1 /*
2  * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
3  *
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:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
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.
22  *
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
30  * SOFTWARE.
31  *
32  */
33 
34 #ifndef _MLX4_EN_H_
35 #define _MLX4_EN_H_
36 
37 #include <linux/bitops.h>
38 #include <linux/compiler.h>
39 #include <linux/list.h>
40 #include <linux/mutex.h>
41 #include <linux/netdevice.h>
42 #include <linux/if_vlan.h>
43 #include <linux/net_tstamp.h>
44 #ifdef CONFIG_MLX4_EN_DCB
45 #include <linux/dcbnl.h>
46 #endif
47 #include <linux/cpu_rmap.h>
48 #include <linux/ptp_clock_kernel.h>
49 
50 #include <linux/mlx4/device.h>
51 #include <linux/mlx4/qp.h>
52 #include <linux/mlx4/cq.h>
53 #include <linux/mlx4/srq.h>
54 #include <linux/mlx4/doorbell.h>
55 #include <linux/mlx4/cmd.h>
56 
57 #include "en_port.h"
58 
59 #define DRV_NAME	"mlx4_en"
60 #define DRV_VERSION	"2.2-1"
61 #define DRV_RELDATE	"Feb 2014"
62 
63 #define MLX4_EN_MSG_LEVEL	(NETIF_MSG_LINK | NETIF_MSG_IFDOWN)
64 
65 /*
66  * Device constants
67  */
68 
69 
70 #define MLX4_EN_PAGE_SHIFT	12
71 #define MLX4_EN_PAGE_SIZE	(1 << MLX4_EN_PAGE_SHIFT)
72 #define DEF_RX_RINGS		16
73 #define MAX_RX_RINGS		128
74 #define MIN_RX_RINGS		4
75 #define TXBB_SIZE		64
76 #define HEADROOM		(2048 / TXBB_SIZE + 1)
77 #define STAMP_STRIDE		64
78 #define STAMP_DWORDS		(STAMP_STRIDE / 4)
79 #define STAMP_SHIFT		31
80 #define STAMP_VAL		0x7fffffff
81 #define STATS_DELAY		(HZ / 4)
82 #define SERVICE_TASK_DELAY	(HZ / 4)
83 #define MAX_NUM_OF_FS_RULES	256
84 
85 #define MLX4_EN_FILTER_HASH_SHIFT 4
86 #define MLX4_EN_FILTER_EXPIRY_QUOTA 60
87 
88 /* Typical TSO descriptor with 16 gather entries is 352 bytes... */
89 #define MAX_DESC_SIZE		512
90 #define MAX_DESC_TXBBS		(MAX_DESC_SIZE / TXBB_SIZE)
91 
92 /*
93  * OS related constants and tunables
94  */
95 
96 #define MLX4_EN_WATCHDOG_TIMEOUT	(15 * HZ)
97 
98 /* Use the maximum between 16384 and a single page */
99 #define MLX4_EN_ALLOC_SIZE	PAGE_ALIGN(16384)
100 
101 #define MLX4_EN_ALLOC_PREFER_ORDER	PAGE_ALLOC_COSTLY_ORDER
102 
103 /* Receive fragment sizes; we use at most 3 fragments (for 9600 byte MTU
104  * and 4K allocations) */
105 enum {
106 	FRAG_SZ0 = 1536 - NET_IP_ALIGN,
107 	FRAG_SZ1 = 4096,
108 	FRAG_SZ2 = 4096,
109 	FRAG_SZ3 = MLX4_EN_ALLOC_SIZE
110 };
111 #define MLX4_EN_MAX_RX_FRAGS	4
112 
113 /* Maximum ring sizes */
114 #define MLX4_EN_MAX_TX_SIZE	8192
115 #define MLX4_EN_MAX_RX_SIZE	8192
116 
117 /* Minimum ring size for our page-allocation scheme to work */
118 #define MLX4_EN_MIN_RX_SIZE	(MLX4_EN_ALLOC_SIZE / SMP_CACHE_BYTES)
119 #define MLX4_EN_MIN_TX_SIZE	(4096 / TXBB_SIZE)
120 
121 #define MLX4_EN_SMALL_PKT_SIZE		64
122 #define MLX4_EN_MAX_TX_RING_P_UP	32
123 #define MLX4_EN_NUM_UP			8
124 #define MLX4_EN_DEF_TX_RING_SIZE	512
125 #define MLX4_EN_DEF_RX_RING_SIZE  	1024
126 #define MAX_TX_RINGS			(MLX4_EN_MAX_TX_RING_P_UP * \
127 					 MLX4_EN_NUM_UP)
128 
129 #define MLX4_EN_DEFAULT_TX_WORK		256
130 
131 /* Target number of packets to coalesce with interrupt moderation */
