2 * Copyright (c) 2015-2016, Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
35 #include <linux/if_vlan.h>
36 #include <linux/etherdevice.h>
37 #include <linux/timecounter.h>
38 #include <linux/net_tstamp.h>
39 #include <linux/crash_dump.h>
40 #include <linux/mlx5/driver.h>
41 #include <linux/mlx5/qp.h>
42 #include <linux/mlx5/cq.h>
43 #include <linux/mlx5/port.h>
44 #include <linux/mlx5/vport.h>
45 #include <linux/mlx5/transobj.h>
46 #include <linux/mlx5/fs.h>
47 #include <linux/rhashtable.h>
48 #include <net/udp_tunnel.h>
49 #include <net/switchdev.h>
51 #include <linux/dim.h>
52 #include <linux/bits.h>
54 #include "mlx5_core.h"
59 #include "lib/hv_vhca.h"
60 #include "lib/clock.h"
62 extern const struct net_device_ops mlx5e_netdev_ops;
65 #define MLX5E_METADATA_ETHER_TYPE (0x8CE4)
66 #define MLX5E_METADATA_ETHER_LEN 8
68 #define MLX5_SET_CFG(p, f, v) MLX5_SET(create_flow_group_in, p, f, v)
70 #define MLX5E_ETH_HARD_MTU (ETH_HLEN + VLAN_HLEN + ETH_FCS_LEN)
72 #define MLX5E_HW2SW_MTU(params, hwmtu) ((hwmtu) - ((params)->hard_mtu))
73 #define MLX5E_SW2HW_MTU(params, swmtu) ((swmtu) + ((params)->hard_mtu))
75 #define MLX5E_MAX_NUM_TC 8
77 #define MLX5_RX_HEADROOM NET_SKB_PAD
78 #define MLX5_SKB_FRAG_SZ(len) (SKB_DATA_ALIGN(len) + \
79 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
81 #define MLX5E_RX_MAX_HEAD (256)
83 #define MLX5_MPWRQ_MIN_LOG_STRIDE_SZ(mdev) \
84 (6 + MLX5_CAP_GEN(mdev, cache_line_128byte)) /* HW restriction */
85 #define MLX5_MPWRQ_LOG_STRIDE_SZ(mdev, req) \
86 max_t(u32, MLX5_MPWRQ_MIN_LOG_STRIDE_SZ(mdev), req)
87 #define MLX5_MPWRQ_DEF_LOG_STRIDE_SZ(mdev) \
88 MLX5_MPWRQ_LOG_STRIDE_SZ(mdev, order_base_2(MLX5E_RX_MAX_HEAD))
90 #define MLX5_MPWRQ_LOG_WQE_SZ 18
91 #define MLX5_MPWRQ_WQE_PAGE_ORDER (MLX5_MPWRQ_LOG_WQE_SZ - PAGE_SHIFT > 0 ? \
92 MLX5_MPWRQ_LOG_WQE_SZ - PAGE_SHIFT : 0)
93 #define MLX5_MPWRQ_PAGES_PER_WQE BIT(MLX5_MPWRQ_WQE_PAGE_ORDER)
95 #define MLX5_ALIGN_MTTS(mtts) (ALIGN(mtts, 8))
96 #define MLX5_ALIGNED_MTTS_OCTW(mtts) ((mtts) / 2)
97 #define MLX5_MTT_OCTW(mtts) (MLX5_ALIGNED_MTTS_OCTW(MLX5_ALIGN_MTTS(mtts)))
98 /* Add another page to MLX5E_REQUIRED_WQE_MTTS as a buffer between
99 * WQEs, This page will absorb write overflow by the hardware, when
100 * receiving packets larger than MTU. These oversize packets are
101 * dropped by the driver at a later stage.
103 #define MLX5E_REQUIRED_WQE_MTTS (MLX5_ALIGN_MTTS(MLX5_MPWRQ_PAGES_PER_WQE + 1))
104 #define MLX5E_REQUIRED_MTTS(wqes) (wqes * MLX5E_REQUIRED_WQE_MTTS)
105 #define MLX5E_MAX_RQ_NUM_MTTS \
106 ((1 << 16) * 2) /* So that MLX5_MTT_OCTW(num_mtts) fits into u16 */
107 #define MLX5E_ORDER2_MAX_PACKET_MTU (order_base_2(10 * 1024))
108 #define MLX5E_PARAMS_MAXIMUM_LOG_RQ_SIZE_MPW \
109 (ilog2(MLX5E_MAX_RQ_NUM_MTTS / MLX5E_REQUIRED_WQE_MTTS))
110 #define MLX5E_LOG_MAX_RQ_NUM_PACKETS_MPW \
111 (MLX5E_PARAMS_MAXIMUM_LOG_RQ_SIZE_MPW + \
112 (MLX5_MPWRQ_LOG_WQE_SZ - MLX5E_ORDER2_MAX_PACKET_MTU))
114 #define MLX5E_MIN_SKB_FRAG_SZ (MLX5_SKB_FRAG_SZ(MLX5_RX_HEADROOM))
115 #define MLX5E_LOG_MAX_RX_WQE_BULK \
116 (ilog2(PAGE_SIZE / roundup_pow_of_two(MLX5E_MIN_SKB_FRAG_SZ)))
118 #define MLX5E_PARAMS_MINIMUM_LOG_SQ_SIZE 0x6
119 #define MLX5E_PARAMS_DEFAULT_LOG_SQ_SIZE 0xa
120 #define MLX5E_PARAMS_MAXIMUM_LOG_SQ_SIZE 0xd
122 #define MLX5E_PARAMS_MINIMUM_LOG_RQ_SIZE (1 + MLX5E_LOG_MAX_RX_WQE_BULK)
123 #define MLX5E_PARAMS_DEFAULT_LOG_RQ_SIZE 0xa
124 #define MLX5E_PARAMS_MAXIMUM_LOG_RQ_SIZE min_t(u8, 0xd, \
125 MLX5E_LOG_MAX_RQ_NUM_PACKETS_MPW)
127 #define MLX5E_PARAMS_MINIMUM_LOG_RQ_SIZE_MPW 0x2
129 #define MLX5E_PARAMS_DEFAULT_LRO_WQE_SZ (64 * 1024)
130 #define MLX5E_DEFAULT_LRO_TIMEOUT 32
131 #define MLX5E_LRO_TIMEOUT_ARR_SIZE 4
133 #define MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC 0x10
134 #define MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_USEC_FROM_CQE 0x3
135 #define MLX5E_PARAMS_DEFAULT_RX_CQ_MODERATION_PKTS 0x20
136 #define MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC 0x10
137 #define MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_USEC_FROM_CQE 0x10
