xref: /linux/drivers/net/ethernet/sfc/net_driver.h (revision 3932b9ca55b0be314a36d3e84faff3e823c081f5)
1 /****************************************************************************
2  * Driver for Solarflare network controllers and boards
3  * Copyright 2005-2006 Fen Systems Ltd.
4  * Copyright 2005-2013 Solarflare Communications Inc.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms of the GNU General Public License version 2 as published
8  * by the Free Software Foundation, incorporated herein by reference.
9  */
10 
11 /* Common definitions for all Efx net driver code */
12 
13 #ifndef EFX_NET_DRIVER_H
14 #define EFX_NET_DRIVER_H
15 
16 #include <linux/netdevice.h>
17 #include <linux/etherdevice.h>
18 #include <linux/ethtool.h>
19 #include <linux/if_vlan.h>
20 #include <linux/timer.h>
21 #include <linux/mdio.h>
22 #include <linux/list.h>
23 #include <linux/pci.h>
24 #include <linux/device.h>
25 #include <linux/highmem.h>
26 #include <linux/workqueue.h>
27 #include <linux/mutex.h>
28 #include <linux/vmalloc.h>
29 #include <linux/i2c.h>
30 #include <linux/mtd/mtd.h>
31 #include <net/busy_poll.h>
32 
33 #include "enum.h"
34 #include "bitfield.h"
35 #include "filter.h"
36 
37 /**************************************************************************
38  *
39  * Build definitions
40  *
41  **************************************************************************/
42 
43 #define EFX_DRIVER_VERSION	"4.0"
44 
45 #ifdef DEBUG
46 #define EFX_BUG_ON_PARANOID(x) BUG_ON(x)
47 #define EFX_WARN_ON_PARANOID(x) WARN_ON(x)
48 #else
49 #define EFX_BUG_ON_PARANOID(x) do {} while (0)
50 #define EFX_WARN_ON_PARANOID(x) do {} while (0)
51 #endif
52 
53 /**************************************************************************
54  *
55  * Efx data structures
56  *
57  **************************************************************************/
58 
59 #define EFX_MAX_CHANNELS 32U
60 #define EFX_MAX_RX_QUEUES EFX_MAX_CHANNELS
61 #define EFX_EXTRA_CHANNEL_IOV	0
62 #define EFX_EXTRA_CHANNEL_PTP	1
63 #define EFX_MAX_EXTRA_CHANNELS	2U
64 
65 /* Checksum generation is a per-queue option in hardware, so each
66  * queue visible to the networking core is backed by two hardware TX
67  * queues. */
68 #define EFX_MAX_TX_TC		2
69 #define EFX_MAX_CORE_TX_QUEUES	(EFX_MAX_TX_TC * EFX_MAX_CHANNELS)
70 #define EFX_TXQ_TYPE_OFFLOAD	1	/* flag */
71 #define EFX_TXQ_TYPE_HIGHPRI	2	/* flag */
72 #define EFX_TXQ_TYPES		4
73 #define EFX_MAX_TX_QUEUES	(EFX_TXQ_TYPES * EFX_MAX_CHANNELS)
74 
75 /* Maximum possible MTU the driver supports */
76 #define EFX_MAX_MTU (9 * 1024)
77 
78 /* Size of an RX scatter buffer.  Small enough to pack 2 into a 4K page,
79  * and should be a multiple of the cache line size.
80  */
81 #define EFX_RX_USR_BUF_SIZE	(2048 - 256)
82 
83 /* If possible, we should ensure cache line alignment at start and end
84  * of every buffer.  Otherwise, we just need to ensure 4-byte
85  * alignment of the network header.
86  */
87 #if NET_IP_ALIGN == 0
88 #define EFX_RX_BUF_ALIGNMENT	L1_CACHE_BYTES
89 #else
90 #define EFX_RX_BUF_ALIGNMENT	4
91 #endif
92 
93 /* Forward declare Precision Time Protocol (PTP) support structure. */
94 struct efx_ptp_data;
95 struct hwtstamp_config;
96 
97 struct efx_self_tests;
98 
99 /**
100  * struct efx_buffer - A general-purpose DMA buffer
101  * @addr: host base address of the buffer
102  * @dma_addr: DMA base address of the buffer
103  * @len: Buffer length, in bytes
104  *
105  * The NIC uses these buffers for its interrupt status registers and
106  * MAC stats dumps.
107  */
108 struct efx_buffer {
109 	void *addr;
110 	dma_addr_t dma_addr;
111 	unsigned int len;
112 };
113 
114 /**
115  * struct efx_special_buffer - DMA buffer entered into buffer table
116  * @buf: Standard &struct efx_buffer
117  * @index: Buffer index within controller;s buffer table
118  * @entries: Number of buffer table entries
119  *
120  * The NIC has a buffer table that maps buffers of size %EFX_BUF_SIZE.
121  * Event and descriptor rings are addressed via one or more buffer
122  * table entries (and so can be physically non-contiguous, although we
123  * currently do not take advantage of that).  On Falcon and Siena we
124  * have to take care of allocating and initialising the entries
125  * ourselves.  On later hardware this is managed by the firmware and
126  * @index and @entries are left as 0.
127  */
128 struct efx_special_buffer {
129 	struct efx_buffer buf;
130 	unsigned int index;
131 	unsigned int entries;
132 };
133 
134 /**
135  * struct efx_tx_buffer - buffer state for a TX descriptor
136  * @skb: When @flags & %EFX_TX_BUF_SKB, the associated socket buffer to be
137  *	freed when descriptor completes
138  * @heap_buf: When @flags & %EFX_TX_BUF_HEAP, the associated heap buffer to be
139  *	freed when descriptor completes.
140  * @option: When @flags & %EFX_TX_BUF_OPTION, a NIC-specific option descriptor.
141  * @dma_addr: DMA address of the fragment.
142  * @flags: Flags for allocation and DMA mapping type
143  * @len: Length of this fragment.
144  *	This field is zero when the queue slot is empty.
145  * @unmap_len: Length of this fragment to unmap
146  * @dma_offset: Offset of @dma_addr from the address of the backing DMA mapping.
147  * Only valid if @unmap_len != 0.
148  */
149 struct efx_tx_buffer {
150 	union {
151 		const struct sk_buff *skb;
152 		void *heap_buf;
153 	};
154 	union {
155 		efx_qword_t option;
156 		dma_addr_t dma_addr;
157 	};
158 	unsigned short flags;
159 	unsigned short len;
160 	unsigned short unmap_len;
161 	unsigned short dma_offset;
162 };
163 #define EFX_TX_BUF_CONT		1	/* not last descriptor of packet */
164 #define EFX_TX_BUF_SKB		2	/* buffer is last part of skb */
165 #define EFX_TX_BUF_HEAP		4	/* buffer was allocated with kmalloc() */
166 #define EFX_TX_BUF_MAP_SINGLE	8	/* buffer was mapped with dma_map_single() */
167 #define EFX_TX_BUF_OPTION	0x10	/* empty buffer for option descriptor */
168 
169 /**
170  * struct efx_tx_queue - An Efx TX queue
171  *
172  * This is a ring buffer of TX fragments.
173  * Since the TX completion path always executes on the same
174  * CPU and the xmit path can operate on different CPUs,
175  * performance is increased by ensuring that the completion
176  * path and the xmit path operate on different cache lines.
177  * This is particularly important if the xmit path is always
178  * executing on one CPU which is different from the completion
179  * path.  There is also a cache line for members which are
180  * read but not written on the fast path.
181  *
182  * @efx: The associated Efx NIC
183  * @queue: DMA queue number
184  * @channel: The associated channel
185  * @core_txq: The networking core TX queue structure
186  * @buffer: The software buffer ring
187  * @tsoh_page: Array of pages of TSO header buffers
188  * @txd: The hardware descriptor ring
189  * @ptr_mask: The size of the ring minus 1.
190  * @piobuf: PIO buffer region for this TX queue (shared with its partner).
191  *	Size of the region is efx_piobuf_size.
192  * @piobuf_offset: Buffer offset to be specified in PIO descriptors
193  * @initialised: Has hardware queue been initialised?
194  * @read_count: Current read pointer.
195  *	This is the number of buffers that have been removed from both rings.
196  * @old_write_count: The value of @write_count when last checked.
197  *	This is here for performance reasons.  The xmit path will
198  *	only get the up-to-date value of @write_count if this
199  *	variable indicates that the queue is empty.  This is to
200  *	avoid cache-line ping-pong between the xmit path and the
201  *	completion path.
202  * @merge_events: Number of TX merged completion events
203  * @insert_count: Current insert pointer
204  *	This is the number of buffers that have been added to the
205  *	software ring.
206  * @write_count: Current write pointer
207  *	This is the number of buffers that have been added to the
208  *	hardware ring.
209  * @old_read_count: The value of read_count when last checked.
210  *	This is here for performance reasons.  The xmit path will
211  *	only get the up-to-date value of read_count if this
212  *	variable indicates that the queue is full.  This is to
213  *	avoid cache-line ping-pong between the xmit path and the
214  *	completion path.