132 #define MLX4_EN_RX_COAL_TARGET	44
133 #define MLX4_EN_RX_COAL_TIME	0x10
134 
135 #define MLX4_EN_TX_COAL_PKTS	16
136 #define MLX4_EN_TX_COAL_TIME	0x10
137 
138 #define MLX4_EN_RX_RATE_LOW		400000
139 #define MLX4_EN_RX_COAL_TIME_LOW	0
140 #define MLX4_EN_RX_RATE_HIGH		450000
141 #define MLX4_EN_RX_COAL_TIME_HIGH	128
142 #define MLX4_EN_RX_SIZE_THRESH		1024
143 #define MLX4_EN_RX_RATE_THRESH		(1000000 / MLX4_EN_RX_COAL_TIME_HIGH)
144 #define MLX4_EN_SAMPLE_INTERVAL		0
145 #define MLX4_EN_AVG_PKT_SMALL		256
146 
147 #define MLX4_EN_AUTO_CONF	0xffff
148 
149 #define MLX4_EN_DEF_RX_PAUSE	1
150 #define MLX4_EN_DEF_TX_PAUSE	1
151 
152 /* Interval between successive polls in the Tx routine when polling is used
153    instead of interrupts (in per-core Tx rings) - should be power of 2 */
154 #define MLX4_EN_TX_POLL_MODER	16
155 #define MLX4_EN_TX_POLL_TIMEOUT	(HZ / 4)
156 
157 #define ETH_LLC_SNAP_SIZE	8
158 
159 #define SMALL_PACKET_SIZE      (256 - NET_IP_ALIGN)
160 #define HEADER_COPY_SIZE       (128 - NET_IP_ALIGN)
161 #define MLX4_LOOPBACK_TEST_PAYLOAD (HEADER_COPY_SIZE - ETH_HLEN)
162 
163 #define MLX4_EN_MIN_MTU		46
164 #define ETH_BCAST		0xffffffffffffULL
165 
166 #define MLX4_EN_LOOPBACK_RETRIES	5
167 #define MLX4_EN_LOOPBACK_TIMEOUT	100
168 
169 #ifdef MLX4_EN_PERF_STAT
170 /* Number of samples to 'average' */
171 #define AVG_SIZE			128
172 #define AVG_FACTOR			1024
173 #define NUM_PERF_STATS			NUM_PERF_COUNTERS
174 
175 #define INC_PERF_COUNTER(cnt)		(++(cnt))
176 #define ADD_PERF_COUNTER(cnt, add)	((cnt) += (add))
177 #define AVG_PERF_COUNTER(cnt, sample) \
178 	((cnt) = ((cnt) * (AVG_SIZE - 1) + (sample) * AVG_FACTOR) / AVG_SIZE)
179 #define GET_PERF_COUNTER(cnt)		(cnt)
180 #define GET_AVG_PERF_COUNTER(cnt)	((cnt) / AVG_FACTOR)
181 
182 #else
183 
184 #define NUM_PERF_STATS			0
185 #define INC_PERF_COUNTER(cnt)		do {} while (0)
186 #define ADD_PERF_COUNTER(cnt, add)	do {} while (0)
187 #define AVG_PERF_COUNTER(cnt, sample)	do {} while (0)
188 #define GET_PERF_COUNTER(cnt)		(0)
189 #define GET_AVG_PERF_COUNTER(cnt)	(0)
190 #endif /* MLX4_EN_PERF_STAT */
191 
192 /* Constants for TX flow */
193 enum {
194 	MAX_INLINE = 104, /* 128 - 16 - 4 - 4 */
195 	MAX_BF = 256,
196 	MIN_PKT_LEN = 17,
197 };
198 
199 /*
200  * Configurables
201  */
202 
203 enum cq_type {
204 	RX = 0,
205 	TX = 1,
206 };
207 
208 
209 /*
210  * Useful macros
211  */
212 #define ROUNDUP_LOG2(x)		ilog2(roundup_pow_of_two(x))
213 #define XNOR(x, y)		(!(x) == !(y))
214 
215 
216 struct mlx4_en_tx_info {
217 	struct sk_buff *skb;
218 	u32 nr_txbb;
219 	u32 nr_bytes;
220 	u8 linear;
221 	u8 data_offset;
222 	u8 inl;
223 	u8 ts_requested;
224 };
225 
226 
227 #define MLX4_EN_BIT_DESC_OWN	0x80000000
228 #define CTRL_SIZE	sizeof(struct mlx4_wqe_ctrl_seg)
229 #define MLX4_EN_MEMTYPE_PAD	0x100
230 #define DS_SIZE		sizeof(struct mlx4_wqe_data_seg)
231 
232 
233 struct mlx4_en_tx_desc {
234 	struct mlx4_wqe_ctrl_seg ctrl;
235 	union {
236 		struct mlx4_wqe_data_seg data; /* at least one data segment */
237 		struct mlx4_wqe_lso_seg lso;
238 		struct mlx4_wqe_inline_seg inl;
239 	};
240 };
241 
242 #define MLX4_EN_USE_SRQ		0x01000000
243 
244 #define MLX4_EN_CX3_LOW_ID	0x1000
