void *)eth_dev->data->rx_queues;
1031 bp->tx_queues = (void *)eth_dev->data->tx_queues;
1032 bp->tx_nr_rings = eth_dev->data->nb_tx_queues;
1033 bp->rx_nr_rings = eth_dev->data->nb_rx_queues;
1035 rc = is_bnxt_in_error(bp);
1039 if (BNXT_VF(bp) && (bp->flags & BNXT_FLAG_NEW_RM)) {
1040 rc = bnxt_hwrm_check_vf_rings(bp);
1042 PMD_DRV_LOG(ERR, "HWRM insufficient resources\n");
1046 /* If a resource has already been allocated - in this case
1047 * it is the async completion ring, free it. Reallocate it after
1048 * resource reservation. This will ensure the resource counts
1049 * are calculated correctly.
1052 pthread_mutex_lock(&bp->def_cp_lock);
1054 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1055 bnxt_disable_int(bp);
1056 bnxt_free_cp_ring(bp, bp->async_cp_ring);
1059 rc = bnxt_hwrm_func_reserve_vf_resc(bp, false);
1061 PMD_DRV_LOG(ERR, "HWRM resource alloc fail:%x\n", rc);
1062 pthread_mutex_unlock(&bp->def_cp_lock);
1066 if (!BNXT_HAS_NQ(bp) && bp->async_cp_ring) {
1067 rc = bnxt_alloc_async_cp_ring(bp);
1069 pthread_mutex_unlock(&bp->def_cp_lock);
1072 bnxt_enable_int(bp);
1075 pthread_mutex_unlock(&bp->def_cp_lock);
1078 /* Inherit new configurations */
1079 if (eth_dev->data->nb_rx_queues > bp->max_rx_rings ||
1080 eth_dev->data->nb_tx_queues > bp->max_tx_rings ||
1081 eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues
1082 + BNXT_NUM_ASYNC_CPR(bp) > bp->max_cp_rings ||
1083 eth_dev->data->nb_rx_queues + eth_dev->data->nb_tx_queues >
1085 goto resource_error;
1087 if (BNXT_HAS_RING_GRPS(bp) &&
1088 (uint32_t)(eth_dev->data->nb_rx_queues) > bp->max_ring_grps)
1089 goto resource_error;
1091 if (!(eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS) &&
1092 bp->max_vnics < eth_dev->data->nb_rx_queues)
1093 goto resource_error;
1095 bp->rx_cp_nr_rings = bp->rx_nr_rings;
1096 bp->tx_cp_nr_rings = bp->tx_nr_rings;
1098 if (eth_dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
1099 rx_offloads |= DEV_RX_OFFLOAD_RSS_HASH;
1100 eth_dev->data->dev_conf.rxmode.offloads = rx_offloads;
1102 if (rx_offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
1103 eth_dev->data->mtu =
1104 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len -
1105 RTE_ETHER_HDR_LEN - RTE_ETHER_CRC_LEN - VLAN_TAG_SIZE *
1107 bnxt_mtu_set_op(eth_dev, eth_dev->data->mtu);
1113 "Insufficient resources to support requested config\n");
1115 "Num Queues Requested: Tx %d, Rx %d\n",
1116 eth_dev->data->nb_tx_queues,
1117 eth_dev->data->nb_rx_queues);
1119 "MAX: TxQ %d, RxQ %d, CQ %d Stat %d, Grp %d, Vnic %d\n",
1120 bp->max_tx_rings, bp->max_rx_rings, bp->max_cp_rings,
1121 bp->max_stat_ctx, bp->max_ring_grps, bp->max_vnics);
1125 void bnxt_print_link_info(struct rte_eth_dev *eth_dev)
1127 struct rte_eth_link *link = ð_dev->data->dev_link;
1129 if (link->link_status)
1130 PMD_DRV_LOG(INFO, "Port %d Link Up - speed %u Mbps - %s\n",
1131 eth_dev->data->port_id,
1132 (uint32_t)link->link_speed,
1133 (link->link_duplex == ETH_LINK_FULL_DUPLEX) ?
1134 ("full-duplex") : ("half-duplex\n"));
1136 PMD_DRV_LOG(INFO, "Port %d Link Down\n",
1137 eth_dev->data->port_id);
1141 * Determine whether the current configuration requires support for scattered
1142 * receive; return 1 if scattered receive is required and 0 if not.
1144 static int bnxt_scattered_rx(struct rte_eth_dev *eth_dev)
1149 if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER)
1152 if (eth_dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_TCP_LRO)
1155 for (i = 0; i < eth_dev->data->nb_rx_queues; i++) {
1156 struct bnxt_rx_queue *rxq = eth_dev->data->rx_queues[i];
1158 buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
1159 RTE_PKTMBUF_HEADROOM);
1160 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len > buf_size)
1166 static eth_rx_burst_t
1167 bnxt_receive_function(struct rte_eth_dev *eth_dev)
1169 struct bnxt *bp = eth_dev->data->dev_private;
1171 /* Disable vector mode RX for Stingray2 for now */
1172 if (BNXT_CHIP_SR2(bp)) {
1173 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1174 return bnxt_recv_pkts;
1177 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1178 #ifndef RTE_LIBRTE_IEEE1588
1180 * Vector mode receive can be enabled only if scatter rx is not
1181 * in use and rx offloads are limited to VLAN stripping and
1184 if (!eth_dev->data->scattered_rx &&
1185 !(eth_dev->data->dev_conf.rxmode.offloads &
1186 ~(DEV_RX_OFFLOAD_VLAN_STRIP |
1187 DEV_RX_OFFLOAD_KEEP_CRC |
1188 DEV_RX_OFFLOAD_JUMBO_FRAME |
1189 DEV_RX_OFFLOAD_IPV4_CKSUM |
1190 DEV_RX_OFFLOAD_UDP_CKSUM |
1191 DEV_RX_OFFLOAD_TCP_CKSUM |
1192 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
1193 DEV_RX_OFFLOAD_OUTER_UDP_CKSUM |
1194 DEV_RX_OFFLOAD_RSS_HASH |
1195 DEV_RX_OFFLOAD_VLAN_FILTER)) &&
1196 !BNXT_TRUFLOW_EN(bp) && BNXT_NUM_ASYNC_CPR(bp) &&
1197 rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1198 PMD_DRV_LOG(INFO, "Using vector mode receive for port %d\n",
1199 eth_dev->data->port_id);
1200 bp->flags |= BNXT_FLAG_RX_VECTOR_PKT_MODE;
1201 return bnxt_recv_pkts_vec;
1203 PMD_DRV_LOG(INFO, "Vector mode receive disabled for port %d\n",
1204 eth_dev->data->port_id);
1206 "Port %d scatter: %d rx offload: %" PRIX64 "\n",
1207 eth_dev->data->port_id,
1208 eth_dev->data->scattered_rx,
1209 eth_dev->data->dev_conf.rxmode.offloads);
1212 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1213 return bnxt_recv_pkts;
1216 static eth_tx_burst_t
1217 bnxt_transmit_function(struct rte_eth_dev *eth_dev)
1219 struct bnxt *bp = eth_dev->data->dev_private;
1221 /* Disable vector mode TX for Stingray2 for now */
1222 if (BNXT_CHIP_SR2(bp))
1223 return bnxt_xmit_pkts;
1225 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
1226 #ifndef RTE_LIBRTE_IEEE1588
1227 uint64_t offloads = eth_dev->data->dev_conf.txmode.offloads;
1230 * Vector mode transmit can be enabled only if not using scatter rx
1233 if (!eth_dev->data->scattered_rx &&
1234 !(offloads & ~DEV_TX_OFFLOAD_MBUF_FAST_FREE) &&
1235 !BNXT_TRUFLOW_EN(bp) &&
1236 rte_vect_get_max_simd_bitwidth() >= RTE_VECT_SIMD_128) {
1237 PMD_DRV_LOG(INFO, "Using vector mode transmit for port %d\n",
1238 eth_dev->data->port_id);
1239 return bnxt_xmit_pkts_vec;
1241 PMD_DRV_LOG(INFO, "Vector mode transmit disabled for port %d\n",
1242 eth_dev->data->port_id);
1244 "Port %d scatter: %d tx offload: %" PRIX64 "\n",
1245 eth_dev->data->port_id,
1246 eth_dev->data->scattered_rx,
1250 return bnxt_xmit_pkts;
1253 static int bnxt_handle_if_change_status(struct bnxt *bp)
1257 /* Since fw has undergone a reset and lost all contexts,
1258 * set fatal flag to not issue hwrm during cleanup
1260 bp->flags |= BNXT_FLAG_FATAL_ERROR;
1261 bnxt_uninit_resources(bp, true);
1263 /* clear fatal flag so that re-init happens */
1264 bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
1265 rc = bnxt_init_resources(bp, true);
1267 bp->flags &= ~BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE;
1272 static int bnxt_dev_set_link_up_op(struct rte_eth_dev *eth_dev)
1274 struct bnxt *bp = eth_dev->data->dev_private;
1277 if (!BNXT_SINGLE_PF(bp))
1280 if (!bp->link_info->link_up)
1281 rc = bnxt_set_hwrm_link_config(bp, true);
1283 eth_dev->data->dev_link.link_status = 1;
1285 bnxt_print_link_info(eth_dev);
1289 static int bnxt_dev_set_link_down_op(struct rte_eth_dev *eth_dev)
1291 struct bnxt *bp = eth_dev->data->dev_private;
1293 if (!BNXT_SINGLE_PF(bp))
1296 eth_dev->data->dev_link.link_status = 0;
1297 bnxt_set_hwrm_link_config(bp, false);
1298 bp->link_info->link_up = 0;
1303 static void bnxt_free_switch_domain(struct bnxt *bp)
1307 if (!(BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)))
1310 rc = rte_eth_switch_domain_free(bp->switch_domain_id);
1312 PMD_DRV_LOG(ERR, "free switch domain:%d fail: %d\n",
1313 bp->switch_domain_id, rc);
1316 static void bnxt_ptp_get_current_time(void *arg)
1318 struct bnxt *bp = arg;
1319 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1322 rc = is_bnxt_in_error(bp);
1329 bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1330 &ptp->current_time);
1332 rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1334 PMD_DRV_LOG(ERR, "Failed to re-schedule PTP alarm\n");
1335 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1339 static int bnxt_schedule_ptp_alarm(struct bnxt *bp)
1341 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
1344 if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED)
1347 bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
1348 &ptp->current_time);
1350 rc = rte_eal_alarm_set(US_PER_S, bnxt_ptp_get_current_time, (void *)bp);
1354 static void bnxt_cancel_ptp_alarm(struct bnxt *bp)
1356 if (bp->flags2 & BNXT_FLAGS2_PTP_ALARM_SCHEDULED) {
1357 rte_eal_alarm_cancel(bnxt_ptp_get_current_time, (void *)bp);
1358 bp->flags2 &= ~BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1362 static void bnxt_ptp_stop(struct bnxt *bp)
1364 bnxt_cancel_ptp_alarm(bp);
1365 bp->flags2 &= ~BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1368 static int bnxt_ptp_start(struct bnxt *bp)
1372 rc = bnxt_schedule_ptp_alarm(bp);
1374 PMD_DRV_LOG(ERR, "Failed to schedule PTP alarm\n");
1376 bp->flags2 |= BNXT_FLAGS2_PTP_TIMESYNC_ENABLED;
1377 bp->flags2 |= BNXT_FLAGS2_PTP_ALARM_SCHEDULED;
1383 static int bnxt_dev_stop(struct rte_eth_dev *eth_dev)
1385 struct bnxt *bp = eth_dev->data->dev_private;
1386 struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1387 struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1388 struct rte_eth_link link;
1391 eth_dev->data->dev_started = 0;
1392 eth_dev->data->scattered_rx = 0;
1394 /* Prevent crashes when queues are still in use */
1395 eth_dev->rx_pkt_burst = &bnxt_dummy_recv_pkts;
1396 eth_dev->tx_pkt_burst = &bnxt_dummy_xmit_pkts;
1398 bnxt_disable_int(bp);
1400 /* disable uio/vfio intr/eventfd mapping */
1401 rte_intr_disable(intr_handle);
1403 /* Stop the child representors for this device */
1404 ret = bnxt_rep_stop_all(bp);
1408 /* delete the bnxt ULP port details */
1409 bnxt_ulp_port_deinit(bp);
1411 bnxt_cancel_fw_health_check(bp);
1413 if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1414 bnxt_cancel_ptp_alarm(bp);
1416 /* Do not bring link down during reset recovery */
1417 if (!is_bnxt_in_error(bp)) {
1418 bnxt_dev_set_link_down_op(eth_dev);
1419 /* Wait for link to be reset */
1420 if (BNXT_SINGLE_PF(bp))
1422 /* clear the recorded link status */
1423 memset(&link, 0, sizeof(link));
1424 rte_eth_linkstatus_set(eth_dev, &link);
1427 /* Clean queue intr-vector mapping */
1428 rte_intr_efd_disable(intr_handle);
1429 if (intr_handle->intr_vec != NULL) {
1430 rte_free(intr_handle->intr_vec);
1431 intr_handle->intr_vec = NULL;
1434 bnxt_hwrm_port_clr_stats(bp);
1435 bnxt_free_tx_mbufs(bp);
1436 bnxt_free_rx_mbufs(bp);
1437 /* Process any remaining notifications in default completion queue */
1438 bnxt_int_handler(eth_dev);
1439 bnxt_shutdown_nic(bp);
1440 bnxt_hwrm_if_change(bp, false);
1442 rte_free(bp->mark_table);
1443 bp->mark_table = NULL;
1445 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
1446 bp->rx_cosq_cnt = 0;
1447 /* All filters are deleted on a port stop. */
1448 if (BNXT_FLOW_XSTATS_EN(bp))
1449 bp->flow_stat->flow_count = 0;
1454 /* Unload the driver, release resources */
1455 static int bnxt_dev_stop_op(struct rte_eth_dev *eth_dev)
1457 struct bnxt *bp = eth_dev->data->dev_private;
1459 pthread_mutex_lock(&bp->err_recovery_lock);
1460 if (bp->flags & BNXT_FLAG_FW_RESET) {
1462 "Adapter recovering from error..Please retry\n");
1463 pthread_mutex_unlock(&bp->err_recovery_lock);
1466 pthread_mutex_unlock(&bp->err_recovery_lock);
1468 return bnxt_dev_stop(eth_dev);
1471 static int bnxt_dev_start_op(struct rte_eth_dev *eth_dev)
1473 struct bnxt *bp = eth_dev->data->dev_private;
1474 uint64_t rx_offloads = eth_dev->data->dev_conf.rxmode.offloads;
1476 int rc, retry_cnt = BNXT_IF_CHANGE_RETRY_COUNT;
1478 if (!eth_dev->data->nb_tx_queues || !eth_dev->data->nb_rx_queues) {
1479 PMD_DRV_LOG(ERR, "Queues are not configured yet!\n");
1483 if (bp->rx_cp_nr_rings > RTE_ETHDEV_QUEUE_STAT_CNTRS)
1485 "RxQ cnt %d > RTE_ETHDEV_QUEUE_STAT_CNTRS %d\n",
1486 bp->rx_cp_nr_rings, RTE_ETHDEV_QUEUE_STAT_CNTRS);
1489 rc = bnxt_hwrm_if_change(bp, true);
1490 if (rc == 0 || rc != -EAGAIN)
1493 rte_delay_ms(BNXT_IF_CHANGE_RETRY_INTERVAL);
1494 } while (retry_cnt--);
1499 if (bp->flags & BNXT_FLAG_IF_CHANGE_HOT_FW_RESET_DONE) {
1500 rc = bnxt_handle_if_change_status(bp);
1505 bnxt_enable_int(bp);
1507 eth_dev->data->scattered_rx = bnxt_scattered_rx(eth_dev);
1509 rc = bnxt_start_nic(bp);
1513 eth_dev->data->dev_started = 1;
1515 bnxt_link_update_op(eth_dev, 1);
1517 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1518 vlan_mask |= ETH_VLAN_FILTER_MASK;
1519 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1520 vlan_mask |= ETH_VLAN_STRIP_MASK;
1521 rc = bnxt_vlan_offload_set_op(eth_dev, vlan_mask);
1525 /* Initialize bnxt ULP port details */
1526 rc = bnxt_ulp_port_init(bp);
1530 eth_dev->rx_pkt_burst = bnxt_receive_function(eth_dev);
1531 eth_dev->tx_pkt_burst = bnxt_transmit_function(eth_dev);
1533 bnxt_schedule_fw_health_check(bp);
1535 if (BNXT_P5_PTP_TIMESYNC_ENABLED(bp))
1536 bnxt_schedule_ptp_alarm(bp);
1541 bnxt_dev_stop(eth_dev);
1546 bnxt_uninit_locks(struct bnxt *bp)
1548 pthread_mutex_destroy(&bp->flow_lock);
1549 pthread_mutex_destroy(&bp->def_cp_lock);
1550 pthread_mutex_destroy(&bp->health_check_lock);
1551 pthread_mutex_destroy(&bp->err_recovery_lock);
1553 pthread_mutex_destroy(&bp->rep_info->vfr_lock);
1554 pthread_mutex_destroy(&bp->rep_info->vfr_start_lock);
1558 static void bnxt_drv_uninit(struct bnxt *bp)
1560 bnxt_free_leds_info(bp);
1561 bnxt_free_cos_queues(bp);
1562 bnxt_free_link_info(bp);
1563 bnxt_free_parent_info(bp);
1564 bnxt_uninit_locks(bp);
1566 rte_memzone_free((const struct rte_memzone *)bp->tx_mem_zone);
1567 bp->tx_mem_zone = NULL;
1568 rte_memzone_free((const struct rte_memzone *)bp->rx_mem_zone);
1569 bp->rx_mem_zone = NULL;
1571 bnxt_free_vf_info(bp);
1572 bnxt_free_pf_info(bp);
1574 rte_free(bp->grp_info);
1575 bp->grp_info = NULL;
1578 static int bnxt_dev_close_op(struct rte_eth_dev *eth_dev)
1580 struct bnxt *bp = eth_dev->data->dev_private;
1583 if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1586 pthread_mutex_lock(&bp->err_recovery_lock);
1587 if (bp->flags & BNXT_FLAG_FW_RESET) {
1589 "Adapter recovering from error...Please retry\n");
1590 pthread_mutex_unlock(&bp->err_recovery_lock);
1593 pthread_mutex_unlock(&bp->err_recovery_lock);
1595 /* cancel the recovery handler before remove dev */
1596 rte_eal_alarm_cancel(bnxt_dev_reset_and_resume, (void *)bp);
1597 rte_eal_alarm_cancel(bnxt_dev_recover, (void *)bp);
1598 bnxt_cancel_fc_thread(bp);
1600 if (eth_dev->data->dev_started)
1601 ret = bnxt_dev_stop(eth_dev);
1603 bnxt_uninit_resources(bp, false);
1605 bnxt_drv_uninit(bp);
1610 static void bnxt_mac_addr_remove_op(struct rte_eth_dev *eth_dev,
1613 struct bnxt *bp = eth_dev->data->dev_private;
1614 uint64_t pool_mask = eth_dev->data->mac_pool_sel[index];
1615 struct bnxt_vnic_info *vnic;
1616 struct bnxt_filter_info *filter, *temp_filter;
1619 if (is_bnxt_in_error(bp))
1623 * Loop through all VNICs from the specified filter flow pools to
1624 * remove the corresponding MAC addr filter
1626 for (i = 0; i < bp->nr_vnics; i++) {
1627 if (!(pool_mask & (1ULL << i)))
1630 vnic = &bp->vnic_info[i];
1631 filter = STAILQ_FIRST(&vnic->filter);
1633 temp_filter = STAILQ_NEXT(filter, next);
1634 if (filter->mac_index == index) {
1635 STAILQ_REMOVE(&vnic->filter, filter,
1636 bnxt_filter_info, next);
1637 bnxt_hwrm_clear_l2_filter(bp, filter);
1638 bnxt_free_filter(bp, filter);
1640 filter = temp_filter;
1645 static int bnxt_add_mac_filter(struct bnxt *bp, struct bnxt_vnic_info *vnic,
1646 struct rte_ether_addr *mac_addr, uint32_t index,
1649 struct bnxt_filter_info *filter;
1652 /* Attach requested MAC address to the new l2_filter */
1653 STAILQ_FOREACH(filter, &vnic->filter, next) {
1654 if (filter->mac_index == index) {
1656 "MAC addr already existed for pool %d\n",
1662 filter = bnxt_alloc_filter(bp);
1664 PMD_DRV_LOG(ERR, "L2 filter alloc failed\n");
1668 /* bnxt_alloc_filter copies default MAC to filter->l2_addr. So,
1669 * if the MAC that's been programmed now is a different one, then,
1670 * copy that addr to filter->l2_addr
1673 memcpy(filter->l2_addr, mac_addr, RTE_ETHER_ADDR_LEN);
1674 filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
1676 rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
1678 filter->mac_index = index;
1679 if (filter->mac_index == 0)
1680 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
1682 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
1684 bnxt_free_filter(bp, filter);
1690 static int bnxt_mac_addr_add_op(struct rte_eth_dev *eth_dev,
1691 struct rte_ether_addr *mac_addr,
1692 uint32_t index, uint32_t pool)
1694 struct bnxt *bp = eth_dev->data->dev_private;
1695 struct bnxt_vnic_info *vnic = &bp->vnic_info[pool];
1698 rc = is_bnxt_in_error(bp);
1702 if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp)) {
1703 PMD_DRV_LOG(ERR, "Cannot add MAC address to a VF interface\n");
1708 PMD_DRV_LOG(ERR, "VNIC not found for pool %d!\n", pool);
1712 /* Filter settings will get applied when port is started */
1713 if (!eth_dev->data->dev_started)
1716 rc = bnxt_add_mac_filter(bp, vnic, mac_addr, index, pool);
1721 int bnxt_link_update_op(struct rte_eth_dev *eth_dev, int wait_to_complete)
1724 struct bnxt *bp = eth_dev->data->dev_private;
1725 struct rte_eth_link new;
1726 int cnt = wait_to_complete ? BNXT_MAX_LINK_WAIT_CNT :
1727 BNXT_MIN_LINK_WAIT_CNT;
1729 rc = is_bnxt_in_error(bp);
1733 memset(&new, 0, sizeof(new));
1735 if (bp->link_info == NULL)
1739 /* Retrieve link info from hardware */
1740 rc = bnxt_get_hwrm_link_config(bp, &new);
1742 new.link_speed = ETH_LINK_SPEED_100M;
1743 new.link_duplex = ETH_LINK_FULL_DUPLEX;
1745 "Failed to retrieve link rc = 0x%x!\n", rc);
1749 if (!wait_to_complete || new.link_status)
1752 rte_delay_ms(BNXT_LINK_WAIT_INTERVAL);
1755 /* Only single function PF can bring phy down.