138 #define MLX5E_PARAMS_DEFAULT_TX_CQ_MODERATION_PKTS 0x20
139 #define MLX5E_PARAMS_DEFAULT_MIN_RX_WQES 0x80
140 #define MLX5E_PARAMS_DEFAULT_MIN_RX_WQES_MPW 0x2
142 #define MLX5E_LOG_INDIR_RQT_SIZE 0x8
143 #define MLX5E_INDIR_RQT_SIZE BIT(MLX5E_LOG_INDIR_RQT_SIZE)
144 #define MLX5E_MIN_NUM_CHANNELS 0x1
145 #define MLX5E_MAX_NUM_CHANNELS (MLX5E_INDIR_RQT_SIZE / 2)
146 #define MLX5E_MAX_NUM_SQS (MLX5E_MAX_NUM_CHANNELS * MLX5E_MAX_NUM_TC)
147 #define MLX5E_TX_CQ_POLL_BUDGET 128
148 #define MLX5E_TX_XSK_POLL_BUDGET 64
149 #define MLX5E_SQ_RECOVER_MIN_INTERVAL 500 /* msecs */
151 #define MLX5E_UMR_WQE_INLINE_SZ \
152 (sizeof(struct mlx5e_umr_wqe) + \
153 ALIGN(MLX5_MPWRQ_PAGES_PER_WQE * sizeof(struct mlx5_mtt), \
154 MLX5_UMR_MTT_ALIGNMENT))
155 #define MLX5E_UMR_WQEBBS \
156 (DIV_ROUND_UP(MLX5E_UMR_WQE_INLINE_SZ, MLX5_SEND_WQE_BB))
158 #define MLX5E_MSG_LEVEL NETIF_MSG_LINK
160 #define mlx5e_dbg(mlevel, priv, format, ...) \
162 if (NETIF_MSG_##mlevel & (priv)->msglevel) \
163 netdev_warn(priv->netdev, format, \
167 #define mlx5e_state_dereference(priv, p) \
168 rcu_dereference_protected((p), lockdep_is_held(&(priv)->state_lock))
170 enum mlx5e_rq_group {
171 MLX5E_RQ_GROUP_REGULAR,
173 #define MLX5E_NUM_RQ_GROUPS(g) (1 + MLX5E_RQ_GROUP_##g)
176 static inline u8 mlx5e_get_num_lag_ports(struct mlx5_core_dev *mdev)
178 if (mlx5_lag_is_lacp_owner(mdev))
181 return clamp_t(u8, MLX5_CAP_GEN(mdev, num_lag_ports), 1, MLX5_MAX_PORTS);
184 static inline u16 mlx5_min_rx_wqes(int wq_type, u32 wq_size)
187 case MLX5_WQ_TYPE_LINKED_LIST_STRIDING_RQ:
188 return min_t(u16, MLX5E_PARAMS_DEFAULT_MIN_RX_WQES_MPW,
191 return min_t(u16, MLX5E_PARAMS_DEFAULT_MIN_RX_WQES,
196 /* Use this function to get max num channels (rxqs/txqs) only to create netdev */
197 static inline int mlx5e_get_max_num_channels(struct mlx5_core_dev *mdev)
199 return is_kdump_kernel() ?
200 MLX5E_MIN_NUM_CHANNELS :
201 min_t(int, mlx5_comp_vectors_count(mdev), MLX5E_MAX_NUM_CHANNELS);
204 struct mlx5e_tx_wqe {
205 struct mlx5_wqe_ctrl_seg ctrl;
206 struct mlx5_wqe_eth_seg eth;
207 struct mlx5_wqe_data_seg data[0];
210 struct mlx5e_rx_wqe_ll {
211 struct mlx5_wqe_srq_next_seg next;
212 struct mlx5_wqe_data_seg data[];
215 struct mlx5e_rx_wqe_cyc {
216 struct mlx5_wqe_data_seg data[0];
219 struct mlx5e_umr_wqe {
220 struct mlx5_wqe_ctrl_seg ctrl;
221 struct mlx5_wqe_umr_ctrl_seg uctrl;
222 struct mlx5_mkey_seg mkc;
223 struct mlx5_mtt inline_mtts[0];
226 extern const char mlx5e_self_tests[][ETH_GSTRING_LEN];
228 enum mlx5e_priv_flag {
229 MLX5E_PFLAG_RX_CQE_BASED_MODER,
230 MLX5E_PFLAG_TX_CQE_BASED_MODER,
231 MLX5E_PFLAG_RX_CQE_COMPRESS,
232 MLX5E_PFLAG_RX_STRIDING_RQ,
233 MLX5E_PFLAG_RX_NO_CSUM_COMPLETE,
234 MLX5E_PFLAG_XDP_TX_MPWQE,
235 MLX5E_PFLAG_SKB_TX_MPWQE,
236 MLX5E_PFLAG_TX_PORT_TS,
237 MLX5E_NUM_PFLAGS, /* Keep last */
240 #define MLX5E_SET_PFLAG(params, pflag, enable) \
243 (params)->pflags |= BIT(pflag); \
245 (params)->pflags &= ~(BIT(pflag)); \
248 #define MLX5E_GET_PFLAG(params, pflag) (!!((params)->pflags & (BIT(pflag))))
250 struct mlx5e_params {
253 u8 log_rq_mtu_frames;
256 bool rx_cqe_compress_def;
257 bool tunneled_offload_en;
258 struct dim_cq_moder rx_cq_moderation;
259 struct dim_cq_moder tx_cq_moderation;
261 u8 tx_min_inline_mode;
262 bool vlan_strip_disable;
268 struct bpf_prog *xdp_prog;
269 struct mlx5e_xsk *xsk;
276 MLX5E_RQ_STATE_ENABLED,
277 MLX5E_RQ_STATE_RECOVERING,
279 MLX5E_RQ_STATE_NO_CSUM_COMPLETE,
280 MLX5E_RQ_STATE_CSUM_FULL, /* cqe_csum_full hw bit is set */
281 MLX5E_RQ_STATE_FPGA_TLS, /* FPGA TLS enabled */
282 MLX5E_RQ_STATE_MINI_CQE_HW_STRIDX /* set when mini_cqe_resp_stride_index cap is used */
286 /* data path - accessed per cqe */
289 /* data path - accessed per napi poll */
291 struct napi_struct *napi;
292 struct mlx5_core_cq mcq;
293 struct mlx5e_ch_stats *ch_stats;
296 struct net_device *netdev;
297 struct mlx5_core_dev *mdev;
298 struct mlx5e_priv *priv;
299 struct mlx5_wq_ctrl wq_ctrl;