215  * @tso_bursts: Number of times TSO xmit invoked by kernel
216  * @tso_long_headers: Number of packets with headers too long for standard
217  *	blocks
218  * @tso_packets: Number of packets via the TSO xmit path
219  * @pushes: Number of times the TX push feature has been used
220  * @pio_packets: Number of times the TX PIO feature has been used
221  * @empty_read_count: If the completion path has seen the queue as empty
222  *	and the transmission path has not yet checked this, the value of
223  *	@read_count bitwise-added to %EFX_EMPTY_COUNT_VALID; otherwise 0.
224  */
225 struct efx_tx_queue {
226 	/* Members which don't change on the fast path */
227 	struct efx_nic *efx ____cacheline_aligned_in_smp;
228 	unsigned queue;
229 	struct efx_channel *channel;
230 	struct netdev_queue *core_txq;
231 	struct efx_tx_buffer *buffer;
232 	struct efx_buffer *tsoh_page;
233 	struct efx_special_buffer txd;
234 	unsigned int ptr_mask;
235 	void __iomem *piobuf;
236 	unsigned int piobuf_offset;
237 	bool initialised;
238 
239 	/* Members used mainly on the completion path */
240 	unsigned int read_count ____cacheline_aligned_in_smp;
241 	unsigned int old_write_count;
242 	unsigned int merge_events;
243 
244 	/* Members used only on the xmit path */
245 	unsigned int insert_count ____cacheline_aligned_in_smp;
246 	unsigned int write_count;
247 	unsigned int old_read_count;
248 	unsigned int tso_bursts;
249 	unsigned int tso_long_headers;
250 	unsigned int tso_packets;
251 	unsigned int pushes;
252 	unsigned int pio_packets;
253 	/* Statistics to supplement MAC stats */
254 	unsigned long tx_packets;
255 
256 	/* Members shared between paths and sometimes updated */
257 	unsigned int empty_read_count ____cacheline_aligned_in_smp;
258 #define EFX_EMPTY_COUNT_VALID 0x80000000
259 	atomic_t flush_outstanding;
260 };
261 
262 /**
263  * struct efx_rx_buffer - An Efx RX data buffer
264  * @dma_addr: DMA base address of the buffer
265  * @page: The associated page buffer.
266  *	Will be %NULL if the buffer slot is currently free.
267  * @page_offset: If pending: offset in @page of DMA base address.
268  *	If completed: offset in @page of Ethernet header.
269  * @len: If pending: length for DMA descriptor.
270  *	If completed: received length, excluding hash prefix.
271  * @flags: Flags for buffer and packet state.  These are only set on the
272  *	first buffer of a scattered packet.
273  */
274 struct efx_rx_buffer {
275 	dma_addr_t dma_addr;
276 	struct page *page;
277 	u16 page_offset;
278 	u16 len;
279 	u16 flags;
280 };
281 #define EFX_RX_BUF_LAST_IN_PAGE	0x0001
282 #define EFX_RX_PKT_CSUMMED	0x0002
283 #define EFX_RX_PKT_DISCARD	0x0004
284 #define EFX_RX_PKT_TCP		0x0040
285 #define EFX_RX_PKT_PREFIX_LEN	0x0080	/* length is in prefix only */
286 
287 /**
288  * struct efx_rx_page_state - Page-based rx buffer state
289  *
290  * Inserted at the start of every page allocated for receive buffers.
291  * Used to facilitate sharing dma mappings between recycled rx buffers
292  * and those passed up to the kernel.
293  *
294  * @dma_addr: The dma address of this page.
295  */
296 struct efx_rx_page_state {
297 	dma_addr_t dma_addr;
298 
299 	unsigned int __pad[0] ____cacheline_aligned;
300 };
301 
302 /**
303  * struct efx_rx_queue - An Efx RX queue
304  * @efx: The associated Efx NIC
305  * @core_index:  Index of network core RX queue.  Will be >= 0 iff this
306  *	is associated with a real RX queue.
307  * @buffer: The software buffer ring
308  * @rxd: The hardware descriptor ring
309  * @ptr_mask: The size of the ring minus 1.
310  * @refill_enabled: Enable refill whenever fill level is low
311  * @flush_pending: Set when a RX flush is pending. Has the same lifetime as
312  *	@rxq_flush_pending.
313  * @added_count: Number of buffers added to the receive queue.
314  * @notified_count: Number of buffers given to NIC (<= @added_count).
315  * @removed_count: Number of buffers removed from the receive queue.
316  * @scatter_n: Used by NIC specific receive code.
317  * @scatter_len: Used by NIC specific receive code.
318  * @page_ring: The ring to store DMA mapped pages for reuse.
319  * @page_add: Counter to calculate the write pointer for the recycle ring.
320  * @page_remove: Counter to calculate the read pointer for the recycle ring.
321  * @page_recycle_count: The number of pages that have been recycled.
322  * @page_recycle_failed: The number of pages that couldn't be recycled because
323  *      the kernel still held a reference to them.
324  * @page_recycle_full: The number of pages that were released because the
325  *      recycle ring was full.
326  * @page_ptr_mask: The number of pages in the RX recycle ring minus 1.
327  * @max_fill: RX descriptor maximum fill level (<= ring size)
328  * @fast_fill_trigger: RX descriptor fill level that will trigger a fast fill
329  *	(<= @max_fill)
330  * @min_fill: RX descriptor minimum non-zero fill level.
331  *	This records the minimum fill level observed when a ring
332  *	refill was triggered.
333  * @recycle_count: RX buffer recycle counter.
334  * @slow_fill: Timer used to defer efx_nic_generate_fill_event().
335  */
336 struct efx_rx_queue {
337 	struct efx_nic *efx;
338 	int core_index;
339 	struct efx_rx_buffer *buffer;
340 	struct efx_special_buffer rxd;
341 	unsigned int ptr_mask;
342 	bool refill_enabled;
343 	bool flush_pending;
344 
345 	unsigned int added_count;
346 	unsigned int notified_count;
347 	unsigned int removed_count;
348 	unsigned int scatter_n;
349 	unsigned int scatter_len;
350 	struct page **page_ring;
351 	unsigned int page_add;
352 	unsigned int page_remove;
353 	unsigned int page_recycle_count;
354 	unsigned int page_recycle_failed;
355 	unsigned int page_recycle_full;
356 	unsigned int page_ptr_mask;
357 	unsigned int max_fill;
358 	unsigned int fast_fill_trigger;
359 	unsigned int min_fill;
360 	unsigned int min_overfill;
361 	unsigned int recycle_count;
362 	struct timer_list slow_fill;
363 	unsigned int slow_fill_count;
364 	/* Statistics to supplement MAC stats */
365 	unsigned long rx_packets;
366 };
367 
368 enum efx_sync_events_state {
369 	SYNC_EVENTS_DISABLED = 0,
370 	SYNC_EVENTS_QUIESCENT,
371 	SYNC_EVENTS_REQUESTED,
372 	SYNC_EVENTS_VALID,
373 };
374 
375 /**
376  * struct efx_channel - An Efx channel
377  *
378  * A channel comprises an event queue, at least one TX queue, at least
379  * one RX queue, and an associated tasklet for processing the event
380  * queue.