1756 * When port is stopped, report link down for VF/MH/NPAR functions.
1758 if (!BNXT_SINGLE_PF(bp) && !eth_dev->data->dev_started)
1759 memset(&new, 0, sizeof(new));
1762 /* Timed out or success */
1763 if (new.link_status != eth_dev->data->dev_link.link_status ||
1764 new.link_speed != eth_dev->data->dev_link.link_speed) {
1765 rte_eth_linkstatus_set(eth_dev, &new);
1767 rte_eth_dev_callback_process(eth_dev,
1768 RTE_ETH_EVENT_INTR_LSC,
1771 bnxt_print_link_info(eth_dev);
1777 static int bnxt_promiscuous_enable_op(struct rte_eth_dev *eth_dev)
1779 struct bnxt *bp = eth_dev->data->dev_private;
1780 struct bnxt_vnic_info *vnic;
1784 rc = is_bnxt_in_error(bp);
1788 /* Filter settings will get applied when port is started */
1789 if (!eth_dev->data->dev_started)
1792 if (bp->vnic_info == NULL)
1795 vnic = BNXT_GET_DEFAULT_VNIC(bp);
1797 old_flags = vnic->flags;
1798 vnic->flags |= BNXT_VNIC_INFO_PROMISC;
1799 rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1801 vnic->flags = old_flags;
1806 static int bnxt_promiscuous_disable_op(struct rte_eth_dev *eth_dev)
1808 struct bnxt *bp = eth_dev->data->dev_private;
1809 struct bnxt_vnic_info *vnic;
1813 rc = is_bnxt_in_error(bp);
1817 /* Filter settings will get applied when port is started */
1818 if (!eth_dev->data->dev_started)
1821 if (bp->vnic_info == NULL)
1824 vnic = BNXT_GET_DEFAULT_VNIC(bp);
1826 old_flags = vnic->flags;
1827 vnic->flags &= ~BNXT_VNIC_INFO_PROMISC;
1828 rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1830 vnic->flags = old_flags;
1835 static int bnxt_allmulticast_enable_op(struct rte_eth_dev *eth_dev)
1837 struct bnxt *bp = eth_dev->data->dev_private;
1838 struct bnxt_vnic_info *vnic;
1842 rc = is_bnxt_in_error(bp);
1846 /* Filter settings will get applied when port is started */
1847 if (!eth_dev->data->dev_started)
1850 if (bp->vnic_info == NULL)
1853 vnic = BNXT_GET_DEFAULT_VNIC(bp);
1855 old_flags = vnic->flags;
1856 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
1857 rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1859 vnic->flags = old_flags;
1864 static int bnxt_allmulticast_disable_op(struct rte_eth_dev *eth_dev)
1866 struct bnxt *bp = eth_dev->data->dev_private;
1867 struct bnxt_vnic_info *vnic;
1871 rc = is_bnxt_in_error(bp);
1875 /* Filter settings will get applied when port is started */
1876 if (!eth_dev->data->dev_started)
1879 if (bp->vnic_info == NULL)
1882 vnic = BNXT_GET_DEFAULT_VNIC(bp);
1884 old_flags = vnic->flags;
1885 vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
1886 rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
1888 vnic->flags = old_flags;
1893 /* Return bnxt_rx_queue pointer corresponding to a given rxq. */
1894 static struct bnxt_rx_queue *bnxt_qid_to_rxq(struct bnxt *bp, uint16_t qid)
1896 if (qid >= bp->rx_nr_rings)
1899 return bp->eth_dev->data->rx_queues[qid];
1902 /* Return rxq corresponding to a given rss table ring/group ID. */
1903 static uint16_t bnxt_rss_to_qid(struct bnxt *bp, uint16_t fwr)
1905 struct bnxt_rx_queue *rxq;
1908 if (!BNXT_HAS_RING_GRPS(bp)) {
1909 for (i = 0; i < bp->rx_nr_rings; i++) {
1910 rxq = bp->eth_dev->data->rx_queues[i];
1911 if (rxq->rx_ring->rx_ring_struct->fw_ring_id == fwr)
1915 for (i = 0; i < bp->rx_nr_rings; i++) {
1916 if (bp->grp_info[i].fw_grp_id == fwr)
1921 return INVALID_HW_RING_ID;
1924 static int bnxt_reta_update_op(struct rte_eth_dev *eth_dev,
1925 struct rte_eth_rss_reta_entry64 *reta_conf,
1928 struct bnxt *bp = eth_dev->data->dev_private;
1929 struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
1930 struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1931 uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1935 rc = is_bnxt_in_error(bp);
1939 if (!vnic->rss_table)
1942 if (!(dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG))
1945 if (reta_size != tbl_size) {
1946 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
1947 "(%d) must equal the size supported by the hardware "
1948 "(%d)\n", reta_size, tbl_size);
1952 for (i = 0; i < reta_size; i++) {
1953 struct bnxt_rx_queue *rxq;
1955 idx = i / RTE_RETA_GROUP_SIZE;
1956 sft = i % RTE_RETA_GROUP_SIZE;
1958 if (!(reta_conf[idx].mask & (1ULL << sft)))
1961 rxq = bnxt_qid_to_rxq(bp, reta_conf[idx].reta[sft]);
1963 PMD_DRV_LOG(ERR, "Invalid ring in reta_conf.\n");
1967 if (BNXT_CHIP_P5(bp)) {
1968 vnic->rss_table[i * 2] =
1969 rxq->rx_ring->rx_ring_struct->fw_ring_id;
1970 vnic->rss_table[i * 2 + 1] =
1971 rxq->cp_ring->cp_ring_struct->fw_ring_id;
1973 vnic->rss_table[i] =
1974 vnic->fw_grp_ids[reta_conf[idx].reta[sft]];
1978 rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
1982 static int bnxt_reta_query_op(struct rte_eth_dev *eth_dev,
1983 struct rte_eth_rss_reta_entry64 *reta_conf,
1986 struct bnxt *bp = eth_dev->data->dev_private;
1987 struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
1988 uint16_t tbl_size = bnxt_rss_hash_tbl_size(bp);
1989 uint16_t idx, sft, i;
1992 rc = is_bnxt_in_error(bp);
1996 /* Retrieve from the default VNIC */
1999 if (!vnic->rss_table)
2002 if (reta_size != tbl_size) {
2003 PMD_DRV_LOG(ERR, "The configured hash table lookup size "
2004 "(%d) must equal the size supported by the hardware "
2005 "(%d)\n", reta_size, tbl_size);
2009 for (idx = 0, i = 0; i < reta_size; i++) {
2010 idx = i / RTE_RETA_GROUP_SIZE;
2011 sft = i % RTE_RETA_GROUP_SIZE;
2013 if (reta_conf[idx].mask & (1ULL << sft)) {
2016 if (BNXT_CHIP_P5(bp))
2017 qid = bnxt_rss_to_qid(bp,
2018 vnic->rss_table[i * 2]);
2020 qid = bnxt_rss_to_qid(bp, vnic->rss_table[i]);
2022 if (qid == INVALID_HW_RING_ID) {
2023 PMD_DRV_LOG(ERR, "Inv. entry in rss table.\n");
2026 reta_conf[idx].reta[sft] = qid;
2033 static int bnxt_rss_hash_update_op(struct rte_eth_dev *eth_dev,
2034 struct rte_eth_rss_conf *rss_conf)
2036 struct bnxt *bp = eth_dev->data->dev_private;
2037 struct rte_eth_conf *dev_conf = &bp->eth_dev->data->dev_conf;
2038 struct bnxt_vnic_info *vnic;
2041 rc = is_bnxt_in_error(bp);
2046 * If RSS enablement were different than dev_configure,
2047 * then return -EINVAL
2049 if (dev_conf->rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
2050 if (!rss_conf->rss_hf)
2051 PMD_DRV_LOG(ERR, "Hash type NONE\n");
2053 if (rss_conf->rss_hf & BNXT_ETH_RSS_SUPPORT)
2057 bp->flags |= BNXT_FLAG_UPDATE_HASH;
2058 memcpy(ð_dev->data->dev_conf.rx_adv_conf.rss_conf,
2062 /* Update the default RSS VNIC(s) */
2063 vnic = BNXT_GET_DEFAULT_VNIC(bp);
2064 vnic->hash_type = bnxt_rte_to_hwrm_hash_types(rss_conf->rss_hf);
2066 bnxt_rte_to_hwrm_hash_level(bp, rss_conf->rss_hf,
2067 ETH_RSS_LEVEL(rss_conf->rss_hf));
2070 * If hashkey is not specified, use the previously configured
2073 if (!rss_conf->rss_key)
2076 if (rss_conf->rss_key_len != HW_HASH_KEY_SIZE) {
2078 "Invalid hashkey length, should be 16 bytes\n");
2081 memcpy(vnic->rss_hash_key, rss_conf->rss_key, rss_conf->rss_key_len);
2084 rc = bnxt_hwrm_vnic_rss_cfg(bp, vnic);
2088 static int bnxt_rss_hash_conf_get_op(struct rte_eth_dev *eth_dev,
2089 struct rte_eth_rss_conf *rss_conf)
2091 struct bnxt *bp = eth_dev->data->dev_private;
2092 struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2094 uint32_t hash_types;
2096 rc = is_bnxt_in_error(bp);
2100 /* RSS configuration is the same for all VNICs */
2101 if (vnic && vnic->rss_hash_key) {
2102 if (rss_conf->rss_key) {
2103 len = rss_conf->rss_key_len <= HW_HASH_KEY_SIZE ?
2104 rss_conf->rss_key_len : HW_HASH_KEY_SIZE;
2105 memcpy(rss_conf->rss_key, vnic->rss_hash_key, len);
2108 hash_types = vnic->hash_type;
2109 rss_conf->rss_hf = 0;
2110 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4) {
2111 rss_conf->rss_hf |= ETH_RSS_IPV4;
2112 hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV4;
2114 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4) {
2115 rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_TCP;
2117 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV4;
2119 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4) {
2120 rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV4_UDP;
2122 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV4;
2124 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6) {
2125 rss_conf->rss_hf |= ETH_RSS_IPV6;
2126 hash_types &= ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_IPV6;
2128 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6) {
2129 rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_TCP;
2131 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_TCP_IPV6;
2133 if (hash_types & HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6) {
2134 rss_conf->rss_hf |= ETH_RSS_NONFRAG_IPV6_UDP;
2136 ~HWRM_VNIC_RSS_CFG_INPUT_HASH_TYPE_UDP_IPV6;
2140 bnxt_hwrm_to_rte_rss_level(bp, vnic->hash_mode);
2144 "Unknown RSS config from firmware (%08x), RSS disabled",
2149 rss_conf->rss_hf = 0;
2154 static int bnxt_flow_ctrl_get_op(struct rte_eth_dev *dev,
2155 struct rte_eth_fc_conf *fc_conf)
2157 struct bnxt *bp = dev->data->dev_private;
2158 struct rte_eth_link link_info;
2161 rc = is_bnxt_in_error(bp);
2165 rc = bnxt_get_hwrm_link_config(bp, &link_info);
2169 memset(fc_conf, 0, sizeof(*fc_conf));
2170 if (bp->link_info->auto_pause)
2171 fc_conf->autoneg = 1;
2172 switch (bp->link_info->pause) {
2174 fc_conf->mode = RTE_FC_NONE;
2176 case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX:
2177 fc_conf->mode = RTE_FC_TX_PAUSE;
2179 case HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX:
2180 fc_conf->mode = RTE_FC_RX_PAUSE;
2182 case (HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_TX |
2183 HWRM_PORT_PHY_QCFG_OUTPUT_PAUSE_RX):
2184 fc_conf->mode = RTE_FC_FULL;
2190 static int bnxt_flow_ctrl_set_op(struct rte_eth_dev *dev,
2191 struct rte_eth_fc_conf *fc_conf)
2193 struct bnxt *bp = dev->data->dev_private;
2196 rc = is_bnxt_in_error(bp);
2200 if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2201 PMD_DRV_LOG(ERR, "Flow Control Settings cannot be modified\n");
2205 switch (fc_conf->mode) {
2207 bp->link_info->auto_pause = 0;
2208 bp->link_info->force_pause = 0;
2210 case RTE_FC_RX_PAUSE:
2211 if (fc_conf->autoneg) {
2212 bp->link_info->auto_pause =
2213 HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2214 bp->link_info->force_pause = 0;
2216 bp->link_info->auto_pause = 0;
2217 bp->link_info->force_pause =
2218 HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2221 case RTE_FC_TX_PAUSE:
2222 if (fc_conf->autoneg) {
2223 bp->link_info->auto_pause =
2224 HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX;
2225 bp->link_info->force_pause = 0;
2227 bp->link_info->auto_pause = 0;
2228 bp->link_info->force_pause =
2229 HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX;
2233 if (fc_conf->autoneg) {
2234 bp->link_info->auto_pause =
2235 HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_TX |
2236 HWRM_PORT_PHY_CFG_INPUT_AUTO_PAUSE_RX;
2237 bp->link_info->force_pause = 0;
2239 bp->link_info->auto_pause = 0;
2240 bp->link_info->force_pause =
2241 HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_TX |
2242 HWRM_PORT_PHY_CFG_INPUT_FORCE_PAUSE_RX;
2246 return bnxt_set_hwrm_link_config(bp, true);
2249 /* Add UDP tunneling port */
2251 bnxt_udp_tunnel_port_add_op(struct rte_eth_dev *eth_dev,
2252 struct rte_eth_udp_tunnel *udp_tunnel)
2254 struct bnxt *bp = eth_dev->data->dev_private;
2255 uint16_t tunnel_type = 0;
2258 rc = is_bnxt_in_error(bp);
2262 switch (udp_tunnel->prot_type) {
2263 case RTE_TUNNEL_TYPE_VXLAN:
2264 if (bp->vxlan_port_cnt) {
2265 PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2266 udp_tunnel->udp_port);
2267 if (bp->vxlan_port != udp_tunnel->udp_port) {
2268 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2271 bp->vxlan_port_cnt++;
2275 HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_VXLAN;
2276 bp->vxlan_port_cnt++;
2278 case RTE_TUNNEL_TYPE_GENEVE:
2279 if (bp->geneve_port_cnt) {
2280 PMD_DRV_LOG(ERR, "Tunnel Port %d already programmed\n",
2281 udp_tunnel->udp_port);
2282 if (bp->geneve_port != udp_tunnel->udp_port) {
2283 PMD_DRV_LOG(ERR, "Only one port allowed\n");
2286 bp->geneve_port_cnt++;
2290 HWRM_TUNNEL_DST_PORT_ALLOC_INPUT_TUNNEL_TYPE_GENEVE;
2291 bp->geneve_port_cnt++;
2294 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2297 rc = bnxt_hwrm_tunnel_dst_port_alloc(bp, udp_tunnel->udp_port,
2303 bnxt_udp_tunnel_port_del_op(struct rte_eth_dev *eth_dev,
2304 struct rte_eth_udp_tunnel *udp_tunnel)
2306 struct bnxt *bp = eth_dev->data->dev_private;
2307 uint16_t tunnel_type = 0;
2311 rc = is_bnxt_in_error(bp);
2315 switch (udp_tunnel->prot_type) {
2316 case RTE_TUNNEL_TYPE_VXLAN:
2317 if (!bp->vxlan_port_cnt) {
2318 PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2321 if (bp->vxlan_port != udp_tunnel->udp_port) {
2322 PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2323 udp_tunnel->udp_port, bp->vxlan_port);
2326 if (--bp->vxlan_port_cnt)
2330 HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_VXLAN;
2331 port = bp->vxlan_fw_dst_port_id;
2333 case RTE_TUNNEL_TYPE_GENEVE:
2334 if (!bp->geneve_port_cnt) {
2335 PMD_DRV_LOG(ERR, "No Tunnel port configured yet\n");
2338 if (bp->geneve_port != udp_tunnel->udp_port) {
2339 PMD_DRV_LOG(ERR, "Req Port: %d. Configured port: %d\n",
2340 udp_tunnel->udp_port, bp->geneve_port);
2343 if (--bp->geneve_port_cnt)
2347 HWRM_TUNNEL_DST_PORT_FREE_INPUT_TUNNEL_TYPE_GENEVE;
2348 port = bp->geneve_fw_dst_port_id;
2351 PMD_DRV_LOG(ERR, "Tunnel type is not supported\n");
2355 rc = bnxt_hwrm_tunnel_dst_port_free(bp, port, tunnel_type);
2359 static int bnxt_del_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2361 struct bnxt_filter_info *filter;
2362 struct bnxt_vnic_info *vnic;
2364 uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2366 vnic = BNXT_GET_DEFAULT_VNIC(bp);
2367 filter = STAILQ_FIRST(&vnic->filter);
2369 /* Search for this matching MAC+VLAN filter */
2370 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id)) {
2371 /* Delete the filter */
2372 rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2375 STAILQ_REMOVE(&vnic->filter, filter,
2376 bnxt_filter_info, next);
2377 bnxt_free_filter(bp, filter);
2379 "Deleted vlan filter for %d\n",
2383 filter = STAILQ_NEXT(filter, next);
2388 static int bnxt_add_vlan_filter(struct bnxt *bp, uint16_t vlan_id)
2390 struct bnxt_filter_info *filter;
2391 struct bnxt_vnic_info *vnic;
2393 uint32_t en = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN |
2394 HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN_MASK;
2395 uint32_t chk = HWRM_CFA_L2_FILTER_ALLOC_INPUT_ENABLES_L2_IVLAN;
2397 /* Implementation notes on the use of VNIC in this command:
2399 * By default, these filters belong to default vnic for the function.