300 } ____cacheline_aligned_in_smp;
302 struct mlx5e_cq_decomp {
303 /* cqe decompression */
304 struct mlx5_cqe64 title;
305 struct mlx5_mini_cqe8 mini_arr[MLX5_MINI_CQE_ARRAY_SIZE];
309 } ____cacheline_aligned_in_smp;
311 enum mlx5e_dma_map_type {
312 MLX5E_DMA_MAP_SINGLE,
316 struct mlx5e_sq_dma {
319 enum mlx5e_dma_map_type type;
323 MLX5E_SQ_STATE_ENABLED,
324 MLX5E_SQ_STATE_MPWQE,
325 MLX5E_SQ_STATE_RECOVERING,
326 MLX5E_SQ_STATE_IPSEC,
328 MLX5E_SQ_STATE_VLAN_NEED_L2_INLINE,
329 MLX5E_SQ_STATE_PENDING_XSK_TX,
330 MLX5E_SQ_STATE_PENDING_TLS_RX_RESYNC,
333 struct mlx5e_tx_mpwqe {
334 /* Current MPWQE session */
335 struct mlx5e_tx_wqe *wqe;
342 struct mlx5e_skb_fifo {
343 struct sk_buff **fifo;
354 /* dirtied @completion */
358 struct dim dim; /* Adaptive Moderation */
361 u16 pc ____cacheline_aligned_in_smp;
364 struct mlx5e_tx_mpwqe mpwqe;
369 struct mlx5_wq_cyc wq;
371 struct mlx5e_sq_stats *stats;
373 struct mlx5e_sq_dma *dma_fifo;
374 struct mlx5e_skb_fifo skb_fifo;
375 struct mlx5e_tx_wqe_info *wqe_info;
377 void __iomem *uar_map;
378 struct netdev_queue *txq;
386 struct hwtstamp_config *tstamp;
387 struct mlx5_clock *clock;
388 struct net_device *netdev;
389 struct mlx5_core_dev *mdev;
390 struct mlx5e_priv *priv;
393 struct mlx5_wq_ctrl wq_ctrl;
397 struct work_struct recover_work;
398 struct mlx5e_ptpsq *ptpsq;
399 cqe_ts_to_ns ptp_cyc2time;
400 } ____cacheline_aligned_in_smp;
402 struct mlx5e_dma_info {
406 struct xdp_buff *xsk;
410 /* XDP packets can be transmitted in different ways. On completion, we need to
411 * distinguish between them to clean up things in a proper way.
413 enum mlx5e_xdp_xmit_mode {
414 /* An xdp_frame was transmitted due to either XDP_REDIRECT from another
415 * device or XDP_TX from an XSK RQ. The frame has to be unmapped and
418 MLX5E_XDP_XMIT_MODE_FRAME,
420 /* The xdp_frame was created in place as a result of XDP_TX from a
421 * regular RQ. No DMA remapping happened, and the page belongs to us.
423 MLX5E_XDP_XMIT_MODE_PAGE,
425 /* No xdp_frame was created at all, the transmit happened from a UMEM
426 * page. The UMEM Completion Ring producer pointer has to be increased.
428 MLX5E_XDP_XMIT_MODE_XSK,
431 struct mlx5e_xdp_info {
432 enum mlx5e_xdp_xmit_mode mode;
435 struct xdp_frame *xdpf;
440 struct mlx5e_dma_info di;
445 struct mlx5e_xmit_data {
451 struct mlx5e_xdp_info_fifo {
452 struct mlx5e_xdp_info *xi;
459 typedef int (*mlx5e_fp_xmit_xdp_frame_check)(struct mlx5e_xdpsq *);
460 typedef bool (*mlx5e_fp_xmit_xdp_frame)(struct mlx5e_xdpsq *,
461 struct mlx5e_xmit_data *,
462 struct mlx5e_xdp_info *,
468 /* dirtied @completion */
473 u32 xdpi_fifo_pc ____cacheline_aligned_in_smp;
475 struct mlx5_wqe_ctrl_seg *doorbell_cseg;
476 struct mlx5e_tx_mpwqe mpwqe;
481 struct xsk_buff_pool *xsk_pool;
482 struct mlx5_wq_cyc wq;
483 struct mlx5e_xdpsq_stats *stats;
484 mlx5e_fp_xmit_xdp_frame_check xmit_xdp_frame_check;
485 mlx5e_fp_xmit_xdp_frame xmit_xdp_frame;
487 struct mlx5e_xdp_wqe_info *wqe_info;
488 struct mlx5e_xdp_info_fifo xdpi_fifo;
490 void __iomem *uar_map;
499 struct mlx5_wq_ctrl wq_ctrl;
500 struct mlx5e_channel *channel;
501 } ____cacheline_aligned_in_smp;
503 struct mlx5e_ktls_resync_resp;
510 struct mlx5_wqe_ctrl_seg *doorbell_cseg;
513 /* write@xmit, read@completion */
515 struct mlx5e_icosq_wqe_info *wqe_info;
519 struct mlx5_wq_cyc wq;
520 void __iomem *uar_map;
524 struct mlx5e_ktls_resync_resp *ktls_resync;
527 struct mlx5_wq_ctrl wq_ctrl;
528 struct mlx5e_channel *channel;
530 struct work_struct recover_work;
531 } ____cacheline_aligned_in_smp;
533 struct mlx5e_wqe_frag_info {
534 struct mlx5e_dma_info *di;
539 struct mlx5e_umr_dma_info {
540 struct mlx5e_dma_info dma_info[MLX5_MPWRQ_PAGES_PER_WQE];
543 struct mlx5e_mpw_info {
544 struct mlx5e_umr_dma_info umr;
545 u16 consumed_strides;
546 DECLARE_BITMAP(xdp_xmit_bitmap, MLX5_MPWRQ_PAGES_PER_WQE);
549 #define MLX5E_MAX_RX_FRAGS 4
551 /* a single cache unit is capable to serve one napi call (for non-striding rq)
552 * or a MPWQE (for striding rq).