381  *
382  * @efx: Associated Efx NIC
383  * @channel: Channel instance number
384  * @type: Channel type definition
385  * @eventq_init: Event queue initialised flag
386  * @enabled: Channel enabled indicator
387  * @irq: IRQ number (MSI and MSI-X only)
388  * @irq_moderation: IRQ moderation value (in hardware ticks)
389  * @napi_dev: Net device used with NAPI
390  * @napi_str: NAPI control structure
391  * @state: state for NAPI vs busy polling
392  * @state_lock: lock protecting @state
393  * @eventq: Event queue buffer
394  * @eventq_mask: Event queue pointer mask
395  * @eventq_read_ptr: Event queue read pointer
396  * @event_test_cpu: Last CPU to handle interrupt or test event for this channel
397  * @irq_count: Number of IRQs since last adaptive moderation decision
398  * @irq_mod_score: IRQ moderation score
399  * @n_rx_tobe_disc: Count of RX_TOBE_DISC errors
400  * @n_rx_ip_hdr_chksum_err: Count of RX IP header checksum errors
401  * @n_rx_tcp_udp_chksum_err: Count of RX TCP and UDP checksum errors
402  * @n_rx_mcast_mismatch: Count of unmatched multicast frames
403  * @n_rx_frm_trunc: Count of RX_FRM_TRUNC errors
404  * @n_rx_overlength: Count of RX_OVERLENGTH errors
405  * @n_skbuff_leaks: Count of skbuffs leaked due to RX overrun
406  * @n_rx_nodesc_trunc: Number of RX packets truncated and then dropped due to
407  *	lack of descriptors
408  * @n_rx_merge_events: Number of RX merged completion events
409  * @n_rx_merge_packets: Number of RX packets completed by merged events
410  * @rx_pkt_n_frags: Number of fragments in next packet to be delivered by
411  *	__efx_rx_packet(), or zero if there is none
412  * @rx_pkt_index: Ring index of first buffer for next packet to be delivered
413  *	by __efx_rx_packet(), if @rx_pkt_n_frags != 0
414  * @rx_queue: RX queue for this channel
415  * @tx_queue: TX queues for this channel
416  * @sync_events_state: Current state of sync events on this channel
417  * @sync_timestamp_major: Major part of the last ptp sync event
418  * @sync_timestamp_minor: Minor part of the last ptp sync event
419  */
420 struct efx_channel {
421 	struct efx_nic *efx;
422 	int channel;
423 	const struct efx_channel_type *type;
424 	bool eventq_init;
425 	bool enabled;
426 	int irq;
427 	unsigned int irq_moderation;
428 	struct net_device *napi_dev;
429 	struct napi_struct napi_str;
430 #ifdef CONFIG_NET_RX_BUSY_POLL
431 	unsigned int state;
432 	spinlock_t state_lock;
433 #define EFX_CHANNEL_STATE_IDLE		0
434 #define EFX_CHANNEL_STATE_NAPI		(1 << 0)  /* NAPI owns this channel */
435 #define EFX_CHANNEL_STATE_POLL		(1 << 1)  /* poll owns this channel */
436 #define EFX_CHANNEL_STATE_DISABLED	(1 << 2)  /* channel is disabled */
437 #define EFX_CHANNEL_STATE_NAPI_YIELD	(1 << 3)  /* NAPI yielded this channel */
438 #define EFX_CHANNEL_STATE_POLL_YIELD	(1 << 4)  /* poll yielded this channel */
439 #define EFX_CHANNEL_OWNED \
440 	(EFX_CHANNEL_STATE_NAPI | EFX_CHANNEL_STATE_POLL)
441 #define EFX_CHANNEL_LOCKED \
442 	(EFX_CHANNEL_OWNED | EFX_CHANNEL_STATE_DISABLED)
443 #define EFX_CHANNEL_USER_PEND \
444 	(EFX_CHANNEL_STATE_POLL | EFX_CHANNEL_STATE_POLL_YIELD)
445 #endif /* CONFIG_NET_RX_BUSY_POLL */
446 	struct efx_special_buffer eventq;
447 	unsigned int eventq_mask;
448 	unsigned int eventq_read_ptr;
449 	int event_test_cpu;
450 
451 	unsigned int irq_count;
452 	unsigned int irq_mod_score;
453 #ifdef CONFIG_RFS_ACCEL
454 	unsigned int rfs_filters_added;
455 #endif
456 
457 	unsigned n_rx_tobe_disc;
458 	unsigned n_rx_ip_hdr_chksum_err;
459 	unsigned n_rx_tcp_udp_chksum_err;
460 	unsigned n_rx_mcast_mismatch;
461 	unsigned n_rx_frm_trunc;
462 	unsigned n_rx_overlength;
463 	unsigned n_skbuff_leaks;
464 	unsigned int n_rx_nodesc_trunc;
465 	unsigned int n_rx_merge_events;
466 	unsigned int n_rx_merge_packets;
467 
468 	unsigned int rx_pkt_n_frags;
469 	unsigned int rx_pkt_index;
470 
471 	struct efx_rx_queue rx_queue;
472 	struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
473 
474 	enum efx_sync_events_state sync_events_state;
475 	u32 sync_timestamp_major;
476 	u32 sync_timestamp_minor;
477 };
478 
479 #ifdef CONFIG_NET_RX_BUSY_POLL
480 static inline void efx_channel_init_lock(struct efx_channel *channel)
481 {
482 	spin_lock_init(&channel->state_lock);
483 }
484 
485 /* Called from the device poll routine to get ownership of a channel. */
486 static inline bool efx_channel_lock_napi(struct efx_channel *channel)
487 {
488 	bool rc = true;
489 
490 	spin_lock_bh(&channel->state_lock);
491 	if (channel->state & EFX_CHANNEL_LOCKED) {
492 		WARN_ON(channel->state & EFX_CHANNEL_STATE_NAPI);
493 		channel->state |= EFX_CHANNEL_STATE_NAPI_YIELD;
494 		rc = false;
495 	} else {
496 		/* we don't care if someone yielded */
497 		channel->state = EFX_CHANNEL_STATE_NAPI;
498 	}
499 	spin_unlock_bh(&channel->state_lock);
500 	return rc;
501 }
502 
503 static inline void efx_channel_unlock_napi(struct efx_channel *channel)
504 {
505 	spin_lock_bh(&channel->state_lock);
506 	WARN_ON(channel->state &
507 		(EFX_CHANNEL_STATE_POLL | EFX_CHANNEL_STATE_NAPI_YIELD));
508 
509 	channel->state &= EFX_CHANNEL_STATE_DISABLED;
510 	spin_unlock_bh(&channel->state_lock);
511 }
512 
513 /* Called from efx_busy_poll(). */
514 static inline bool efx_channel_lock_poll(struct efx_channel *channel)
515 {
516 	bool rc = true;
517 
518 	spin_lock_bh(&channel->state_lock);
519 	if ((channel->state & EFX_CHANNEL_LOCKED)) {
520 		channel->state |= EFX_CHANNEL_STATE_POLL_YIELD;
521 		rc = false;
522 	} else {
523 		/* preserve yield marks */
524 		channel->state |= EFX_CHANNEL_STATE_POLL;
525 	}
526 	spin_unlock_bh(&channel->state_lock);
527 	return rc;
528 }
529 
530 /* Returns true if NAPI tried to get the channel while it was locked. */
531 static inline void efx_channel_unlock_poll(struct efx_channel *channel)
532 {
533 	spin_lock_bh(&channel->state_lock);
534 	WARN_ON(channel->state & EFX_CHANNEL_STATE_NAPI);
535 
536 	/* will reset state to idle, unless channel is disabled */
537 	channel->state &= EFX_CHANNEL_STATE_DISABLED;
538 	spin_unlock_bh(&channel->state_lock);
539 }
540 
541 /* True if a socket is polling, even if it did not get the lock. */
542 static inline bool efx_channel_busy_polling(struct efx_channel *channel)
543 {
544 	WARN_ON(!(channel->state & EFX_CHANNEL_OWNED));
545 	return channel->state & EFX_CHANNEL_USER_PEND;
546 }
547 
548 static inline void efx_channel_enable(struct efx_channel *channel)
549 {
550 	spin_lock_bh(&channel->state_lock);
551 	channel->state = EFX_CHANNEL_STATE_IDLE;
552 	spin_unlock_bh(&channel->state_lock);
553 }
554 
555 /* False if the channel is currently owned. */
556 static inline bool efx_channel_disable(struct efx_channel *channel)
557 {
558 	bool rc = true;
559 
560 	spin_lock_bh(&channel->state_lock);
561 	if (channel->state & EFX_CHANNEL_OWNED)
562 		rc = false;
563 	channel->state |= EFX_CHANNEL_STATE_DISABLED;
564 	spin_unlock_bh(&channel->state_lock);
565 
566 	return rc;
567 }
568 
569 #else /* CONFIG_NET_RX_BUSY_POLL */
570 
571 static inline void efx_channel_init_lock(struct efx_channel *channel)
572 {
573 }
574 
575 static inline bool efx_channel_lock_napi(struct efx_channel *channel)
576 {
577 	return true;
578 }
579 
580 static inline void efx_channel_unlock_napi(struct efx_channel *channel)
581 {
582 }
583 
584 static inline bool efx_channel_lock_poll(struct efx_channel *channel)
585 {
586 	return false;
587 }
588 
589 static inline void efx_channel_unlock_poll(struct efx_channel *channel)
590 {
591 }
592 
593 static inline bool efx_channel_busy_polling(struct efx_channel *channel)
594 {
595 	return false;
596 }
597 
598 static inline void efx_channel_enable(struct efx_channel *channel)
599 {
600 }
601 
602 static inline bool efx_channel_disable(struct efx_channel *channel)
603 {
604 	return true;
605 }
606 #endif /* CONFIG_NET_RX_BUSY_POLL */
607 
608 /**
609  * struct efx_msi_context - Context for each MSI
610  * @efx: The associated NIC
611  * @index: Index of the channel/IRQ
612  * @name: Name of the channel/IRQ
613  *
614  * Unlike &struct efx_channel, this is never reallocated and is always
615  * safe for the IRQ handler to access.
616  */
617 struct efx_msi_context {
618 	struct efx_nic *efx;
619 	unsigned int index;
620 	char name[IFNAMSIZ + 6];
621 };
622 
623 /**
624  * struct efx_channel_type - distinguishes traffic and extra channels
625  * @handle_no_channel: Handle failure to allocate an extra channel
626  * @pre_probe: Set up extra state prior to initialisation
627  * @post_remove: Tear down extra state after finalisation, if allocated.
628  *	May be called on channels that have not been probed.