245 #define MLX4_EN_CX3_HIGH_ID	0x1005
246 
247 struct mlx4_en_rx_alloc {
248 	struct page	*page;
249 	dma_addr_t	dma;
250 	u32		page_offset;
251 	u32		page_size;
252 };
253 
254 struct mlx4_en_tx_ring {
255 	struct mlx4_hwq_resources wqres;
256 	u32 size ; /* number of TXBBs */
257 	u32 size_mask;
258 	u16 stride;
259 	u16 cqn;	/* index of port CQ associated with this ring */
260 	u32 prod;
261 	u32 cons;
262 	u32 buf_size;
263 	u32 doorbell_qpn;
264 	void *buf;
265 	u16 poll_cnt;
266 	struct mlx4_en_tx_info *tx_info;
267 	u8 *bounce_buf;
268 	u8 queue_index;
269 	cpumask_t affinity_mask;
270 	u32 last_nr_txbb;
271 	struct mlx4_qp qp;
272 	struct mlx4_qp_context context;
273 	int qpn;
274 	enum mlx4_qp_state qp_state;
275 	struct mlx4_srq dummy;
276 	unsigned long bytes;
277 	unsigned long packets;
278 	unsigned long tx_csum;
279 	unsigned long queue_stopped;
280 	unsigned long wake_queue;
281 	struct mlx4_bf bf;
282 	bool bf_enabled;
283 	struct netdev_queue *tx_queue;
284 	int hwtstamp_tx_type;
285 	int inline_thold;
286 };
287 
288 struct mlx4_en_rx_desc {
289 	/* actual number of entries depends on rx ring stride */
290 	struct mlx4_wqe_data_seg data[0];
291 };
292 
293 struct mlx4_en_rx_ring {
294 	struct mlx4_hwq_resources wqres;
295 	struct mlx4_en_rx_alloc page_alloc[MLX4_EN_MAX_RX_FRAGS];
296 	u32 size ;	/* number of Rx descs*/
297 	u32 actual_size;
298 	u32 size_mask;
299 	u16 stride;
300 	u16 log_stride;
301 	u16 cqn;	/* index of port CQ associated with this ring */
302 	u32 prod;
303 	u32 cons;
304 	u32 buf_size;
305 	u8  fcs_del;
306 	void *buf;
307 	void *rx_info;
308 	unsigned long bytes;
309 	unsigned long packets;
310 #ifdef CONFIG_NET_RX_BUSY_POLL
311 	unsigned long yields;
312 	unsigned long misses;
313 	unsigned long cleaned;
314 #endif
315 	unsigned long csum_ok;
316 	unsigned long csum_none;
317 	int hwtstamp_rx_filter;
318 	cpumask_var_t affinity_mask;
319 };
320 
321 struct mlx4_en_cq {
322 	struct mlx4_cq          mcq;
323 	struct mlx4_hwq_resources wqres;
324 	int                     ring;
325 	struct net_device      *dev;
326 	struct napi_struct	napi;
327 	int size;
328 	int buf_size;
329 	unsigned vector;
330 	enum cq_type is_tx;
331 	u16 moder_time;
332 	u16 moder_cnt;
333 	struct mlx4_cqe *buf;
334 #define MLX4_EN_OPCODE_ERROR	0x1e
335 
336 #ifdef CONFIG_NET_RX_BUSY_POLL
337 	unsigned int state;
338 #define MLX4_EN_CQ_STATE_IDLE        0
339 #define MLX4_EN_CQ_STATE_NAPI     1    /* NAPI owns this CQ */
340 #define MLX4_EN_CQ_STATE_POLL     2    /* poll owns this CQ */
341 #define MLX4_CQ_LOCKED (MLX4_EN_CQ_STATE_NAPI | MLX4_EN_CQ_STATE_POLL)
342 #define MLX4_EN_CQ_STATE_NAPI_YIELD  4    /* NAPI yielded this CQ */
343 #define MLX4_EN_CQ_STATE_POLL_YIELD  8    /* poll yielded this CQ */
344 #define CQ_YIELD (MLX4_EN_CQ_STATE_NAPI_YIELD | MLX4_EN_CQ_STATE_POLL_YIELD)
345 #define CQ_USER_PEND (MLX4_EN_CQ_STATE_POLL | MLX4_EN_CQ_STATE_POLL_YIELD)
346 	spinlock_t poll_lock; /* protects from LLS/napi conflicts */
347 #endif  /* CONFIG_NET_RX_BUSY_POLL */
348 	struct irq_desc *irq_desc;
349 };
350 
351 struct mlx4_en_port_profile {
352 	u32 flags;
353 	u32 tx_ring_num;
354 	u32 rx_ring_num;