554 #define MLX5E_CACHE_UNIT (MLX5_MPWRQ_PAGES_PER_WQE > NAPI_POLL_WEIGHT ? \
555 MLX5_MPWRQ_PAGES_PER_WQE : NAPI_POLL_WEIGHT)
556 #define MLX5E_CACHE_SIZE (4 * roundup_pow_of_two(MLX5E_CACHE_UNIT))
557 struct mlx5e_page_cache {
560 struct mlx5e_dma_info page_cache[MLX5E_CACHE_SIZE];
564 typedef void (*mlx5e_fp_handle_rx_cqe)(struct mlx5e_rq*, struct mlx5_cqe64*);
565 typedef struct sk_buff *
566 (*mlx5e_fp_skb_from_cqe_mpwrq)(struct mlx5e_rq *rq, struct mlx5e_mpw_info *wi,
567 u16 cqe_bcnt, u32 head_offset, u32 page_idx);
568 typedef struct sk_buff *
569 (*mlx5e_fp_skb_from_cqe)(struct mlx5e_rq *rq, struct mlx5_cqe64 *cqe,
570 struct mlx5e_wqe_frag_info *wi, u32 cqe_bcnt);
571 typedef bool (*mlx5e_fp_post_rx_wqes)(struct mlx5e_rq *rq);
572 typedef void (*mlx5e_fp_dealloc_wqe)(struct mlx5e_rq*, u16);
574 int mlx5e_rq_set_handlers(struct mlx5e_rq *rq, struct mlx5e_params *params, bool xsk);
575 void mlx5e_rq_set_trap_handlers(struct mlx5e_rq *rq, struct mlx5e_params *params);
578 MLX5E_RQ_FLAG_XDP_XMIT,
579 MLX5E_RQ_FLAG_XDP_REDIRECT,
582 struct mlx5e_rq_frag_info {
587 struct mlx5e_rq_frags_info {
588 struct mlx5e_rq_frag_info arr[MLX5E_MAX_RX_FRAGS];
598 struct mlx5_wq_cyc wq;
599 struct mlx5e_wqe_frag_info *frags;
600 struct mlx5e_dma_info *di;
601 struct mlx5e_rq_frags_info info;
602 mlx5e_fp_skb_from_cqe skb_from_cqe;
605 struct mlx5_wq_ll wq;
606 struct mlx5e_umr_wqe umr_wqe;
607 struct mlx5e_mpw_info *info;
608 mlx5e_fp_skb_from_cqe_mpwrq skb_from_cqe_mpwrq;
620 u8 map_dir; /* dma map direction */
624 struct net_device *netdev;
625 struct mlx5e_rq_stats *stats;
627 struct mlx5e_cq_decomp cqd;
628 struct mlx5e_page_cache page_cache;
629 struct hwtstamp_config *tstamp;
630 struct mlx5_clock *clock;
631 struct mlx5e_icosq *icosq;
632 struct mlx5e_priv *priv;
634 mlx5e_fp_handle_rx_cqe handle_rx_cqe;
635 mlx5e_fp_post_rx_wqes post_wqes;
636 mlx5e_fp_dealloc_wqe dealloc_wqe;
642 struct dim dim; /* Dynamic Interrupt Moderation */
645 struct bpf_prog __rcu *xdp_prog;
646 struct mlx5e_xdpsq *xdpsq;
647 DECLARE_BITMAP(flags, 8);
648 struct page_pool *page_pool;
650 /* AF_XDP zero-copy */
651 struct xsk_buff_pool *xsk_pool;
653 struct work_struct recover_work;
656 struct mlx5_wq_ctrl wq_ctrl;
660 struct mlx5_core_dev *mdev;
661 struct mlx5_core_mkey umr_mkey;
662 struct mlx5e_dma_info wqe_overflow;
664 /* XDP read-mostly */
665 struct xdp_rxq_info xdp_rxq;
666 cqe_ts_to_ns ptp_cyc2time;
667 } ____cacheline_aligned_in_smp;
669 enum mlx5e_channel_state {
670 MLX5E_CHANNEL_STATE_XSK,
671 MLX5E_CHANNEL_NUM_STATES
674 struct mlx5e_channel {
677 struct mlx5e_xdpsq rq_xdpsq;
678 struct mlx5e_txqsq sq[MLX5E_MAX_NUM_TC];
679 struct mlx5e_icosq icosq; /* internal control operations */
680 struct mlx5e_txqsq __rcu * __rcu *qos_sqs;
682 struct napi_struct napi;
684 struct net_device *netdev;
691 struct mlx5e_xdpsq xdpsq;
693 /* AF_XDP zero-copy */
694 struct mlx5e_rq xskrq;
695 struct mlx5e_xdpsq xsksq;
698 struct mlx5e_icosq async_icosq;
699 /* async_icosq can be accessed from any CPU - the spinlock protects it. */
700 spinlock_t async_icosq_lock;
702 /* data path - accessed per napi poll */
703 const struct cpumask *aff_mask;
704 struct mlx5e_ch_stats *stats;
707 struct mlx5e_priv *priv;
708 struct mlx5_core_dev *mdev;
709 struct hwtstamp_config *tstamp;
710 DECLARE_BITMAP(state, MLX5E_CHANNEL_NUM_STATES);
717 struct mlx5e_channels {
718 struct mlx5e_channel **c;
719 struct mlx5e_ptp *ptp;
721 struct mlx5e_params params;
724 struct mlx5e_channel_stats {
725 struct mlx5e_ch_stats ch;
726 struct mlx5e_sq_stats sq[MLX5E_MAX_NUM_TC];
727 struct mlx5e_rq_stats rq;
728 struct mlx5e_rq_stats xskrq;
729 struct mlx5e_xdpsq_stats rq_xdpsq;
730 struct mlx5e_xdpsq_stats xdpsq;
731 struct mlx5e_xdpsq_stats xsksq;
732 } ____cacheline_aligned_in_smp;
734 struct mlx5e_ptp_stats {
735 struct mlx5e_ch_stats ch;
736 struct mlx5e_sq_stats sq[MLX5E_MAX_NUM_TC];
737 struct mlx5e_ptp_cq_stats cq[MLX5E_MAX_NUM_TC];
738 struct mlx5e_rq_stats rq;
739 } ____cacheline_aligned_in_smp;
743 MLX5E_STATE_DESTROYING,
744 MLX5E_STATE_XDP_TX_ENABLED,
745 MLX5E_STATE_XDP_ACTIVE,