629  * @get_name: Generate the channel's name (used for its IRQ handler)
630  * @copy: Copy the channel state prior to reallocation.  May be %NULL if
631  *	reallocation is not supported.
632  * @receive_skb: Handle an skb ready to be passed to netif_receive_skb()
633  * @keep_eventq: Flag for whether event queue should be kept initialised
634  *	while the device is stopped
635  */
636 struct efx_channel_type {
637 	void (*handle_no_channel)(struct efx_nic *);
638 	int (*pre_probe)(struct efx_channel *);
639 	void (*post_remove)(struct efx_channel *);
640 	void (*get_name)(struct efx_channel *, char *buf, size_t len);
641 	struct efx_channel *(*copy)(const struct efx_channel *);
642 	bool (*receive_skb)(struct efx_channel *, struct sk_buff *);
643 	bool keep_eventq;
644 };
645 
646 enum efx_led_mode {
647 	EFX_LED_OFF	= 0,
648 	EFX_LED_ON	= 1,
649 	EFX_LED_DEFAULT	= 2
650 };
651 
652 #define STRING_TABLE_LOOKUP(val, member) \
653 	((val) < member ## _max) ? member ## _names[val] : "(invalid)"
654 
655 extern const char *const efx_loopback_mode_names[];
656 extern const unsigned int efx_loopback_mode_max;
657 #define LOOPBACK_MODE(efx) \
658 	STRING_TABLE_LOOKUP((efx)->loopback_mode, efx_loopback_mode)
659 
660 extern const char *const efx_reset_type_names[];
661 extern const unsigned int efx_reset_type_max;
662 #define RESET_TYPE(type) \
663 	STRING_TABLE_LOOKUP(type, efx_reset_type)
664 
665 enum efx_int_mode {
666 	/* Be careful if altering to correct macro below */
667 	EFX_INT_MODE_MSIX = 0,
668 	EFX_INT_MODE_MSI = 1,
669 	EFX_INT_MODE_LEGACY = 2,
670 	EFX_INT_MODE_MAX	/* Insert any new items before this */
671 };
672 #define EFX_INT_MODE_USE_MSI(x) (((x)->interrupt_mode) <= EFX_INT_MODE_MSI)
673 
674 enum nic_state {
675 	STATE_UNINIT = 0,	/* device being probed/removed or is frozen */
676 	STATE_READY = 1,	/* hardware ready and netdev registered */
677 	STATE_DISABLED = 2,	/* device disabled due to hardware errors */
678 	STATE_RECOVERY = 3,	/* device recovering from PCI error */
679 };
680 
681 /* Forward declaration */
682 struct efx_nic;
683 
684 /* Pseudo bit-mask flow control field */
685 #define EFX_FC_RX	FLOW_CTRL_RX
686 #define EFX_FC_TX	FLOW_CTRL_TX
687 #define EFX_FC_AUTO	4
688 
689 /**
690  * struct efx_link_state - Current state of the link
691  * @up: Link is up
692  * @fd: Link is full-duplex
693  * @fc: Actual flow control flags
694  * @speed: Link speed (Mbps)
695  */
696 struct efx_link_state {
697 	bool up;
698 	bool fd;
699 	u8 fc;
700 	unsigned int speed;
701 };
702 
703 static inline bool efx_link_state_equal(const struct efx_link_state *left,
704 					const struct efx_link_state *right)
705 {
706 	return left->up == right->up && left->fd == right->fd &&
707 		left->fc == right->fc && left->speed == right->speed;
708 }
709 
710 /**
711  * struct efx_phy_operations - Efx PHY operations table
712  * @probe: Probe PHY and initialise efx->mdio.mode_support, efx->mdio.mmds,
713  *	efx->loopback_modes.
714  * @init: Initialise PHY
715  * @fini: Shut down PHY
716  * @reconfigure: Reconfigure PHY (e.g. for new link parameters)
717  * @poll: Update @link_state and report whether it changed.
718  *	Serialised by the mac_lock.
719  * @get_settings: Get ethtool settings. Serialised by the mac_lock.
720  * @set_settings: Set ethtool settings. Serialised by the mac_lock.
721  * @set_npage_adv: Set abilities advertised in (Extended) Next Page
722  *	(only needed where AN bit is set in mmds)
723  * @test_alive: Test that PHY is 'alive' (online)
724  * @test_name: Get the name of a PHY-specific test/result
725  * @run_tests: Run tests and record results as appropriate (offline).
726  *	Flags are the ethtool tests flags.
727  */
728 struct efx_phy_operations {
729 	int (*probe) (struct efx_nic *efx);
730 	int (*init) (struct efx_nic *efx);
731 	void (*fini) (struct efx_nic *efx);
732 	void (*remove) (struct efx_nic *efx);
733 	int (*reconfigure) (struct efx_nic *efx);
734 	bool (*poll) (struct efx_nic *efx);
735 	void (*get_settings) (struct efx_nic *efx,
736 			      struct ethtool_cmd *ecmd);
737 	int (*set_settings) (struct efx_nic *efx,
738 			     struct ethtool_cmd *ecmd);
739 	void (*set_npage_adv) (struct efx_nic *efx, u32);
740 	int (*test_alive) (struct efx_nic *efx);
741 	const char *(*test_name) (struct efx_nic *efx, unsigned int index);
742 	int (*run_tests) (struct efx_nic *efx, int *results, unsigned flags);
743 	int (*get_module_eeprom) (struct efx_nic *efx,
744 			       struct ethtool_eeprom *ee,
745 			       u8 *data);
746 	int (*get_module_info) (struct efx_nic *efx,
747 				struct ethtool_modinfo *modinfo);
748 };
749 
750 /**
751  * enum efx_phy_mode - PHY operating mode flags
752  * @PHY_MODE_NORMAL: on and should pass traffic
753  * @PHY_MODE_TX_DISABLED: on with TX disabled
754  * @PHY_MODE_LOW_POWER: set to low power through MDIO
755  * @PHY_MODE_OFF: switched off through external control
756  * @PHY_MODE_SPECIAL: on but will not pass traffic
757  */
758 enum efx_phy_mode {
759 	PHY_MODE_NORMAL		= 0,
760 	PHY_MODE_TX_DISABLED	= 1,
761 	PHY_MODE_LOW_POWER	= 2,
762 	PHY_MODE_OFF		= 4,
763 	PHY_MODE_SPECIAL	= 8,
764 };
765 
766 static inline bool efx_phy_mode_disabled(enum efx_phy_mode mode)
767 {
768 	return !!(mode & ~PHY_MODE_TX_DISABLED);
769 }
770 
771 /**
772  * struct efx_hw_stat_desc - Description of a hardware statistic
773  * @name: Name of the statistic as visible through ethtool, or %NULL if
774  *	it should not be exposed
775  * @dma_width: Width in bits (0 for non-DMA statistics)
776  * @offset: Offset within stats (ignored for non-DMA statistics)
777  */
778 struct efx_hw_stat_desc {
779 	const char *name;
780 	u16 dma_width;
781 	u16 offset;
782 };
783 
784 /* Number of bits used in a multicast filter hash address */
785 #define EFX_MCAST_HASH_BITS 8
786 
787 /* Number of (single-bit) entries in a multicast filter hash */
788 #define EFX_MCAST_HASH_ENTRIES (1 << EFX_MCAST_HASH_BITS)
789 
790 /* An Efx multicast filter hash */
791 union efx_multicast_hash {
792 	u8 byte[EFX_MCAST_HASH_ENTRIES / 8];
793 	efx_oword_t oword[EFX_MCAST_HASH_ENTRIES / sizeof(efx_oword_t) / 8];
794 };
795 
796 struct efx_vf;
797 struct vfdi_status;
798 
799 /**
800  * struct efx_nic - an Efx NIC
801  * @name: Device name (net device name or bus id before net device registered)
802  * @pci_dev: The PCI device
803  * @node: List node for maintaning primary/secondary function lists
804  * @primary: &struct efx_nic instance for the primary function of this
805  *	controller.  May be the same structure, and may be %NULL if no
806  *	primary function is bound.  Serialised by rtnl_lock.
807  * @secondary_list: List of &struct efx_nic instances for the secondary PCI
808  *	functions of the controller, if this is for the primary function.
809  *	Serialised by rtnl_lock.
810  * @type: Controller type attributes
811  * @legacy_irq: IRQ number
812  * @workqueue: Workqueue for port reconfigures and the HW monitor.
813  *	Work items do not hold and must not acquire RTNL.
814  * @workqueue_name: Name of workqueue
815  * @reset_work: Scheduled reset workitem
816  * @membase_phys: Memory BAR value as physical address
817  * @membase: Memory BAR value
818  * @interrupt_mode: Interrupt mode
819  * @timer_quantum_ns: Interrupt timer quantum, in nanoseconds
820  * @irq_rx_adaptive: Adaptive IRQ moderation enabled for RX event queues
821  * @irq_rx_moderation: IRQ moderation time for RX event queues
822  * @msg_enable: Log message enable flags
823  * @state: Device state number (%STATE_*). Serialised by the rtnl_lock.