2400 * Once these filters are set up, only destination VNIC can be modified.
2401 * If the destination VNIC is not specified in this command,
2402 * then the HWRM shall only create an l2 context id.
2405 vnic = BNXT_GET_DEFAULT_VNIC(bp);
2406 filter = STAILQ_FIRST(&vnic->filter);
2407 /* Check if the VLAN has already been added */
2409 if (bnxt_vlan_filter_exists(bp, filter, chk, vlan_id))
2412 filter = STAILQ_NEXT(filter, next);
2415 /* No match found. Alloc a fresh filter and issue the L2_FILTER_ALLOC
2416 * command to create MAC+VLAN filter with the right flags, enables set.
2418 filter = bnxt_alloc_filter(bp);
2421 "MAC/VLAN filter alloc failed\n");
2424 /* MAC + VLAN ID filter */
2425 /* If l2_ivlan == 0 and l2_ivlan_mask != 0, only
2426 * untagged packets are received
2428 * If l2_ivlan != 0 and l2_ivlan_mask != 0, untagged
2429 * packets and only the programmed vlan's packets are received
2431 filter->l2_ivlan = vlan_id;
2432 filter->l2_ivlan_mask = 0x0FFF;
2433 filter->enables |= en;
2434 filter->flags |= HWRM_CFA_L2_FILTER_ALLOC_INPUT_FLAGS_OUTERMOST;
2436 rc = bnxt_hwrm_set_l2_filter(bp, vnic->fw_vnic_id, filter);
2438 /* Free the newly allocated filter as we were
2439 * not able to create the filter in hardware.
2441 bnxt_free_filter(bp, filter);
2445 filter->mac_index = 0;
2446 /* Add this new filter to the list */
2448 STAILQ_INSERT_HEAD(&vnic->filter, filter, next);
2450 STAILQ_INSERT_TAIL(&vnic->filter, filter, next);
2453 "Added Vlan filter for %d\n", vlan_id);
2457 static int bnxt_vlan_filter_set_op(struct rte_eth_dev *eth_dev,
2458 uint16_t vlan_id, int on)
2460 struct bnxt *bp = eth_dev->data->dev_private;
2463 rc = is_bnxt_in_error(bp);
2467 if (!eth_dev->data->dev_started) {
2468 PMD_DRV_LOG(ERR, "port must be started before setting vlan\n");
2472 /* These operations apply to ALL existing MAC/VLAN filters */
2474 return bnxt_add_vlan_filter(bp, vlan_id);
2476 return bnxt_del_vlan_filter(bp, vlan_id);
2479 static int bnxt_del_dflt_mac_filter(struct bnxt *bp,
2480 struct bnxt_vnic_info *vnic)
2482 struct bnxt_filter_info *filter;
2485 filter = STAILQ_FIRST(&vnic->filter);
2487 if (filter->mac_index == 0 &&
2488 !memcmp(filter->l2_addr, bp->mac_addr,
2489 RTE_ETHER_ADDR_LEN)) {
2490 rc = bnxt_hwrm_clear_l2_filter(bp, filter);
2492 STAILQ_REMOVE(&vnic->filter, filter,
2493 bnxt_filter_info, next);
2494 bnxt_free_filter(bp, filter);
2498 filter = STAILQ_NEXT(filter, next);
2504 bnxt_config_vlan_hw_filter(struct bnxt *bp, uint64_t rx_offloads)
2506 struct bnxt_vnic_info *vnic;
2510 vnic = BNXT_GET_DEFAULT_VNIC(bp);
2511 if (!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)) {
2512 /* Remove any VLAN filters programmed */
2513 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2514 bnxt_del_vlan_filter(bp, i);
2516 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2520 /* Default filter will allow packets that match the
2521 * dest mac. So, it has to be deleted, otherwise, we
2522 * will endup receiving vlan packets for which the
2523 * filter is not programmed, when hw-vlan-filter
2524 * configuration is ON
2526 bnxt_del_dflt_mac_filter(bp, vnic);
2527 /* This filter will allow only untagged packets */
2528 bnxt_add_vlan_filter(bp, 0);
2530 PMD_DRV_LOG(DEBUG, "VLAN Filtering: %d\n",
2531 !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER));
2536 static int bnxt_free_one_vnic(struct bnxt *bp, uint16_t vnic_id)
2538 struct bnxt_vnic_info *vnic = &bp->vnic_info[vnic_id];
2542 /* Destroy vnic filters and vnic */
2543 if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2544 DEV_RX_OFFLOAD_VLAN_FILTER) {
2545 for (i = 0; i < RTE_ETHER_MAX_VLAN_ID; i++)
2546 bnxt_del_vlan_filter(bp, i);
2548 bnxt_del_dflt_mac_filter(bp, vnic);
2550 rc = bnxt_hwrm_vnic_ctx_free(bp, vnic);
2554 rc = bnxt_hwrm_vnic_free(bp, vnic);
2558 rte_free(vnic->fw_grp_ids);
2559 vnic->fw_grp_ids = NULL;
2561 vnic->rx_queue_cnt = 0;
2567 bnxt_config_vlan_hw_stripping(struct bnxt *bp, uint64_t rx_offloads)
2569 struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2572 /* Destroy, recreate and reconfigure the default vnic */
2573 rc = bnxt_free_one_vnic(bp, 0);
2577 /* default vnic 0 */
2578 rc = bnxt_setup_one_vnic(bp, 0);
2582 if (bp->eth_dev->data->dev_conf.rxmode.offloads &
2583 DEV_RX_OFFLOAD_VLAN_FILTER) {
2584 rc = bnxt_add_vlan_filter(bp, 0);
2587 rc = bnxt_restore_vlan_filters(bp);
2591 rc = bnxt_add_mac_filter(bp, vnic, NULL, 0, 0);
2596 rc = bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2600 PMD_DRV_LOG(DEBUG, "VLAN Strip Offload: %d\n",
2601 !!(rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP));
2607 bnxt_vlan_offload_set_op(struct rte_eth_dev *dev, int mask)
2609 uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
2610 struct bnxt *bp = dev->data->dev_private;
2613 rc = is_bnxt_in_error(bp);
2617 /* Filter settings will get applied when port is started */
2618 if (!dev->data->dev_started)
2621 if (mask & ETH_VLAN_FILTER_MASK) {
2622 /* Enable or disable VLAN filtering */
2623 rc = bnxt_config_vlan_hw_filter(bp, rx_offloads);
2628 if (mask & ETH_VLAN_STRIP_MASK) {
2629 /* Enable or disable VLAN stripping */
2630 rc = bnxt_config_vlan_hw_stripping(bp, rx_offloads);
2635 if (mask & ETH_VLAN_EXTEND_MASK) {
2636 if (rx_offloads & DEV_RX_OFFLOAD_VLAN_EXTEND)
2637 PMD_DRV_LOG(DEBUG, "Extend VLAN supported\n");
2639 PMD_DRV_LOG(INFO, "Extend VLAN unsupported\n");
2646 bnxt_vlan_tpid_set_op(struct rte_eth_dev *dev, enum rte_vlan_type vlan_type,
2649 struct bnxt *bp = dev->data->dev_private;
2650 int qinq = dev->data->dev_conf.rxmode.offloads &
2651 DEV_RX_OFFLOAD_VLAN_EXTEND;
2653 if (vlan_type != ETH_VLAN_TYPE_INNER &&
2654 vlan_type != ETH_VLAN_TYPE_OUTER) {
2656 "Unsupported vlan type.");
2661 "QinQ not enabled. Needs to be ON as we can "
2662 "accelerate only outer vlan\n");
2666 if (vlan_type == ETH_VLAN_TYPE_OUTER) {
2668 case RTE_ETHER_TYPE_QINQ:
2670 TX_BD_LONG_CFA_META_VLAN_TPID_TPID88A8;
2672 case RTE_ETHER_TYPE_VLAN:
2674 TX_BD_LONG_CFA_META_VLAN_TPID_TPID8100;
2676 case RTE_ETHER_TYPE_QINQ1:
2678 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9100;
2680 case RTE_ETHER_TYPE_QINQ2:
2682 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9200;
2684 case RTE_ETHER_TYPE_QINQ3:
2686 TX_BD_LONG_CFA_META_VLAN_TPID_TPID9300;
2689 PMD_DRV_LOG(ERR, "Invalid TPID: %x\n", tpid);
2692 bp->outer_tpid_bd |= tpid;
2693 PMD_DRV_LOG(INFO, "outer_tpid_bd = %x\n", bp->outer_tpid_bd);
2694 } else if (vlan_type == ETH_VLAN_TYPE_INNER) {
2696 "Can accelerate only outer vlan in QinQ\n");
2704 bnxt_set_default_mac_addr_op(struct rte_eth_dev *dev,
2705 struct rte_ether_addr *addr)
2707 struct bnxt *bp = dev->data->dev_private;
2708 /* Default Filter is tied to VNIC 0 */
2709 struct bnxt_vnic_info *vnic = BNXT_GET_DEFAULT_VNIC(bp);
2712 rc = is_bnxt_in_error(bp);
2716 if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
2719 if (rte_is_zero_ether_addr(addr))
2722 /* Filter settings will get applied when port is started */
2723 if (!dev->data->dev_started)
2726 /* Check if the requested MAC is already added */
2727 if (memcmp(addr, bp->mac_addr, RTE_ETHER_ADDR_LEN) == 0)
2730 /* Destroy filter and re-create it */
2731 bnxt_del_dflt_mac_filter(bp, vnic);
2733 memcpy(bp->mac_addr, addr, RTE_ETHER_ADDR_LEN);
2734 if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_VLAN_FILTER) {
2735 /* This filter will allow only untagged packets */
2736 rc = bnxt_add_vlan_filter(bp, 0);
2738 rc = bnxt_add_mac_filter(bp, vnic, addr, 0, 0);
2741 PMD_DRV_LOG(DEBUG, "Set MAC addr\n");
2746 bnxt_dev_set_mc_addr_list_op(struct rte_eth_dev *eth_dev,
2747 struct rte_ether_addr *mc_addr_set,
2748 uint32_t nb_mc_addr)
2750 struct bnxt *bp = eth_dev->data->dev_private;
2751 char *mc_addr_list = (char *)mc_addr_set;
2752 struct bnxt_vnic_info *vnic;
2753 uint32_t off = 0, i = 0;
2756 rc = is_bnxt_in_error(bp);
2760 vnic = BNXT_GET_DEFAULT_VNIC(bp);
2762 if (nb_mc_addr > BNXT_MAX_MC_ADDRS) {
2763 vnic->flags |= BNXT_VNIC_INFO_ALLMULTI;
2767 /* TODO Check for Duplicate mcast addresses */
2768 vnic->flags &= ~BNXT_VNIC_INFO_ALLMULTI;
2769 for (i = 0; i < nb_mc_addr; i++) {
2770 memcpy(vnic->mc_list + off, &mc_addr_list[i],
2771 RTE_ETHER_ADDR_LEN);
2772 off += RTE_ETHER_ADDR_LEN;
2775 vnic->mc_addr_cnt = i;
2776 if (vnic->mc_addr_cnt)
2777 vnic->flags |= BNXT_VNIC_INFO_MCAST;
2779 vnic->flags &= ~BNXT_VNIC_INFO_MCAST;
2782 return bnxt_hwrm_cfa_l2_set_rx_mask(bp, vnic, 0, NULL);
2786 bnxt_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size)
2788 struct bnxt *bp = dev->data->dev_private;
2789 uint8_t fw_major = (bp->fw_ver >> 24) & 0xff;
2790 uint8_t fw_minor = (bp->fw_ver >> 16) & 0xff;
2791 uint8_t fw_updt = (bp->fw_ver >> 8) & 0xff;
2792 uint8_t fw_rsvd = bp->fw_ver & 0xff;
2795 ret = snprintf(fw_version, fw_size, "%d.%d.%d.%d",
2796 fw_major, fw_minor, fw_updt, fw_rsvd);
2800 ret += 1; /* add the size of '\0' */
2801 if (fw_size < (size_t)ret)
2808 bnxt_rxq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2809 struct rte_eth_rxq_info *qinfo)
2811 struct bnxt *bp = dev->data->dev_private;
2812 struct bnxt_rx_queue *rxq;
2814 if (is_bnxt_in_error(bp))
2817 rxq = dev->data->rx_queues[queue_id];
2819 qinfo->mp = rxq->mb_pool;
2820 qinfo->scattered_rx = dev->data->scattered_rx;
2821 qinfo->nb_desc = rxq->nb_rx_desc;
2823 qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
2824 qinfo->conf.rx_drop_en = rxq->drop_en;
2825 qinfo->conf.rx_deferred_start = rxq->rx_deferred_start;
2826 qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads;
2830 bnxt_txq_info_get_op(struct rte_eth_dev *dev, uint16_t queue_id,
2831 struct rte_eth_txq_info *qinfo)
2833 struct bnxt *bp = dev->data->dev_private;
2834 struct bnxt_tx_queue *txq;
2836 if (is_bnxt_in_error(bp))
2839 txq = dev->data->tx_queues[queue_id];
2841 qinfo->nb_desc = txq->nb_tx_desc;
2843 qinfo->conf.tx_thresh.pthresh = txq->pthresh;
2844 qinfo->conf.tx_thresh.hthresh = txq->hthresh;
2845 qinfo->conf.tx_thresh.wthresh = txq->wthresh;
2847 qinfo->conf.tx_free_thresh = txq->tx_free_thresh;
2848 qinfo->conf.tx_rs_thresh = 0;
2849 qinfo->conf.tx_deferred_start = txq->tx_deferred_start;
2850 qinfo->conf.offloads = txq->offloads;
2853 static const struct {
2854 eth_rx_burst_t pkt_burst;
2856 } bnxt_rx_burst_info[] = {
2857 {bnxt_recv_pkts, "Scalar"},
2858 #if defined(RTE_ARCH_X86)
2859 {bnxt_recv_pkts_vec, "Vector SSE"},
2860 #elif defined(RTE_ARCH_ARM64)
2861 {bnxt_recv_pkts_vec, "Vector Neon"},
2866 bnxt_rx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2867 struct rte_eth_burst_mode *mode)
2869 eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
2872 for (i = 0; i < RTE_DIM(bnxt_rx_burst_info); i++) {
2873 if (pkt_burst == bnxt_rx_burst_info[i].pkt_burst) {
2874 snprintf(mode->info, sizeof(mode->info), "%s",
2875 bnxt_rx_burst_info[i].info);
2883 static const struct {
2884 eth_tx_burst_t pkt_burst;
2886 } bnxt_tx_burst_info[] = {
2887 {bnxt_xmit_pkts, "Scalar"},
2888 #if defined(RTE_ARCH_X86)
2889 {bnxt_xmit_pkts_vec, "Vector SSE"},
2890 #elif defined(RTE_ARCH_ARM64)
2891 {bnxt_xmit_pkts_vec, "Vector Neon"},
2896 bnxt_tx_burst_mode_get(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id,
2897 struct rte_eth_burst_mode *mode)
2899 eth_tx_burst_t pkt_burst = dev->tx_pkt_burst;
2902 for (i = 0; i < RTE_DIM(bnxt_tx_burst_info); i++) {
2903 if (pkt_burst == bnxt_tx_burst_info[i].pkt_burst) {
2904 snprintf(mode->info, sizeof(mode->info), "%s",
2905 bnxt_tx_burst_info[i].info);
2913 int bnxt_mtu_set_op(struct rte_eth_dev *eth_dev, uint16_t new_mtu)
2915 struct bnxt *bp = eth_dev->data->dev_private;
2916 uint32_t new_pkt_size;
2920 rc = is_bnxt_in_error(bp);
2924 /* Exit if receive queues are not configured yet */
2925 if (!eth_dev->data->nb_rx_queues)
2928 new_pkt_size = new_mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
2929 VLAN_TAG_SIZE * BNXT_NUM_VLANS;
2932 * Disallow any MTU change that would require scattered receive support
2933 * if it is not already enabled.