755 struct mlx5e_rqt rqt;
756 struct list_head list;
764 struct mlx5e_rss_params {
765 u32 indirection_rqt[MLX5E_INDIR_RQT_SIZE];
766 u32 rx_hash_fields[MLX5E_NUM_INDIR_TIRS];
767 u8 toeplitz_hash_key[40];
771 struct mlx5e_modify_sq_param {
777 u16 qos_queue_group_id;
780 #if IS_ENABLED(CONFIG_PCI_HYPERV_INTERFACE)
781 struct mlx5e_hv_vhca_stats_agent {
782 struct mlx5_hv_vhca_agent *agent;
783 struct delayed_work work;
790 /* XSK buffer pools are stored separately from channels,
791 * because we don't want to lose them when channels are
792 * recreated. The kernel also stores buffer pool, but it doesn't
793 * distinguish between zero-copy and non-zero-copy UMEMs, so
794 * rely on our mechanism.
796 struct xsk_buff_pool **pools;
801 /* Temporary storage for variables that are allocated when struct mlx5e_priv is
802 * initialized, and used where we can't allocate them because that functions
803 * must not fail. Use with care and make sure the same variable is not used
804 * simultaneously by multiple users.
806 struct mlx5e_scratchpad {
807 cpumask_var_t cpumask;
811 DECLARE_HASHTABLE(qos_tc2node, order_base_2(MLX5E_QOS_MAX_LEAF_NODES));
812 DECLARE_BITMAP(qos_used_qids, MLX5E_QOS_MAX_LEAF_NODES);
813 struct mlx5e_sq_stats **qos_sq_stats;
822 /* priv data path fields - start */
823 /* +1 for port ptp ts */
824 struct mlx5e_txqsq *txq2sq[(MLX5E_MAX_NUM_CHANNELS + 1) * MLX5E_MAX_NUM_TC +
825 MLX5E_QOS_MAX_LEAF_NODES];
826 int channel_tc2realtxq[MLX5E_MAX_NUM_CHANNELS][MLX5E_MAX_NUM_TC];
827 int port_ptp_tc2realtxq[MLX5E_MAX_NUM_TC];
828 #ifdef CONFIG_MLX5_CORE_EN_DCB
829 struct mlx5e_dcbx_dp dcbx_dp;
831 /* priv data path fields - end */
835 struct mutex state_lock; /* Protects Interface state */
836 struct mlx5e_rq drop_rq;
838 struct mlx5e_channels channels;
839 u32 tisn[MLX5_MAX_PORTS][MLX5E_MAX_NUM_TC];
840 struct mlx5e_rqt indir_rqt;
841 struct mlx5e_tir indir_tir[MLX5E_NUM_INDIR_TIRS];
842 struct mlx5e_tir inner_indir_tir[MLX5E_NUM_INDIR_TIRS];
843 struct mlx5e_tir direct_tir[MLX5E_MAX_NUM_CHANNELS];
844 struct mlx5e_tir xsk_tir[MLX5E_MAX_NUM_CHANNELS];
845 struct mlx5e_tir ptp_tir;
846 struct mlx5e_rss_params rss_params;
847 u32 tx_rates[MLX5E_MAX_NUM_SQS];
849 struct mlx5e_flow_steering fs;
851 struct workqueue_struct *wq;
852 struct work_struct update_carrier_work;
853 struct work_struct set_rx_mode_work;
854 struct work_struct tx_timeout_work;
855 struct work_struct update_stats_work;
856 struct work_struct monitor_counters_work;
857 struct mlx5_nb monitor_counters_nb;
859 struct mlx5_core_dev *mdev;
860 struct net_device *netdev;
861 struct mlx5e_trap *en_trap;
862 struct mlx5e_stats stats;
863 struct mlx5e_channel_stats channel_stats[MLX5E_MAX_NUM_CHANNELS];
864 struct mlx5e_channel_stats trap_stats;
865 struct mlx5e_ptp_stats ptp_stats;
870 struct hwtstamp_config tstamp;
872 u16 drop_rq_q_counter;
873 struct notifier_block events_nb;
874 struct notifier_block blocking_events_nb;
877 struct udp_tunnel_nic_info nic_info;
878 #ifdef CONFIG_MLX5_CORE_EN_DCB
879 struct mlx5e_dcbx dcbx;
882 const struct mlx5e_profile *profile;
884 #ifdef CONFIG_MLX5_EN_IPSEC
885 struct mlx5e_ipsec *ipsec;
887 #ifdef CONFIG_MLX5_EN_TLS
888 struct mlx5e_tls *tls;
890 struct devlink_health_reporter *tx_reporter;
891 struct devlink_health_reporter *rx_reporter;
892 struct mlx5e_xsk xsk;
893 #if IS_ENABLED(CONFIG_PCI_HYPERV_INTERFACE)
894 struct mlx5e_hv_vhca_stats_agent stats_agent;
896 struct mlx5e_scratchpad scratchpad;
897 struct mlx5e_htb htb;
900 struct mlx5e_rx_handlers {
901 mlx5e_fp_handle_rx_cqe handle_rx_cqe;
902 mlx5e_fp_handle_rx_cqe handle_rx_cqe_mpwqe;
905 extern const struct mlx5e_rx_handlers mlx5e_rx_handlers_nic;
907 struct mlx5e_profile {
908 int (*init)(struct mlx5_core_dev *mdev,
909 struct net_device *netdev);
910 void (*cleanup)(struct mlx5e_priv *priv);
911 int (*init_rx)(struct mlx5e_priv *priv);
912 void (*cleanup_rx)(struct mlx5e_priv *priv);
913 int (*init_tx)(struct mlx5e_priv *priv);
914 void (*cleanup_tx)(struct mlx5e_priv *priv);
915 void (*enable)(struct mlx5e_priv *priv);
916 void (*disable)(struct mlx5e_priv *priv);