824  * @reset_pending: Bitmask for pending resets
825  * @tx_queue: TX DMA queues
826  * @rx_queue: RX DMA queues
827  * @channel: Channels
828  * @msi_context: Context for each MSI
829  * @extra_channel_types: Types of extra (non-traffic) channels that
830  *	should be allocated for this NIC
831  * @rxq_entries: Size of receive queues requested by user.
832  * @txq_entries: Size of transmit queues requested by user.
833  * @txq_stop_thresh: TX queue fill level at or above which we stop it.
834  * @txq_wake_thresh: TX queue fill level at or below which we wake it.
835  * @tx_dc_base: Base qword address in SRAM of TX queue descriptor caches
836  * @rx_dc_base: Base qword address in SRAM of RX queue descriptor caches
837  * @sram_lim_qw: Qword address limit of SRAM
838  * @next_buffer_table: First available buffer table id
839  * @n_channels: Number of channels in use
840  * @n_rx_channels: Number of channels used for RX (= number of RX queues)
841  * @n_tx_channels: Number of channels used for TX
842  * @rx_ip_align: RX DMA address offset to have IP header aligned in
843  *	in accordance with NET_IP_ALIGN
844  * @rx_dma_len: Current maximum RX DMA length
845  * @rx_buffer_order: Order (log2) of number of pages for each RX buffer
846  * @rx_buffer_truesize: Amortised allocation size of an RX buffer,
847  *	for use in sk_buff::truesize
848  * @rx_prefix_size: Size of RX prefix before packet data
849  * @rx_packet_hash_offset: Offset of RX flow hash from start of packet data
850  *	(valid only if @rx_prefix_size != 0; always negative)
851  * @rx_packet_len_offset: Offset of RX packet length from start of packet data
852  *	(valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
853  * @rx_packet_ts_offset: Offset of timestamp from start of packet data
854  *	(valid only if channel->sync_timestamps_enabled; always negative)
855  * @rx_hash_key: Toeplitz hash key for RSS
856  * @rx_indir_table: Indirection table for RSS
857  * @rx_scatter: Scatter mode enabled for receives
858  * @int_error_count: Number of internal errors seen recently
859  * @int_error_expire: Time at which error count will be expired
860  * @irq_soft_enabled: Are IRQs soft-enabled? If not, IRQ handler will
861  *	acknowledge but do nothing else.
862  * @irq_status: Interrupt status buffer
863  * @irq_zero_count: Number of legacy IRQs seen with queue flags == 0
864  * @irq_level: IRQ level/index for IRQs not triggered by an event queue
865  * @selftest_work: Work item for asynchronous self-test
866  * @mtd_list: List of MTDs attached to the NIC
867  * @nic_data: Hardware dependent state
868  * @mcdi: Management-Controller-to-Driver Interface state
869  * @mac_lock: MAC access lock. Protects @port_enabled, @phy_mode,
870  *	efx_monitor() and efx_reconfigure_port()
871  * @port_enabled: Port enabled indicator.
872  *	Serialises efx_stop_all(), efx_start_all(), efx_monitor() and
873  *	efx_mac_work() with kernel interfaces. Safe to read under any
874  *	one of the rtnl_lock, mac_lock, or netif_tx_lock, but all three must
875  *	be held to modify it.
876  * @port_initialized: Port initialized?
877  * @net_dev: Operating system network device. Consider holding the rtnl lock
878  * @stats_buffer: DMA buffer for statistics
879  * @phy_type: PHY type
880  * @phy_op: PHY interface
881  * @phy_data: PHY private data (including PHY-specific stats)
882  * @mdio: PHY MDIO interface
883  * @mdio_bus: PHY MDIO bus ID (only used by Siena)
884  * @phy_mode: PHY operating mode. Serialised by @mac_lock.
885  * @link_advertising: Autonegotiation advertising flags
886  * @link_state: Current state of the link
887  * @n_link_state_changes: Number of times the link has changed state
888  * @unicast_filter: Flag for Falcon-arch simple unicast filter.
889  *	Protected by @mac_lock.
890  * @multicast_hash: Multicast hash table for Falcon-arch.
891  *	Protected by @mac_lock.
892  * @wanted_fc: Wanted flow control flags
893  * @fc_disable: When non-zero flow control is disabled. Typically used to
894  *	ensure that network back pressure doesn't delay dma queue flushes.
895  *	Serialised by the rtnl lock.
896  * @mac_work: Work item for changing MAC promiscuity and multicast hash
897  * @loopback_mode: Loopback status
898  * @loopback_modes: Supported loopback mode bitmask
899  * @loopback_selftest: Offline self-test private state
900  * @filter_lock: Filter table lock
901  * @filter_state: Architecture-dependent filter table state
902  * @rps_flow_id: Flow IDs of filters allocated for accelerated RFS,
903  *	indexed by filter ID
904  * @rps_expire_index: Next index to check for expiry in @rps_flow_id
905  * @active_queues: Count of RX and TX queues that haven't been flushed and drained.
906  * @rxq_flush_pending: Count of number of receive queues that need to be flushed.
907  *	Decremented when the efx_flush_rx_queue() is called.
908  * @rxq_flush_outstanding: Count of number of RX flushes started but not yet
909  *	completed (either success or failure). Not used when MCDI is used to
910  *	flush receive queues.
911  * @flush_wq: wait queue used by efx_nic_flush_queues() to wait for flush completions.
912  * @vf: Array of &struct efx_vf objects.
913  * @vf_count: Number of VFs intended to be enabled.
914  * @vf_init_count: Number of VFs that have been fully initialised.
915  * @vi_scale: log2 number of vnics per VF.
916  * @vf_buftbl_base: The zeroth buffer table index used to back VF queues.
917  * @vfdi_status: Common VFDI status page to be dmad to VF address space.
918  * @local_addr_list: List of local addresses. Protected by %local_lock.
919  * @local_page_list: List of DMA addressable pages used to broadcast
920  *	%local_addr_list. Protected by %local_lock.
921  * @local_lock: Mutex protecting %local_addr_list and %local_page_list.
922  * @peer_work: Work item to broadcast peer addresses to VMs.
923  * @ptp_data: PTP state data
924  * @vpd_sn: Serial number read from VPD
925  * @monitor_work: Hardware monitor workitem
926  * @biu_lock: BIU (bus interface unit) lock
927  * @last_irq_cpu: Last CPU to handle a possible test interrupt.  This
928  *	field is used by efx_test_interrupts() to verify that an
929  *	interrupt has occurred.
930  * @stats_lock: Statistics update lock. Must be held when calling
931  *	efx_nic_type::{update,start,stop}_stats.
932  * @n_rx_noskb_drops: Count of RX packets dropped due to failure to allocate an skb
933  *
934  * This is stored in the private area of the &struct net_device.