2935 if (eth_dev->data->dev_started &&
2936 !eth_dev->data->scattered_rx &&
2938 eth_dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
2940 "MTU change would require scattered rx support. ");
2941 PMD_DRV_LOG(ERR, "Stop port before changing MTU.\n");
2945 if (new_mtu > RTE_ETHER_MTU) {
2946 bp->flags |= BNXT_FLAG_JUMBO;
2947 bp->eth_dev->data->dev_conf.rxmode.offloads |=
2948 DEV_RX_OFFLOAD_JUMBO_FRAME;
2950 bp->eth_dev->data->dev_conf.rxmode.offloads &=
2951 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
2952 bp->flags &= ~BNXT_FLAG_JUMBO;
2955 /* Is there a change in mtu setting? */
2956 if (eth_dev->data->dev_conf.rxmode.max_rx_pkt_len == new_pkt_size)
2959 for (i = 0; i < bp->nr_vnics; i++) {
2960 struct bnxt_vnic_info *vnic = &bp->vnic_info[i];
2963 vnic->mru = BNXT_VNIC_MRU(new_mtu);
2964 rc = bnxt_hwrm_vnic_cfg(bp, vnic);
2968 size = rte_pktmbuf_data_room_size(bp->rx_queues[0]->mb_pool);
2969 size -= RTE_PKTMBUF_HEADROOM;
2971 if (size < new_mtu) {
2972 rc = bnxt_hwrm_vnic_plcmode_cfg(bp, vnic);
2979 eth_dev->data->dev_conf.rxmode.max_rx_pkt_len = new_pkt_size;
2981 PMD_DRV_LOG(INFO, "New MTU is %d\n", new_mtu);
2987 bnxt_vlan_pvid_set_op(struct rte_eth_dev *dev, uint16_t pvid, int on)
2989 struct bnxt *bp = dev->data->dev_private;
2990 uint16_t vlan = bp->vlan;
2993 rc = is_bnxt_in_error(bp);
2997 if (!BNXT_SINGLE_PF(bp) || BNXT_VF(bp)) {
2999 "PVID cannot be modified for this function\n");
3002 bp->vlan = on ? pvid : 0;
3004 rc = bnxt_hwrm_set_default_vlan(bp, 0, 0);
3011 bnxt_dev_led_on_op(struct rte_eth_dev *dev)
3013 struct bnxt *bp = dev->data->dev_private;
3016 rc = is_bnxt_in_error(bp);
3020 return bnxt_hwrm_port_led_cfg(bp, true);
3024 bnxt_dev_led_off_op(struct rte_eth_dev *dev)
3026 struct bnxt *bp = dev->data->dev_private;
3029 rc = is_bnxt_in_error(bp);
3033 return bnxt_hwrm_port_led_cfg(bp, false);
3037 bnxt_rx_queue_count_op(struct rte_eth_dev *dev, uint16_t rx_queue_id)
3039 struct bnxt *bp = (struct bnxt *)dev->data->dev_private;
3040 struct bnxt_cp_ring_info *cpr;
3041 uint32_t desc = 0, raw_cons;
3042 struct bnxt_rx_queue *rxq;
3043 struct rx_pkt_cmpl *rxcmp;
3046 rc = is_bnxt_in_error(bp);
3050 rxq = dev->data->rx_queues[rx_queue_id];
3052 raw_cons = cpr->cp_raw_cons;
3055 uint32_t agg_cnt, cons, cmpl_type;
3057 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3058 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3060 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3063 cmpl_type = CMP_TYPE(rxcmp);
3065 switch (cmpl_type) {
3066 case CMPL_BASE_TYPE_RX_L2:
3067 case CMPL_BASE_TYPE_RX_L2_V2:
3068 agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3069 raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3073 case CMPL_BASE_TYPE_RX_TPA_END:
3074 if (BNXT_CHIP_P5(rxq->bp)) {
3075 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3077 p5_tpa_end = (void *)rxcmp;
3078 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3080 struct rx_tpa_end_cmpl *tpa_end;
3082 tpa_end = (void *)rxcmp;
3083 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3086 raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3091 raw_cons += CMP_LEN(cmpl_type);
3099 bnxt_rx_descriptor_status_op(void *rx_queue, uint16_t offset)
3101 struct bnxt_rx_queue *rxq = rx_queue;
3102 struct bnxt_cp_ring_info *cpr;
3103 struct bnxt_rx_ring_info *rxr;
3104 uint32_t desc, raw_cons;
3105 struct bnxt *bp = rxq->bp;
3106 struct rx_pkt_cmpl *rxcmp;
3109 rc = is_bnxt_in_error(bp);
3113 if (offset >= rxq->nb_rx_desc)
3120 * For the vector receive case, the completion at the requested
3121 * offset can be indexed directly.
3123 #if defined(RTE_ARCH_X86) || defined(RTE_ARCH_ARM64)
3124 if (bp->flags & BNXT_FLAG_RX_VECTOR_PKT_MODE) {
3125 struct rx_pkt_cmpl *rxcmp;
3128 /* Check status of completion descriptor. */
3129 raw_cons = cpr->cp_raw_cons +
3130 offset * CMP_LEN(CMPL_BASE_TYPE_RX_L2);
3131 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3132 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3134 if (CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3135 return RTE_ETH_RX_DESC_DONE;
3137 /* Check whether rx desc has an mbuf attached. */
3138 cons = RING_CMP(rxr->rx_ring_struct, raw_cons / 2);
3139 if (cons >= rxq->rxrearm_start &&
3140 cons < rxq->rxrearm_start + rxq->rxrearm_nb) {
3141 return RTE_ETH_RX_DESC_UNAVAIL;
3144 return RTE_ETH_RX_DESC_AVAIL;
3149 * For the non-vector receive case, scan the completion ring to
3150 * locate the completion descriptor for the requested offset.
3152 raw_cons = cpr->cp_raw_cons;
3155 uint32_t agg_cnt, cons, cmpl_type;
3157 cons = RING_CMP(cpr->cp_ring_struct, raw_cons);
3158 rxcmp = (struct rx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3160 if (!CMP_VALID(rxcmp, raw_cons, cpr->cp_ring_struct))
3163 cmpl_type = CMP_TYPE(rxcmp);
3165 switch (cmpl_type) {
3166 case CMPL_BASE_TYPE_RX_L2:
3167 case CMPL_BASE_TYPE_RX_L2_V2:
3168 if (desc == offset) {
3169 cons = rxcmp->opaque;
3170 if (rxr->rx_buf_ring[cons])
3171 return RTE_ETH_RX_DESC_DONE;
3173 return RTE_ETH_RX_DESC_UNAVAIL;
3175 agg_cnt = BNXT_RX_L2_AGG_BUFS(rxcmp);
3176 raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3180 case CMPL_BASE_TYPE_RX_TPA_END:
3182 return RTE_ETH_RX_DESC_DONE;
3184 if (BNXT_CHIP_P5(rxq->bp)) {
3185 struct rx_tpa_v2_end_cmpl_hi *p5_tpa_end;
3187 p5_tpa_end = (void *)rxcmp;
3188 agg_cnt = BNXT_TPA_END_AGG_BUFS_TH(p5_tpa_end);
3190 struct rx_tpa_end_cmpl *tpa_end;
3192 tpa_end = (void *)rxcmp;
3193 agg_cnt = BNXT_TPA_END_AGG_BUFS(tpa_end);
3196 raw_cons = raw_cons + CMP_LEN(cmpl_type) + agg_cnt;
3201 raw_cons += CMP_LEN(cmpl_type);
3205 return RTE_ETH_RX_DESC_AVAIL;
3209 bnxt_tx_descriptor_status_op(void *tx_queue, uint16_t offset)
3211 struct bnxt_tx_queue *txq = (struct bnxt_tx_queue *)tx_queue;
3212 struct bnxt_tx_ring_info *txr;
3213 struct bnxt_cp_ring_info *cpr;
3214 struct rte_mbuf **tx_buf;
3215 struct tx_pkt_cmpl *txcmp;
3216 uint32_t cons, cp_cons;
3222 rc = is_bnxt_in_error(txq->bp);
3229 if (offset >= txq->nb_tx_desc)
3232 cons = RING_CMP(cpr->cp_ring_struct, offset);
3233 txcmp = (struct tx_pkt_cmpl *)&cpr->cp_desc_ring[cons];
3234 cp_cons = cpr->cp_raw_cons;
3236 if (cons > cp_cons) {
3237 if (CMPL_VALID(txcmp, cpr->valid))
3238 return RTE_ETH_TX_DESC_UNAVAIL;
3240 if (CMPL_VALID(txcmp, !cpr->valid))
3241 return RTE_ETH_TX_DESC_UNAVAIL;
3243 tx_buf = &txr->tx_buf_ring[cons];
3244 if (*tx_buf == NULL)
3245 return RTE_ETH_TX_DESC_DONE;
3247 return RTE_ETH_TX_DESC_FULL;
3251 bnxt_flow_ops_get_op(struct rte_eth_dev *dev,
3252 const struct rte_flow_ops **ops)
3254 struct bnxt *bp = dev->data->dev_private;
3260 if (BNXT_ETH_DEV_IS_REPRESENTOR(dev)) {
3261 struct bnxt_representor *vfr = dev->data->dev_private;
3262 bp = vfr->parent_dev->data->dev_private;
3263 /* parent is deleted while children are still valid */
3265 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR Error\n",
3266 dev->data->port_id);
3271 ret = is_bnxt_in_error(bp);
3275 /* PMD supports thread-safe flow operations. rte_flow API
3276 * functions can avoid mutex for multi-thread safety.
3278 dev->data->dev_flags |= RTE_ETH_DEV_FLOW_OPS_THREAD_SAFE;
3280 if (BNXT_TRUFLOW_EN(bp))
3281 *ops = &bnxt_ulp_rte_flow_ops;
3283 *ops = &bnxt_flow_ops;
3288 static const uint32_t *
3289 bnxt_dev_supported_ptypes_get_op(struct rte_eth_dev *dev)
3291 static const uint32_t ptypes[] = {
3292 RTE_PTYPE_L2_ETHER_VLAN,
3293 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
3294 RTE_PTYPE_L3_IPV6_EXT_UNKNOWN,
3298 RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN,
3299 RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN,
3300 RTE_PTYPE_INNER_L4_ICMP,
3301 RTE_PTYPE_INNER_L4_TCP,
3302 RTE_PTYPE_INNER_L4_UDP,
3306 if (!dev->rx_pkt_burst)
3312 static int bnxt_map_regs(struct bnxt *bp, uint32_t *reg_arr, int count,
3315 uint32_t reg_base = *reg_arr & 0xfffff000;
3319 for (i = 0; i < count; i++) {
3320 if ((reg_arr[i] & 0xfffff000) != reg_base)
3323 win_off = BNXT_GRCPF_REG_WINDOW_BASE_OUT + (reg_win - 1) * 4;
3324 rte_write32(reg_base, (uint8_t *)bp->bar0 + win_off);
3328 static int bnxt_map_ptp_regs(struct bnxt *bp)
3330 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3334 reg_arr = ptp->rx_regs;
3335 rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_RX_REGS, 5);
3339 reg_arr = ptp->tx_regs;
3340 rc = bnxt_map_regs(bp, reg_arr, BNXT_PTP_TX_REGS, 6);
3344 for (i = 0; i < BNXT_PTP_RX_REGS; i++)
3345 ptp->rx_mapped_regs[i] = 0x5000 + (ptp->rx_regs[i] & 0xfff);
3347 for (i = 0; i < BNXT_PTP_TX_REGS; i++)
3348 ptp->tx_mapped_regs[i] = 0x6000 + (ptp->tx_regs[i] & 0xfff);
3353 static void bnxt_unmap_ptp_regs(struct bnxt *bp)
3355 rte_write32(0, (uint8_t *)bp->bar0 +
3356 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 16);
3357 rte_write32(0, (uint8_t *)bp->bar0 +
3358 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 20);
3361 static uint64_t bnxt_cc_read(struct bnxt *bp)
3365 ns = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3366 BNXT_GRCPF_REG_SYNC_TIME));
3367 ns |= (uint64_t)(rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3368 BNXT_GRCPF_REG_SYNC_TIME + 4))) << 32;
3372 static int bnxt_get_tx_ts(struct bnxt *bp, uint64_t *ts)
3374 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3377 fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3378 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3379 if (fifo & BNXT_PTP_TX_FIFO_EMPTY)
3382 fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3383 ptp->tx_mapped_regs[BNXT_PTP_TX_FIFO]));
3384 *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3385 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_L]));
3386 *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3387 ptp->tx_mapped_regs[BNXT_PTP_TX_TS_H])) << 32;
3388 rte_read32((uint8_t *)bp->bar0 + ptp->tx_mapped_regs[BNXT_PTP_TX_SEQ]);
3393 static int bnxt_get_rx_ts(struct bnxt *bp, uint64_t *ts)
3395 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3396 struct bnxt_pf_info *pf = bp->pf;
3400 fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3401 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3402 if (!(fifo & BNXT_PTP_RX_FIFO_PENDING))
3405 port_id = pf->port_id;
3406 rte_write32(1 << port_id, (uint8_t *)bp->bar0 +
3407 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO_ADV]);
3409 fifo = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3410 ptp->rx_mapped_regs[BNXT_PTP_RX_FIFO]));
3411 if (fifo & BNXT_PTP_RX_FIFO_PENDING) {
3412 /* bnxt_clr_rx_ts(bp); TBD */
3416 *ts = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3417 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_L]));
3418 *ts |= (uint64_t)rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
3419 ptp->rx_mapped_regs[BNXT_PTP_RX_TS_H])) << 32;
3425 bnxt_timesync_write_time(struct rte_eth_dev *dev, const struct timespec *ts)
3428 struct bnxt *bp = dev->data->dev_private;
3429 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3434 ns = rte_timespec_to_ns(ts);
3435 /* Set the timecounters to a new value. */
3437 ptp->tx_tstamp_tc.nsec = ns;
3438 ptp->rx_tstamp_tc.nsec = ns;
3444 bnxt_timesync_read_time(struct rte_eth_dev *dev, struct timespec *ts)
3446 struct bnxt *bp = dev->data->dev_private;
3447 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3448 uint64_t ns, systime_cycles = 0;
3454 if (BNXT_CHIP_P5(bp))
3455 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_CURRENT_TIME,
3458 systime_cycles = bnxt_cc_read(bp);
3460 ns = rte_timecounter_update(&ptp->tc, systime_cycles);
3461 *ts = rte_ns_to_timespec(ns);
3466 bnxt_timesync_enable(struct rte_eth_dev *dev)
3468 struct bnxt *bp = dev->data->dev_private;
3469 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3477 ptp->tx_tstamp_en = 1;
3478 ptp->rxctl = BNXT_PTP_MSG_EVENTS;
3480 rc = bnxt_hwrm_ptp_cfg(bp);
3484 memset(&ptp->tc, 0, sizeof(struct rte_timecounter));
3485 memset(&ptp->rx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3486 memset(&ptp->tx_tstamp_tc, 0, sizeof(struct rte_timecounter));
3488 ptp->tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3489 ptp->tc.cc_shift = shift;
3490 ptp->tc.nsec_mask = (1ULL << shift) - 1;
3492 ptp->rx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3493 ptp->rx_tstamp_tc.cc_shift = shift;
3494 ptp->rx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3496 ptp->tx_tstamp_tc.cc_mask = BNXT_CYCLECOUNTER_MASK;
3497 ptp->tx_tstamp_tc.cc_shift = shift;
3498 ptp->tx_tstamp_tc.nsec_mask = (1ULL << shift) - 1;
3500 if (!BNXT_CHIP_P5(bp))
3501 bnxt_map_ptp_regs(bp);
3503 rc = bnxt_ptp_start(bp);
3509 bnxt_timesync_disable(struct rte_eth_dev *dev)
3511 struct bnxt *bp = dev->data->dev_private;
3512 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3518 ptp->tx_tstamp_en = 0;
3521 bnxt_hwrm_ptp_cfg(bp);
3523 if (!BNXT_CHIP_P5(bp))
3524 bnxt_unmap_ptp_regs(bp);
3532 bnxt_timesync_read_rx_timestamp(struct rte_eth_dev *dev,
3533 struct timespec *timestamp,
3534 uint32_t flags __rte_unused)
3536 struct bnxt *bp = dev->data->dev_private;
3537 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3538 uint64_t rx_tstamp_cycles = 0;
3544 if (BNXT_CHIP_P5(bp))
3545 rx_tstamp_cycles = ptp->rx_timestamp;
3547 bnxt_get_rx_ts(bp, &rx_tstamp_cycles);
3549 ns = rte_timecounter_update(&ptp->rx_tstamp_tc, rx_tstamp_cycles);
3550 *timestamp = rte_ns_to_timespec(ns);
3555 bnxt_timesync_read_tx_timestamp(struct rte_eth_dev *dev,
3556 struct timespec *timestamp)
3558 struct bnxt *bp = dev->data->dev_private;
3559 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3560 uint64_t tx_tstamp_cycles = 0;
3567 if (BNXT_CHIP_P5(bp))
3568 rc = bnxt_hwrm_port_ts_query(bp, BNXT_PTP_FLAGS_PATH_TX,
3571 rc = bnxt_get_tx_ts(bp, &tx_tstamp_cycles);
3573 ns = rte_timecounter_update(&ptp->tx_tstamp_tc, tx_tstamp_cycles);
3574 *timestamp = rte_ns_to_timespec(ns);
3580 bnxt_timesync_adjust_time(struct rte_eth_dev *dev, int64_t delta)
3582 struct bnxt *bp = dev->data->dev_private;
3583 struct bnxt_ptp_cfg *ptp = bp->ptp_cfg;
3588 ptp->tc.nsec += delta;
3589 ptp->tx_tstamp_tc.nsec += delta;
3590 ptp->rx_tstamp_tc.nsec += delta;
3596 bnxt_get_eeprom_length_op(struct rte_eth_dev *dev)
3598 struct bnxt *bp = dev->data->dev_private;
3600 uint32_t dir_entries;
3601 uint32_t entry_length;
3603 rc = is_bnxt_in_error(bp);
3607 PMD_DRV_LOG(INFO, PCI_PRI_FMT "\n",
3608 bp->pdev->addr.domain, bp->pdev->addr.bus,
3609 bp->pdev->addr.devid, bp->pdev->addr.function);
3611 rc = bnxt_hwrm_nvm_get_dir_info(bp, &dir_entries, &entry_length);
3615 return dir_entries * entry_length;
3619 bnxt_get_eeprom_op(struct rte_eth_dev *dev,
3620 struct rte_dev_eeprom_info *in_eeprom)
3622 struct bnxt *bp = dev->data->dev_private;