917 int (*update_rx)(struct mlx5e_priv *priv);
918 void (*update_stats)(struct mlx5e_priv *priv);
919 void (*update_carrier)(struct mlx5e_priv *priv);
920 unsigned int (*stats_grps_num)(struct mlx5e_priv *priv);
921 mlx5e_stats_grp_t *stats_grps;
922 const struct mlx5e_rx_handlers *rx_handlers;
928 void mlx5e_build_ptys2ethtool_map(void);
930 bool mlx5e_check_fragmented_striding_rq_cap(struct mlx5_core_dev *mdev);
932 void mlx5e_get_stats(struct net_device *dev, struct rtnl_link_stats64 *stats);
933 void mlx5e_fold_sw_stats64(struct mlx5e_priv *priv, struct rtnl_link_stats64 *s);
935 void mlx5e_init_l2_addr(struct mlx5e_priv *priv);
936 int mlx5e_self_test_num(struct mlx5e_priv *priv);
937 void mlx5e_self_test(struct net_device *ndev, struct ethtool_test *etest,
939 void mlx5e_set_rx_mode_work(struct work_struct *work);
941 int mlx5e_hwstamp_set(struct mlx5e_priv *priv, struct ifreq *ifr);
942 int mlx5e_hwstamp_get(struct mlx5e_priv *priv, struct ifreq *ifr);
943 int mlx5e_modify_rx_cqe_compression_locked(struct mlx5e_priv *priv, bool val);
945 int mlx5e_vlan_rx_add_vid(struct net_device *dev, __always_unused __be16 proto,
947 int mlx5e_vlan_rx_kill_vid(struct net_device *dev, __always_unused __be16 proto,
949 void mlx5e_timestamp_init(struct mlx5e_priv *priv);
951 struct mlx5e_redirect_rqt_param {
954 u32 rqn; /* Direct RQN (Non-RSS) */
957 struct mlx5e_channels *channels;
958 } rss; /* RSS data */
962 int mlx5e_redirect_rqt(struct mlx5e_priv *priv, u32 rqtn, int sz,
963 struct mlx5e_redirect_rqt_param rrp);
964 void mlx5e_build_indir_tir_ctx_hash(struct mlx5e_rss_params *rss_params,
965 const struct mlx5e_tirc_config *ttconfig,
966 void *tirc, bool inner);
967 void mlx5e_modify_tirs_hash(struct mlx5e_priv *priv, void *in);
968 struct mlx5e_tirc_config mlx5e_tirc_get_default_config(enum mlx5e_traffic_types tt);
970 struct mlx5e_xsk_param;
972 struct mlx5e_rq_param;
973 int mlx5e_open_rq(struct mlx5e_params *params, struct mlx5e_rq_param *param,
974 struct mlx5e_xsk_param *xsk, int node,
975 struct mlx5e_rq *rq);
976 int mlx5e_wait_for_min_rx_wqes(struct mlx5e_rq *rq, int wait_time);
977 void mlx5e_close_rq(struct mlx5e_rq *rq);
978 int mlx5e_create_rq(struct mlx5e_rq *rq, struct mlx5e_rq_param *param);
979 void mlx5e_destroy_rq(struct mlx5e_rq *rq);
981 struct mlx5e_sq_param;
982 int mlx5e_open_icosq(struct mlx5e_channel *c, struct mlx5e_params *params,
983 struct mlx5e_sq_param *param, struct mlx5e_icosq *sq);
984 void mlx5e_close_icosq(struct mlx5e_icosq *sq);
985 int mlx5e_open_xdpsq(struct mlx5e_channel *c, struct mlx5e_params *params,
986 struct mlx5e_sq_param *param, struct xsk_buff_pool *xsk_pool,
987 struct mlx5e_xdpsq *sq, bool is_redirect);
988 void mlx5e_close_xdpsq(struct mlx5e_xdpsq *sq);
990 struct mlx5e_create_cq_param {
991 struct napi_struct *napi;
992 struct mlx5e_ch_stats *ch_stats;
997 struct mlx5e_cq_param;
998 int mlx5e_open_cq(struct mlx5e_priv *priv, struct dim_cq_moder moder,
999 struct mlx5e_cq_param *param, struct mlx5e_create_cq_param *ccp,
1000 struct mlx5e_cq *cq);
1001 void mlx5e_close_cq(struct mlx5e_cq *cq);
1003 int mlx5e_open_locked(struct net_device *netdev);
1004 int mlx5e_close_locked(struct net_device *netdev);
1006 int mlx5e_open_channels(struct mlx5e_priv *priv,
1007 struct mlx5e_channels *chs);
1008 void mlx5e_close_channels(struct mlx5e_channels *chs);
1010 /* Function pointer to be used to modify HW or kernel settings while
1011 * switching channels
1013 typedef int (*mlx5e_fp_preactivate)(struct mlx5e_priv *priv, void *context);
1014 #define MLX5E_DEFINE_PREACTIVATE_WRAPPER_CTX(fn) \
1015 int fn##_ctx(struct mlx5e_priv *priv, void *context) \
1019 int mlx5e_safe_reopen_channels(struct mlx5e_priv *priv);
1020 int mlx5e_safe_switch_params(struct mlx5e_priv *priv,
1021 struct mlx5e_params *new_params,
1022 mlx5e_fp_preactivate preactivate,
1023 void *context, bool reset);
1024 int mlx5e_update_tx_netdev_queues(struct mlx5e_priv *priv);
1025 int mlx5e_num_channels_changed(struct mlx5e_priv *priv);
1026 int mlx5e_num_channels_changed_ctx(struct mlx5e_priv *priv, void *context);
1027 void mlx5e_activate_priv_channels(struct mlx5e_priv *priv);