935  */
936 struct efx_nic {
937 	/* The following fields should be written very rarely */
938 
939 	char name[IFNAMSIZ];
940 	struct list_head node;
941 	struct efx_nic *primary;
942 	struct list_head secondary_list;
943 	struct pci_dev *pci_dev;
944 	unsigned int port_num;
945 	const struct efx_nic_type *type;
946 	int legacy_irq;
947 	bool eeh_disabled_legacy_irq;
948 	struct workqueue_struct *workqueue;
949 	char workqueue_name[16];
950 	struct work_struct reset_work;
951 	resource_size_t membase_phys;
952 	void __iomem *membase;
953 
954 	enum efx_int_mode interrupt_mode;
955 	unsigned int timer_quantum_ns;
956 	bool irq_rx_adaptive;
957 	unsigned int irq_rx_moderation;
958 	u32 msg_enable;
959 
960 	enum nic_state state;
961 	unsigned long reset_pending;
962 
963 	struct efx_channel *channel[EFX_MAX_CHANNELS];
964 	struct efx_msi_context msi_context[EFX_MAX_CHANNELS];
965 	const struct efx_channel_type *
966 	extra_channel_type[EFX_MAX_EXTRA_CHANNELS];
967 
968 	unsigned rxq_entries;
969 	unsigned txq_entries;
970 	unsigned int txq_stop_thresh;
971 	unsigned int txq_wake_thresh;
972 
973 	unsigned tx_dc_base;
974 	unsigned rx_dc_base;
975 	unsigned sram_lim_qw;
976 	unsigned next_buffer_table;
977 
978 	unsigned int max_channels;
979 	unsigned n_channels;
980 	unsigned n_rx_channels;
981 	unsigned rss_spread;
982 	unsigned tx_channel_offset;
983 	unsigned n_tx_channels;
984 	unsigned int rx_ip_align;
985 	unsigned int rx_dma_len;
986 	unsigned int rx_buffer_order;
987 	unsigned int rx_buffer_truesize;
988 	unsigned int rx_page_buf_step;
989 	unsigned int rx_bufs_per_page;
990 	unsigned int rx_pages_per_batch;
991 	unsigned int rx_prefix_size;
992 	int rx_packet_hash_offset;
993 	int rx_packet_len_offset;
994 	int rx_packet_ts_offset;
995 	u8 rx_hash_key[40];
996 	u32 rx_indir_table[128];
997 	bool rx_scatter;
998 
999 	unsigned int_error_count;
1000 	unsigned long int_error_expire;
1001 
1002 	bool irq_soft_enabled;
1003 	struct efx_buffer irq_status;
1004 	unsigned irq_zero_count;
1005 	unsigned irq_level;
1006 	struct delayed_work selftest_work;
1007 
1008 #ifdef CONFIG_SFC_MTD
1009 	struct list_head mtd_list;
1010 #endif
1011 
1012 	void *nic_data;
1013 	struct efx_mcdi_data *mcdi;
1014 
1015 	struct mutex mac_lock;
1016 	struct work_struct mac_work;
1017 	bool port_enabled;
1018 
1019 	bool mc_bist_for_other_fn;
1020 	bool port_initialized;
1021 	struct net_device *net_dev;
1022 
1023 	struct efx_buffer stats_buffer;
1024 	u64 rx_nodesc_drops_total;
1025 	u64 rx_nodesc_drops_while_down;
1026 	bool rx_nodesc_drops_prev_state;
1027 
1028 	unsigned int phy_type;
1029 	const struct efx_phy_operations *phy_op;
1030 	void *phy_data;
1031 	struct mdio_if_info mdio;
1032 	unsigned int mdio_bus;
1033 	enum efx_phy_mode phy_mode;
1034 
1035 	u32 link_advertising;
1036 	struct efx_link_state link_state;
1037 	unsigned int n_link_state_changes;
1038 
1039 	bool unicast_filter;
1040 	union efx_multicast_hash multicast_hash;
1041 	u8 wanted_fc;
1042 	unsigned fc_disable;
1043 
1044 	atomic_t rx_reset;
1045 	enum efx_loopback_mode loopback_mode;
1046 	u64 loopback_modes;
1047 
1048 	void *loopback_selftest;
1049 
1050 	spinlock_t filter_lock;
1051 	void *filter_state;
1052 #ifdef CONFIG_RFS_ACCEL
1053 	u32 *rps_flow_id;
1054 	unsigned int rps_expire_index;
1055 #endif
1056 
1057 	atomic_t active_queues;
1058 	atomic_t rxq_flush_pending;
1059 	atomic_t rxq_flush_outstanding;
1060 	wait_queue_head_t flush_wq;
1061 
1062 #ifdef CONFIG_SFC_SRIOV
1063 	struct efx_channel *vfdi_channel;
1064 	struct efx_vf *vf;
1065 	unsigned vf_count;
1066 	unsigned vf_init_count;
1067 	unsigned vi_scale;
1068 	unsigned vf_buftbl_base;
1069 	struct efx_buffer vfdi_status;
1070 	struct list_head local_addr_list;
1071 	struct list_head local_page_list;
1072 	struct mutex local_lock;
1073 	struct work_struct peer_work;
1074 #endif
1075 
1076 	struct efx_ptp_data *ptp_data;
1077 
1078 	char *vpd_sn;
1079 
1080 	/* The following fields may be written more often */
1081 
1082 	struct delayed_work monitor_work ____cacheline_aligned_in_smp;
1083 	spinlock_t biu_lock;
1084 	int last_irq_cpu;
1085 	spinlock_t stats_lock;
1086 	atomic_t n_rx_noskb_drops;
1087 };
1088 
1089 static inline int efx_dev_registered(struct efx_nic *efx)
1090 {
1091 	return efx->net_dev->reg_state == NETREG_REGISTERED;
1092 }
1093 
1094 static inline unsigned int efx_port_num(struct efx_nic *efx)
1095 {
1096 	return efx->port_num;
1097 }
1098 
1099 struct efx_mtd_partition {
1100 	struct list_head node;
1101 	struct mtd_info mtd;
1102 	const char *dev_type_name;
1103 	const char *type_name;
1104 	char name[IFNAMSIZ + 20];
1105 };
1106 
1107 /**
1108  * struct efx_nic_type - Efx device type definition
1109  * @mem_map_size: Get memory BAR mapped size
1110  * @probe: Probe the controller
1111  * @remove: Free resources allocated by probe()
1112  * @init: Initialise the controller
1113  * @dimension_resources: Dimension controller resources (buffer table,
1114  *	and VIs once the available interrupt resources are clear)
1115  * @fini: Shut down the controller
1116  * @monitor: Periodic function for polling link state and hardware monitor
1117  * @map_reset_reason: Map ethtool reset reason to a reset method
1118  * @map_reset_flags: Map ethtool reset flags to a reset method, if possible
1119  * @reset: Reset the controller hardware and possibly the PHY.  This will
1120  *	be called while the controller is uninitialised.
1121  * @probe_port: Probe the MAC and PHY
1122  * @remove_port: Free resources allocated by probe_port()
1123  * @handle_global_event: Handle a "global" event (may be %NULL)
1124  * @fini_dmaq: Flush and finalise DMA queues (RX and TX queues)
1125  * @prepare_flush: Prepare the hardware for flushing the DMA queues
1126  *	(for Falcon architecture)
1127  * @finish_flush: Clean up after flushing the DMA queues (for Falcon
1128  *	architecture)
1129  * @prepare_flr: Prepare for an FLR
1130  * @finish_flr: Clean up after an FLR
1131  * @describe_stats: Describe statistics for ethtool
1132  * @update_stats: Update statistics not provided by event handling.
1133  *	Either argument may be %NULL.
1134  * @start_stats: Start the regular fetching of statistics
1135  * @pull_stats: Pull stats from the NIC and wait until they arrive.
1136  * @stop_stats: Stop the regular fetching of statistics
1137  * @set_id_led: Set state of identifying LED or revert to automatic function
1138  * @push_irq_moderation: Apply interrupt moderation value
1139  * @reconfigure_port: Push loopback/power/txdis changes to the MAC and PHY
1140  * @prepare_enable_fc_tx: Prepare MAC to enable pause frame TX (may be %NULL)
1141  * @reconfigure_mac: Push MAC address, MTU, flow control and filter settings
1142  *	to the hardware.  Serialised by the mac_lock.
1143  * @check_mac_fault: Check MAC fault state. True if fault present.
1144  * @get_wol: Get WoL configuration from driver state
1145  * @set_wol: Push WoL configuration to the NIC
1146  * @resume_wol: Synchronise WoL state between driver and MC (e.g. after resume)
1147  * @test_chip: Test registers.  May use efx_farch_test_registers(), and is
1148  *	expected to reset the NIC.
1149  * @test_nvram: Test validity of NVRAM contents
1150  * @mcdi_request: Send an MCDI request with the given header and SDU.
1151  *	The SDU length may be any value from 0 up to the protocol-
1152  *	defined maximum, but its buffer will be padded to a multiple
1153  *	of 4 bytes.
1154  * @mcdi_poll_response: Test whether an MCDI response is available.
1155  * @mcdi_read_response: Read the MCDI response PDU.  The offset will
1156  *	be a multiple of 4.  The length may not be, but the buffer
1157  *	will be padded so it is safe to round up.
1158  * @mcdi_poll_reboot: Test whether the MCDI has rebooted.  If so,
1159  *	return an appropriate error code for aborting any current
1160  *	request; otherwise return 0.
1161  * @irq_enable_master: Enable IRQs on the NIC.  Each event queue must
1162  *	be separately enabled after this.
1163  * @irq_test_generate: Generate a test IRQ
1164  * @irq_disable_non_ev: Disable non-event IRQs on the NIC.  Each event
1165  *	queue must be separately disabled before this.
1166  * @irq_handle_msi: Handle MSI for a channel.  The @dev_id argument is
1167  *	a pointer to the &struct efx_msi_context for the channel.
1168  * @irq_handle_legacy: Handle legacy interrupt.  The @dev_id argument
1169  *	is a pointer to the &struct efx_nic.