355 	u32 tx_ring_size;
356 	u32 rx_ring_size;
357 	u8 rx_pause;
358 	u8 rx_ppp;
359 	u8 tx_pause;
360 	u8 tx_ppp;
361 	int rss_rings;
362 	int inline_thold;
363 };
364 
365 struct mlx4_en_profile {
366 	int rss_xor;
367 	int udp_rss;
368 	u8 rss_mask;
369 	u32 active_ports;
370 	u32 small_pkt_int;
371 	u8 no_reset;
372 	u8 num_tx_rings_p_up;
373 	struct mlx4_en_port_profile prof[MLX4_MAX_PORTS + 1];
374 };
375 
376 struct mlx4_en_dev {
377 	struct mlx4_dev         *dev;
378 	struct pci_dev		*pdev;
379 	struct mutex		state_lock;
380 	struct net_device       *pndev[MLX4_MAX_PORTS + 1];
381 	u32                     port_cnt;
382 	bool			device_up;
383 	struct mlx4_en_profile  profile;
384 	u32			LSO_support;
385 	struct workqueue_struct *workqueue;
386 	struct device           *dma_device;
387 	void __iomem            *uar_map;
388 	struct mlx4_uar         priv_uar;
389 	struct mlx4_mr		mr;
390 	u32                     priv_pdn;
391 	spinlock_t              uar_lock;
392 	u8			mac_removed[MLX4_MAX_PORTS + 1];
393 	rwlock_t		clock_lock;
394 	u32			nominal_c_mult;
395 	struct cyclecounter	cycles;
396 	struct timecounter	clock;
397 	unsigned long		last_overflow_check;
398 	unsigned long		overflow_period;
399 	struct ptp_clock	*ptp_clock;
400 	struct ptp_clock_info	ptp_clock_info;
401 };
402 
403 
404 struct mlx4_en_rss_map {
405 	int base_qpn;
406 	struct mlx4_qp qps[MAX_RX_RINGS];
407 	enum mlx4_qp_state state[MAX_RX_RINGS];
408 	struct mlx4_qp indir_qp;
409 	enum mlx4_qp_state indir_state;
410 };
411 
412 struct mlx4_en_port_state {
413 	int link_state;
414 	int link_speed;
415 	int transciver;
416 };
417 
418 struct mlx4_en_pkt_stats {
419 	unsigned long broadcast;
420 	unsigned long rx_prio[8];
421 	unsigned long tx_prio[8];
422 #define NUM_PKT_STATS		17
423 };
424 
425 struct mlx4_en_port_stats {
426 	unsigned long tso_packets;
427 	unsigned long queue_stopped;
428 	unsigned long wake_queue;
429 	unsigned long tx_timeout;
430 	unsigned long rx_alloc_failed;
431 	unsigned long rx_chksum_good;
432 	unsigned long rx_chksum_none;
433 	unsigned long tx_chksum_offload;
434 #define NUM_PORT_STATS		8
435 };
436 
437 struct mlx4_en_perf_stats {
438 	u32 tx_poll;
439 	u64 tx_pktsz_avg;
440 	u32 inflight_avg;
441 	u16 tx_coal_avg;
442 	u16 rx_coal_avg;
443 	u32 napi_quota;
444 #define NUM_PERF_COUNTERS		6
445 };
446 
447 enum mlx4_en_mclist_act {
448 	MCLIST_NONE,
449 	MCLIST_REM,
450 	MCLIST_ADD,
451 };
452 
453 struct mlx4_en_mc_list {
454 	struct list_head	list;
455 	enum mlx4_en_mclist_act	action;
456 	u8			addr[ETH_ALEN];
457 	u64			reg_id;
458 	u64			tunnel_reg_id;
459 };
460 
461 struct mlx4_en_frag_info {
462 	u16 frag_size;
463 	u16 frag_prefix_size;
464 	u16 frag_stride;
465 	u16 frag_align;
466 };
467 
468 #ifdef CONFIG_MLX4_EN_DCB
469 /* Minimal TC BW - setting to 0 will block traffic */
470 #define MLX4_EN_BW_MIN 1
471 #define MLX4_EN_BW_MAX 100 /* Utilize 100% of the line */
472 
473 #define MLX4_EN_TC_ETS 7
474 
475 #endif
476 
477 struct ethtool_flow_id {
478 	struct list_head list;
479 	struct ethtool_rx_flow_spec flow_spec;
480 	u64 id;
481 };
482 
483 enum {
484 	MLX4_EN_FLAG_PROMISC		= (1 << 0),