1028 void mlx5e_deactivate_priv_channels(struct mlx5e_priv *priv);
1029 int mlx5e_ptp_rx_manage_fs_ctx(struct mlx5e_priv *priv, void *ctx);
1031 void mlx5e_build_default_indir_rqt(u32 *indirection_rqt, int len,
1034 int mlx5e_modify_rq_state(struct mlx5e_rq *rq, int curr_state, int next_state);
1035 void mlx5e_activate_rq(struct mlx5e_rq *rq);
1036 void mlx5e_deactivate_rq(struct mlx5e_rq *rq);
1037 void mlx5e_activate_icosq(struct mlx5e_icosq *icosq);
1038 void mlx5e_deactivate_icosq(struct mlx5e_icosq *icosq);
1040 int mlx5e_modify_sq(struct mlx5_core_dev *mdev, u32 sqn,
1041 struct mlx5e_modify_sq_param *p);
1042 int mlx5e_open_txqsq(struct mlx5e_channel *c, u32 tisn, int txq_ix,
1043 struct mlx5e_params *params, struct mlx5e_sq_param *param,
1044 struct mlx5e_txqsq *sq, int tc, u16 qos_queue_group_id, u16 qos_qid);
1045 void mlx5e_activate_txqsq(struct mlx5e_txqsq *sq);
1046 void mlx5e_deactivate_txqsq(struct mlx5e_txqsq *sq);
1047 void mlx5e_free_txqsq(struct mlx5e_txqsq *sq);
1048 void mlx5e_tx_disable_queue(struct netdev_queue *txq);
1049 int mlx5e_alloc_txqsq_db(struct mlx5e_txqsq *sq, int numa);
1050 void mlx5e_free_txqsq_db(struct mlx5e_txqsq *sq);
1051 struct mlx5e_create_sq_param;
1052 int mlx5e_create_sq_rdy(struct mlx5_core_dev *mdev,
1053 struct mlx5e_sq_param *param,
1054 struct mlx5e_create_sq_param *csp,
1055 u16 qos_queue_group_id,
1057 void mlx5e_tx_err_cqe_work(struct work_struct *recover_work);
1058 void mlx5e_close_txqsq(struct mlx5e_txqsq *sq);
1060 static inline bool mlx5_tx_swp_supported(struct mlx5_core_dev *mdev)
1062 return MLX5_CAP_ETH(mdev, swp) &&
1063 MLX5_CAP_ETH(mdev, swp_csum) && MLX5_CAP_ETH(mdev, swp_lso);
1066 extern const struct ethtool_ops mlx5e_ethtool_ops;
1068 int mlx5e_create_tir(struct mlx5_core_dev *mdev, struct mlx5e_tir *tir,
1070 void mlx5e_destroy_tir(struct mlx5_core_dev *mdev,
1071 struct mlx5e_tir *tir);
1072 int mlx5e_create_mdev_resources(struct mlx5_core_dev *mdev);
1073 void mlx5e_destroy_mdev_resources(struct mlx5_core_dev *mdev);
1074 int mlx5e_refresh_tirs(struct mlx5e_priv *priv, bool enable_uc_lb,
1076 void mlx5e_mkey_set_relaxed_ordering(struct mlx5_core_dev *mdev, void *mkc);
1078 /* common netdev helpers */
1079 void mlx5e_create_q_counters(struct mlx5e_priv *priv);
1080 void mlx5e_destroy_q_counters(struct mlx5e_priv *priv);
1081 int mlx5e_open_drop_rq(struct mlx5e_priv *priv,
1082 struct mlx5e_rq *drop_rq);
1083 void mlx5e_close_drop_rq(struct mlx5e_rq *drop_rq);
1084 int mlx5e_init_di_list(struct mlx5e_rq *rq, int wq_sz, int node);
1085 void mlx5e_free_di_list(struct mlx5e_rq *rq);
1087 int mlx5e_create_indirect_rqt(struct mlx5e_priv *priv);
1089 int mlx5e_create_indirect_tirs(struct mlx5e_priv *priv, bool inner_ttc);
1090 void mlx5e_destroy_indirect_tirs(struct mlx5e_priv *priv);
1092 int mlx5e_create_direct_rqts(struct mlx5e_priv *priv, struct mlx5e_tir *tirs, int n);
1093 void mlx5e_destroy_direct_rqts(struct mlx5e_priv *priv, struct mlx5e_tir *tirs, int n);
1094 int mlx5e_create_direct_tirs(struct mlx5e_priv *priv, struct mlx5e_tir *tirs, int n);
1095 void mlx5e_destroy_direct_tirs(struct mlx5e_priv *priv, struct mlx5e_tir *tirs, int n);
1096 void mlx5e_destroy_rqt(struct mlx5e_priv *priv, struct mlx5e_rqt *rqt);
1098 int mlx5e_create_tis(struct mlx5_core_dev *mdev, void *in, u32 *tisn);
1099 void mlx5e_destroy_tis(struct mlx5_core_dev *mdev, u32 tisn);
1101 int mlx5e_create_tises(struct mlx5e_priv *priv);
1102 void mlx5e_destroy_tises(struct mlx5e_priv *priv);
1103 int mlx5e_update_nic_rx(struct mlx5e_priv *priv);
1104 void mlx5e_update_carrier(struct mlx5e_priv *priv);
1105 int mlx5e_close(struct net_device *netdev);
1106 int mlx5e_open(struct net_device *netdev);
1108 void mlx5e_queue_update_stats(struct mlx5e_priv *priv);
1109 int mlx5e_bits_invert(unsigned long a, int size);
1111 int mlx5e_set_dev_port_mtu(struct mlx5e_priv *priv);
1112 int mlx5e_set_dev_port_mtu_ctx(struct mlx5e_priv *priv, void *context);
1113 int mlx5e_change_mtu(struct net_device *netdev, int new_mtu,
1114 mlx5e_fp_preactivate preactivate);
1115 void mlx5e_vxlan_set_netdev_info(struct mlx5e_priv *priv);