1170  * @tx_probe: Allocate resources for TX queue
1171  * @tx_init: Initialise TX queue on the NIC
1172  * @tx_remove: Free resources for TX queue
1173  * @tx_write: Write TX descriptors and doorbell
1174  * @rx_push_rss_config: Write RSS hash key and indirection table to the NIC
1175  * @rx_probe: Allocate resources for RX queue
1176  * @rx_init: Initialise RX queue on the NIC
1177  * @rx_remove: Free resources for RX queue
1178  * @rx_write: Write RX descriptors and doorbell
1179  * @rx_defer_refill: Generate a refill reminder event
1180  * @ev_probe: Allocate resources for event queue
1181  * @ev_init: Initialise event queue on the NIC
1182  * @ev_fini: Deinitialise event queue on the NIC
1183  * @ev_remove: Free resources for event queue
1184  * @ev_process: Process events for a queue, up to the given NAPI quota
1185  * @ev_read_ack: Acknowledge read events on a queue, rearming its IRQ
1186  * @ev_test_generate: Generate a test event
1187  * @filter_table_probe: Probe filter capabilities and set up filter software state
1188  * @filter_table_restore: Restore filters removed from hardware
1189  * @filter_table_remove: Remove filters from hardware and tear down software state
1190  * @filter_update_rx_scatter: Update filters after change to rx scatter setting
1191  * @filter_insert: add or replace a filter
1192  * @filter_remove_safe: remove a filter by ID, carefully
1193  * @filter_get_safe: retrieve a filter by ID, carefully
1194  * @filter_clear_rx: Remove all RX filters whose priority is less than or
1195  *	equal to the given priority and is not %EFX_FILTER_PRI_AUTO
1196  * @filter_count_rx_used: Get the number of filters in use at a given priority
1197  * @filter_get_rx_id_limit: Get maximum value of a filter id, plus 1
1198  * @filter_get_rx_ids: Get list of RX filters at a given priority
1199  * @filter_rfs_insert: Add or replace a filter for RFS.  This must be
1200  *	atomic.  The hardware change may be asynchronous but should
1201  *	not be delayed for long.  It may fail if this can't be done
1202  *	atomically.
1203  * @filter_rfs_expire_one: Consider expiring a filter inserted for RFS.
1204  *	This must check whether the specified table entry is used by RFS
1205  *	and that rps_may_expire_flow() returns true for it.
1206  * @mtd_probe: Probe and add MTD partitions associated with this net device,
1207  *	 using efx_mtd_add()
1208  * @mtd_rename: Set an MTD partition name using the net device name
1209  * @mtd_read: Read from an MTD partition
1210  * @mtd_erase: Erase part of an MTD partition
1211  * @mtd_write: Write to an MTD partition
1212  * @mtd_sync: Wait for write-back to complete on MTD partition.  This
1213  *	also notifies the driver that a writer has finished using this
1214  *	partition.
1215  * @ptp_write_host_time: Send host time to MC as part of sync protocol
1216  * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
1217  *	timestamping, possibly only temporarily for the purposes of a reset.
1218  * @ptp_set_ts_config: Set hardware timestamp configuration.  The flags
1219  *	and tx_type will already have been validated but this operation
1220  *	must validate and update rx_filter.
1221  * @revision: Hardware architecture revision
1222  * @txd_ptr_tbl_base: TX descriptor ring base address
1223  * @rxd_ptr_tbl_base: RX descriptor ring base address
1224  * @buf_tbl_base: Buffer table base address
1225  * @evq_ptr_tbl_base: Event queue pointer table base address
1226  * @evq_rptr_tbl_base: Event queue read-pointer table base address
1227  * @max_dma_mask: Maximum possible DMA mask
1228  * @rx_prefix_size: Size of RX prefix before packet data
1229  * @rx_hash_offset: Offset of RX flow hash within prefix
1230  * @rx_ts_offset: Offset of timestamp within prefix
1231  * @rx_buffer_padding: Size of padding at end of RX packet
1232  * @can_rx_scatter: NIC is able to scatter packets to multiple buffers
1233  * @always_rx_scatter: NIC will always scatter packets to multiple buffers
1234  * @max_interrupt_mode: Highest capability interrupt mode supported
1235  *	from &enum efx_init_mode.
1236  * @timer_period_max: Maximum period of interrupt timer (in ticks)
1237  * @offload_features: net_device feature flags for protocol offload
1238  *	features implemented in hardware
1239  * @mcdi_max_ver: Maximum MCDI version supported
1240  * @hwtstamp_filters: Mask of hardware timestamp filter types supported
1241  */
1242 struct efx_nic_type {
1243 	unsigned int (*mem_map_size)(struct efx_nic *efx);
1244 	int (*probe)(struct efx_nic *efx);
1245 	void (*remove)(struct efx_nic *efx);
1246 	int (*init)(struct efx_nic *efx);
1247 	int (*dimension_resources)(struct efx_nic *efx);
1248 	void (*fini)(struct efx_nic *efx);
1249 	void (*monitor)(struct efx_nic *efx);
1250 	enum reset_type (*map_reset_reason)(enum reset_type reason);
1251 	int (*map_reset_flags)(u32 *flags);
1252 	int (*reset)(struct efx_nic *efx, enum reset_type method);
1253 	int (*probe_port)(struct efx_nic *efx);
1254 	void (*remove_port)(struct efx_nic *efx);
1255 	bool (*handle_global_event)(struct efx_channel *channel, efx_qword_t *);
1256 	int (*fini_dmaq)(struct efx_nic *efx);
1257 	void (*prepare_flush)(struct efx_nic *efx);
1258 	void (*finish_flush)(struct efx_nic *efx);
1259 	void (*prepare_flr)(struct efx_nic *efx);
1260 	void (*finish_flr)(struct efx_nic *efx);
1261 	size_t (*describe_stats)(struct efx_nic *efx, u8 *names);
1262 	size_t (*update_stats)(struct efx_nic *efx, u64 *full_stats,
1263 			       struct rtnl_link_stats64 *core_stats);
1264 	void (*start_stats)(struct efx_nic *efx);
1265 	void (*pull_stats)(struct efx_nic *efx);
1266 	void (*stop_stats)(struct efx_nic *efx);
1267 	void (*set_id_led)(struct efx_nic *efx, enum efx_led_mode mode);
1268 	void (*push_irq_moderation)(struct efx_channel *channel);
1269 	int (*reconfigure_port)(struct efx_nic *efx);
1270 	void (*prepare_enable_fc_tx)(struct efx_nic *efx);
1271 	int (*reconfigure_mac)(struct efx_nic *efx);
1272 	bool (*check_mac_fault)(struct efx_nic *efx);
1273 	void (*get_wol)(struct efx_nic *efx, struct ethtool_wolinfo *wol);
1274 	int (*set_wol)(struct efx_nic *efx, u32 type);
1275 	void (*resume_wol)(struct efx_nic *efx);
1276 	int (*test_chip)(struct efx_nic *efx, struct efx_self_tests *tests);
1277 	int (*test_nvram)(struct efx_nic *efx);
1278 	void (*mcdi_request)(struct efx_nic *efx,
1279 			     const efx_dword_t *hdr, size_t hdr_len,
1280 			     const efx_dword_t *sdu, size_t sdu_len);
1281 	bool (*mcdi_poll_response)(struct efx_nic *efx);
1282 	void (*mcdi_read_response)(struct efx_nic *efx, efx_dword_t *pdu,
1283 				   size_t pdu_offset, size_t pdu_len);
1284 	int (*mcdi_poll_reboot)(struct efx_nic *efx);
1285 	void (*irq_enable_master)(struct efx_nic *efx);
1286 	void (*irq_test_generate)(struct efx_nic *efx);
1287 	void (*irq_disable_non_ev)(struct efx_nic *efx);
1288 	irqreturn_t (*irq_handle_msi)(int irq, void *dev_id);
1289 	irqreturn_t (*irq_handle_legacy)(int irq, void *dev_id);
1290 	int (*tx_probe)(struct efx_tx_queue *tx_queue);
1291 	void (*tx_init)(struct efx_tx_queue *tx_queue);
1292 	void (*tx_remove)(struct efx_tx_queue *tx_queue);
1293 	void (*tx_write)(struct efx_tx_queue *tx_queue);
1294 	void (*rx_push_rss_config)(struct efx_nic *efx);
1295 	int (*rx_probe)(struct efx_rx_queue *rx_queue);
1296 	void (*rx_init)(struct efx_rx_queue *rx_queue);
1297 	void (*rx_remove)(struct efx_rx_queue *rx_queue);
1298 	void (*rx_write)(struct efx_rx_queue *rx_queue);
1299 	void (*rx_defer_refill)(struct efx_rx_queue *rx_queue);
1300 	int (*ev_probe)(struct efx_channel *channel);
1301 	int (*ev_init)(struct efx_channel *channel);
1302 	void (*ev_fini)(struct efx_channel *channel);
1303 	void (*ev_remove)(struct efx_channel *channel);
1304 	int (*ev_process)(struct efx_channel *channel, int quota);
1305 	void (*ev_read_ack)(struct efx_channel *channel);
1306 	void (*ev_test_generate)(struct efx_channel *channel);