485 	MLX4_EN_FLAG_MC_PROMISC		= (1 << 1),
486 	/* whether we need to enable hardware loopback by putting dmac
487 	 * in Tx WQE
488 	 */
489 	MLX4_EN_FLAG_ENABLE_HW_LOOPBACK	= (1 << 2),
490 	/* whether we need to drop packets that hardware loopback-ed */
491 	MLX4_EN_FLAG_RX_FILTER_NEEDED	= (1 << 3),
492 	MLX4_EN_FLAG_FORCE_PROMISC	= (1 << 4)
493 };
494 
495 #define MLX4_EN_MAC_HASH_SIZE (1 << BITS_PER_BYTE)
496 #define MLX4_EN_MAC_HASH_IDX 5
497 
498 struct mlx4_en_priv {
499 	struct mlx4_en_dev *mdev;
500 	struct mlx4_en_port_profile *prof;
501 	struct net_device *dev;
502 	unsigned long active_vlans[BITS_TO_LONGS(VLAN_N_VID)];
503 	struct net_device_stats stats;
504 	struct net_device_stats ret_stats;
505 	struct mlx4_en_port_state port_state;
506 	spinlock_t stats_lock;
507 	struct ethtool_flow_id ethtool_rules[MAX_NUM_OF_FS_RULES];
508 	/* To allow rules removal while port is going down */
509 	struct list_head ethtool_list;
510 
511 	unsigned long last_moder_packets[MAX_RX_RINGS];
512 	unsigned long last_moder_tx_packets;
513 	unsigned long last_moder_bytes[MAX_RX_RINGS];
514 	unsigned long last_moder_jiffies;
515 	int last_moder_time[MAX_RX_RINGS];
516 	u16 rx_usecs;
517 	u16 rx_frames;
518 	u16 tx_usecs;
519 	u16 tx_frames;
520 	u32 pkt_rate_low;
521 	u16 rx_usecs_low;
522 	u32 pkt_rate_high;
523 	u16 rx_usecs_high;
524 	u16 sample_interval;
525 	u16 adaptive_rx_coal;
526 	u32 msg_enable;
527 	u32 loopback_ok;
528 	u32 validate_loopback;
529 
530 	struct mlx4_hwq_resources res;
531 	int link_state;
532 	int last_link_state;
533 	bool port_up;
534 	int port;
535 	int registered;
536 	int allocated;
537 	int stride;
538 	unsigned char prev_mac[ETH_ALEN + 2];
539 	int mac_index;
540 	unsigned max_mtu;
541 	int base_qpn;
542 	int cqe_factor;
543 
544 	struct mlx4_en_rss_map rss_map;
545 	__be32 ctrl_flags;
546 	u32 flags;
547 	u8 num_tx_rings_p_up;
548 	u32 tx_work_limit;
549 	u32 tx_ring_num;
550 	u32 rx_ring_num;
551 	u32 rx_skb_size;
552 	struct mlx4_en_frag_info frag_info[MLX4_EN_MAX_RX_FRAGS];
553 	u16 num_frags;
554 	u16 log_rx_info;
555 
556 	struct mlx4_en_tx_ring **tx_ring;
557 	struct mlx4_en_rx_ring *rx_ring[MAX_RX_RINGS];
558 	struct mlx4_en_cq **tx_cq;
559 	struct mlx4_en_cq *rx_cq[MAX_RX_RINGS];
560 	struct mlx4_qp drop_qp;
561 	struct work_struct rx_mode_task;
562 	struct work_struct watchdog_task;
563 	struct work_struct linkstate_task;
564 	struct delayed_work stats_task;
565 	struct delayed_work service_task;
566 #ifdef CONFIG_MLX4_EN_VXLAN
567 	struct work_struct vxlan_add_task;
568 	struct work_struct vxlan_del_task;
569 #endif
570 	struct mlx4_en_perf_stats pstats;
571 	struct mlx4_en_pkt_stats pkstats;
572 	struct mlx4_en_port_stats port_stats;
573 	u64 stats_bitmap;
574 	struct list_head mc_list;
575 	struct list_head curr_list;
576 	u64 broadcast_id;
577 	struct mlx4_en_stat_out_mbox hw_stats;
578 	int vids[128];
579 	bool wol;
580 	struct device *ddev;
581 	int base_tx_qpn;
582 	struct hlist_head mac_hash[MLX4_EN_MAC_HASH_SIZE];
583 	struct hwtstamp_config hwtstamp_config;
584 
585 #ifdef CONFIG_MLX4_EN_DCB
586 	struct ieee_ets ets;
587 	u16 maxrate[IEEE_8021QAZ_MAX_TCS];
588 #endif
589 #ifdef CONFIG_RFS_ACCEL