3627 rc = is_bnxt_in_error(bp);
3631 PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3632 bp->pdev->addr.domain, bp->pdev->addr.bus,
3633 bp->pdev->addr.devid, bp->pdev->addr.function,
3634 in_eeprom->offset, in_eeprom->length);
3636 if (in_eeprom->offset == 0) /* special offset value to get directory */
3637 return bnxt_get_nvram_directory(bp, in_eeprom->length,
3640 index = in_eeprom->offset >> 24;
3641 offset = in_eeprom->offset & 0xffffff;
3644 return bnxt_hwrm_get_nvram_item(bp, index - 1, offset,
3645 in_eeprom->length, in_eeprom->data);
3650 static bool bnxt_dir_type_is_ape_bin_format(uint16_t dir_type)
3653 case BNX_DIR_TYPE_CHIMP_PATCH:
3654 case BNX_DIR_TYPE_BOOTCODE:
3655 case BNX_DIR_TYPE_BOOTCODE_2:
3656 case BNX_DIR_TYPE_APE_FW:
3657 case BNX_DIR_TYPE_APE_PATCH:
3658 case BNX_DIR_TYPE_KONG_FW:
3659 case BNX_DIR_TYPE_KONG_PATCH:
3660 case BNX_DIR_TYPE_BONO_FW:
3661 case BNX_DIR_TYPE_BONO_PATCH:
3669 static bool bnxt_dir_type_is_other_exec_format(uint16_t dir_type)
3672 case BNX_DIR_TYPE_AVS:
3673 case BNX_DIR_TYPE_EXP_ROM_MBA:
3674 case BNX_DIR_TYPE_PCIE:
3675 case BNX_DIR_TYPE_TSCF_UCODE:
3676 case BNX_DIR_TYPE_EXT_PHY:
3677 case BNX_DIR_TYPE_CCM:
3678 case BNX_DIR_TYPE_ISCSI_BOOT:
3679 case BNX_DIR_TYPE_ISCSI_BOOT_IPV6:
3680 case BNX_DIR_TYPE_ISCSI_BOOT_IPV4N6:
3688 static bool bnxt_dir_type_is_executable(uint16_t dir_type)
3690 return bnxt_dir_type_is_ape_bin_format(dir_type) ||
3691 bnxt_dir_type_is_other_exec_format(dir_type);
3695 bnxt_set_eeprom_op(struct rte_eth_dev *dev,
3696 struct rte_dev_eeprom_info *in_eeprom)
3698 struct bnxt *bp = dev->data->dev_private;
3699 uint8_t index, dir_op;
3700 uint16_t type, ext, ordinal, attr;
3703 rc = is_bnxt_in_error(bp);
3707 PMD_DRV_LOG(INFO, PCI_PRI_FMT " in_eeprom->offset = %d len = %d\n",
3708 bp->pdev->addr.domain, bp->pdev->addr.bus,
3709 bp->pdev->addr.devid, bp->pdev->addr.function,
3710 in_eeprom->offset, in_eeprom->length);
3713 PMD_DRV_LOG(ERR, "NVM write not supported from a VF\n");
3717 type = in_eeprom->magic >> 16;
3719 if (type == 0xffff) { /* special value for directory operations */
3720 index = in_eeprom->magic & 0xff;
3721 dir_op = in_eeprom->magic >> 8;
3725 case 0x0e: /* erase */
3726 if (in_eeprom->offset != ~in_eeprom->magic)
3728 return bnxt_hwrm_erase_nvram_directory(bp, index - 1);
3734 /* Create or re-write an NVM item: */
3735 if (bnxt_dir_type_is_executable(type) == true)
3737 ext = in_eeprom->magic & 0xffff;
3738 ordinal = in_eeprom->offset >> 16;
3739 attr = in_eeprom->offset & 0xffff;
3741 return bnxt_hwrm_flash_nvram(bp, type, ordinal, ext, attr,
3742 in_eeprom->data, in_eeprom->length);
3749 static const struct eth_dev_ops bnxt_dev_ops = {
3750 .dev_infos_get = bnxt_dev_info_get_op,
3751 .dev_close = bnxt_dev_close_op,
3752 .dev_configure = bnxt_dev_configure_op,
3753 .dev_start = bnxt_dev_start_op,
3754 .dev_stop = bnxt_dev_stop_op,
3755 .dev_set_link_up = bnxt_dev_set_link_up_op,
3756 .dev_set_link_down = bnxt_dev_set_link_down_op,
3757 .stats_get = bnxt_stats_get_op,
3758 .stats_reset = bnxt_stats_reset_op,
3759 .rx_queue_setup = bnxt_rx_queue_setup_op,
3760 .rx_queue_release = bnxt_rx_queue_release_op,
3761 .tx_queue_setup = bnxt_tx_queue_setup_op,
3762 .tx_queue_release = bnxt_tx_queue_release_op,
3763 .rx_queue_intr_enable = bnxt_rx_queue_intr_enable_op,
3764 .rx_queue_intr_disable = bnxt_rx_queue_intr_disable_op,
3765 .reta_update = bnxt_reta_update_op,
3766 .reta_query = bnxt_reta_query_op,
3767 .rss_hash_update = bnxt_rss_hash_update_op,
3768 .rss_hash_conf_get = bnxt_rss_hash_conf_get_op,
3769 .link_update = bnxt_link_update_op,
3770 .promiscuous_enable = bnxt_promiscuous_enable_op,
3771 .promiscuous_disable = bnxt_promiscuous_disable_op,
3772 .allmulticast_enable = bnxt_allmulticast_enable_op,
3773 .allmulticast_disable = bnxt_allmulticast_disable_op,
3774 .mac_addr_add = bnxt_mac_addr_add_op,
3775 .mac_addr_remove = bnxt_mac_addr_remove_op,
3776 .flow_ctrl_get = bnxt_flow_ctrl_get_op,
3777 .flow_ctrl_set = bnxt_flow_ctrl_set_op,
3778 .udp_tunnel_port_add = bnxt_udp_tunnel_port_add_op,
3779 .udp_tunnel_port_del = bnxt_udp_tunnel_port_del_op,
3780 .vlan_filter_set = bnxt_vlan_filter_set_op,
3781 .vlan_offload_set = bnxt_vlan_offload_set_op,
3782 .vlan_tpid_set = bnxt_vlan_tpid_set_op,
3783 .vlan_pvid_set = bnxt_vlan_pvid_set_op,
3784 .mtu_set = bnxt_mtu_set_op,
3785 .mac_addr_set = bnxt_set_default_mac_addr_op,
3786 .xstats_get = bnxt_dev_xstats_get_op,
3787 .xstats_get_names = bnxt_dev_xstats_get_names_op,
3788 .xstats_reset = bnxt_dev_xstats_reset_op,
3789 .fw_version_get = bnxt_fw_version_get,
3790 .set_mc_addr_list = bnxt_dev_set_mc_addr_list_op,
3791 .rxq_info_get = bnxt_rxq_info_get_op,
3792 .txq_info_get = bnxt_txq_info_get_op,
3793 .rx_burst_mode_get = bnxt_rx_burst_mode_get,
3794 .tx_burst_mode_get = bnxt_tx_burst_mode_get,
3795 .dev_led_on = bnxt_dev_led_on_op,
3796 .dev_led_off = bnxt_dev_led_off_op,
3797 .rx_queue_start = bnxt_rx_queue_start,
3798 .rx_queue_stop = bnxt_rx_queue_stop,
3799 .tx_queue_start = bnxt_tx_queue_start,
3800 .tx_queue_stop = bnxt_tx_queue_stop,
3801 .flow_ops_get = bnxt_flow_ops_get_op,
3802 .dev_supported_ptypes_get = bnxt_dev_supported_ptypes_get_op,
3803 .get_eeprom_length = bnxt_get_eeprom_length_op,
3804 .get_eeprom = bnxt_get_eeprom_op,
3805 .set_eeprom = bnxt_set_eeprom_op,
3806 .timesync_enable = bnxt_timesync_enable,
3807 .timesync_disable = bnxt_timesync_disable,
3808 .timesync_read_time = bnxt_timesync_read_time,
3809 .timesync_write_time = bnxt_timesync_write_time,
3810 .timesync_adjust_time = bnxt_timesync_adjust_time,
3811 .timesync_read_rx_timestamp = bnxt_timesync_read_rx_timestamp,
3812 .timesync_read_tx_timestamp = bnxt_timesync_read_tx_timestamp,
3815 static uint32_t bnxt_map_reset_regs(struct bnxt *bp, uint32_t reg)
3819 /* Only pre-map the reset GRC registers using window 3 */
3820 rte_write32(reg & 0xfffff000, (uint8_t *)bp->bar0 +
3821 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 8);
3823 offset = BNXT_GRCP_WINDOW_3_BASE + (reg & 0xffc);
3828 int bnxt_map_fw_health_status_regs(struct bnxt *bp)
3830 struct bnxt_error_recovery_info *info = bp->recovery_info;
3831 uint32_t reg_base = 0xffffffff;
3834 /* Only pre-map the monitoring GRC registers using window 2 */
3835 for (i = 0; i < BNXT_FW_STATUS_REG_CNT; i++) {
3836 uint32_t reg = info->status_regs[i];
3838 if (BNXT_FW_STATUS_REG_TYPE(reg) != BNXT_FW_STATUS_REG_TYPE_GRC)
3841 if (reg_base == 0xffffffff)
3842 reg_base = reg & 0xfffff000;
3843 if ((reg & 0xfffff000) != reg_base)
3846 /* Use mask 0xffc as the Lower 2 bits indicates
3847 * address space location
3849 info->mapped_status_regs[i] = BNXT_GRCP_WINDOW_2_BASE +
3853 if (reg_base == 0xffffffff)
3856 rte_write32(reg_base, (uint8_t *)bp->bar0 +
3857 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
3862 static void bnxt_write_fw_reset_reg(struct bnxt *bp, uint32_t index)
3864 struct bnxt_error_recovery_info *info = bp->recovery_info;
3865 uint32_t delay = info->delay_after_reset[index];
3866 uint32_t val = info->reset_reg_val[index];
3867 uint32_t reg = info->reset_reg[index];
3868 uint32_t type, offset;
3871 type = BNXT_FW_STATUS_REG_TYPE(reg);
3872 offset = BNXT_FW_STATUS_REG_OFF(reg);
3875 case BNXT_FW_STATUS_REG_TYPE_CFG:
3876 ret = rte_pci_write_config(bp->pdev, &val, sizeof(val), offset);
3878 PMD_DRV_LOG(ERR, "Failed to write %#x at PCI offset %#x",
3883 case BNXT_FW_STATUS_REG_TYPE_GRC:
3884 offset = bnxt_map_reset_regs(bp, offset);
3885 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3887 case BNXT_FW_STATUS_REG_TYPE_BAR0:
3888 rte_write32(val, (uint8_t *)bp->bar0 + offset);
3891 /* wait on a specific interval of time until core reset is complete */
3893 rte_delay_ms(delay);
3896 static void bnxt_dev_cleanup(struct bnxt *bp)
3898 bp->eth_dev->data->dev_link.link_status = 0;
3899 bp->link_info->link_up = 0;
3900 if (bp->eth_dev->data->dev_started)
3901 bnxt_dev_stop(bp->eth_dev);
3903 bnxt_uninit_resources(bp, true);
3907 bnxt_check_fw_reset_done(struct bnxt *bp)
3909 int timeout = bp->fw_reset_max_msecs;
3914 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
3916 PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
3922 } while (timeout--);
3924 if (val == 0xffff) {
3925 PMD_DRV_LOG(ERR, "Firmware reset aborted, PCI config space invalid\n");
3932 static int bnxt_restore_vlan_filters(struct bnxt *bp)
3934 struct rte_eth_dev *dev = bp->eth_dev;
3935 struct rte_vlan_filter_conf *vfc;
3939 for (vlan_id = 1; vlan_id <= RTE_ETHER_MAX_VLAN_ID; vlan_id++) {
3940 vfc = &dev->data->vlan_filter_conf;
3941 vidx = vlan_id / 64;
3942 vbit = vlan_id % 64;
3944 /* Each bit corresponds to a VLAN id */
3945 if (vfc->ids[vidx] & (UINT64_C(1) << vbit)) {
3946 rc = bnxt_add_vlan_filter(bp, vlan_id);
3955 static int bnxt_restore_mac_filters(struct bnxt *bp)
3957 struct rte_eth_dev *dev = bp->eth_dev;
3958 struct rte_eth_dev_info dev_info;
3959 struct rte_ether_addr *addr;
3965 if (BNXT_VF(bp) && !BNXT_VF_IS_TRUSTED(bp))
3968 rc = bnxt_dev_info_get_op(dev, &dev_info);
3972 /* replay MAC address configuration */
3973 for (i = 1; i < dev_info.max_mac_addrs; i++) {
3974 addr = &dev->data->mac_addrs[i];
3976 /* skip zero address */
3977 if (rte_is_zero_ether_addr(addr))
3981 pool_mask = dev->data->mac_pool_sel[i];
3984 if (pool_mask & 1ULL) {
3985 rc = bnxt_mac_addr_add_op(dev, addr, i, pool);
3991 } while (pool_mask);
3997 static int bnxt_restore_filters(struct bnxt *bp)
3999 struct rte_eth_dev *dev = bp->eth_dev;
4002 if (dev->data->all_multicast) {
4003 ret = bnxt_allmulticast_enable_op(dev);
4007 if (dev->data->promiscuous) {
4008 ret = bnxt_promiscuous_enable_op(dev);
4013 ret = bnxt_restore_mac_filters(bp);
4017 ret = bnxt_restore_vlan_filters(bp);
4018 /* TODO restore other filters as well */
4022 static int bnxt_check_fw_ready(struct bnxt *bp)
4024 int timeout = bp->fw_reset_max_msecs;
4028 rc = bnxt_hwrm_poll_ver_get(bp);
4031 rte_delay_ms(BNXT_FW_READY_WAIT_INTERVAL);
4032 timeout -= BNXT_FW_READY_WAIT_INTERVAL;
4033 } while (rc && timeout > 0);
4036 PMD_DRV_LOG(ERR, "FW is not Ready after reset\n");
4041 static void bnxt_dev_recover(void *arg)
4043 struct bnxt *bp = arg;
4046 pthread_mutex_lock(&bp->err_recovery_lock);
4048 if (!bp->fw_reset_min_msecs) {
4049 rc = bnxt_check_fw_reset_done(bp);
4054 /* Clear Error flag so that device re-init should happen */
4055 bp->flags &= ~BNXT_FLAG_FATAL_ERROR;
4057 rc = bnxt_check_fw_ready(bp);
4061 rc = bnxt_init_resources(bp, true);
4064 "Failed to initialize resources after reset\n");
4067 /* clear reset flag as the device is initialized now */
4068 bp->flags &= ~BNXT_FLAG_FW_RESET;
4070 rc = bnxt_dev_start_op(bp->eth_dev);
4072 PMD_DRV_LOG(ERR, "Failed to start port after reset\n");
4076 rc = bnxt_restore_filters(bp);
4080 PMD_DRV_LOG(INFO, "Recovered from FW reset\n");
4081 pthread_mutex_unlock(&bp->err_recovery_lock);
4085 bnxt_dev_stop(bp->eth_dev);
4087 bp->flags |= BNXT_FLAG_FATAL_ERROR;
4088 bnxt_uninit_resources(bp, false);
4089 pthread_mutex_unlock(&bp->err_recovery_lock);
4090 PMD_DRV_LOG(ERR, "Failed to recover from FW reset\n");
4093 void bnxt_dev_reset_and_resume(void *arg)
4095 struct bnxt *bp = arg;
4096 uint32_t us = US_PER_MS * bp->fw_reset_min_msecs;
4100 bnxt_dev_cleanup(bp);
4102 bnxt_wait_for_device_shutdown(bp);
4104 /* During some fatal firmware error conditions, the PCI config space
4105 * register 0x2e which normally contains the subsystem ID will become
4106 * 0xffff. This register will revert back to the normal value after
4107 * the chip has completed core reset. If we detect this condition,
4108 * we can poll this config register immediately for the value to revert.
4110 if (bp->flags & BNXT_FLAG_FATAL_ERROR) {
4111 rc = rte_pci_read_config(bp->pdev, &val, sizeof(val), PCI_SUBSYSTEM_ID_OFFSET);
4113 PMD_DRV_LOG(ERR, "Failed to read PCI offset 0x%x", PCI_SUBSYSTEM_ID_OFFSET);
4116 if (val == 0xffff) {
4117 bp->fw_reset_min_msecs = 0;
4122 rc = rte_eal_alarm_set(us, bnxt_dev_recover, (void *)bp);
4124 PMD_DRV_LOG(ERR, "Error setting recovery alarm");
4127 uint32_t bnxt_read_fw_status_reg(struct bnxt *bp, uint32_t index)
4129 struct bnxt_error_recovery_info *info = bp->recovery_info;
4130 uint32_t reg = info->status_regs[index];
4131 uint32_t type, offset, val = 0;
4133 type = BNXT_FW_STATUS_REG_TYPE(reg);
4134 offset = BNXT_FW_STATUS_REG_OFF(reg);
4137 case BNXT_FW_STATUS_REG_TYPE_CFG:
4138 rte_pci_read_config(bp->pdev, &val, sizeof(val), offset);
4140 case BNXT_FW_STATUS_REG_TYPE_GRC:
4141 offset = info->mapped_status_regs[index];
4143 case BNXT_FW_STATUS_REG_TYPE_BAR0:
4144 val = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4152 static int bnxt_fw_reset_all(struct bnxt *bp)
4154 struct bnxt_error_recovery_info *info = bp->recovery_info;
4158 if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4159 /* Reset through master function driver */
4160 for (i = 0; i < info->reg_array_cnt; i++)
4161 bnxt_write_fw_reset_reg(bp, i);
4162 /* Wait for time specified by FW after triggering reset */
4163 rte_delay_ms(info->master_func_wait_period_after_reset);
4164 } else if (info->flags & BNXT_FLAG_ERROR_RECOVERY_CO_CPU) {
4165 /* Reset with the help of Kong processor */
4166 rc = bnxt_hwrm_fw_reset(bp);
4168 PMD_DRV_LOG(ERR, "Failed to reset FW\n");
4174 static void bnxt_fw_reset_cb(void *arg)
4176 struct bnxt *bp = arg;
4177 struct bnxt_error_recovery_info *info = bp->recovery_info;
4180 /* Only Master function can do FW reset */
4181 if (bnxt_is_master_func(bp) &&
4182 bnxt_is_recovery_enabled(bp)) {
4183 rc = bnxt_fw_reset_all(bp);
4185 PMD_DRV_LOG(ERR, "Adapter recovery failed\n");
4190 /* if recovery method is ERROR_RECOVERY_CO_CPU, KONG will send
4191 * EXCEPTION_FATAL_ASYNC event to all the functions
4192 * (including MASTER FUNC). After receiving this Async, all the active
4193 * drivers should treat this case as FW initiated recovery
4195 if (info->flags & BNXT_FLAG_ERROR_RECOVERY_HOST) {
4196 bp->fw_reset_min_msecs = BNXT_MIN_FW_READY_TIMEOUT;
4197 bp->fw_reset_max_msecs = BNXT_MAX_FW_RESET_TIMEOUT;
4199 /* To recover from error */
4200 rte_eal_alarm_set(US_PER_MS, bnxt_dev_reset_and_resume,
4205 /* Driver should poll FW heartbeat, reset_counter with the frequency
4206 * advertised by FW in HWRM_ERROR_RECOVERY_QCFG.
4207 * When the driver detects heartbeat stop or change in reset_counter,
4208 * it has to trigger a reset to recover from the error condition.
4209 * A “master PF” is the function who will have the privilege to
4210 * initiate the chimp reset. The master PF will be elected by the
4211 * firmware and will be notified through async message.