1307 	int (*filter_table_probe)(struct efx_nic *efx);
1308 	void (*filter_table_restore)(struct efx_nic *efx);
1309 	void (*filter_table_remove)(struct efx_nic *efx);
1310 	void (*filter_update_rx_scatter)(struct efx_nic *efx);
1311 	s32 (*filter_insert)(struct efx_nic *efx,
1312 			     struct efx_filter_spec *spec, bool replace);
1313 	int (*filter_remove_safe)(struct efx_nic *efx,
1314 				  enum efx_filter_priority priority,
1315 				  u32 filter_id);
1316 	int (*filter_get_safe)(struct efx_nic *efx,
1317 			       enum efx_filter_priority priority,
1318 			       u32 filter_id, struct efx_filter_spec *);
1319 	int (*filter_clear_rx)(struct efx_nic *efx,
1320 			       enum efx_filter_priority priority);
1321 	u32 (*filter_count_rx_used)(struct efx_nic *efx,
1322 				    enum efx_filter_priority priority);
1323 	u32 (*filter_get_rx_id_limit)(struct efx_nic *efx);
1324 	s32 (*filter_get_rx_ids)(struct efx_nic *efx,
1325 				 enum efx_filter_priority priority,
1326 				 u32 *buf, u32 size);
1327 #ifdef CONFIG_RFS_ACCEL
1328 	s32 (*filter_rfs_insert)(struct efx_nic *efx,
1329 				 struct efx_filter_spec *spec);
1330 	bool (*filter_rfs_expire_one)(struct efx_nic *efx, u32 flow_id,
1331 				      unsigned int index);
1332 #endif
1333 #ifdef CONFIG_SFC_MTD
1334 	int (*mtd_probe)(struct efx_nic *efx);
1335 	void (*mtd_rename)(struct efx_mtd_partition *part);
1336 	int (*mtd_read)(struct mtd_info *mtd, loff_t start, size_t len,
1337 			size_t *retlen, u8 *buffer);
1338 	int (*mtd_erase)(struct mtd_info *mtd, loff_t start, size_t len);
1339 	int (*mtd_write)(struct mtd_info *mtd, loff_t start, size_t len,
1340 			 size_t *retlen, const u8 *buffer);
1341 	int (*mtd_sync)(struct mtd_info *mtd);
1342 #endif
1343 	void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
1344 	int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
1345 	int (*ptp_set_ts_config)(struct efx_nic *efx,
1346 				 struct hwtstamp_config *init);
1347 
1348 	int revision;
1349 	unsigned int txd_ptr_tbl_base;
1350 	unsigned int rxd_ptr_tbl_base;
1351 	unsigned int buf_tbl_base;
1352 	unsigned int evq_ptr_tbl_base;
1353 	unsigned int evq_rptr_tbl_base;
1354 	u64 max_dma_mask;
1355 	unsigned int rx_prefix_size;
1356 	unsigned int rx_hash_offset;
1357 	unsigned int rx_ts_offset;
1358 	unsigned int rx_buffer_padding;
1359 	bool can_rx_scatter;
1360 	bool always_rx_scatter;
1361 	unsigned int max_interrupt_mode;
1362 	unsigned int timer_period_max;
1363 	netdev_features_t offload_features;
1364 	int mcdi_max_ver;
1365 	unsigned int max_rx_ip_filters;
1366 	u32 hwtstamp_filters;
1367 };
1368 
1369 /**************************************************************************
1370  *
1371  * Prototypes and inline functions
1372  *
1373  *************************************************************************/
1374 
1375 static inline struct efx_channel *
1376 efx_get_channel(struct efx_nic *efx, unsigned index)
1377 {
1378 	EFX_BUG_ON_PARANOID(index >= efx->n_channels);
1379 	return efx->channel[index];
1380 }
1381 
1382 /* Iterate over all used channels */
1383 #define efx_for_each_channel(_channel, _efx)				\
1384 	for (_channel = (_efx)->channel[0];				\
1385 	     _channel;							\
1386 	     _channel = (_channel->channel + 1 < (_efx)->n_channels) ?	\
1387 		     (_efx)->channel[_channel->channel + 1] : NULL)
1388 
1389 /* Iterate over all used channels in reverse */
1390 #define efx_for_each_channel_rev(_channel, _efx)			\
1391 	for (_channel = (_efx)->channel[(_efx)->n_channels - 1];	\
1392 	     _channel;							\
1393 	     _channel = _channel->channel ?				\
1394 		     (_efx)->channel[_channel->channel - 1] : NULL)
1395 
1396 static inline struct efx_tx_queue *
1397 efx_get_tx_queue(struct efx_nic *efx, unsigned index, unsigned type)
1398 {
1399 	EFX_BUG_ON_PARANOID(index >= efx->n_tx_channels ||
1400 			    type >= EFX_TXQ_TYPES);
1401 	return &efx->channel[efx->tx_channel_offset + index]->tx_queue[type];
1402 }
1403 
1404 static inline bool efx_channel_has_tx_queues(struct efx_channel *channel)
1405 {
1406 	return channel->channel - channel->efx->tx_channel_offset <
1407 		channel->efx->n_tx_channels;
1408 }
1409 
1410 static inline struct efx_tx_queue *
1411 efx_channel_get_tx_queue(struct efx_channel *channel, unsigned type)
1412 {
1413 	EFX_BUG_ON_PARANOID(!efx_channel_has_tx_queues(channel) ||
1414 			    type >= EFX_TXQ_TYPES);
1415 	return &channel->tx_queue[type];
1416 }
1417 
1418 static inline bool efx_tx_queue_used(struct efx_tx_queue *tx_queue)
1419 {
1420 	return !(tx_queue->efx->net_dev->num_tc < 2 &&
1421 		 tx_queue->queue & EFX_TXQ_TYPE_HIGHPRI);
1422 }
1423 
1424 /* Iterate over all TX queues belonging to a channel */
1425 #define efx_for_each_channel_tx_queue(_tx_queue, _channel)		\
1426 	if (!efx_channel_has_tx_queues(_channel))			\
1427 		;							\
1428 	else								\
1429 		for (_tx_queue = (_channel)->tx_queue;			\
1430 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES && \
1431 			     efx_tx_queue_used(_tx_queue);		\
1432 		     _tx_queue++)
1433 
1434 /* Iterate over all possible TX queues belonging to a channel */
1435 #define efx_for_each_possible_channel_tx_queue(_tx_queue, _channel)	\
1436 	if (!efx_channel_has_tx_queues(_channel))			\
1437 		;							\
1438 	else								\
1439 		for (_tx_queue = (_channel)->tx_queue;			\
1440 		     _tx_queue < (_channel)->tx_queue + EFX_TXQ_TYPES;	\
1441 		     _tx_queue++)
1442 
1443 static inline bool efx_channel_has_rx_queue(struct efx_channel *channel)
1444 {
1445 	return channel->rx_queue.core_index >= 0;
1446 }
1447 
1448 static inline struct efx_rx_queue *
1449 efx_channel_get_rx_queue(struct efx_channel *channel)
1450 {
1451 	EFX_BUG_ON_PARANOID(!efx_channel_has_rx_queue(channel));
1452 	return &channel->rx_queue;
1453 }
1454 
1455 /* Iterate over all RX queues belonging to a channel */
1456 #define efx_for_each_channel_rx_queue(_rx_queue, _channel)		\
1457 	if (!efx_channel_has_rx_queue(_channel))			\
1458 		;							\
1459 	else								\
1460 		for (_rx_queue = &(_channel)->rx_queue;			\
1461 		     _rx_queue;						\
1462 		     _rx_queue = NULL)
1463 
1464 static inline struct efx_channel *
1465 efx_rx_queue_channel(struct efx_rx_queue *rx_queue)
1466 {
1467 	return container_of(rx_queue, struct efx_channel, rx_queue);
1468 }
1469 
1470 static inline int efx_rx_queue_index(struct efx_rx_queue *rx_queue)
1471 {
1472 	return efx_rx_queue_channel(rx_queue)->channel;
1473 }
1474 
1475 /* Returns a pointer to the specified receive buffer in the RX
1476  * descriptor queue.
1477  */
1478 static inline struct efx_rx_buffer *efx_rx_buffer(struct efx_rx_queue *rx_queue,
1479 						  unsigned int index)
1480 {
1481 	return &rx_queue->buffer[index];
1482 }
1483 
1484 /**
1485  * EFX_MAX_FRAME_LEN - calculate maximum frame length
1486  *
1487  * This calculates the maximum frame length that will be used for a
1488  * given MTU.  The frame length will be equal to the MTU plus a
1489  * constant amount of header space and padding.  This is the quantity
1490  * that the net driver will program into the MAC as the maximum frame
1491  * length.
1492  *
1493  * The 10G MAC requires 8-byte alignment on the frame
1494  * length, so we round up to the nearest 8.
1495  *
1496  * Re-clocking by the XGXS on RX can reduce an IPG to 32 bits (half an
1497  * XGMII cycle).  If the frame length reaches the maximum value in the
1498  * same cycle, the XMAC can miss the IPG altogether.  We work around
1499  * this by adding a further 16 bytes.
1500  */
1501 #define EFX_MAX_FRAME_LEN(mtu) \
1502 	((((mtu) + ETH_HLEN + VLAN_HLEN + 4/* FCS */ + 7) & ~7) + 16)
1503 
1504 static inline bool efx_xmit_with_hwtstamp(struct sk_buff *skb)
1505 {
1506 	return skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP;
1507 }
1508 static inline void efx_xmit_hwtstamp_pending(struct sk_buff *skb)
1509 {
1510 	skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
1511 }
1512 
1513 #endif /* EFX_NET_DRIVER_H */
1514