590 	spinlock_t filters_lock;
591 	int last_filter_id;
592 	struct list_head filters;
593 	struct hlist_head filter_hash[1 << MLX4_EN_FILTER_HASH_SHIFT];
594 #endif
595 	u64 tunnel_reg_id;
596 	__be16 vxlan_port;
597 };
598 
599 enum mlx4_en_wol {
600 	MLX4_EN_WOL_MAGIC = (1ULL << 61),
601 	MLX4_EN_WOL_ENABLED = (1ULL << 62),
602 };
603 
604 struct mlx4_mac_entry {
605 	struct hlist_node hlist;
606 	unsigned char mac[ETH_ALEN + 2];
607 	u64 reg_id;
608 	struct rcu_head rcu;
609 };
610 
611 #ifdef CONFIG_NET_RX_BUSY_POLL
612 static inline void mlx4_en_cq_init_lock(struct mlx4_en_cq *cq)
613 {
614 	spin_lock_init(&cq->poll_lock);
615 	cq->state = MLX4_EN_CQ_STATE_IDLE;
616 }
617 
618 /* called from the device poll rutine to get ownership of a cq */
619 static inline bool mlx4_en_cq_lock_napi(struct mlx4_en_cq *cq)
620 {
621 	int rc = true;
622 	spin_lock(&cq->poll_lock);
623 	if (cq->state & MLX4_CQ_LOCKED) {
624 		WARN_ON(cq->state & MLX4_EN_CQ_STATE_NAPI);
625 		cq->state |= MLX4_EN_CQ_STATE_NAPI_YIELD;
626 		rc = false;
627 	} else
628 		/* we don't care if someone yielded */
629 		cq->state = MLX4_EN_CQ_STATE_NAPI;
630 	spin_unlock(&cq->poll_lock);
631 	return rc;
632 }
633 
634 /* returns true is someone tried to get the cq while napi had it */
635 static inline bool mlx4_en_cq_unlock_napi(struct mlx4_en_cq *cq)
636 {
637 	int rc = false;
638 	spin_lock(&cq->poll_lock);
639 	WARN_ON(cq->state & (MLX4_EN_CQ_STATE_POLL |
640 			       MLX4_EN_CQ_STATE_NAPI_YIELD));
641 
642 	if (cq->state & MLX4_EN_CQ_STATE_POLL_YIELD)
643 		rc = true;
644 	cq->state = MLX4_EN_CQ_STATE_IDLE;
645 	spin_unlock(&cq->poll_lock);
646 	return rc;
647 }
648 
649 /* called from mlx4_en_low_latency_poll() */
650 static inline bool mlx4_en_cq_lock_poll(struct mlx4_en_cq *cq)
651 {
652 	int rc = true;
653 	spin_lock_bh(&cq->poll_lock);
654 	if ((cq->state & MLX4_CQ_LOCKED)) {
655 		struct net_device *dev = cq->dev;
656 		struct mlx4_en_priv *priv = netdev_priv(dev);
657 		struct mlx4_en_rx_ring *rx_ring = priv->rx_ring[cq->ring];
658 
659 		cq->state |= MLX4_EN_CQ_STATE_POLL_YIELD;
660 		rc = false;
661 		rx_ring->yields++;
662 	} else
663 		/* preserve yield marks */
664 		cq->state |= MLX4_EN_CQ_STATE_POLL;
665 	spin_unlock_bh(&cq->poll_lock);
666 	return rc;
667 }
668 
669 /* returns true if someone tried to get the cq while it was locked */
670 static inline bool mlx4_en_cq_unlock_poll(struct mlx4_en_cq *cq)
671 {
672 	int rc = false;
673 	spin_lock_bh(&cq->poll_lock);
674 	WARN_ON(cq->state & (MLX4_EN_CQ_STATE_NAPI));
675 
676 	if (cq->state & MLX4_EN_CQ_STATE_POLL_YIELD)
677 		rc = true;
678 	cq->state = MLX4_EN_CQ_STATE_IDLE;
679 	spin_unlock_bh(&cq->poll_lock);
680 	return rc;
681 }
682 
683 /* true if a socket is polling, even if it did not get the lock */
684 static inline bool mlx4_en_cq_busy_polling(struct mlx4_en_cq *cq)
685 {
686 	WARN_ON(!(cq->state & MLX4_CQ_LOCKED));
687 	return cq->state & CQ_USER_PEND;
688 }
689 #else
690 static inline void mlx4_en_cq_init_lock(struct mlx4_en_cq *cq)
691 {
692 }
693 
694 static inline bool mlx4_en_cq_lock_napi(struct mlx4_en_cq *cq)
695 {
696 	return true;
697 }
698 