4213 static void bnxt_check_fw_health(void *arg)
4215 struct bnxt *bp = arg;
4216 struct bnxt_error_recovery_info *info = bp->recovery_info;
4217 uint32_t val = 0, wait_msec;
4219 if (!info || !bnxt_is_recovery_enabled(bp) ||
4220 is_bnxt_in_error(bp))
4223 val = bnxt_read_fw_status_reg(bp, BNXT_FW_HEARTBEAT_CNT_REG);
4224 if (val == info->last_heart_beat)
4227 info->last_heart_beat = val;
4229 val = bnxt_read_fw_status_reg(bp, BNXT_FW_RECOVERY_CNT_REG);
4230 if (val != info->last_reset_counter)
4233 info->last_reset_counter = val;
4235 rte_eal_alarm_set(US_PER_MS * info->driver_polling_freq,
4236 bnxt_check_fw_health, (void *)bp);
4240 /* Stop DMA to/from device */
4241 bp->flags |= BNXT_FLAG_FATAL_ERROR;
4242 bp->flags |= BNXT_FLAG_FW_RESET;
4244 PMD_DRV_LOG(ERR, "Detected FW dead condition\n");
4246 if (bnxt_is_master_func(bp))
4247 wait_msec = info->master_func_wait_period;
4249 wait_msec = info->normal_func_wait_period;
4251 rte_eal_alarm_set(US_PER_MS * wait_msec,
4252 bnxt_fw_reset_cb, (void *)bp);
4255 void bnxt_schedule_fw_health_check(struct bnxt *bp)
4257 uint32_t polling_freq;
4259 pthread_mutex_lock(&bp->health_check_lock);
4261 if (!bnxt_is_recovery_enabled(bp))
4264 if (bp->flags & BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED)
4267 polling_freq = bp->recovery_info->driver_polling_freq;
4269 rte_eal_alarm_set(US_PER_MS * polling_freq,
4270 bnxt_check_fw_health, (void *)bp);
4271 bp->flags |= BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4274 pthread_mutex_unlock(&bp->health_check_lock);
4277 static void bnxt_cancel_fw_health_check(struct bnxt *bp)
4279 rte_eal_alarm_cancel(bnxt_check_fw_health, (void *)bp);
4280 bp->flags &= ~BNXT_FLAG_FW_HEALTH_CHECK_SCHEDULED;
4283 static bool bnxt_vf_pciid(uint16_t device_id)
4285 switch (device_id) {
4286 case BROADCOM_DEV_ID_57304_VF:
4287 case BROADCOM_DEV_ID_57406_VF:
4288 case BROADCOM_DEV_ID_5731X_VF:
4289 case BROADCOM_DEV_ID_5741X_VF:
4290 case BROADCOM_DEV_ID_57414_VF:
4291 case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4292 case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4293 case BROADCOM_DEV_ID_58802_VF:
4294 case BROADCOM_DEV_ID_57500_VF1:
4295 case BROADCOM_DEV_ID_57500_VF2:
4296 case BROADCOM_DEV_ID_58818_VF:
4304 /* Phase 5 device */
4305 static bool bnxt_p5_device(uint16_t device_id)
4307 switch (device_id) {
4308 case BROADCOM_DEV_ID_57508:
4309 case BROADCOM_DEV_ID_57504:
4310 case BROADCOM_DEV_ID_57502:
4311 case BROADCOM_DEV_ID_57508_MF1:
4312 case BROADCOM_DEV_ID_57504_MF1:
4313 case BROADCOM_DEV_ID_57502_MF1:
4314 case BROADCOM_DEV_ID_57508_MF2:
4315 case BROADCOM_DEV_ID_57504_MF2:
4316 case BROADCOM_DEV_ID_57502_MF2:
4317 case BROADCOM_DEV_ID_57500_VF1:
4318 case BROADCOM_DEV_ID_57500_VF2:
4319 case BROADCOM_DEV_ID_58812:
4320 case BROADCOM_DEV_ID_58814:
4321 case BROADCOM_DEV_ID_58818:
4322 case BROADCOM_DEV_ID_58818_VF:
4330 bool bnxt_stratus_device(struct bnxt *bp)
4332 uint16_t device_id = bp->pdev->id.device_id;
4334 switch (device_id) {
4335 case BROADCOM_DEV_ID_STRATUS_NIC:
4336 case BROADCOM_DEV_ID_STRATUS_NIC_VF1:
4337 case BROADCOM_DEV_ID_STRATUS_NIC_VF2:
4345 static int bnxt_map_pci_bars(struct rte_eth_dev *eth_dev)
4347 struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
4348 struct bnxt *bp = eth_dev->data->dev_private;
4350 /* enable device (incl. PCI PM wakeup), and bus-mastering */
4351 bp->bar0 = (void *)pci_dev->mem_resource[0].addr;
4352 bp->doorbell_base = (void *)pci_dev->mem_resource[2].addr;
4353 if (!bp->bar0 || !bp->doorbell_base) {
4354 PMD_DRV_LOG(ERR, "Unable to access Hardware\n");
4358 bp->eth_dev = eth_dev;
4364 static int bnxt_alloc_ctx_mem_blk(struct bnxt *bp,
4365 struct bnxt_ctx_pg_info *ctx_pg,
4370 struct bnxt_ring_mem_info *rmem = &ctx_pg->ring_mem;
4371 const struct rte_memzone *mz = NULL;
4372 char mz_name[RTE_MEMZONE_NAMESIZE];
4373 rte_iova_t mz_phys_addr;
4374 uint64_t valid_bits = 0;
4381 rmem->nr_pages = RTE_ALIGN_MUL_CEIL(mem_size, BNXT_PAGE_SIZE) /
4383 rmem->page_size = BNXT_PAGE_SIZE;
4384 rmem->pg_arr = ctx_pg->ctx_pg_arr;
4385 rmem->dma_arr = ctx_pg->ctx_dma_arr;
4386 rmem->flags = BNXT_RMEM_VALID_PTE_FLAG;
4388 valid_bits = PTU_PTE_VALID;
4390 if (rmem->nr_pages > 1) {
4391 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4392 "bnxt_ctx_pg_tbl%s_%x_%d",
4393 suffix, idx, bp->eth_dev->data->port_id);
4394 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4395 mz = rte_memzone_lookup(mz_name);
4397 mz = rte_memzone_reserve_aligned(mz_name,
4401 RTE_MEMZONE_SIZE_HINT_ONLY |
4402 RTE_MEMZONE_IOVA_CONTIG,
4408 memset(mz->addr, 0, mz->len);
4409 mz_phys_addr = mz->iova;
4411 rmem->pg_tbl = mz->addr;
4412 rmem->pg_tbl_map = mz_phys_addr;
4413 rmem->pg_tbl_mz = mz;
4416 snprintf(mz_name, RTE_MEMZONE_NAMESIZE, "bnxt_ctx_%s_%x_%d",
4417 suffix, idx, bp->eth_dev->data->port_id);
4418 mz = rte_memzone_lookup(mz_name);
4420 mz = rte_memzone_reserve_aligned(mz_name,
4424 RTE_MEMZONE_SIZE_HINT_ONLY |
4425 RTE_MEMZONE_IOVA_CONTIG,
4431 memset(mz->addr, 0, mz->len);
4432 mz_phys_addr = mz->iova;
4434 for (sz = 0, i = 0; sz < mem_size; sz += BNXT_PAGE_SIZE, i++) {
4435 rmem->pg_arr[i] = ((char *)mz->addr) + sz;
4436 rmem->dma_arr[i] = mz_phys_addr + sz;
4438 if (rmem->nr_pages > 1) {
4439 if (i == rmem->nr_pages - 2 &&
4440 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4441 valid_bits |= PTU_PTE_NEXT_TO_LAST;
4442 else if (i == rmem->nr_pages - 1 &&
4443 (rmem->flags & BNXT_RMEM_RING_PTE_FLAG))
4444 valid_bits |= PTU_PTE_LAST;
4446 rmem->pg_tbl[i] = rte_cpu_to_le_64(rmem->dma_arr[i] |
4452 if (rmem->vmem_size)
4453 rmem->vmem = (void **)mz->addr;
4454 rmem->dma_arr[0] = mz_phys_addr;
4458 static void bnxt_free_ctx_mem(struct bnxt *bp)
4462 if (!bp->ctx || !(bp->ctx->flags & BNXT_CTX_FLAG_INITED))
4465 bp->ctx->flags &= ~BNXT_CTX_FLAG_INITED;
4466 rte_memzone_free(bp->ctx->qp_mem.ring_mem.mz);
4467 rte_memzone_free(bp->ctx->srq_mem.ring_mem.mz);
4468 rte_memzone_free(bp->ctx->cq_mem.ring_mem.mz);
4469 rte_memzone_free(bp->ctx->vnic_mem.ring_mem.mz);
4470 rte_memzone_free(bp->ctx->stat_mem.ring_mem.mz);
4471 rte_memzone_free(bp->ctx->qp_mem.ring_mem.pg_tbl_mz);
4472 rte_memzone_free(bp->ctx->srq_mem.ring_mem.pg_tbl_mz);
4473 rte_memzone_free(bp->ctx->cq_mem.ring_mem.pg_tbl_mz);
4474 rte_memzone_free(bp->ctx->vnic_mem.ring_mem.pg_tbl_mz);
4475 rte_memzone_free(bp->ctx->stat_mem.ring_mem.pg_tbl_mz);
4477 for (i = 0; i < bp->ctx->tqm_fp_rings_count + 1; i++) {
4478 if (bp->ctx->tqm_mem[i])
4479 rte_memzone_free(bp->ctx->tqm_mem[i]->ring_mem.mz);
4486 #define bnxt_roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
4488 #define min_t(type, x, y) ({ \
4489 type __min1 = (x); \
4490 type __min2 = (y); \
4491 __min1 < __min2 ? __min1 : __min2; })
4493 #define max_t(type, x, y) ({ \
4494 type __max1 = (x); \
4495 type __max2 = (y); \
4496 __max1 > __max2 ? __max1 : __max2; })
4498 #define clamp_t(type, _x, min, max) min_t(type, max_t(type, _x, min), max)
4500 int bnxt_alloc_ctx_mem(struct bnxt *bp)
4502 struct bnxt_ctx_pg_info *ctx_pg;
4503 struct bnxt_ctx_mem_info *ctx;
4504 uint32_t mem_size, ena, entries;
4505 uint32_t entries_sp, min;
4508 rc = bnxt_hwrm_func_backing_store_qcaps(bp);
4510 PMD_DRV_LOG(ERR, "Query context mem capability failed\n");
4514 if (!ctx || (ctx->flags & BNXT_CTX_FLAG_INITED))
4517 ctx_pg = &ctx->qp_mem;
4518 ctx_pg->entries = ctx->qp_min_qp1_entries + ctx->qp_max_l2_entries;
4519 if (ctx->qp_entry_size) {
4520 mem_size = ctx->qp_entry_size * ctx_pg->entries;
4521 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "qp_mem", 0);
4526 ctx_pg = &ctx->srq_mem;
4527 ctx_pg->entries = ctx->srq_max_l2_entries;
4528 if (ctx->srq_entry_size) {
4529 mem_size = ctx->srq_entry_size * ctx_pg->entries;
4530 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "srq_mem", 0);
4535 ctx_pg = &ctx->cq_mem;
4536 ctx_pg->entries = ctx->cq_max_l2_entries;
4537 if (ctx->cq_entry_size) {
4538 mem_size = ctx->cq_entry_size * ctx_pg->entries;
4539 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "cq_mem", 0);
4544 ctx_pg = &ctx->vnic_mem;
4545 ctx_pg->entries = ctx->vnic_max_vnic_entries +
4546 ctx->vnic_max_ring_table_entries;
4547 if (ctx->vnic_entry_size) {
4548 mem_size = ctx->vnic_entry_size * ctx_pg->entries;
4549 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "vnic_mem", 0);
4554 ctx_pg = &ctx->stat_mem;
4555 ctx_pg->entries = ctx->stat_max_entries;
4556 if (ctx->stat_entry_size) {
4557 mem_size = ctx->stat_entry_size * ctx_pg->entries;
4558 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size, "stat_mem", 0);
4563 min = ctx->tqm_min_entries_per_ring;
4565 entries_sp = ctx->qp_max_l2_entries +
4566 ctx->vnic_max_vnic_entries +
4567 2 * ctx->qp_min_qp1_entries + min;
4568 entries_sp = bnxt_roundup(entries_sp, ctx->tqm_entries_multiple);
4570 entries = ctx->qp_max_l2_entries + ctx->qp_min_qp1_entries;
4571 entries = bnxt_roundup(entries, ctx->tqm_entries_multiple);
4572 entries = clamp_t(uint32_t, entries, min,
4573 ctx->tqm_max_entries_per_ring);
4574 for (i = 0, ena = 0; i < ctx->tqm_fp_rings_count + 1; i++) {
4575 /* i=0 is for TQM_SP. i=1 to i=8 applies to RING0 to RING7.
4576 * i > 8 is other ext rings.
4578 ctx_pg = ctx->tqm_mem[i];
4579 ctx_pg->entries = i ? entries : entries_sp;
4580 if (ctx->tqm_entry_size) {
4581 mem_size = ctx->tqm_entry_size * ctx_pg->entries;
4582 rc = bnxt_alloc_ctx_mem_blk(bp, ctx_pg, mem_size,
4587 if (i < BNXT_MAX_TQM_LEGACY_RINGS)
4588 ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_SP << i;
4590 ena |= HWRM_FUNC_BACKING_STORE_CFG_INPUT_ENABLES_TQM_RING8;
4593 ena |= FUNC_BACKING_STORE_CFG_INPUT_DFLT_ENABLES;
4594 rc = bnxt_hwrm_func_backing_store_cfg(bp, ena);
4597 "Failed to configure context mem: rc = %d\n", rc);
4599 ctx->flags |= BNXT_CTX_FLAG_INITED;
4604 static int bnxt_alloc_stats_mem(struct bnxt *bp)
4606 struct rte_pci_device *pci_dev = bp->pdev;
4607 char mz_name[RTE_MEMZONE_NAMESIZE];
4608 const struct rte_memzone *mz = NULL;
4609 uint32_t total_alloc_len;
4610 rte_iova_t mz_phys_addr;
4612 if (pci_dev->id.device_id == BROADCOM_DEV_ID_NS2)
4615 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4616 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4617 pci_dev->addr.bus, pci_dev->addr.devid,
4618 pci_dev->addr.function, "rx_port_stats");
4619 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4620 mz = rte_memzone_lookup(mz_name);
4622 RTE_CACHE_LINE_ROUNDUP(sizeof(struct rx_port_stats) +
4623 sizeof(struct rx_port_stats_ext) + 512);
4625 mz = rte_memzone_reserve(mz_name, total_alloc_len,
4628 RTE_MEMZONE_SIZE_HINT_ONLY |
4629 RTE_MEMZONE_IOVA_CONTIG);
4633 memset(mz->addr, 0, mz->len);
4634 mz_phys_addr = mz->iova;
4636 bp->rx_mem_zone = (const void *)mz;
4637 bp->hw_rx_port_stats = mz->addr;
4638 bp->hw_rx_port_stats_map = mz_phys_addr;
4640 snprintf(mz_name, RTE_MEMZONE_NAMESIZE,
4641 "bnxt_" PCI_PRI_FMT "-%s", pci_dev->addr.domain,
4642 pci_dev->addr.bus, pci_dev->addr.devid,
4643 pci_dev->addr.function, "tx_port_stats");
4644 mz_name[RTE_MEMZONE_NAMESIZE - 1] = 0;
4645 mz = rte_memzone_lookup(mz_name);
4647 RTE_CACHE_LINE_ROUNDUP(sizeof(struct tx_port_stats) +
4648 sizeof(struct tx_port_stats_ext) + 512);
4650 mz = rte_memzone_reserve(mz_name,
4654 RTE_MEMZONE_SIZE_HINT_ONLY |
4655 RTE_MEMZONE_IOVA_CONTIG);
4659 memset(mz->addr, 0, mz->len);
4660 mz_phys_addr = mz->iova;
4662 bp->tx_mem_zone = (const void *)mz;
4663 bp->hw_tx_port_stats = mz->addr;
4664 bp->hw_tx_port_stats_map = mz_phys_addr;
4665 bp->flags |= BNXT_FLAG_PORT_STATS;
4667 /* Display extended statistics if FW supports it */
4668 if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_8_4 ||
4669 bp->hwrm_spec_code == HWRM_SPEC_CODE_1_9_0 ||
4670 !(bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED))
4673 bp->hw_rx_port_stats_ext = (void *)
4674 ((uint8_t *)bp->hw_rx_port_stats +
4675 sizeof(struct rx_port_stats));
4676 bp->hw_rx_port_stats_ext_map = bp->hw_rx_port_stats_map +
4677 sizeof(struct rx_port_stats);
4678 bp->flags |= BNXT_FLAG_EXT_RX_PORT_STATS;
4680 if (bp->hwrm_spec_code < HWRM_SPEC_CODE_1_9_2 ||
4681 bp->flags & BNXT_FLAG_EXT_STATS_SUPPORTED) {
4682 bp->hw_tx_port_stats_ext = (void *)
4683 ((uint8_t *)bp->hw_tx_port_stats +
4684 sizeof(struct tx_port_stats));
4685 bp->hw_tx_port_stats_ext_map =
4686 bp->hw_tx_port_stats_map +
4687 sizeof(struct tx_port_stats);
4688 bp->flags |= BNXT_FLAG_EXT_TX_PORT_STATS;
4694 static int bnxt_setup_mac_addr(struct rte_eth_dev *eth_dev)
4696 struct bnxt *bp = eth_dev->data->dev_private;
4699 eth_dev->data->mac_addrs = rte_zmalloc("bnxt_mac_addr_tbl",
4700 RTE_ETHER_ADDR_LEN *
4703 if (eth_dev->data->mac_addrs == NULL) {
4704 PMD_DRV_LOG(ERR, "Failed to alloc MAC addr tbl\n");
4708 if (!BNXT_HAS_DFLT_MAC_SET(bp)) {
4712 /* Generate a random MAC address, if none was assigned by PF */
4713 PMD_DRV_LOG(INFO, "VF MAC address not assigned by Host PF\n");
4714 bnxt_eth_hw_addr_random(bp->mac_addr);
4716 "Assign random MAC:%02X:%02X:%02X:%02X:%02X:%02X\n",
4717 bp->mac_addr[0], bp->mac_addr[1], bp->mac_addr[2],
4718 bp->mac_addr[3], bp->mac_addr[4], bp->mac_addr[5]);
4720 rc = bnxt_hwrm_set_mac(bp);
4725 /* Copy the permanent MAC from the FUNC_QCAPS response */
4726 memcpy(ð_dev->data->mac_addrs[0], bp->mac_addr, RTE_ETHER_ADDR_LEN);
4731 static int bnxt_restore_dflt_mac(struct bnxt *bp)
4735 /* MAC is already configured in FW */
4736 if (BNXT_HAS_DFLT_MAC_SET(bp))
4739 /* Restore the old MAC configured */
4740 rc = bnxt_hwrm_set_mac(bp);
4742 PMD_DRV_LOG(ERR, "Failed to restore MAC address\n");
4747 static void bnxt_config_vf_req_fwd(struct bnxt *bp)
4752 memset(bp->pf->vf_req_fwd, 0, sizeof(bp->pf->vf_req_fwd));
4754 if (!(bp->fw_cap & BNXT_FW_CAP_LINK_ADMIN))
4755 BNXT_HWRM_CMD_TO_FORWARD(HWRM_PORT_PHY_QCFG);
4756 BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_CFG);
4757 BNXT_HWRM_CMD_TO_FORWARD(HWRM_FUNC_VF_CFG);
4758 BNXT_HWRM_CMD_TO_FORWARD(HWRM_CFA_L2_FILTER_ALLOC);
4759 BNXT_HWRM_CMD_TO_FORWARD(HWRM_OEM_CMD);
4763 bnxt_get_svif(uint16_t port_id, bool func_svif,
4764 enum bnxt_ulp_intf_type type)
4766 struct rte_eth_dev *eth_dev;
4769 eth_dev = &rte_eth_devices[port_id];