1117 /* ethtool helpers */
1118 void mlx5e_ethtool_get_drvinfo(struct mlx5e_priv *priv,
1119 struct ethtool_drvinfo *drvinfo);
1120 void mlx5e_ethtool_get_strings(struct mlx5e_priv *priv,
1121 uint32_t stringset, uint8_t *data);
1122 int mlx5e_ethtool_get_sset_count(struct mlx5e_priv *priv, int sset);
1123 void mlx5e_ethtool_get_ethtool_stats(struct mlx5e_priv *priv,
1124 struct ethtool_stats *stats, u64 *data);
1125 void mlx5e_ethtool_get_ringparam(struct mlx5e_priv *priv,
1126 struct ethtool_ringparam *param);
1127 int mlx5e_ethtool_set_ringparam(struct mlx5e_priv *priv,
1128 struct ethtool_ringparam *param);
1129 void mlx5e_ethtool_get_channels(struct mlx5e_priv *priv,
1130 struct ethtool_channels *ch);
1131 int mlx5e_ethtool_set_channels(struct mlx5e_priv *priv,
1132 struct ethtool_channels *ch);
1133 int mlx5e_ethtool_get_coalesce(struct mlx5e_priv *priv,
1134 struct ethtool_coalesce *coal);
1135 int mlx5e_ethtool_set_coalesce(struct mlx5e_priv *priv,
1136 struct ethtool_coalesce *coal);
1137 int mlx5e_ethtool_get_link_ksettings(struct mlx5e_priv *priv,
1138 struct ethtool_link_ksettings *link_ksettings);
1139 int mlx5e_ethtool_set_link_ksettings(struct mlx5e_priv *priv,
1140 const struct ethtool_link_ksettings *link_ksettings);
1141 int mlx5e_get_rxfh(struct net_device *netdev, u32 *indir, u8 *key, u8 *hfunc);
1142 int mlx5e_set_rxfh(struct net_device *dev, const u32 *indir, const u8 *key,
1144 int mlx5e_get_rxnfc(struct net_device *dev, struct ethtool_rxnfc *info,
1146 int mlx5e_set_rxnfc(struct net_device *dev, struct ethtool_rxnfc *cmd);
1147 u32 mlx5e_ethtool_get_rxfh_key_size(struct mlx5e_priv *priv);
1148 u32 mlx5e_ethtool_get_rxfh_indir_size(struct mlx5e_priv *priv);
1149 int mlx5e_ethtool_get_ts_info(struct mlx5e_priv *priv,
1150 struct ethtool_ts_info *info);
1151 int mlx5e_ethtool_flash_device(struct mlx5e_priv *priv,
1152 struct ethtool_flash *flash);
1153 void mlx5e_ethtool_get_pauseparam(struct mlx5e_priv *priv,
1154 struct ethtool_pauseparam *pauseparam);
1155 int mlx5e_ethtool_set_pauseparam(struct mlx5e_priv *priv,
1156 struct ethtool_pauseparam *pauseparam);
1158 /* mlx5e generic netdev management API */
1159 static inline unsigned int
1160 mlx5e_calc_max_nch(struct mlx5e_priv *priv, const struct mlx5e_profile *profile)
1162 return priv->netdev->num_rx_queues / max_t(u8, profile->rq_groups, 1);
1166 mlx5e_tx_mpwqe_supported(struct mlx5_core_dev *mdev)
1168 return !is_kdump_kernel() &&
1169 MLX5_CAP_ETH(mdev, enhanced_multi_pkt_send_wqe);
1172 int mlx5e_priv_init(struct mlx5e_priv *priv,
1173 struct net_device *netdev,
1174 struct mlx5_core_dev *mdev);
1175 void mlx5e_priv_cleanup(struct mlx5e_priv *priv);
1177 mlx5e_create_netdev(struct mlx5_core_dev *mdev, unsigned int txqs, unsigned int rxqs);
1178 int mlx5e_attach_netdev(struct mlx5e_priv *priv);
1179 void mlx5e_detach_netdev(struct mlx5e_priv *priv);
1180 void mlx5e_destroy_netdev(struct mlx5e_priv *priv);
1181 int mlx5e_netdev_change_profile(struct mlx5e_priv *priv,
1182 const struct mlx5e_profile *new_profile, void *new_ppriv);
1183 void mlx5e_netdev_attach_nic_profile(struct mlx5e_priv *priv);
1184 void mlx5e_set_netdev_mtu_boundaries(struct mlx5e_priv *priv);
1185 void mlx5e_build_nic_params(struct mlx5e_priv *priv, struct mlx5e_xsk *xsk, u16 mtu);
1186 void mlx5e_build_rss_params(struct mlx5e_rss_params *rss_params,
1188 void mlx5e_rx_dim_work(struct work_struct *work);
1189 void mlx5e_tx_dim_work(struct work_struct *work);
1191 netdev_features_t mlx5e_features_check(struct sk_buff *skb,
1192 struct net_device *netdev,
1193 netdev_features_t features);
1194 int mlx5e_set_features(struct net_device *netdev, netdev_features_t features);
1195 #ifdef CONFIG_MLX5_ESWITCH
1196 int mlx5e_set_vf_mac(struct net_device *dev, int vf, u8 *mac);
1197 int mlx5e_set_vf_rate(struct net_device *dev, int vf, int min_tx_rate, int max_tx_rate);
1198 int mlx5e_get_vf_config(struct net_device *dev, int vf, struct ifla_vf_info *ivi);
1199 int mlx5e_get_vf_stats(struct net_device *dev, int vf, struct ifla_vf_stats *vf_stats);
1201 #endif /* __MLX5_EN_H__ */