699 static inline bool mlx4_en_cq_unlock_napi(struct mlx4_en_cq *cq)
700 {
701 	return false;
702 }
703 
704 static inline bool mlx4_en_cq_lock_poll(struct mlx4_en_cq *cq)
705 {
706 	return false;
707 }
708 
709 static inline bool mlx4_en_cq_unlock_poll(struct mlx4_en_cq *cq)
710 {
711 	return false;
712 }
713 
714 static inline bool mlx4_en_cq_busy_polling(struct mlx4_en_cq *cq)
715 {
716 	return false;
717 }
718 #endif /* CONFIG_NET_RX_BUSY_POLL */
719 
720 #define MLX4_EN_WOL_DO_MODIFY (1ULL << 63)
721 
722 void mlx4_en_update_loopback_state(struct net_device *dev,
723 				   netdev_features_t features);
724 
725 void mlx4_en_destroy_netdev(struct net_device *dev);
726 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
727 			struct mlx4_en_port_profile *prof);
728 
729 int mlx4_en_start_port(struct net_device *dev);
730 void mlx4_en_stop_port(struct net_device *dev, int detach);
731 
732 void mlx4_en_free_resources(struct mlx4_en_priv *priv);
733 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv);
734 
735 int mlx4_en_create_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq **pcq,
736 		      int entries, int ring, enum cq_type mode, int node);
737 void mlx4_en_destroy_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq **pcq);
738 int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
739 			int cq_idx);
740 void mlx4_en_deactivate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
741 int mlx4_en_set_cq_moder(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
742 int mlx4_en_arm_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
743 
744 void mlx4_en_tx_irq(struct mlx4_cq *mcq);
745 u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb,
746 			 void *accel_priv, select_queue_fallback_t fallback);
747 netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev);
748 
749 int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
750 			   struct mlx4_en_tx_ring **pring,
751 			   int qpn, u32 size, u16 stride,
752 			   int node, int queue_index);
753 void mlx4_en_destroy_tx_ring(struct mlx4_en_priv *priv,
754 			     struct mlx4_en_tx_ring **pring);
755 int mlx4_en_activate_tx_ring(struct mlx4_en_priv *priv,
756 			     struct mlx4_en_tx_ring *ring,
757 			     int cq, int user_prio);
758 void mlx4_en_deactivate_tx_ring(struct mlx4_en_priv *priv,
759 				struct mlx4_en_tx_ring *ring);
760 void mlx4_en_set_num_rx_rings(struct mlx4_en_dev *mdev);
761 int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,
762 			   struct mlx4_en_rx_ring **pring,
763 			   u32 size, u16 stride, int node);
764 void mlx4_en_destroy_rx_ring(struct mlx4_en_priv *priv,
765 			     struct mlx4_en_rx_ring **pring,
766 			     u32 size, u16 stride);
767 int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv);
768 void mlx4_en_deactivate_rx_ring(struct mlx4_en_priv *priv,
769 				struct mlx4_en_rx_ring *ring);
770 int mlx4_en_process_rx_cq(struct net_device *dev,
771 			  struct mlx4_en_cq *cq,
772 			  int budget);
773 int mlx4_en_poll_rx_cq(struct napi_struct *napi, int budget);
774 int mlx4_en_poll_tx_cq(struct napi_struct *napi, int budget);
775 void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
776 		int is_tx, int rss, int qpn, int cqn, int user_prio,