4770 if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4771 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4775 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4778 eth_dev = vfr->parent_dev;
4781 bp = eth_dev->data->dev_private;
4783 return func_svif ? bp->func_svif : bp->port_svif;
4787 bnxt_get_vnic_id(uint16_t port, enum bnxt_ulp_intf_type type)
4789 struct rte_eth_dev *eth_dev;
4790 struct bnxt_vnic_info *vnic;
4793 eth_dev = &rte_eth_devices[port];
4794 if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4795 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4799 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4800 return vfr->dflt_vnic_id;
4802 eth_dev = vfr->parent_dev;
4805 bp = eth_dev->data->dev_private;
4807 vnic = BNXT_GET_DEFAULT_VNIC(bp);
4809 return vnic->fw_vnic_id;
4813 bnxt_get_fw_func_id(uint16_t port, enum bnxt_ulp_intf_type type)
4815 struct rte_eth_dev *eth_dev;
4818 eth_dev = &rte_eth_devices[port];
4819 if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4820 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4824 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4827 eth_dev = vfr->parent_dev;
4830 bp = eth_dev->data->dev_private;
4835 enum bnxt_ulp_intf_type
4836 bnxt_get_interface_type(uint16_t port)
4838 struct rte_eth_dev *eth_dev;
4841 eth_dev = &rte_eth_devices[port];
4842 if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev))
4843 return BNXT_ULP_INTF_TYPE_VF_REP;
4845 bp = eth_dev->data->dev_private;
4847 return BNXT_ULP_INTF_TYPE_PF;
4848 else if (BNXT_VF_IS_TRUSTED(bp))
4849 return BNXT_ULP_INTF_TYPE_TRUSTED_VF;
4850 else if (BNXT_VF(bp))
4851 return BNXT_ULP_INTF_TYPE_VF;
4853 return BNXT_ULP_INTF_TYPE_INVALID;
4857 bnxt_get_phy_port_id(uint16_t port_id)
4859 struct bnxt_representor *vfr;
4860 struct rte_eth_dev *eth_dev;
4863 eth_dev = &rte_eth_devices[port_id];
4864 if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4865 vfr = eth_dev->data->dev_private;
4869 eth_dev = vfr->parent_dev;
4872 bp = eth_dev->data->dev_private;
4874 return BNXT_PF(bp) ? bp->pf->port_id : bp->parent->port_id;
4878 bnxt_get_parif(uint16_t port_id, enum bnxt_ulp_intf_type type)
4880 struct rte_eth_dev *eth_dev;
4883 eth_dev = &rte_eth_devices[port_id];
4884 if (BNXT_ETH_DEV_IS_REPRESENTOR(eth_dev)) {
4885 struct bnxt_representor *vfr = eth_dev->data->dev_private;
4889 if (type == BNXT_ULP_INTF_TYPE_VF_REP)
4890 return vfr->fw_fid - 1;
4892 eth_dev = vfr->parent_dev;
4895 bp = eth_dev->data->dev_private;
4897 return BNXT_PF(bp) ? bp->fw_fid - 1 : bp->parent->fid - 1;
4901 bnxt_get_vport(uint16_t port_id)
4903 return (1 << bnxt_get_phy_port_id(port_id));
4906 static void bnxt_alloc_error_recovery_info(struct bnxt *bp)
4908 struct bnxt_error_recovery_info *info = bp->recovery_info;
4911 if (!(bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS))
4912 memset(info, 0, sizeof(*info));
4916 if (!(bp->fw_cap & BNXT_FW_CAP_ERROR_RECOVERY))
4919 info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4922 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
4924 bp->recovery_info = info;
4927 static void bnxt_check_fw_status(struct bnxt *bp)
4931 if (!(bp->recovery_info &&
4932 (bp->fw_cap & BNXT_FW_CAP_HCOMM_FW_STATUS)))
4935 fw_status = bnxt_read_fw_status_reg(bp, BNXT_FW_STATUS_REG);
4936 if (fw_status != BNXT_FW_STATUS_HEALTHY)
4937 PMD_DRV_LOG(ERR, "Firmware not responding, status: %#x\n",
4941 static int bnxt_map_hcomm_fw_status_reg(struct bnxt *bp)
4943 struct bnxt_error_recovery_info *info = bp->recovery_info;
4944 uint32_t status_loc;
4947 rte_write32(HCOMM_STATUS_STRUCT_LOC, (uint8_t *)bp->bar0 +
4948 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4949 sig_ver = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4950 BNXT_GRCP_WINDOW_2_BASE +
4951 offsetof(struct hcomm_status,
4953 /* If the signature is absent, then FW does not support this feature */
4954 if ((sig_ver & HCOMM_STATUS_SIGNATURE_MASK) !=
4955 HCOMM_STATUS_SIGNATURE_VAL)
4959 info = rte_zmalloc("bnxt_hwrm_error_recovery_qcfg",
4963 bp->recovery_info = info;
4965 memset(info, 0, sizeof(*info));
4968 status_loc = rte_le_to_cpu_32(rte_read32((uint8_t *)bp->bar0 +
4969 BNXT_GRCP_WINDOW_2_BASE +
4970 offsetof(struct hcomm_status,
4973 /* Only pre-map the FW health status GRC register */
4974 if (BNXT_FW_STATUS_REG_TYPE(status_loc) != BNXT_FW_STATUS_REG_TYPE_GRC)
4977 info->status_regs[BNXT_FW_STATUS_REG] = status_loc;
4978 info->mapped_status_regs[BNXT_FW_STATUS_REG] =
4979 BNXT_GRCP_WINDOW_2_BASE + (status_loc & BNXT_GRCP_OFFSET_MASK);
4981 rte_write32((status_loc & BNXT_GRCP_BASE_MASK), (uint8_t *)bp->bar0 +
4982 BNXT_GRCPF_REG_WINDOW_BASE_OUT + 4);
4984 bp->fw_cap |= BNXT_FW_CAP_HCOMM_FW_STATUS;
4989 /* This function gets the FW version along with the
4990 * capabilities(MAX and current) of the function, vnic,
4991 * error recovery, phy and other chip related info
4993 static int bnxt_get_config(struct bnxt *bp)
5000 rc = bnxt_map_hcomm_fw_status_reg(bp);
5004 rc = bnxt_hwrm_ver_get(bp, DFLT_HWRM_CMD_TIMEOUT);
5006 bnxt_check_fw_status(bp);
5010 rc = bnxt_hwrm_func_reset(bp);
5014 rc = bnxt_hwrm_vnic_qcaps(bp);
5018 rc = bnxt_hwrm_queue_qportcfg(bp);
5022 /* Get the MAX capabilities for this function.
5023 * This function also allocates context memory for TQM rings and
5024 * informs the firmware about this allocated backing store memory.
5026 rc = bnxt_hwrm_func_qcaps(bp);
5030 rc = bnxt_hwrm_func_qcfg(bp, &mtu);
5034 rc = bnxt_hwrm_cfa_adv_flow_mgmt_qcaps(bp);
5038 bnxt_hwrm_port_mac_qcfg(bp);
5040 bnxt_hwrm_parent_pf_qcfg(bp);
5042 bnxt_hwrm_port_phy_qcaps(bp);
5044 bnxt_alloc_error_recovery_info(bp);
5045 /* Get the adapter error recovery support info */
5046 rc = bnxt_hwrm_error_recovery_qcfg(bp);
5048 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5050 bnxt_hwrm_port_led_qcaps(bp);
5056 bnxt_init_locks(struct bnxt *bp)
5060 err = pthread_mutex_init(&bp->flow_lock, NULL);
5062 PMD_DRV_LOG(ERR, "Unable to initialize flow_lock\n");
5066 err = pthread_mutex_init(&bp->def_cp_lock, NULL);
5068 PMD_DRV_LOG(ERR, "Unable to initialize def_cp_lock\n");
5072 err = pthread_mutex_init(&bp->health_check_lock, NULL);
5074 PMD_DRV_LOG(ERR, "Unable to initialize health_check_lock\n");
5078 err = pthread_mutex_init(&bp->err_recovery_lock, NULL);
5080 PMD_DRV_LOG(ERR, "Unable to initialize err_recovery_lock\n");
5085 static int bnxt_init_resources(struct bnxt *bp, bool reconfig_dev)
5089 rc = bnxt_get_config(bp);
5093 if (!reconfig_dev) {
5094 rc = bnxt_setup_mac_addr(bp->eth_dev);
5098 rc = bnxt_restore_dflt_mac(bp);
5103 bnxt_config_vf_req_fwd(bp);
5105 rc = bnxt_hwrm_func_driver_register(bp);
5107 PMD_DRV_LOG(ERR, "Failed to register driver");
5112 if (bp->pdev->max_vfs) {
5113 rc = bnxt_hwrm_allocate_vfs(bp, bp->pdev->max_vfs);
5115 PMD_DRV_LOG(ERR, "Failed to allocate VFs\n");
5119 rc = bnxt_hwrm_allocate_pf_only(bp);
5122 "Failed to allocate PF resources");
5128 rc = bnxt_alloc_mem(bp, reconfig_dev);
5132 rc = bnxt_setup_int(bp);
5136 rc = bnxt_request_int(bp);
5140 rc = bnxt_init_ctx_mem(bp);
5142 PMD_DRV_LOG(ERR, "Failed to init adv_flow_counters\n");
5150 bnxt_parse_devarg_truflow(__rte_unused const char *key,
5151 const char *value, void *opaque_arg)
5153 struct bnxt *bp = opaque_arg;
5154 unsigned long truflow;
5157 if (!value || !opaque_arg) {
5159 "Invalid parameter passed to truflow devargs.\n");
5163 truflow = strtoul(value, &end, 10);
5164 if (end == NULL || *end != '\0' ||
5165 (truflow == ULONG_MAX && errno == ERANGE)) {
5167 "Invalid parameter passed to truflow devargs.\n");
5171 if (BNXT_DEVARG_TRUFLOW_INVALID(truflow)) {
5173 "Invalid value passed to truflow devargs.\n");
5178 bp->flags |= BNXT_FLAG_TRUFLOW_EN;
5179 PMD_DRV_LOG(INFO, "Host-based truflow feature enabled.\n");
5181 bp->flags &= ~BNXT_FLAG_TRUFLOW_EN;
5182 PMD_DRV_LOG(INFO, "Host-based truflow feature disabled.\n");
5189 bnxt_parse_devarg_flow_xstat(__rte_unused const char *key,
5190 const char *value, void *opaque_arg)
5192 struct bnxt *bp = opaque_arg;
5193 unsigned long flow_xstat;
5196 if (!value || !opaque_arg) {
5198 "Invalid parameter passed to flow_xstat devarg.\n");
5202 flow_xstat = strtoul(value, &end, 10);
5203 if (end == NULL || *end != '\0' ||
5204 (flow_xstat == ULONG_MAX && errno == ERANGE)) {
5206 "Invalid parameter passed to flow_xstat devarg.\n");
5210 if (BNXT_DEVARG_FLOW_XSTAT_INVALID(flow_xstat)) {
5212 "Invalid value passed to flow_xstat devarg.\n");
5216 bp->flags |= BNXT_FLAG_FLOW_XSTATS_EN;
5217 if (BNXT_FLOW_XSTATS_EN(bp))
5218 PMD_DRV_LOG(INFO, "flow_xstat feature enabled.\n");
5224 bnxt_parse_devarg_max_num_kflows(__rte_unused const char *key,
5225 const char *value, void *opaque_arg)
5227 struct bnxt *bp = opaque_arg;
5228 unsigned long max_num_kflows;
5231 if (!value || !opaque_arg) {
5233 "Invalid parameter passed to max_num_kflows devarg.\n");
5237 max_num_kflows = strtoul(value, &end, 10);
5238 if (end == NULL || *end != '\0' ||
5239 (max_num_kflows == ULONG_MAX && errno == ERANGE)) {
5241 "Invalid parameter passed to max_num_kflows devarg.\n");
5245 if (bnxt_devarg_max_num_kflow_invalid(max_num_kflows)) {
5247 "Invalid value passed to max_num_kflows devarg.\n");
5251 bp->max_num_kflows = max_num_kflows;
5252 if (bp->max_num_kflows)
5253 PMD_DRV_LOG(INFO, "max_num_kflows set as %ldK.\n",
5260 bnxt_parse_devarg_rep_is_pf(__rte_unused const char *key,
5261 const char *value, void *opaque_arg)
5263 struct bnxt_representor *vfr_bp = opaque_arg;
5264 unsigned long rep_is_pf;
5267 if (!value || !opaque_arg) {
5269 "Invalid parameter passed to rep_is_pf devargs.\n");
5273 rep_is_pf = strtoul(value, &end, 10);
5274 if (end == NULL || *end != '\0' ||
5275 (rep_is_pf == ULONG_MAX && errno == ERANGE)) {
5277 "Invalid parameter passed to rep_is_pf devargs.\n");
5281 if (BNXT_DEVARG_REP_IS_PF_INVALID(rep_is_pf)) {
5283 "Invalid value passed to rep_is_pf devargs.\n");
5287 vfr_bp->flags |= rep_is_pf;
5288 if (BNXT_REP_PF(vfr_bp))
5289 PMD_DRV_LOG(INFO, "PF representor\n");
5291 PMD_DRV_LOG(INFO, "VF representor\n");
5297 bnxt_parse_devarg_rep_based_pf(__rte_unused const char *key,
5298 const char *value, void *opaque_arg)
5300 struct bnxt_representor *vfr_bp = opaque_arg;
5301 unsigned long rep_based_pf;
5304 if (!value || !opaque_arg) {
5306 "Invalid parameter passed to rep_based_pf "
5311 rep_based_pf = strtoul(value, &end, 10);
5312 if (end == NULL || *end != '\0' ||
5313 (rep_based_pf == ULONG_MAX && errno == ERANGE)) {
5315 "Invalid parameter passed to rep_based_pf "
5320 if (BNXT_DEVARG_REP_BASED_PF_INVALID(rep_based_pf)) {
5322 "Invalid value passed to rep_based_pf devargs.\n");
5326 vfr_bp->rep_based_pf = rep_based_pf;
5327 vfr_bp->flags |= BNXT_REP_BASED_PF_VALID;
5329 PMD_DRV_LOG(INFO, "rep-based-pf = %d\n", vfr_bp->rep_based_pf);
5335 bnxt_parse_devarg_rep_q_r2f(__rte_unused const char *key,
5336 const char *value, void *opaque_arg)
5338 struct bnxt_representor *vfr_bp = opaque_arg;
5339 unsigned long rep_q_r2f;
5342 if (!value || !opaque_arg) {
5344 "Invalid parameter passed to rep_q_r2f "
5349 rep_q_r2f = strtoul(value, &end, 10);
5350 if (end == NULL || *end != '\0' ||
5351 (rep_q_r2f == ULONG_MAX && errno == ERANGE)) {
5353 "Invalid parameter passed to rep_q_r2f "
5358 if (BNXT_DEVARG_REP_Q_R2F_INVALID(rep_q_r2f)) {
5360 "Invalid value passed to rep_q_r2f devargs.\n");
5364 vfr_bp->rep_q_r2f = rep_q_r2f;
5365 vfr_bp->flags |= BNXT_REP_Q_R2F_VALID;
5366 PMD_DRV_LOG(INFO, "rep-q-r2f = %d\n", vfr_bp->rep_q_r2f);
5372 bnxt_parse_devarg_rep_q_f2r(__rte_unused const char *key,
5373 const char *value, void *opaque_arg)
5375 struct bnxt_representor *vfr_bp = opaque_arg;
5376 unsigned long rep_q_f2r;
5379 if (!value || !opaque_arg) {
5381 "Invalid parameter passed to rep_q_f2r "
5386 rep_q_f2r = strtoul(value, &end, 10);
5387 if (end == NULL || *end != '\0' ||
5388 (rep_q_f2r == ULONG_MAX && errno == ERANGE)) {
5390 "Invalid parameter passed to rep_q_f2r "
5395 if (BNXT_DEVARG_REP_Q_F2R_INVALID(rep_q_f2r)) {
5397 "Invalid value passed to rep_q_f2r devargs.\n");
5401 vfr_bp->rep_q_f2r = rep_q_f2r;
5402 vfr_bp->flags |= BNXT_REP_Q_F2R_VALID;
5403 PMD_DRV_LOG(INFO, "rep-q-f2r = %d\n", vfr_bp->rep_q_f2r);
5409 bnxt_parse_devarg_rep_fc_r2f(__rte_unused const char *key,
5410 const char *value, void *opaque_arg)
5412 struct bnxt_representor *vfr_bp = opaque_arg;
5413 unsigned long rep_fc_r2f;
5416 if (!value || !opaque_arg) {
5418 "Invalid parameter passed to rep_fc_r2f "
5423 rep_fc_r2f = strtoul(value, &end, 10);
5424 if (end == NULL || *end != '\0' ||
5425 (rep_fc_r2f == ULONG_MAX && errno == ERANGE)) {
5427 "Invalid parameter passed to rep_fc_r2f "
5432 if (BNXT_DEVARG_REP_FC_R2F_INVALID(rep_fc_r2f)) {
5434 "Invalid value passed to rep_fc_r2f devargs.\n");
5438 vfr_bp->flags |= BNXT_REP_FC_R2F_VALID;
5439 vfr_bp->rep_fc_r2f = rep_fc_r2f;
5440 PMD_DRV_LOG(INFO, "rep-fc-r2f = %lu\n", rep_fc_r2f);
5446 bnxt_parse_devarg_rep_fc_f2r(__rte_unused const char *key,
5447 const char *value, void *opaque_arg)
5449 struct bnxt_representor *vfr_bp = opaque_arg;
5450 unsigned long rep_fc_f2r;
5453 if (!value || !opaque_arg) {
5455 "Invalid parameter passed to rep_fc_f2r "
5460 rep_fc_f2r = strtoul(value, &end, 10);
5461 if (end == NULL || *end != '\0' ||
5462 (rep_fc_f2r == ULONG_MAX && errno == ERANGE)) {
5464 "Invalid parameter passed to rep_fc_f2r "
5469 if (BNXT_DEVARG_REP_FC_F2R_INVALID(rep_fc_f2r)) {
5471 "Invalid value passed to rep_fc_f2r devargs.\n");
5475 vfr_bp->flags |= BNXT_REP_FC_F2R_VALID;
5476 vfr_bp->rep_fc_f2r = rep_fc_f2r;
5477 PMD_DRV_LOG(INFO, "rep-fc-f2r = %lu\n", rep_fc_f2r);
5483 bnxt_parse_dev_args(struct bnxt *bp, struct rte_devargs *devargs)
5485 struct rte_kvargs *kvlist;
5488 if (devargs == NULL)
5491 kvlist = rte_kvargs_parse(devargs->args, bnxt_dev_args);
5496 * Handler for "truflow" devarg.
5497 * Invoked as for ex: "-a 0000:00:0d.0,host-based-truflow=1"
5499 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_TRUFLOW,
5500 bnxt_parse_devarg_truflow, bp);
5505 * Handler for "flow_xstat" devarg.
5506 * Invoked as for ex: "-a 0000:00:0d.0,flow_xstat=1"
5508 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_FLOW_XSTAT,
5509 bnxt_parse_devarg_flow_xstat, bp);
5514 * Handler for "max_num_kflows" devarg.