777 		struct mlx4_qp_context *context);
778 void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event);
779 int mlx4_en_map_buffer(struct mlx4_buf *buf);
780 void mlx4_en_unmap_buffer(struct mlx4_buf *buf);
781 
782 void mlx4_en_calc_rx_buf(struct net_device *dev);
783 int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv);
784 void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv);
785 int mlx4_en_create_drop_qp(struct mlx4_en_priv *priv);
786 void mlx4_en_destroy_drop_qp(struct mlx4_en_priv *priv);
787 int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring);
788 void mlx4_en_rx_irq(struct mlx4_cq *mcq);
789 
790 int mlx4_SET_MCAST_FLTR(struct mlx4_dev *dev, u8 port, u64 mac, u64 clear, u8 mode);
791 int mlx4_SET_VLAN_FLTR(struct mlx4_dev *dev, struct mlx4_en_priv *priv);
792 
793 int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset);
794 int mlx4_en_QUERY_PORT(struct mlx4_en_dev *mdev, u8 port);
795 
796 #ifdef CONFIG_MLX4_EN_DCB
797 extern const struct dcbnl_rtnl_ops mlx4_en_dcbnl_ops;
798 extern const struct dcbnl_rtnl_ops mlx4_en_dcbnl_pfc_ops;
799 #endif
800 
801 int mlx4_en_setup_tc(struct net_device *dev, u8 up);
802 
803 #ifdef CONFIG_RFS_ACCEL
804 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv);
805 #endif
806 
807 #define MLX4_EN_NUM_SELF_TEST	5
808 void mlx4_en_ex_selftest(struct net_device *dev, u32 *flags, u64 *buf);
809 void mlx4_en_ptp_overflow_check(struct mlx4_en_dev *mdev);
810 
811 /*
812  * Functions for time stamping
813  */
814 u64 mlx4_en_get_cqe_ts(struct mlx4_cqe *cqe);
815 void mlx4_en_fill_hwtstamps(struct mlx4_en_dev *mdev,
816 			    struct skb_shared_hwtstamps *hwts,
817 			    u64 timestamp);
818 void mlx4_en_init_timestamp(struct mlx4_en_dev *mdev);
819 void mlx4_en_remove_timestamp(struct mlx4_en_dev *mdev);
820 int mlx4_en_timestamp_config(struct net_device *dev,
821 			     int tx_type,
822 			     int rx_filter);
823 
824 /* Globals
825  */
826 extern const struct ethtool_ops mlx4_en_ethtool_ops;
827 
828 
829 
830 /*
831  * printk / logging functions
832  */
833 
834 __printf(3, 4)
835 int en_print(const char *level, const struct mlx4_en_priv *priv,
836 	     const char *format, ...);
837 
838 #define en_dbg(mlevel, priv, format, ...)				\
839 do {									\
840 	if (NETIF_MSG_##mlevel & (priv)->msg_enable)			\
841 		en_print(KERN_DEBUG, priv, format, ##__VA_ARGS__);	\
842 } while (0)
843 #define en_warn(priv, format, ...)					\
844 	en_print(KERN_WARNING, priv, format, ##__VA_ARGS__)
845 #define en_err(priv, format, ...)					\
846 	en_print(KERN_ERR, priv, format, ##__VA_ARGS__)
847 #define en_info(priv, format, ...)					\
848 	en_print(KERN_INFO, priv, format, ##__VA_ARGS__)
849 
850 #define mlx4_err(mdev, format, ...)					\
851 	pr_err(DRV_NAME " %s: " format,					\
852 	       dev_name(&(mdev)->pdev->dev), ##__VA_ARGS__)
853 #define mlx4_info(mdev, format, ...)					\
854 	pr_info(DRV_NAME " %s: " format,				\
855 		dev_name(&(mdev)->pdev->dev), ##__VA_ARGS__)
856 #define mlx4_warn(mdev, format, ...)					\
857 	pr_warn(DRV_NAME " %s: " format,				\
858 		dev_name(&(mdev)->pdev->dev), ##__VA_ARGS__)
859 
860 #endif
861