5515 * Invoked as for ex: "-a 000:00:0d.0,max_num_kflows=32"
5517 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_MAX_NUM_KFLOWS,
5518 bnxt_parse_devarg_max_num_kflows, bp);
5523 rte_kvargs_free(kvlist);
5527 static int bnxt_alloc_switch_domain(struct bnxt *bp)
5531 if (BNXT_PF(bp) || BNXT_VF_IS_TRUSTED(bp)) {
5532 rc = rte_eth_switch_domain_alloc(&bp->switch_domain_id);
5535 "Failed to alloc switch domain: %d\n", rc);
5538 "Switch domain allocated %d\n",
5539 bp->switch_domain_id);
5545 /* Allocate and initialize various fields in bnxt struct that
5546 * need to be allocated/destroyed only once in the lifetime of the driver
5548 static int bnxt_drv_init(struct rte_eth_dev *eth_dev)
5550 struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5551 struct bnxt *bp = eth_dev->data->dev_private;
5554 bp->flags &= ~BNXT_FLAG_RX_VECTOR_PKT_MODE;
5556 if (bnxt_vf_pciid(pci_dev->id.device_id))
5557 bp->flags |= BNXT_FLAG_VF;
5559 if (bnxt_p5_device(pci_dev->id.device_id))
5560 bp->flags |= BNXT_FLAG_CHIP_P5;
5562 if (pci_dev->id.device_id == BROADCOM_DEV_ID_58802 ||
5563 pci_dev->id.device_id == BROADCOM_DEV_ID_58804 ||
5564 pci_dev->id.device_id == BROADCOM_DEV_ID_58808 ||
5565 pci_dev->id.device_id == BROADCOM_DEV_ID_58802_VF)
5566 bp->flags |= BNXT_FLAG_STINGRAY;
5568 if (BNXT_TRUFLOW_EN(bp)) {
5569 /* extra mbuf field is required to store CFA code from mark */
5570 static const struct rte_mbuf_dynfield bnxt_cfa_code_dynfield_desc = {
5571 .name = RTE_PMD_BNXT_CFA_CODE_DYNFIELD_NAME,
5572 .size = sizeof(bnxt_cfa_code_dynfield_t),
5573 .align = __alignof__(bnxt_cfa_code_dynfield_t),
5575 bnxt_cfa_code_dynfield_offset =
5576 rte_mbuf_dynfield_register(&bnxt_cfa_code_dynfield_desc);
5577 if (bnxt_cfa_code_dynfield_offset < 0) {
5579 "Failed to register mbuf field for TruFlow mark\n");
5584 rc = bnxt_map_pci_bars(eth_dev);
5587 "Failed to initialize board rc: %x\n", rc);
5591 rc = bnxt_alloc_pf_info(bp);
5595 rc = bnxt_alloc_link_info(bp);
5599 rc = bnxt_alloc_parent_info(bp);
5603 rc = bnxt_alloc_hwrm_resources(bp);
5606 "Failed to allocate response buffer rc: %x\n", rc);
5609 rc = bnxt_alloc_leds_info(bp);
5613 rc = bnxt_alloc_cos_queues(bp);
5617 rc = bnxt_init_locks(bp);
5621 rc = bnxt_alloc_switch_domain(bp);
5629 bnxt_dev_init(struct rte_eth_dev *eth_dev, void *params __rte_unused)
5631 struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
5632 static int version_printed;
5636 if (version_printed++ == 0)
5637 PMD_DRV_LOG(INFO, "%s\n", bnxt_version);
5639 eth_dev->dev_ops = &bnxt_dev_ops;
5640 eth_dev->rx_queue_count = bnxt_rx_queue_count_op;
5641 eth_dev->rx_descriptor_status = bnxt_rx_descriptor_status_op;
5642 eth_dev->tx_descriptor_status = bnxt_tx_descriptor_status_op;
5643 eth_dev->rx_pkt_burst = &bnxt_recv_pkts;
5644 eth_dev->tx_pkt_burst = &bnxt_xmit_pkts;
5647 * For secondary processes, we don't initialise any further
5648 * as primary has already done this work.
5650 if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5653 rte_eth_copy_pci_info(eth_dev, pci_dev);
5654 eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
5656 bp = eth_dev->data->dev_private;
5658 /* Parse dev arguments passed on when starting the DPDK application. */
5659 rc = bnxt_parse_dev_args(bp, pci_dev->device.devargs);
5663 rc = bnxt_drv_init(eth_dev);
5667 rc = bnxt_init_resources(bp, false);
5671 rc = bnxt_alloc_stats_mem(bp);
5676 DRV_MODULE_NAME "found at mem %" PRIX64 ", node addr %pM\n",
5677 pci_dev->mem_resource[0].phys_addr,
5678 pci_dev->mem_resource[0].addr);
5683 bnxt_dev_uninit(eth_dev);
5688 static void bnxt_free_ctx_mem_buf(struct bnxt_ctx_mem_buf_info *ctx)
5697 ctx->dma = RTE_BAD_IOVA;
5698 ctx->ctx_id = BNXT_CTX_VAL_INVAL;
5701 static void bnxt_unregister_fc_ctx_mem(struct bnxt *bp)
5703 bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_RX,
5704 CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5705 bp->flow_stat->rx_fc_out_tbl.ctx_id,
5706 bp->flow_stat->max_fc,
5709 bnxt_hwrm_cfa_counter_cfg(bp, BNXT_DIR_TX,
5710 CFA_COUNTER_CFG_IN_COUNTER_TYPE_FC,
5711 bp->flow_stat->tx_fc_out_tbl.ctx_id,
5712 bp->flow_stat->max_fc,
5715 if (bp->flow_stat->rx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5716 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_in_tbl.ctx_id);
5717 bp->flow_stat->rx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5719 if (bp->flow_stat->rx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5720 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->rx_fc_out_tbl.ctx_id);
5721 bp->flow_stat->rx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5723 if (bp->flow_stat->tx_fc_in_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5724 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_in_tbl.ctx_id);
5725 bp->flow_stat->tx_fc_in_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5727 if (bp->flow_stat->tx_fc_out_tbl.ctx_id != BNXT_CTX_VAL_INVAL)
5728 bnxt_hwrm_ctx_unrgtr(bp, bp->flow_stat->tx_fc_out_tbl.ctx_id);
5729 bp->flow_stat->tx_fc_out_tbl.ctx_id = BNXT_CTX_VAL_INVAL;
5732 static void bnxt_uninit_fc_ctx_mem(struct bnxt *bp)
5734 bnxt_unregister_fc_ctx_mem(bp);
5736 bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_in_tbl);
5737 bnxt_free_ctx_mem_buf(&bp->flow_stat->rx_fc_out_tbl);
5738 bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_in_tbl);
5739 bnxt_free_ctx_mem_buf(&bp->flow_stat->tx_fc_out_tbl);
5742 static void bnxt_uninit_ctx_mem(struct bnxt *bp)
5744 if (BNXT_FLOW_XSTATS_EN(bp))
5745 bnxt_uninit_fc_ctx_mem(bp);
5749 bnxt_free_error_recovery_info(struct bnxt *bp)
5751 rte_free(bp->recovery_info);
5752 bp->recovery_info = NULL;
5753 bp->fw_cap &= ~BNXT_FW_CAP_ERROR_RECOVERY;
5757 bnxt_uninit_resources(struct bnxt *bp, bool reconfig_dev)
5762 bnxt_free_mem(bp, reconfig_dev);
5764 bnxt_hwrm_func_buf_unrgtr(bp);
5765 if (bp->pf != NULL) {
5766 rte_free(bp->pf->vf_req_buf);
5767 bp->pf->vf_req_buf = NULL;
5770 rc = bnxt_hwrm_func_driver_unregister(bp, 0);
5771 bp->flags &= ~BNXT_FLAG_REGISTERED;
5772 bnxt_free_ctx_mem(bp);
5773 if (!reconfig_dev) {
5774 bnxt_free_hwrm_resources(bp);
5775 bnxt_free_error_recovery_info(bp);
5778 bnxt_uninit_ctx_mem(bp);
5780 bnxt_free_flow_stats_info(bp);
5781 if (bp->rep_info != NULL)
5782 bnxt_free_switch_domain(bp);
5783 bnxt_free_rep_info(bp);
5784 rte_free(bp->ptp_cfg);
5790 bnxt_dev_uninit(struct rte_eth_dev *eth_dev)
5792 if (rte_eal_process_type() != RTE_PROC_PRIMARY)
5795 PMD_DRV_LOG(DEBUG, "Calling Device uninit\n");
5797 if (eth_dev->state != RTE_ETH_DEV_UNUSED)
5798 bnxt_dev_close_op(eth_dev);
5803 static int bnxt_pci_remove_dev_with_reps(struct rte_eth_dev *eth_dev)
5805 struct bnxt *bp = eth_dev->data->dev_private;
5806 struct rte_eth_dev *vf_rep_eth_dev;
5812 for (i = 0; i < bp->num_reps; i++) {
5813 vf_rep_eth_dev = bp->rep_info[i].vfr_eth_dev;
5814 if (!vf_rep_eth_dev)
5816 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci remove\n",
5817 vf_rep_eth_dev->data->port_id);
5818 rte_eth_dev_destroy(vf_rep_eth_dev, bnxt_representor_uninit);
5820 PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n",
5821 eth_dev->data->port_id);
5822 ret = rte_eth_dev_destroy(eth_dev, bnxt_dev_uninit);
5827 static void bnxt_free_rep_info(struct bnxt *bp)
5829 rte_free(bp->rep_info);
5830 bp->rep_info = NULL;
5831 rte_free(bp->cfa_code_map);
5832 bp->cfa_code_map = NULL;
5835 static int bnxt_init_rep_info(struct bnxt *bp)
5842 bp->rep_info = rte_zmalloc("bnxt_rep_info",
5843 sizeof(bp->rep_info[0]) * BNXT_MAX_VF_REPS,
5845 if (!bp->rep_info) {
5846 PMD_DRV_LOG(ERR, "Failed to alloc memory for rep info\n");
5849 bp->cfa_code_map = rte_zmalloc("bnxt_cfa_code_map",
5850 sizeof(*bp->cfa_code_map) *
5851 BNXT_MAX_CFA_CODE, 0);
5852 if (!bp->cfa_code_map) {
5853 PMD_DRV_LOG(ERR, "Failed to alloc memory for cfa_code_map\n");
5854 bnxt_free_rep_info(bp);
5858 for (i = 0; i < BNXT_MAX_CFA_CODE; i++)
5859 bp->cfa_code_map[i] = BNXT_VF_IDX_INVALID;
5861 rc = pthread_mutex_init(&bp->rep_info->vfr_lock, NULL);
5863 PMD_DRV_LOG(ERR, "Unable to initialize vfr_lock\n");
5864 bnxt_free_rep_info(bp);
5868 rc = pthread_mutex_init(&bp->rep_info->vfr_start_lock, NULL);
5870 PMD_DRV_LOG(ERR, "Unable to initialize vfr_start_lock\n");
5871 bnxt_free_rep_info(bp);
5878 static int bnxt_rep_port_probe(struct rte_pci_device *pci_dev,
5879 struct rte_eth_devargs *eth_da,
5880 struct rte_eth_dev *backing_eth_dev,
5881 const char *dev_args)
5883 struct rte_eth_dev *vf_rep_eth_dev;
5884 char name[RTE_ETH_NAME_MAX_LEN];
5885 struct bnxt *backing_bp;
5888 struct rte_kvargs *kvlist = NULL;
5890 if (eth_da->type == RTE_ETH_REPRESENTOR_NONE)
5892 if (eth_da->type != RTE_ETH_REPRESENTOR_VF) {
5893 PMD_DRV_LOG(ERR, "unsupported representor type %d\n",
5897 num_rep = eth_da->nb_representor_ports;
5898 if (num_rep > BNXT_MAX_VF_REPS) {
5899 PMD_DRV_LOG(ERR, "nb_representor_ports = %d > %d MAX VF REPS\n",
5900 num_rep, BNXT_MAX_VF_REPS);
5904 if (num_rep >= RTE_MAX_ETHPORTS) {
5906 "nb_representor_ports = %d > %d MAX ETHPORTS\n",
5907 num_rep, RTE_MAX_ETHPORTS);
5911 backing_bp = backing_eth_dev->data->dev_private;
5913 if (!(BNXT_PF(backing_bp) || BNXT_VF_IS_TRUSTED(backing_bp))) {
5915 "Not a PF or trusted VF. No Representor support\n");
5916 /* Returning an error is not an option.
5917 * Applications are not handling this correctly
5922 if (bnxt_init_rep_info(backing_bp))
5925 for (i = 0; i < num_rep; i++) {
5926 struct bnxt_representor representor = {
5927 .vf_id = eth_da->representor_ports[i],
5928 .switch_domain_id = backing_bp->switch_domain_id,
5929 .parent_dev = backing_eth_dev
5932 if (representor.vf_id >= BNXT_MAX_VF_REPS) {
5933 PMD_DRV_LOG(ERR, "VF-Rep id %d >= %d MAX VF ID\n",
5934 representor.vf_id, BNXT_MAX_VF_REPS);
5938 /* representor port net_bdf_port */
5939 snprintf(name, sizeof(name), "net_%s_representor_%d",
5940 pci_dev->device.name, eth_da->representor_ports[i]);
5942 kvlist = rte_kvargs_parse(dev_args, bnxt_dev_args);
5945 * Handler for "rep_is_pf" devarg.
5946 * Invoked as for ex: "-a 000:00:0d.0,
5947 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5949 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_IS_PF,
5950 bnxt_parse_devarg_rep_is_pf,
5951 (void *)&representor);
5957 * Handler for "rep_based_pf" devarg.
5958 * Invoked as for ex: "-a 000:00:0d.0,
5959 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5961 ret = rte_kvargs_process(kvlist,
5962 BNXT_DEVARG_REP_BASED_PF,
5963 bnxt_parse_devarg_rep_based_pf,
5964 (void *)&representor);
5970 * Handler for "rep_based_pf" devarg.
5971 * Invoked as for ex: "-a 000:00:0d.0,
5972 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5974 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_R2F,
5975 bnxt_parse_devarg_rep_q_r2f,
5976 (void *)&representor);
5982 * Handler for "rep_based_pf" devarg.
5983 * Invoked as for ex: "-a 000:00:0d.0,
5984 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5986 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_Q_F2R,
5987 bnxt_parse_devarg_rep_q_f2r,
5988 (void *)&representor);
5994 * Handler for "rep_based_pf" devarg.
5995 * Invoked as for ex: "-a 000:00:0d.0,
5996 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
5998 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_R2F,
5999 bnxt_parse_devarg_rep_fc_r2f,
6000 (void *)&representor);
6006 * Handler for "rep_based_pf" devarg.
6007 * Invoked as for ex: "-a 000:00:0d.0,
6008 * rep-based-pf=<pf index> rep-is-pf=<VF=0 or PF=1>"
6010 ret = rte_kvargs_process(kvlist, BNXT_DEVARG_REP_FC_F2R,
6011 bnxt_parse_devarg_rep_fc_f2r,
6012 (void *)&representor);
6019 ret = rte_eth_dev_create(&pci_dev->device, name,
6020 sizeof(struct bnxt_representor),
6022 bnxt_representor_init,
6025 PMD_DRV_LOG(ERR, "failed to create bnxt vf "
6026 "representor %s.", name);
6030 vf_rep_eth_dev = rte_eth_dev_allocated(name);
6031 if (!vf_rep_eth_dev) {
6032 PMD_DRV_LOG(ERR, "Failed to find the eth_dev"
6033 " for VF-Rep: %s.", name);
6038 PMD_DRV_LOG(DEBUG, "BNXT Port:%d VFR pci probe\n",
6039 backing_eth_dev->data->port_id);
6040 backing_bp->rep_info[representor.vf_id].vfr_eth_dev =
6042 backing_bp->num_reps++;
6046 rte_kvargs_free(kvlist);
6050 /* If num_rep > 1, then rollback already created
6051 * ports, since we'll be failing the probe anyway
6054 bnxt_pci_remove_dev_with_reps(backing_eth_dev);
6056 rte_kvargs_free(kvlist);
6061 static int bnxt_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
6062 struct rte_pci_device *pci_dev)
6064 struct rte_eth_devargs eth_da = { .nb_representor_ports = 0 };
6065 struct rte_eth_dev *backing_eth_dev;
6069 if (pci_dev->device.devargs) {
6070 ret = rte_eth_devargs_parse(pci_dev->device.devargs->args,
6076 num_rep = eth_da.nb_representor_ports;
6077 PMD_DRV_LOG(DEBUG, "nb_representor_ports = %d\n",
6080 /* We could come here after first level of probe is already invoked
6081 * as part of an application bringup(OVS-DPDK vswitchd), so first check
6082 * for already allocated eth_dev for the backing device (PF/Trusted VF)
6084 backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6085 if (backing_eth_dev == NULL) {
6086 ret = rte_eth_dev_create(&pci_dev->device, pci_dev->device.name,
6087 sizeof(struct bnxt),
6088 eth_dev_pci_specific_init, pci_dev,
6089 bnxt_dev_init, NULL);
6091 if (ret || !num_rep)
6094 backing_eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6096 PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci probe\n",
6097 backing_eth_dev->data->port_id);
6102 /* probe representor ports now */
6103 ret = bnxt_rep_port_probe(pci_dev, ð_da, backing_eth_dev,
6104 pci_dev->device.devargs->args);
6109 static int bnxt_pci_remove(struct rte_pci_device *pci_dev)
6111 struct rte_eth_dev *eth_dev;
6113 eth_dev = rte_eth_dev_allocated(pci_dev->device.name);
6115 return 0; /* Invoked typically only by OVS-DPDK, by the
6116 * time it comes here the eth_dev is already
6117 * deleted by rte_eth_dev_close(), so returning
6118 * +ve value will at least help in proper cleanup
6121 PMD_DRV_LOG(DEBUG, "BNXT Port:%d pci remove\n", eth_dev->data->port_id);
6122 if (rte_eal_process_type() == RTE_PROC_PRIMARY) {
6123 if (eth_dev->data->dev_flags & RTE_ETH_DEV_REPRESENTOR)
6124 return rte_eth_dev_destroy(eth_dev,
6125 bnxt_representor_uninit);
6127 return rte_eth_dev_destroy(eth_dev,
6130 return rte_eth_dev_pci_generic_remove(pci_dev, NULL);
6134 static struct rte_pci_driver bnxt_rte_pmd = {
6135 .id_table = bnxt_pci_id_map,
6136 .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
6137 RTE_PCI_DRV_PROBE_AGAIN, /* Needed in case of VF-REPs
6140 .probe = bnxt_pci_probe,
6141 .remove = bnxt_pci_remove,
6145 is_device_supported(struct rte_eth_dev *dev, struct rte_pci_driver *drv)
6147 if (strcmp(dev->device->driver->name, drv->driver.name))
6153 bool is_bnxt_supported(struct rte_eth_dev *dev)
6155 return is_device_supported(dev, &bnxt_rte_pmd);
6158 RTE_LOG_REGISTER(bnxt_logtype_driver, pmd.net.bnxt.driver, NOTICE);
6159 RTE_PMD_REGISTER_PCI(net_bnxt, bnxt_rte_pmd);
6160 RTE_PMD_REGISTER_PCI_TABLE(net_bnxt, bnxt_pci_id_map);
6161 RTE_PMD_REGISTER_KMOD_DEP(net_bnxt, "* igb_uio | uio_pci_generic | vfio-pci");